Attachment 114936 Details for Bug 89944 – R600_DEBUG=ps,vs,gs output from crashing program

R600_DEBUG=ps,vs,gs output from crashing program

civ5xp_r600_debug_psvsgs_mesa_10.6.0-git-490621f_bad.log (text/plain), 1.93 MB, created by Sami Liedes on 2015-04-07 20:19:28 UTC

(hide)

Description:

Filename:

MIME Type:

Creator: Sami Liedes

Created: 2015-04-07 20:19:28 UTC

Size: 1.93 MB

patch

obsolete

>[S_API FAIL] SteamAPI_Init() failed; SteamAPI_IsSteamRunning() failed.
>[S_API FAIL] SteamAPI_Init() failed; unable to locate a running instance of Steam, or a local steamclient.so.
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>GUID Assets\dlc\dlc_01\mongol.civ5pkg 7a036b7fb9a80e8dea7b73fb58c5a288
>GUID Assets\dlc\dlc_02\spaininca.civ5pkg 4f75e72761f4c6019b55a0a7b90444a2
>GUID Assets\dlc\dlc_03\polynesia.civ5pkg 99ac9d5f6ca4b5bed0ab89c0fd3b9e6d
>GUID Assets\dlc\dlc_04\denmark.civ5pkg 0efb155307bd6d14c9290b49b5364a3e
>GUID Assets\dlc\dlc_05\korea.civ5pkg 9f4df81cf712ae9480737f816bf6f4c8
>GUID Assets\dlc\dlc_06\ancientwonders.civ5pkg 92b102db9a3c7dc068030c3ce33bbb48
>GUID Assets\dlc\dlc_07\civcomplete.civ5pkg eb01a0be4d8e5312f53b042c8a7c30b5
>GUID Assets\dlc\dlc_deluxe\babylon.civ5pkg 712495341921f2b288746c6d44fd6867
>GUID Assets\dlc\dlc_sp_maps\dlc_sp_maps.civ5pkg 52b285c37939913e0a5b72933bb06067
>GUID Assets\dlc\dlc_sp_maps_2\dlc_sp_maps_2.civ5pkg 16a61e7a2a7bb4bc2d1f677b5bb58ff4
>GUID Assets\dlc\dlc_sp_maps_3\dlc_sp_maps_3.civ5pkg 1954db58e0a60b018969c49440fa01ef
>GUID Assets\dlc\expansion\expansion1.civ5pkg 8bc30c58378345cb0911c5848926f1ff
>GUID Assets\dlc\expansion2\expansion2.civ5pkg 31dfaa9838c5b051d4c2112ddd9e7eb3
>GUID Assets\dlc\shared\upgrade1.civ5pkg e818fa28902977b42ee5e3426f5112e6
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[0]
>  0: MOV OUT[0], IN[0]
>  1: MOV OUT[1], IN[1]
>  2: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = add i32 %5, %7
>  %14 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %12, i32 0, i32 %13)
>  %15 = extractelement <4 x float> %14, i32 0
>  %16 = extractelement <4 x float> %14, i32 1
>  %17 = extractelement <4 x float> %14, i32 2
>  %18 = extractelement <4 x float> %14, i32 3
>  %19 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %20 = load <16 x i8> addrspace(2)* %19, !tbaa !0
>  %21 = add i32 %5, %7
>  %22 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %20, i32 0, i32 %21)
>  %23 = extractelement <4 x float> %22, i32 0
>  %24 = extractelement <4 x float> %22, i32 1
>  %25 = extractelement <4 x float> %22, i32 2
>  %26 = extractelement <4 x float> %22, i32 3
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %23, float %24, float %25, float %26)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %15, float %16, float %17, float %18)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                           ; 4A00000A
>	s_load_dwordx4 s[0:3], s[8:9], 0x0                  ; C0800900
>	s_load_dwordx4 s[4:7], s[8:9], 0x4                  ; C0820904
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[4:7], 0 idxen ; E00C2000 80010100
>	s_waitcnt vmcnt(0)                                  ; BF8C0770
>	exp 15, 32, 0, 0, 0, v1, v2, v3, v4                 ; F800020F 04030201
>	s_waitcnt expcnt(0)                                 ; BF8C070F
>	buffer_load_format_xyzw v[0:3], v0, s[0:3], 0 idxen ; E00C2000 80000000
>	s_waitcnt vmcnt(0)                                  ; BF8C0770
>	exp 15, 12, 0, 1, 0, v0, v1, v2, v3                 ; F80008CF 03020100
>	s_endpgm                                            ; BF810000
>
>SGPRS: 16
>VGPRS: 8
>Code Size: 64 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>PROPERTY FS_COLOR0_WRITES_ALL_CBUFS 1
>DCL IN[0], GENERIC[0], CONSTANT
>DCL OUT[0], COLOR
>  0: MOV OUT[0], IN[0]
>  1: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = call float @llvm.SI.fs.constant(i32 0, i32 0, i32 %5)
>  %23 = call float @llvm.SI.fs.constant(i32 1, i32 0, i32 %5)
>  %24 = call float @llvm.SI.fs.constant(i32 2, i32 0, i32 %5)
>  %25 = call float @llvm.SI.fs.constant(i32 3, i32 0, i32 %5)
>  %26 = call i32 @llvm.SI.packf16(float %22, float %23)
>  %27 = bitcast i32 %26 to float
>  %28 = call i32 @llvm.SI.packf16(float %24, float %25)
>  %29 = bitcast i32 %28 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %27, float %29, float %27, float %29)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.constant(i32, i32, i32) #1
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                ; BEFE0A7E
>	s_mov_b32 m0, s9                    ; BEFC0309
>	v_interp_mov_f32 v0, P0, 3, 0, [m0] ; C8020302
>	v_interp_mov_f32 v1, P0, 2, 0, [m0] ; C8060202
>	v_cvt_pkrtz_f16_f32_e32 v0, v1, v0  ; 5E000101
>	v_interp_mov_f32 v1, P0, 1, 0, [m0] ; C8060102
>	v_interp_mov_f32 v2, P0, 0, 0, [m0] ; C80A0002
>	v_cvt_pkrtz_f16_f32_e32 v1, v2, v1  ; 5E020302
>	exp 15, 0, 1, 1, 1, v1, v0, v1, v0  ; F8001C0F 00010001
>	s_endpgm                            ; BF810000
>
>SGPRS: 16
>VGPRS: 4
>Code Size: 44 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[0], CONSTANT
>DCL OUT[0], COLOR
>  0: MOV OUT[0], IN[0]
>  1: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = call float @llvm.SI.fs.constant(i32 0, i32 0, i32 %5)
>  %23 = call float @llvm.SI.fs.constant(i32 1, i32 0, i32 %5)
>  %24 = call float @llvm.SI.fs.constant(i32 2, i32 0, i32 %5)
>  %25 = call float @llvm.SI.fs.constant(i32 3, i32 0, i32 %5)
>  %26 = call i32 @llvm.SI.packf16(float %22, float %23)
>  %27 = bitcast i32 %26 to float
>  %28 = call i32 @llvm.SI.packf16(float %24, float %25)
>  %29 = bitcast i32 %28 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %27, float %29, float %27, float %29)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.constant(i32, i32, i32) #1
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                ; BEFE0A7E
>	s_mov_b32 m0, s9                    ; BEFC0309
>	v_interp_mov_f32 v0, P0, 3, 0, [m0] ; C8020302
>	v_interp_mov_f32 v1, P0, 2, 0, [m0] ; C8060202
>	v_cvt_pkrtz_f16_f32_e32 v0, v1, v0  ; 5E000101
>	v_interp_mov_f32 v1, P0, 1, 0, [m0] ; C8060102
>	v_interp_mov_f32 v2, P0, 0, 0, [m0] ; C80A0002
>	v_cvt_pkrtz_f16_f32_e32 v1, v2, v1  ; 5E020302
>	exp 15, 0, 1, 1, 1, v1, v0, v1, v0  ; F8001C0F 00010001
>	s_endpgm                            ; BF810000
>
>SGPRS: 16
>VGPRS: 4
>Code Size: 44 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[0], LINEAR
>DCL OUT[0], COLOR
>DCL SAMP[0]
>  0: TEX OUT[0], IN[0], SAMP[0], 2D
>  1: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %23 = load <8 x i32> addrspace(2)* %22, !tbaa !0
>  %24 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %25 = load <4 x i32> addrspace(2)* %24, !tbaa !0
>  %26 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %11)
>  %27 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %11)
>  %28 = bitcast float %26 to i32
>  %29 = bitcast float %27 to i32
>  %30 = insertelement <2 x i32> undef, i32 %28, i32 0
>  %31 = insertelement <2 x i32> %30, i32 %29, i32 1
>  %32 = bitcast <8 x i32> %23 to <32 x i8>
>  %33 = bitcast <4 x i32> %25 to <16 x i8>
>  %34 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %31, <32 x i8> %32, <16 x i8> %33, i32 2)
>  %35 = extractelement <4 x float> %34, i32 0
>  %36 = extractelement <4 x float> %34, i32 1
>  %37 = extractelement <4 x float> %34, i32 2
>  %38 = extractelement <4 x float> %34, i32 3
>  %39 = call i32 @llvm.SI.packf16(float %35, float %36)
>  %40 = bitcast i32 %39 to float
>  %41 = call i32 @llvm.SI.packf16(float %37, float %38)
>  %42 = bitcast i32 %41 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %40, float %42, float %40, float %42)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                  ; BEFE0A7E
>	s_mov_b32 m0, s9                                                      ; BEFC0309
>	v_interp_p1_f32 v3, v0, 1, 0, [m0]                                    ; C80C0100
>	v_interp_p2_f32 v3, [v3], v1, 1, 0, [m0]                              ; C80D0101
>	v_interp_p1_f32 v2, v0, 0, 0, [m0]                                    ; C8080000
>	v_interp_p2_f32 v2, [v2], v1, 0, 0, [m0]                              ; C8090001
>	s_load_dwordx4 s[0:3], s[4:5], 0x0                                    ; C0800500
>	s_load_dwordx8 s[4:11], s[6:7], 0x0                                   ; C0C20700
>	s_waitcnt lgkmcnt(0)                                                  ; BF8C007F
>	image_sample v[0:3], 15, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[4:11], s[0:3] ; F0800F00 00010002
>	s_waitcnt vmcnt(0)                                                    ; BF8C0770
>	v_cvt_pkrtz_f16_f32_e32 v4, v2, v3                                    ; 5E080702
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v1                                    ; 5E000300
>	exp 15, 0, 1, 1, 1, v0, v4, v0, v4                                    ; F8001C0F 04000400
>	s_endpgm                                                              ; BF810000
>
>SGPRS: 16
>VGPRS: 8
>Code Size: 68 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL TEMP[0..1], LOCAL
>IMM[0] FLT32 {    0.0000,     1.0000,     0.0000,     0.0000}
>  0: MOV TEMP[0].zw, IMM[0].yyxy
>  1: MOV TEMP[0].x, IN[0].xxxx
>  2: MOV TEMP[0].y, IN[0].yyyy
>  3: MOV TEMP[1].xy, IN[1].xyxx
>  4: MOV OUT[1], TEMP[1]
>  5: MOV OUT[0], TEMP[0]
>  6: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = add i32 %5, %7
>  %14 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %12, i32 0, i32 %13)
>  %15 = extractelement <4 x float> %14, i32 0
>  %16 = extractelement <4 x float> %14, i32 1
>  %17 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %18 = load <16 x i8> addrspace(2)* %17, !tbaa !0
>  %19 = add i32 %5, %7
>  %20 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %18, i32 0, i32 %19)
>  %21 = extractelement <4 x float> %20, i32 0
>  %22 = extractelement <4 x float> %20, i32 1
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %21, float %22, float 0.000000e+00, float 0.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %15, float %16, float 0.000000e+00, float 1.000000e+00)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                           ; 4A00000A
>	s_load_dwordx4 s[0:3], s[8:9], 0x0                  ; C0800900
>	s_load_dwordx4 s[4:7], s[8:9], 0x4                  ; C0820904
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[4:7], 0 idxen ; E00C2000 80010100
>	v_mov_b32_e32 v5, 0                                 ; 7E0A0280
>	s_waitcnt vmcnt(0)                                  ; BF8C0770
>	exp 15, 32, 0, 0, 0, v1, v2, v5, v5                 ; F800020F 05050201
>	s_waitcnt expcnt(0)                                 ; BF8C070F
>	buffer_load_format_xyzw v[0:3], v0, s[0:3], 0 idxen ; E00C2000 80000000
>	v_mov_b32_e32 v4, 1.0                               ; 7E0802F2
>	s_waitcnt vmcnt(0)                                  ; BF8C0770
>	exp 15, 12, 0, 1, 0, v0, v1, v5, v4                 ; F80008CF 04050100
>	s_endpgm                                            ; BF810000
>
>SGPRS: 16
>VGPRS: 8
>Code Size: 72 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>PROPERTY FS_COLOR0_WRITES_ALL_CBUFS 1
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL TEMP[0], LOCAL
>  0: MOV TEMP[0].xy, IN[0].xyyy
>  1: TEX TEMP[0], TEMP[0], SAMP[0], 2D
>  2: MOV OUT[0], TEMP[0]
>  3: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %23 = load <8 x i32> addrspace(2)* %22, !tbaa !0
>  %24 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %25 = load <4 x i32> addrspace(2)* %24, !tbaa !0
>  %26 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %27 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %28 = bitcast float %26 to i32
>  %29 = bitcast float %27 to i32
>  %30 = insertelement <2 x i32> undef, i32 %28, i32 0
>  %31 = insertelement <2 x i32> %30, i32 %29, i32 1
>  %32 = bitcast <8 x i32> %23 to <32 x i8>
>  %33 = bitcast <4 x i32> %25 to <16 x i8>
>  %34 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %31, <32 x i8> %32, <16 x i8> %33, i32 2)
>  %35 = extractelement <4 x float> %34, i32 0
>  %36 = extractelement <4 x float> %34, i32 1
>  %37 = extractelement <4 x float> %34, i32 2
>  %38 = extractelement <4 x float> %34, i32 3
>  %39 = call i32 @llvm.SI.packf16(float %35, float %36)
>  %40 = bitcast i32 %39 to float
>  %41 = call i32 @llvm.SI.packf16(float %37, float %38)
>  %42 = bitcast i32 %41 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %40, float %42, float %40, float %42)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                  ; BEFE0A7E
>	s_mov_b32 m0, s9                                                      ; BEFC0309
>	v_interp_p1_f32 v3, v0, 1, 0, [m0]                                    ; C80C0100
>	v_interp_p2_f32 v3, [v3], v1, 1, 0, [m0]                              ; C80D0101
>	v_interp_p1_f32 v2, v0, 0, 0, [m0]                                    ; C8080000
>	v_interp_p2_f32 v2, [v2], v1, 0, 0, [m0]                              ; C8090001
>	s_load_dwordx4 s[0:3], s[4:5], 0x0                                    ; C0800500
>	s_load_dwordx8 s[4:11], s[6:7], 0x0                                   ; C0C20700
>	s_waitcnt lgkmcnt(0)                                                  ; BF8C007F
>	image_sample v[0:3], 15, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[4:11], s[0:3] ; F0800F00 00010002
>	s_waitcnt vmcnt(0)                                                    ; BF8C0770
>	v_cvt_pkrtz_f16_f32_e32 v4, v2, v3                                    ; 5E080702
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v1                                    ; 5E000300
>	exp 15, 0, 1, 1, 1, v0, v4, v0, v4                                    ; F8001C0F 04000400
>	s_endpgm                                                              ; BF810000
>
>SGPRS: 16
>VGPRS: 8
>Code Size: 68 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x0
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[0], LINEAR
>DCL OUT[0], COLOR
>DCL SAMP[0]
>  0: TEX OUT[0], IN[0], SAMP[0], 2D
>  1: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %23 = load <8 x i32> addrspace(2)* %22, !tbaa !0
>  %24 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %25 = load <4 x i32> addrspace(2)* %24, !tbaa !0
>  %26 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %11)
>  %27 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %11)
>  %28 = bitcast float %26 to i32
>  %29 = bitcast float %27 to i32
>  %30 = insertelement <2 x i32> undef, i32 %28, i32 0
>  %31 = insertelement <2 x i32> %30, i32 %29, i32 1
>  %32 = bitcast <8 x i32> %23 to <32 x i8>
>  %33 = bitcast <4 x i32> %25 to <16 x i8>
>  %34 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %31, <32 x i8> %32, <16 x i8> %33, i32 2)
>  %35 = extractelement <4 x float> %34, i32 0
>  %36 = extractelement <4 x float> %34, i32 1
>  %37 = extractelement <4 x float> %34, i32 2
>  %38 = extractelement <4 x float> %34, i32 3
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 0, float %35, float %36, float %37, float %38)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                  ; BEFE0A7E
>	s_mov_b32 m0, s9                                                      ; BEFC0309
>	v_interp_p1_f32 v3, v0, 1, 0, [m0]                                    ; C80C0100
>	v_interp_p2_f32 v3, [v3], v1, 1, 0, [m0]                              ; C80D0101
>	v_interp_p1_f32 v2, v0, 0, 0, [m0]                                    ; C8080000
>	v_interp_p2_f32 v2, [v2], v1, 0, 0, [m0]                              ; C8090001
>	s_load_dwordx4 s[0:3], s[4:5], 0x0                                    ; C0800500
>	s_load_dwordx8 s[4:11], s[6:7], 0x0                                   ; C0C20700
>	s_waitcnt lgkmcnt(0)                                                  ; BF8C007F
>	image_sample v[0:3], 15, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[4:11], s[0:3] ; F0800F00 00010002
>	s_waitcnt vmcnt(0)                                                    ; BF8C0770
>	exp 15, 0, 0, 1, 1, v0, v1, v2, v3                                    ; F800180F 03020100
>	s_endpgm                                                              ; BF810000
>
>SGPRS: 16
>VGPRS: 4
>Code Size: 60 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL IN[2]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL OUT[2], GENERIC[10]
>DCL CONST[0..8]
>DCL TEMP[0..3], LOCAL
>IMM[0] FLT32 {    0.0000,     1.0000,     0.0000,     0.0000}
>  0: MAD TEMP[0].xy, IN[0], CONST[0], CONST[0].zwzw
>  1: MUL TEMP[1], TEMP[0].yyyy, CONST[6]
>  2: MAD TEMP[0], TEMP[0].xxxx, CONST[5], TEMP[1]
>  3: ADD TEMP[0], TEMP[0], CONST[8]
>  4: ADD TEMP[1].x, CONST[4].xxxx, CONST[4].yyyy
>  5: MUL TEMP[1], TEMP[0], TEMP[1].xxxx
>  6: MAD TEMP[2].xy, IN[1], CONST[1], CONST[1].zwzw
>  7: MOV TEMP[2].xy, TEMP[2].xyxx
>  8: MUL TEMP[0], CONST[2], IN[2]
>  9: MUL TEMP[3].w, TEMP[0].wwww, CONST[3].wwww
> 10: MOV TEMP[3].w, TEMP[3].wwww
> 11: MOV TEMP[3].xyz, TEMP[0].xyzx
> 12: MOV TEMP[2].zw, IMM[0].yyxy
> 13: MOV OUT[1], TEMP[3]
> 14: MOV OUT[2], TEMP[2]
> 15: MOV OUT[0], TEMP[1]
> 16: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 0)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 4)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %12, i32 8)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %12, i32 12)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %12, i32 16)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %12, i32 20)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %12, i32 24)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %12, i32 28)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %12, i32 32)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %12, i32 36)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %12, i32 40)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %12, i32 44)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %12, i32 60)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %12, i32 64)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %12, i32 68)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %12, i32 80)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %12, i32 84)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %12, i32 88)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %12, i32 92)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %12, i32 96)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %12, i32 100)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %12, i32 104)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %12, i32 108)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %12, i32 128)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %12, i32 132)
>  %38 = call float @llvm.SI.load.const(<16 x i8> %12, i32 136)
>  %39 = call float @llvm.SI.load.const(<16 x i8> %12, i32 140)
>  %40 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %41 = load <16 x i8> addrspace(2)* %40, !tbaa !0
>  %42 = add i32 %5, %7
>  %43 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %41, i32 0, i32 %42)
>  %44 = extractelement <4 x float> %43, i32 0
>  %45 = extractelement <4 x float> %43, i32 1
>  %46 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %47 = load <16 x i8> addrspace(2)* %46, !tbaa !0
>  %48 = add i32 %5, %7
>  %49 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %47, i32 0, i32 %48)
>  %50 = extractelement <4 x float> %49, i32 0
>  %51 = extractelement <4 x float> %49, i32 1
>  %52 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 2
>  %53 = load <16 x i8> addrspace(2)* %52, !tbaa !0
>  %54 = add i32 %5, %7
>  %55 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %53, i32 0, i32 %54)
>  %56 = extractelement <4 x float> %55, i32 0
>  %57 = extractelement <4 x float> %55, i32 1
>  %58 = extractelement <4 x float> %55, i32 2
>  %59 = extractelement <4 x float> %55, i32 3
>  %60 = fmul float %44, %13
>  %61 = fadd float %60, %15
>  %62 = fmul float %45, %14
>  %63 = fadd float %62, %16
>  %64 = fmul float %63, %32
>  %65 = fmul float %63, %33
>  %66 = fmul float %63, %34
>  %67 = fmul float %63, %35
>  %68 = fmul float %61, %28
>  %69 = fadd float %68, %64
>  %70 = fmul float %61, %29
>  %71 = fadd float %70, %65
>  %72 = fmul float %61, %30
>  %73 = fadd float %72, %66
>  %74 = fmul float %61, %31
>  %75 = fadd float %74, %67
>  %76 = fadd float %69, %36
>  %77 = fadd float %71, %37
>  %78 = fadd float %73, %38
>  %79 = fadd float %75, %39
>  %80 = fadd float %26, %27
>  %81 = fmul float %76, %80
>  %82 = fmul float %77, %80
>  %83 = fmul float %78, %80
>  %84 = fmul float %79, %80
>  %85 = fmul float %50, %17
>  %86 = fadd float %85, %19
>  %87 = fmul float %51, %18
>  %88 = fadd float %87, %20
>  %89 = fmul float %21, %56
>  %90 = fmul float %22, %57
>  %91 = fmul float %23, %58
>  %92 = fmul float %24, %59
>  %93 = fmul float %92, %25
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %89, float %90, float %91, float %93)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %86, float %88, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %81, float %82, float %83, float %84)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                             ; 4A00000A
>	s_load_dwordx4 s[4:7], s[8:9], 0x0                    ; C0820900
>	s_load_dwordx4 s[12:15], s[8:9], 0x4                  ; C0860904
>	s_load_dwordx4 s[8:11], s[8:9], 0x8                   ; C0840908
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[8:11], 0 idxen  ; E00C2000 80020100
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                    ; C0800300
>	s_waitcnt vmcnt(0) lgkmcnt(0)                         ; BF8C0070
>	s_buffer_load_dword s8, s[0:3], 0xa                   ; C204010A
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v5, s8, v3                              ; 100A0608
>	s_buffer_load_dword s8, s[0:3], 0x9                   ; C2040109
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v6, s8, v2                              ; 100C0408
>	s_buffer_load_dword s8, s[0:3], 0x8                   ; C2040108
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v7, s8, v1                              ; 100E0208
>	s_buffer_load_dword s8, s[0:3], 0xb                   ; C204010B
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v1, s8, v4                              ; 10020808
>	s_buffer_load_dword s8, s[0:3], 0xf                   ; C204010F
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v1, s8, v1                              ; 10020208
>	exp 15, 32, 0, 0, 0, v7, v6, v5, v1                   ; F800020F 01050607
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	buffer_load_format_xyzw v[1:4], v0, s[12:15], 0 idxen ; E00C2000 80030100
>	s_buffer_load_dword s8, s[0:3], 0x7                   ; C2040107
>	s_buffer_load_dword s9, s[0:3], 0x5                   ; C2048105
>	s_waitcnt vmcnt(0) lgkmcnt(0)                         ; BF8C0070
>	v_mov_b32_e32 v5, s8                                  ; 7E0A0208
>	v_mad_f32 v5, s9, v2, v5                              ; D2820005 04160409
>	s_buffer_load_dword s8, s[0:3], 0x6                   ; C2040106
>	s_buffer_load_dword s9, s[0:3], 0x4                   ; C2048104
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mov_b32_e32 v6, s8                                  ; 7E0C0208
>	v_mad_f32 v1, s9, v1, v6                              ; D2820001 041A0209
>	v_mov_b32_e32 v2, 1.0                                 ; 7E0402F2
>	v_mov_b32_e32 v3, 0                                   ; 7E060280
>	exp 15, 33, 0, 0, 0, v1, v5, v3, v2                   ; F800021F 02030501
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	buffer_load_format_xyzw v[0:3], v0, s[4:7], 0 idxen   ; E00C2000 80010000
>	s_buffer_load_dword s4, s[0:3], 0x2                   ; C2020102
>	s_buffer_load_dword s5, s[0:3], 0x0                   ; C2028100
>	s_waitcnt vmcnt(0) lgkmcnt(0)                         ; BF8C0070
>	v_mov_b32_e32 v4, s4                                  ; 7E080204
>	v_mad_f32 v4, s5, v0, v4                              ; D2820004 04120005
>	s_buffer_load_dword s4, s[0:3], 0x3                   ; C2020103
>	s_buffer_load_dword s5, s[0:3], 0x1                   ; C2028101
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mov_b32_e32 v5, s4                                  ; 7E0A0204
>	v_mad_f32 v0, s5, v1, v5                              ; D2820000 04160205
>	s_buffer_load_dword s4, s[0:3], 0x1b                  ; C202011B
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v1, s4, v0                              ; 10020004
>	s_buffer_load_dword s4, s[0:3], 0x17                  ; C2020117
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v1, v4, s4, v1                              ; D2820001 04040904
>	s_buffer_load_dword s4, s[0:3], 0x23                  ; C2020123
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v1, s4, v1                              ; 06020204
>	s_buffer_load_dword s4, s[0:3], 0x11                  ; C2020111
>	s_buffer_load_dword s5, s[0:3], 0x10                  ; C2028110
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mov_b32_e32 v2, s4                                  ; 7E040204
>	v_add_f32_e32 v2, s5, v2                              ; 06040405
>	v_mul_f32_e32 v1, v2, v1                              ; 10020302
>	s_buffer_load_dword s4, s[0:3], 0x1a                  ; C202011A
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v3, s4, v0                              ; 10060004
>	s_buffer_load_dword s4, s[0:3], 0x16                  ; C2020116
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v3, v4, s4, v3                              ; D2820003 040C0904
>	s_buffer_load_dword s4, s[0:3], 0x22                  ; C2020122
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v3, s4, v3                              ; 06060604
>	v_mul_f32_e32 v3, v2, v3                              ; 10060702
>	s_buffer_load_dword s4, s[0:3], 0x19                  ; C2020119
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v5, s4, v0                              ; 100A0004
>	s_buffer_load_dword s4, s[0:3], 0x15                  ; C2020115
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v5, v4, s4, v5                              ; D2820005 04140904
>	s_buffer_load_dword s4, s[0:3], 0x21                  ; C2020121
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v5, s4, v5                              ; 060A0A04
>	v_mul_f32_e32 v5, v2, v5                              ; 100A0B02
>	s_buffer_load_dword s4, s[0:3], 0x18                  ; C2020118
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v0, s4, v0                              ; 10000004
>	s_buffer_load_dword s4, s[0:3], 0x14                  ; C2020114
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v0, v4, s4, v0                              ; D2820000 04000904
>	s_buffer_load_dword s0, s[0:3], 0x20                  ; C2000120
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v0, s0, v0                              ; 06000000
>	v_mul_f32_e32 v0, v2, v0                              ; 10000102
>	exp 15, 12, 0, 1, 0, v0, v5, v3, v1                   ; F80008CF 01030500
>	s_endpgm                                              ; BF810000
>
>SGPRS: 24
>VGPRS: 8
>Code Size: 448 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL IN[1], GENERIC[10], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SAMP[1]
>DCL CONST[0..3]
>DCL TEMP[0..7], LOCAL
>IMM[0] FLT32 {    0.5000,  1024.0000,     0.0000,   255.0000}
>IMM[1] FLT32 {  256.0000,     0.0000,     0.0000,     0.0000}
>  0: MOV TEMP[0].xz, IMM[0].xxyx
>  1: FSLT TEMP[1].x, IMM[0].xxxx, CONST[0].xxxx
>  2: UIF TEMP[1].xxxx :0
>  3:   MUL TEMP[1].yw, CONST[2].xxzy, IN[1].xxzy
>  4:   MOV TEMP[0].yw, TEMP[1].wyww
>  5:   FRC TEMP[1].xy, TEMP[0].ywzw
>  6:   MOV TEMP[1].xy, TEMP[1].xyxx
>  7:   ADD TEMP[2].yw, TEMP[0], -TEMP[1].xxzy
>  8:   MOV TEMP[0].yw, TEMP[2].wyww
>  9:   MAD TEMP[2].xy, IN[1], CONST[2], -TEMP[0].ywzw
> 10:   MOV TEMP[1].xy, TEMP[2].xyxx
> 11:   ADD TEMP[2].yw, TEMP[0], IMM[0].xxxx
> 12:   MOV TEMP[0].yw, TEMP[2].wyww
> 13:   RCP TEMP[2].x, CONST[2].xxxx
> 14:   RCP TEMP[3].x, CONST[2].yyyy
> 15:   MOV TEMP[2].y, TEMP[3].xxxx
> 16:   MUL TEMP[3].xy, TEMP[0].ywzw, TEMP[2]
> 17:   MOV TEMP[3].xy, TEMP[3].xyyy
> 18:   MOV TEMP[3].w, IMM[0].zzzz
> 19:   TXL TEMP[3], TEMP[3], SAMP[0], 2D
> 20:   MOV TEMP[2].zw, TEMP[3]
> 21:   ADD TEMP[4].y, IMM[0].yyyy, -CONST[1].xxxx
> 22:   MOV TEMP[0].y, TEMP[4].yyyy
> 23:   MUL TEMP[5].z, TEMP[3].xxxx, IMM[0].wwww
> 24:   MAD TEMP[6].w, TEMP[3].wwww, IMM[0].wwww, IMM[0].xxxx
> 25:   FRC TEMP[7].z, TEMP[6].wwww
> 26:   ADD TEMP[6].w, TEMP[6].wwww, -TEMP[7].zzzz
> 27:   MAD TEMP[5].z, TEMP[6].wwww, IMM[1].xxxx, TEMP[5].zzzz
> 28:   ADD TEMP[5].z, TEMP[5].zzzz, IMM[0].xxxx
> 29:   FRC TEMP[6].w, TEMP[5].zzzz
> 30:   ADD TEMP[5].z, -TEMP[6].wwww, TEMP[5].zzzz
> 31:   MOV TEMP[0].z, TEMP[5].zzzz
> 32:   MAD TEMP[3].w, CONST[1].xxxx, TEMP[3].xxxx, IMM[0].xxxx
> 33:   FRC TEMP[5].z, TEMP[3].wwww
> 34:   MOV TEMP[1].z, TEMP[5].zzzz
> 35:   ADD TEMP[3].w, TEMP[3].wwww, -TEMP[5].zzzz
> 36:   MOV TEMP[0].w, TEMP[3].wwww
> 37:   ADD TEMP[3].zw, TEMP[0], IMM[0].xxxx
> 38:   FSGE TEMP[4].x, TEMP[4].yyyy, IMM[0].zzzz
> 39:   UIF TEMP[4].xxxx :0
> 40:     MOV TEMP[4].x, TEMP[3].wwww
> 41:   ELSE :0
> 42:     MOV TEMP[4].x, TEMP[3].zzzz
> 43:   ENDIF
> 44:   MOV TEMP[0].y, TEMP[4].xxxx
> 45:   RCP TEMP[3].x, CONST[0].xxxx
> 46:   MAD TEMP[3].x, CONST[3].xxxx, TEMP[3].xxxx, IMM[0].xxxx
> 47:   FRC TEMP[5].z, TEMP[3].xxxx
> 48:   ADD TEMP[3].x, -TEMP[5].zzzz, TEMP[3].xxxx
> 49:   MUL TEMP[5].z, TEMP[3].xxxx, TEMP[4].xxxx
> 50:   MOV TEMP[6].x, -TEMP[3].xxxx
> 51:   FSGE TEMP[5].x, TEMP[5].zzzz, IMM[0].zzzz
> 52:   UIF TEMP[5].xxxx :0
> 53:     MOV TEMP[5].x, TEMP[3].xxxx
> 54:   ELSE :0
> 55:     MOV TEMP[5].x, TEMP[6].xxxx
> 56:   ENDIF
> 57:   MOV TEMP[0].z, TEMP[5].xxxx
> 58:   RCP TEMP[5].x, TEMP[5].xxxx
> 59:   MUL TEMP[4].w, TEMP[5].xxxx, TEMP[4].xxxx
> 60:   FRC TEMP[4].w, TEMP[4].wwww
> 61:   MOV TEMP[0].w, TEMP[4].wwww
> 62:   RCP TEMP[0].x, TEMP[3].xxxx
> 63:   MUL TEMP[3].xy, TEMP[0].wxzw, TEMP[0].zyzw
> 64:   MOV TEMP[2].xy, TEMP[3].xyxx
> 65:   FRC TEMP[3].xy, TEMP[2]
> 66:   MOV TEMP[0].xy, TEMP[3].xyxx
> 67:   ADD TEMP[2].xy, -TEMP[0], TEMP[2]
> 68:   MOV TEMP[0].xy, TEMP[2].xyxx
> 69:   ADD TEMP[2].xy, TEMP[1], TEMP[0]
> 70:   MOV TEMP[0].xy, TEMP[2].xyxx
> 71:   MUL TEMP[2].xy, TEMP[0], CONST[0].xxxx
> 72:   MOV TEMP[0].xy, TEMP[2].xyxx
> 73:   FRC TEMP[2].zw, TEMP[2].xyxy
> 74:   MOV TEMP[0].zw, TEMP[2].wwzw
> 75:   ADD TEMP[2].xy, -TEMP[2].zwzw, TEMP[0]
> 76:   MOV TEMP[0].xy, TEMP[2].xyxx
> 77:   ADD TEMP[2].xy, TEMP[0], IMM[0].xxxx
> 78:   MOV TEMP[0].xy, TEMP[2].xyxx
> 79:   RCP TEMP[1].x, CONST[3].xxxx
> 80:   RCP TEMP[2].x, CONST[3].yyyy
> 81:   MOV TEMP[1].y, TEMP[2].xxxx
> 82:   MUL TEMP[1].xy, TEMP[0], TEMP[1]
> 83:   MOV TEMP[1].xy, TEMP[1].xyyy
> 84:   MOV TEMP[1].w, IMM[0].zzzz
> 85:   TXL TEMP[1], TEMP[1], SAMP[1], 2D
> 86:   MOV TEMP[0], TEMP[1]
> 87: ELSE :0
> 88:   MOV TEMP[1].xy, IN[1].xyyy
> 89:   TEX TEMP[1], TEMP[1], SAMP[1], 2D
> 90:   MOV TEMP[0], TEMP[1]
> 91: ENDIF
> 92: MUL TEMP[0], TEMP[0], IN[0]
> 93: MOV OUT[0], TEMP[0]
> 94: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %23 = load <16 x i8> addrspace(2)* %22, !tbaa !0
>  %24 = call float @llvm.SI.load.const(<16 x i8> %23, i32 0)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %23, i32 16)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %23, i32 32)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %23, i32 36)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %23, i32 48)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %23, i32 52)
>  %30 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %31 = load <8 x i32> addrspace(2)* %30, !tbaa !0
>  %32 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %33 = load <4 x i32> addrspace(2)* %32, !tbaa !0
>  %34 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 1
>  %35 = load <8 x i32> addrspace(2)* %34, !tbaa !0
>  %36 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 1
>  %37 = load <4 x i32> addrspace(2)* %36, !tbaa !0
>  %38 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %39 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %40 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %5, <2 x i32> %7)
>  %41 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %5, <2 x i32> %7)
>  %42 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %5, <2 x i32> %7)
>  %43 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %5, <2 x i32> %7)
>  %44 = fcmp olt float 5.000000e-01, %24
>  %45 = sext i1 %44 to i32
>  %46 = bitcast i32 %45 to float
>  %47 = bitcast float %46 to i32
>  %48 = icmp ne i32 %47, 0
>  br i1 %48, label %IF, label %ELSE
>
>IF:                                               ; preds = %main_body
>  %49 = fmul float %26, %42
>  %50 = fmul float %27, %43
>  %51 = call float @llvm.AMDIL.fraction.(float %49)
>  %52 = call float @llvm.AMDIL.fraction.(float %50)
>  %53 = fsub float -0.000000e+00, %51
>  %54 = fadd float %49, %53
>  %55 = fsub float -0.000000e+00, %52
>  %56 = fadd float %50, %55
>  %57 = fsub float -0.000000e+00, %54
>  %58 = fmul float %42, %26
>  %59 = fadd float %58, %57
>  %60 = fsub float -0.000000e+00, %56
>  %61 = fmul float %43, %27
>  %62 = fadd float %61, %60
>  %63 = fadd float %54, 5.000000e-01
>  %64 = fadd float %56, 5.000000e-01
>  %65 = fdiv float 1.000000e+00, %26
>  %66 = fdiv float 1.000000e+00, %27
>  %67 = fmul float %63, %65
>  %68 = fmul float %64, %66
>  %69 = bitcast float %67 to i32
>  %70 = bitcast float %68 to i32
>  %71 = bitcast float 0.000000e+00 to i32
>  %72 = insertelement <4 x i32> undef, i32 %69, i32 0
>  %73 = insertelement <4 x i32> %72, i32 %70, i32 1
>  %74 = insertelement <4 x i32> %73, i32 %71, i32 2
>  %75 = insertelement <4 x i32> %74, i32 undef, i32 3
>  %76 = bitcast <8 x i32> %31 to <32 x i8>
>  %77 = bitcast <4 x i32> %33 to <16 x i8>
>  %78 = call <4 x float> @llvm.SI.samplel.v4i32(<4 x i32> %75, <32 x i8> %76, <16 x i8> %77, i32 2)
>  %79 = extractelement <4 x float> %78, i32 0
>  %80 = extractelement <4 x float> %78, i32 3
>  %81 = fsub float -0.000000e+00, %25
>  %82 = fadd float 1.024000e+03, %81
>  %83 = fmul float %79, 2.550000e+02
>  %84 = fmul float %80, 2.550000e+02
>  %85 = fadd float %84, 5.000000e-01
>  %86 = call float @llvm.AMDIL.fraction.(float %85)
>  %87 = fsub float -0.000000e+00, %86
>  %88 = fadd float %85, %87
>  %89 = fmul float %88, 2.560000e+02
>  %90 = fadd float %89, %83
>  %91 = fadd float %90, 5.000000e-01
>  %92 = call float @llvm.AMDIL.fraction.(float %91)
>  %93 = fsub float -0.000000e+00, %92
>  %94 = fadd float %93, %91
>  %95 = fmul float %25, %79
>  %96 = fadd float %95, 5.000000e-01
>  %97 = call float @llvm.AMDIL.fraction.(float %96)
>  %98 = fsub float -0.000000e+00, %97
>  %99 = fadd float %96, %98
>  %100 = fadd float %94, 5.000000e-01
>  %101 = fadd float %99, 5.000000e-01
>  %102 = fcmp oge float %82, 0.000000e+00
>  %103 = sext i1 %102 to i32
>  %104 = bitcast i32 %103 to float
>  %105 = bitcast float %104 to i32
>  %106 = icmp ne i32 %105, 0
>  %. = select i1 %106, float %101, float %100
>  %107 = fdiv float 1.000000e+00, %24
>  %108 = fmul float %28, %107
>  %109 = fadd float %108, 5.000000e-01
>  %110 = call float @llvm.AMDIL.fraction.(float %109)
>  %111 = fsub float -0.000000e+00, %110
>  %112 = fadd float %111, %109
>  %113 = fmul float %112, %.
>  %114 = fsub float -0.000000e+00, %112
>  %115 = fcmp oge float %113, 0.000000e+00
>  %116 = sext i1 %115 to i32
>  %117 = bitcast i32 %116 to float
>  %118 = bitcast float %117 to i32
>  %119 = icmp ne i32 %118, 0
>  %temp20.0 = select i1 %119, float %112, float %114
>  %120 = fdiv float 1.000000e+00, %temp20.0
>  %121 = fmul float %120, %.
>  %122 = call float @llvm.AMDIL.fraction.(float %121)
>  %123 = fdiv float 1.000000e+00, %112
>  %124 = fmul float %122, %temp20.0
>  %125 = fmul float %123, %.
>  %126 = call float @llvm.AMDIL.fraction.(float %124)
>  %127 = call float @llvm.AMDIL.fraction.(float %125)
>  %128 = fsub float -0.000000e+00, %126
>  %129 = fadd float %128, %124
>  %130 = fsub float -0.000000e+00, %127
>  %131 = fadd float %130, %125
>  %132 = fadd float %59, %129
>  %133 = fadd float %62, %131
>  %134 = fmul float %132, %24
>  %135 = fmul float %133, %24
>  %136 = call float @llvm.AMDIL.fraction.(float %134)
>  %137 = call float @llvm.AMDIL.fraction.(float %135)
>  %138 = fsub float -0.000000e+00, %136
>  %139 = fadd float %138, %134
>  %140 = fsub float -0.000000e+00, %137
>  %141 = fadd float %140, %135
>  %142 = fadd float %139, 5.000000e-01
>  %143 = fadd float %141, 5.000000e-01
>  %144 = fdiv float 1.000000e+00, %28
>  %145 = fdiv float 1.000000e+00, %29
>  %146 = fmul float %142, %144
>  %147 = fmul float %143, %145
>  %148 = bitcast float %146 to i32
>  %149 = bitcast float %147 to i32
>  %150 = bitcast float 0.000000e+00 to i32
>  %151 = insertelement <4 x i32> undef, i32 %148, i32 0
>  %152 = insertelement <4 x i32> %151, i32 %149, i32 1
>  %153 = insertelement <4 x i32> %152, i32 %150, i32 2
>  %154 = insertelement <4 x i32> %153, i32 undef, i32 3
>  %155 = bitcast <8 x i32> %35 to <32 x i8>
>  %156 = bitcast <4 x i32> %37 to <16 x i8>
>  %157 = call <4 x float> @llvm.SI.samplel.v4i32(<4 x i32> %154, <32 x i8> %155, <16 x i8> %156, i32 2)
>  br label %ENDIF
>
>ELSE:                                             ; preds = %main_body
>  %158 = bitcast float %42 to i32
>  %159 = bitcast float %43 to i32
>  %160 = insertelement <2 x i32> undef, i32 %158, i32 0
>  %161 = insertelement <2 x i32> %160, i32 %159, i32 1
>  %162 = bitcast <8 x i32> %35 to <32 x i8>
>  %163 = bitcast <4 x i32> %37 to <16 x i8>
>  %164 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %161, <32 x i8> %162, <16 x i8> %163, i32 2)
>  br label %ENDIF
>
>ENDIF:                                            ; preds = %ELSE, %IF
>  %.sink = phi <4 x float> [ %164, %ELSE ], [ %157, %IF ]
>  %165 = extractelement <4 x float> %.sink, i32 0
>  %166 = extractelement <4 x float> %.sink, i32 1
>  %167 = extractelement <4 x float> %.sink, i32 2
>  %168 = extractelement <4 x float> %.sink, i32 3
>  %169 = fmul float %165, %38
>  %170 = fmul float %166, %39
>  %171 = fmul float %167, %40
>  %172 = fmul float %168, %41
>  %173 = call i32 @llvm.SI.packf16(float %169, float %170)
>  %174 = bitcast i32 %173 to float
>  %175 = call i32 @llvm.SI.packf16(float %171, float %172)
>  %176 = bitcast i32 %175 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %174, float %176, float %174, float %176)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.fraction.(float) #2
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.samplel.v4i32(<4 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                         ; BEFE0A7E
>	s_mov_b32 m0, s9                                                             ; BEFC0309
>	v_interp_p1_f32 v7, v0, 1, 1, [m0]                                           ; C81C0500
>	v_interp_p2_f32 v7, [v7], v1, 1, 1, [m0]                                     ; C81D0501
>	v_interp_p1_f32 v6, v0, 0, 1, [m0]                                           ; C8180400
>	v_interp_p2_f32 v6, [v6], v1, 0, 1, [m0]                                     ; C8190401
>	v_interp_p1_f32 v2, v0, 3, 0, [m0]                                           ; C8080300
>	v_interp_p2_f32 v2, [v2], v1, 3, 0, [m0]                                     ; C8090301
>	v_interp_p1_f32 v3, v0, 2, 0, [m0]                                           ; C80C0200
>	v_interp_p2_f32 v3, [v3], v1, 2, 0, [m0]                                     ; C80D0201
>	v_interp_p1_f32 v4, v0, 1, 0, [m0]                                           ; C8100100
>	v_interp_p2_f32 v4, [v4], v1, 1, 0, [m0]                                     ; C8110101
>	v_interp_p1_f32 v5, v0, 0, 0, [m0]                                           ; C8140000
>	v_interp_p2_f32 v5, [v5], v1, 0, 0, [m0]                                     ; C8150001
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                                           ; C0800300
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	s_buffer_load_dword s8, s[0:3], 0x0                                          ; C2040100
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_cmp_gt_f32_e64 s[10:11], s8, 0.5                                           ; D008000A 0001E008
>	v_cndmask_b32_e64 v0, 0, -1, s[10:11]                                        ; D2000000 00298280
>	v_cmp_eq_i32_e64 s[10:11], v0, 0                                             ; D104000A 00010100
>	s_load_dwordx4 s[12:15], s[4:5], 0x4                                         ; C0860504
>	s_load_dwordx8 s[16:23], s[6:7], 0x8                                         ; C0C80708
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	s_and_saveexec_b64 s[10:11], s[10:11]                                        ; BE8A240A
>	s_xor_b64 s[10:11], exec, s[10:11]                                           ; 898A0A7E
>	image_sample v[8:11], 15, 0, 0, 0, 0, 0, 0, 0, v[6:7], s[16:23], s[12:15]    ; F0800F00 00640806
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	s_or_saveexec_b64 s[10:11], s[10:11]                                         ; BE8A250A
>	s_xor_b64 exec, exec, s[10:11]                                               ; 89FE0A7E
>	s_cbranch_execz BB0_4                                                        ; BF880000
>	s_buffer_load_dword s9, s[0:3], 0xd                                          ; C204810D
>	s_buffer_load_dword s24, s[0:3], 0xc                                         ; C20C010C
>	s_buffer_load_dword s25, s[0:3], 0x9                                         ; C20C8109
>	s_buffer_load_dword s26, s[0:3], 0x8                                         ; C20D0108
>	s_buffer_load_dword s27, s[0:3], 0x4                                         ; C20D8104
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v0, s25, v7                                                    ; 10000E19
>	v_fract_f32_e32 v0, v0                                                       ; 7E004100
>	v_mad_f32 v0, s25, v7, -v0                                                   ; D2820000 84020E19
>	v_add_f32_e32 v1, 0.5, v0                                                    ; 060200F0
>	v_rcp_f32_e32 v8, s25                                                        ; 7E105419
>	v_mul_f32_e32 v9, v8, v1                                                     ; 10120308
>	v_mul_f32_e32 v1, s26, v6                                                    ; 10020C1A
>	v_fract_f32_e32 v1, v1                                                       ; 7E024101
>	v_mad_f32 v1, s26, v6, -v1                                                   ; D2820001 84060C1A
>	v_add_f32_e32 v12, 0.5, v1                                                   ; 061802F0
>	v_rcp_f32_e32 v13, s26                                                       ; 7E1A541A
>	v_mul_f32_e32 v8, v13, v12                                                   ; 1010190D
>	v_mov_b32_e32 v10, 0                                                         ; 7E140280
>	s_load_dwordx4 s[28:31], s[4:5], 0x0                                         ; C08E0500
>	s_load_dwordx8 s[32:39], s[6:7], 0x0                                         ; C0D00700
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample_l v[12:13], 9, 0, 0, 0, 0, 0, 0, 0, v[8:11], s[32:39], s[28:31] ; F0900900 00E80C08
>	v_mov_b32_e32 v14, 0x437f0000                                                ; 7E1C02FF 437F0000
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v14, v13, v14, 0.5                                                 ; D282000E 03C21D0D
>	v_fract_f32_e32 v15, v14                                                     ; 7E1E410E
>	v_subrev_f32_e32 v14, v15, v14                                               ; 0A1C1D0F
>	v_mul_f32_e32 v15, 0x437f0000, v12                                           ; 101E18FF 437F0000
>	v_mov_b32_e32 v16, 0x43800000                                                ; 7E2002FF 43800000
>	v_mad_f32 v14, v16, v14, v15                                                 ; D282000E 043E1D10
>	v_add_f32_e32 v14, 0.5, v14                                                  ; 061C1CF0
>	v_fract_f32_e32 v15, v14                                                     ; 7E1E410E
>	v_subrev_f32_e32 v14, v15, v14                                               ; 0A1C1D0F
>	v_add_f32_e32 v14, 0.5, v14                                                  ; 061C1CF0
>	v_mad_f32 v12, s27, v12, 0.5                                                 ; D282000C 03C2181B
>	v_fract_f32_e32 v13, v12                                                     ; 7E1A410C
>	v_subrev_f32_e32 v12, v13, v12                                               ; 0A18190D
>	v_add_f32_e32 v12, 0.5, v12                                                  ; 061818F0
>	v_mov_b32_e32 v13, 0x44800000                                                ; 7E1A02FF 44800000
>	v_subrev_f32_e32 v13, s27, v13                                               ; 0A1A1A1B
>	v_cmp_ge_f32_e64 s[28:29], v13, 0                                            ; D00C001C 0001010D
>	v_cndmask_b32_e64 v13, 0, -1, s[28:29]                                       ; D200080D 00718280
>	v_cmp_ne_i32_e64 s[28:29], v13, 0                                            ; D10A001C 0001010D
>	v_cndmask_b32_e64 v12, v14, v12, s[28:29]                                    ; D200000C 1072190E
>	v_rcp_f32_e32 v13, s8                                                        ; 7E1A5408
>	v_mad_f32 v13, s24, v13, 0.5                                                 ; D282000D 03C21A18
>	v_fract_f32_e32 v14, v13                                                     ; 7E1C410D
>	v_subrev_f32_e32 v13, v14, v13                                               ; 0A1A1B0E
>	v_mul_f32_e32 v14, v12, v13                                                  ; 101C1B0C
>	v_cmp_ge_f32_e64 s[28:29], v14, 0                                            ; D00C001C 0001010E
>	v_cndmask_b32_e64 v14, 0, -1, s[28:29]                                       ; D200000E 00718280
>	v_cmp_ne_i32_e64 s[28:29], v14, 0                                            ; D10A001C 0001010E
>	v_xor_b32_e32 v14, 0x80000000, v13                                           ; 3A1C1AFF 80000000
>	v_cndmask_b32_e64 v14, v14, v13, s[28:29]                                    ; D200000E 10721B0E
>	v_rcp_f32_e32 v15, v14                                                       ; 7E1E550E
>	v_mul_f32_e32 v15, v12, v15                                                  ; 101E1F0C
>	v_fract_f32_e32 v15, v15                                                     ; 7E1E410F
>	v_mul_f32_e32 v16, v14, v15                                                  ; 10201F0E
>	v_fract_f32_e32 v16, v16                                                     ; 7E204110
>	v_mad_f32 v14, v15, v14, -v16                                                ; D282000E 84421D0F
>	v_mad_f32 v1, s26, v6, -v1                                                   ; D2820001 84060C1A
>	v_add_f32_e32 v1, v14, v1                                                    ; 0602030E
>	v_mul_f32_e32 v14, s8, v1                                                    ; 101C0208
>	v_fract_f32_e32 v14, v14                                                     ; 7E1C410E
>	v_mad_f32 v1, v1, s8, -v14                                                   ; D2820001 84381101
>	v_add_f32_e32 v1, 0.5, v1                                                    ; 060202F0
>	v_rcp_f32_e32 v14, s24                                                       ; 7E1C5418
>	v_mul_f32_e32 v8, v14, v1                                                    ; 1010030E
>	v_rcp_f32_e32 v1, v13                                                        ; 7E02550D
>	v_mul_f32_e32 v13, v12, v1                                                   ; 101A030C
>	v_fract_f32_e32 v13, v13                                                     ; 7E1A410D
>	v_mad_f32 v1, v1, v12, -v13                                                  ; D2820001 84361901
>	v_mad_f32 v0, s25, v7, -v0                                                   ; D2820000 84020E19
>	v_add_f32_e32 v0, v1, v0                                                     ; 06000101
>	v_mul_f32_e32 v1, s8, v0                                                     ; 10020008
>	v_fract_f32_e32 v1, v1                                                       ; 7E024101
>	v_mad_f32 v0, v0, s8, -v1                                                    ; D2820000 84041100
>	v_add_f32_e32 v0, 0.5, v0                                                    ; 060000F0
>	v_rcp_f32_e32 v1, s9                                                         ; 7E025409
>	v_mul_f32_e32 v9, v1, v0                                                     ; 10120101
>	image_sample_l v[8:11], 15, 0, 0, 0, 0, 0, 0, 0, v[8:11], s[16:23], s[12:15] ; F0900F00 00640808
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	s_or_b64 exec, exec, s[10:11]                                                ; 88FE0A7E
>	v_mul_f32_e32 v0, v2, v11                                                    ; 10001702
>	v_mul_f32_e32 v1, v3, v10                                                    ; 10021503
>	v_cvt_pkrtz_f16_f32_e32 v0, v1, v0                                           ; 5E000101
>	v_mul_f32_e32 v1, v4, v9                                                     ; 10021304
>	v_mul_f32_e32 v2, v5, v8                                                     ; 10041105
>	v_cvt_pkrtz_f16_f32_e32 v1, v2, v1                                           ; 5E020302
>	exp 15, 0, 1, 1, 1, v1, v0, v1, v0                                           ; F8001C0F 00010001
>	s_endpgm                                                                     ; BF810000
>
>SGPRS: 40
>VGPRS: 20
>Code Size: 620 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL IN[2]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL OUT[2], GENERIC[10]
>DCL CONST[0..8]
>DCL TEMP[0..3], LOCAL
>IMM[0] FLT32 {    0.0000,     1.0000,     0.0000,     0.0000}
>  0: MAD TEMP[0].xy, IN[0], CONST[0], CONST[0].zwzw
>  1: MUL TEMP[1], TEMP[0].yyyy, CONST[6]
>  2: MAD TEMP[0], TEMP[0].xxxx, CONST[5], TEMP[1]
>  3: ADD TEMP[0], TEMP[0], CONST[8]
>  4: ADD TEMP[1].x, CONST[4].xxxx, CONST[4].yyyy
>  5: MUL TEMP[1], TEMP[0], TEMP[1].xxxx
>  6: MAD TEMP[2].xy, IN[1], CONST[1], CONST[1].zwzw
>  7: MOV TEMP[2].xy, TEMP[2].xyxx
>  8: MUL TEMP[0], CONST[2], IN[2]
>  9: MUL TEMP[3].w, TEMP[0].wwww, CONST[3].wwww
> 10: MOV TEMP[3].w, TEMP[3].wwww
> 11: MOV TEMP[3].xyz, TEMP[0].xyzx
> 12: MOV TEMP[2].zw, IMM[0].yyxy
> 13: MOV OUT[1], TEMP[3]
> 14: MOV OUT[2], TEMP[2]
> 15: MOV OUT[0], TEMP[1]
> 16: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 0)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 4)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %12, i32 8)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %12, i32 12)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %12, i32 16)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %12, i32 20)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %12, i32 24)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %12, i32 28)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %12, i32 32)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %12, i32 36)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %12, i32 40)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %12, i32 44)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %12, i32 60)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %12, i32 64)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %12, i32 68)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %12, i32 80)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %12, i32 84)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %12, i32 88)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %12, i32 92)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %12, i32 96)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %12, i32 100)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %12, i32 104)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %12, i32 108)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %12, i32 128)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %12, i32 132)
>  %38 = call float @llvm.SI.load.const(<16 x i8> %12, i32 136)
>  %39 = call float @llvm.SI.load.const(<16 x i8> %12, i32 140)
>  %40 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %41 = load <16 x i8> addrspace(2)* %40, !tbaa !0
>  %42 = add i32 %5, %7
>  %43 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %41, i32 0, i32 %42)
>  %44 = extractelement <4 x float> %43, i32 0
>  %45 = extractelement <4 x float> %43, i32 1
>  %46 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %47 = load <16 x i8> addrspace(2)* %46, !tbaa !0
>  %48 = add i32 %5, %7
>  %49 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %47, i32 0, i32 %48)
>  %50 = extractelement <4 x float> %49, i32 0
>  %51 = extractelement <4 x float> %49, i32 1
>  %52 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 2
>  %53 = load <16 x i8> addrspace(2)* %52, !tbaa !0
>  %54 = add i32 %5, %7
>  %55 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %53, i32 0, i32 %54)
>  %56 = extractelement <4 x float> %55, i32 0
>  %57 = extractelement <4 x float> %55, i32 1
>  %58 = extractelement <4 x float> %55, i32 2
>  %59 = extractelement <4 x float> %55, i32 3
>  %60 = fmul float %44, %13
>  %61 = fadd float %60, %15
>  %62 = fmul float %45, %14
>  %63 = fadd float %62, %16
>  %64 = fmul float %63, %32
>  %65 = fmul float %63, %33
>  %66 = fmul float %63, %34
>  %67 = fmul float %63, %35
>  %68 = fmul float %61, %28
>  %69 = fadd float %68, %64
>  %70 = fmul float %61, %29
>  %71 = fadd float %70, %65
>  %72 = fmul float %61, %30
>  %73 = fadd float %72, %66
>  %74 = fmul float %61, %31
>  %75 = fadd float %74, %67
>  %76 = fadd float %69, %36
>  %77 = fadd float %71, %37
>  %78 = fadd float %73, %38
>  %79 = fadd float %75, %39
>  %80 = fadd float %26, %27
>  %81 = fmul float %76, %80
>  %82 = fmul float %77, %80
>  %83 = fmul float %78, %80
>  %84 = fmul float %79, %80
>  %85 = fmul float %50, %17
>  %86 = fadd float %85, %19
>  %87 = fmul float %51, %18
>  %88 = fadd float %87, %20
>  %89 = fmul float %21, %56
>  %90 = fmul float %22, %57
>  %91 = fmul float %23, %58
>  %92 = fmul float %24, %59
>  %93 = fmul float %92, %25
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %89, float %90, float %91, float %93)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %86, float %88, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %81, float %82, float %83, float %84)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                             ; 4A00000A
>	s_load_dwordx4 s[4:7], s[8:9], 0x0                    ; C0820900
>	s_load_dwordx4 s[12:15], s[8:9], 0x4                  ; C0860904
>	s_load_dwordx4 s[8:11], s[8:9], 0x8                   ; C0840908
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[8:11], 0 idxen  ; E00C2000 80020100
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                    ; C0800300
>	s_waitcnt vmcnt(0) lgkmcnt(0)                         ; BF8C0070
>	s_buffer_load_dword s8, s[0:3], 0xa                   ; C204010A
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v5, s8, v3                              ; 100A0608
>	s_buffer_load_dword s8, s[0:3], 0x9                   ; C2040109
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v6, s8, v2                              ; 100C0408
>	s_buffer_load_dword s8, s[0:3], 0x8                   ; C2040108
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v7, s8, v1                              ; 100E0208
>	s_buffer_load_dword s8, s[0:3], 0xb                   ; C204010B
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v1, s8, v4                              ; 10020808
>	s_buffer_load_dword s8, s[0:3], 0xf                   ; C204010F
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v1, s8, v1                              ; 10020208
>	exp 15, 32, 0, 0, 0, v7, v6, v5, v1                   ; F800020F 01050607
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	buffer_load_format_xyzw v[1:4], v0, s[12:15], 0 idxen ; E00C2000 80030100
>	s_buffer_load_dword s8, s[0:3], 0x7                   ; C2040107
>	s_buffer_load_dword s9, s[0:3], 0x5                   ; C2048105
>	s_waitcnt vmcnt(0) lgkmcnt(0)                         ; BF8C0070
>	v_mov_b32_e32 v5, s8                                  ; 7E0A0208
>	v_mad_f32 v5, s9, v2, v5                              ; D2820005 04160409
>	s_buffer_load_dword s8, s[0:3], 0x6                   ; C2040106
>	s_buffer_load_dword s9, s[0:3], 0x4                   ; C2048104
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mov_b32_e32 v6, s8                                  ; 7E0C0208
>	v_mad_f32 v1, s9, v1, v6                              ; D2820001 041A0209
>	v_mov_b32_e32 v2, 1.0                                 ; 7E0402F2
>	v_mov_b32_e32 v3, 0                                   ; 7E060280
>	exp 15, 33, 0, 0, 0, v1, v5, v3, v2                   ; F800021F 02030501
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	buffer_load_format_xyzw v[0:3], v0, s[4:7], 0 idxen   ; E00C2000 80010000
>	s_buffer_load_dword s4, s[0:3], 0x2                   ; C2020102
>	s_buffer_load_dword s5, s[0:3], 0x0                   ; C2028100
>	s_waitcnt vmcnt(0) lgkmcnt(0)                         ; BF8C0070
>	v_mov_b32_e32 v4, s4                                  ; 7E080204
>	v_mad_f32 v4, s5, v0, v4                              ; D2820004 04120005
>	s_buffer_load_dword s4, s[0:3], 0x3                   ; C2020103
>	s_buffer_load_dword s5, s[0:3], 0x1                   ; C2028101
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mov_b32_e32 v5, s4                                  ; 7E0A0204
>	v_mad_f32 v0, s5, v1, v5                              ; D2820000 04160205
>	s_buffer_load_dword s4, s[0:3], 0x1b                  ; C202011B
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v1, s4, v0                              ; 10020004
>	s_buffer_load_dword s4, s[0:3], 0x17                  ; C2020117
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v1, v4, s4, v1                              ; D2820001 04040904
>	s_buffer_load_dword s4, s[0:3], 0x23                  ; C2020123
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v1, s4, v1                              ; 06020204
>	s_buffer_load_dword s4, s[0:3], 0x11                  ; C2020111
>	s_buffer_load_dword s5, s[0:3], 0x10                  ; C2028110
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mov_b32_e32 v2, s4                                  ; 7E040204
>	v_add_f32_e32 v2, s5, v2                              ; 06040405
>	v_mul_f32_e32 v1, v2, v1                              ; 10020302
>	s_buffer_load_dword s4, s[0:3], 0x1a                  ; C202011A
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v3, s4, v0                              ; 10060004
>	s_buffer_load_dword s4, s[0:3], 0x16                  ; C2020116
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v3, v4, s4, v3                              ; D2820003 040C0904
>	s_buffer_load_dword s4, s[0:3], 0x22                  ; C2020122
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v3, s4, v3                              ; 06060604
>	v_mul_f32_e32 v3, v2, v3                              ; 10060702
>	s_buffer_load_dword s4, s[0:3], 0x19                  ; C2020119
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v5, s4, v0                              ; 100A0004
>	s_buffer_load_dword s4, s[0:3], 0x15                  ; C2020115
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v5, v4, s4, v5                              ; D2820005 04140904
>	s_buffer_load_dword s4, s[0:3], 0x21                  ; C2020121
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v5, s4, v5                              ; 060A0A04
>	v_mul_f32_e32 v5, v2, v5                              ; 100A0B02
>	s_buffer_load_dword s4, s[0:3], 0x18                  ; C2020118
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v0, s4, v0                              ; 10000004
>	s_buffer_load_dword s4, s[0:3], 0x14                  ; C2020114
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v0, v4, s4, v0                              ; D2820000 04000904
>	s_buffer_load_dword s0, s[0:3], 0x20                  ; C2000120
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v0, s0, v0                              ; 06000000
>	v_mul_f32_e32 v0, v2, v0                              ; 10000102
>	exp 15, 12, 0, 1, 0, v0, v5, v3, v1                   ; F80008CF 01030500
>	s_endpgm                                              ; BF810000
>
>SGPRS: 24
>VGPRS: 8
>Code Size: 448 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL IN[1], GENERIC[10], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SAMP[1]
>DCL CONST[0..3]
>DCL TEMP[0..7], LOCAL
>IMM[0] FLT32 {    0.5000,  1024.0000,     0.0000,   255.0000}
>IMM[1] FLT32 {  256.0000,     0.0000,     0.0000,     0.0000}
>  0: MOV TEMP[0].xz, IMM[0].xxyx
>  1: FSLT TEMP[1].x, IMM[0].xxxx, CONST[0].xxxx
>  2: UIF TEMP[1].xxxx :0
>  3:   MUL TEMP[1].yw, CONST[2].xxzy, IN[1].xxzy
>  4:   MOV TEMP[0].yw, TEMP[1].wyww
>  5:   FRC TEMP[1].xy, TEMP[0].ywzw
>  6:   MOV TEMP[1].xy, TEMP[1].xyxx
>  7:   ADD TEMP[2].yw, TEMP[0], -TEMP[1].xxzy
>  8:   MOV TEMP[0].yw, TEMP[2].wyww
>  9:   MAD TEMP[2].xy, IN[1], CONST[2], -TEMP[0].ywzw
> 10:   MOV TEMP[1].xy, TEMP[2].xyxx
> 11:   ADD TEMP[2].yw, TEMP[0], IMM[0].xxxx
> 12:   MOV TEMP[0].yw, TEMP[2].wyww
> 13:   RCP TEMP[2].x, CONST[2].xxxx
> 14:   RCP TEMP[3].x, CONST[2].yyyy
> 15:   MOV TEMP[2].y, TEMP[3].xxxx
> 16:   MUL TEMP[3].xy, TEMP[0].ywzw, TEMP[2]
> 17:   MOV TEMP[3].xy, TEMP[3].xyyy
> 18:   MOV TEMP[3].w, IMM[0].zzzz
> 19:   TXL TEMP[3], TEMP[3], SAMP[0], 2D
> 20:   MOV TEMP[2].zw, TEMP[3]
> 21:   ADD TEMP[4].y, IMM[0].yyyy, -CONST[1].xxxx
> 22:   MOV TEMP[0].y, TEMP[4].yyyy
> 23:   MUL TEMP[5].z, TEMP[3].xxxx, IMM[0].wwww
> 24:   MAD TEMP[6].w, TEMP[3].wwww, IMM[0].wwww, IMM[0].xxxx
> 25:   FRC TEMP[7].z, TEMP[6].wwww
> 26:   ADD TEMP[6].w, TEMP[6].wwww, -TEMP[7].zzzz
> 27:   MAD TEMP[5].z, TEMP[6].wwww, IMM[1].xxxx, TEMP[5].zzzz
> 28:   ADD TEMP[5].z, TEMP[5].zzzz, IMM[0].xxxx
> 29:   FRC TEMP[6].w, TEMP[5].zzzz
> 30:   ADD TEMP[5].z, -TEMP[6].wwww, TEMP[5].zzzz
> 31:   MOV TEMP[0].z, TEMP[5].zzzz
> 32:   MAD TEMP[3].w, CONST[1].xxxx, TEMP[3].xxxx, IMM[0].xxxx
> 33:   FRC TEMP[5].z, TEMP[3].wwww
> 34:   MOV TEMP[1].z, TEMP[5].zzzz
> 35:   ADD TEMP[3].w, TEMP[3].wwww, -TEMP[5].zzzz
> 36:   MOV TEMP[0].w, TEMP[3].wwww
> 37:   ADD TEMP[3].zw, TEMP[0], IMM[0].xxxx
> 38:   FSGE TEMP[4].x, TEMP[4].yyyy, IMM[0].zzzz
> 39:   UIF TEMP[4].xxxx :0
> 40:     MOV TEMP[4].x, TEMP[3].wwww
> 41:   ELSE :0
> 42:     MOV TEMP[4].x, TEMP[3].zzzz
> 43:   ENDIF
> 44:   MOV TEMP[0].y, TEMP[4].xxxx
> 45:   RCP TEMP[3].x, CONST[0].xxxx
> 46:   MAD TEMP[3].x, CONST[3].xxxx, TEMP[3].xxxx, IMM[0].xxxx
> 47:   FRC TEMP[5].z, TEMP[3].xxxx
> 48:   ADD TEMP[3].x, -TEMP[5].zzzz, TEMP[3].xxxx
> 49:   MUL TEMP[5].z, TEMP[3].xxxx, TEMP[4].xxxx
> 50:   MOV TEMP[6].x, -TEMP[3].xxxx
> 51:   FSGE TEMP[5].x, TEMP[5].zzzz, IMM[0].zzzz
> 52:   UIF TEMP[5].xxxx :0
> 53:     MOV TEMP[5].x, TEMP[3].xxxx
> 54:   ELSE :0
> 55:     MOV TEMP[5].x, TEMP[6].xxxx
> 56:   ENDIF
> 57:   MOV TEMP[0].z, TEMP[5].xxxx
> 58:   RCP TEMP[5].x, TEMP[5].xxxx
> 59:   MUL TEMP[4].w, TEMP[5].xxxx, TEMP[4].xxxx
> 60:   FRC TEMP[4].w, TEMP[4].wwww
> 61:   MOV TEMP[0].w, TEMP[4].wwww
> 62:   RCP TEMP[0].x, TEMP[3].xxxx
> 63:   MUL TEMP[3].xy, TEMP[0].wxzw, TEMP[0].zyzw
> 64:   MOV TEMP[2].xy, TEMP[3].xyxx
> 65:   FRC TEMP[3].xy, TEMP[2]
> 66:   MOV TEMP[0].xy, TEMP[3].xyxx
> 67:   ADD TEMP[2].xy, -TEMP[0], TEMP[2]
> 68:   MOV TEMP[0].xy, TEMP[2].xyxx
> 69:   ADD TEMP[2].xy, TEMP[1], TEMP[0]
> 70:   MOV TEMP[0].xy, TEMP[2].xyxx
> 71:   MUL TEMP[2].xy, TEMP[0], CONST[0].xxxx
> 72:   MOV TEMP[0].xy, TEMP[2].xyxx
> 73:   FRC TEMP[2].zw, TEMP[2].xyxy
> 74:   MOV TEMP[0].zw, TEMP[2].wwzw
> 75:   ADD TEMP[2].xy, -TEMP[2].zwzw, TEMP[0]
> 76:   MOV TEMP[0].xy, TEMP[2].xyxx
> 77:   ADD TEMP[2].xy, TEMP[0], IMM[0].xxxx
> 78:   MOV TEMP[0].xy, TEMP[2].xyxx
> 79:   RCP TEMP[1].x, CONST[3].xxxx
> 80:   RCP TEMP[2].x, CONST[3].yyyy
> 81:   MOV TEMP[1].y, TEMP[2].xxxx
> 82:   MUL TEMP[1].xy, TEMP[0], TEMP[1]
> 83:   MOV TEMP[1].xy, TEMP[1].xyyy
> 84:   MOV TEMP[1].w, IMM[0].zzzz
> 85:   TXL TEMP[1], TEMP[1], SAMP[1], 2D
> 86:   MOV TEMP[0], TEMP[1]
> 87: ELSE :0
> 88:   MOV TEMP[1].xy, IN[1].xyyy
> 89:   TEX TEMP[1], TEMP[1], SAMP[1], 2D
> 90:   MOV TEMP[0], TEMP[1]
> 91: ENDIF
> 92: MUL TEMP[0], TEMP[0], IN[0]
> 93: MOV OUT[0], TEMP[0]
> 94: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %23 = load <16 x i8> addrspace(2)* %22, !tbaa !0
>  %24 = call float @llvm.SI.load.const(<16 x i8> %23, i32 0)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %23, i32 16)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %23, i32 32)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %23, i32 36)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %23, i32 48)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %23, i32 52)
>  %30 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %31 = load <8 x i32> addrspace(2)* %30, !tbaa !0
>  %32 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %33 = load <4 x i32> addrspace(2)* %32, !tbaa !0
>  %34 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 1
>  %35 = load <8 x i32> addrspace(2)* %34, !tbaa !0
>  %36 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 1
>  %37 = load <4 x i32> addrspace(2)* %36, !tbaa !0
>  %38 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %39 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %40 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %5, <2 x i32> %7)
>  %41 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %5, <2 x i32> %7)
>  %42 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %5, <2 x i32> %7)
>  %43 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %5, <2 x i32> %7)
>  %44 = fcmp olt float 5.000000e-01, %24
>  %45 = sext i1 %44 to i32
>  %46 = bitcast i32 %45 to float
>  %47 = bitcast float %46 to i32
>  %48 = icmp ne i32 %47, 0
>  br i1 %48, label %IF, label %ELSE
>
>IF:                                               ; preds = %main_body
>  %49 = fmul float %26, %42
>  %50 = fmul float %27, %43
>  %51 = call float @llvm.AMDIL.fraction.(float %49)
>  %52 = call float @llvm.AMDIL.fraction.(float %50)
>  %53 = fsub float -0.000000e+00, %51
>  %54 = fadd float %49, %53
>  %55 = fsub float -0.000000e+00, %52
>  %56 = fadd float %50, %55
>  %57 = fsub float -0.000000e+00, %54
>  %58 = fmul float %42, %26
>  %59 = fadd float %58, %57
>  %60 = fsub float -0.000000e+00, %56
>  %61 = fmul float %43, %27
>  %62 = fadd float %61, %60
>  %63 = fadd float %54, 5.000000e-01
>  %64 = fadd float %56, 5.000000e-01
>  %65 = fdiv float 1.000000e+00, %26
>  %66 = fdiv float 1.000000e+00, %27
>  %67 = fmul float %63, %65
>  %68 = fmul float %64, %66
>  %69 = bitcast float %67 to i32
>  %70 = bitcast float %68 to i32
>  %71 = bitcast float 0.000000e+00 to i32
>  %72 = insertelement <4 x i32> undef, i32 %69, i32 0
>  %73 = insertelement <4 x i32> %72, i32 %70, i32 1
>  %74 = insertelement <4 x i32> %73, i32 %71, i32 2
>  %75 = insertelement <4 x i32> %74, i32 undef, i32 3
>  %76 = bitcast <8 x i32> %31 to <32 x i8>
>  %77 = bitcast <4 x i32> %33 to <16 x i8>
>  %78 = call <4 x float> @llvm.SI.samplel.v4i32(<4 x i32> %75, <32 x i8> %76, <16 x i8> %77, i32 2)
>  %79 = extractelement <4 x float> %78, i32 0
>  %80 = extractelement <4 x float> %78, i32 3
>  %81 = fsub float -0.000000e+00, %25
>  %82 = fadd float 1.024000e+03, %81
>  %83 = fmul float %79, 2.550000e+02
>  %84 = fmul float %80, 2.550000e+02
>  %85 = fadd float %84, 5.000000e-01
>  %86 = call float @llvm.AMDIL.fraction.(float %85)
>  %87 = fsub float -0.000000e+00, %86
>  %88 = fadd float %85, %87
>  %89 = fmul float %88, 2.560000e+02
>  %90 = fadd float %89, %83
>  %91 = fadd float %90, 5.000000e-01
>  %92 = call float @llvm.AMDIL.fraction.(float %91)
>  %93 = fsub float -0.000000e+00, %92
>  %94 = fadd float %93, %91
>  %95 = fmul float %25, %79
>  %96 = fadd float %95, 5.000000e-01
>  %97 = call float @llvm.AMDIL.fraction.(float %96)
>  %98 = fsub float -0.000000e+00, %97
>  %99 = fadd float %96, %98
>  %100 = fadd float %94, 5.000000e-01
>  %101 = fadd float %99, 5.000000e-01
>  %102 = fcmp oge float %82, 0.000000e+00
>  %103 = sext i1 %102 to i32
>  %104 = bitcast i32 %103 to float
>  %105 = bitcast float %104 to i32
>  %106 = icmp ne i32 %105, 0
>  %. = select i1 %106, float %101, float %100
>  %107 = fdiv float 1.000000e+00, %24
>  %108 = fmul float %28, %107
>  %109 = fadd float %108, 5.000000e-01
>  %110 = call float @llvm.AMDIL.fraction.(float %109)
>  %111 = fsub float -0.000000e+00, %110
>  %112 = fadd float %111, %109
>  %113 = fmul float %112, %.
>  %114 = fsub float -0.000000e+00, %112
>  %115 = fcmp oge float %113, 0.000000e+00
>  %116 = sext i1 %115 to i32
>  %117 = bitcast i32 %116 to float
>  %118 = bitcast float %117 to i32
>  %119 = icmp ne i32 %118, 0
>  %temp20.0 = select i1 %119, float %112, float %114
>  %120 = fdiv float 1.000000e+00, %temp20.0
>  %121 = fmul float %120, %.
>  %122 = call float @llvm.AMDIL.fraction.(float %121)
>  %123 = fdiv float 1.000000e+00, %112
>  %124 = fmul float %122, %temp20.0
>  %125 = fmul float %123, %.
>  %126 = call float @llvm.AMDIL.fraction.(float %124)
>  %127 = call float @llvm.AMDIL.fraction.(float %125)
>  %128 = fsub float -0.000000e+00, %126
>  %129 = fadd float %128, %124
>  %130 = fsub float -0.000000e+00, %127
>  %131 = fadd float %130, %125
>  %132 = fadd float %59, %129
>  %133 = fadd float %62, %131
>  %134 = fmul float %132, %24
>  %135 = fmul float %133, %24
>  %136 = call float @llvm.AMDIL.fraction.(float %134)
>  %137 = call float @llvm.AMDIL.fraction.(float %135)
>  %138 = fsub float -0.000000e+00, %136
>  %139 = fadd float %138, %134
>  %140 = fsub float -0.000000e+00, %137
>  %141 = fadd float %140, %135
>  %142 = fadd float %139, 5.000000e-01
>  %143 = fadd float %141, 5.000000e-01
>  %144 = fdiv float 1.000000e+00, %28
>  %145 = fdiv float 1.000000e+00, %29
>  %146 = fmul float %142, %144
>  %147 = fmul float %143, %145
>  %148 = bitcast float %146 to i32
>  %149 = bitcast float %147 to i32
>  %150 = bitcast float 0.000000e+00 to i32
>  %151 = insertelement <4 x i32> undef, i32 %148, i32 0
>  %152 = insertelement <4 x i32> %151, i32 %149, i32 1
>  %153 = insertelement <4 x i32> %152, i32 %150, i32 2
>  %154 = insertelement <4 x i32> %153, i32 undef, i32 3
>  %155 = bitcast <8 x i32> %35 to <32 x i8>
>  %156 = bitcast <4 x i32> %37 to <16 x i8>
>  %157 = call <4 x float> @llvm.SI.samplel.v4i32(<4 x i32> %154, <32 x i8> %155, <16 x i8> %156, i32 2)
>  br label %ENDIF
>
>ELSE:                                             ; preds = %main_body
>  %158 = bitcast float %42 to i32
>  %159 = bitcast float %43 to i32
>  %160 = insertelement <2 x i32> undef, i32 %158, i32 0
>  %161 = insertelement <2 x i32> %160, i32 %159, i32 1
>  %162 = bitcast <8 x i32> %35 to <32 x i8>
>  %163 = bitcast <4 x i32> %37 to <16 x i8>
>  %164 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %161, <32 x i8> %162, <16 x i8> %163, i32 2)
>  br label %ENDIF
>
>ENDIF:                                            ; preds = %ELSE, %IF
>  %.sink = phi <4 x float> [ %164, %ELSE ], [ %157, %IF ]
>  %165 = extractelement <4 x float> %.sink, i32 0
>  %166 = extractelement <4 x float> %.sink, i32 1
>  %167 = extractelement <4 x float> %.sink, i32 2
>  %168 = extractelement <4 x float> %.sink, i32 3
>  %169 = fmul float %165, %38
>  %170 = fmul float %166, %39
>  %171 = fmul float %167, %40
>  %172 = fmul float %168, %41
>  %173 = call i32 @llvm.SI.packf16(float %169, float %170)
>  %174 = bitcast i32 %173 to float
>  %175 = call i32 @llvm.SI.packf16(float %171, float %172)
>  %176 = bitcast i32 %175 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %174, float %176, float %174, float %176)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.fraction.(float) #2
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.samplel.v4i32(<4 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                         ; BEFE0A7E
>	s_mov_b32 m0, s9                                                             ; BEFC0309
>	v_interp_p1_f32 v7, v0, 1, 1, [m0]                                           ; C81C0500
>	v_interp_p2_f32 v7, [v7], v1, 1, 1, [m0]                                     ; C81D0501
>	v_interp_p1_f32 v6, v0, 0, 1, [m0]                                           ; C8180400
>	v_interp_p2_f32 v6, [v6], v1, 0, 1, [m0]                                     ; C8190401
>	v_interp_p1_f32 v2, v0, 3, 0, [m0]                                           ; C8080300
>	v_interp_p2_f32 v2, [v2], v1, 3, 0, [m0]                                     ; C8090301
>	v_interp_p1_f32 v3, v0, 2, 0, [m0]                                           ; C80C0200
>	v_interp_p2_f32 v3, [v3], v1, 2, 0, [m0]                                     ; C80D0201
>	v_interp_p1_f32 v4, v0, 1, 0, [m0]                                           ; C8100100
>	v_interp_p2_f32 v4, [v4], v1, 1, 0, [m0]                                     ; C8110101
>	v_interp_p1_f32 v5, v0, 0, 0, [m0]                                           ; C8140000
>	v_interp_p2_f32 v5, [v5], v1, 0, 0, [m0]                                     ; C8150001
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                                           ; C0800300
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	s_buffer_load_dword s8, s[0:3], 0x0                                          ; C2040100
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_cmp_gt_f32_e64 s[10:11], s8, 0.5                                           ; D008000A 0001E008
>	v_cndmask_b32_e64 v0, 0, -1, s[10:11]                                        ; D2000000 00298280
>	v_cmp_eq_i32_e64 s[10:11], v0, 0                                             ; D104000A 00010100
>	s_load_dwordx4 s[12:15], s[4:5], 0x4                                         ; C0860504
>	s_load_dwordx8 s[16:23], s[6:7], 0x8                                         ; C0C80708
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	s_and_saveexec_b64 s[10:11], s[10:11]                                        ; BE8A240A
>	s_xor_b64 s[10:11], exec, s[10:11]                                           ; 898A0A7E
>	image_sample v[8:11], 15, 0, 0, 0, 0, 0, 0, 0, v[6:7], s[16:23], s[12:15]    ; F0800F00 00640806
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	s_or_saveexec_b64 s[10:11], s[10:11]                                         ; BE8A250A
>	s_xor_b64 exec, exec, s[10:11]                                               ; 89FE0A7E
>	s_cbranch_execz BB0_4                                                        ; BF880000
>	s_buffer_load_dword s9, s[0:3], 0xd                                          ; C204810D
>	s_buffer_load_dword s24, s[0:3], 0xc                                         ; C20C010C
>	s_buffer_load_dword s25, s[0:3], 0x9                                         ; C20C8109
>	s_buffer_load_dword s26, s[0:3], 0x8                                         ; C20D0108
>	s_buffer_load_dword s27, s[0:3], 0x4                                         ; C20D8104
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v0, s25, v7                                                    ; 10000E19
>	v_fract_f32_e32 v0, v0                                                       ; 7E004100
>	v_mad_f32 v0, s25, v7, -v0                                                   ; D2820000 84020E19
>	v_add_f32_e32 v1, 0.5, v0                                                    ; 060200F0
>	v_rcp_f32_e32 v8, s25                                                        ; 7E105419
>	v_mul_f32_e32 v9, v8, v1                                                     ; 10120308
>	v_mul_f32_e32 v1, s26, v6                                                    ; 10020C1A
>	v_fract_f32_e32 v1, v1                                                       ; 7E024101
>	v_mad_f32 v1, s26, v6, -v1                                                   ; D2820001 84060C1A
>	v_add_f32_e32 v12, 0.5, v1                                                   ; 061802F0
>	v_rcp_f32_e32 v13, s26                                                       ; 7E1A541A
>	v_mul_f32_e32 v8, v13, v12                                                   ; 1010190D
>	v_mov_b32_e32 v10, 0                                                         ; 7E140280
>	s_load_dwordx4 s[28:31], s[4:5], 0x0                                         ; C08E0500
>	s_load_dwordx8 s[32:39], s[6:7], 0x0                                         ; C0D00700
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample_l v[12:13], 9, 0, 0, 0, 0, 0, 0, 0, v[8:11], s[32:39], s[28:31] ; F0900900 00E80C08
>	v_mov_b32_e32 v14, 0x437f0000                                                ; 7E1C02FF 437F0000
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v14, v13, v14, 0.5                                                 ; D282000E 03C21D0D
>	v_fract_f32_e32 v15, v14                                                     ; 7E1E410E
>	v_subrev_f32_e32 v14, v15, v14                                               ; 0A1C1D0F
>	v_mul_f32_e32 v15, 0x437f0000, v12                                           ; 101E18FF 437F0000
>	v_mov_b32_e32 v16, 0x43800000                                                ; 7E2002FF 43800000
>	v_mad_f32 v14, v16, v14, v15                                                 ; D282000E 043E1D10
>	v_add_f32_e32 v14, 0.5, v14                                                  ; 061C1CF0
>	v_fract_f32_e32 v15, v14                                                     ; 7E1E410E
>	v_subrev_f32_e32 v14, v15, v14                                               ; 0A1C1D0F
>	v_add_f32_e32 v14, 0.5, v14                                                  ; 061C1CF0
>	v_mad_f32 v12, s27, v12, 0.5                                                 ; D282000C 03C2181B
>	v_fract_f32_e32 v13, v12                                                     ; 7E1A410C
>	v_subrev_f32_e32 v12, v13, v12                                               ; 0A18190D
>	v_add_f32_e32 v12, 0.5, v12                                                  ; 061818F0
>	v_mov_b32_e32 v13, 0x44800000                                                ; 7E1A02FF 44800000
>	v_subrev_f32_e32 v13, s27, v13                                               ; 0A1A1A1B
>	v_cmp_ge_f32_e64 s[28:29], v13, 0                                            ; D00C001C 0001010D
>	v_cndmask_b32_e64 v13, 0, -1, s[28:29]                                       ; D200080D 00718280
>	v_cmp_ne_i32_e64 s[28:29], v13, 0                                            ; D10A001C 0001010D
>	v_cndmask_b32_e64 v12, v14, v12, s[28:29]                                    ; D200000C 1072190E
>	v_rcp_f32_e32 v13, s8                                                        ; 7E1A5408
>	v_mad_f32 v13, s24, v13, 0.5                                                 ; D282000D 03C21A18
>	v_fract_f32_e32 v14, v13                                                     ; 7E1C410D
>	v_subrev_f32_e32 v13, v14, v13                                               ; 0A1A1B0E
>	v_mul_f32_e32 v14, v12, v13                                                  ; 101C1B0C
>	v_cmp_ge_f32_e64 s[28:29], v14, 0                                            ; D00C001C 0001010E
>	v_cndmask_b32_e64 v14, 0, -1, s[28:29]                                       ; D200000E 00718280
>	v_cmp_ne_i32_e64 s[28:29], v14, 0                                            ; D10A001C 0001010E
>	v_xor_b32_e32 v14, 0x80000000, v13                                           ; 3A1C1AFF 80000000
>	v_cndmask_b32_e64 v14, v14, v13, s[28:29]                                    ; D200000E 10721B0E
>	v_rcp_f32_e32 v15, v14                                                       ; 7E1E550E
>	v_mul_f32_e32 v15, v12, v15                                                  ; 101E1F0C
>	v_fract_f32_e32 v15, v15                                                     ; 7E1E410F
>	v_mul_f32_e32 v16, v14, v15                                                  ; 10201F0E
>	v_fract_f32_e32 v16, v16                                                     ; 7E204110
>	v_mad_f32 v14, v15, v14, -v16                                                ; D282000E 84421D0F
>	v_mad_f32 v1, s26, v6, -v1                                                   ; D2820001 84060C1A
>	v_add_f32_e32 v1, v14, v1                                                    ; 0602030E
>	v_mul_f32_e32 v14, s8, v1                                                    ; 101C0208
>	v_fract_f32_e32 v14, v14                                                     ; 7E1C410E
>	v_mad_f32 v1, v1, s8, -v14                                                   ; D2820001 84381101
>	v_add_f32_e32 v1, 0.5, v1                                                    ; 060202F0
>	v_rcp_f32_e32 v14, s24                                                       ; 7E1C5418
>	v_mul_f32_e32 v8, v14, v1                                                    ; 1010030E
>	v_rcp_f32_e32 v1, v13                                                        ; 7E02550D
>	v_mul_f32_e32 v13, v12, v1                                                   ; 101A030C
>	v_fract_f32_e32 v13, v13                                                     ; 7E1A410D
>	v_mad_f32 v1, v1, v12, -v13                                                  ; D2820001 84361901
>	v_mad_f32 v0, s25, v7, -v0                                                   ; D2820000 84020E19
>	v_add_f32_e32 v0, v1, v0                                                     ; 06000101
>	v_mul_f32_e32 v1, s8, v0                                                     ; 10020008
>	v_fract_f32_e32 v1, v1                                                       ; 7E024101
>	v_mad_f32 v0, v0, s8, -v1                                                    ; D2820000 84041100
>	v_add_f32_e32 v0, 0.5, v0                                                    ; 060000F0
>	v_rcp_f32_e32 v1, s9                                                         ; 7E025409
>	v_mul_f32_e32 v9, v1, v0                                                     ; 10120101
>	image_sample_l v[8:11], 15, 0, 0, 0, 0, 0, 0, 0, v[8:11], s[16:23], s[12:15] ; F0900F00 00640808
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	s_or_b64 exec, exec, s[10:11]                                                ; 88FE0A7E
>	v_mul_f32_e32 v0, v2, v11                                                    ; 10001702
>	v_mul_f32_e32 v1, v3, v10                                                    ; 10021503
>	v_cvt_pkrtz_f16_f32_e32 v0, v1, v0                                           ; 5E000101
>	v_mul_f32_e32 v1, v4, v9                                                     ; 10021304
>	v_mul_f32_e32 v2, v5, v8                                                     ; 10041105
>	v_cvt_pkrtz_f16_f32_e32 v1, v2, v1                                           ; 5E020302
>	exp 15, 0, 1, 1, 1, v1, v0, v1, v0                                           ; F8001C0F 00010001
>	s_endpgm                                                                     ; BF810000
>
>SGPRS: 40
>VGPRS: 20
>Code Size: 620 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>number - table
>/media/ssd_plain/sliedes/_steam/steam/SteamApps/common/Sid Meier's Civilization V/steamassets/controller.vdf
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL CONST[0..7]
>DCL TEMP[0..2], LOCAL
>  0: MAD TEMP[0].xy, IN[0], CONST[0], CONST[0].zwzw
>  1: MUL TEMP[1], TEMP[0].yyyy, CONST[5]
>  2: MAD TEMP[0], TEMP[0].xxxx, CONST[4], TEMP[1]
>  3: ADD TEMP[0], TEMP[0], CONST[7]
>  4: ADD TEMP[1].x, CONST[3].xxxx, CONST[3].yyyy
>  5: MUL TEMP[1], TEMP[0], TEMP[1].xxxx
>  6: MUL TEMP[0], CONST[1], IN[1]
>  7: MUL TEMP[2].w, TEMP[0].wwww, CONST[2].wwww
>  8: MOV TEMP[2].w, TEMP[2].wwww
>  9: MOV TEMP[2].xyz, TEMP[0].xyzx
> 10: MOV OUT[1], TEMP[2]
> 11: MOV OUT[0], TEMP[1]
> 12: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 0)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 4)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %12, i32 8)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %12, i32 12)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %12, i32 16)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %12, i32 20)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %12, i32 24)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %12, i32 28)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %12, i32 44)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %12, i32 48)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %12, i32 52)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %12, i32 64)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %12, i32 68)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %12, i32 72)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %12, i32 76)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %12, i32 80)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %12, i32 84)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %12, i32 88)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %12, i32 92)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %12, i32 112)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %12, i32 116)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %12, i32 120)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %12, i32 124)
>  %36 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %37 = load <16 x i8> addrspace(2)* %36, !tbaa !0
>  %38 = add i32 %5, %7
>  %39 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %37, i32 0, i32 %38)
>  %40 = extractelement <4 x float> %39, i32 0
>  %41 = extractelement <4 x float> %39, i32 1
>  %42 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %43 = load <16 x i8> addrspace(2)* %42, !tbaa !0
>  %44 = add i32 %5, %7
>  %45 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %43, i32 0, i32 %44)
>  %46 = extractelement <4 x float> %45, i32 0
>  %47 = extractelement <4 x float> %45, i32 1
>  %48 = extractelement <4 x float> %45, i32 2
>  %49 = extractelement <4 x float> %45, i32 3
>  %50 = fmul float %40, %13
>  %51 = fadd float %50, %15
>  %52 = fmul float %41, %14
>  %53 = fadd float %52, %16
>  %54 = fmul float %53, %28
>  %55 = fmul float %53, %29
>  %56 = fmul float %53, %30
>  %57 = fmul float %53, %31
>  %58 = fmul float %51, %24
>  %59 = fadd float %58, %54
>  %60 = fmul float %51, %25
>  %61 = fadd float %60, %55
>  %62 = fmul float %51, %26
>  %63 = fadd float %62, %56
>  %64 = fmul float %51, %27
>  %65 = fadd float %64, %57
>  %66 = fadd float %59, %32
>  %67 = fadd float %61, %33
>  %68 = fadd float %63, %34
>  %69 = fadd float %65, %35
>  %70 = fadd float %22, %23
>  %71 = fmul float %66, %70
>  %72 = fmul float %67, %70
>  %73 = fmul float %68, %70
>  %74 = fmul float %69, %70
>  %75 = fmul float %17, %46
>  %76 = fmul float %18, %47
>  %77 = fmul float %19, %48
>  %78 = fmul float %20, %49
>  %79 = fmul float %78, %21
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %75, float %76, float %77, float %79)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %71, float %72, float %73, float %74)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                            ; 4A00000A
>	s_load_dwordx4 s[4:7], s[8:9], 0x0                   ; C0820900
>	s_load_dwordx4 s[8:11], s[8:9], 0x4                  ; C0840904
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[8:11], 0 idxen ; E00C2000 80020100
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                   ; C0800300
>	s_waitcnt vmcnt(0) lgkmcnt(0)                        ; BF8C0070
>	s_buffer_load_dword s8, s[0:3], 0x6                  ; C2040106
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v5, s8, v3                             ; 100A0608
>	s_buffer_load_dword s8, s[0:3], 0x5                  ; C2040105
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v6, s8, v2                             ; 100C0408
>	s_buffer_load_dword s8, s[0:3], 0x4                  ; C2040104
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v7, s8, v1                             ; 100E0208
>	s_buffer_load_dword s8, s[0:3], 0x7                  ; C2040107
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v1, s8, v4                             ; 10020808
>	s_buffer_load_dword s8, s[0:3], 0xb                  ; C204010B
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v1, s8, v1                             ; 10020208
>	exp 15, 32, 0, 0, 0, v7, v6, v5, v1                  ; F800020F 01050607
>	s_waitcnt expcnt(0)                                  ; BF8C070F
>	buffer_load_format_xyzw v[0:3], v0, s[4:7], 0 idxen  ; E00C2000 80010000
>	s_buffer_load_dword s4, s[0:3], 0x2                  ; C2020102
>	s_buffer_load_dword s5, s[0:3], 0x0                  ; C2028100
>	s_waitcnt vmcnt(0) lgkmcnt(0)                        ; BF8C0070
>	v_mov_b32_e32 v4, s4                                 ; 7E080204
>	v_mad_f32 v4, s5, v0, v4                             ; D2820004 04120005
>	s_buffer_load_dword s4, s[0:3], 0x3                  ; C2020103
>	s_buffer_load_dword s5, s[0:3], 0x1                  ; C2028101
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mov_b32_e32 v5, s4                                 ; 7E0A0204
>	v_mad_f32 v0, s5, v1, v5                             ; D2820000 04160205
>	s_buffer_load_dword s4, s[0:3], 0x17                 ; C2020117
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v1, s4, v0                             ; 10020004
>	s_buffer_load_dword s4, s[0:3], 0x13                 ; C2020113
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v1, v4, s4, v1                             ; D2820001 04040904
>	s_buffer_load_dword s4, s[0:3], 0x1f                 ; C202011F
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_add_f32_e32 v1, s4, v1                             ; 06020204
>	s_buffer_load_dword s4, s[0:3], 0xd                  ; C202010D
>	s_buffer_load_dword s5, s[0:3], 0xc                  ; C202810C
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mov_b32_e32 v2, s4                                 ; 7E040204
>	v_add_f32_e32 v2, s5, v2                             ; 06040405
>	v_mul_f32_e32 v1, v2, v1                             ; 10020302
>	s_buffer_load_dword s4, s[0:3], 0x16                 ; C2020116
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v3, s4, v0                             ; 10060004
>	s_buffer_load_dword s4, s[0:3], 0x12                 ; C2020112
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v3, v4, s4, v3                             ; D2820003 040C0904
>	s_buffer_load_dword s4, s[0:3], 0x1e                 ; C202011E
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_add_f32_e32 v3, s4, v3                             ; 06060604
>	v_mul_f32_e32 v3, v2, v3                             ; 10060702
>	s_buffer_load_dword s4, s[0:3], 0x15                 ; C2020115
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v5, s4, v0                             ; 100A0004
>	s_buffer_load_dword s4, s[0:3], 0x11                 ; C2020111
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v5, v4, s4, v5                             ; D2820005 04140904
>	s_buffer_load_dword s4, s[0:3], 0x1d                 ; C202011D
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_add_f32_e32 v5, s4, v5                             ; 060A0A04
>	v_mul_f32_e32 v5, v2, v5                             ; 100A0B02
>	s_buffer_load_dword s4, s[0:3], 0x14                 ; C2020114
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v0, s4, v0                             ; 10000004
>	s_buffer_load_dword s4, s[0:3], 0x10                 ; C2020110
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v0, v4, s4, v0                             ; D2820000 04000904
>	s_buffer_load_dword s0, s[0:3], 0x1c                 ; C200011C
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_add_f32_e32 v0, s0, v0                             ; 06000000
>	v_mul_f32_e32 v0, v2, v0                             ; 10000102
>	exp 15, 12, 0, 1, 0, v0, v5, v3, v1                  ; F80008CF 01030500
>	s_endpgm                                             ; BF810000
>
>SGPRS: 16
>VGPRS: 8
>Code Size: 368 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL OUT[0], COLOR
>  0: MOV OUT[0], IN[0]
>  1: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %23 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %24 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %5, <2 x i32> %7)
>  %25 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %5, <2 x i32> %7)
>  %26 = call i32 @llvm.SI.packf16(float %22, float %23)
>  %27 = bitcast i32 %26 to float
>  %28 = call i32 @llvm.SI.packf16(float %24, float %25)
>  %29 = bitcast i32 %28 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %27, float %29, float %27, float %29)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                     ; BEFE0A7E
>	s_mov_b32 m0, s9                         ; BEFC0309
>	v_interp_p1_f32 v2, v0, 3, 0, [m0]       ; C8080300
>	v_interp_p2_f32 v2, [v2], v1, 3, 0, [m0] ; C8090301
>	v_interp_p1_f32 v3, v0, 2, 0, [m0]       ; C80C0200
>	v_interp_p2_f32 v3, [v3], v1, 2, 0, [m0] ; C80D0201
>	v_cvt_pkrtz_f16_f32_e32 v2, v3, v2       ; 5E040503
>	v_interp_p1_f32 v3, v0, 1, 0, [m0]       ; C80C0100
>	v_interp_p2_f32 v3, [v3], v1, 1, 0, [m0] ; C80D0101
>	v_interp_p1_f32 v4, v0, 0, 0, [m0]       ; C8100000
>	v_interp_p2_f32 v4, [v4], v1, 0, 0, [m0] ; C8110001
>	v_cvt_pkrtz_f16_f32_e32 v0, v4, v3       ; 5E000704
>	exp 15, 0, 1, 1, 1, v0, v2, v0, v2       ; F8001C0F 02000200
>	s_endpgm                                 ; BF810000
>
>SGPRS: 16
>VGPRS: 8
>Code Size: 60 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL IN[2]
>DCL IN[3]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL OUT[2], GENERIC[10]
>DCL OUT[3], GENERIC[11]
>DCL CONST[0..24]
>DCL TEMP[0..3], LOCAL
>DCL ADDR[0]
>IMM[0] FLT32 {    0.0000,     1.0000,     0.0000,     0.0000}
>  0: F2I TEMP[0].x, IN[3].xxxx
>  1: UARL ADDR[0].x, TEMP[0].xxxx
>  2: UARL ADDR[0].x, TEMP[0].xxxx
>  3: MOV TEMP[1], CONST[ADDR[0].x+4].zwzw
>  4: UARL ADDR[0].x, TEMP[0].xxxx
>  5: MAD TEMP[1].xy, IN[0], CONST[ADDR[0].x+4], TEMP[1]
>  6: UARL ADDR[0].x, TEMP[0].xxxx
>  7: UARL ADDR[0].x, TEMP[0].xxxx
>  8: MOV TEMP[2], CONST[ADDR[0].x+13].zwzw
>  9: UARL ADDR[0].x, TEMP[0].xxxx
> 10: MAD TEMP[2].xy, IN[1], CONST[ADDR[0].x+13], TEMP[2]
> 11: MOV TEMP[2].xy, TEMP[2].xyxx
> 12: MUL TEMP[3], TEMP[1].yyyy, CONST[1]
> 13: MAD TEMP[0], TEMP[1].xxxx, CONST[0], TEMP[3]
> 14: ADD TEMP[0], TEMP[0], CONST[3]
> 15: ADD TEMP[1].x, CONST[24].xxxx, CONST[24].yyyy
> 16: MUL TEMP[1], TEMP[0], TEMP[1].xxxx
> 17: MUL TEMP[0], CONST[22], IN[2]
> 18: MUL TEMP[3].w, TEMP[0].wwww, CONST[23].wwww
> 19: MOV TEMP[3].w, TEMP[3].wwww
> 20: MOV TEMP[3].xyz, TEMP[0].xyzx
> 21: MOV TEMP[0].xy, IN[3].yzyy
> 22: MOV TEMP[2].zw, IMM[0].yyxy
> 23: MOV TEMP[0].zw, IMM[0].yyxy
> 24: MOV OUT[1], TEMP[3]
> 25: MOV OUT[2], TEMP[2]
> 26: MOV OUT[0], TEMP[1]
> 27: MOV OUT[3], TEMP[0]
> 28: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 0)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 4)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %12, i32 8)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %12, i32 12)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %12, i32 16)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %12, i32 20)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %12, i32 24)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %12, i32 28)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %12, i32 48)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %12, i32 52)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %12, i32 56)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %12, i32 60)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %12, i32 352)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %12, i32 356)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %12, i32 360)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %12, i32 364)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %12, i32 380)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %12, i32 384)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %12, i32 388)
>  %32 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %33 = load <16 x i8> addrspace(2)* %32, !tbaa !0
>  %34 = add i32 %5, %7
>  %35 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %33, i32 0, i32 %34)
>  %36 = extractelement <4 x float> %35, i32 0
>  %37 = extractelement <4 x float> %35, i32 1
>  %38 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %39 = load <16 x i8> addrspace(2)* %38, !tbaa !0
>  %40 = add i32 %5, %7
>  %41 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %39, i32 0, i32 %40)
>  %42 = extractelement <4 x float> %41, i32 0
>  %43 = extractelement <4 x float> %41, i32 1
>  %44 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 2
>  %45 = load <16 x i8> addrspace(2)* %44, !tbaa !0
>  %46 = add i32 %5, %7
>  %47 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %45, i32 0, i32 %46)
>  %48 = extractelement <4 x float> %47, i32 0
>  %49 = extractelement <4 x float> %47, i32 1
>  %50 = extractelement <4 x float> %47, i32 2
>  %51 = extractelement <4 x float> %47, i32 3
>  %52 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 3
>  %53 = load <16 x i8> addrspace(2)* %52, !tbaa !0
>  %54 = add i32 %5, %7
>  %55 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %53, i32 0, i32 %54)
>  %56 = extractelement <4 x float> %55, i32 0
>  %57 = extractelement <4 x float> %55, i32 1
>  %58 = extractelement <4 x float> %55, i32 2
>  %59 = fptosi float %56 to i32
>  %60 = bitcast i32 %59 to float
>  %61 = bitcast float %60 to i32
>  %62 = shl i32 %61, 4
>  %63 = add i32 %62, 72
>  %64 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %63)
>  %65 = shl i32 %61, 4
>  %66 = add i32 %65, 76
>  %67 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %66)
>  %68 = bitcast float %60 to i32
>  %69 = shl i32 %68, 4
>  %70 = add i32 %69, 64
>  %71 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %70)
>  %72 = fmul float %36, %71
>  %73 = fadd float %72, %64
>  %74 = shl i32 %68, 4
>  %75 = add i32 %74, 68
>  %76 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %75)
>  %77 = fmul float %37, %76
>  %78 = fadd float %77, %67
>  %79 = bitcast float %60 to i32
>  %80 = shl i32 %79, 4
>  %81 = add i32 %80, 216
>  %82 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %81)
>  %83 = shl i32 %79, 4
>  %84 = add i32 %83, 220
>  %85 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %84)
>  %86 = bitcast float %60 to i32
>  %87 = shl i32 %86, 4
>  %88 = add i32 %87, 208
>  %89 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %88)
>  %90 = fmul float %42, %89
>  %91 = fadd float %90, %82
>  %92 = shl i32 %86, 4
>  %93 = add i32 %92, 212
>  %94 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %93)
>  %95 = fmul float %43, %94
>  %96 = fadd float %95, %85
>  %97 = fmul float %78, %17
>  %98 = fmul float %78, %18
>  %99 = fmul float %78, %19
>  %100 = fmul float %78, %20
>  %101 = fmul float %73, %13
>  %102 = fadd float %101, %97
>  %103 = fmul float %73, %14
>  %104 = fadd float %103, %98
>  %105 = fmul float %73, %15
>  %106 = fadd float %105, %99
>  %107 = fmul float %73, %16
>  %108 = fadd float %107, %100
>  %109 = fadd float %102, %21
>  %110 = fadd float %104, %22
>  %111 = fadd float %106, %23
>  %112 = fadd float %108, %24
>  %113 = fadd float %30, %31
>  %114 = fmul float %109, %113
>  %115 = fmul float %110, %113
>  %116 = fmul float %111, %113
>  %117 = fmul float %112, %113
>  %118 = fmul float %25, %48
>  %119 = fmul float %26, %49
>  %120 = fmul float %27, %50
>  %121 = fmul float %28, %51
>  %122 = fmul float %121, %29
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %118, float %119, float %120, float %122)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %91, float %96, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 34, i32 0, float %57, float %58, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %114, float %115, float %116, float %117)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                              ; 4A00000A
>	s_load_dwordx4 s[4:7], s[8:9], 0x0                     ; C0820900
>	s_load_dwordx4 s[12:15], s[8:9], 0x4                   ; C0860904
>	s_load_dwordx4 s[16:19], s[8:9], 0x8                   ; C0880908
>	s_load_dwordx4 s[8:11], s[8:9], 0xc                    ; C084090C
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[16:19], 0 idxen  ; E00C2000 80040100
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                     ; C0800300
>	s_waitcnt vmcnt(0) lgkmcnt(0)                          ; BF8C0070
>	s_buffer_load_dword s16, s[0:3], 0x5a                  ; C208015A
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_mul_f32_e32 v5, s16, v3                              ; 100A0610
>	s_buffer_load_dword s16, s[0:3], 0x59                  ; C2080159
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_mul_f32_e32 v6, s16, v2                              ; 100C0410
>	s_buffer_load_dword s16, s[0:3], 0x58                  ; C2080158
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_mul_f32_e32 v7, s16, v1                              ; 100E0210
>	s_buffer_load_dword s16, s[0:3], 0x5b                  ; C208015B
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_mul_f32_e32 v1, s16, v4                              ; 10020810
>	s_buffer_load_dword s16, s[0:3], 0x5f                  ; C208015F
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_mul_f32_e32 v1, s16, v1                              ; 10020210
>	exp 15, 32, 0, 0, 0, v7, v6, v5, v1                    ; F800020F 01050607
>	s_waitcnt expcnt(0)                                    ; BF8C070F
>	buffer_load_format_xyzw v[1:4], v0, s[8:11], 0 idxen   ; E00C2000 80020100
>	s_waitcnt vmcnt(0)                                     ; BF8C0770
>	v_cvt_i32_f32_e32 v5, v1                               ; 7E0A1101
>	v_lshlrev_b32_e32 v5, 4, v5                            ; 340A0A84
>	v_add_i32_e32 v6, 0xdc, v5                             ; 4A0C0AFF 000000DC
>	buffer_load_dword v6, v6, s[0:3], 0 offen              ; E0301000 80000606
>	v_add_i32_e32 v7, 0xd4, v5                             ; 4A0E0AFF 000000D4
>	buffer_load_dword v7, v7, s[0:3], 0 offen              ; E0301000 80000707
>	buffer_load_format_xyzw v[8:11], v0, s[12:15], 0 idxen ; E00C2000 80030800
>	s_waitcnt vmcnt(0)                                     ; BF8C0770
>	v_mad_f32 v6, v9, v7, v6                               ; D2820006 041A0F09
>	v_add_i32_e32 v7, 0xd8, v5                             ; 4A0E0AFF 000000D8
>	buffer_load_dword v7, v7, s[0:3], 0 offen              ; E0301000 80000707
>	v_add_i32_e32 v12, 0xd0, v5                            ; 4A180AFF 000000D0
>	buffer_load_dword v12, v12, s[0:3], 0 offen            ; E0301000 80000C0C
>	s_waitcnt vmcnt(0)                                     ; BF8C0770
>	v_mad_f32 v7, v8, v12, v7                              ; D2820007 041E1908
>	v_mov_b32_e32 v8, 1.0                                  ; 7E1002F2
>	v_mov_b32_e32 v9, 0                                    ; 7E120280
>	exp 15, 33, 0, 0, 0, v7, v6, v9, v8                    ; F800021F 08090607
>	exp 15, 34, 0, 0, 0, v2, v3, v9, v8                    ; F800022F 08090302
>	s_waitcnt expcnt(0)                                    ; BF8C070F
>	v_add_i32_e32 v1, 0x48, v5                             ; 4A020AFF 00000048
>	buffer_load_dword v1, v1, s[0:3], 0 offen              ; E0301000 80000101
>	v_add_i32_e32 v2, 64, v5                               ; 4A040AC0
>	buffer_load_dword v2, v2, s[0:3], 0 offen              ; E0301000 80000202
>	buffer_load_format_xyzw v[6:9], v0, s[4:7], 0 idxen    ; E00C2000 80010600
>	s_waitcnt vmcnt(0)                                     ; BF8C0770
>	v_mad_f32 v0, v6, v2, v1                               ; D2820000 04060506
>	v_add_i32_e32 v1, 0x4c, v5                             ; 4A020AFF 0000004C
>	buffer_load_dword v1, v1, s[0:3], 0 offen              ; E0301000 80000101
>	v_add_i32_e32 v2, 0x44, v5                             ; 4A040AFF 00000044
>	buffer_load_dword v2, v2, s[0:3], 0 offen              ; E0301000 80000202
>	s_waitcnt vmcnt(0)                                     ; BF8C0770
>	v_mad_f32 v1, v7, v2, v1                               ; D2820001 04060507
>	s_buffer_load_dword s4, s[0:3], 0x7                    ; C2020107
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_mul_f32_e32 v2, s4, v1                               ; 10040204
>	s_buffer_load_dword s4, s[0:3], 0x3                    ; C2020103
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_mad_f32 v2, v0, s4, v2                               ; D2820002 04080900
>	s_buffer_load_dword s4, s[0:3], 0xf                    ; C202010F
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_add_f32_e32 v2, s4, v2                               ; 06040404
>	s_buffer_load_dword s4, s[0:3], 0x61                   ; C2020161
>	s_buffer_load_dword s5, s[0:3], 0x60                   ; C2028160
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_mov_b32_e32 v3, s4                                   ; 7E060204
>	v_add_f32_e32 v3, s5, v3                               ; 06060605
>	v_mul_f32_e32 v2, v3, v2                               ; 10040503
>	s_buffer_load_dword s4, s[0:3], 0x6                    ; C2020106
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_mul_f32_e32 v4, s4, v1                               ; 10080204
>	s_buffer_load_dword s4, s[0:3], 0x2                    ; C2020102
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_mad_f32 v4, v0, s4, v4                               ; D2820004 04100900
>	s_buffer_load_dword s4, s[0:3], 0xe                    ; C202010E
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_add_f32_e32 v4, s4, v4                               ; 06080804
>	v_mul_f32_e32 v4, v3, v4                               ; 10080903
>	s_buffer_load_dword s4, s[0:3], 0x5                    ; C2020105
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_mul_f32_e32 v5, s4, v1                               ; 100A0204
>	s_buffer_load_dword s4, s[0:3], 0x1                    ; C2020101
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_mad_f32 v5, v0, s4, v5                               ; D2820005 04140900
>	s_buffer_load_dword s4, s[0:3], 0xd                    ; C202010D
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_add_f32_e32 v5, s4, v5                               ; 060A0A04
>	v_mul_f32_e32 v5, v3, v5                               ; 100A0B03
>	s_buffer_load_dword s4, s[0:3], 0x4                    ; C2020104
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_mul_f32_e32 v1, s4, v1                               ; 10020204
>	s_buffer_load_dword s4, s[0:3], 0x0                    ; C2020100
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_mad_f32 v0, v0, s4, v1                               ; D2820000 04040900
>	s_buffer_load_dword s0, s[0:3], 0xc                    ; C200010C
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_add_f32_e32 v0, s0, v0                               ; 06000000
>	v_mul_f32_e32 v0, v3, v0                               ; 10000103
>	exp 15, 12, 0, 1, 0, v0, v5, v4, v2                    ; F80008CF 02040500
>	s_endpgm                                               ; BF810000
>
>SGPRS: 24
>VGPRS: 16
>Code Size: 556 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL IN[1], GENERIC[10], PERSPECTIVE
>DCL IN[2], GENERIC[11], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SAMP[1]
>DCL TEMP[0..1], LOCAL
>  0: MOV TEMP[0].xy, IN[1].xyyy
>  1: TEX TEMP[0], TEMP[0], SAMP[1], 2D
>  2: MUL TEMP[0], TEMP[0], IN[2].yyyy
>  3: MOV TEMP[1].xy, IN[1].xyyy
>  4: TEX TEMP[1], TEMP[1], SAMP[0], 2D
>  5: MAD TEMP[0], IN[2].xxxx, TEMP[1], TEMP[0]
>  6: MUL TEMP[0], TEMP[0], IN[0]
>  7: MOV OUT[0], TEMP[0]
>  8: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %23 = load <8 x i32> addrspace(2)* %22, !tbaa !0
>  %24 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %25 = load <4 x i32> addrspace(2)* %24, !tbaa !0
>  %26 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 1
>  %27 = load <8 x i32> addrspace(2)* %26, !tbaa !0
>  %28 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 1
>  %29 = load <4 x i32> addrspace(2)* %28, !tbaa !0
>  %30 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %31 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %32 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %5, <2 x i32> %7)
>  %33 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %5, <2 x i32> %7)
>  %34 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %5, <2 x i32> %7)
>  %35 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %5, <2 x i32> %7)
>  %36 = call float @llvm.SI.fs.interp(i32 0, i32 2, i32 %5, <2 x i32> %7)
>  %37 = call float @llvm.SI.fs.interp(i32 1, i32 2, i32 %5, <2 x i32> %7)
>  %38 = bitcast float %34 to i32
>  %39 = bitcast float %35 to i32
>  %40 = insertelement <2 x i32> undef, i32 %38, i32 0
>  %41 = insertelement <2 x i32> %40, i32 %39, i32 1
>  %42 = bitcast <8 x i32> %27 to <32 x i8>
>  %43 = bitcast <4 x i32> %29 to <16 x i8>
>  %44 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %41, <32 x i8> %42, <16 x i8> %43, i32 2)
>  %45 = extractelement <4 x float> %44, i32 0
>  %46 = extractelement <4 x float> %44, i32 1
>  %47 = extractelement <4 x float> %44, i32 2
>  %48 = extractelement <4 x float> %44, i32 3
>  %49 = fmul float %45, %37
>  %50 = fmul float %46, %37
>  %51 = fmul float %47, %37
>  %52 = fmul float %48, %37
>  %53 = bitcast float %34 to i32
>  %54 = bitcast float %35 to i32
>  %55 = insertelement <2 x i32> undef, i32 %53, i32 0
>  %56 = insertelement <2 x i32> %55, i32 %54, i32 1
>  %57 = bitcast <8 x i32> %23 to <32 x i8>
>  %58 = bitcast <4 x i32> %25 to <16 x i8>
>  %59 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %56, <32 x i8> %57, <16 x i8> %58, i32 2)
>  %60 = extractelement <4 x float> %59, i32 0
>  %61 = extractelement <4 x float> %59, i32 1
>  %62 = extractelement <4 x float> %59, i32 2
>  %63 = extractelement <4 x float> %59, i32 3
>  %64 = fmul float %36, %60
>  %65 = fadd float %64, %49
>  %66 = fmul float %36, %61
>  %67 = fadd float %66, %50
>  %68 = fmul float %36, %62
>  %69 = fadd float %68, %51
>  %70 = fmul float %36, %63
>  %71 = fadd float %70, %52
>  %72 = fmul float %65, %30
>  %73 = fmul float %67, %31
>  %74 = fmul float %69, %32
>  %75 = fmul float %71, %33
>  %76 = call i32 @llvm.SI.packf16(float %72, float %73)
>  %77 = bitcast i32 %76 to float
>  %78 = call i32 @llvm.SI.packf16(float %74, float %75)
>  %79 = bitcast i32 %78 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %77, float %79, float %77, float %79)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                     ; BEFE0A7E
>	s_mov_b32 m0, s9                                                         ; BEFC0309
>	v_interp_p1_f32 v3, v0, 1, 1, [m0]                                       ; C80C0500
>	v_interp_p2_f32 v3, [v3], v1, 1, 1, [m0]                                 ; C80D0501
>	v_interp_p1_f32 v2, v0, 0, 1, [m0]                                       ; C8080400
>	v_interp_p2_f32 v2, [v2], v1, 0, 1, [m0]                                 ; C8090401
>	s_load_dwordx4 s[0:3], s[4:5], 0x0                                       ; C0800500
>	s_load_dwordx4 s[8:11], s[4:5], 0x4                                      ; C0840504
>	s_load_dwordx8 s[12:19], s[6:7], 0x0                                     ; C0C60700
>	s_load_dwordx8 s[20:27], s[6:7], 0x8                                     ; C0CA0708
>	s_waitcnt lgkmcnt(0)                                                     ; BF8C007F
>	image_sample v[4:7], 15, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[20:27], s[8:11]  ; F0800F00 00450402
>	v_interp_p1_f32 v8, v0, 1, 2, [m0]                                       ; C8200900
>	v_interp_p2_f32 v8, [v8], v1, 1, 2, [m0]                                 ; C8210901
>	s_waitcnt vmcnt(0)                                                       ; BF8C0770
>	v_mul_f32_e32 v9, v8, v7                                                 ; 10120F08
>	image_sample v[10:13], 15, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[12:19], s[0:3] ; F0800F00 00030A02
>	v_interp_p1_f32 v2, v0, 0, 2, [m0]                                       ; C8080800
>	v_interp_p2_f32 v2, [v2], v1, 0, 2, [m0]                                 ; C8090801
>	s_waitcnt vmcnt(0)                                                       ; BF8C0770
>	v_mad_f32 v3, v2, v13, v9                                                ; D2820003 04261B02
>	v_interp_p1_f32 v9, v0, 3, 0, [m0]                                       ; C8240300
>	v_interp_p2_f32 v9, [v9], v1, 3, 0, [m0]                                 ; C8250301
>	v_mul_f32_e32 v3, v9, v3                                                 ; 10060709
>	v_mul_f32_e32 v9, v8, v6                                                 ; 10120D08
>	v_mad_f32 v9, v2, v12, v9                                                ; D2820009 04261902
>	v_interp_p1_f32 v14, v0, 2, 0, [m0]                                      ; C8380200
>	v_interp_p2_f32 v14, [v14], v1, 2, 0, [m0]                               ; C8390201
>	v_mul_f32_e32 v9, v14, v9                                                ; 1012130E
>	v_cvt_pkrtz_f16_f32_e32 v3, v9, v3                                       ; 5E060709
>	v_mul_f32_e32 v9, v8, v5                                                 ; 10120B08
>	v_mad_f32 v9, v2, v11, v9                                                ; D2820009 04261702
>	v_interp_p1_f32 v14, v0, 1, 0, [m0]                                      ; C8380100
>	v_interp_p2_f32 v14, [v14], v1, 1, 0, [m0]                               ; C8390101
>	v_mul_f32_e32 v9, v14, v9                                                ; 1012130E
>	v_mul_f32_e32 v4, v8, v4                                                 ; 10080908
>	v_mad_f32 v2, v2, v10, v4                                                ; D2820002 04121502
>	v_interp_p1_f32 v4, v0, 0, 0, [m0]                                       ; C8100000
>	v_interp_p2_f32 v4, [v4], v1, 0, 0, [m0]                                 ; C8110001
>	v_mul_f32_e32 v0, v4, v2                                                 ; 10000504
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v9                                       ; 5E001300
>	exp 15, 0, 1, 1, 1, v0, v3, v0, v3                                       ; F8001C0F 03000300
>	s_endpgm                                                                 ; BF810000
>
>SGPRS: 32
>VGPRS: 16
>Code Size: 200 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL IN[2]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL OUT[2], GENERIC[10]
>DCL CONST[0..6]
>DCL TEMP[0..2], LOCAL
>IMM[0] FLT32 {    0.0000,     1.0000,     0.0000,     0.0000}
>  0: ADD TEMP[0].xy, CONST[4].zwzw, IN[0]
>  1: MUL TEMP[1], TEMP[0].yyyy, CONST[1]
>  2: MAD TEMP[0], TEMP[0].xxxx, CONST[0], TEMP[1]
>  3: ADD TEMP[0], TEMP[0], CONST[3]
>  4: ADD TEMP[1].x, CONST[6].xxxx, CONST[6].yyyy
>  5: MUL TEMP[0], TEMP[0], TEMP[1].xxxx
>  6: MUL TEMP[1].w, CONST[5].wwww, IN[2].wwww
>  7: MOV TEMP[1].w, TEMP[1].wwww
>  8: MOV TEMP[2].xy, IN[1].xyxx
>  9: MOV TEMP[1].xyz, IN[2].xyzx
> 10: MOV TEMP[2].zw, IMM[0].yyxy
> 11: MOV OUT[1], TEMP[1]
> 12: MOV OUT[2], TEMP[2]
> 13: MOV OUT[0], TEMP[0]
> 14: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 0)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 4)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %12, i32 8)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %12, i32 12)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %12, i32 16)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %12, i32 20)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %12, i32 24)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %12, i32 28)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %12, i32 48)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %12, i32 52)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %12, i32 56)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %12, i32 60)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %12, i32 72)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %12, i32 76)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %12, i32 92)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %12, i32 96)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %12, i32 100)
>  %30 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %31 = load <16 x i8> addrspace(2)* %30, !tbaa !0
>  %32 = add i32 %5, %7
>  %33 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %31, i32 0, i32 %32)
>  %34 = extractelement <4 x float> %33, i32 0
>  %35 = extractelement <4 x float> %33, i32 1
>  %36 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %37 = load <16 x i8> addrspace(2)* %36, !tbaa !0
>  %38 = add i32 %5, %7
>  %39 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %37, i32 0, i32 %38)
>  %40 = extractelement <4 x float> %39, i32 0
>  %41 = extractelement <4 x float> %39, i32 1
>  %42 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 2
>  %43 = load <16 x i8> addrspace(2)* %42, !tbaa !0
>  %44 = add i32 %5, %7
>  %45 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %43, i32 0, i32 %44)
>  %46 = extractelement <4 x float> %45, i32 0
>  %47 = extractelement <4 x float> %45, i32 1
>  %48 = extractelement <4 x float> %45, i32 2
>  %49 = extractelement <4 x float> %45, i32 3
>  %50 = fadd float %25, %34
>  %51 = fadd float %26, %35
>  %52 = fmul float %51, %17
>  %53 = fmul float %51, %18
>  %54 = fmul float %51, %19
>  %55 = fmul float %51, %20
>  %56 = fmul float %50, %13
>  %57 = fadd float %56, %52
>  %58 = fmul float %50, %14
>  %59 = fadd float %58, %53
>  %60 = fmul float %50, %15
>  %61 = fadd float %60, %54
>  %62 = fmul float %50, %16
>  %63 = fadd float %62, %55
>  %64 = fadd float %57, %21
>  %65 = fadd float %59, %22
>  %66 = fadd float %61, %23
>  %67 = fadd float %63, %24
>  %68 = fadd float %28, %29
>  %69 = fmul float %64, %68
>  %70 = fmul float %65, %68
>  %71 = fmul float %66, %68
>  %72 = fmul float %67, %68
>  %73 = fmul float %27, %49
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %46, float %47, float %48, float %73)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %40, float %41, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %69, float %70, float %71, float %72)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                             ; 4A00000A
>	s_load_dwordx4 s[4:7], s[8:9], 0x0                    ; C0820900
>	s_load_dwordx4 s[12:15], s[8:9], 0x4                  ; C0860904
>	s_load_dwordx4 s[8:11], s[8:9], 0x8                   ; C0840908
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[8:11], 0 idxen  ; E00C2000 80020100
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                    ; C0800300
>	s_waitcnt vmcnt(0) lgkmcnt(0)                         ; BF8C0070
>	s_buffer_load_dword s8, s[0:3], 0x17                  ; C2040117
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v5, s8, v4                              ; 100A0808
>	exp 15, 32, 0, 0, 0, v1, v2, v3, v5                   ; F800020F 05030201
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	buffer_load_format_xyzw v[1:4], v0, s[12:15], 0 idxen ; E00C2000 80030100
>	v_mov_b32_e32 v5, 1.0                                 ; 7E0A02F2
>	v_mov_b32_e32 v6, 0                                   ; 7E0C0280
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	exp 15, 33, 0, 0, 0, v1, v2, v6, v5                   ; F800021F 05060201
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	buffer_load_format_xyzw v[0:3], v0, s[4:7], 0 idxen   ; E00C2000 80010000
>	s_buffer_load_dword s4, s[0:3], 0x12                  ; C2020112
>	s_waitcnt vmcnt(0) lgkmcnt(0)                         ; BF8C0070
>	v_add_f32_e32 v4, s4, v0                              ; 06080004
>	s_buffer_load_dword s4, s[0:3], 0x13                  ; C2020113
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v0, s4, v1                              ; 06000204
>	s_buffer_load_dword s4, s[0:3], 0x7                   ; C2020107
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v1, s4, v0                              ; 10020004
>	s_buffer_load_dword s4, s[0:3], 0x3                   ; C2020103
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v1, v4, s4, v1                              ; D2820001 04040904
>	s_buffer_load_dword s4, s[0:3], 0xf                   ; C202010F
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v1, s4, v1                              ; 06020204
>	s_buffer_load_dword s4, s[0:3], 0x19                  ; C2020119
>	s_buffer_load_dword s5, s[0:3], 0x18                  ; C2028118
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mov_b32_e32 v2, s4                                  ; 7E040204
>	v_add_f32_e32 v2, s5, v2                              ; 06040405
>	v_mul_f32_e32 v1, v2, v1                              ; 10020302
>	s_buffer_load_dword s4, s[0:3], 0x6                   ; C2020106
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v3, s4, v0                              ; 10060004
>	s_buffer_load_dword s4, s[0:3], 0x2                   ; C2020102
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v3, v4, s4, v3                              ; D2820003 040C0904
>	s_buffer_load_dword s4, s[0:3], 0xe                   ; C202010E
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v3, s4, v3                              ; 06060604
>	v_mul_f32_e32 v3, v2, v3                              ; 10060702
>	s_buffer_load_dword s4, s[0:3], 0x5                   ; C2020105
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v5, s4, v0                              ; 100A0004
>	s_buffer_load_dword s4, s[0:3], 0x1                   ; C2020101
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v5, v4, s4, v5                              ; D2820005 04140904
>	s_buffer_load_dword s4, s[0:3], 0xd                   ; C202010D
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v5, s4, v5                              ; 060A0A04
>	v_mul_f32_e32 v5, v2, v5                              ; 100A0B02
>	s_buffer_load_dword s4, s[0:3], 0x4                   ; C2020104
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v0, s4, v0                              ; 10000004
>	s_buffer_load_dword s4, s[0:3], 0x0                   ; C2020100
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v0, v4, s4, v0                              ; D2820000 04000904
>	s_buffer_load_dword s0, s[0:3], 0xc                   ; C200010C
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v0, s0, v0                              ; 06000000
>	v_mul_f32_e32 v0, v2, v0                              ; 10000102
>	exp 15, 12, 0, 1, 0, v0, v5, v3, v1                   ; F80008CF 01030500
>	s_endpgm                                              ; BF810000
>
>SGPRS: 24
>VGPRS: 8
>Code Size: 332 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL IN[1], GENERIC[10], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SAMP[1]
>DCL CONST[0..4]
>DCL TEMP[0..8], LOCAL
>IMM[0] FLT32 {    0.5000,  1024.0000,     0.0000, 65280.0000}
>IMM[1] FLT32 {65280.0000,   255.0000,     0.0000,     0.0000}
>  0: MOV TEMP[0].y, IMM[0].xxxx
>  1: ADD TEMP[1].x, IMM[0].yyyy, -CONST[2].xxxx
>  2: MOV TEMP[0].x, TEMP[1].xxxx
>  3: FSNE TEMP[2].x, CONST[0].xxxx, -CONST[0].xxxx
>  4: UIF TEMP[2].xxxx :0
>  5:   FSLT TEMP[2].x, IMM[0].xxxx, CONST[1].xxxx
>  6:   UIF TEMP[2].xxxx :0
>  7:     MUL TEMP[2].zw, CONST[3].xyxy, IN[1].xyxy
>  8:     MOV TEMP[0].zw, TEMP[2].wwzw
>  9:     FRC TEMP[2].xy, TEMP[2].zwzw
> 10:     ADD TEMP[3].zw, TEMP[0], -TEMP[2].xyxy
> 11:     MOV TEMP[0].zw, TEMP[3].wwzw
> 12:     MAD TEMP[3].xy, IN[1], CONST[3], -TEMP[3].zwzw
> 13:     ADD TEMP[4].zw, TEMP[0], IMM[0].xxxx
> 14:     RCP TEMP[5].x, CONST[3].xxxx
> 15:     RCP TEMP[6].x, CONST[3].yyyy
> 16:     MOV TEMP[5].y, TEMP[6].xxxx
> 17:     MUL TEMP[4].xy, TEMP[4].zwzw, TEMP[5]
> 18:     MOV TEMP[4].xy, TEMP[4].xyyy
> 19:     MOV TEMP[4].w, IMM[0].zzzz
> 20:     TXL TEMP[4], TEMP[4], SAMP[0], 2D
> 21:     MOV TEMP[5].zw, TEMP[4]
> 22:     DP2 TEMP[6].x, TEMP[4].wxxx, IMM[1].xyyy
> 23:     ADD TEMP[6].x, TEMP[6].xxxx, IMM[0].xxxx
> 24:     FRC TEMP[7].w, TEMP[6].xxxx
> 25:     ADD TEMP[6].z, -TEMP[7].wwww, TEMP[6].xxxx
> 26:     MOV TEMP[0].z, TEMP[6].zzzz
> 27:     MAD TEMP[4].w, CONST[2].xxxx, TEMP[4].xxxx, IMM[0].xxxx
> 28:     FRC TEMP[6].z, TEMP[4].wwww
> 29:     ADD TEMP[4].w, TEMP[4].wwww, -TEMP[6].zzzz
> 30:     MOV TEMP[0].w, TEMP[4].wwww
> 31:     ADD TEMP[4].zw, TEMP[0], IMM[0].xxxx
> 32:     FSGE TEMP[6].x, TEMP[1].xxxx, IMM[0].zzzz
> 33:     UIF TEMP[6].xxxx :0
> 34:       MOV TEMP[6].x, TEMP[4].wwww
> 35:     ELSE :0
> 36:       MOV TEMP[6].x, TEMP[4].zzzz
> 37:     ENDIF
> 38:     RCP TEMP[4].x, CONST[1].xxxx
> 39:     MAD TEMP[4].w, CONST[4].xxxx, TEMP[4].xxxx, IMM[0].xxxx
> 40:     FRC TEMP[7].z, TEMP[4].wwww
> 41:     ADD TEMP[4].w, TEMP[4].wwww, -TEMP[7].zzzz
> 42:     MUL TEMP[7].z, TEMP[4].wwww, TEMP[6].xxxx
> 43:     MOV TEMP[8].x, -TEMP[4].wwww
> 44:     FSGE TEMP[7].x, TEMP[7].zzzz, IMM[0].zzzz
> 45:     UIF TEMP[7].xxxx :0
> 46:       MOV TEMP[7].x, TEMP[4].wwww
> 47:     ELSE :0
> 48:       MOV TEMP[7].x, TEMP[8].xxxx
> 49:     ENDIF
> 50:     MOV TEMP[2].z, TEMP[7].xxxx
> 51:     RCP TEMP[8].x, TEMP[7].xxxx
> 52:     MUL TEMP[8].w, TEMP[6].xxxx, TEMP[8].xxxx
> 53:     FRC TEMP[8].w, TEMP[8].wwww
> 54:     MOV TEMP[2].w, TEMP[8].wwww
> 55:     MUL TEMP[7].x, TEMP[8].wwww, TEMP[7].xxxx
> 56:     MOV TEMP[5].x, TEMP[7].xxxx
> 57:     RCP TEMP[4].x, TEMP[4].wwww
> 58:     MUL TEMP[4].y, TEMP[4].xxxx, TEMP[6].xxxx
> 59:     MOV TEMP[5].y, TEMP[4].yyyy
> 60:     FRC TEMP[4].zw, TEMP[5].xyxy
> 61:     MOV TEMP[0].zw, TEMP[4].wwzw
> 62:     ADD TEMP[4].zw, -TEMP[0], TEMP[5].xyxy
> 63:     MOV TEMP[0].zw, TEMP[4].wwzw
> 64:     ADD TEMP[3].zw, TEMP[3].xyxy, TEMP[0]
> 65:     MOV TEMP[0].zw, TEMP[3].wwzw
> 66:     MUL TEMP[3].zw, TEMP[0], CONST[1].xxxx
> 67:     MOV TEMP[0].zw, TEMP[3].wwzw
> 68:     FRC TEMP[3].xy, TEMP[3].zwzw
> 69:     ADD TEMP[3].zw, TEMP[0], -TEMP[3].xyxy
> 70:     MOV TEMP[0].zw, TEMP[3].wwzw
> 71:     ADD TEMP[3].zw, TEMP[0], IMM[0].xxxx
> 72:     MOV TEMP[0].w, TEMP[3].wwzw
> 73:     RCP TEMP[2].x, CONST[4].xxxx
> 74:     RCP TEMP[4].x, CONST[4].yyyy
> 75:     MOV TEMP[2].y, TEMP[4].xxxx
> 76:     MUL TEMP[3].xy, TEMP[3].zwzw, TEMP[2]
> 77:     MOV TEMP[3].xy, TEMP[3].xyyy
> 78:     MOV TEMP[3].w, IMM[0].zzzz
> 79:     TXL TEMP[3], TEMP[3], SAMP[1], 2D
> 80:     MOV TEMP[2], TEMP[3]
> 81:     MOV TEMP[0].z, TEMP[3].wwww
> 82:   ELSE :0
> 83:     MOV TEMP[3].xy, IN[1].xyyy
> 84:     TEX TEMP[3], TEMP[3], SAMP[1], 2D
> 85:     MOV TEMP[2], TEMP[3]
> 86:     MOV TEMP[0].z, TEMP[3].wwww
> 87:   ENDIF
> 88:   MUL TEMP[3].w, TEMP[0].zzzz, IN[0].wwww
> 89:   MOV TEMP[3].w, TEMP[3].wwww
> 90:   MOV TEMP[3].xyz, IN[0].xyzx
> 91: ELSE :0
> 92:   FSLT TEMP[4].x, IMM[0].xxxx, CONST[1].xxxx
> 93:   UIF TEMP[4].xxxx :0
> 94:     MUL TEMP[4].zw, CONST[3].xyxy, IN[1].xyxy
> 95:     MOV TEMP[0].zw, TEMP[4].wwzw
> 96:     FRC TEMP[4].xy, TEMP[4].zwzw
> 97:     MOV TEMP[2].xy, TEMP[4].xyxx
> 98:     ADD TEMP[4].zw, TEMP[0], -TEMP[4].xyxy
> 99:     MOV TEMP[0].zw, TEMP[4].wwzw
>100:     MAD TEMP[4].xy, IN[1], CONST[3], -TEMP[4].zwzw
>101:     MOV TEMP[2].xy, TEMP[4].xyxx
>102:     ADD TEMP[4].zw, TEMP[0], IMM[0].xxxx
>103:     RCP TEMP[5].x, CONST[3].xxxx
>104:     RCP TEMP[6].x, CONST[3].yyyy
>105:     MOV TEMP[5].y, TEMP[6].xxxx
>106:     MUL TEMP[4].xy, TEMP[4].zwzw, TEMP[5]
>107:     MOV TEMP[4].xy, TEMP[4].xyyy
>108:     MOV TEMP[4].w, IMM[0].zzzz
>109:     TXL TEMP[4], TEMP[4], SAMP[0], 2D
>110:     MOV TEMP[5].zw, TEMP[4]
>111:     DP2 TEMP[6].x, TEMP[4].wxxx, IMM[1].xyyy
>112:     ADD TEMP[6].x, TEMP[6].xxxx, IMM[0].xxxx
>113:     FRC TEMP[7].w, TEMP[6].xxxx
>114:     ADD TEMP[6].z, -TEMP[7].wwww, TEMP[6].xxxx
>115:     MOV TEMP[0].z, TEMP[6].zzzz
>116:     MAD TEMP[4].w, CONST[2].xxxx, TEMP[4].xxxx, IMM[0].xxxx
>117:     FRC TEMP[6].z, TEMP[4].wwww
>118:     MOV TEMP[2].z, TEMP[6].zzzz
>119:     ADD TEMP[4].w, TEMP[4].wwww, -TEMP[6].zzzz
>120:     MOV TEMP[0].w, TEMP[4].wwww
>121:     ADD TEMP[4].zw, TEMP[0], IMM[0].xxxx
>122:     FSGE TEMP[1].x, TEMP[1].xxxx, IMM[0].zzzz
>123:     UIF TEMP[1].xxxx :0
>124:       MOV TEMP[1].x, TEMP[4].wwww
>125:     ELSE :0
>126:       MOV TEMP[1].x, TEMP[4].zzzz
>127:     ENDIF
>128:     MOV TEMP[0].x, TEMP[1].xxxx
>129:     RCP TEMP[4].x, CONST[1].xxxx
>130:     MAD TEMP[4].y, CONST[4].xxxx, TEMP[4].xxxx, IMM[0].xxxx
>131:     FRC TEMP[6].z, TEMP[4].yyyy
>132:     ADD TEMP[4].y, -TEMP[6].zzzz, TEMP[4].yyyy
>133:     MUL TEMP[6].z, TEMP[4].yyyy, TEMP[1].xxxx
>134:     MOV TEMP[7].x, -TEMP[4].yyyy
>135:     FSGE TEMP[6].x, TEMP[6].zzzz, IMM[0].zzzz
>136:     UIF TEMP[6].xxxx :0
>137:       MOV TEMP[6].x, TEMP[4].yyyy
>138:     ELSE :0
>139:       MOV TEMP[6].x, TEMP[7].xxxx
>140:     ENDIF
>141:     MOV TEMP[0].z, TEMP[6].xxxx
>142:     RCP TEMP[6].x, TEMP[6].xxxx
>143:     MUL TEMP[1].w, TEMP[6].xxxx, TEMP[1].xxxx
>144:     FRC TEMP[1].w, TEMP[1].wwww
>145:     MOV TEMP[0].w, TEMP[1].wwww
>146:     RCP TEMP[1].x, TEMP[4].yyyy
>147:     MOV TEMP[0].y, TEMP[1].xxxx
>148:     MUL TEMP[1].xy, TEMP[0].wyzw, TEMP[0].zxzw
>149:     MOV TEMP[5].xy, TEMP[1].xyxx
>150:     FRC TEMP[1].xy, TEMP[5]
>151:     MOV TEMP[0].xy, TEMP[1].xyxx
>152:     ADD TEMP[1].xy, -TEMP[0], TEMP[5]
>153:     MOV TEMP[0].xy, TEMP[1].xyxx
>154:     ADD TEMP[1].xy, TEMP[2], TEMP[0]
>155:     MOV TEMP[0].xy, TEMP[1].xyxx
>156:     MUL TEMP[1].xy, TEMP[0], CONST[1].xxxx
>157:     MOV TEMP[0].xy, TEMP[1].xyxx
>158:     FRC TEMP[1].zw, TEMP[1].xyxy
>159:     MOV TEMP[0].zw, TEMP[1].wwzw
>160:     ADD TEMP[1].xy, -TEMP[1].zwzw, TEMP[0]
>161:     MOV TEMP[0].xy, TEMP[1].xyxx
>162:     ADD TEMP[1].xy, TEMP[0], IMM[0].xxxx
>163:     MOV TEMP[0].xy, TEMP[1].xyxx
>164:     RCP TEMP[2].x, CONST[4].xxxx
>165:     RCP TEMP[1].x, CONST[4].yyyy
>166:     MOV TEMP[2].y, TEMP[1].xxxx
>167:     MUL TEMP[1].xy, TEMP[0], TEMP[2]
>168:     MOV TEMP[1].xy, TEMP[1].xyyy
>169:     MOV TEMP[1].w, IMM[0].zzzz
>170:     TXL TEMP[1], TEMP[1], SAMP[1], 2D
>171:     MOV TEMP[0].xyz, TEMP[1]
>172:     MOV TEMP[2].x, TEMP[1].wwww
>173:   ELSE :0
>174:     MOV TEMP[1].xy, IN[1].xyyy
>175:     TEX TEMP[1], TEMP[1], SAMP[1], 2D
>176:     MOV TEMP[0].xyz, TEMP[1]
>177:     MOV TEMP[2].x, TEMP[1].wwww
>178:   ENDIF
>179:   MUL TEMP[1].w, TEMP[2].xxxx, IN[0].wwww
>180:   MOV TEMP[0].w, TEMP[1].wwww
>181:   MOV TEMP[3], TEMP[0]
>182: ENDIF
>183: MOV OUT[0], TEMP[3]
>184: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %23 = load <16 x i8> addrspace(2)* %22, !tbaa !0
>  %24 = call float @llvm.SI.load.const(<16 x i8> %23, i32 0)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %23, i32 16)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %23, i32 32)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %23, i32 48)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %23, i32 52)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %23, i32 64)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %23, i32 68)
>  %31 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %32 = load <8 x i32> addrspace(2)* %31, !tbaa !0
>  %33 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %34 = load <4 x i32> addrspace(2)* %33, !tbaa !0
>  %35 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 1
>  %36 = load <8 x i32> addrspace(2)* %35, !tbaa !0
>  %37 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 1
>  %38 = load <4 x i32> addrspace(2)* %37, !tbaa !0
>  %39 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %40 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %41 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %5, <2 x i32> %7)
>  %42 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %5, <2 x i32> %7)
>  %43 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %5, <2 x i32> %7)
>  %44 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %5, <2 x i32> %7)
>  %45 = fsub float -0.000000e+00, %26
>  %46 = fadd float 1.024000e+03, %45
>  %47 = fsub float -0.000000e+00, %24
>  %48 = fcmp une float %24, %47
>  %49 = sext i1 %48 to i32
>  %50 = bitcast i32 %49 to float
>  %51 = bitcast float %50 to i32
>  %52 = icmp ne i32 %51, 0
>  %53 = fcmp olt float 5.000000e-01, %25
>  %54 = sext i1 %53 to i32
>  %55 = bitcast i32 %54 to float
>  %56 = bitcast float %55 to i32
>  %57 = icmp ne i32 %56, 0
>  br i1 %52, label %IF, label %ELSE
>
>IF:                                               ; preds = %main_body
>  br i1 %57, label %IF37, label %ELSE38
>
>ELSE:                                             ; preds = %main_body
>  br i1 %57, label %IF46, label %ELSE47
>
>ENDIF:                                            ; preds = %IF37, %ELSE38, %ENDIF45
>  %.sink56.sink = phi <4 x float> [ %.sink56, %ENDIF45 ], [ %172, %ELSE38 ], [ %165, %IF37 ]
>  %temp12.0 = phi float [ %282, %ENDIF45 ], [ %39, %ELSE38 ], [ %39, %IF37 ]
>  %temp13.0 = phi float [ %283, %ENDIF45 ], [ %40, %ELSE38 ], [ %40, %IF37 ]
>  %temp14.0 = phi float [ %284, %ENDIF45 ], [ %41, %ELSE38 ], [ %41, %IF37 ]
>  %58 = extractelement <4 x float> %.sink56.sink, i32 3
>  %59 = fmul float %58, %42
>  %60 = call i32 @llvm.SI.packf16(float %temp12.0, float %temp13.0)
>  %61 = bitcast i32 %60 to float
>  %62 = call i32 @llvm.SI.packf16(float %temp14.0, float %59)
>  %63 = bitcast i32 %62 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %61, float %63, float %61, float %63)
>  ret void
>
>IF37:                                             ; preds = %IF
>  %64 = fmul float %27, %43
>  %65 = fmul float %28, %44
>  %66 = call float @llvm.AMDIL.fraction.(float %64)
>  %67 = call float @llvm.AMDIL.fraction.(float %65)
>  %68 = fsub float -0.000000e+00, %66
>  %69 = fadd float %64, %68
>  %70 = fsub float -0.000000e+00, %67
>  %71 = fadd float %65, %70
>  %72 = fsub float -0.000000e+00, %69
>  %73 = fmul float %43, %27
>  %74 = fadd float %73, %72
>  %75 = fsub float -0.000000e+00, %71
>  %76 = fmul float %44, %28
>  %77 = fadd float %76, %75
>  %78 = fadd float %69, 5.000000e-01
>  %79 = fadd float %71, 5.000000e-01
>  %80 = fdiv float 1.000000e+00, %27
>  %81 = fdiv float 1.000000e+00, %28
>  %82 = fmul float %78, %80
>  %83 = fmul float %79, %81
>  %84 = bitcast float %82 to i32
>  %85 = bitcast float %83 to i32
>  %86 = bitcast float 0.000000e+00 to i32
>  %87 = insertelement <4 x i32> undef, i32 %84, i32 0
>  %88 = insertelement <4 x i32> %87, i32 %85, i32 1
>  %89 = insertelement <4 x i32> %88, i32 %86, i32 2
>  %90 = insertelement <4 x i32> %89, i32 undef, i32 3
>  %91 = bitcast <8 x i32> %32 to <32 x i8>
>  %92 = bitcast <4 x i32> %34 to <16 x i8>
>  %93 = call <4 x float> @llvm.SI.samplel.v4i32(<4 x i32> %90, <32 x i8> %91, <16 x i8> %92, i32 2)
>  %94 = extractelement <4 x float> %93, i32 0
>  %95 = extractelement <4 x float> %93, i32 3
>  %96 = fmul float %95, 6.528000e+04
>  %97 = fmul float %94, 2.550000e+02
>  %98 = fadd float %96, %97
>  %99 = fadd float %98, 5.000000e-01
>  %100 = call float @llvm.AMDIL.fraction.(float %99)
>  %101 = fsub float -0.000000e+00, %100
>  %102 = fadd float %101, %99
>  %103 = fmul float %26, %94
>  %104 = fadd float %103, 5.000000e-01
>  %105 = call float @llvm.AMDIL.fraction.(float %104)
>  %106 = fsub float -0.000000e+00, %105
>  %107 = fadd float %104, %106
>  %108 = fadd float %102, 5.000000e-01
>  %109 = fadd float %107, 5.000000e-01
>  %110 = fcmp oge float %46, 0.000000e+00
>  %111 = sext i1 %110 to i32
>  %112 = bitcast i32 %111 to float
>  %113 = bitcast float %112 to i32
>  %114 = icmp ne i32 %113, 0
>  %. = select i1 %114, float %109, float %108
>  %115 = fdiv float 1.000000e+00, %25
>  %116 = fmul float %29, %115
>  %117 = fadd float %116, 5.000000e-01
>  %118 = call float @llvm.AMDIL.fraction.(float %117)
>  %119 = fsub float -0.000000e+00, %118
>  %120 = fadd float %117, %119
>  %121 = fmul float %120, %.
>  %122 = fsub float -0.000000e+00, %120
>  %123 = fcmp oge float %121, 0.000000e+00
>  %124 = sext i1 %123 to i32
>  %125 = bitcast i32 %124 to float
>  %126 = bitcast float %125 to i32
>  %127 = icmp ne i32 %126, 0
>  %temp28.0 = select i1 %127, float %120, float %122
>  %128 = fdiv float 1.000000e+00, %temp28.0
>  %129 = fmul float %., %128
>  %130 = call float @llvm.AMDIL.fraction.(float %129)
>  %131 = fmul float %130, %temp28.0
>  %132 = fdiv float 1.000000e+00, %120
>  %133 = fmul float %132, %.
>  %134 = call float @llvm.AMDIL.fraction.(float %131)
>  %135 = call float @llvm.AMDIL.fraction.(float %133)
>  %136 = fsub float -0.000000e+00, %134
>  %137 = fadd float %136, %131
>  %138 = fsub float -0.000000e+00, %135
>  %139 = fadd float %138, %133
>  %140 = fadd float %74, %137
>  %141 = fadd float %77, %139
>  %142 = fmul float %140, %25
>  %143 = fmul float %141, %25
>  %144 = call float @llvm.AMDIL.fraction.(float %142)
>  %145 = call float @llvm.AMDIL.fraction.(float %143)
>  %146 = fsub float -0.000000e+00, %144
>  %147 = fadd float %142, %146
>  %148 = fsub float -0.000000e+00, %145
>  %149 = fadd float %143, %148
>  %150 = fadd float %147, 5.000000e-01
>  %151 = fadd float %149, 5.000000e-01
>  %152 = fdiv float 1.000000e+00, %29
>  %153 = fdiv float 1.000000e+00, %30
>  %154 = fmul float %150, %152
>  %155 = fmul float %151, %153
>  %156 = bitcast float %154 to i32
>  %157 = bitcast float %155 to i32
>  %158 = bitcast float 0.000000e+00 to i32
>  %159 = insertelement <4 x i32> undef, i32 %156, i32 0
>  %160 = insertelement <4 x i32> %159, i32 %157, i32 1
>  %161 = insertelement <4 x i32> %160, i32 %158, i32 2
>  %162 = insertelement <4 x i32> %161, i32 undef, i32 3
>  %163 = bitcast <8 x i32> %36 to <32 x i8>
>  %164 = bitcast <4 x i32> %38 to <16 x i8>
>  %165 = call <4 x float> @llvm.SI.samplel.v4i32(<4 x i32> %162, <32 x i8> %163, <16 x i8> %164, i32 2)
>  br label %ENDIF
>
>ELSE38:                                           ; preds = %IF
>  %166 = bitcast float %43 to i32
>  %167 = bitcast float %44 to i32
>  %168 = insertelement <2 x i32> undef, i32 %166, i32 0
>  %169 = insertelement <2 x i32> %168, i32 %167, i32 1
>  %170 = bitcast <8 x i32> %36 to <32 x i8>
>  %171 = bitcast <4 x i32> %38 to <16 x i8>
>  %172 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %169, <32 x i8> %170, <16 x i8> %171, i32 2)
>  br label %ENDIF
>
>IF46:                                             ; preds = %ELSE
>  %173 = fmul float %27, %43
>  %174 = fmul float %28, %44
>  %175 = call float @llvm.AMDIL.fraction.(float %173)
>  %176 = call float @llvm.AMDIL.fraction.(float %174)
>  %177 = fsub float -0.000000e+00, %175
>  %178 = fadd float %173, %177
>  %179 = fsub float -0.000000e+00, %176
>  %180 = fadd float %174, %179
>  %181 = fsub float -0.000000e+00, %178
>  %182 = fmul float %43, %27
>  %183 = fadd float %182, %181
>  %184 = fsub float -0.000000e+00, %180
>  %185 = fmul float %44, %28
>  %186 = fadd float %185, %184
>  %187 = fadd float %178, 5.000000e-01
>  %188 = fadd float %180, 5.000000e-01
>  %189 = fdiv float 1.000000e+00, %27
>  %190 = fdiv float 1.000000e+00, %28
>  %191 = fmul float %187, %189
>  %192 = fmul float %188, %190
>  %193 = bitcast float %191 to i32
>  %194 = bitcast float %192 to i32
>  %195 = bitcast float 0.000000e+00 to i32
>  %196 = insertelement <4 x i32> undef, i32 %193, i32 0
>  %197 = insertelement <4 x i32> %196, i32 %194, i32 1
>  %198 = insertelement <4 x i32> %197, i32 %195, i32 2
>  %199 = insertelement <4 x i32> %198, i32 undef, i32 3
>  %200 = bitcast <8 x i32> %32 to <32 x i8>
>  %201 = bitcast <4 x i32> %34 to <16 x i8>
>  %202 = call <4 x float> @llvm.SI.samplel.v4i32(<4 x i32> %199, <32 x i8> %200, <16 x i8> %201, i32 2)
>  %203 = extractelement <4 x float> %202, i32 0
>  %204 = extractelement <4 x float> %202, i32 3
>  %205 = fmul float %204, 6.528000e+04
>  %206 = fmul float %203, 2.550000e+02
>  %207 = fadd float %205, %206
>  %208 = fadd float %207, 5.000000e-01
>  %209 = call float @llvm.AMDIL.fraction.(float %208)
>  %210 = fsub float -0.000000e+00, %209
>  %211 = fadd float %210, %208
>  %212 = fmul float %26, %203
>  %213 = fadd float %212, 5.000000e-01
>  %214 = call float @llvm.AMDIL.fraction.(float %213)
>  %215 = fsub float -0.000000e+00, %214
>  %216 = fadd float %213, %215
>  %217 = fadd float %211, 5.000000e-01
>  %218 = fadd float %216, 5.000000e-01
>  %219 = fcmp oge float %46, 0.000000e+00
>  %220 = sext i1 %219 to i32
>  %221 = bitcast i32 %220 to float
>  %222 = bitcast float %221 to i32
>  %223 = icmp ne i32 %222, 0
>  %.55 = select i1 %223, float %218, float %217
>  %224 = fdiv float 1.000000e+00, %25
>  %225 = fmul float %29, %224
>  %226 = fadd float %225, 5.000000e-01
>  %227 = call float @llvm.AMDIL.fraction.(float %226)
>  %228 = fsub float -0.000000e+00, %227
>  %229 = fadd float %228, %226
>  %230 = fmul float %229, %.55
>  %231 = fsub float -0.000000e+00, %229
>  %232 = fcmp oge float %230, 0.000000e+00
>  %233 = sext i1 %232 to i32
>  %234 = bitcast i32 %233 to float
>  %235 = bitcast float %234 to i32
>  %236 = icmp ne i32 %235, 0
>  %temp24.1 = select i1 %236, float %229, float %231
>  %237 = fdiv float 1.000000e+00, %temp24.1
>  %238 = fmul float %237, %.55
>  %239 = call float @llvm.AMDIL.fraction.(float %238)
>  %240 = fdiv float 1.000000e+00, %229
>  %241 = fmul float %239, %temp24.1
>  %242 = fmul float %240, %.55
>  %243 = call float @llvm.AMDIL.fraction.(float %241)
>  %244 = call float @llvm.AMDIL.fraction.(float %242)
>  %245 = fsub float -0.000000e+00, %243
>  %246 = fadd float %245, %241
>  %247 = fsub float -0.000000e+00, %244
>  %248 = fadd float %247, %242
>  %249 = fadd float %183, %246
>  %250 = fadd float %186, %248
>  %251 = fmul float %249, %25
>  %252 = fmul float %250, %25
>  %253 = call float @llvm.AMDIL.fraction.(float %251)
>  %254 = call float @llvm.AMDIL.fraction.(float %252)
>  %255 = fsub float -0.000000e+00, %253
>  %256 = fadd float %255, %251
>  %257 = fsub float -0.000000e+00, %254
>  %258 = fadd float %257, %252
>  %259 = fadd float %256, 5.000000e-01
>  %260 = fadd float %258, 5.000000e-01
>  %261 = fdiv float 1.000000e+00, %29
>  %262 = fdiv float 1.000000e+00, %30
>  %263 = fmul float %259, %261
>  %264 = fmul float %260, %262
>  %265 = bitcast float %263 to i32
>  %266 = bitcast float %264 to i32
>  %267 = bitcast float 0.000000e+00 to i32
>  %268 = insertelement <4 x i32> undef, i32 %265, i32 0
>  %269 = insertelement <4 x i32> %268, i32 %266, i32 1
>  %270 = insertelement <4 x i32> %269, i32 %267, i32 2
>  %271 = insertelement <4 x i32> %270, i32 undef, i32 3
>  %272 = bitcast <8 x i32> %36 to <32 x i8>
>  %273 = bitcast <4 x i32> %38 to <16 x i8>
>  %274 = call <4 x float> @llvm.SI.samplel.v4i32(<4 x i32> %271, <32 x i8> %272, <16 x i8> %273, i32 2)
>  br label %ENDIF45
>
>ELSE47:                                           ; preds = %ELSE
>  %275 = bitcast float %43 to i32
>  %276 = bitcast float %44 to i32
>  %277 = insertelement <2 x i32> undef, i32 %275, i32 0
>  %278 = insertelement <2 x i32> %277, i32 %276, i32 1
>  %279 = bitcast <8 x i32> %36 to <32 x i8>
>  %280 = bitcast <4 x i32> %38 to <16 x i8>
>  %281 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %278, <32 x i8> %279, <16 x i8> %280, i32 2)
>  br label %ENDIF45
>
>ENDIF45:                                          ; preds = %ELSE47, %IF46
>  %.sink56 = phi <4 x float> [ %281, %ELSE47 ], [ %274, %IF46 ]
>  %282 = extractelement <4 x float> %.sink56, i32 0
>  %283 = extractelement <4 x float> %.sink56, i32 1
>  %284 = extractelement <4 x float> %.sink56, i32 2
>  br label %ENDIF
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.fraction.(float) #2
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.samplel.v4i32(<4 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                          ; BEFE0A7E
>	s_mov_b32 m0, s9                                                              ; BEFC0309
>	v_interp_p1_f32 v4, v0, 1, 1, [m0]                                            ; C8100500
>	v_interp_p2_f32 v4, [v4], v1, 1, 1, [m0]                                      ; C8110501
>	v_interp_p1_f32 v3, v0, 0, 1, [m0]                                            ; C80C0400
>	v_interp_p2_f32 v3, [v3], v1, 0, 1, [m0]                                      ; C80D0401
>	v_interp_p1_f32 v2, v0, 3, 0, [m0]                                            ; C8080300
>	v_interp_p2_f32 v2, [v2], v1, 3, 0, [m0]                                      ; C8090301
>	v_interp_p1_f32 v5, v0, 2, 0, [m0]                                            ; C8140200
>	v_interp_p2_f32 v5, [v5], v1, 2, 0, [m0]                                      ; C8150201
>	v_interp_p1_f32 v6, v0, 1, 0, [m0]                                            ; C8180100
>	v_interp_p2_f32 v6, [v6], v1, 1, 0, [m0]                                      ; C8190101
>	v_interp_p1_f32 v7, v0, 0, 0, [m0]                                            ; C81C0000
>	v_interp_p2_f32 v7, [v7], v1, 0, 0, [m0]                                      ; C81D0001
>	s_load_dwordx4 s[32:35], s[2:3], 0x0                                          ; C0900300
>	s_waitcnt lgkmcnt(0)                                                          ; BF8C007F
>	s_buffer_load_dword s0, s[32:35], 0x0                                         ; C2002100
>	s_waitcnt lgkmcnt(0)                                                          ; BF8C007F
>	v_cmp_neq_f32_e64 s[0:1], s0, -s0                                             ; D01A0000 40000000
>	v_cndmask_b32_e64 v0, 0, -1, s[0:1]                                           ; D2000000 00018280
>	v_cmp_eq_i32_e64 s[36:37], v0, 0                                              ; D1040024 00010100
>	s_buffer_load_dword s0, s[32:35], 0x4                                         ; C2002104
>	s_waitcnt lgkmcnt(0)                                                          ; BF8C007F
>	v_cmp_gt_f32_e64 s[2:3], s0, 0.5                                              ; D0080002 0001E000
>	v_cndmask_b32_e64 v1, 0, -1, s[2:3]                                           ; D2000801 00098280
>	s_buffer_load_dword s1, s[32:35], 0x8                                         ; C200A108
>	v_mov_b32_e32 v0, 0x44800000                                                  ; 7E0002FF 44800000
>	s_waitcnt lgkmcnt(0)                                                          ; BF8C007F
>	v_subrev_f32_e32 v0, s1, v0                                                   ; 0A000001
>	s_load_dwordx4 s[8:11], s[4:5], 0x0                                           ; C0840500
>	s_load_dwordx4 s[12:15], s[4:5], 0x4                                          ; C0860504
>	s_load_dwordx8 s[16:23], s[6:7], 0x0                                          ; C0C80700
>	s_load_dwordx8 s[24:31], s[6:7], 0x8                                          ; C0CC0708
>	s_buffer_load_dword s2, s[32:35], 0x11                                        ; C2012111
>	s_buffer_load_dword s3, s[32:35], 0x10                                        ; C201A110
>	s_buffer_load_dword s6, s[32:35], 0xd                                         ; C203210D
>	s_buffer_load_dword s7, s[32:35], 0xc                                         ; C203A10C
>	s_waitcnt lgkmcnt(0)                                                          ; BF8C007F
>	s_and_saveexec_b64 s[4:5], s[36:37]                                           ; BE842424
>	s_xor_b64 s[4:5], exec, s[4:5]                                                ; 8984047E
>	s_cbranch_execz BB0_11                                                        ; BF880000
>	v_cmp_eq_i32_e64 s[32:33], v1, 0                                              ; D1040020 00010101
>	s_and_saveexec_b64 s[32:33], s[32:33]                                         ; BEA02420
>	s_xor_b64 s[32:33], exec, s[32:33]                                            ; 89A0207E
>	image_sample v[8:11], 15, 0, 0, 0, 0, 0, 0, 0, v[3:4], s[24:31], s[12:15]     ; F0800F00 00660803
>	s_waitcnt vmcnt(0)                                                            ; BF8C0770
>	s_or_saveexec_b64 s[32:33], s[32:33]                                          ; BEA02520
>	s_xor_b64 exec, exec, s[32:33]                                                ; 89FE207E
>	s_cbranch_execz BB0_12                                                        ; BF880000
>	v_mul_f32_e32 v8, s6, v4                                                      ; 10100806
>	v_fract_f32_e32 v8, v8                                                        ; 7E104108
>	v_mad_f32 v8, s6, v4, -v8                                                     ; D2820008 84220806
>	v_add_f32_e32 v9, 0.5, v8                                                     ; 061210F0
>	v_rcp_f32_e32 v10, s6                                                         ; 7E145406
>	v_mul_f32_e32 v10, v10, v9                                                    ; 1014130A
>	v_mul_f32_e32 v13, s7, v3                                                     ; 101A0607
>	v_fract_f32_e32 v13, v13                                                      ; 7E1A410D
>	v_mad_f32 v13, s7, v3, -v13                                                   ; D282000D 84360607
>	v_add_f32_e32 v14, 0.5, v13                                                   ; 061C1AF0
>	v_rcp_f32_e32 v15, s7                                                         ; 7E1E5407
>	v_mul_f32_e32 v9, v15, v14                                                    ; 10121D0F
>	v_mov_b32_e32 v11, 0                                                          ; 7E160280
>	image_sample_l v[14:15], 9, 0, 0, 0, 0, 0, 0, 0, v[9:12], s[16:23], s[8:11]   ; F0900900 00440E09
>	s_waitcnt vmcnt(0)                                                            ; BF8C0770
>	v_mul_f32_e32 v16, 0x437f0000, v14                                            ; 10201CFF 437F0000
>	v_mov_b32_e32 v17, 0x477f0000                                                 ; 7E2202FF 477F0000
>	v_mad_f32 v16, v17, v15, v16                                                  ; D2820010 04421F11
>	v_add_f32_e32 v16, 0.5, v16                                                   ; 062020F0
>	v_fract_f32_e32 v17, v16                                                      ; 7E224110
>	v_subrev_f32_e32 v16, v17, v16                                                ; 0A202111
>	v_add_f32_e32 v16, 0.5, v16                                                   ; 062020F0
>	v_mad_f32 v14, s1, v14, 0.5                                                   ; D282000E 03C21C01
>	v_fract_f32_e32 v15, v14                                                      ; 7E1E410E
>	v_subrev_f32_e32 v14, v15, v14                                                ; 0A1C1D0F
>	v_add_f32_e32 v14, 0.5, v14                                                   ; 061C1CF0
>	v_cmp_ge_f32_e64 s[34:35], v0, 0                                              ; D00C0022 00010100
>	v_cndmask_b32_e64 v15, 0, -1, s[34:35]                                        ; D200080F 00898280
>	v_cmp_ne_i32_e64 s[34:35], v15, 0                                             ; D10A0022 0001010F
>	v_cndmask_b32_e64 v14, v16, v14, s[34:35]                                     ; D200000E 008A1D10
>	v_rcp_f32_e32 v15, s0                                                         ; 7E1E5400
>	v_mad_f32 v15, s3, v15, 0.5                                                   ; D282000F 03C21E03
>	v_fract_f32_e32 v16, v15                                                      ; 7E20410F
>	v_subrev_f32_e32 v15, v16, v15                                                ; 0A1E1F10
>	v_mul_f32_e32 v16, v14, v15                                                   ; 10201F0E
>	v_cmp_ge_f32_e64 s[34:35], v16, 0                                             ; D00C0022 00010110
>	v_cndmask_b32_e64 v16, 0, -1, s[34:35]                                        ; D2000010 00898280
>	v_cmp_ne_i32_e64 s[34:35], v16, 0                                             ; D10A0022 00010110
>	v_xor_b32_e32 v16, 0x80000000, v15                                            ; 3A201EFF 80000000
>	v_cndmask_b32_e64 v16, v16, v15, s[34:35]                                     ; D2000010 008A1F10
>	v_rcp_f32_e32 v17, v16                                                        ; 7E225510
>	v_mul_f32_e32 v17, v14, v17                                                   ; 1022230E
>	v_fract_f32_e32 v17, v17                                                      ; 7E224111
>	v_mul_f32_e32 v18, v16, v17                                                   ; 10242310
>	v_fract_f32_e32 v18, v18                                                      ; 7E244112
>	v_mad_f32 v16, v17, v16, -v18                                                 ; D2820010 844A2111
>	v_mad_f32 v13, s7, v3, -v13                                                   ; D282000D 84360607
>	v_add_f32_e32 v13, v16, v13                                                   ; 061A1B10
>	v_mul_f32_e32 v16, s0, v13                                                    ; 10201A00
>	v_fract_f32_e32 v16, v16                                                      ; 7E204110
>	v_mad_f32 v13, v13, s0, -v16                                                  ; D282000D 8440010D
>	v_add_f32_e32 v13, 0.5, v13                                                   ; 061A1AF0
>	v_rcp_f32_e32 v16, s3                                                         ; 7E205403
>	v_mul_f32_e32 v9, v16, v13                                                    ; 10121B10
>	v_rcp_f32_e32 v13, v15                                                        ; 7E1A550F
>	v_mul_f32_e32 v15, v14, v13                                                   ; 101E1B0E
>	v_fract_f32_e32 v15, v15                                                      ; 7E1E410F
>	v_mad_f32 v13, v13, v14, -v15                                                 ; D282000D 843E1D0D
>	v_mad_f32 v8, s6, v4, -v8                                                     ; D2820008 84220806
>	v_add_f32_e32 v8, v13, v8                                                     ; 0610110D
>	v_mul_f32_e32 v13, s0, v8                                                     ; 101A1000
>	v_fract_f32_e32 v13, v13                                                      ; 7E1A410D
>	v_mad_f32 v8, v8, s0, -v13                                                    ; D2820008 84340108
>	v_add_f32_e32 v8, 0.5, v8                                                     ; 061010F0
>	v_rcp_f32_e32 v13, s2                                                         ; 7E1A5402
>	v_mul_f32_e32 v10, v13, v8                                                    ; 1014110D
>	image_sample_l v[8:11], 15, 0, 0, 0, 0, 0, 0, 0, v[9:12], s[24:31], s[12:15]  ; F0900F00 00660809
>	s_waitcnt vmcnt(0)                                                            ; BF8C0770
>	s_or_b64 exec, exec, s[32:33]                                                 ; 88FE207E
>	v_mov_b32_e32 v12, v8                                                         ; 7E180308
>	v_mov_b32_e32 v13, v9                                                         ; 7E1A0309
>	v_mov_b32_e32 v14, v10                                                        ; 7E1C030A
>	v_mov_b32_e32 v15, v11                                                        ; 7E1E030B
>	s_or_saveexec_b64 s[4:5], s[4:5]                                              ; BE842504
>	s_xor_b64 exec, exec, s[4:5]                                                  ; 89FE047E
>	s_cbranch_execz BB0_4                                                         ; BF880000
>	v_cmp_eq_i32_e64 s[32:33], v1, 0                                              ; D1040020 00010101
>	s_and_saveexec_b64 s[32:33], s[32:33]                                         ; BEA02420
>	s_xor_b64 s[32:33], exec, s[32:33]                                            ; 89A0207E
>	image_sample v[12:15], 15, 0, 0, 0, 0, 0, 0, 0, v[3:4], s[24:31], s[12:15]    ; F0800F00 00660C03
>	v_mov_b32_e32 v10, v5                                                         ; 7E140305
>	v_mov_b32_e32 v9, v6                                                          ; 7E120306
>	v_mov_b32_e32 v8, v7                                                          ; 7E100307
>	s_waitcnt vmcnt(0)                                                            ; BF8C0770
>	s_or_saveexec_b64 s[32:33], s[32:33]                                          ; BEA02520
>	s_waitcnt expcnt(0)                                                           ; BF8C070F
>	s_xor_b64 exec, exec, s[32:33]                                                ; 89FE207E
>	s_cbranch_execz BB0_3                                                         ; BF880000
>	v_mul_f32_e32 v1, s6, v4                                                      ; 10020806
>	v_fract_f32_e32 v1, v1                                                        ; 7E024101
>	v_mad_f32 v1, s6, v4, -v1                                                     ; D2820001 84060806
>	v_add_f32_e32 v8, 0.5, v1                                                     ; 061002F0
>	v_rcp_f32_e32 v9, s6                                                          ; 7E125406
>	v_mul_f32_e32 v9, v9, v8                                                      ; 10121109
>	v_mul_f32_e32 v12, s7, v3                                                     ; 10180607
>	v_fract_f32_e32 v12, v12                                                      ; 7E18410C
>	v_mad_f32 v12, s7, v3, -v12                                                   ; D282000C 84320607
>	v_add_f32_e32 v13, 0.5, v12                                                   ; 061A18F0
>	v_rcp_f32_e32 v14, s7                                                         ; 7E1C5407
>	v_mul_f32_e32 v8, v14, v13                                                    ; 10101B0E
>	v_mov_b32_e32 v10, 0                                                          ; 7E140280
>	image_sample_l v[13:14], 9, 0, 0, 0, 0, 0, 0, 0, v[8:11], s[16:23], s[8:11]   ; F0900900 00440D08
>	s_waitcnt vmcnt(0)                                                            ; BF8C0770
>	v_mul_f32_e32 v15, 0x437f0000, v13                                            ; 101E1AFF 437F0000
>	v_mov_b32_e32 v16, 0x477f0000                                                 ; 7E2002FF 477F0000
>	v_mad_f32 v15, v16, v14, v15                                                  ; D282000F 043E1D10
>	v_add_f32_e32 v15, 0.5, v15                                                   ; 061E1EF0
>	v_fract_f32_e32 v16, v15                                                      ; 7E20410F
>	v_subrev_f32_e32 v15, v16, v15                                                ; 0A1E1F10
>	v_add_f32_e32 v15, 0.5, v15                                                   ; 061E1EF0
>	v_mad_f32 v13, s1, v13, 0.5                                                   ; D282000D 03C21A01
>	v_fract_f32_e32 v14, v13                                                      ; 7E1C410D
>	v_subrev_f32_e32 v13, v14, v13                                                ; 0A1A1B0E
>	v_add_f32_e32 v13, 0.5, v13                                                   ; 061A1AF0
>	v_cmp_ge_f32_e64 s[34:35], v0, 0                                              ; D00C0022 00010100
>	v_cndmask_b32_e64 v0, 0, -1, s[34:35]                                         ; D2000000 00898280
>	v_cmp_ne_i32_e64 s[34:35], v0, 0                                              ; D10A0022 00010100
>	v_cndmask_b32_e64 v0, v15, v13, s[34:35]                                      ; D2000000 188A1B0F
>	v_rcp_f32_e32 v13, s0                                                         ; 7E1A5400
>	v_mad_f32 v13, s3, v13, 0.5                                                   ; D282000D 03C21A03
>	v_fract_f32_e32 v14, v13                                                      ; 7E1C410D
>	v_subrev_f32_e32 v13, v14, v13                                                ; 0A1A1B0E
>	v_mul_f32_e32 v14, v0, v13                                                    ; 101C1B00
>	v_cmp_ge_f32_e64 s[34:35], v14, 0                                             ; D00C0022 0001010E
>	v_cndmask_b32_e64 v14, 0, -1, s[34:35]                                        ; D200000E 00898280
>	v_cmp_ne_i32_e64 s[34:35], v14, 0                                             ; D10A0022 0001010E
>	v_xor_b32_e32 v14, 0x80000000, v13                                            ; 3A1C1AFF 80000000
>	v_cndmask_b32_e64 v14, v14, v13, s[34:35]                                     ; D200000E 108A1B0E
>	v_rcp_f32_e32 v15, v14                                                        ; 7E1E550E
>	v_mul_f32_e32 v15, v15, v0                                                    ; 101E010F
>	v_fract_f32_e32 v15, v15                                                      ; 7E1E410F
>	v_mul_f32_e32 v16, v14, v15                                                   ; 10201F0E
>	v_fract_f32_e32 v16, v16                                                      ; 7E204110
>	v_mad_f32 v14, v15, v14, -v16                                                 ; D282000E 84421D0F
>	v_mad_f32 v12, s7, v3, -v12                                                   ; D282000C 84320607
>	v_add_f32_e32 v12, v14, v12                                                   ; 0618190E
>	v_mul_f32_e32 v14, s0, v12                                                    ; 101C1800
>	v_fract_f32_e32 v14, v14                                                      ; 7E1C410E
>	v_mad_f32 v12, v12, s0, -v14                                                  ; D282000C 8438010C
>	v_add_f32_e32 v12, 0.5, v12                                                   ; 061818F0
>	v_rcp_f32_e32 v14, s3                                                         ; 7E1C5403
>	v_mul_f32_e32 v8, v14, v12                                                    ; 1010190E
>	v_rcp_f32_e32 v12, v13                                                        ; 7E18550D
>	v_mul_f32_e32 v13, v0, v12                                                    ; 101A1900
>	v_fract_f32_e32 v13, v13                                                      ; 7E1A410D
>	v_mad_f32 v0, v12, v0, -v13                                                   ; D2820000 8436010C
>	v_mad_f32 v1, s6, v4, -v1                                                     ; D2820001 84060806
>	v_add_f32_e32 v0, v0, v1                                                      ; 06000300
>	v_mul_f32_e32 v1, s0, v0                                                      ; 10020000
>	v_fract_f32_e32 v1, v1                                                        ; 7E024101
>	v_mad_f32 v0, v0, s0, -v1                                                     ; D2820000 84040100
>	v_add_f32_e32 v0, 0.5, v0                                                     ; 060000F0
>	v_rcp_f32_e32 v1, s2                                                          ; 7E025402
>	v_mul_f32_e32 v9, v1, v0                                                      ; 10120101
>	image_sample_l v[12:15], 15, 0, 0, 0, 0, 0, 0, 0, v[8:11], s[24:31], s[12:15] ; F0900F00 00660C08
>	v_mov_b32_e32 v10, v5                                                         ; 7E140305
>	v_mov_b32_e32 v9, v6                                                          ; 7E120306
>	v_mov_b32_e32 v8, v7                                                          ; 7E100307
>	s_waitcnt vmcnt(0)                                                            ; BF8C0770
>	s_or_b64 exec, exec, s[32:33]                                                 ; 88FE207E
>	s_or_b64 exec, exec, s[4:5]                                                   ; 88FE047E
>	v_cvt_pkrtz_f16_f32_e32 v0, v8, v9                                            ; 5E001308
>	v_mul_f32_e32 v1, v2, v15                                                     ; 10021F02
>	v_cvt_pkrtz_f16_f32_e32 v1, v10, v1                                           ; 5E02030A
>	exp 15, 0, 1, 1, 1, v0, v1, v0, v1                                            ; F8001C0F 01000100
>	s_endpgm                                                                      ; BF810000
>
>SGPRS: 40
>VGPRS: 20
>Code Size: 1096 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL IN[2]
>DCL IN[3]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL OUT[2], GENERIC[10]
>DCL OUT[3], GENERIC[11]
>DCL CONST[0..24]
>DCL TEMP[0..3], LOCAL
>DCL ADDR[0]
>IMM[0] FLT32 {    0.0000,     1.0000,     0.0000,     0.0000}
>  0: F2I TEMP[0].x, IN[3].xxxx
>  1: UARL ADDR[0].x, TEMP[0].xxxx
>  2: UARL ADDR[0].x, TEMP[0].xxxx
>  3: MOV TEMP[1], CONST[ADDR[0].x+4].zwzw
>  4: UARL ADDR[0].x, TEMP[0].xxxx
>  5: MAD TEMP[1].xy, IN[0], CONST[ADDR[0].x+4], TEMP[1]
>  6: UARL ADDR[0].x, TEMP[0].xxxx
>  7: UARL ADDR[0].x, TEMP[0].xxxx
>  8: MOV TEMP[2], CONST[ADDR[0].x+13].zwzw
>  9: UARL ADDR[0].x, TEMP[0].xxxx
> 10: MAD TEMP[2].xy, IN[1], CONST[ADDR[0].x+13], TEMP[2]
> 11: MOV TEMP[2].xy, TEMP[2].xyxx
> 12: MUL TEMP[3], TEMP[1].yyyy, CONST[1]
> 13: MAD TEMP[0], TEMP[1].xxxx, CONST[0], TEMP[3]
> 14: ADD TEMP[0], TEMP[0], CONST[3]
> 15: ADD TEMP[1].x, CONST[24].xxxx, CONST[24].yyyy
> 16: MUL TEMP[1], TEMP[0], TEMP[1].xxxx
> 17: MUL TEMP[0], CONST[22], IN[2]
> 18: MUL TEMP[3].w, TEMP[0].wwww, CONST[23].wwww
> 19: MOV TEMP[3].w, TEMP[3].wwww
> 20: MOV TEMP[3].xyz, TEMP[0].xyzx
> 21: MOV TEMP[0].xy, IN[3].yzyy
> 22: MOV TEMP[2].zw, IMM[0].yyxy
> 23: MOV TEMP[0].zw, IMM[0].yyxy
> 24: MOV OUT[1], TEMP[3]
> 25: MOV OUT[2], TEMP[2]
> 26: MOV OUT[0], TEMP[1]
> 27: MOV OUT[3], TEMP[0]
> 28: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 0)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 4)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %12, i32 8)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %12, i32 12)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %12, i32 16)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %12, i32 20)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %12, i32 24)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %12, i32 28)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %12, i32 48)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %12, i32 52)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %12, i32 56)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %12, i32 60)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %12, i32 352)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %12, i32 356)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %12, i32 360)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %12, i32 364)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %12, i32 380)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %12, i32 384)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %12, i32 388)
>  %32 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %33 = load <16 x i8> addrspace(2)* %32, !tbaa !0
>  %34 = add i32 %5, %7
>  %35 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %33, i32 0, i32 %34)
>  %36 = extractelement <4 x float> %35, i32 0
>  %37 = extractelement <4 x float> %35, i32 1
>  %38 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %39 = load <16 x i8> addrspace(2)* %38, !tbaa !0
>  %40 = add i32 %5, %7
>  %41 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %39, i32 0, i32 %40)
>  %42 = extractelement <4 x float> %41, i32 0
>  %43 = extractelement <4 x float> %41, i32 1
>  %44 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 2
>  %45 = load <16 x i8> addrspace(2)* %44, !tbaa !0
>  %46 = add i32 %5, %7
>  %47 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %45, i32 0, i32 %46)
>  %48 = extractelement <4 x float> %47, i32 0
>  %49 = extractelement <4 x float> %47, i32 1
>  %50 = extractelement <4 x float> %47, i32 2
>  %51 = extractelement <4 x float> %47, i32 3
>  %52 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 3
>  %53 = load <16 x i8> addrspace(2)* %52, !tbaa !0
>  %54 = add i32 %5, %7
>  %55 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %53, i32 0, i32 %54)
>  %56 = extractelement <4 x float> %55, i32 0
>  %57 = extractelement <4 x float> %55, i32 1
>  %58 = extractelement <4 x float> %55, i32 2
>  %59 = fptosi float %56 to i32
>  %60 = bitcast i32 %59 to float
>  %61 = bitcast float %60 to i32
>  %62 = shl i32 %61, 4
>  %63 = add i32 %62, 72
>  %64 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %63)
>  %65 = shl i32 %61, 4
>  %66 = add i32 %65, 76
>  %67 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %66)
>  %68 = bitcast float %60 to i32
>  %69 = shl i32 %68, 4
>  %70 = add i32 %69, 64
>  %71 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %70)
>  %72 = fmul float %36, %71
>  %73 = fadd float %72, %64
>  %74 = shl i32 %68, 4
>  %75 = add i32 %74, 68
>  %76 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %75)
>  %77 = fmul float %37, %76
>  %78 = fadd float %77, %67
>  %79 = bitcast float %60 to i32
>  %80 = shl i32 %79, 4
>  %81 = add i32 %80, 216
>  %82 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %81)
>  %83 = shl i32 %79, 4
>  %84 = add i32 %83, 220
>  %85 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %84)
>  %86 = bitcast float %60 to i32
>  %87 = shl i32 %86, 4
>  %88 = add i32 %87, 208
>  %89 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %88)
>  %90 = fmul float %42, %89
>  %91 = fadd float %90, %82
>  %92 = shl i32 %86, 4
>  %93 = add i32 %92, 212
>  %94 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %93)
>  %95 = fmul float %43, %94
>  %96 = fadd float %95, %85
>  %97 = fmul float %78, %17
>  %98 = fmul float %78, %18
>  %99 = fmul float %78, %19
>  %100 = fmul float %78, %20
>  %101 = fmul float %73, %13
>  %102 = fadd float %101, %97
>  %103 = fmul float %73, %14
>  %104 = fadd float %103, %98
>  %105 = fmul float %73, %15
>  %106 = fadd float %105, %99
>  %107 = fmul float %73, %16
>  %108 = fadd float %107, %100
>  %109 = fadd float %102, %21
>  %110 = fadd float %104, %22
>  %111 = fadd float %106, %23
>  %112 = fadd float %108, %24
>  %113 = fadd float %30, %31
>  %114 = fmul float %109, %113
>  %115 = fmul float %110, %113
>  %116 = fmul float %111, %113
>  %117 = fmul float %112, %113
>  %118 = fmul float %25, %48
>  %119 = fmul float %26, %49
>  %120 = fmul float %27, %50
>  %121 = fmul float %28, %51
>  %122 = fmul float %121, %29
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %118, float %119, float %120, float %122)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %91, float %96, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 34, i32 0, float %57, float %58, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %114, float %115, float %116, float %117)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                              ; 4A00000A
>	s_load_dwordx4 s[4:7], s[8:9], 0x0                     ; C0820900
>	s_load_dwordx4 s[12:15], s[8:9], 0x4                   ; C0860904
>	s_load_dwordx4 s[16:19], s[8:9], 0x8                   ; C0880908
>	s_load_dwordx4 s[8:11], s[8:9], 0xc                    ; C084090C
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[16:19], 0 idxen  ; E00C2000 80040100
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                     ; C0800300
>	s_waitcnt vmcnt(0) lgkmcnt(0)                          ; BF8C0070
>	s_buffer_load_dword s16, s[0:3], 0x5a                  ; C208015A
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_mul_f32_e32 v5, s16, v3                              ; 100A0610
>	s_buffer_load_dword s16, s[0:3], 0x59                  ; C2080159
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_mul_f32_e32 v6, s16, v2                              ; 100C0410
>	s_buffer_load_dword s16, s[0:3], 0x58                  ; C2080158
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_mul_f32_e32 v7, s16, v1                              ; 100E0210
>	s_buffer_load_dword s16, s[0:3], 0x5b                  ; C208015B
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_mul_f32_e32 v1, s16, v4                              ; 10020810
>	s_buffer_load_dword s16, s[0:3], 0x5f                  ; C208015F
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_mul_f32_e32 v1, s16, v1                              ; 10020210
>	exp 15, 32, 0, 0, 0, v7, v6, v5, v1                    ; F800020F 01050607
>	s_waitcnt expcnt(0)                                    ; BF8C070F
>	buffer_load_format_xyzw v[1:4], v0, s[8:11], 0 idxen   ; E00C2000 80020100
>	s_waitcnt vmcnt(0)                                     ; BF8C0770
>	v_cvt_i32_f32_e32 v5, v1                               ; 7E0A1101
>	v_lshlrev_b32_e32 v5, 4, v5                            ; 340A0A84
>	v_add_i32_e32 v6, 0xdc, v5                             ; 4A0C0AFF 000000DC
>	buffer_load_dword v6, v6, s[0:3], 0 offen              ; E0301000 80000606
>	v_add_i32_e32 v7, 0xd4, v5                             ; 4A0E0AFF 000000D4
>	buffer_load_dword v7, v7, s[0:3], 0 offen              ; E0301000 80000707
>	buffer_load_format_xyzw v[8:11], v0, s[12:15], 0 idxen ; E00C2000 80030800
>	s_waitcnt vmcnt(0)                                     ; BF8C0770
>	v_mad_f32 v6, v9, v7, v6                               ; D2820006 041A0F09
>	v_add_i32_e32 v7, 0xd8, v5                             ; 4A0E0AFF 000000D8
>	buffer_load_dword v7, v7, s[0:3], 0 offen              ; E0301000 80000707
>	v_add_i32_e32 v12, 0xd0, v5                            ; 4A180AFF 000000D0
>	buffer_load_dword v12, v12, s[0:3], 0 offen            ; E0301000 80000C0C
>	s_waitcnt vmcnt(0)                                     ; BF8C0770
>	v_mad_f32 v7, v8, v12, v7                              ; D2820007 041E1908
>	v_mov_b32_e32 v8, 1.0                                  ; 7E1002F2
>	v_mov_b32_e32 v9, 0                                    ; 7E120280
>	exp 15, 33, 0, 0, 0, v7, v6, v9, v8                    ; F800021F 08090607
>	exp 15, 34, 0, 0, 0, v2, v3, v9, v8                    ; F800022F 08090302
>	s_waitcnt expcnt(0)                                    ; BF8C070F
>	v_add_i32_e32 v1, 0x48, v5                             ; 4A020AFF 00000048
>	buffer_load_dword v1, v1, s[0:3], 0 offen              ; E0301000 80000101
>	v_add_i32_e32 v2, 64, v5                               ; 4A040AC0
>	buffer_load_dword v2, v2, s[0:3], 0 offen              ; E0301000 80000202
>	buffer_load_format_xyzw v[6:9], v0, s[4:7], 0 idxen    ; E00C2000 80010600
>	s_waitcnt vmcnt(0)                                     ; BF8C0770
>	v_mad_f32 v0, v6, v2, v1                               ; D2820000 04060506
>	v_add_i32_e32 v1, 0x4c, v5                             ; 4A020AFF 0000004C
>	buffer_load_dword v1, v1, s[0:3], 0 offen              ; E0301000 80000101
>	v_add_i32_e32 v2, 0x44, v5                             ; 4A040AFF 00000044
>	buffer_load_dword v2, v2, s[0:3], 0 offen              ; E0301000 80000202
>	s_waitcnt vmcnt(0)                                     ; BF8C0770
>	v_mad_f32 v1, v7, v2, v1                               ; D2820001 04060507
>	s_buffer_load_dword s4, s[0:3], 0x7                    ; C2020107
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_mul_f32_e32 v2, s4, v1                               ; 10040204
>	s_buffer_load_dword s4, s[0:3], 0x3                    ; C2020103
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_mad_f32 v2, v0, s4, v2                               ; D2820002 04080900
>	s_buffer_load_dword s4, s[0:3], 0xf                    ; C202010F
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_add_f32_e32 v2, s4, v2                               ; 06040404
>	s_buffer_load_dword s4, s[0:3], 0x61                   ; C2020161
>	s_buffer_load_dword s5, s[0:3], 0x60                   ; C2028160
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_mov_b32_e32 v3, s4                                   ; 7E060204
>	v_add_f32_e32 v3, s5, v3                               ; 06060605
>	v_mul_f32_e32 v2, v3, v2                               ; 10040503
>	s_buffer_load_dword s4, s[0:3], 0x6                    ; C2020106
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_mul_f32_e32 v4, s4, v1                               ; 10080204
>	s_buffer_load_dword s4, s[0:3], 0x2                    ; C2020102
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_mad_f32 v4, v0, s4, v4                               ; D2820004 04100900
>	s_buffer_load_dword s4, s[0:3], 0xe                    ; C202010E
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_add_f32_e32 v4, s4, v4                               ; 06080804
>	v_mul_f32_e32 v4, v3, v4                               ; 10080903
>	s_buffer_load_dword s4, s[0:3], 0x5                    ; C2020105
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_mul_f32_e32 v5, s4, v1                               ; 100A0204
>	s_buffer_load_dword s4, s[0:3], 0x1                    ; C2020101
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_mad_f32 v5, v0, s4, v5                               ; D2820005 04140900
>	s_buffer_load_dword s4, s[0:3], 0xd                    ; C202010D
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_add_f32_e32 v5, s4, v5                               ; 060A0A04
>	v_mul_f32_e32 v5, v3, v5                               ; 100A0B03
>	s_buffer_load_dword s4, s[0:3], 0x4                    ; C2020104
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_mul_f32_e32 v1, s4, v1                               ; 10020204
>	s_buffer_load_dword s4, s[0:3], 0x0                    ; C2020100
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_mad_f32 v0, v0, s4, v1                               ; D2820000 04040900
>	s_buffer_load_dword s0, s[0:3], 0xc                    ; C200010C
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_add_f32_e32 v0, s0, v0                               ; 06000000
>	v_mul_f32_e32 v0, v3, v0                               ; 10000103
>	exp 15, 12, 0, 1, 0, v0, v5, v4, v2                    ; F80008CF 02040500
>	s_endpgm                                               ; BF810000
>
>SGPRS: 24
>VGPRS: 16
>Code Size: 556 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], POSITION, LINEAR
>DCL IN[1], GENERIC[9], PERSPECTIVE
>DCL IN[2], GENERIC[10], PERSPECTIVE
>DCL IN[3], GENERIC[11], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SAMP[1]
>DCL CONST[4]
>DCL CONST[0..1]
>DCL TEMP[0]
>DCL TEMP[1..5], LOCAL
>IMM[0] FLT32 {    0.0000,     1.0000,     0.0000,     0.0000}
>  0: MOV TEMP[0], IN[0]
>  1: MAD TEMP[0].y, IN[0], CONST[4].xxxx, CONST[4].yyyy
>  2: MAD TEMP[1].xy, TEMP[0], CONST[1].xyxx, CONST[1].zwzz
>  3: MOV TEMP[2].xz, -CONST[0]
>  4: ADD TEMP[3].xyz, TEMP[2].xxxx, TEMP[1].xxxx
>  5: FSLT TEMP[4].xyz, TEMP[3].xyzz, IMM[0].xxxx
>  6: OR TEMP[5].x, TEMP[4].xxxx, TEMP[4].zzzz
>  7: OR TEMP[4].x, TEMP[5].xxxx, TEMP[4].yyyy
>  8: AND TEMP[4].x, TEMP[4].xxxx, IMM[0].yyyy
>  9: KILL_IF -TEMP[4].xxxx
> 10: ADD TEMP[3].xyz, CONST[0].yyyy, -TEMP[1].xxxx
> 11: FSLT TEMP[4].xyz, TEMP[3].xyzz, IMM[0].xxxx
> 12: OR TEMP[5].x, TEMP[4].xxxx, TEMP[4].zzzz
> 13: OR TEMP[4].x, TEMP[5].xxxx, TEMP[4].yyyy
> 14: AND TEMP[4].x, TEMP[4].xxxx, IMM[0].yyyy
> 15: KILL_IF -TEMP[4].xxxx
> 16: ADD TEMP[3].xyz, TEMP[2].zzzz, TEMP[1].yyyy
> 17: FSLT TEMP[2].xyz, TEMP[3].xyzz, IMM[0].xxxx
> 18: OR TEMP[4].x, TEMP[2].xxxx, TEMP[2].zzzz
> 19: OR TEMP[2].x, TEMP[4].xxxx, TEMP[2].yyyy
> 20: AND TEMP[2].x, TEMP[2].xxxx, IMM[0].yyyy
> 21: KILL_IF -TEMP[2].xxxx
> 22: ADD TEMP[3].xyz, CONST[0].wwww, -TEMP[1].yyyy
> 23: FSLT TEMP[1].xyz, TEMP[3].xyzz, IMM[0].xxxx
> 24: OR TEMP[2].x, TEMP[1].xxxx, TEMP[1].zzzz
> 25: OR TEMP[1].x, TEMP[2].xxxx, TEMP[1].yyyy
> 26: AND TEMP[1].x, TEMP[1].xxxx, IMM[0].yyyy
> 27: KILL_IF -TEMP[1].xxxx
> 28: MOV TEMP[1].xy, IN[2].xyyy
> 29: TEX TEMP[1], TEMP[1], SAMP[1], 2D
> 30: MUL TEMP[3], TEMP[1], IN[3].yyyy
> 31: MOV TEMP[1].xy, IN[2].xyyy
> 32: TEX TEMP[1], TEMP[1], SAMP[0], 2D
> 33: MAD TEMP[3], IN[3].xxxx, TEMP[1], TEMP[3]
> 34: MUL TEMP[1], TEMP[3], IN[1]
> 35: MOV OUT[0], TEMP[1]
> 36: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %23 = load <16 x i8> addrspace(2)* %22, !tbaa !0
>  %24 = call float @llvm.SI.load.const(<16 x i8> %23, i32 0)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %23, i32 4)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %23, i32 8)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %23, i32 12)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %23, i32 16)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %23, i32 20)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %23, i32 24)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %23, i32 28)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %23, i32 64)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %23, i32 68)
>  %34 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %35 = load <8 x i32> addrspace(2)* %34, !tbaa !0
>  %36 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %37 = load <4 x i32> addrspace(2)* %36, !tbaa !0
>  %38 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 1
>  %39 = load <8 x i32> addrspace(2)* %38, !tbaa !0
>  %40 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 1
>  %41 = load <4 x i32> addrspace(2)* %40, !tbaa !0
>  %42 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %43 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %44 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %5, <2 x i32> %7)
>  %45 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %5, <2 x i32> %7)
>  %46 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %5, <2 x i32> %7)
>  %47 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %5, <2 x i32> %7)
>  %48 = call float @llvm.SI.fs.interp(i32 0, i32 2, i32 %5, <2 x i32> %7)
>  %49 = call float @llvm.SI.fs.interp(i32 1, i32 2, i32 %5, <2 x i32> %7)
>  %50 = fmul float %15, %32
>  %51 = fadd float %50, %33
>  %52 = fmul float %14, %28
>  %53 = fadd float %52, %30
>  %54 = fmul float %51, %29
>  %55 = fadd float %54, %31
>  %56 = fsub float -0.000000e+00, %24
>  %57 = fsub float -0.000000e+00, %26
>  %58 = fadd float %56, %53
>  %59 = fadd float %56, %53
>  %60 = fadd float %56, %53
>  %61 = fcmp olt float %58, 0.000000e+00
>  %62 = sext i1 %61 to i32
>  %63 = fcmp olt float %59, 0.000000e+00
>  %64 = sext i1 %63 to i32
>  %65 = fcmp olt float %60, 0.000000e+00
>  %66 = sext i1 %65 to i32
>  %67 = bitcast i32 %62 to float
>  %68 = bitcast i32 %64 to float
>  %69 = bitcast i32 %66 to float
>  %70 = bitcast float %67 to i32
>  %71 = bitcast float %69 to i32
>  %72 = or i32 %70, %71
>  %73 = bitcast i32 %72 to float
>  %74 = bitcast float %73 to i32
>  %75 = bitcast float %68 to i32
>  %76 = or i32 %74, %75
>  %77 = bitcast i32 %76 to float
>  %78 = bitcast float %77 to i32
>  %79 = and i32 %78, 1065353216
>  %80 = bitcast i32 %79 to float
>  %81 = fsub float -0.000000e+00, %80
>  %82 = fsub float -0.000000e+00, %80
>  %83 = fsub float -0.000000e+00, %80
>  %84 = fsub float -0.000000e+00, %80
>  call void @llvm.AMDGPU.kill(float %81)
>  call void @llvm.AMDGPU.kill(float %82)
>  call void @llvm.AMDGPU.kill(float %83)
>  call void @llvm.AMDGPU.kill(float %84)
>  %85 = fsub float -0.000000e+00, %53
>  %86 = fadd float %25, %85
>  %87 = fsub float -0.000000e+00, %53
>  %88 = fadd float %25, %87
>  %89 = fsub float -0.000000e+00, %53
>  %90 = fadd float %25, %89
>  %91 = fcmp olt float %86, 0.000000e+00
>  %92 = sext i1 %91 to i32
>  %93 = fcmp olt float %88, 0.000000e+00
>  %94 = sext i1 %93 to i32
>  %95 = fcmp olt float %90, 0.000000e+00
>  %96 = sext i1 %95 to i32
>  %97 = bitcast i32 %92 to float
>  %98 = bitcast i32 %94 to float
>  %99 = bitcast i32 %96 to float
>  %100 = bitcast float %97 to i32
>  %101 = bitcast float %99 to i32
>  %102 = or i32 %100, %101
>  %103 = bitcast i32 %102 to float
>  %104 = bitcast float %103 to i32
>  %105 = bitcast float %98 to i32
>  %106 = or i32 %104, %105
>  %107 = bitcast i32 %106 to float
>  %108 = bitcast float %107 to i32
>  %109 = and i32 %108, 1065353216
>  %110 = bitcast i32 %109 to float
>  %111 = fsub float -0.000000e+00, %110
>  %112 = fsub float -0.000000e+00, %110
>  %113 = fsub float -0.000000e+00, %110
>  %114 = fsub float -0.000000e+00, %110
>  call void @llvm.AMDGPU.kill(float %111)
>  call void @llvm.AMDGPU.kill(float %112)
>  call void @llvm.AMDGPU.kill(float %113)
>  call void @llvm.AMDGPU.kill(float %114)
>  %115 = fadd float %57, %55
>  %116 = fadd float %57, %55
>  %117 = fadd float %57, %55
>  %118 = fcmp olt float %115, 0.000000e+00
>  %119 = sext i1 %118 to i32
>  %120 = fcmp olt float %116, 0.000000e+00
>  %121 = sext i1 %120 to i32
>  %122 = fcmp olt float %117, 0.000000e+00
>  %123 = sext i1 %122 to i32
>  %124 = bitcast i32 %119 to float
>  %125 = bitcast i32 %121 to float
>  %126 = bitcast i32 %123 to float
>  %127 = bitcast float %124 to i32
>  %128 = bitcast float %126 to i32
>  %129 = or i32 %127, %128
>  %130 = bitcast i32 %129 to float
>  %131 = bitcast float %130 to i32
>  %132 = bitcast float %125 to i32
>  %133 = or i32 %131, %132
>  %134 = bitcast i32 %133 to float
>  %135 = bitcast float %134 to i32
>  %136 = and i32 %135, 1065353216
>  %137 = bitcast i32 %136 to float
>  %138 = fsub float -0.000000e+00, %137
>  %139 = fsub float -0.000000e+00, %137
>  %140 = fsub float -0.000000e+00, %137
>  %141 = fsub float -0.000000e+00, %137
>  call void @llvm.AMDGPU.kill(float %138)
>  call void @llvm.AMDGPU.kill(float %139)
>  call void @llvm.AMDGPU.kill(float %140)
>  call void @llvm.AMDGPU.kill(float %141)
>  %142 = fsub float -0.000000e+00, %55
>  %143 = fadd float %27, %142
>  %144 = fsub float -0.000000e+00, %55
>  %145 = fadd float %27, %144
>  %146 = fsub float -0.000000e+00, %55
>  %147 = fadd float %27, %146
>  %148 = fcmp olt float %143, 0.000000e+00
>  %149 = sext i1 %148 to i32
>  %150 = fcmp olt float %145, 0.000000e+00
>  %151 = sext i1 %150 to i32
>  %152 = fcmp olt float %147, 0.000000e+00
>  %153 = sext i1 %152 to i32
>  %154 = bitcast i32 %149 to float
>  %155 = bitcast i32 %151 to float
>  %156 = bitcast i32 %153 to float
>  %157 = bitcast float %154 to i32
>  %158 = bitcast float %156 to i32
>  %159 = or i32 %157, %158
>  %160 = bitcast i32 %159 to float
>  %161 = bitcast float %160 to i32
>  %162 = bitcast float %155 to i32
>  %163 = or i32 %161, %162
>  %164 = bitcast i32 %163 to float
>  %165 = bitcast float %164 to i32
>  %166 = and i32 %165, 1065353216
>  %167 = bitcast i32 %166 to float
>  %168 = fsub float -0.000000e+00, %167
>  %169 = fsub float -0.000000e+00, %167
>  %170 = fsub float -0.000000e+00, %167
>  %171 = fsub float -0.000000e+00, %167
>  call void @llvm.AMDGPU.kill(float %168)
>  call void @llvm.AMDGPU.kill(float %169)
>  call void @llvm.AMDGPU.kill(float %170)
>  call void @llvm.AMDGPU.kill(float %171)
>  %172 = bitcast float %46 to i32
>  %173 = bitcast float %47 to i32
>  %174 = insertelement <2 x i32> undef, i32 %172, i32 0
>  %175 = insertelement <2 x i32> %174, i32 %173, i32 1
>  %176 = bitcast <8 x i32> %39 to <32 x i8>
>  %177 = bitcast <4 x i32> %41 to <16 x i8>
>  %178 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %175, <32 x i8> %176, <16 x i8> %177, i32 2)
>  %179 = extractelement <4 x float> %178, i32 0
>  %180 = extractelement <4 x float> %178, i32 1
>  %181 = extractelement <4 x float> %178, i32 2
>  %182 = extractelement <4 x float> %178, i32 3
>  %183 = fmul float %179, %49
>  %184 = fmul float %180, %49
>  %185 = fmul float %181, %49
>  %186 = fmul float %182, %49
>  %187 = bitcast float %46 to i32
>  %188 = bitcast float %47 to i32
>  %189 = insertelement <2 x i32> undef, i32 %187, i32 0
>  %190 = insertelement <2 x i32> %189, i32 %188, i32 1
>  %191 = bitcast <8 x i32> %35 to <32 x i8>
>  %192 = bitcast <4 x i32> %37 to <16 x i8>
>  %193 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %190, <32 x i8> %191, <16 x i8> %192, i32 2)
>  %194 = extractelement <4 x float> %193, i32 0
>  %195 = extractelement <4 x float> %193, i32 1
>  %196 = extractelement <4 x float> %193, i32 2
>  %197 = extractelement <4 x float> %193, i32 3
>  %198 = fmul float %48, %194
>  %199 = fadd float %198, %183
>  %200 = fmul float %48, %195
>  %201 = fadd float %200, %184
>  %202 = fmul float %48, %196
>  %203 = fadd float %202, %185
>  %204 = fmul float %48, %197
>  %205 = fadd float %204, %186
>  %206 = fmul float %199, %42
>  %207 = fmul float %201, %43
>  %208 = fmul float %203, %44
>  %209 = fmul float %205, %45
>  %210 = call i32 @llvm.SI.packf16(float %206, float %207)
>  %211 = bitcast i32 %210 to float
>  %212 = call i32 @llvm.SI.packf16(float %208, float %209)
>  %213 = bitcast i32 %212 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %211, float %213, float %211, float %213)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>declare void @llvm.AMDGPU.kill(float)
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                     ; BEFE0A7E
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                                       ; C0800300
>	s_waitcnt lgkmcnt(0)                                                     ; BF8C007F
>	s_buffer_load_dword s8, s[0:3], 0x6                                      ; C2040106
>	s_buffer_load_dword s10, s[0:3], 0x4                                     ; C2050104
>	s_waitcnt lgkmcnt(0)                                                     ; BF8C007F
>	v_mov_b32_e32 v4, s8                                                     ; 7E080208
>	v_mad_f32 v2, s10, v2, v4                                                ; D2820002 0412040A
>	s_buffer_load_dword s8, s[0:3], 0x0                                      ; C2040100
>	s_waitcnt lgkmcnt(0)                                                     ; BF8C007F
>	v_subrev_f32_e32 v4, s8, v2                                              ; 0A080408
>	v_cmp_lt_f32_e64 s[10:11], v4, 0                                         ; D002000A 00010104
>	v_cndmask_b32_e64 v4, 0, -1, s[10:11]                                    ; D2000004 00298280
>	v_and_b32_e32 v4, 1.0, v4                                                ; 360808F2
>	v_mov_b32_e32 v5, 0x80000000                                             ; 7E0A02FF 80000000
>	v_xor_b32_e32 v4, v4, v5                                                 ; 3A080B04
>	v_cmpx_le_f32_e32 vcc, 0, v4                                             ; 7C260880
>	v_cmpx_le_f32_e32 vcc, 0, v4                                             ; 7C260880
>	v_cmpx_le_f32_e32 vcc, 0, v4                                             ; 7C260880
>	v_cmpx_le_f32_e32 vcc, 0, v4                                             ; 7C260880
>	s_buffer_load_dword s8, s[0:3], 0x1                                      ; C2040101
>	s_waitcnt lgkmcnt(0)                                                     ; BF8C007F
>	v_sub_f32_e32 v2, s8, v2                                                 ; 08040408
>	v_cmp_lt_f32_e64 s[10:11], v2, 0                                         ; D002000A 00010102
>	v_cndmask_b32_e64 v2, 0, -1, s[10:11]                                    ; D2000002 00298280
>	v_and_b32_e32 v2, 1.0, v2                                                ; 360404F2
>	v_xor_b32_e32 v2, v2, v5                                                 ; 3A040B02
>	v_cmpx_le_f32_e32 vcc, 0, v2                                             ; 7C260480
>	v_cmpx_le_f32_e32 vcc, 0, v2                                             ; 7C260480
>	v_cmpx_le_f32_e32 vcc, 0, v2                                             ; 7C260480
>	v_cmpx_le_f32_e32 vcc, 0, v2                                             ; 7C260480
>	s_buffer_load_dword s8, s[0:3], 0x11                                     ; C2040111
>	s_buffer_load_dword s10, s[0:3], 0x10                                    ; C2050110
>	s_waitcnt lgkmcnt(0)                                                     ; BF8C007F
>	v_mov_b32_e32 v2, s8                                                     ; 7E040208
>	v_mad_f32 v2, s10, v3, v2                                                ; D2820002 040A060A
>	s_buffer_load_dword s8, s[0:3], 0x7                                      ; C2040107
>	s_buffer_load_dword s10, s[0:3], 0x5                                     ; C2050105
>	s_waitcnt lgkmcnt(0)                                                     ; BF8C007F
>	v_mov_b32_e32 v3, s8                                                     ; 7E060208
>	v_mad_f32 v2, s10, v2, v3                                                ; D2820002 040E040A
>	s_buffer_load_dword s8, s[0:3], 0x2                                      ; C2040102
>	s_waitcnt lgkmcnt(0)                                                     ; BF8C007F
>	v_subrev_f32_e32 v3, s8, v2                                              ; 0A060408
>	v_cmp_lt_f32_e64 s[10:11], v3, 0                                         ; D002000A 00010103
>	v_cndmask_b32_e64 v3, 0, -1, s[10:11]                                    ; D2000803 00298280
>	v_and_b32_e32 v3, 1.0, v3                                                ; 360606F2
>	v_xor_b32_e32 v3, v3, v5                                                 ; 3A060B03
>	v_cmpx_le_f32_e32 vcc, 0, v3                                             ; 7C260680
>	v_cmpx_le_f32_e32 vcc, 0, v3                                             ; 7C260680
>	v_cmpx_le_f32_e32 vcc, 0, v3                                             ; 7C260680
>	v_cmpx_le_f32_e32 vcc, 0, v3                                             ; 7C260680
>	s_buffer_load_dword s0, s[0:3], 0x3                                      ; C2000103
>	s_waitcnt lgkmcnt(0)                                                     ; BF8C007F
>	v_sub_f32_e32 v2, s0, v2                                                 ; 08040400
>	v_cmp_lt_f32_e64 s[0:1], v2, 0                                           ; D0020000 00010102
>	v_cndmask_b32_e64 v2, 0, -1, s[0:1]                                      ; D2000002 00018280
>	v_and_b32_e32 v2, 1.0, v2                                                ; 360404F2
>	v_xor_b32_e32 v2, v2, v5                                                 ; 3A040B02
>	v_cmpx_le_f32_e32 vcc, 0, v2                                             ; 7C260480
>	v_cmpx_le_f32_e32 vcc, 0, v2                                             ; 7C260480
>	v_cmpx_le_f32_e32 vcc, 0, v2                                             ; 7C260480
>	v_cmpx_le_f32_e32 vcc, 0, v2                                             ; 7C260480
>	s_mov_b32 m0, s9                                                         ; BEFC0309
>	v_interp_p1_f32 v3, v0, 1, 1, [m0]                                       ; C80C0500
>	v_interp_p2_f32 v3, [v3], v1, 1, 1, [m0]                                 ; C80D0501
>	v_interp_p1_f32 v2, v0, 0, 1, [m0]                                       ; C8080400
>	v_interp_p2_f32 v2, [v2], v1, 0, 1, [m0]                                 ; C8090401
>	s_load_dwordx4 s[0:3], s[4:5], 0x0                                       ; C0800500
>	s_load_dwordx4 s[8:11], s[4:5], 0x4                                      ; C0840504
>	s_load_dwordx8 s[12:19], s[6:7], 0x0                                     ; C0C60700
>	s_load_dwordx8 s[20:27], s[6:7], 0x8                                     ; C0CA0708
>	s_waitcnt lgkmcnt(0)                                                     ; BF8C007F
>	image_sample v[4:7], 15, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[20:27], s[8:11]  ; F0800F00 00450402
>	v_interp_p1_f32 v8, v0, 1, 2, [m0]                                       ; C8200900
>	v_interp_p2_f32 v8, [v8], v1, 1, 2, [m0]                                 ; C8210901
>	s_waitcnt vmcnt(0)                                                       ; BF8C0770
>	v_mul_f32_e32 v9, v8, v7                                                 ; 10120F08
>	image_sample v[10:13], 15, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[12:19], s[0:3] ; F0800F00 00030A02
>	v_interp_p1_f32 v2, v0, 0, 2, [m0]                                       ; C8080800
>	v_interp_p2_f32 v2, [v2], v1, 0, 2, [m0]                                 ; C8090801
>	s_waitcnt vmcnt(0)                                                       ; BF8C0770
>	v_mad_f32 v3, v2, v13, v9                                                ; D2820003 04261B02
>	v_interp_p1_f32 v9, v0, 3, 0, [m0]                                       ; C8240300
>	v_interp_p2_f32 v9, [v9], v1, 3, 0, [m0]                                 ; C8250301
>	v_mul_f32_e32 v3, v9, v3                                                 ; 10060709
>	v_mul_f32_e32 v9, v8, v6                                                 ; 10120D08
>	v_mad_f32 v9, v2, v12, v9                                                ; D2820009 04261902
>	v_interp_p1_f32 v14, v0, 2, 0, [m0]                                      ; C8380200
>	v_interp_p2_f32 v14, [v14], v1, 2, 0, [m0]                               ; C8390201
>	v_mul_f32_e32 v9, v14, v9                                                ; 1012130E
>	v_cvt_pkrtz_f16_f32_e32 v3, v9, v3                                       ; 5E060709
>	v_mul_f32_e32 v9, v8, v5                                                 ; 10120B08
>	v_mad_f32 v9, v2, v11, v9                                                ; D2820009 04261702
>	v_interp_p1_f32 v14, v0, 1, 0, [m0]                                      ; C8380100
>	v_interp_p2_f32 v14, [v14], v1, 1, 0, [m0]                               ; C8390101
>	v_mul_f32_e32 v9, v14, v9                                                ; 1012130E
>	v_mul_f32_e32 v4, v8, v4                                                 ; 10080908
>	v_mad_f32 v2, v2, v10, v4                                                ; D2820002 04121502
>	v_interp_p1_f32 v4, v0, 0, 0, [m0]                                       ; C8100000
>	v_interp_p2_f32 v4, [v4], v1, 0, 0, [m0]                                 ; C8110001
>	v_mul_f32_e32 v0, v4, v2                                                 ; 10000504
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v9                                       ; 5E001300
>	exp 15, 0, 1, 1, 1, v0, v3, v0, v3                                       ; F8001C0F 03000300
>	s_endpgm                                                                 ; BF810000
>
>SGPRS: 32
>VGPRS: 16
>Code Size: 496 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL IN[2]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL OUT[2], GENERIC[10]
>DCL CONST[0..6]
>DCL TEMP[0..2], LOCAL
>IMM[0] FLT32 {    0.0000,     1.0000,     0.0000,     0.0000}
>  0: ADD TEMP[0].xy, CONST[4].zwzw, IN[0]
>  1: MUL TEMP[1], TEMP[0].yyyy, CONST[1]
>  2: MAD TEMP[0], TEMP[0].xxxx, CONST[0], TEMP[1]
>  3: ADD TEMP[0], TEMP[0], CONST[3]
>  4: ADD TEMP[1].x, CONST[6].xxxx, CONST[6].yyyy
>  5: MUL TEMP[0], TEMP[0], TEMP[1].xxxx
>  6: MUL TEMP[1].w, CONST[5].wwww, IN[2].wwww
>  7: MOV TEMP[1].w, TEMP[1].wwww
>  8: MOV TEMP[2].xy, IN[1].xyxx
>  9: MOV TEMP[1].xyz, IN[2].xyzx
> 10: MOV TEMP[2].zw, IMM[0].yyxy
> 11: MOV OUT[1], TEMP[1]
> 12: MOV OUT[2], TEMP[2]
> 13: MOV OUT[0], TEMP[0]
> 14: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 0)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 4)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %12, i32 8)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %12, i32 12)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %12, i32 16)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %12, i32 20)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %12, i32 24)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %12, i32 28)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %12, i32 48)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %12, i32 52)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %12, i32 56)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %12, i32 60)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %12, i32 72)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %12, i32 76)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %12, i32 92)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %12, i32 96)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %12, i32 100)
>  %30 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %31 = load <16 x i8> addrspace(2)* %30, !tbaa !0
>  %32 = add i32 %5, %7
>  %33 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %31, i32 0, i32 %32)
>  %34 = extractelement <4 x float> %33, i32 0
>  %35 = extractelement <4 x float> %33, i32 1
>  %36 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %37 = load <16 x i8> addrspace(2)* %36, !tbaa !0
>  %38 = add i32 %5, %7
>  %39 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %37, i32 0, i32 %38)
>  %40 = extractelement <4 x float> %39, i32 0
>  %41 = extractelement <4 x float> %39, i32 1
>  %42 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 2
>  %43 = load <16 x i8> addrspace(2)* %42, !tbaa !0
>  %44 = add i32 %5, %7
>  %45 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %43, i32 0, i32 %44)
>  %46 = extractelement <4 x float> %45, i32 0
>  %47 = extractelement <4 x float> %45, i32 1
>  %48 = extractelement <4 x float> %45, i32 2
>  %49 = extractelement <4 x float> %45, i32 3
>  %50 = fadd float %25, %34
>  %51 = fadd float %26, %35
>  %52 = fmul float %51, %17
>  %53 = fmul float %51, %18
>  %54 = fmul float %51, %19
>  %55 = fmul float %51, %20
>  %56 = fmul float %50, %13
>  %57 = fadd float %56, %52
>  %58 = fmul float %50, %14
>  %59 = fadd float %58, %53
>  %60 = fmul float %50, %15
>  %61 = fadd float %60, %54
>  %62 = fmul float %50, %16
>  %63 = fadd float %62, %55
>  %64 = fadd float %57, %21
>  %65 = fadd float %59, %22
>  %66 = fadd float %61, %23
>  %67 = fadd float %63, %24
>  %68 = fadd float %28, %29
>  %69 = fmul float %64, %68
>  %70 = fmul float %65, %68
>  %71 = fmul float %66, %68
>  %72 = fmul float %67, %68
>  %73 = fmul float %27, %49
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %46, float %47, float %48, float %73)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %40, float %41, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %69, float %70, float %71, float %72)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                             ; 4A00000A
>	s_load_dwordx4 s[4:7], s[8:9], 0x0                    ; C0820900
>	s_load_dwordx4 s[12:15], s[8:9], 0x4                  ; C0860904
>	s_load_dwordx4 s[8:11], s[8:9], 0x8                   ; C0840908
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[8:11], 0 idxen  ; E00C2000 80020100
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                    ; C0800300
>	s_waitcnt vmcnt(0) lgkmcnt(0)                         ; BF8C0070
>	s_buffer_load_dword s8, s[0:3], 0x17                  ; C2040117
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v5, s8, v4                              ; 100A0808
>	exp 15, 32, 0, 0, 0, v1, v2, v3, v5                   ; F800020F 05030201
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	buffer_load_format_xyzw v[1:4], v0, s[12:15], 0 idxen ; E00C2000 80030100
>	v_mov_b32_e32 v5, 1.0                                 ; 7E0A02F2
>	v_mov_b32_e32 v6, 0                                   ; 7E0C0280
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	exp 15, 33, 0, 0, 0, v1, v2, v6, v5                   ; F800021F 05060201
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	buffer_load_format_xyzw v[0:3], v0, s[4:7], 0 idxen   ; E00C2000 80010000
>	s_buffer_load_dword s4, s[0:3], 0x12                  ; C2020112
>	s_waitcnt vmcnt(0) lgkmcnt(0)                         ; BF8C0070
>	v_add_f32_e32 v4, s4, v0                              ; 06080004
>	s_buffer_load_dword s4, s[0:3], 0x13                  ; C2020113
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v0, s4, v1                              ; 06000204
>	s_buffer_load_dword s4, s[0:3], 0x7                   ; C2020107
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v1, s4, v0                              ; 10020004
>	s_buffer_load_dword s4, s[0:3], 0x3                   ; C2020103
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v1, v4, s4, v1                              ; D2820001 04040904
>	s_buffer_load_dword s4, s[0:3], 0xf                   ; C202010F
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v1, s4, v1                              ; 06020204
>	s_buffer_load_dword s4, s[0:3], 0x19                  ; C2020119
>	s_buffer_load_dword s5, s[0:3], 0x18                  ; C2028118
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mov_b32_e32 v2, s4                                  ; 7E040204
>	v_add_f32_e32 v2, s5, v2                              ; 06040405
>	v_mul_f32_e32 v1, v2, v1                              ; 10020302
>	s_buffer_load_dword s4, s[0:3], 0x6                   ; C2020106
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v3, s4, v0                              ; 10060004
>	s_buffer_load_dword s4, s[0:3], 0x2                   ; C2020102
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v3, v4, s4, v3                              ; D2820003 040C0904
>	s_buffer_load_dword s4, s[0:3], 0xe                   ; C202010E
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v3, s4, v3                              ; 06060604
>	v_mul_f32_e32 v3, v2, v3                              ; 10060702
>	s_buffer_load_dword s4, s[0:3], 0x5                   ; C2020105
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v5, s4, v0                              ; 100A0004
>	s_buffer_load_dword s4, s[0:3], 0x1                   ; C2020101
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v5, v4, s4, v5                              ; D2820005 04140904
>	s_buffer_load_dword s4, s[0:3], 0xd                   ; C202010D
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v5, s4, v5                              ; 060A0A04
>	v_mul_f32_e32 v5, v2, v5                              ; 100A0B02
>	s_buffer_load_dword s4, s[0:3], 0x4                   ; C2020104
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v0, s4, v0                              ; 10000004
>	s_buffer_load_dword s4, s[0:3], 0x0                   ; C2020100
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v0, v4, s4, v0                              ; D2820000 04000904
>	s_buffer_load_dword s0, s[0:3], 0xc                   ; C200010C
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v0, s0, v0                              ; 06000000
>	v_mul_f32_e32 v0, v2, v0                              ; 10000102
>	exp 15, 12, 0, 1, 0, v0, v5, v3, v1                   ; F80008CF 01030500
>	s_endpgm                                              ; BF810000
>
>SGPRS: 24
>VGPRS: 8
>Code Size: 332 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], POSITION, LINEAR
>DCL IN[1], GENERIC[9], PERSPECTIVE
>DCL IN[2], GENERIC[10], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SAMP[1]
>DCL CONST[9]
>DCL CONST[0..6]
>DCL TEMP[0]
>DCL TEMP[1..9], LOCAL
>IMM[0] FLT32 {    0.0000,     1.0000,     0.5000,  1024.0000}
>IMM[1] FLT32 {65280.0000,   255.0000,     0.0000,     0.0000}
>  0: MOV TEMP[0], IN[0]
>  1: MAD TEMP[0].y, IN[0], CONST[9].xxxx, CONST[9].yyyy
>  2: MAD TEMP[1].xy, TEMP[0], CONST[6].xyxx, CONST[6].zwzz
>  3: MOV TEMP[2].xz, -CONST[0]
>  4: ADD TEMP[3].xyz, TEMP[2].xxxx, TEMP[1].xxxx
>  5: FSLT TEMP[4].xyz, TEMP[3].xyzz, IMM[0].xxxx
>  6: OR TEMP[5].x, TEMP[4].xxxx, TEMP[4].zzzz
>  7: OR TEMP[4].x, TEMP[5].xxxx, TEMP[4].yyyy
>  8: AND TEMP[4].x, TEMP[4].xxxx, IMM[0].yyyy
>  9: KILL_IF -TEMP[4].xxxx
> 10: ADD TEMP[3].xyz, CONST[0].yyyy, -TEMP[1].xxxx
> 11: FSLT TEMP[4].xyz, TEMP[3].xyzz, IMM[0].xxxx
> 12: OR TEMP[5].x, TEMP[4].xxxx, TEMP[4].zzzz
> 13: OR TEMP[4].x, TEMP[5].xxxx, TEMP[4].yyyy
> 14: AND TEMP[4].x, TEMP[4].xxxx, IMM[0].yyyy
> 15: KILL_IF -TEMP[4].xxxx
> 16: ADD TEMP[3].xyz, TEMP[2].zzzz, TEMP[1].yyyy
> 17: FSLT TEMP[2].xyz, TEMP[3].xyzz, IMM[0].xxxx
> 18: OR TEMP[4].x, TEMP[2].xxxx, TEMP[2].zzzz
> 19: OR TEMP[2].x, TEMP[4].xxxx, TEMP[2].yyyy
> 20: AND TEMP[2].x, TEMP[2].xxxx, IMM[0].yyyy
> 21: KILL_IF -TEMP[2].xxxx
> 22: ADD TEMP[3], CONST[0].wwww, -TEMP[1].yyyy
> 23: FSLT TEMP[1].xyz, TEMP[3].xyzz, IMM[0].xxxx
> 24: OR TEMP[2].x, TEMP[1].xxxx, TEMP[1].zzzz
> 25: OR TEMP[1].x, TEMP[2].xxxx, TEMP[1].yyyy
> 26: AND TEMP[1].x, TEMP[1].xxxx, IMM[0].yyyy
> 27: KILL_IF -TEMP[1].xxxx
> 28: MOV TEMP[3].y, IMM[0].zzzz
> 29: ADD TEMP[1].x, IMM[0].wwww, -CONST[3].xxxx
> 30: MOV TEMP[3].x, TEMP[1].xxxx
> 31: FSNE TEMP[2].x, CONST[1].xxxx, -CONST[1].xxxx
> 32: UIF TEMP[2].xxxx :2
> 33:   FSLT TEMP[2].x, IMM[0].zzzz, CONST[2].xxxx
> 34:   UIF TEMP[2].xxxx :2
> 35:     MUL TEMP[2].zw, CONST[4].xyxy, IN[2].xyxy
> 36:     MOV TEMP[3].zw, TEMP[2].wwzw
> 37:     FRC TEMP[2].xy, TEMP[2].zwzw
> 38:     ADD TEMP[4].zw, TEMP[3], -TEMP[2].xyxy
> 39:     MOV TEMP[3].zw, TEMP[4].wwzw
> 40:     MAD TEMP[4].xy, IN[2], CONST[4], -TEMP[4].zwzw
> 41:     ADD TEMP[5].zw, TEMP[3], IMM[0].zzzz
> 42:     RCP TEMP[6].x, CONST[4].xxxx
> 43:     RCP TEMP[7].x, CONST[4].yyyy
> 44:     MOV TEMP[6].y, TEMP[7].xxxx
> 45:     MUL TEMP[5].xy, TEMP[5].zwzw, TEMP[6]
> 46:     MOV TEMP[5].xy, TEMP[5].xyyy
> 47:     MOV TEMP[5].w, IMM[0].xxxx
> 48:     TXL TEMP[5], TEMP[5], SAMP[0], 2D
> 49:     MOV TEMP[6].zw, TEMP[5]
> 50:     DP2 TEMP[7].x, TEMP[5].wxxx, IMM[1].xyyy
> 51:     ADD TEMP[7].x, TEMP[7].xxxx, IMM[0].zzzz
> 52:     FRC TEMP[8].w, TEMP[7].xxxx
> 53:     ADD TEMP[7].z, -TEMP[8].wwww, TEMP[7].xxxx
> 54:     MOV TEMP[3].z, TEMP[7].zzzz
> 55:     MAD TEMP[5].w, CONST[3].xxxx, TEMP[5].xxxx, IMM[0].zzzz
> 56:     FRC TEMP[7].z, TEMP[5].wwww
> 57:     ADD TEMP[5].w, TEMP[5].wwww, -TEMP[7].zzzz
> 58:     MOV TEMP[3].w, TEMP[5].wwww
> 59:     ADD TEMP[5].zw, TEMP[3], IMM[0].zzzz
> 60:     FSGE TEMP[7].x, TEMP[1].xxxx, IMM[0].xxxx
> 61:     UIF TEMP[7].xxxx :2
> 62:       MOV TEMP[7].x, TEMP[5].wwww
> 63:     ELSE :2
> 64:       MOV TEMP[7].x, TEMP[5].zzzz
> 65:     ENDIF
> 66:     RCP TEMP[5].x, CONST[2].xxxx
> 67:     MAD TEMP[5].w, CONST[5].xxxx, TEMP[5].xxxx, IMM[0].zzzz
> 68:     FRC TEMP[8].z, TEMP[5].wwww
> 69:     ADD TEMP[5].w, TEMP[5].wwww, -TEMP[8].zzzz
> 70:     MUL TEMP[8].z, TEMP[5].wwww, TEMP[7].xxxx
> 71:     MOV TEMP[9].x, -TEMP[5].wwww
> 72:     FSGE TEMP[8].x, TEMP[8].zzzz, IMM[0].xxxx
> 73:     UIF TEMP[8].xxxx :2
> 74:       MOV TEMP[8].x, TEMP[5].wwww
> 75:     ELSE :2
> 76:       MOV TEMP[8].x, TEMP[9].xxxx
> 77:     ENDIF
> 78:     MOV TEMP[2].z, TEMP[8].xxxx
> 79:     RCP TEMP[9].x, TEMP[8].xxxx
> 80:     MUL TEMP[9].w, TEMP[7].xxxx, TEMP[9].xxxx
> 81:     FRC TEMP[9].w, TEMP[9].wwww
> 82:     MOV TEMP[2].w, TEMP[9].wwww
> 83:     MUL TEMP[8].x, TEMP[9].wwww, TEMP[8].xxxx
> 84:     MOV TEMP[6].x, TEMP[8].xxxx
> 85:     RCP TEMP[5].x, TEMP[5].wwww
> 86:     MUL TEMP[5].y, TEMP[5].xxxx, TEMP[7].xxxx
> 87:     MOV TEMP[6].y, TEMP[5].yyyy
> 88:     FRC TEMP[5].zw, TEMP[6].xyxy
> 89:     MOV TEMP[3].zw, TEMP[5].wwzw
> 90:     ADD TEMP[5].zw, -TEMP[3], TEMP[6].xyxy
> 91:     MOV TEMP[3].zw, TEMP[5].wwzw
> 92:     ADD TEMP[4].zw, TEMP[4].xyxy, TEMP[3]
> 93:     MOV TEMP[3].zw, TEMP[4].wwzw
> 94:     MUL TEMP[4].zw, TEMP[3], CONST[2].xxxx
> 95:     MOV TEMP[3].zw, TEMP[4].wwzw
> 96:     FRC TEMP[4].xy, TEMP[4].zwzw
> 97:     ADD TEMP[4].zw, TEMP[3], -TEMP[4].xyxy
> 98:     MOV TEMP[3].zw, TEMP[4].wwzw
> 99:     ADD TEMP[4].zw, TEMP[3], IMM[0].zzzz
>100:     MOV TEMP[3].w, TEMP[4].wwzw
>101:     RCP TEMP[2].x, CONST[5].xxxx
>102:     RCP TEMP[5].x, CONST[5].yyyy
>103:     MOV TEMP[2].y, TEMP[5].xxxx
>104:     MUL TEMP[4].xy, TEMP[4].zwzw, TEMP[2]
>105:     MOV TEMP[4].xy, TEMP[4].xyyy
>106:     MOV TEMP[4].w, IMM[0].xxxx
>107:     TXL TEMP[4], TEMP[4], SAMP[1], 2D
>108:     MOV TEMP[2], TEMP[4]
>109:     MOV TEMP[3].z, TEMP[4].wwww
>110:   ELSE :2
>111:     MOV TEMP[4].xy, IN[2].xyyy
>112:     TEX TEMP[4], TEMP[4], SAMP[1], 2D
>113:     MOV TEMP[2], TEMP[4]
>114:     MOV TEMP[3].z, TEMP[4].wwww
>115:   ENDIF
>116:   MUL TEMP[4].w, TEMP[3].zzzz, IN[1].wwww
>117:   MOV TEMP[4].w, TEMP[4].wwww
>118:   MOV TEMP[4].xyz, IN[1].xyzx
>119: ELSE :2
>120:   FSLT TEMP[5].x, IMM[0].zzzz, CONST[2].xxxx
>121:   UIF TEMP[5].xxxx :2
>122:     MUL TEMP[5].zw, CONST[4].xyxy, IN[2].xyxy
>123:     MOV TEMP[3].zw, TEMP[5].wwzw
>124:     FRC TEMP[5].xy, TEMP[5].zwzw
>125:     MOV TEMP[2].xy, TEMP[5].xyxx
>126:     ADD TEMP[5].zw, TEMP[3], -TEMP[5].xyxy
>127:     MOV TEMP[3].zw, TEMP[5].wwzw
>128:     MAD TEMP[5].xy, IN[2], CONST[4], -TEMP[5].zwzw
>129:     MOV TEMP[2].xy, TEMP[5].xyxx
>130:     ADD TEMP[5].zw, TEMP[3], IMM[0].zzzz
>131:     RCP TEMP[6].x, CONST[4].xxxx
>132:     RCP TEMP[7].x, CONST[4].yyyy
>133:     MOV TEMP[6].y, TEMP[7].xxxx
>134:     MUL TEMP[5].xy, TEMP[5].zwzw, TEMP[6]
>135:     MOV TEMP[5].xy, TEMP[5].xyyy
>136:     MOV TEMP[5].w, IMM[0].xxxx
>137:     TXL TEMP[5], TEMP[5], SAMP[0], 2D
>138:     MOV TEMP[6].zw, TEMP[5]
>139:     DP2 TEMP[7].x, TEMP[5].wxxx, IMM[1].xyyy
>140:     ADD TEMP[7].x, TEMP[7].xxxx, IMM[0].zzzz
>141:     FRC TEMP[8].w, TEMP[7].xxxx
>142:     ADD TEMP[7].z, -TEMP[8].wwww, TEMP[7].xxxx
>143:     MOV TEMP[3].z, TEMP[7].zzzz
>144:     MAD TEMP[5].w, CONST[3].xxxx, TEMP[5].xxxx, IMM[0].zzzz
>145:     FRC TEMP[7].z, TEMP[5].wwww
>146:     MOV TEMP[2].z, TEMP[7].zzzz
>147:     ADD TEMP[5].w, TEMP[5].wwww, -TEMP[7].zzzz
>148:     MOV TEMP[3].w, TEMP[5].wwww
>149:     ADD TEMP[5].zw, TEMP[3], IMM[0].zzzz
>150:     FSGE TEMP[1].x, TEMP[1].xxxx, IMM[0].xxxx
>151:     UIF TEMP[1].xxxx :2
>152:       MOV TEMP[1].x, TEMP[5].wwww
>153:     ELSE :2
>154:       MOV TEMP[1].x, TEMP[5].zzzz
>155:     ENDIF
>156:     MOV TEMP[3].x, TEMP[1].xxxx
>157:     RCP TEMP[5].x, CONST[2].xxxx
>158:     MAD TEMP[5].y, CONST[5].xxxx, TEMP[5].xxxx, IMM[0].zzzz
>159:     FRC TEMP[7].z, TEMP[5].yyyy
>160:     ADD TEMP[5].y, -TEMP[7].zzzz, TEMP[5].yyyy
>161:     MUL TEMP[7].z, TEMP[5].yyyy, TEMP[1].xxxx
>162:     MOV TEMP[8].x, -TEMP[5].yyyy
>163:     FSGE TEMP[7].x, TEMP[7].zzzz, IMM[0].xxxx
>164:     UIF TEMP[7].xxxx :2
>165:       MOV TEMP[7].x, TEMP[5].yyyy
>166:     ELSE :2
>167:       MOV TEMP[7].x, TEMP[8].xxxx
>168:     ENDIF
>169:     MOV TEMP[3].z, TEMP[7].xxxx
>170:     RCP TEMP[7].x, TEMP[7].xxxx
>171:     MUL TEMP[1].w, TEMP[7].xxxx, TEMP[1].xxxx
>172:     FRC TEMP[1].w, TEMP[1].wwww
>173:     MOV TEMP[3].w, TEMP[1].wwww
>174:     RCP TEMP[1].x, TEMP[5].yyyy
>175:     MOV TEMP[3].y, TEMP[1].xxxx
>176:     MUL TEMP[1].xy, TEMP[3].wyzw, TEMP[3].zxzw
>177:     MOV TEMP[6].xy, TEMP[1].xyxx
>178:     FRC TEMP[1].xy, TEMP[6]
>179:     MOV TEMP[3].xy, TEMP[1].xyxx
>180:     ADD TEMP[1].xy, -TEMP[3], TEMP[6]
>181:     MOV TEMP[3].xy, TEMP[1].xyxx
>182:     ADD TEMP[1].xy, TEMP[2], TEMP[3]
>183:     MOV TEMP[3].xy, TEMP[1].xyxx
>184:     MUL TEMP[1].xy, TEMP[3], CONST[2].xxxx
>185:     MOV TEMP[3].xy, TEMP[1].xyxx
>186:     FRC TEMP[1].zw, TEMP[1].xyxy
>187:     MOV TEMP[3].zw, TEMP[1].wwzw
>188:     ADD TEMP[1].xy, -TEMP[1].zwzw, TEMP[3]
>189:     MOV TEMP[3].xy, TEMP[1].xyxx
>190:     ADD TEMP[1].xy, TEMP[3], IMM[0].zzzz
>191:     MOV TEMP[3].xy, TEMP[1].xyxx
>192:     RCP TEMP[2].x, CONST[5].xxxx
>193:     RCP TEMP[1].x, CONST[5].yyyy
>194:     MOV TEMP[2].y, TEMP[1].xxxx
>195:     MUL TEMP[1].xy, TEMP[3], TEMP[2]
>196:     MOV TEMP[1].xy, TEMP[1].xyyy
>197:     MOV TEMP[1].w, IMM[0].xxxx
>198:     TXL TEMP[1], TEMP[1], SAMP[1], 2D
>199:     MOV TEMP[3].xyz, TEMP[1]
>200:     MOV TEMP[2].x, TEMP[1].wwww
>201:   ELSE :2
>202:     MOV TEMP[1].xy, IN[2].xyyy
>203:     TEX TEMP[1], TEMP[1], SAMP[1], 2D
>204:     MOV TEMP[3].xyz, TEMP[1]
>205:     MOV TEMP[2].x, TEMP[1].wwww
>206:   ENDIF
>207:   MUL TEMP[1].w, TEMP[2].xxxx, IN[1].wwww
>208:   MOV TEMP[3].w, TEMP[1].wwww
>209:   MOV TEMP[4], TEMP[3]
>210: ENDIF
>211: MOV OUT[0], TEMP[4]
>212: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %23 = load <16 x i8> addrspace(2)* %22, !tbaa !0
>  %24 = call float @llvm.SI.load.const(<16 x i8> %23, i32 0)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %23, i32 4)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %23, i32 8)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %23, i32 12)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %23, i32 16)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %23, i32 32)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %23, i32 48)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %23, i32 64)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %23, i32 68)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %23, i32 80)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %23, i32 84)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %23, i32 96)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %23, i32 100)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %23, i32 104)
>  %38 = call float @llvm.SI.load.const(<16 x i8> %23, i32 108)
>  %39 = call float @llvm.SI.load.const(<16 x i8> %23, i32 144)
>  %40 = call float @llvm.SI.load.const(<16 x i8> %23, i32 148)
>  %41 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %42 = load <8 x i32> addrspace(2)* %41, !tbaa !0
>  %43 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %44 = load <4 x i32> addrspace(2)* %43, !tbaa !0
>  %45 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 1
>  %46 = load <8 x i32> addrspace(2)* %45, !tbaa !0
>  %47 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 1
>  %48 = load <4 x i32> addrspace(2)* %47, !tbaa !0
>  %49 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %50 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %51 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %5, <2 x i32> %7)
>  %52 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %5, <2 x i32> %7)
>  %53 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %5, <2 x i32> %7)
>  %54 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %5, <2 x i32> %7)
>  %55 = fmul float %15, %39
>  %56 = fadd float %55, %40
>  %57 = fmul float %14, %35
>  %58 = fadd float %57, %37
>  %59 = fmul float %56, %36
>  %60 = fadd float %59, %38
>  %61 = fsub float -0.000000e+00, %24
>  %62 = fsub float -0.000000e+00, %26
>  %63 = fadd float %61, %58
>  %64 = fadd float %61, %58
>  %65 = fadd float %61, %58
>  %66 = fcmp olt float %63, 0.000000e+00
>  %67 = sext i1 %66 to i32
>  %68 = fcmp olt float %64, 0.000000e+00
>  %69 = sext i1 %68 to i32
>  %70 = fcmp olt float %65, 0.000000e+00
>  %71 = sext i1 %70 to i32
>  %72 = bitcast i32 %67 to float
>  %73 = bitcast i32 %69 to float
>  %74 = bitcast i32 %71 to float
>  %75 = bitcast float %72 to i32
>  %76 = bitcast float %74 to i32
>  %77 = or i32 %75, %76
>  %78 = bitcast i32 %77 to float
>  %79 = bitcast float %78 to i32
>  %80 = bitcast float %73 to i32
>  %81 = or i32 %79, %80
>  %82 = bitcast i32 %81 to float
>  %83 = bitcast float %82 to i32
>  %84 = and i32 %83, 1065353216
>  %85 = bitcast i32 %84 to float
>  %86 = fsub float -0.000000e+00, %85
>  %87 = fsub float -0.000000e+00, %85
>  %88 = fsub float -0.000000e+00, %85
>  %89 = fsub float -0.000000e+00, %85
>  call void @llvm.AMDGPU.kill(float %86)
>  call void @llvm.AMDGPU.kill(float %87)
>  call void @llvm.AMDGPU.kill(float %88)
>  call void @llvm.AMDGPU.kill(float %89)
>  %90 = fsub float -0.000000e+00, %58
>  %91 = fadd float %25, %90
>  %92 = fsub float -0.000000e+00, %58
>  %93 = fadd float %25, %92
>  %94 = fsub float -0.000000e+00, %58
>  %95 = fadd float %25, %94
>  %96 = fcmp olt float %91, 0.000000e+00
>  %97 = sext i1 %96 to i32
>  %98 = fcmp olt float %93, 0.000000e+00
>  %99 = sext i1 %98 to i32
>  %100 = fcmp olt float %95, 0.000000e+00
>  %101 = sext i1 %100 to i32
>  %102 = bitcast i32 %97 to float
>  %103 = bitcast i32 %99 to float
>  %104 = bitcast i32 %101 to float
>  %105 = bitcast float %102 to i32
>  %106 = bitcast float %104 to i32
>  %107 = or i32 %105, %106
>  %108 = bitcast i32 %107 to float
>  %109 = bitcast float %108 to i32
>  %110 = bitcast float %103 to i32
>  %111 = or i32 %109, %110
>  %112 = bitcast i32 %111 to float
>  %113 = bitcast float %112 to i32
>  %114 = and i32 %113, 1065353216
>  %115 = bitcast i32 %114 to float
>  %116 = fsub float -0.000000e+00, %115
>  %117 = fsub float -0.000000e+00, %115
>  %118 = fsub float -0.000000e+00, %115
>  %119 = fsub float -0.000000e+00, %115
>  call void @llvm.AMDGPU.kill(float %116)
>  call void @llvm.AMDGPU.kill(float %117)
>  call void @llvm.AMDGPU.kill(float %118)
>  call void @llvm.AMDGPU.kill(float %119)
>  %120 = fadd float %62, %60
>  %121 = fadd float %62, %60
>  %122 = fadd float %62, %60
>  %123 = fcmp olt float %120, 0.000000e+00
>  %124 = sext i1 %123 to i32
>  %125 = fcmp olt float %121, 0.000000e+00
>  %126 = sext i1 %125 to i32
>  %127 = fcmp olt float %122, 0.000000e+00
>  %128 = sext i1 %127 to i32
>  %129 = bitcast i32 %124 to float
>  %130 = bitcast i32 %126 to float
>  %131 = bitcast i32 %128 to float
>  %132 = bitcast float %129 to i32
>  %133 = bitcast float %131 to i32
>  %134 = or i32 %132, %133
>  %135 = bitcast i32 %134 to float
>  %136 = bitcast float %135 to i32
>  %137 = bitcast float %130 to i32
>  %138 = or i32 %136, %137
>  %139 = bitcast i32 %138 to float
>  %140 = bitcast float %139 to i32
>  %141 = and i32 %140, 1065353216
>  %142 = bitcast i32 %141 to float
>  %143 = fsub float -0.000000e+00, %142
>  %144 = fsub float -0.000000e+00, %142
>  %145 = fsub float -0.000000e+00, %142
>  %146 = fsub float -0.000000e+00, %142
>  call void @llvm.AMDGPU.kill(float %143)
>  call void @llvm.AMDGPU.kill(float %144)
>  call void @llvm.AMDGPU.kill(float %145)
>  call void @llvm.AMDGPU.kill(float %146)
>  %147 = fsub float -0.000000e+00, %60
>  %148 = fadd float %27, %147
>  %149 = fsub float -0.000000e+00, %60
>  %150 = fadd float %27, %149
>  %151 = fsub float -0.000000e+00, %60
>  %152 = fadd float %27, %151
>  %153 = fcmp olt float %148, 0.000000e+00
>  %154 = sext i1 %153 to i32
>  %155 = fcmp olt float %150, 0.000000e+00
>  %156 = sext i1 %155 to i32
>  %157 = fcmp olt float %152, 0.000000e+00
>  %158 = sext i1 %157 to i32
>  %159 = bitcast i32 %154 to float
>  %160 = bitcast i32 %156 to float
>  %161 = bitcast i32 %158 to float
>  %162 = bitcast float %159 to i32
>  %163 = bitcast float %161 to i32
>  %164 = or i32 %162, %163
>  %165 = bitcast i32 %164 to float
>  %166 = bitcast float %165 to i32
>  %167 = bitcast float %160 to i32
>  %168 = or i32 %166, %167
>  %169 = bitcast i32 %168 to float
>  %170 = bitcast float %169 to i32
>  %171 = and i32 %170, 1065353216
>  %172 = bitcast i32 %171 to float
>  %173 = fsub float -0.000000e+00, %172
>  %174 = fsub float -0.000000e+00, %172
>  %175 = fsub float -0.000000e+00, %172
>  %176 = fsub float -0.000000e+00, %172
>  call void @llvm.AMDGPU.kill(float %173)
>  call void @llvm.AMDGPU.kill(float %174)
>  call void @llvm.AMDGPU.kill(float %175)
>  call void @llvm.AMDGPU.kill(float %176)
>  %177 = fsub float -0.000000e+00, %30
>  %178 = fadd float 1.024000e+03, %177
>  %179 = fsub float -0.000000e+00, %28
>  %180 = fcmp une float %28, %179
>  %181 = sext i1 %180 to i32
>  %182 = bitcast i32 %181 to float
>  %183 = bitcast float %182 to i32
>  %184 = icmp ne i32 %183, 0
>  %185 = fcmp olt float 5.000000e-01, %29
>  %186 = sext i1 %185 to i32
>  %187 = bitcast i32 %186 to float
>  %188 = bitcast float %187 to i32
>  %189 = icmp ne i32 %188, 0
>  br i1 %184, label %IF, label %ELSE
>
>IF:                                               ; preds = %main_body
>  br i1 %189, label %IF41, label %ELSE42
>
>ELSE:                                             ; preds = %main_body
>  br i1 %189, label %IF50, label %ELSE51
>
>ENDIF:                                            ; preds = %IF41, %ELSE42, %ENDIF49
>  %.sink60.sink = phi <4 x float> [ %.sink60, %ENDIF49 ], [ %304, %ELSE42 ], [ %297, %IF41 ]
>  %temp16.0 = phi float [ %414, %ENDIF49 ], [ %49, %ELSE42 ], [ %49, %IF41 ]
>  %temp17.0 = phi float [ %415, %ENDIF49 ], [ %50, %ELSE42 ], [ %50, %IF41 ]
>  %temp18.0 = phi float [ %416, %ENDIF49 ], [ %51, %ELSE42 ], [ %51, %IF41 ]
>  %190 = extractelement <4 x float> %.sink60.sink, i32 3
>  %191 = fmul float %190, %52
>  %192 = call i32 @llvm.SI.packf16(float %temp16.0, float %temp17.0)
>  %193 = bitcast i32 %192 to float
>  %194 = call i32 @llvm.SI.packf16(float %temp18.0, float %191)
>  %195 = bitcast i32 %194 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %193, float %195, float %193, float %195)
>  ret void
>
>IF41:                                             ; preds = %IF
>  %196 = fmul float %31, %53
>  %197 = fmul float %32, %54
>  %198 = call float @llvm.AMDIL.fraction.(float %196)
>  %199 = call float @llvm.AMDIL.fraction.(float %197)
>  %200 = fsub float -0.000000e+00, %198
>  %201 = fadd float %196, %200
>  %202 = fsub float -0.000000e+00, %199
>  %203 = fadd float %197, %202
>  %204 = fsub float -0.000000e+00, %201
>  %205 = fmul float %53, %31
>  %206 = fadd float %205, %204
>  %207 = fsub float -0.000000e+00, %203
>  %208 = fmul float %54, %32
>  %209 = fadd float %208, %207
>  %210 = fadd float %201, 5.000000e-01
>  %211 = fadd float %203, 5.000000e-01
>  %212 = fdiv float 1.000000e+00, %31
>  %213 = fdiv float 1.000000e+00, %32
>  %214 = fmul float %210, %212
>  %215 = fmul float %211, %213
>  %216 = bitcast float %214 to i32
>  %217 = bitcast float %215 to i32
>  %218 = bitcast float 0.000000e+00 to i32
>  %219 = insertelement <4 x i32> undef, i32 %216, i32 0
>  %220 = insertelement <4 x i32> %219, i32 %217, i32 1
>  %221 = insertelement <4 x i32> %220, i32 %218, i32 2
>  %222 = insertelement <4 x i32> %221, i32 undef, i32 3
>  %223 = bitcast <8 x i32> %42 to <32 x i8>
>  %224 = bitcast <4 x i32> %44 to <16 x i8>
>  %225 = call <4 x float> @llvm.SI.samplel.v4i32(<4 x i32> %222, <32 x i8> %223, <16 x i8> %224, i32 2)
>  %226 = extractelement <4 x float> %225, i32 0
>  %227 = extractelement <4 x float> %225, i32 3
>  %228 = fmul float %227, 6.528000e+04
>  %229 = fmul float %226, 2.550000e+02
>  %230 = fadd float %228, %229
>  %231 = fadd float %230, 5.000000e-01
>  %232 = call float @llvm.AMDIL.fraction.(float %231)
>  %233 = fsub float -0.000000e+00, %232
>  %234 = fadd float %233, %231
>  %235 = fmul float %30, %226
>  %236 = fadd float %235, 5.000000e-01
>  %237 = call float @llvm.AMDIL.fraction.(float %236)
>  %238 = fsub float -0.000000e+00, %237
>  %239 = fadd float %236, %238
>  %240 = fadd float %234, 5.000000e-01
>  %241 = fadd float %239, 5.000000e-01
>  %242 = fcmp oge float %178, 0.000000e+00
>  %243 = sext i1 %242 to i32
>  %244 = bitcast i32 %243 to float
>  %245 = bitcast float %244 to i32
>  %246 = icmp ne i32 %245, 0
>  %. = select i1 %246, float %241, float %240
>  %247 = fdiv float 1.000000e+00, %29
>  %248 = fmul float %33, %247
>  %249 = fadd float %248, 5.000000e-01
>  %250 = call float @llvm.AMDIL.fraction.(float %249)
>  %251 = fsub float -0.000000e+00, %250
>  %252 = fadd float %249, %251
>  %253 = fmul float %252, %.
>  %254 = fsub float -0.000000e+00, %252
>  %255 = fcmp oge float %253, 0.000000e+00
>  %256 = sext i1 %255 to i32
>  %257 = bitcast i32 %256 to float
>  %258 = bitcast float %257 to i32
>  %259 = icmp ne i32 %258, 0
>  %temp32.0 = select i1 %259, float %252, float %254
>  %260 = fdiv float 1.000000e+00, %temp32.0
>  %261 = fmul float %., %260
>  %262 = call float @llvm.AMDIL.fraction.(float %261)
>  %263 = fmul float %262, %temp32.0
>  %264 = fdiv float 1.000000e+00, %252
>  %265 = fmul float %264, %.
>  %266 = call float @llvm.AMDIL.fraction.(float %263)
>  %267 = call float @llvm.AMDIL.fraction.(float %265)
>  %268 = fsub float -0.000000e+00, %266
>  %269 = fadd float %268, %263
>  %270 = fsub float -0.000000e+00, %267
>  %271 = fadd float %270, %265
>  %272 = fadd float %206, %269
>  %273 = fadd float %209, %271
>  %274 = fmul float %272, %29
>  %275 = fmul float %273, %29
>  %276 = call float @llvm.AMDIL.fraction.(float %274)
>  %277 = call float @llvm.AMDIL.fraction.(float %275)
>  %278 = fsub float -0.000000e+00, %276
>  %279 = fadd float %274, %278
>  %280 = fsub float -0.000000e+00, %277
>  %281 = fadd float %275, %280
>  %282 = fadd float %279, 5.000000e-01
>  %283 = fadd float %281, 5.000000e-01
>  %284 = fdiv float 1.000000e+00, %33
>  %285 = fdiv float 1.000000e+00, %34
>  %286 = fmul float %282, %284
>  %287 = fmul float %283, %285
>  %288 = bitcast float %286 to i32
>  %289 = bitcast float %287 to i32
>  %290 = bitcast float 0.000000e+00 to i32
>  %291 = insertelement <4 x i32> undef, i32 %288, i32 0
>  %292 = insertelement <4 x i32> %291, i32 %289, i32 1
>  %293 = insertelement <4 x i32> %292, i32 %290, i32 2
>  %294 = insertelement <4 x i32> %293, i32 undef, i32 3
>  %295 = bitcast <8 x i32> %46 to <32 x i8>
>  %296 = bitcast <4 x i32> %48 to <16 x i8>
>  %297 = call <4 x float> @llvm.SI.samplel.v4i32(<4 x i32> %294, <32 x i8> %295, <16 x i8> %296, i32 2)
>  br label %ENDIF
>
>ELSE42:                                           ; preds = %IF
>  %298 = bitcast float %53 to i32
>  %299 = bitcast float %54 to i32
>  %300 = insertelement <2 x i32> undef, i32 %298, i32 0
>  %301 = insertelement <2 x i32> %300, i32 %299, i32 1
>  %302 = bitcast <8 x i32> %46 to <32 x i8>
>  %303 = bitcast <4 x i32> %48 to <16 x i8>
>  %304 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %301, <32 x i8> %302, <16 x i8> %303, i32 2)
>  br label %ENDIF
>
>IF50:                                             ; preds = %ELSE
>  %305 = fmul float %31, %53
>  %306 = fmul float %32, %54
>  %307 = call float @llvm.AMDIL.fraction.(float %305)
>  %308 = call float @llvm.AMDIL.fraction.(float %306)
>  %309 = fsub float -0.000000e+00, %307
>  %310 = fadd float %305, %309
>  %311 = fsub float -0.000000e+00, %308
>  %312 = fadd float %306, %311
>  %313 = fsub float -0.000000e+00, %310
>  %314 = fmul float %53, %31
>  %315 = fadd float %314, %313
>  %316 = fsub float -0.000000e+00, %312
>  %317 = fmul float %54, %32
>  %318 = fadd float %317, %316
>  %319 = fadd float %310, 5.000000e-01
>  %320 = fadd float %312, 5.000000e-01
>  %321 = fdiv float 1.000000e+00, %31
>  %322 = fdiv float 1.000000e+00, %32
>  %323 = fmul float %319, %321
>  %324 = fmul float %320, %322
>  %325 = bitcast float %323 to i32
>  %326 = bitcast float %324 to i32
>  %327 = bitcast float 0.000000e+00 to i32
>  %328 = insertelement <4 x i32> undef, i32 %325, i32 0
>  %329 = insertelement <4 x i32> %328, i32 %326, i32 1
>  %330 = insertelement <4 x i32> %329, i32 %327, i32 2
>  %331 = insertelement <4 x i32> %330, i32 undef, i32 3
>  %332 = bitcast <8 x i32> %42 to <32 x i8>
>  %333 = bitcast <4 x i32> %44 to <16 x i8>
>  %334 = call <4 x float> @llvm.SI.samplel.v4i32(<4 x i32> %331, <32 x i8> %332, <16 x i8> %333, i32 2)
>  %335 = extractelement <4 x float> %334, i32 0
>  %336 = extractelement <4 x float> %334, i32 3
>  %337 = fmul float %336, 6.528000e+04
>  %338 = fmul float %335, 2.550000e+02
>  %339 = fadd float %337, %338
>  %340 = fadd float %339, 5.000000e-01
>  %341 = call float @llvm.AMDIL.fraction.(float %340)
>  %342 = fsub float -0.000000e+00, %341
>  %343 = fadd float %342, %340
>  %344 = fmul float %30, %335
>  %345 = fadd float %344, 5.000000e-01
>  %346 = call float @llvm.AMDIL.fraction.(float %345)
>  %347 = fsub float -0.000000e+00, %346
>  %348 = fadd float %345, %347
>  %349 = fadd float %343, 5.000000e-01
>  %350 = fadd float %348, 5.000000e-01
>  %351 = fcmp oge float %178, 0.000000e+00
>  %352 = sext i1 %351 to i32
>  %353 = bitcast i32 %352 to float
>  %354 = bitcast float %353 to i32
>  %355 = icmp ne i32 %354, 0
>  %.59 = select i1 %355, float %350, float %349
>  %356 = fdiv float 1.000000e+00, %29
>  %357 = fmul float %33, %356
>  %358 = fadd float %357, 5.000000e-01
>  %359 = call float @llvm.AMDIL.fraction.(float %358)
>  %360 = fsub float -0.000000e+00, %359
>  %361 = fadd float %360, %358
>  %362 = fmul float %361, %.59
>  %363 = fsub float -0.000000e+00, %361
>  %364 = fcmp oge float %362, 0.000000e+00
>  %365 = sext i1 %364 to i32
>  %366 = bitcast i32 %365 to float
>  %367 = bitcast float %366 to i32
>  %368 = icmp ne i32 %367, 0
>  %temp28.1 = select i1 %368, float %361, float %363
>  %369 = fdiv float 1.000000e+00, %temp28.1
>  %370 = fmul float %369, %.59
>  %371 = call float @llvm.AMDIL.fraction.(float %370)
>  %372 = fdiv float 1.000000e+00, %361
>  %373 = fmul float %371, %temp28.1
>  %374 = fmul float %372, %.59
>  %375 = call float @llvm.AMDIL.fraction.(float %373)
>  %376 = call float @llvm.AMDIL.fraction.(float %374)
>  %377 = fsub float -0.000000e+00, %375
>  %378 = fadd float %377, %373
>  %379 = fsub float -0.000000e+00, %376
>  %380 = fadd float %379, %374
>  %381 = fadd float %315, %378
>  %382 = fadd float %318, %380
>  %383 = fmul float %381, %29
>  %384 = fmul float %382, %29
>  %385 = call float @llvm.AMDIL.fraction.(float %383)
>  %386 = call float @llvm.AMDIL.fraction.(float %384)
>  %387 = fsub float -0.000000e+00, %385
>  %388 = fadd float %387, %383
>  %389 = fsub float -0.000000e+00, %386
>  %390 = fadd float %389, %384
>  %391 = fadd float %388, 5.000000e-01
>  %392 = fadd float %390, 5.000000e-01
>  %393 = fdiv float 1.000000e+00, %33
>  %394 = fdiv float 1.000000e+00, %34
>  %395 = fmul float %391, %393
>  %396 = fmul float %392, %394
>  %397 = bitcast float %395 to i32
>  %398 = bitcast float %396 to i32
>  %399 = bitcast float 0.000000e+00 to i32
>  %400 = insertelement <4 x i32> undef, i32 %397, i32 0
>  %401 = insertelement <4 x i32> %400, i32 %398, i32 1
>  %402 = insertelement <4 x i32> %401, i32 %399, i32 2
>  %403 = insertelement <4 x i32> %402, i32 undef, i32 3
>  %404 = bitcast <8 x i32> %46 to <32 x i8>
>  %405 = bitcast <4 x i32> %48 to <16 x i8>
>  %406 = call <4 x float> @llvm.SI.samplel.v4i32(<4 x i32> %403, <32 x i8> %404, <16 x i8> %405, i32 2)
>  br label %ENDIF49
>
>ELSE51:                                           ; preds = %ELSE
>  %407 = bitcast float %53 to i32
>  %408 = bitcast float %54 to i32
>  %409 = insertelement <2 x i32> undef, i32 %407, i32 0
>  %410 = insertelement <2 x i32> %409, i32 %408, i32 1
>  %411 = bitcast <8 x i32> %46 to <32 x i8>
>  %412 = bitcast <4 x i32> %48 to <16 x i8>
>  %413 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %410, <32 x i8> %411, <16 x i8> %412, i32 2)
>  br label %ENDIF49
>
>ENDIF49:                                          ; preds = %ELSE51, %IF50
>  %.sink60 = phi <4 x float> [ %413, %ELSE51 ], [ %406, %IF50 ]
>  %414 = extractelement <4 x float> %.sink60, i32 0
>  %415 = extractelement <4 x float> %.sink60, i32 1
>  %416 = extractelement <4 x float> %.sink60, i32 2
>  br label %ENDIF
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>declare void @llvm.AMDGPU.kill(float)
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.fraction.(float) #2
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.samplel.v4i32(<4 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                          ; BEFE0A7E
>	s_load_dwordx4 s[32:35], s[2:3], 0x0                                          ; C0900300
>	s_waitcnt lgkmcnt(0)                                                          ; BF8C007F
>	s_buffer_load_dword s0, s[32:35], 0x1a                                        ; C200211A
>	s_buffer_load_dword s1, s[32:35], 0x18                                        ; C200A118
>	s_waitcnt lgkmcnt(0)                                                          ; BF8C007F
>	v_mov_b32_e32 v4, s0                                                          ; 7E080200
>	v_mad_f32 v2, s1, v2, v4                                                      ; D2820002 04120401
>	s_buffer_load_dword s0, s[32:35], 0x0                                         ; C2002100
>	s_waitcnt lgkmcnt(0)                                                          ; BF8C007F
>	v_subrev_f32_e32 v4, s0, v2                                                   ; 0A080400
>	v_cmp_lt_f32_e64 s[0:1], v4, 0                                                ; D0020000 00010104
>	v_cndmask_b32_e64 v4, 0, -1, s[0:1]                                           ; D2000004 00018280
>	v_and_b32_e32 v4, 1.0, v4                                                     ; 360808F2
>	v_mov_b32_e32 v5, 0x80000000                                                  ; 7E0A02FF 80000000
>	v_xor_b32_e32 v4, v4, v5                                                      ; 3A080B04
>	v_cmpx_le_f32_e32 vcc, 0, v4                                                  ; 7C260880
>	s_cbranch_execnz 3                                                            ; BF890003
>	exp 0, 9, 0, 1, 1, v0, v0, v0, v0                                             ; F8001890 00000000
>	s_endpgm                                                                      ; BF810000
>	v_cmpx_le_f32_e32 vcc, 0, v4                                                  ; 7C260880
>	s_cbranch_execnz 3                                                            ; BF890003
>	exp 0, 9, 0, 1, 1, v0, v0, v0, v0                                             ; F8001890 00000000
>	s_endpgm                                                                      ; BF810000
>	v_cmpx_le_f32_e32 vcc, 0, v4                                                  ; 7C260880
>	s_cbranch_execnz 3                                                            ; BF890003
>	exp 0, 9, 0, 1, 1, v0, v0, v0, v0                                             ; F8001890 00000000
>	s_endpgm                                                                      ; BF810000
>	v_cmpx_le_f32_e32 vcc, 0, v4                                                  ; 7C260880
>	s_cbranch_execnz 3                                                            ; BF890003
>	exp 0, 9, 0, 1, 1, v0, v0, v0, v0                                             ; F8001890 00000000
>	s_endpgm                                                                      ; BF810000
>	s_buffer_load_dword s0, s[32:35], 0x1                                         ; C2002101
>	s_waitcnt lgkmcnt(0)                                                          ; BF8C007F
>	v_sub_f32_e32 v2, s0, v2                                                      ; 08040400
>	v_cmp_lt_f32_e64 s[0:1], v2, 0                                                ; D0020000 00010102
>	v_cndmask_b32_e64 v2, 0, -1, s[0:1]                                           ; D2000002 00018280
>	v_and_b32_e32 v2, 1.0, v2                                                     ; 360404F2
>	v_xor_b32_e32 v2, v2, v5                                                      ; 3A040B02
>	v_cmpx_le_f32_e32 vcc, 0, v2                                                  ; 7C260480
>	s_cbranch_execnz 3                                                            ; BF890003
>	exp 0, 9, 0, 1, 1, v0, v0, v0, v0                                             ; F8001890 00000000
>	s_endpgm                                                                      ; BF810000
>	v_cmpx_le_f32_e32 vcc, 0, v2                                                  ; 7C260480
>	s_cbranch_execnz 3                                                            ; BF890003
>	exp 0, 9, 0, 1, 1, v0, v0, v0, v0                                             ; F8001890 00000000
>	s_endpgm                                                                      ; BF810000
>	v_cmpx_le_f32_e32 vcc, 0, v2                                                  ; 7C260480
>	s_cbranch_execnz 3                                                            ; BF890003
>	exp 0, 9, 0, 1, 1, v0, v0, v0, v0                                             ; F8001890 00000000
>	s_endpgm                                                                      ; BF810000
>	v_cmpx_le_f32_e32 vcc, 0, v2                                                  ; 7C260480
>	s_cbranch_execnz 3                                                            ; BF890003
>	exp 0, 9, 0, 1, 1, v0, v0, v0, v0                                             ; F8001890 00000000
>	s_endpgm                                                                      ; BF810000
>	s_buffer_load_dword s0, s[32:35], 0x25                                        ; C2002125
>	s_buffer_load_dword s1, s[32:35], 0x24                                        ; C200A124
>	s_waitcnt lgkmcnt(0)                                                          ; BF8C007F
>	v_mov_b32_e32 v2, s0                                                          ; 7E040200
>	v_mad_f32 v2, s1, v3, v2                                                      ; D2820002 040A0601
>	s_buffer_load_dword s0, s[32:35], 0x1b                                        ; C200211B
>	s_buffer_load_dword s1, s[32:35], 0x19                                        ; C200A119
>	s_waitcnt lgkmcnt(0)                                                          ; BF8C007F
>	v_mov_b32_e32 v3, s0                                                          ; 7E060200
>	v_mad_f32 v2, s1, v2, v3                                                      ; D2820002 040E0401
>	s_buffer_load_dword s0, s[32:35], 0x2                                         ; C2002102
>	s_waitcnt lgkmcnt(0)                                                          ; BF8C007F
>	v_subrev_f32_e32 v3, s0, v2                                                   ; 0A060400
>	v_cmp_lt_f32_e64 s[0:1], v3, 0                                                ; D0020000 00010103
>	v_cndmask_b32_e64 v3, 0, -1, s[0:1]                                           ; D2000803 00018280
>	v_and_b32_e32 v3, 1.0, v3                                                     ; 360606F2
>	v_xor_b32_e32 v3, v3, v5                                                      ; 3A060B03
>	v_cmpx_le_f32_e32 vcc, 0, v3                                                  ; 7C260680
>	s_cbranch_execnz 3                                                            ; BF890003
>	exp 0, 9, 0, 1, 1, v0, v0, v0, v0                                             ; F8001890 00000000
>	s_endpgm                                                                      ; BF810000
>	v_cmpx_le_f32_e32 vcc, 0, v3                                                  ; 7C260680
>	s_cbranch_execnz 3                                                            ; BF890003
>	exp 0, 9, 0, 1, 1, v0, v0, v0, v0                                             ; F8001890 00000000
>	s_endpgm                                                                      ; BF810000
>	v_cmpx_le_f32_e32 vcc, 0, v3                                                  ; 7C260680
>	s_cbranch_execnz 3                                                            ; BF890003
>	exp 0, 9, 0, 1, 1, v0, v0, v0, v0                                             ; F8001890 00000000
>	s_endpgm                                                                      ; BF810000
>	v_cmpx_le_f32_e32 vcc, 0, v3                                                  ; 7C260680
>	s_cbranch_execnz 3                                                            ; BF890003
>	exp 0, 9, 0, 1, 1, v0, v0, v0, v0                                             ; F8001890 00000000
>	s_endpgm                                                                      ; BF810000
>	s_buffer_load_dword s0, s[32:35], 0x3                                         ; C2002103
>	s_waitcnt lgkmcnt(0)                                                          ; BF8C007F
>	v_sub_f32_e32 v2, s0, v2                                                      ; 08040400
>	v_cmp_lt_f32_e64 s[0:1], v2, 0                                                ; D0020000 00010102
>	v_cndmask_b32_e64 v2, 0, -1, s[0:1]                                           ; D2000002 00018280
>	v_and_b32_e32 v2, 1.0, v2                                                     ; 360404F2
>	v_xor_b32_e32 v2, v2, v5                                                      ; 3A040B02
>	v_cmpx_le_f32_e32 vcc, 0, v2                                                  ; 7C260480
>	s_cbranch_execnz 3                                                            ; BF890003
>	exp 0, 9, 0, 1, 1, v0, v0, v0, v0                                             ; F8001890 00000000
>	s_endpgm                                                                      ; BF810000
>	v_cmpx_le_f32_e32 vcc, 0, v2                                                  ; 7C260480
>	s_cbranch_execnz 3                                                            ; BF890003
>	exp 0, 9, 0, 1, 1, v0, v0, v0, v0                                             ; F8001890 00000000
>	s_endpgm                                                                      ; BF810000
>	v_cmpx_le_f32_e32 vcc, 0, v2                                                  ; 7C260480
>	s_cbranch_execnz 3                                                            ; BF890003
>	exp 0, 9, 0, 1, 1, v0, v0, v0, v0                                             ; F8001890 00000000
>	s_endpgm                                                                      ; BF810000
>	v_cmpx_le_f32_e32 vcc, 0, v2                                                  ; 7C260480
>	s_cbranch_execnz 3                                                            ; BF890003
>	exp 0, 9, 0, 1, 1, v0, v0, v0, v0                                             ; F8001890 00000000
>	s_endpgm                                                                      ; BF810000
>	s_mov_b32 m0, s9                                                              ; BEFC0309
>	v_interp_p1_f32 v4, v0, 1, 1, [m0]                                            ; C8100500
>	v_interp_p2_f32 v4, [v4], v1, 1, 1, [m0]                                      ; C8110501
>	v_interp_p1_f32 v3, v0, 0, 1, [m0]                                            ; C80C0400
>	v_interp_p2_f32 v3, [v3], v1, 0, 1, [m0]                                      ; C80D0401
>	v_interp_p1_f32 v2, v0, 3, 0, [m0]                                            ; C8080300
>	v_interp_p2_f32 v2, [v2], v1, 3, 0, [m0]                                      ; C8090301
>	v_interp_p1_f32 v5, v0, 2, 0, [m0]                                            ; C8140200
>	v_interp_p2_f32 v5, [v5], v1, 2, 0, [m0]                                      ; C8150201
>	v_interp_p1_f32 v6, v0, 1, 0, [m0]                                            ; C8180100
>	v_interp_p2_f32 v6, [v6], v1, 1, 0, [m0]                                      ; C8190101
>	v_interp_p1_f32 v7, v0, 0, 0, [m0]                                            ; C81C0000
>	v_interp_p2_f32 v7, [v7], v1, 0, 0, [m0]                                      ; C81D0001
>	s_buffer_load_dword s0, s[32:35], 0x4                                         ; C2002104
>	s_waitcnt lgkmcnt(0)                                                          ; BF8C007F
>	v_cmp_neq_f32_e64 s[0:1], s0, -s0                                             ; D01A0000 40000000
>	v_cndmask_b32_e64 v0, 0, -1, s[0:1]                                           ; D2000000 00018280
>	v_cmp_eq_i32_e64 s[36:37], v0, 0                                              ; D1040024 00010100
>	s_buffer_load_dword s0, s[32:35], 0x8                                         ; C2002108
>	s_waitcnt lgkmcnt(0)                                                          ; BF8C007F
>	v_cmp_gt_f32_e64 s[2:3], s0, 0.5                                              ; D0080002 0001E000
>	v_cndmask_b32_e64 v1, 0, -1, s[2:3]                                           ; D2000801 00098280
>	s_buffer_load_dword s1, s[32:35], 0xc                                         ; C200A10C
>	v_mov_b32_e32 v0, 0x44800000                                                  ; 7E0002FF 44800000
>	s_waitcnt lgkmcnt(0)                                                          ; BF8C007F
>	v_subrev_f32_e32 v0, s1, v0                                                   ; 0A000001
>	s_load_dwordx4 s[8:11], s[4:5], 0x0                                           ; C0840500
>	s_load_dwordx4 s[12:15], s[4:5], 0x4                                          ; C0860504
>	s_load_dwordx8 s[16:23], s[6:7], 0x0                                          ; C0C80700
>	s_load_dwordx8 s[24:31], s[6:7], 0x8                                          ; C0CC0708
>	s_buffer_load_dword s2, s[32:35], 0x15                                        ; C2012115
>	s_buffer_load_dword s3, s[32:35], 0x14                                        ; C201A114
>	s_buffer_load_dword s6, s[32:35], 0x11                                        ; C2032111
>	s_buffer_load_dword s7, s[32:35], 0x10                                        ; C203A110
>	s_waitcnt lgkmcnt(0)                                                          ; BF8C007F
>	s_and_saveexec_b64 s[4:5], s[36:37]                                           ; BE842424
>	s_xor_b64 s[4:5], exec, s[4:5]                                                ; 8984047E
>	s_cbranch_execz BB0_11                                                        ; BF880000
>	v_cmp_eq_i32_e64 s[32:33], v1, 0                                              ; D1040020 00010101
>	s_and_saveexec_b64 s[32:33], s[32:33]                                         ; BEA02420
>	s_xor_b64 s[32:33], exec, s[32:33]                                            ; 89A0207E
>	image_sample v[8:11], 15, 0, 0, 0, 0, 0, 0, 0, v[3:4], s[24:31], s[12:15]     ; F0800F00 00660803
>	s_waitcnt vmcnt(0)                                                            ; BF8C0770
>	s_or_saveexec_b64 s[32:33], s[32:33]                                          ; BEA02520
>	s_xor_b64 exec, exec, s[32:33]                                                ; 89FE207E
>	s_cbranch_execz BB0_12                                                        ; BF880000
>	v_mul_f32_e32 v8, s6, v4                                                      ; 10100806
>	v_fract_f32_e32 v8, v8                                                        ; 7E104108
>	v_mad_f32 v8, s6, v4, -v8                                                     ; D2820008 84220806
>	v_add_f32_e32 v9, 0.5, v8                                                     ; 061210F0
>	v_rcp_f32_e32 v10, s6                                                         ; 7E145406
>	v_mul_f32_e32 v10, v10, v9                                                    ; 1014130A
>	v_mul_f32_e32 v13, s7, v3                                                     ; 101A0607
>	v_fract_f32_e32 v13, v13                                                      ; 7E1A410D
>	v_mad_f32 v13, s7, v3, -v13                                                   ; D282000D 84360607
>	v_add_f32_e32 v14, 0.5, v13                                                   ; 061C1AF0
>	v_rcp_f32_e32 v15, s7                                                         ; 7E1E5407
>	v_mul_f32_e32 v9, v15, v14                                                    ; 10121D0F
>	v_mov_b32_e32 v11, 0                                                          ; 7E160280
>	image_sample_l v[14:15], 9, 0, 0, 0, 0, 0, 0, 0, v[9:12], s[16:23], s[8:11]   ; F0900900 00440E09
>	s_waitcnt vmcnt(0)                                                            ; BF8C0770
>	v_mul_f32_e32 v16, 0x437f0000, v14                                            ; 10201CFF 437F0000
>	v_mov_b32_e32 v17, 0x477f0000                                                 ; 7E2202FF 477F0000
>	v_mad_f32 v16, v17, v15, v16                                                  ; D2820010 04421F11
>	v_add_f32_e32 v16, 0.5, v16                                                   ; 062020F0
>	v_fract_f32_e32 v17, v16                                                      ; 7E224110
>	v_subrev_f32_e32 v16, v17, v16                                                ; 0A202111
>	v_add_f32_e32 v16, 0.5, v16                                                   ; 062020F0
>	v_mad_f32 v14, s1, v14, 0.5                                                   ; D282000E 03C21C01
>	v_fract_f32_e32 v15, v14                                                      ; 7E1E410E
>	v_subrev_f32_e32 v14, v15, v14                                                ; 0A1C1D0F
>	v_add_f32_e32 v14, 0.5, v14                                                   ; 061C1CF0
>	v_cmp_ge_f32_e64 s[34:35], v0, 0                                              ; D00C0022 00010100
>	v_cndmask_b32_e64 v15, 0, -1, s[34:35]                                        ; D200080F 00898280
>	v_cmp_ne_i32_e64 s[34:35], v15, 0                                             ; D10A0022 0001010F
>	v_cndmask_b32_e64 v14, v16, v14, s[34:35]                                     ; D200000E 008A1D10
>	v_rcp_f32_e32 v15, s0                                                         ; 7E1E5400
>	v_mad_f32 v15, s3, v15, 0.5                                                   ; D282000F 03C21E03
>	v_fract_f32_e32 v16, v15                                                      ; 7E20410F
>	v_subrev_f32_e32 v15, v16, v15                                                ; 0A1E1F10
>	v_mul_f32_e32 v16, v14, v15                                                   ; 10201F0E
>	v_cmp_ge_f32_e64 s[34:35], v16, 0                                             ; D00C0022 00010110
>	v_cndmask_b32_e64 v16, 0, -1, s[34:35]                                        ; D2000010 00898280
>	v_cmp_ne_i32_e64 s[34:35], v16, 0                                             ; D10A0022 00010110
>	v_xor_b32_e32 v16, 0x80000000, v15                                            ; 3A201EFF 80000000
>	v_cndmask_b32_e64 v16, v16, v15, s[34:35]                                     ; D2000010 008A1F10
>	v_rcp_f32_e32 v17, v16                                                        ; 7E225510
>	v_mul_f32_e32 v17, v14, v17                                                   ; 1022230E
>	v_fract_f32_e32 v17, v17                                                      ; 7E224111
>	v_mul_f32_e32 v18, v16, v17                                                   ; 10242310
>	v_fract_f32_e32 v18, v18                                                      ; 7E244112
>	v_mad_f32 v16, v17, v16, -v18                                                 ; D2820010 844A2111
>	v_mad_f32 v13, s7, v3, -v13                                                   ; D282000D 84360607
>	v_add_f32_e32 v13, v16, v13                                                   ; 061A1B10
>	v_mul_f32_e32 v16, s0, v13                                                    ; 10201A00
>	v_fract_f32_e32 v16, v16                                                      ; 7E204110
>	v_mad_f32 v13, v13, s0, -v16                                                  ; D282000D 8440010D
>	v_add_f32_e32 v13, 0.5, v13                                                   ; 061A1AF0
>	v_rcp_f32_e32 v16, s3                                                         ; 7E205403
>	v_mul_f32_e32 v9, v16, v13                                                    ; 10121B10
>	v_rcp_f32_e32 v13, v15                                                        ; 7E1A550F
>	v_mul_f32_e32 v15, v14, v13                                                   ; 101E1B0E
>	v_fract_f32_e32 v15, v15                                                      ; 7E1E410F
>	v_mad_f32 v13, v13, v14, -v15                                                 ; D282000D 843E1D0D
>	v_mad_f32 v8, s6, v4, -v8                                                     ; D2820008 84220806
>	v_add_f32_e32 v8, v13, v8                                                     ; 0610110D
>	v_mul_f32_e32 v13, s0, v8                                                     ; 101A1000
>	v_fract_f32_e32 v13, v13                                                      ; 7E1A410D
>	v_mad_f32 v8, v8, s0, -v13                                                    ; D2820008 84340108
>	v_add_f32_e32 v8, 0.5, v8                                                     ; 061010F0
>	v_rcp_f32_e32 v13, s2                                                         ; 7E1A5402
>	v_mul_f32_e32 v10, v13, v8                                                    ; 1014110D
>	image_sample_l v[8:11], 15, 0, 0, 0, 0, 0, 0, 0, v[9:12], s[24:31], s[12:15]  ; F0900F00 00660809
>	s_waitcnt vmcnt(0)                                                            ; BF8C0770
>	s_or_b64 exec, exec, s[32:33]                                                 ; 88FE207E
>	v_mov_b32_e32 v12, v8                                                         ; 7E180308
>	v_mov_b32_e32 v13, v9                                                         ; 7E1A0309
>	v_mov_b32_e32 v14, v10                                                        ; 7E1C030A
>	v_mov_b32_e32 v15, v11                                                        ; 7E1E030B
>	s_or_saveexec_b64 s[4:5], s[4:5]                                              ; BE842504
>	s_xor_b64 exec, exec, s[4:5]                                                  ; 89FE047E
>	s_cbranch_execz BB0_4                                                         ; BF880000
>	v_cmp_eq_i32_e64 s[32:33], v1, 0                                              ; D1040020 00010101
>	s_and_saveexec_b64 s[32:33], s[32:33]                                         ; BEA02420
>	s_xor_b64 s[32:33], exec, s[32:33]                                            ; 89A0207E
>	image_sample v[12:15], 15, 0, 0, 0, 0, 0, 0, 0, v[3:4], s[24:31], s[12:15]    ; F0800F00 00660C03
>	v_mov_b32_e32 v10, v5                                                         ; 7E140305
>	v_mov_b32_e32 v9, v6                                                          ; 7E120306
>	v_mov_b32_e32 v8, v7                                                          ; 7E100307
>	s_waitcnt vmcnt(0)                                                            ; BF8C0770
>	s_or_saveexec_b64 s[32:33], s[32:33]                                          ; BEA02520
>	s_waitcnt expcnt(0)                                                           ; BF8C070F
>	s_xor_b64 exec, exec, s[32:33]                                                ; 89FE207E
>	s_cbranch_execz BB0_3                                                         ; BF880000
>	v_mul_f32_e32 v1, s6, v4                                                      ; 10020806
>	v_fract_f32_e32 v1, v1                                                        ; 7E024101
>	v_mad_f32 v1, s6, v4, -v1                                                     ; D2820001 84060806
>	v_add_f32_e32 v8, 0.5, v1                                                     ; 061002F0
>	v_rcp_f32_e32 v9, s6                                                          ; 7E125406
>	v_mul_f32_e32 v9, v9, v8                                                      ; 10121109
>	v_mul_f32_e32 v12, s7, v3                                                     ; 10180607
>	v_fract_f32_e32 v12, v12                                                      ; 7E18410C
>	v_mad_f32 v12, s7, v3, -v12                                                   ; D282000C 84320607
>	v_add_f32_e32 v13, 0.5, v12                                                   ; 061A18F0
>	v_rcp_f32_e32 v14, s7                                                         ; 7E1C5407
>	v_mul_f32_e32 v8, v14, v13                                                    ; 10101B0E
>	v_mov_b32_e32 v10, 0                                                          ; 7E140280
>	image_sample_l v[13:14], 9, 0, 0, 0, 0, 0, 0, 0, v[8:11], s[16:23], s[8:11]   ; F0900900 00440D08
>	s_waitcnt vmcnt(0)                                                            ; BF8C0770
>	v_mul_f32_e32 v15, 0x437f0000, v13                                            ; 101E1AFF 437F0000
>	v_mov_b32_e32 v16, 0x477f0000                                                 ; 7E2002FF 477F0000
>	v_mad_f32 v15, v16, v14, v15                                                  ; D282000F 043E1D10
>	v_add_f32_e32 v15, 0.5, v15                                                   ; 061E1EF0
>	v_fract_f32_e32 v16, v15                                                      ; 7E20410F
>	v_subrev_f32_e32 v15, v16, v15                                                ; 0A1E1F10
>	v_add_f32_e32 v15, 0.5, v15                                                   ; 061E1EF0
>	v_mad_f32 v13, s1, v13, 0.5                                                   ; D282000D 03C21A01
>	v_fract_f32_e32 v14, v13                                                      ; 7E1C410D
>	v_subrev_f32_e32 v13, v14, v13                                                ; 0A1A1B0E
>	v_add_f32_e32 v13, 0.5, v13                                                   ; 061A1AF0
>	v_cmp_ge_f32_e64 s[34:35], v0, 0                                              ; D00C0022 00010100
>	v_cndmask_b32_e64 v0, 0, -1, s[34:35]                                         ; D2000000 00898280
>	v_cmp_ne_i32_e64 s[34:35], v0, 0                                              ; D10A0022 00010100
>	v_cndmask_b32_e64 v0, v15, v13, s[34:35]                                      ; D2000000 188A1B0F
>	v_rcp_f32_e32 v13, s0                                                         ; 7E1A5400
>	v_mad_f32 v13, s3, v13, 0.5                                                   ; D282000D 03C21A03
>	v_fract_f32_e32 v14, v13                                                      ; 7E1C410D
>	v_subrev_f32_e32 v13, v14, v13                                                ; 0A1A1B0E
>	v_mul_f32_e32 v14, v0, v13                                                    ; 101C1B00
>	v_cmp_ge_f32_e64 s[34:35], v14, 0                                             ; D00C0022 0001010E
>	v_cndmask_b32_e64 v14, 0, -1, s[34:35]                                        ; D200000E 00898280
>	v_cmp_ne_i32_e64 s[34:35], v14, 0                                             ; D10A0022 0001010E
>	v_xor_b32_e32 v14, 0x80000000, v13                                            ; 3A1C1AFF 80000000
>	v_cndmask_b32_e64 v14, v14, v13, s[34:35]                                     ; D200000E 108A1B0E
>	v_rcp_f32_e32 v15, v14                                                        ; 7E1E550E
>	v_mul_f32_e32 v15, v15, v0                                                    ; 101E010F
>	v_fract_f32_e32 v15, v15                                                      ; 7E1E410F
>	v_mul_f32_e32 v16, v14, v15                                                   ; 10201F0E
>	v_fract_f32_e32 v16, v16                                                      ; 7E204110
>	v_mad_f32 v14, v15, v14, -v16                                                 ; D282000E 84421D0F
>	v_mad_f32 v12, s7, v3, -v12                                                   ; D282000C 84320607
>	v_add_f32_e32 v12, v14, v12                                                   ; 0618190E
>	v_mul_f32_e32 v14, s0, v12                                                    ; 101C1800
>	v_fract_f32_e32 v14, v14                                                      ; 7E1C410E
>	v_mad_f32 v12, v12, s0, -v14                                                  ; D282000C 8438010C
>	v_add_f32_e32 v12, 0.5, v12                                                   ; 061818F0
>	v_rcp_f32_e32 v14, s3                                                         ; 7E1C5403
>	v_mul_f32_e32 v8, v14, v12                                                    ; 1010190E
>	v_rcp_f32_e32 v12, v13                                                        ; 7E18550D
>	v_mul_f32_e32 v13, v0, v12                                                    ; 101A1900
>	v_fract_f32_e32 v13, v13                                                      ; 7E1A410D
>	v_mad_f32 v0, v12, v0, -v13                                                   ; D2820000 8436010C
>	v_mad_f32 v1, s6, v4, -v1                                                     ; D2820001 84060806
>	v_add_f32_e32 v0, v0, v1                                                      ; 06000300
>	v_mul_f32_e32 v1, s0, v0                                                      ; 10020000
>	v_fract_f32_e32 v1, v1                                                        ; 7E024101
>	v_mad_f32 v0, v0, s0, -v1                                                     ; D2820000 84040100
>	v_add_f32_e32 v0, 0.5, v0                                                     ; 060000F0
>	v_rcp_f32_e32 v1, s2                                                          ; 7E025402
>	v_mul_f32_e32 v9, v1, v0                                                      ; 10120101
>	image_sample_l v[12:15], 15, 0, 0, 0, 0, 0, 0, 0, v[8:11], s[24:31], s[12:15] ; F0900F00 00660C08
>	v_mov_b32_e32 v10, v5                                                         ; 7E140305
>	v_mov_b32_e32 v9, v6                                                          ; 7E120306
>	v_mov_b32_e32 v8, v7                                                          ; 7E100307
>	s_waitcnt vmcnt(0)                                                            ; BF8C0770
>	s_or_b64 exec, exec, s[32:33]                                                 ; 88FE207E
>	s_or_b64 exec, exec, s[4:5]                                                   ; 88FE047E
>	v_cvt_pkrtz_f16_f32_e32 v0, v8, v9                                            ; 5E001308
>	v_mul_f32_e32 v1, v2, v15                                                     ; 10021F02
>	v_cvt_pkrtz_f16_f32_e32 v1, v10, v1                                           ; 5E02030A
>	exp 15, 0, 1, 1, 1, v0, v1, v0, v1                                            ; F8001C0F 01000100
>	s_endpgm                                                                      ; BF810000
>
>SGPRS: 40
>VGPRS: 20
>Code Size: 1640 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL IN[2]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL OUT[2], GENERIC[10]
>DCL CONST[0..8]
>DCL TEMP[0..3], LOCAL
>IMM[0] FLT32 {    0.0000,     1.0000,     0.0000,     0.0000}
>  0: MAD TEMP[0].xy, IN[0], CONST[0], CONST[0].zwzw
>  1: MUL TEMP[1], TEMP[0].yyyy, CONST[6]
>  2: MAD TEMP[0], TEMP[0].xxxx, CONST[5], TEMP[1]
>  3: ADD TEMP[0], TEMP[0], CONST[8]
>  4: ADD TEMP[1].x, CONST[4].xxxx, CONST[4].yyyy
>  5: MUL TEMP[1], TEMP[0], TEMP[1].xxxx
>  6: MAD TEMP[2].xy, IN[1], CONST[1], CONST[1].zwzw
>  7: MOV TEMP[2].xy, TEMP[2].xyxx
>  8: MUL TEMP[0], CONST[2], IN[2]
>  9: MUL TEMP[3].w, TEMP[0].wwww, CONST[3].wwww
> 10: MOV TEMP[3].w, TEMP[3].wwww
> 11: MOV TEMP[3].xyz, TEMP[0].xyzx
> 12: MOV TEMP[2].zw, IMM[0].yyxy
> 13: MOV OUT[1], TEMP[3]
> 14: MOV OUT[2], TEMP[2]
> 15: MOV OUT[0], TEMP[1]
> 16: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 0)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 4)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %12, i32 8)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %12, i32 12)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %12, i32 16)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %12, i32 20)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %12, i32 24)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %12, i32 28)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %12, i32 32)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %12, i32 36)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %12, i32 40)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %12, i32 44)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %12, i32 60)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %12, i32 64)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %12, i32 68)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %12, i32 80)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %12, i32 84)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %12, i32 88)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %12, i32 92)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %12, i32 96)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %12, i32 100)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %12, i32 104)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %12, i32 108)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %12, i32 128)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %12, i32 132)
>  %38 = call float @llvm.SI.load.const(<16 x i8> %12, i32 136)
>  %39 = call float @llvm.SI.load.const(<16 x i8> %12, i32 140)
>  %40 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %41 = load <16 x i8> addrspace(2)* %40, !tbaa !0
>  %42 = add i32 %5, %7
>  %43 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %41, i32 0, i32 %42)
>  %44 = extractelement <4 x float> %43, i32 0
>  %45 = extractelement <4 x float> %43, i32 1
>  %46 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %47 = load <16 x i8> addrspace(2)* %46, !tbaa !0
>  %48 = add i32 %5, %7
>  %49 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %47, i32 0, i32 %48)
>  %50 = extractelement <4 x float> %49, i32 0
>  %51 = extractelement <4 x float> %49, i32 1
>  %52 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 2
>  %53 = load <16 x i8> addrspace(2)* %52, !tbaa !0
>  %54 = add i32 %5, %7
>  %55 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %53, i32 0, i32 %54)
>  %56 = extractelement <4 x float> %55, i32 0
>  %57 = extractelement <4 x float> %55, i32 1
>  %58 = extractelement <4 x float> %55, i32 2
>  %59 = extractelement <4 x float> %55, i32 3
>  %60 = fmul float %44, %13
>  %61 = fadd float %60, %15
>  %62 = fmul float %45, %14
>  %63 = fadd float %62, %16
>  %64 = fmul float %63, %32
>  %65 = fmul float %63, %33
>  %66 = fmul float %63, %34
>  %67 = fmul float %63, %35
>  %68 = fmul float %61, %28
>  %69 = fadd float %68, %64
>  %70 = fmul float %61, %29
>  %71 = fadd float %70, %65
>  %72 = fmul float %61, %30
>  %73 = fadd float %72, %66
>  %74 = fmul float %61, %31
>  %75 = fadd float %74, %67
>  %76 = fadd float %69, %36
>  %77 = fadd float %71, %37
>  %78 = fadd float %73, %38
>  %79 = fadd float %75, %39
>  %80 = fadd float %26, %27
>  %81 = fmul float %76, %80
>  %82 = fmul float %77, %80
>  %83 = fmul float %78, %80
>  %84 = fmul float %79, %80
>  %85 = fmul float %50, %17
>  %86 = fadd float %85, %19
>  %87 = fmul float %51, %18
>  %88 = fadd float %87, %20
>  %89 = fmul float %21, %56
>  %90 = fmul float %22, %57
>  %91 = fmul float %23, %58
>  %92 = fmul float %24, %59
>  %93 = fmul float %92, %25
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %89, float %90, float %91, float %93)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %86, float %88, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %81, float %82, float %83, float %84)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                             ; 4A00000A
>	s_load_dwordx4 s[4:7], s[8:9], 0x0                    ; C0820900
>	s_load_dwordx4 s[12:15], s[8:9], 0x4                  ; C0860904
>	s_load_dwordx4 s[8:11], s[8:9], 0x8                   ; C0840908
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[8:11], 0 idxen  ; E00C2000 80020100
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                    ; C0800300
>	s_waitcnt vmcnt(0) lgkmcnt(0)                         ; BF8C0070
>	s_buffer_load_dword s8, s[0:3], 0xa                   ; C204010A
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v5, s8, v3                              ; 100A0608
>	s_buffer_load_dword s8, s[0:3], 0x9                   ; C2040109
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v6, s8, v2                              ; 100C0408
>	s_buffer_load_dword s8, s[0:3], 0x8                   ; C2040108
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v7, s8, v1                              ; 100E0208
>	s_buffer_load_dword s8, s[0:3], 0xb                   ; C204010B
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v1, s8, v4                              ; 10020808
>	s_buffer_load_dword s8, s[0:3], 0xf                   ; C204010F
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v1, s8, v1                              ; 10020208
>	exp 15, 32, 0, 0, 0, v7, v6, v5, v1                   ; F800020F 01050607
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	buffer_load_format_xyzw v[1:4], v0, s[12:15], 0 idxen ; E00C2000 80030100
>	s_buffer_load_dword s8, s[0:3], 0x7                   ; C2040107
>	s_buffer_load_dword s9, s[0:3], 0x5                   ; C2048105
>	s_waitcnt vmcnt(0) lgkmcnt(0)                         ; BF8C0070
>	v_mov_b32_e32 v5, s8                                  ; 7E0A0208
>	v_mad_f32 v5, s9, v2, v5                              ; D2820005 04160409
>	s_buffer_load_dword s8, s[0:3], 0x6                   ; C2040106
>	s_buffer_load_dword s9, s[0:3], 0x4                   ; C2048104
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mov_b32_e32 v6, s8                                  ; 7E0C0208
>	v_mad_f32 v1, s9, v1, v6                              ; D2820001 041A0209
>	v_mov_b32_e32 v2, 1.0                                 ; 7E0402F2
>	v_mov_b32_e32 v3, 0                                   ; 7E060280
>	exp 15, 33, 0, 0, 0, v1, v5, v3, v2                   ; F800021F 02030501
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	buffer_load_format_xyzw v[0:3], v0, s[4:7], 0 idxen   ; E00C2000 80010000
>	s_buffer_load_dword s4, s[0:3], 0x2                   ; C2020102
>	s_buffer_load_dword s5, s[0:3], 0x0                   ; C2028100
>	s_waitcnt vmcnt(0) lgkmcnt(0)                         ; BF8C0070
>	v_mov_b32_e32 v4, s4                                  ; 7E080204
>	v_mad_f32 v4, s5, v0, v4                              ; D2820004 04120005
>	s_buffer_load_dword s4, s[0:3], 0x3                   ; C2020103
>	s_buffer_load_dword s5, s[0:3], 0x1                   ; C2028101
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mov_b32_e32 v5, s4                                  ; 7E0A0204
>	v_mad_f32 v0, s5, v1, v5                              ; D2820000 04160205
>	s_buffer_load_dword s4, s[0:3], 0x1b                  ; C202011B
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v1, s4, v0                              ; 10020004
>	s_buffer_load_dword s4, s[0:3], 0x17                  ; C2020117
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v1, v4, s4, v1                              ; D2820001 04040904
>	s_buffer_load_dword s4, s[0:3], 0x23                  ; C2020123
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v1, s4, v1                              ; 06020204
>	s_buffer_load_dword s4, s[0:3], 0x11                  ; C2020111
>	s_buffer_load_dword s5, s[0:3], 0x10                  ; C2028110
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mov_b32_e32 v2, s4                                  ; 7E040204
>	v_add_f32_e32 v2, s5, v2                              ; 06040405
>	v_mul_f32_e32 v1, v2, v1                              ; 10020302
>	s_buffer_load_dword s4, s[0:3], 0x1a                  ; C202011A
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v3, s4, v0                              ; 10060004
>	s_buffer_load_dword s4, s[0:3], 0x16                  ; C2020116
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v3, v4, s4, v3                              ; D2820003 040C0904
>	s_buffer_load_dword s4, s[0:3], 0x22                  ; C2020122
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v3, s4, v3                              ; 06060604
>	v_mul_f32_e32 v3, v2, v3                              ; 10060702
>	s_buffer_load_dword s4, s[0:3], 0x19                  ; C2020119
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v5, s4, v0                              ; 100A0004
>	s_buffer_load_dword s4, s[0:3], 0x15                  ; C2020115
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v5, v4, s4, v5                              ; D2820005 04140904
>	s_buffer_load_dword s4, s[0:3], 0x21                  ; C2020121
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v5, s4, v5                              ; 060A0A04
>	v_mul_f32_e32 v5, v2, v5                              ; 100A0B02
>	s_buffer_load_dword s4, s[0:3], 0x18                  ; C2020118
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v0, s4, v0                              ; 10000004
>	s_buffer_load_dword s4, s[0:3], 0x14                  ; C2020114
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v0, v4, s4, v0                              ; D2820000 04000904
>	s_buffer_load_dword s0, s[0:3], 0x20                  ; C2000120
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v0, s0, v0                              ; 06000000
>	v_mul_f32_e32 v0, v2, v0                              ; 10000102
>	exp 15, 12, 0, 1, 0, v0, v5, v3, v1                   ; F80008CF 01030500
>	s_endpgm                                              ; BF810000
>
>SGPRS: 24
>VGPRS: 8
>Code Size: 448 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL IN[1], GENERIC[10], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SAMP[1]
>DCL CONST[0..3]
>DCL TEMP[0..7], LOCAL
>IMM[0] FLT32 {    0.5000,  1024.0000,     0.0000,   255.0000}
>IMM[1] FLT32 {  256.0000,     0.0000,     0.0000,     0.0000}
>  0: MOV TEMP[0].xz, IMM[0].xxyx
>  1: FSLT TEMP[1].x, IMM[0].xxxx, CONST[0].xxxx
>  2: UIF TEMP[1].xxxx :0
>  3:   MUL TEMP[1].yw, CONST[2].xxzy, IN[1].xxzy
>  4:   MOV TEMP[0].yw, TEMP[1].wyww
>  5:   FRC TEMP[1].xy, TEMP[0].ywzw
>  6:   MOV TEMP[1].xy, TEMP[1].xyxx
>  7:   ADD TEMP[2].yw, TEMP[0], -TEMP[1].xxzy
>  8:   MOV TEMP[0].yw, TEMP[2].wyww
>  9:   MAD TEMP[2].xy, IN[1], CONST[2], -TEMP[0].ywzw
> 10:   MOV TEMP[1].xy, TEMP[2].xyxx
> 11:   ADD TEMP[2].yw, TEMP[0], IMM[0].xxxx
> 12:   MOV TEMP[0].yw, TEMP[2].wyww
> 13:   RCP TEMP[2].x, CONST[2].xxxx
> 14:   RCP TEMP[3].x, CONST[2].yyyy
> 15:   MOV TEMP[2].y, TEMP[3].xxxx
> 16:   MUL TEMP[3].xy, TEMP[0].ywzw, TEMP[2]
> 17:   MOV TEMP[3].xy, TEMP[3].xyyy
> 18:   MOV TEMP[3].w, IMM[0].zzzz
> 19:   TXL TEMP[3], TEMP[3], SAMP[0], 2D
> 20:   MOV TEMP[2].zw, TEMP[3]
> 21:   ADD TEMP[4].y, IMM[0].yyyy, -CONST[1].xxxx
> 22:   MOV TEMP[0].y, TEMP[4].yyyy
> 23:   MUL TEMP[5].z, TEMP[3].xxxx, IMM[0].wwww
> 24:   MAD TEMP[6].w, TEMP[3].wwww, IMM[0].wwww, IMM[0].xxxx
> 25:   FRC TEMP[7].z, TEMP[6].wwww
> 26:   ADD TEMP[6].w, TEMP[6].wwww, -TEMP[7].zzzz
> 27:   MAD TEMP[5].z, TEMP[6].wwww, IMM[1].xxxx, TEMP[5].zzzz
> 28:   ADD TEMP[5].z, TEMP[5].zzzz, IMM[0].xxxx
> 29:   FRC TEMP[6].w, TEMP[5].zzzz
> 30:   ADD TEMP[5].z, -TEMP[6].wwww, TEMP[5].zzzz
> 31:   MOV TEMP[0].z, TEMP[5].zzzz
> 32:   MAD TEMP[3].w, CONST[1].xxxx, TEMP[3].xxxx, IMM[0].xxxx
> 33:   FRC TEMP[5].z, TEMP[3].wwww
> 34:   MOV TEMP[1].z, TEMP[5].zzzz
> 35:   ADD TEMP[3].w, TEMP[3].wwww, -TEMP[5].zzzz
> 36:   MOV TEMP[0].w, TEMP[3].wwww
> 37:   ADD TEMP[3].zw, TEMP[0], IMM[0].xxxx
> 38:   FSGE TEMP[4].x, TEMP[4].yyyy, IMM[0].zzzz
> 39:   UIF TEMP[4].xxxx :0
> 40:     MOV TEMP[4].x, TEMP[3].wwww
> 41:   ELSE :0
> 42:     MOV TEMP[4].x, TEMP[3].zzzz
> 43:   ENDIF
> 44:   MOV TEMP[0].y, TEMP[4].xxxx
> 45:   RCP TEMP[3].x, CONST[0].xxxx
> 46:   MAD TEMP[3].x, CONST[3].xxxx, TEMP[3].xxxx, IMM[0].xxxx
> 47:   FRC TEMP[5].z, TEMP[3].xxxx
> 48:   ADD TEMP[3].x, -TEMP[5].zzzz, TEMP[3].xxxx
> 49:   MUL TEMP[5].z, TEMP[3].xxxx, TEMP[4].xxxx
> 50:   MOV TEMP[6].x, -TEMP[3].xxxx
> 51:   FSGE TEMP[5].x, TEMP[5].zzzz, IMM[0].zzzz
> 52:   UIF TEMP[5].xxxx :0
> 53:     MOV TEMP[5].x, TEMP[3].xxxx
> 54:   ELSE :0
> 55:     MOV TEMP[5].x, TEMP[6].xxxx
> 56:   ENDIF
> 57:   MOV TEMP[0].z, TEMP[5].xxxx
> 58:   RCP TEMP[5].x, TEMP[5].xxxx
> 59:   MUL TEMP[4].w, TEMP[5].xxxx, TEMP[4].xxxx
> 60:   FRC TEMP[4].w, TEMP[4].wwww
> 61:   MOV TEMP[0].w, TEMP[4].wwww
> 62:   RCP TEMP[0].x, TEMP[3].xxxx
> 63:   MUL TEMP[3].xy, TEMP[0].wxzw, TEMP[0].zyzw
> 64:   MOV TEMP[2].xy, TEMP[3].xyxx
> 65:   FRC TEMP[3].xy, TEMP[2]
> 66:   MOV TEMP[0].xy, TEMP[3].xyxx
> 67:   ADD TEMP[2].xy, -TEMP[0], TEMP[2]
> 68:   MOV TEMP[0].xy, TEMP[2].xyxx
> 69:   ADD TEMP[2].xy, TEMP[1], TEMP[0]
> 70:   MOV TEMP[0].xy, TEMP[2].xyxx
> 71:   MUL TEMP[2].xy, TEMP[0], CONST[0].xxxx
> 72:   MOV TEMP[0].xy, TEMP[2].xyxx
> 73:   FRC TEMP[2].zw, TEMP[2].xyxy
> 74:   MOV TEMP[0].zw, TEMP[2].wwzw
> 75:   ADD TEMP[2].xy, -TEMP[2].zwzw, TEMP[0]
> 76:   MOV TEMP[0].xy, TEMP[2].xyxx
> 77:   ADD TEMP[2].xy, TEMP[0], IMM[0].xxxx
> 78:   MOV TEMP[0].xy, TEMP[2].xyxx
> 79:   RCP TEMP[1].x, CONST[3].xxxx
> 80:   RCP TEMP[2].x, CONST[3].yyyy
> 81:   MOV TEMP[1].y, TEMP[2].xxxx
> 82:   MUL TEMP[1].xy, TEMP[0], TEMP[1]
> 83:   MOV TEMP[1].xy, TEMP[1].xyyy
> 84:   MOV TEMP[1].w, IMM[0].zzzz
> 85:   TXL TEMP[1], TEMP[1], SAMP[1], 2D
> 86:   MOV TEMP[0], TEMP[1]
> 87: ELSE :0
> 88:   MOV TEMP[1].xy, IN[1].xyyy
> 89:   TEX TEMP[1], TEMP[1], SAMP[1], 2D
> 90:   MOV TEMP[0], TEMP[1]
> 91: ENDIF
> 92: MUL TEMP[0], TEMP[0], IN[0]
> 93: MOV OUT[0], TEMP[0]
> 94: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %23 = load <16 x i8> addrspace(2)* %22, !tbaa !0
>  %24 = call float @llvm.SI.load.const(<16 x i8> %23, i32 0)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %23, i32 16)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %23, i32 32)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %23, i32 36)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %23, i32 48)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %23, i32 52)
>  %30 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %31 = load <8 x i32> addrspace(2)* %30, !tbaa !0
>  %32 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %33 = load <4 x i32> addrspace(2)* %32, !tbaa !0
>  %34 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 1
>  %35 = load <8 x i32> addrspace(2)* %34, !tbaa !0
>  %36 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 1
>  %37 = load <4 x i32> addrspace(2)* %36, !tbaa !0
>  %38 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %39 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %40 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %5, <2 x i32> %7)
>  %41 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %5, <2 x i32> %7)
>  %42 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %5, <2 x i32> %7)
>  %43 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %5, <2 x i32> %7)
>  %44 = fcmp olt float 5.000000e-01, %24
>  %45 = sext i1 %44 to i32
>  %46 = bitcast i32 %45 to float
>  %47 = bitcast float %46 to i32
>  %48 = icmp ne i32 %47, 0
>  br i1 %48, label %IF, label %ELSE
>
>IF:                                               ; preds = %main_body
>  %49 = fmul float %26, %42
>  %50 = fmul float %27, %43
>  %51 = call float @llvm.AMDIL.fraction.(float %49)
>  %52 = call float @llvm.AMDIL.fraction.(float %50)
>  %53 = fsub float -0.000000e+00, %51
>  %54 = fadd float %49, %53
>  %55 = fsub float -0.000000e+00, %52
>  %56 = fadd float %50, %55
>  %57 = fsub float -0.000000e+00, %54
>  %58 = fmul float %42, %26
>  %59 = fadd float %58, %57
>  %60 = fsub float -0.000000e+00, %56
>  %61 = fmul float %43, %27
>  %62 = fadd float %61, %60
>  %63 = fadd float %54, 5.000000e-01
>  %64 = fadd float %56, 5.000000e-01
>  %65 = fdiv float 1.000000e+00, %26
>  %66 = fdiv float 1.000000e+00, %27
>  %67 = fmul float %63, %65
>  %68 = fmul float %64, %66
>  %69 = bitcast float %67 to i32
>  %70 = bitcast float %68 to i32
>  %71 = bitcast float 0.000000e+00 to i32
>  %72 = insertelement <4 x i32> undef, i32 %69, i32 0
>  %73 = insertelement <4 x i32> %72, i32 %70, i32 1
>  %74 = insertelement <4 x i32> %73, i32 %71, i32 2
>  %75 = insertelement <4 x i32> %74, i32 undef, i32 3
>  %76 = bitcast <8 x i32> %31 to <32 x i8>
>  %77 = bitcast <4 x i32> %33 to <16 x i8>
>  %78 = call <4 x float> @llvm.SI.samplel.v4i32(<4 x i32> %75, <32 x i8> %76, <16 x i8> %77, i32 2)
>  %79 = extractelement <4 x float> %78, i32 0
>  %80 = extractelement <4 x float> %78, i32 3
>  %81 = fsub float -0.000000e+00, %25
>  %82 = fadd float 1.024000e+03, %81
>  %83 = fmul float %79, 2.550000e+02
>  %84 = fmul float %80, 2.550000e+02
>  %85 = fadd float %84, 5.000000e-01
>  %86 = call float @llvm.AMDIL.fraction.(float %85)
>  %87 = fsub float -0.000000e+00, %86
>  %88 = fadd float %85, %87
>  %89 = fmul float %88, 2.560000e+02
>  %90 = fadd float %89, %83
>  %91 = fadd float %90, 5.000000e-01
>  %92 = call float @llvm.AMDIL.fraction.(float %91)
>  %93 = fsub float -0.000000e+00, %92
>  %94 = fadd float %93, %91
>  %95 = fmul float %25, %79
>  %96 = fadd float %95, 5.000000e-01
>  %97 = call float @llvm.AMDIL.fraction.(float %96)
>  %98 = fsub float -0.000000e+00, %97
>  %99 = fadd float %96, %98
>  %100 = fadd float %94, 5.000000e-01
>  %101 = fadd float %99, 5.000000e-01
>  %102 = fcmp oge float %82, 0.000000e+00
>  %103 = sext i1 %102 to i32
>  %104 = bitcast i32 %103 to float
>  %105 = bitcast float %104 to i32
>  %106 = icmp ne i32 %105, 0
>  %. = select i1 %106, float %101, float %100
>  %107 = fdiv float 1.000000e+00, %24
>  %108 = fmul float %28, %107
>  %109 = fadd float %108, 5.000000e-01
>  %110 = call float @llvm.AMDIL.fraction.(float %109)
>  %111 = fsub float -0.000000e+00, %110
>  %112 = fadd float %111, %109
>  %113 = fmul float %112, %.
>  %114 = fsub float -0.000000e+00, %112
>  %115 = fcmp oge float %113, 0.000000e+00
>  %116 = sext i1 %115 to i32
>  %117 = bitcast i32 %116 to float
>  %118 = bitcast float %117 to i32
>  %119 = icmp ne i32 %118, 0
>  %temp20.0 = select i1 %119, float %112, float %114
>  %120 = fdiv float 1.000000e+00, %temp20.0
>  %121 = fmul float %120, %.
>  %122 = call float @llvm.AMDIL.fraction.(float %121)
>  %123 = fdiv float 1.000000e+00, %112
>  %124 = fmul float %122, %temp20.0
>  %125 = fmul float %123, %.
>  %126 = call float @llvm.AMDIL.fraction.(float %124)
>  %127 = call float @llvm.AMDIL.fraction.(float %125)
>  %128 = fsub float -0.000000e+00, %126
>  %129 = fadd float %128, %124
>  %130 = fsub float -0.000000e+00, %127
>  %131 = fadd float %130, %125
>  %132 = fadd float %59, %129
>  %133 = fadd float %62, %131
>  %134 = fmul float %132, %24
>  %135 = fmul float %133, %24
>  %136 = call float @llvm.AMDIL.fraction.(float %134)
>  %137 = call float @llvm.AMDIL.fraction.(float %135)
>  %138 = fsub float -0.000000e+00, %136
>  %139 = fadd float %138, %134
>  %140 = fsub float -0.000000e+00, %137
>  %141 = fadd float %140, %135
>  %142 = fadd float %139, 5.000000e-01
>  %143 = fadd float %141, 5.000000e-01
>  %144 = fdiv float 1.000000e+00, %28
>  %145 = fdiv float 1.000000e+00, %29
>  %146 = fmul float %142, %144
>  %147 = fmul float %143, %145
>  %148 = bitcast float %146 to i32
>  %149 = bitcast float %147 to i32
>  %150 = bitcast float 0.000000e+00 to i32
>  %151 = insertelement <4 x i32> undef, i32 %148, i32 0
>  %152 = insertelement <4 x i32> %151, i32 %149, i32 1
>  %153 = insertelement <4 x i32> %152, i32 %150, i32 2
>  %154 = insertelement <4 x i32> %153, i32 undef, i32 3
>  %155 = bitcast <8 x i32> %35 to <32 x i8>
>  %156 = bitcast <4 x i32> %37 to <16 x i8>
>  %157 = call <4 x float> @llvm.SI.samplel.v4i32(<4 x i32> %154, <32 x i8> %155, <16 x i8> %156, i32 2)
>  br label %ENDIF
>
>ELSE:                                             ; preds = %main_body
>  %158 = bitcast float %42 to i32
>  %159 = bitcast float %43 to i32
>  %160 = insertelement <2 x i32> undef, i32 %158, i32 0
>  %161 = insertelement <2 x i32> %160, i32 %159, i32 1
>  %162 = bitcast <8 x i32> %35 to <32 x i8>
>  %163 = bitcast <4 x i32> %37 to <16 x i8>
>  %164 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %161, <32 x i8> %162, <16 x i8> %163, i32 2)
>  br label %ENDIF
>
>ENDIF:                                            ; preds = %ELSE, %IF
>  %.sink = phi <4 x float> [ %164, %ELSE ], [ %157, %IF ]
>  %165 = extractelement <4 x float> %.sink, i32 0
>  %166 = extractelement <4 x float> %.sink, i32 1
>  %167 = extractelement <4 x float> %.sink, i32 2
>  %168 = extractelement <4 x float> %.sink, i32 3
>  %169 = fmul float %165, %38
>  %170 = fmul float %166, %39
>  %171 = fmul float %167, %40
>  %172 = fmul float %168, %41
>  %173 = call i32 @llvm.SI.packf16(float %169, float %170)
>  %174 = bitcast i32 %173 to float
>  %175 = call i32 @llvm.SI.packf16(float %171, float %172)
>  %176 = bitcast i32 %175 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %174, float %176, float %174, float %176)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.fraction.(float) #2
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.samplel.v4i32(<4 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                         ; BEFE0A7E
>	s_mov_b32 m0, s9                                                             ; BEFC0309
>	v_interp_p1_f32 v7, v0, 1, 1, [m0]                                           ; C81C0500
>	v_interp_p2_f32 v7, [v7], v1, 1, 1, [m0]                                     ; C81D0501
>	v_interp_p1_f32 v6, v0, 0, 1, [m0]                                           ; C8180400
>	v_interp_p2_f32 v6, [v6], v1, 0, 1, [m0]                                     ; C8190401
>	v_interp_p1_f32 v2, v0, 3, 0, [m0]                                           ; C8080300
>	v_interp_p2_f32 v2, [v2], v1, 3, 0, [m0]                                     ; C8090301
>	v_interp_p1_f32 v3, v0, 2, 0, [m0]                                           ; C80C0200
>	v_interp_p2_f32 v3, [v3], v1, 2, 0, [m0]                                     ; C80D0201
>	v_interp_p1_f32 v4, v0, 1, 0, [m0]                                           ; C8100100
>	v_interp_p2_f32 v4, [v4], v1, 1, 0, [m0]                                     ; C8110101
>	v_interp_p1_f32 v5, v0, 0, 0, [m0]                                           ; C8140000
>	v_interp_p2_f32 v5, [v5], v1, 0, 0, [m0]                                     ; C8150001
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                                           ; C0800300
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	s_buffer_load_dword s8, s[0:3], 0x0                                          ; C2040100
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_cmp_gt_f32_e64 s[10:11], s8, 0.5                                           ; D008000A 0001E008
>	v_cndmask_b32_e64 v0, 0, -1, s[10:11]                                        ; D2000000 00298280
>	v_cmp_eq_i32_e64 s[10:11], v0, 0                                             ; D104000A 00010100
>	s_load_dwordx4 s[12:15], s[4:5], 0x4                                         ; C0860504
>	s_load_dwordx8 s[16:23], s[6:7], 0x8                                         ; C0C80708
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	s_and_saveexec_b64 s[10:11], s[10:11]                                        ; BE8A240A
>	s_xor_b64 s[10:11], exec, s[10:11]                                           ; 898A0A7E
>	image_sample v[8:11], 15, 0, 0, 0, 0, 0, 0, 0, v[6:7], s[16:23], s[12:15]    ; F0800F00 00640806
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	s_or_saveexec_b64 s[10:11], s[10:11]                                         ; BE8A250A
>	s_xor_b64 exec, exec, s[10:11]                                               ; 89FE0A7E
>	s_cbranch_execz BB0_4                                                        ; BF880000
>	s_buffer_load_dword s9, s[0:3], 0xd                                          ; C204810D
>	s_buffer_load_dword s24, s[0:3], 0xc                                         ; C20C010C
>	s_buffer_load_dword s25, s[0:3], 0x9                                         ; C20C8109
>	s_buffer_load_dword s26, s[0:3], 0x8                                         ; C20D0108
>	s_buffer_load_dword s27, s[0:3], 0x4                                         ; C20D8104
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v0, s25, v7                                                    ; 10000E19
>	v_fract_f32_e32 v0, v0                                                       ; 7E004100
>	v_mad_f32 v0, s25, v7, -v0                                                   ; D2820000 84020E19
>	v_add_f32_e32 v1, 0.5, v0                                                    ; 060200F0
>	v_rcp_f32_e32 v8, s25                                                        ; 7E105419
>	v_mul_f32_e32 v9, v8, v1                                                     ; 10120308
>	v_mul_f32_e32 v1, s26, v6                                                    ; 10020C1A
>	v_fract_f32_e32 v1, v1                                                       ; 7E024101
>	v_mad_f32 v1, s26, v6, -v1                                                   ; D2820001 84060C1A
>	v_add_f32_e32 v12, 0.5, v1                                                   ; 061802F0
>	v_rcp_f32_e32 v13, s26                                                       ; 7E1A541A
>	v_mul_f32_e32 v8, v13, v12                                                   ; 1010190D
>	v_mov_b32_e32 v10, 0                                                         ; 7E140280
>	s_load_dwordx4 s[28:31], s[4:5], 0x0                                         ; C08E0500
>	s_load_dwordx8 s[32:39], s[6:7], 0x0                                         ; C0D00700
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample_l v[12:13], 9, 0, 0, 0, 0, 0, 0, 0, v[8:11], s[32:39], s[28:31] ; F0900900 00E80C08
>	v_mov_b32_e32 v14, 0x437f0000                                                ; 7E1C02FF 437F0000
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v14, v13, v14, 0.5                                                 ; D282000E 03C21D0D
>	v_fract_f32_e32 v15, v14                                                     ; 7E1E410E
>	v_subrev_f32_e32 v14, v15, v14                                               ; 0A1C1D0F
>	v_mul_f32_e32 v15, 0x437f0000, v12                                           ; 101E18FF 437F0000
>	v_mov_b32_e32 v16, 0x43800000                                                ; 7E2002FF 43800000
>	v_mad_f32 v14, v16, v14, v15                                                 ; D282000E 043E1D10
>	v_add_f32_e32 v14, 0.5, v14                                                  ; 061C1CF0
>	v_fract_f32_e32 v15, v14                                                     ; 7E1E410E
>	v_subrev_f32_e32 v14, v15, v14                                               ; 0A1C1D0F
>	v_add_f32_e32 v14, 0.5, v14                                                  ; 061C1CF0
>	v_mad_f32 v12, s27, v12, 0.5                                                 ; D282000C 03C2181B
>	v_fract_f32_e32 v13, v12                                                     ; 7E1A410C
>	v_subrev_f32_e32 v12, v13, v12                                               ; 0A18190D
>	v_add_f32_e32 v12, 0.5, v12                                                  ; 061818F0
>	v_mov_b32_e32 v13, 0x44800000                                                ; 7E1A02FF 44800000
>	v_subrev_f32_e32 v13, s27, v13                                               ; 0A1A1A1B
>	v_cmp_ge_f32_e64 s[28:29], v13, 0                                            ; D00C001C 0001010D
>	v_cndmask_b32_e64 v13, 0, -1, s[28:29]                                       ; D200080D 00718280
>	v_cmp_ne_i32_e64 s[28:29], v13, 0                                            ; D10A001C 0001010D
>	v_cndmask_b32_e64 v12, v14, v12, s[28:29]                                    ; D200000C 1072190E
>	v_rcp_f32_e32 v13, s8                                                        ; 7E1A5408
>	v_mad_f32 v13, s24, v13, 0.5                                                 ; D282000D 03C21A18
>	v_fract_f32_e32 v14, v13                                                     ; 7E1C410D
>	v_subrev_f32_e32 v13, v14, v13                                               ; 0A1A1B0E
>	v_mul_f32_e32 v14, v12, v13                                                  ; 101C1B0C
>	v_cmp_ge_f32_e64 s[28:29], v14, 0                                            ; D00C001C 0001010E
>	v_cndmask_b32_e64 v14, 0, -1, s[28:29]                                       ; D200000E 00718280
>	v_cmp_ne_i32_e64 s[28:29], v14, 0                                            ; D10A001C 0001010E
>	v_xor_b32_e32 v14, 0x80000000, v13                                           ; 3A1C1AFF 80000000
>	v_cndmask_b32_e64 v14, v14, v13, s[28:29]                                    ; D200000E 10721B0E
>	v_rcp_f32_e32 v15, v14                                                       ; 7E1E550E
>	v_mul_f32_e32 v15, v12, v15                                                  ; 101E1F0C
>	v_fract_f32_e32 v15, v15                                                     ; 7E1E410F
>	v_mul_f32_e32 v16, v14, v15                                                  ; 10201F0E
>	v_fract_f32_e32 v16, v16                                                     ; 7E204110
>	v_mad_f32 v14, v15, v14, -v16                                                ; D282000E 84421D0F
>	v_mad_f32 v1, s26, v6, -v1                                                   ; D2820001 84060C1A
>	v_add_f32_e32 v1, v14, v1                                                    ; 0602030E
>	v_mul_f32_e32 v14, s8, v1                                                    ; 101C0208
>	v_fract_f32_e32 v14, v14                                                     ; 7E1C410E
>	v_mad_f32 v1, v1, s8, -v14                                                   ; D2820001 84381101
>	v_add_f32_e32 v1, 0.5, v1                                                    ; 060202F0
>	v_rcp_f32_e32 v14, s24                                                       ; 7E1C5418
>	v_mul_f32_e32 v8, v14, v1                                                    ; 1010030E
>	v_rcp_f32_e32 v1, v13                                                        ; 7E02550D
>	v_mul_f32_e32 v13, v12, v1                                                   ; 101A030C
>	v_fract_f32_e32 v13, v13                                                     ; 7E1A410D
>	v_mad_f32 v1, v1, v12, -v13                                                  ; D2820001 84361901
>	v_mad_f32 v0, s25, v7, -v0                                                   ; D2820000 84020E19
>	v_add_f32_e32 v0, v1, v0                                                     ; 06000101
>	v_mul_f32_e32 v1, s8, v0                                                     ; 10020008
>	v_fract_f32_e32 v1, v1                                                       ; 7E024101
>	v_mad_f32 v0, v0, s8, -v1                                                    ; D2820000 84041100
>	v_add_f32_e32 v0, 0.5, v0                                                    ; 060000F0
>	v_rcp_f32_e32 v1, s9                                                         ; 7E025409
>	v_mul_f32_e32 v9, v1, v0                                                     ; 10120101
>	image_sample_l v[8:11], 15, 0, 0, 0, 0, 0, 0, 0, v[8:11], s[16:23], s[12:15] ; F0900F00 00640808
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	s_or_b64 exec, exec, s[10:11]                                                ; 88FE0A7E
>	v_mul_f32_e32 v0, v2, v11                                                    ; 10001702
>	v_mul_f32_e32 v1, v3, v10                                                    ; 10021503
>	v_cvt_pkrtz_f16_f32_e32 v0, v1, v0                                           ; 5E000101
>	v_mul_f32_e32 v1, v4, v9                                                     ; 10021304
>	v_mul_f32_e32 v2, v5, v8                                                     ; 10041105
>	v_cvt_pkrtz_f16_f32_e32 v1, v2, v1                                           ; 5E020302
>	exp 15, 0, 1, 1, 1, v1, v0, v1, v0                                           ; F8001C0F 00010001
>	s_endpgm                                                                     ; BF810000
>
>SGPRS: 40
>VGPRS: 20
>Code Size: 620 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL TEMP[0..1], LOCAL
>IMM[0] FLT32 {    1.0000,     0.0000,     0.0000,     0.0000}
>  0: MAD TEMP[0], IN[0].xyzx, IMM[0].xxxy, IMM[0].yyyx
>  1: MOV TEMP[1].xy, IN[1].xyxx
>  2: MOV TEMP[1].zw, IMM[0].xxyx
>  3: MOV OUT[1], TEMP[1]
>  4: MOV OUT[0], TEMP[0]
>  5: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = add i32 %5, %7
>  %14 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %12, i32 0, i32 %13)
>  %15 = extractelement <4 x float> %14, i32 0
>  %16 = extractelement <4 x float> %14, i32 1
>  %17 = extractelement <4 x float> %14, i32 2
>  %18 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %19 = load <16 x i8> addrspace(2)* %18, !tbaa !0
>  %20 = add i32 %5, %7
>  %21 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %19, i32 0, i32 %20)
>  %22 = extractelement <4 x float> %21, i32 0
>  %23 = extractelement <4 x float> %21, i32 1
>  %24 = fmul float %15, 1.000000e+00
>  %25 = fadd float %24, 0.000000e+00
>  %26 = fmul float %16, 1.000000e+00
>  %27 = fadd float %26, 0.000000e+00
>  %28 = fmul float %17, 1.000000e+00
>  %29 = fadd float %28, 0.000000e+00
>  %30 = fmul float %15, 0.000000e+00
>  %31 = fadd float %30, 1.000000e+00
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %22, float %23, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %25, float %27, float %29, float %31)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                           ; 4A00000A
>	s_load_dwordx4 s[0:3], s[8:9], 0x0                  ; C0800900
>	s_load_dwordx4 s[4:7], s[8:9], 0x4                  ; C0820904
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[4:7], 0 idxen ; E00C2000 80010100
>	v_mov_b32_e32 v5, 1.0                               ; 7E0A02F2
>	v_mov_b32_e32 v6, 0                                 ; 7E0C0280
>	s_waitcnt vmcnt(0)                                  ; BF8C0770
>	exp 15, 32, 0, 0, 0, v1, v2, v6, v5                 ; F800020F 05060201
>	s_waitcnt expcnt(0)                                 ; BF8C070F
>	buffer_load_format_xyzw v[0:3], v0, s[0:3], 0 idxen ; E00C2000 80000000
>	s_waitcnt vmcnt(0)                                  ; BF8C0770
>	v_mad_f32 v4, 0, v0, 1.0                            ; D2820004 03CA0080
>	v_add_f32_e32 v5, 0, v2                             ; 060A0480
>	v_add_f32_e32 v6, 0, v1                             ; 060C0280
>	v_add_f32_e32 v0, 0, v0                             ; 06000080
>	exp 15, 12, 0, 1, 0, v0, v6, v5, v4                 ; F80008CF 04050600
>	s_endpgm                                            ; BF810000
>
>SGPRS: 16
>VGPRS: 8
>Code Size: 92 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL CONST[0..3]
>DCL TEMP[0..8], LOCAL
>IMM[0] FLT32 {    0.0000,     1.0000,    17.0000,  3721.0000}
>IMM[1] FLT32 {   13.0000,   930.2500,  1860.5000,    -0.5000}
>IMM[2] FLT32 {   -0.0000,    -1.0000,    -6.0000,     6.0000}
>IMM[3] FLT32 {   -2.0000,     3.0000,     0.5000,    -1.0000}
>IMM[4] FLT32 {   -1.0000,     1.0000,     0.0000,     0.5000}
>IMM[5] FLT32 {    1.0000,    -1.0000,    -2.0000,     2.0000}
>IMM[6] FLT32 {    0.0000,     0.0000,     0.0000,     0.0000}
>  0: RCP TEMP[0].x, CONST[2].xxxx
>  1: MUL TEMP[0], TEMP[0].xxxx, CONST[0].yyxx
>  2: FRC TEMP[1], TEMP[0].yyww
>  3: MOV TEMP[2].w, TEMP[1].wwww
>  4: ADD TEMP[0], TEMP[0], -TEMP[1]
>  5: RCP TEMP[2].x, TEMP[0].yyyy
>  6: MUL TEMP[1].yz, TEMP[0].xyww, IN[0].xyxw
>  7: MOV TEMP[2].yz, TEMP[1].zyzz
>  8: FRC TEMP[1].xy, TEMP[2].zyzw
>  9: MOV TEMP[3].xy, TEMP[1].xyxx
> 10: MAD TEMP[4], IN[0].yyxx, TEMP[0], -TEMP[1].yyxx
> 11: RCP TEMP[0].x, TEMP[0].wwww
> 12: ADD TEMP[4], TEMP[4], IMM[0].xyxy
> 13: MUL TEMP[5], TEMP[4].xyxy, IMM[0].zzzz
> 14: MUL TEMP[2], TEMP[2].xxxx, TEMP[5]
> 15: FRC TEMP[6], TEMP[2]
> 16: MUL TEMP[2], TEMP[6], TEMP[6]
> 17: MAD TEMP[2], TEMP[2], IMM[0].wwww, TEMP[4].zzww
> 18: MUL TEMP[2], TEMP[2], IMM[1].xxxx
> 19: MUL TEMP[0], TEMP[0].xxxx, TEMP[2]
> 20: FRC TEMP[6], TEMP[0]
> 21: MUL TEMP[0], TEMP[6], TEMP[6]
> 22: MUL TEMP[2], TEMP[0], IMM[1].yyyy
> 23: MUL TEMP[0], TEMP[0], IMM[1].zzzz
> 24: FRC TEMP[6], TEMP[0]
> 25: ADD TEMP[0], TEMP[6], IMM[1].wwww
> 26: FRC TEMP[6], TEMP[2]
> 27: ADD TEMP[2], TEMP[6], IMM[1].wwww
> 28: MOV TEMP[6], -TEMP[2]
> 29: FSGE TEMP[7].x, TEMP[6].xxxx, IMM[0].xxxx
> 30: UIF TEMP[7].xxxx :0
> 31:   MOV TEMP[7].x, IMM[0].xxxx
> 32: ELSE :0
> 33:   MOV TEMP[7].x, IMM[0].yyyy
> 34: ENDIF
> 35: MOV TEMP[7].x, TEMP[7].xxxx
> 36: FSGE TEMP[8].x, TEMP[6].yyyy, IMM[0].xxxx
> 37: UIF TEMP[8].xxxx :0
> 38:   MOV TEMP[8].x, IMM[0].xxxx
> 39: ELSE :0
> 40:   MOV TEMP[8].x, IMM[0].yyyy
> 41: ENDIF
> 42: MOV TEMP[7].y, TEMP[8].xxxx
> 43: FSGE TEMP[8].x, TEMP[6].zzzz, IMM[0].xxxx
> 44: UIF TEMP[8].xxxx :0
> 45:   MOV TEMP[8].x, IMM[0].xxxx
> 46: ELSE :0
> 47:   MOV TEMP[8].x, IMM[0].yyyy
> 48: ENDIF
> 49: MOV TEMP[7].z, TEMP[8].xxxx
> 50: FSGE TEMP[6].x, TEMP[6].wwww, IMM[0].xxxx
> 51: UIF TEMP[6].xxxx :0
> 52:   MOV TEMP[6].x, IMM[0].xxxx
> 53: ELSE :0
> 54:   MOV TEMP[6].x, IMM[0].yyyy
> 55: ENDIF
> 56: MOV TEMP[7].w, TEMP[6].xxxx
> 57: FSGE TEMP[6].x, TEMP[2].xxxx, IMM[0].xxxx
> 58: UIF TEMP[6].xxxx :0
> 59:   MOV TEMP[6].x, IMM[2].xxxx
> 60: ELSE :0
> 61:   MOV TEMP[6].x, IMM[2].yyyy
> 62: ENDIF
> 63: MOV TEMP[6].x, TEMP[6].xxxx
> 64: FSGE TEMP[8].x, TEMP[2].yyyy, IMM[0].xxxx
> 65: UIF TEMP[8].xxxx :0
> 66:   MOV TEMP[8].x, IMM[2].xxxx
> 67: ELSE :0
> 68:   MOV TEMP[8].x, IMM[2].yyyy
> 69: ENDIF
> 70: MOV TEMP[6].y, TEMP[8].xxxx
> 71: FSGE TEMP[8].x, TEMP[2].zzzz, IMM[0].xxxx
> 72: UIF TEMP[8].xxxx :0
> 73:   MOV TEMP[8].x, IMM[2].xxxx
> 74: ELSE :0
> 75:   MOV TEMP[8].x, IMM[2].yyyy
> 76: ENDIF
> 77: MOV TEMP[6].z, TEMP[8].xxxx
> 78: FSGE TEMP[8].x, TEMP[2].wwww, IMM[0].xxxx
> 79: UIF TEMP[8].xxxx :0
> 80:   MOV TEMP[8].x, IMM[2].xxxx
> 81: ELSE :0
> 82:   MOV TEMP[8].x, IMM[2].yyyy
> 83: ENDIF
> 84: MOV TEMP[6].w, TEMP[8].xxxx
> 85: ADD TEMP[2], TEMP[6], TEMP[7]
> 86: ADD TEMP[4], TEMP[1].xxyy, IMM[2].xyxy
> 87: MUL TEMP[5], TEMP[2], TEMP[4].zwzw
> 88: MOV TEMP[6], -TEMP[0]
> 89: FSGE TEMP[7].x, TEMP[6].xxxx, IMM[0].xxxx
> 90: UIF TEMP[7].xxxx :0
> 91:   MOV TEMP[7].x, IMM[0].xxxx
> 92: ELSE :0
> 93:   MOV TEMP[7].x, IMM[0].yyyy
> 94: ENDIF
> 95: MOV TEMP[7].x, TEMP[7].xxxx
> 96: FSGE TEMP[8].x, TEMP[6].yyyy, IMM[0].xxxx
> 97: UIF TEMP[8].xxxx :0
> 98:   MOV TEMP[8].x, IMM[0].xxxx
> 99: ELSE :0
>100:   MOV TEMP[8].x, IMM[0].yyyy
>101: ENDIF
>102: MOV TEMP[7].y, TEMP[8].xxxx
>103: FSGE TEMP[8].x, TEMP[6].zzzz, IMM[0].xxxx
>104: UIF TEMP[8].xxxx :0
>105:   MOV TEMP[8].x, IMM[0].xxxx
>106: ELSE :0
>107:   MOV TEMP[8].x, IMM[0].yyyy
>108: ENDIF
>109: MOV TEMP[7].z, TEMP[8].xxxx
>110: FSGE TEMP[6].x, TEMP[6].wwww, IMM[0].xxxx
>111: UIF TEMP[6].xxxx :0
>112:   MOV TEMP[6].x, IMM[0].xxxx
>113: ELSE :0
>114:   MOV TEMP[6].x, IMM[0].yyyy
>115: ENDIF
>116: MOV TEMP[7].w, TEMP[6].xxxx
>117: FSGE TEMP[6].x, TEMP[0].xxxx, IMM[0].xxxx
>118: UIF TEMP[6].xxxx :0
>119:   MOV TEMP[6].x, IMM[2].xxxx
>120: ELSE :0
>121:   MOV TEMP[6].x, IMM[2].yyyy
>122: ENDIF
>123: MOV TEMP[6].x, TEMP[6].xxxx
>124: FSGE TEMP[8].x, TEMP[0].yyyy, IMM[0].xxxx
>125: UIF TEMP[8].xxxx :0
>126:   MOV TEMP[8].x, IMM[2].xxxx
>127: ELSE :0
>128:   MOV TEMP[8].x, IMM[2].yyyy
>129: ENDIF
>130: MOV TEMP[6].y, TEMP[8].xxxx
>131: FSGE TEMP[8].x, TEMP[0].zzzz, IMM[0].xxxx
>132: UIF TEMP[8].xxxx :0
>133:   MOV TEMP[8].x, IMM[2].xxxx
>134: ELSE :0
>135:   MOV TEMP[8].x, IMM[2].yyyy
>136: ENDIF
>137: MOV TEMP[6].z, TEMP[8].xxxx
>138: FSGE TEMP[8].x, TEMP[0].wwww, IMM[0].xxxx
>139: UIF TEMP[8].xxxx :0
>140:   MOV TEMP[8].x, IMM[2].xxxx
>141: ELSE :0
>142:   MOV TEMP[8].x, IMM[2].yyyy
>143: ENDIF
>144: MOV TEMP[6].w, TEMP[8].xxxx
>145: ADD TEMP[0], TEMP[6], TEMP[7]
>146: MAD TEMP[4], TEMP[4].xxyy, TEMP[0], TEMP[5]
>147: ADD TEMP[4], -TEMP[4].xyxz, TEMP[4].zwyw
>148: MAD TEMP[6].zw, TEMP[1].xyxy, IMM[2].zzzz, IMM[2].wwww
>149: MOV TEMP[3].zw, TEMP[6].wwzw
>150: MUL TEMP[5], TEMP[1].xyxy, TEMP[3]
>151: MAD TEMP[3], TEMP[1].xyxx, IMM[3].xxxx, IMM[3].yyyy
>152: MUL TEMP[3], TEMP[3], TEMP[5].xyxx
>153: LRP TEMP[6].xy, TEMP[3].wwww, TEMP[0].zwzw, TEMP[0]
>154: MOV TEMP[1].xy, TEMP[6].xyxx
>155: LRP TEMP[6].zw, TEMP[3].yyyy, TEMP[2].xyyw, TEMP[2].xyxz
>156: MOV TEMP[1].zw, TEMP[6].wwzw
>157: MAD TEMP[0], TEMP[4], TEMP[5].zzww, TEMP[1]
>158: LRP TEMP[1], TEMP[3], TEMP[0].wyww, TEMP[0].zxzz
>159: MOV TEMP[2].z, TEMP[1].zyzz
>160: MUL TEMP[0], TEMP[1], CONST[1].xxxx
>161: MUL TEMP[0], TEMP[0], IMM[3].zzzz
>162: MAD TEMP[0], TEMP[0], IMM[4].xyzz, IMM[4].wwzy
>163: MOV TEMP[2].w, IMM[0].xxxx
>164: RCP TEMP[2].x, CONST[0].xxxx
>165: ADD TEMP[1].xy, TEMP[2].xwzw, IN[0]
>166: MOV TEMP[1].xy, TEMP[1].xyyy
>167: TEX TEMP[1].xw, TEMP[1], SAMP[0], 2D
>168: MOV TEMP[3].x, TEMP[1].xxxw
>169: MOV TEMP[2].y, -TEMP[2].xxxx
>170: MOV TEMP[4].yw, TEMP[2].xyxx
>171: RCP TEMP[4].x, CONST[0].yyyy
>172: ADD TEMP[1].xy, TEMP[4].yxzw, IN[0]
>173: ADD TEMP[6].yw, TEMP[4].xwzx, IN[0].xxzy
>174: MOV TEMP[6].xy, TEMP[6].ywww
>175: TEX TEMP[6].x, TEMP[6], SAMP[0], 2D
>176: MOV TEMP[3].z, TEMP[6].xxxx
>177: MOV TEMP[1].xy, TEMP[1].xyyy
>178: TEX TEMP[1].x, TEMP[1], SAMP[0], 2D
>179: MOV TEMP[3].y, TEMP[1].xxxx
>180: DP3 TEMP[1].x, TEMP[3].yzxx, IMM[5].xyzz
>181: ADD TEMP[6].yw, TEMP[2], IN[0].xxzy
>182: MOV TEMP[6].xy, TEMP[6].ywww
>183: TEX TEMP[6].x, TEMP[6], SAMP[0], 2D
>184: MOV TEMP[5].z, TEMP[6].xxxx
>185: MOV TEMP[2].z, -TEMP[4].xxxx
>186: ADD TEMP[6].yw, TEMP[2].xxzz, IN[0].xxzy
>187: MOV TEMP[4].yw, TEMP[6].wyww
>188: MOV TEMP[6].xy, TEMP[6].ywww
>189: TEX TEMP[6].x, TEMP[6], SAMP[0], 2D
>190: MOV TEMP[5].y, TEMP[6].xxxx
>191: ADD TEMP[6].xy, TEMP[2].yzzw, IN[0]
>192: MOV TEMP[2].xy, TEMP[6].xyxx
>193: ADD TEMP[7].zw, TEMP[2].xywz, IN[0].xyxy
>194: MOV TEMP[7].xy, TEMP[7].zwww
>195: TEX TEMP[7].x, TEMP[7], SAMP[0], 2D
>196: MOV TEMP[5].w, TEMP[7].xxxx
>197: MOV TEMP[6].xy, TEMP[6].xyyy
>198: TEX TEMP[6].xw, TEMP[6], SAMP[0], 2D
>199: MOV TEMP[2].w, TEMP[6].wwww
>200: MOV TEMP[5].x, TEMP[6].xxxx
>201: DP3 TEMP[6].x, TEMP[5].xyww, IMM[5].xxww
>202: DP3 TEMP[7].x, TEMP[5].xyzz, IMM[5].xyww
>203: ADD TEMP[1].x, TEMP[1].xxxx, TEMP[7].xxxx
>204: MOV TEMP[5].x, TEMP[1].xxxx
>205: MOV TEMP[4].z, IMM[0].xxxx
>206: ADD TEMP[1].xz, TEMP[4].zyxw, IN[0].xyyw
>207: MOV TEMP[1].xy, TEMP[1].xzzz
>208: TEX TEMP[1].x, TEMP[1], SAMP[0], 2D
>209: MOV TEMP[3].w, TEMP[1].xxxx
>210: DP3 TEMP[1].x, TEMP[3].yzww, IMM[3].wwxx
>211: ADD TEMP[1].y, TEMP[1].xxxx, TEMP[6].xxxx
>212: MOV TEMP[5].y, TEMP[1].yyyy
>213: MUL TEMP[1].xy, TEMP[5], CONST[3].xxxx
>214: MOV TEMP[2].xy, TEMP[1].xyxx
>215: MUL TEMP[4].xyz, TEMP[2].yxyw, IMM[3].zzzz
>216: DP2 TEMP[2].x, TEMP[1].xyyy, TEMP[1].xyyy
>217: MAX TEMP[1].x, TEMP[2].xxxx, IMM[6].xxxx
>218: SQRT TEMP[2].x, TEMP[1].xxxx
>219: MAD TEMP[1].yzw, TEMP[4].xxyz, IMM[4].xxyz, IMM[4].xwwz
>220: MOV TEMP[2].yzw, TEMP[1].zyzw
>221: ADD TEMP[3], -TEMP[0], TEMP[2].yzwx
>222: MOV_SAT TEMP[1].x, TEMP[2].xxxx
>223: MAD TEMP[0], TEMP[1].xxxx, TEMP[3], TEMP[0]
>224: MOV OUT[0], TEMP[0]
>225: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %23 = load <16 x i8> addrspace(2)* %22, !tbaa !0
>  %24 = call float @llvm.SI.load.const(<16 x i8> %23, i32 0)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %23, i32 4)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %23, i32 16)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %23, i32 32)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %23, i32 48)
>  %29 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %30 = load <8 x i32> addrspace(2)* %29, !tbaa !0
>  %31 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %32 = load <4 x i32> addrspace(2)* %31, !tbaa !0
>  %33 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %34 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %35 = fdiv float 1.000000e+00, %27
>  %36 = fmul float %35, %25
>  %37 = fmul float %35, %25
>  %38 = fmul float %35, %24
>  %39 = fmul float %35, %24
>  %40 = call float @llvm.AMDIL.fraction.(float %37)
>  %41 = call float @llvm.AMDIL.fraction.(float %37)
>  %42 = call float @llvm.AMDIL.fraction.(float %39)
>  %43 = call float @llvm.AMDIL.fraction.(float %39)
>  %44 = fsub float -0.000000e+00, %40
>  %45 = fadd float %36, %44
>  %46 = fsub float -0.000000e+00, %41
>  %47 = fadd float %37, %46
>  %48 = fsub float -0.000000e+00, %42
>  %49 = fadd float %38, %48
>  %50 = fsub float -0.000000e+00, %43
>  %51 = fadd float %39, %50
>  %52 = fdiv float 1.000000e+00, %47
>  %53 = fmul float %47, %34
>  %54 = fmul float %51, %33
>  %55 = call float @llvm.AMDIL.fraction.(float %54)
>  %56 = call float @llvm.AMDIL.fraction.(float %53)
>  %57 = fsub float -0.000000e+00, %56
>  %58 = fmul float %34, %45
>  %59 = fadd float %58, %57
>  %60 = fsub float -0.000000e+00, %56
>  %61 = fmul float %34, %47
>  %62 = fadd float %61, %60
>  %63 = fsub float -0.000000e+00, %55
>  %64 = fmul float %33, %49
>  %65 = fadd float %64, %63
>  %66 = fsub float -0.000000e+00, %55
>  %67 = fmul float %33, %51
>  %68 = fadd float %67, %66
>  %69 = fdiv float 1.000000e+00, %51
>  %70 = fadd float %59, 0.000000e+00
>  %71 = fadd float %62, 1.000000e+00
>  %72 = fadd float %65, 0.000000e+00
>  %73 = fadd float %68, 1.000000e+00
>  %74 = fmul float %70, 1.700000e+01
>  %75 = fmul float %71, 1.700000e+01
>  %76 = fmul float %70, 1.700000e+01
>  %77 = fmul float %71, 1.700000e+01
>  %78 = fmul float %52, %74
>  %79 = fmul float %52, %75
>  %80 = fmul float %52, %76
>  %81 = fmul float %52, %77
>  %82 = call float @llvm.AMDIL.fraction.(float %78)
>  %83 = call float @llvm.AMDIL.fraction.(float %79)
>  %84 = call float @llvm.AMDIL.fraction.(float %80)
>  %85 = call float @llvm.AMDIL.fraction.(float %81)
>  %86 = fmul float %82, %82
>  %87 = fmul float %83, %83
>  %88 = fmul float %84, %84
>  %89 = fmul float %85, %85
>  %90 = fmul float %86, 3.721000e+03
>  %91 = fadd float %90, %72
>  %92 = fmul float %87, 3.721000e+03
>  %93 = fadd float %92, %72
>  %94 = fmul float %88, 3.721000e+03
>  %95 = fadd float %94, %73
>  %96 = fmul float %89, 3.721000e+03
>  %97 = fadd float %96, %73
>  %98 = fmul float %91, 1.300000e+01
>  %99 = fmul float %93, 1.300000e+01
>  %100 = fmul float %95, 1.300000e+01
>  %101 = fmul float %97, 1.300000e+01
>  %102 = fmul float %69, %98
>  %103 = fmul float %69, %99
>  %104 = fmul float %69, %100
>  %105 = fmul float %69, %101
>  %106 = call float @llvm.AMDIL.fraction.(float %102)
>  %107 = call float @llvm.AMDIL.fraction.(float %103)
>  %108 = call float @llvm.AMDIL.fraction.(float %104)
>  %109 = call float @llvm.AMDIL.fraction.(float %105)
>  %110 = fmul float %106, %106
>  %111 = fmul float %107, %107
>  %112 = fmul float %108, %108
>  %113 = fmul float %109, %109
>  %114 = fmul float %110, 9.302500e+02
>  %115 = fmul float %111, 9.302500e+02
>  %116 = fmul float %112, 9.302500e+02
>  %117 = fmul float %113, 9.302500e+02
>  %118 = fmul float %110, 1.860500e+03
>  %119 = fmul float %111, 1.860500e+03
>  %120 = fmul float %112, 1.860500e+03
>  %121 = fmul float %113, 1.860500e+03
>  %122 = call float @llvm.AMDIL.fraction.(float %118)
>  %123 = call float @llvm.AMDIL.fraction.(float %119)
>  %124 = call float @llvm.AMDIL.fraction.(float %120)
>  %125 = call float @llvm.AMDIL.fraction.(float %121)
>  %126 = fadd float %122, -5.000000e-01
>  %127 = fadd float %123, -5.000000e-01
>  %128 = fadd float %124, -5.000000e-01
>  %129 = fadd float %125, -5.000000e-01
>  %130 = call float @llvm.AMDIL.fraction.(float %114)
>  %131 = call float @llvm.AMDIL.fraction.(float %115)
>  %132 = call float @llvm.AMDIL.fraction.(float %116)
>  %133 = call float @llvm.AMDIL.fraction.(float %117)
>  %134 = fadd float %130, -5.000000e-01
>  %135 = fadd float %131, -5.000000e-01
>  %136 = fadd float %132, -5.000000e-01
>  %137 = fadd float %133, -5.000000e-01
>  %138 = fsub float -0.000000e+00, %134
>  %139 = fsub float -0.000000e+00, %135
>  %140 = fsub float -0.000000e+00, %136
>  %141 = fsub float -0.000000e+00, %137
>  %142 = fcmp oge float %138, 0.000000e+00
>  %143 = sext i1 %142 to i32
>  %144 = bitcast i32 %143 to float
>  %145 = bitcast float %144 to i32
>  %146 = icmp ne i32 %145, 0
>  %. = select i1 %146, float 0.000000e+00, float 1.000000e+00
>  %147 = fcmp oge float %139, 0.000000e+00
>  %148 = sext i1 %147 to i32
>  %149 = bitcast i32 %148 to float
>  %150 = bitcast float %149 to i32
>  %151 = icmp ne i32 %150, 0
>  %temp32.0 = select i1 %151, float 0.000000e+00, float 1.000000e+00
>  %152 = fcmp oge float %140, 0.000000e+00
>  %153 = sext i1 %152 to i32
>  %154 = bitcast i32 %153 to float
>  %155 = bitcast float %154 to i32
>  %156 = icmp ne i32 %155, 0
>  %.81 = select i1 %156, float 0.000000e+00, float 1.000000e+00
>  %157 = fcmp oge float %141, 0.000000e+00
>  %158 = sext i1 %157 to i32
>  %159 = bitcast i32 %158 to float
>  %160 = bitcast float %159 to i32
>  %161 = icmp ne i32 %160, 0
>  %temp24.0 = select i1 %161, float 0.000000e+00, float 1.000000e+00
>  %162 = fcmp oge float %134, 0.000000e+00
>  %163 = sext i1 %162 to i32
>  %164 = bitcast i32 %163 to float
>  %165 = bitcast float %164 to i32
>  %166 = icmp ne i32 %165, 0
>  %.82 = select i1 %166, float -0.000000e+00, float -1.000000e+00
>  %167 = fcmp oge float %135, 0.000000e+00
>  %168 = sext i1 %167 to i32
>  %169 = bitcast i32 %168 to float
>  %170 = bitcast float %169 to i32
>  %171 = icmp ne i32 %170, 0
>  %temp32.2 = select i1 %171, float -0.000000e+00, float -1.000000e+00
>  %172 = fcmp oge float %136, 0.000000e+00
>  %173 = sext i1 %172 to i32
>  %174 = bitcast i32 %173 to float
>  %175 = bitcast float %174 to i32
>  %176 = icmp ne i32 %175, 0
>  %.83 = select i1 %176, float -0.000000e+00, float -1.000000e+00
>  %177 = fcmp oge float %137, 0.000000e+00
>  %178 = sext i1 %177 to i32
>  %179 = bitcast i32 %178 to float
>  %180 = bitcast float %179 to i32
>  %181 = icmp ne i32 %180, 0
>  %temp32.4 = select i1 %181, float -0.000000e+00, float -1.000000e+00
>  %182 = fadd float %.82, %.
>  %183 = fadd float %temp32.2, %temp32.0
>  %184 = fadd float %.83, %.81
>  %185 = fadd float %temp32.4, %temp24.0
>  %186 = fadd float %55, -0.000000e+00
>  %187 = fadd float %55, -1.000000e+00
>  %188 = fadd float %56, -0.000000e+00
>  %189 = fadd float %56, -1.000000e+00
>  %190 = fmul float %182, %188
>  %191 = fmul float %183, %189
>  %192 = fmul float %184, %188
>  %193 = fmul float %185, %189
>  %194 = fsub float -0.000000e+00, %126
>  %195 = fsub float -0.000000e+00, %127
>  %196 = fsub float -0.000000e+00, %128
>  %197 = fsub float -0.000000e+00, %129
>  %198 = fcmp oge float %194, 0.000000e+00
>  %199 = sext i1 %198 to i32
>  %200 = bitcast i32 %199 to float
>  %201 = bitcast float %200 to i32
>  %202 = icmp ne i32 %201, 0
>  %.84 = select i1 %202, float 0.000000e+00, float 1.000000e+00
>  %203 = fcmp oge float %195, 0.000000e+00
>  %204 = sext i1 %203 to i32
>  %205 = bitcast i32 %204 to float
>  %206 = bitcast float %205 to i32
>  %207 = icmp ne i32 %206, 0
>  %temp32.5 = select i1 %207, float 0.000000e+00, float 1.000000e+00
>  %208 = fcmp oge float %196, 0.000000e+00
>  %209 = sext i1 %208 to i32
>  %210 = bitcast i32 %209 to float
>  %211 = bitcast float %210 to i32
>  %212 = icmp ne i32 %211, 0
>  %.85 = select i1 %212, float 0.000000e+00, float 1.000000e+00
>  %213 = fcmp oge float %197, 0.000000e+00
>  %214 = sext i1 %213 to i32
>  %215 = bitcast i32 %214 to float
>  %216 = bitcast float %215 to i32
>  %217 = icmp ne i32 %216, 0
>  %temp24.2 = select i1 %217, float 0.000000e+00, float 1.000000e+00
>  %218 = fcmp oge float %126, 0.000000e+00
>  %219 = sext i1 %218 to i32
>  %220 = bitcast i32 %219 to float
>  %221 = bitcast float %220 to i32
>  %222 = icmp ne i32 %221, 0
>  %.86 = select i1 %222, float -0.000000e+00, float -1.000000e+00
>  %223 = fcmp oge float %127, 0.000000e+00
>  %224 = sext i1 %223 to i32
>  %225 = bitcast i32 %224 to float
>  %226 = bitcast float %225 to i32
>  %227 = icmp ne i32 %226, 0
>  %temp32.7 = select i1 %227, float -0.000000e+00, float -1.000000e+00
>  %228 = fcmp oge float %128, 0.000000e+00
>  %229 = sext i1 %228 to i32
>  %230 = bitcast i32 %229 to float
>  %231 = bitcast float %230 to i32
>  %232 = icmp ne i32 %231, 0
>  %.87 = select i1 %232, float -0.000000e+00, float -1.000000e+00
>  %233 = fcmp oge float %129, 0.000000e+00
>  %234 = sext i1 %233 to i32
>  %235 = bitcast i32 %234 to float
>  %236 = bitcast float %235 to i32
>  %237 = icmp ne i32 %236, 0
>  %temp32.9 = select i1 %237, float -0.000000e+00, float -1.000000e+00
>  %238 = fadd float %.86, %.84
>  %239 = fadd float %temp32.7, %temp32.5
>  %240 = fadd float %.87, %.85
>  %241 = fadd float %temp32.9, %temp24.2
>  %242 = fmul float %186, %238
>  %243 = fadd float %242, %190
>  %244 = fmul float %186, %239
>  %245 = fadd float %244, %191
>  %246 = fmul float %187, %240
>  %247 = fadd float %246, %192
>  %248 = fmul float %187, %241
>  %249 = fadd float %248, %193
>  %250 = fsub float -0.000000e+00, %243
>  %251 = fadd float %250, %247
>  %252 = fsub float -0.000000e+00, %245
>  %253 = fadd float %252, %249
>  %254 = fsub float -0.000000e+00, %243
>  %255 = fadd float %254, %245
>  %256 = fsub float -0.000000e+00, %247
>  %257 = fadd float %256, %249
>  %258 = fmul float %55, -6.000000e+00
>  %259 = fadd float %258, 6.000000e+00
>  %260 = fmul float %56, -6.000000e+00
>  %261 = fadd float %260, 6.000000e+00
>  %262 = fmul float %55, %55
>  %263 = fmul float %56, %56
>  %264 = fmul float %55, %259
>  %265 = fmul float %56, %261
>  %266 = fmul float %55, -2.000000e+00
>  %267 = fadd float %266, 3.000000e+00
>  %268 = fmul float %56, -2.000000e+00
>  %269 = fadd float %268, 3.000000e+00
>  %270 = fmul float %55, -2.000000e+00
>  %271 = fadd float %270, 3.000000e+00
>  %272 = fmul float %55, -2.000000e+00
>  %273 = fadd float %272, 3.000000e+00
>  %274 = fmul float %267, %262
>  %275 = fmul float %269, %263
>  %276 = fmul float %271, %262
>  %277 = fmul float %273, %262
>  %278 = call float @llvm.AMDGPU.lrp(float %277, float %240, float %238)
>  %279 = call float @llvm.AMDGPU.lrp(float %277, float %241, float %239)
>  %280 = call float @llvm.AMDGPU.lrp(float %275, float %183, float %182)
>  %281 = call float @llvm.AMDGPU.lrp(float %275, float %185, float %184)
>  %282 = fmul float %251, %264
>  %283 = fadd float %282, %278
>  %284 = fmul float %253, %264
>  %285 = fadd float %284, %279
>  %286 = fmul float %255, %265
>  %287 = fadd float %286, %280
>  %288 = fmul float %257, %265
>  %289 = fadd float %288, %281
>  %290 = call float @llvm.AMDGPU.lrp(float %274, float %289, float %287)
>  %291 = call float @llvm.AMDGPU.lrp(float %275, float %285, float %283)
>  %292 = call float @llvm.AMDGPU.lrp(float %276, float %289, float %287)
>  %293 = call float @llvm.AMDGPU.lrp(float %277, float %289, float %287)
>  %294 = fmul float %290, %26
>  %295 = fmul float %291, %26
>  %296 = fmul float %292, %26
>  %297 = fmul float %293, %26
>  %298 = fmul float %294, 5.000000e-01
>  %299 = fmul float %295, 5.000000e-01
>  %300 = fmul float %296, 5.000000e-01
>  %301 = fmul float %297, 5.000000e-01
>  %302 = fmul float %298, -1.000000e+00
>  %303 = fadd float %302, 5.000000e-01
>  %304 = fmul float %299, 1.000000e+00
>  %305 = fadd float %304, 5.000000e-01
>  %306 = fmul float %300, 0.000000e+00
>  %307 = fadd float %306, 0.000000e+00
>  %308 = fmul float %301, 0.000000e+00
>  %309 = fadd float %308, 1.000000e+00
>  %310 = fdiv float 1.000000e+00, %24
>  %311 = fadd float %310, %33
>  %312 = fadd float 0.000000e+00, %34
>  %313 = bitcast float %311 to i32
>  %314 = bitcast float %312 to i32
>  %315 = insertelement <2 x i32> undef, i32 %313, i32 0
>  %316 = insertelement <2 x i32> %315, i32 %314, i32 1
>  %317 = bitcast <8 x i32> %30 to <32 x i8>
>  %318 = bitcast <4 x i32> %32 to <16 x i8>
>  %319 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %316, <32 x i8> %317, <16 x i8> %318, i32 2)
>  %320 = extractelement <4 x float> %319, i32 0
>  %321 = fsub float -0.000000e+00, %310
>  %322 = fdiv float 1.000000e+00, %25
>  %323 = fadd float %321, %33
>  %324 = fadd float %322, %34
>  %325 = fadd float %310, %33
>  %326 = fadd float %322, %34
>  %327 = bitcast float %325 to i32
>  %328 = bitcast float %326 to i32
>  %329 = insertelement <2 x i32> undef, i32 %327, i32 0
>  %330 = insertelement <2 x i32> %329, i32 %328, i32 1
>  %331 = bitcast <8 x i32> %30 to <32 x i8>
>  %332 = bitcast <4 x i32> %32 to <16 x i8>
>  %333 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %330, <32 x i8> %331, <16 x i8> %332, i32 2)
>  %334 = extractelement <4 x float> %333, i32 0
>  %335 = bitcast float %323 to i32
>  %336 = bitcast float %324 to i32
>  %337 = insertelement <2 x i32> undef, i32 %335, i32 0
>  %338 = insertelement <2 x i32> %337, i32 %336, i32 1
>  %339 = bitcast <8 x i32> %30 to <32 x i8>
>  %340 = bitcast <4 x i32> %32 to <16 x i8>
>  %341 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %338, <32 x i8> %339, <16 x i8> %340, i32 2)
>  %342 = extractelement <4 x float> %341, i32 0
>  %343 = fmul float %342, 1.000000e+00
>  %344 = fmul float %334, -1.000000e+00
>  %345 = fadd float %344, %343
>  %346 = fmul float %320, -2.000000e+00
>  %347 = fadd float %345, %346
>  %348 = fadd float %321, %33
>  %349 = fadd float 0.000000e+00, %34
>  %350 = bitcast float %348 to i32
>  %351 = bitcast float %349 to i32
>  %352 = insertelement <2 x i32> undef, i32 %350, i32 0
>  %353 = insertelement <2 x i32> %352, i32 %351, i32 1
>  %354 = bitcast <8 x i32> %30 to <32 x i8>
>  %355 = bitcast <4 x i32> %32 to <16 x i8>
>  %356 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %353, <32 x i8> %354, <16 x i8> %355, i32 2)
>  %357 = extractelement <4 x float> %356, i32 0
>  %358 = fsub float -0.000000e+00, %322
>  %359 = fadd float %310, %33
>  %360 = fadd float %358, %34
>  %361 = bitcast float %359 to i32
>  %362 = bitcast float %360 to i32
>  %363 = insertelement <2 x i32> undef, i32 %361, i32 0
>  %364 = insertelement <2 x i32> %363, i32 %362, i32 1
>  %365 = bitcast <8 x i32> %30 to <32 x i8>
>  %366 = bitcast <4 x i32> %32 to <16 x i8>
>  %367 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %364, <32 x i8> %365, <16 x i8> %366, i32 2)
>  %368 = extractelement <4 x float> %367, i32 0
>  %369 = fadd float %321, %33
>  %370 = fadd float %358, %34
>  %371 = fadd float 0.000000e+00, %33
>  %372 = fadd float %358, %34
>  %373 = bitcast float %371 to i32
>  %374 = bitcast float %372 to i32
>  %375 = insertelement <2 x i32> undef, i32 %373, i32 0
>  %376 = insertelement <2 x i32> %375, i32 %374, i32 1
>  %377 = bitcast <8 x i32> %30 to <32 x i8>
>  %378 = bitcast <4 x i32> %32 to <16 x i8>
>  %379 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %376, <32 x i8> %377, <16 x i8> %378, i32 2)
>  %380 = extractelement <4 x float> %379, i32 0
>  %381 = bitcast float %369 to i32
>  %382 = bitcast float %370 to i32
>  %383 = insertelement <2 x i32> undef, i32 %381, i32 0
>  %384 = insertelement <2 x i32> %383, i32 %382, i32 1
>  %385 = bitcast <8 x i32> %30 to <32 x i8>
>  %386 = bitcast <4 x i32> %32 to <16 x i8>
>  %387 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %384, <32 x i8> %385, <16 x i8> %386, i32 2)
>  %388 = extractelement <4 x float> %387, i32 0
>  %389 = fmul float %388, 1.000000e+00
>  %390 = fmul float %368, 1.000000e+00
>  %391 = fadd float %390, %389
>  %392 = fmul float %380, 2.000000e+00
>  %393 = fadd float %391, %392
>  %394 = fmul float %388, 1.000000e+00
>  %395 = fmul float %368, -1.000000e+00
>  %396 = fadd float %395, %394
>  %397 = fmul float %357, 2.000000e+00
>  %398 = fadd float %396, %397
>  %399 = fadd float %347, %398
>  %400 = fadd float 0.000000e+00, %33
>  %401 = fadd float %322, %34
>  %402 = bitcast float %400 to i32
>  %403 = bitcast float %401 to i32
>  %404 = insertelement <2 x i32> undef, i32 %402, i32 0
>  %405 = insertelement <2 x i32> %404, i32 %403, i32 1
>  %406 = bitcast <8 x i32> %30 to <32 x i8>
>  %407 = bitcast <4 x i32> %32 to <16 x i8>
>  %408 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %405, <32 x i8> %406, <16 x i8> %407, i32 2)
>  %409 = extractelement <4 x float> %408, i32 0
>  %410 = fmul float %342, -1.000000e+00
>  %411 = fmul float %334, -1.000000e+00
>  %412 = fadd float %411, %410
>  %413 = fmul float %409, -2.000000e+00
>  %414 = fadd float %412, %413
>  %415 = fadd float %414, %393
>  %416 = fmul float %399, %28
>  %417 = fmul float %415, %28
>  %418 = fmul float %417, 5.000000e-01
>  %419 = fmul float %416, 5.000000e-01
>  %420 = fmul float %417, 5.000000e-01
>  %421 = fmul float %416, %416
>  %422 = fmul float %417, %417
>  %423 = fadd float %421, %422
>  %424 = call float @llvm.maxnum.f32(float %423, float 0x3E7AD7F2A0000000)
>  %425 = call float @llvm.sqrt.f32(float %424)
>  %426 = fmul float %418, -1.000000e+00
>  %427 = fadd float %426, 5.000000e-01
>  %428 = fmul float %419, 1.000000e+00
>  %429 = fadd float %428, 5.000000e-01
>  %430 = fmul float %420, 0.000000e+00
>  %431 = fadd float %430, 0.000000e+00
>  %432 = fsub float -0.000000e+00, %303
>  %433 = fadd float %432, %427
>  %434 = fsub float -0.000000e+00, %305
>  %435 = fadd float %434, %429
>  %436 = fsub float -0.000000e+00, %307
>  %437 = fadd float %436, %431
>  %438 = fsub float -0.000000e+00, %309
>  %439 = fadd float %438, %425
>  %440 = call float @llvm.AMDIL.clamp.(float %425, float 0.000000e+00, float 1.000000e+00)
>  %441 = fmul float %440, %433
>  %442 = fadd float %441, %303
>  %443 = fmul float %440, %435
>  %444 = fadd float %443, %305
>  %445 = fmul float %440, %437
>  %446 = fadd float %445, %307
>  %447 = fmul float %440, %439
>  %448 = fadd float %447, %309
>  %449 = call i32 @llvm.SI.packf16(float %442, float %444)
>  %450 = bitcast i32 %449 to float
>  %451 = call i32 @llvm.SI.packf16(float %446, float %448)
>  %452 = bitcast i32 %451 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %450, float %452, float %450, float %452)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.fraction.(float) #2
>
>; Function Attrs: readnone
>declare float @llvm.AMDGPU.lrp(float, float, float) #2
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.maxnum.f32(float, float) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.sqrt.f32(float) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.clamp.(float, float, float) #2
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                   ; BEFE0A7E
>	s_mov_b32 m0, s9                                                       ; BEFC0309
>	v_interp_p1_f32 v2, v0, 0, 0, [m0]                                     ; C8080000
>	v_interp_p2_f32 v2, [v2], v1, 0, 0, [m0]                               ; C8090001
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                                     ; C0800300
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	s_buffer_load_dword s8, s[0:3], 0x0                                    ; C2040100
>	s_buffer_load_dword s9, s[0:3], 0x8                                    ; C2048108
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_rcp_f32_e32 v3, s9                                                   ; 7E065409
>	v_mul_f32_e32 v4, s8, v3                                               ; 10080608
>	v_fract_f32_e32 v4, v4                                                 ; 7E084104
>	v_mad_f32 v4, v3, s8, -v4                                              ; D2820004 84101103
>	v_mul_f32_e32 v5, v2, v4                                               ; 100A0902
>	v_fract_f32_e32 v5, v5                                                 ; 7E0A4105
>	v_mad_f32 v6, v4, v2, -v5                                              ; D2820006 84160504
>	v_add_f32_e32 v7, 0, v6                                                ; 060E0C80
>	v_interp_p1_f32 v8, v0, 1, 0, [m0]                                     ; C8200100
>	v_interp_p2_f32 v8, [v8], v1, 1, 0, [m0]                               ; C8210101
>	s_buffer_load_dword s9, s[0:3], 0x1                                    ; C2048101
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mul_f32_e32 v0, s9, v3                                               ; 10000609
>	v_fract_f32_e32 v0, v0                                                 ; 7E004100
>	v_mad_f32 v0, v3, s9, -v0                                              ; D2820000 84001303
>	v_mul_f32_e32 v1, v8, v0                                               ; 10020108
>	v_fract_f32_e32 v1, v1                                                 ; 7E024101
>	v_mad_f32 v3, v0, v8, -v1                                              ; D2820003 84061100
>	v_add_f32_e32 v9, 1.0, v3                                              ; 061206F2
>	v_mul_f32_e32 v9, 0x41880000, v9                                       ; 101212FF 41880000
>	v_rcp_f32_e32 v0, v0                                                   ; 7E005500
>	v_mul_f32_e32 v9, v9, v0                                               ; 10120109
>	v_fract_f32_e32 v9, v9                                                 ; 7E124109
>	v_mul_f32_e32 v9, v9, v9                                               ; 10121309
>	v_mov_b32_e32 v10, 0x45689000                                          ; 7E1402FF 45689000
>	v_mad_f32 v11, v9, v10, v7                                             ; D282000B 041E1509
>	v_mul_f32_e32 v11, 0x41500000, v11                                     ; 101616FF 41500000
>	v_rcp_f32_e32 v4, v4                                                   ; 7E085504
>	v_mul_f32_e32 v11, v11, v4                                             ; 1016090B
>	v_fract_f32_e32 v11, v11                                               ; 7E16410B
>	v_mul_f32_e32 v11, v11, v11                                            ; 1016170B
>	v_mul_f32_e32 v12, 0x44e89000, v11                                     ; 101816FF 44E89000
>	v_fract_f32_e32 v12, v12                                               ; 7E18410C
>	v_add_f32_e32 v12, -0.5, v12                                           ; 061818F1
>	v_cmp_ge_f32_e64 s[10:11], -v12, 0                                     ; D00C000A 2001010C
>	v_cndmask_b32_e64 v13, 0, -1, s[10:11]                                 ; D200080D 00298280
>	v_cmp_ne_i32_e64 s[10:11], v13, 0                                      ; D10A000A 0001010D
>	v_cndmask_b32_e64 v13, 1.0, 0, s[10:11]                                ; D200080D 102900F2
>	v_cmp_ge_f32_e64 s[10:11], v12, 0                                      ; D00C000A 0001010C
>	v_cndmask_b32_e64 v12, 0, -1, s[10:11]                                 ; D200000C 00298280
>	v_cmp_ne_i32_e64 s[10:11], v12, 0                                      ; D10A000A 0001010C
>	v_mov_b32_e32 v12, 0x80000000                                          ; 7E1802FF 80000000
>	v_cndmask_b32_e64 v14, -1.0, v12, s[10:11]                             ; D200000E 182A18F3
>	v_add_f32_e32 v13, v13, v14                                            ; 061A1D0D
>	v_mul_f32_e32 v11, 0x44689000, v11                                     ; 101616FF 44689000
>	v_fract_f32_e32 v11, v11                                               ; 7E16410B
>	v_add_f32_e32 v11, -0.5, v11                                           ; 061616F1
>	v_cmp_ge_f32_e64 s[10:11], -v11, 0                                     ; D00C000A 2001010B
>	v_cndmask_b32_e64 v14, 0, -1, s[10:11]                                 ; D200000E 00298280
>	v_cmp_ne_i32_e64 s[10:11], v14, 0                                      ; D10A000A 0001010E
>	v_cndmask_b32_e64 v14, 1.0, 0, s[10:11]                                ; D200000E 102900F2
>	v_cmp_ge_f32_e64 s[10:11], v11, 0                                      ; D00C000A 0001010B
>	v_cndmask_b32_e64 v11, 0, -1, s[10:11]                                 ; D200080B 00298280
>	v_cmp_ne_i32_e64 s[10:11], v11, 0                                      ; D10A000A 0001010B
>	v_cndmask_b32_e64 v11, -1.0, v12, s[10:11]                             ; D200080B 182A18F3
>	v_add_f32_e32 v11, v14, v11                                            ; 0616170E
>	v_add_f32_e32 v14, -1.0, v1                                            ; 061C02F3
>	v_mul_f32_e32 v15, v14, v11                                            ; 101E170E
>	v_add_f32_e32 v16, v12, v5                                             ; 06200B0C
>	v_mad_f32 v15, v16, v13, v15                                           ; D282000F 043E1B10
>	v_add_f32_e32 v6, 1.0, v6                                              ; 060C0CF2
>	v_mad_f32 v9, v9, v10, v6                                              ; D2820009 041A1509
>	v_mul_f32_e32 v9, 0x41500000, v9                                       ; 101212FF 41500000
>	v_mul_f32_e32 v9, v9, v4                                               ; 10120909
>	v_fract_f32_e32 v9, v9                                                 ; 7E124109
>	v_mul_f32_e32 v9, v9, v9                                               ; 10121309
>	v_mul_f32_e32 v17, 0x44e89000, v9                                      ; 102212FF 44E89000
>	v_fract_f32_e32 v17, v17                                               ; 7E224111
>	v_add_f32_e32 v17, -0.5, v17                                           ; 062222F1
>	v_cmp_ge_f32_e64 s[10:11], -v17, 0                                     ; D00C000A 20010111
>	v_cndmask_b32_e64 v18, 0, -1, s[10:11]                                 ; D2000012 00298280
>	v_cmp_ne_i32_e64 s[10:11], v18, 0                                      ; D10A000A 00010112
>	v_cndmask_b32_e64 v18, 1.0, 0, s[10:11]                                ; D2000012 102900F2
>	v_cmp_ge_f32_e64 s[10:11], v17, 0                                      ; D00C000A 00010111
>	v_cndmask_b32_e64 v17, 0, -1, s[10:11]                                 ; D2000811 00298280
>	v_cmp_ne_i32_e64 s[10:11], v17, 0                                      ; D10A000A 00010111
>	v_cndmask_b32_e64 v17, -1.0, v12, s[10:11]                             ; D2000811 182A18F3
>	v_add_f32_e32 v17, v18, v17                                            ; 06222312
>	v_mul_f32_e32 v9, 0x44689000, v9                                       ; 101212FF 44689000
>	v_fract_f32_e32 v9, v9                                                 ; 7E124109
>	v_add_f32_e32 v9, -0.5, v9                                             ; 061212F1
>	v_cmp_ge_f32_e64 s[10:11], -v9, 0                                      ; D00C000A 20010109
>	v_cndmask_b32_e64 v18, 0, -1, s[10:11]                                 ; D2000012 00298280
>	v_cmp_ne_i32_e64 s[10:11], v18, 0                                      ; D10A000A 00010112
>	v_cndmask_b32_e64 v18, 1.0, 0, s[10:11]                                ; D2000012 102900F2
>	v_cmp_ge_f32_e64 s[10:11], v9, 0                                       ; D00C000A 00010109
>	v_cndmask_b32_e64 v9, 0, -1, s[10:11]                                  ; D2000809 00298280
>	v_cmp_ne_i32_e64 s[10:11], v9, 0                                       ; D10A000A 00010109
>	v_cndmask_b32_e64 v9, -1.0, v12, s[10:11]                              ; D2000809 182A18F3
>	v_add_f32_e32 v9, v18, v9                                              ; 06121312
>	v_mul_f32_e32 v14, v14, v9                                             ; 101C130E
>	v_add_f32_e32 v18, -1.0, v5                                            ; 06240AF3
>	v_mad_f32 v14, v18, v17, v14                                           ; D282000E 043A2312
>	v_subrev_f32_e32 v19, v15, v14                                         ; 0A261D0F
>	v_mov_b32_e32 v20, 0x40400000                                          ; 7E2802FF 40400000
>	v_mad_f32 v21, -2.0, v5, v20                                           ; D2820015 04520AF5
>	v_mul_f32_e32 v22, v5, v5                                              ; 102C0B05
>	v_mad_f32 v23, -v21, v22, 1.0                                          ; D2820017 23CA2D15
>	v_mul_f32_e32 v13, v13, v23                                            ; 101A2F0D
>	v_mul_f32_e32 v21, v22, v21                                            ; 102A2B16
>	v_mad_f32 v13, v21, v17, v13                                           ; D282000D 04362315
>	v_mov_b32_e32 v17, 0x40c00000                                          ; 7E2202FF 40C00000
>	v_mov_b32_e32 v22, 0xc0c00000                                          ; 7E2C02FF C0C00000
>	v_mad_f32 v24, v5, v22, v17                                            ; D2820018 04462D05
>	v_mul_f32_e32 v5, v24, v5                                              ; 100A0B18
>	v_mad_f32 v13, v19, v5, v13                                            ; D282000D 04360B13
>	v_add_f32_e32 v3, 0, v3                                                ; 06060680
>	v_mul_f32_e32 v3, 0x41880000, v3                                       ; 100606FF 41880000
>	v_mul_f32_e32 v0, v3, v0                                               ; 10000103
>	v_fract_f32_e32 v0, v0                                                 ; 7E004100
>	v_mul_f32_e32 v0, v0, v0                                               ; 10000100
>	v_mad_f32 v3, v0, v10, v7                                              ; D2820003 041E1500
>	v_mul_f32_e32 v3, 0x41500000, v3                                       ; 100606FF 41500000
>	v_mul_f32_e32 v3, v3, v4                                               ; 10060903
>	v_fract_f32_e32 v3, v3                                                 ; 7E064103
>	v_mul_f32_e32 v3, v3, v3                                               ; 10060703
>	v_mul_f32_e32 v7, 0x44e89000, v3                                       ; 100E06FF 44E89000
>	v_fract_f32_e32 v7, v7                                                 ; 7E0E4107
>	v_add_f32_e32 v7, -0.5, v7                                             ; 060E0EF1
>	v_cmp_ge_f32_e64 s[10:11], -v7, 0                                      ; D00C000A 20010107
>	v_cndmask_b32_e64 v19, 0, -1, s[10:11]                                 ; D2000813 00298280
>	v_cmp_ne_i32_e64 s[10:11], v19, 0                                      ; D10A000A 00010113
>	v_cndmask_b32_e64 v19, 1.0, 0, s[10:11]                                ; D2000813 102900F2
>	v_cmp_ge_f32_e64 s[10:11], v7, 0                                       ; D00C000A 00010107
>	v_cndmask_b32_e64 v7, 0, -1, s[10:11]                                  ; D2000807 00298280
>	v_cmp_ne_i32_e64 s[10:11], v7, 0                                       ; D10A000A 00010107
>	v_cndmask_b32_e64 v7, -1.0, v12, s[10:11]                              ; D2000807 182A18F3
>	v_add_f32_e32 v7, v19, v7                                              ; 060E0F13
>	v_mul_f32_e32 v3, 0x44689000, v3                                       ; 100606FF 44689000
>	v_fract_f32_e32 v3, v3                                                 ; 7E064103
>	v_add_f32_e32 v3, -0.5, v3                                             ; 060606F1
>	v_cmp_ge_f32_e64 s[10:11], -v3, 0                                      ; D00C000A 20010103
>	v_cndmask_b32_e64 v19, 0, -1, s[10:11]                                 ; D2000813 00298280
>	v_cmp_ne_i32_e64 s[10:11], v19, 0                                      ; D10A000A 00010113
>	v_cndmask_b32_e64 v19, 1.0, 0, s[10:11]                                ; D2000813 102900F2
>	v_cmp_ge_f32_e64 s[10:11], v3, 0                                       ; D00C000A 00010103
>	v_cndmask_b32_e64 v3, 0, -1, s[10:11]                                  ; D2000803 00298280
>	v_cmp_ne_i32_e64 s[10:11], v3, 0                                       ; D10A000A 00010103
>	v_cndmask_b32_e64 v3, -1.0, v12, s[10:11]                              ; D2000803 182A18F3
>	v_add_f32_e32 v3, v19, v3                                              ; 06060713
>	v_add_f32_e32 v19, v12, v1                                             ; 0626030C
>	v_mul_f32_e32 v24, v19, v3                                             ; 10300713
>	v_mad_f32 v16, v16, v7, v24                                            ; D2820010 04620F10
>	v_mad_f32 v0, v0, v10, v6                                              ; D2820000 041A1500
>	v_mul_f32_e32 v0, 0x41500000, v0                                       ; 100000FF 41500000
>	v_mul_f32_e32 v0, v0, v4                                               ; 10000900
>	v_fract_f32_e32 v0, v0                                                 ; 7E004100
>	v_mul_f32_e32 v0, v0, v0                                               ; 10000100
>	v_mul_f32_e32 v4, 0x44e89000, v0                                       ; 100800FF 44E89000
>	v_fract_f32_e32 v4, v4                                                 ; 7E084104
>	v_add_f32_e32 v4, -0.5, v4                                             ; 060808F1
>	v_cmp_ge_f32_e64 s[10:11], -v4, 0                                      ; D00C000A 20010104
>	v_cndmask_b32_e64 v6, 0, -1, s[10:11]                                  ; D2000006 00298280
>	v_cmp_ne_i32_e64 s[10:11], v6, 0                                       ; D10A000A 00010106
>	v_cndmask_b32_e64 v6, 1.0, 0, s[10:11]                                 ; D2000006 102900F2
>	v_cmp_ge_f32_e64 s[10:11], v4, 0                                       ; D00C000A 00010104
>	v_cndmask_b32_e64 v4, 0, -1, s[10:11]                                  ; D2000004 00298280
>	v_cmp_ne_i32_e64 s[10:11], v4, 0                                       ; D10A000A 00010104
>	v_cndmask_b32_e64 v4, -1.0, v12, s[10:11]                              ; D2000004 182A18F3
>	v_add_f32_e32 v4, v6, v4                                               ; 06080906
>	v_mul_f32_e32 v0, 0x44689000, v0                                       ; 100000FF 44689000
>	v_fract_f32_e32 v0, v0                                                 ; 7E004100
>	v_add_f32_e32 v0, -0.5, v0                                             ; 060000F1
>	v_cmp_ge_f32_e64 s[10:11], -v0, 0                                      ; D00C000A 20010100
>	v_cndmask_b32_e64 v6, 0, -1, s[10:11]                                  ; D2000006 00298280
>	v_cmp_ne_i32_e64 s[10:11], v6, 0                                       ; D10A000A 00010106
>	v_cndmask_b32_e64 v6, 1.0, 0, s[10:11]                                 ; D2000006 102900F2
>	v_cmp_ge_f32_e64 s[10:11], v0, 0                                       ; D00C000A 00010100
>	v_cndmask_b32_e64 v0, 0, -1, s[10:11]                                  ; D2000000 00298280
>	v_cmp_ne_i32_e64 s[10:11], v0, 0                                       ; D10A000A 00010100
>	v_cndmask_b32_e64 v0, -1.0, v12, s[10:11]                              ; D2000000 182A18F3
>	v_add_f32_e32 v0, v6, v0                                               ; 06000106
>	v_mul_f32_e32 v6, v19, v0                                              ; 100C0113
>	v_mad_f32 v6, v18, v4, v6                                              ; D2820006 041A0912
>	v_subrev_f32_e32 v10, v16, v6                                          ; 0A140D10
>	v_mul_f32_e32 v7, v7, v23                                              ; 100E2F07
>	v_mad_f32 v4, v21, v4, v7                                              ; D2820004 041E0915
>	v_mad_f32 v4, v10, v5, v4                                              ; D2820004 04120B0A
>	v_mad_f32 v5, -2.0, v1, v20                                            ; D2820005 045202F5
>	v_mul_f32_e32 v7, v1, v1                                               ; 100E0301
>	v_mad_f32 v10, -v5, v7, 1.0                                            ; D282000A 23CA0F05
>	v_mul_f32_e32 v4, v4, v10                                              ; 10081504
>	v_mul_f32_e32 v5, v7, v5                                               ; 100A0B07
>	v_mad_f32 v4, v5, v13, v4                                              ; D2820004 04121B05
>	s_buffer_load_dword s10, s[0:3], 0x4                                   ; C2050104
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mul_f32_e32 v4, s10, v4                                              ; 1008080A
>	v_mad_f32 v4, 0.5, v4, 0.5                                             ; D2820004 03C208F0
>	v_rcp_f32_e32 v7, s9                                                   ; 7E0E5409
>	v_subrev_f32_e32 v13, v7, v8                                           ; 0A1A1107
>	v_rcp_f32_e32 v18, s8                                                  ; 7E245408
>	v_add_f32_e32 v12, v2, v18                                             ; 06182502
>	s_load_dwordx4 s[12:15], s[4:5], 0x0                                   ; C0860500
>	s_load_dwordx8 s[16:23], s[6:7], 0x0                                   ; C0C80700
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	image_sample v19, 1, 0, 0, 0, 0, 0, 0, 0, v[12:13], s[16:23], s[12:15] ; F0800100 0064130C
>	v_subrev_f32_e32 v24, v18, v2                                          ; 0A300512
>	v_mov_b32_e32 v25, v13                                                 ; 7E32030D
>	image_sample v18, 1, 0, 0, 0, 0, 0, 0, 0, v[24:25], s[16:23], s[12:15] ; F0800100 00641218
>	s_waitcnt vmcnt(0)                                                     ; BF8C0770
>	v_subrev_f32_e32 v20, v19, v18                                         ; 0A282513
>	v_add_f32_e32 v26, 0, v8                                               ; 06341080
>	v_mov_b32_e32 v25, v26                                                 ; 7E32031A
>	image_sample v27, 1, 0, 0, 0, 0, 0, 0, 0, v[24:25], s[16:23], s[12:15] ; F0800100 00641B18
>	s_waitcnt vmcnt(0)                                                     ; BF8C0770
>	v_mad_f32 v20, 2.0, v27, v20                                           ; D2820014 045236F4
>	v_add_f32_e32 v25, v8, v7                                              ; 06320F08
>	v_mov_b32_e32 v7, v12                                                  ; 7E0E030C
>	v_mov_b32_e32 v8, v13                                                  ; 7E10030D
>	v_mov_b32_e32 v8, v25                                                  ; 7E100319
>	image_sample v7, 1, 0, 0, 0, 0, 0, 0, 0, v[7:8], s[16:23], s[12:15]    ; F0800100 00640707
>	image_sample v8, 1, 0, 0, 0, 0, 0, 0, 0, v[24:25], s[16:23], s[12:15]  ; F0800100 00640818
>	s_waitcnt vmcnt(0)                                                     ; BF8C0770
>	v_subrev_f32_e32 v27, v7, v8                                           ; 0A361107
>	v_mov_b32_e32 v28, v12                                                 ; 7E38030C
>	v_mov_b32_e32 v29, v13                                                 ; 7E3A030D
>	v_mov_b32_e32 v29, v26                                                 ; 7E3A031A
>	image_sample v26, 1, 0, 0, 0, 0, 0, 0, 0, v[28:29], s[16:23], s[12:15] ; F0800100 00641A1C
>	s_waitcnt vmcnt(0)                                                     ; BF8C0770
>	v_mad_f32 v26, -2.0, v26, v27                                          ; D282001A 046E34F5
>	v_add_f32_e32 v20, v20, v26                                            ; 06283514
>	s_buffer_load_dword s0, s[0:3], 0xc                                    ; C200010C
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mul_f32_e32 v20, s0, v20                                             ; 10282800
>	v_mad_f32 v26, 0.5, v20, 0.5                                           ; D282001A 03C228F0
>	v_subrev_f32_e32 v26, v4, v26                                          ; 0A343504
>	v_add_f32_e32 v18, v18, v19                                            ; 06242712
>	v_add_f32_e32 v12, 0, v2                                               ; 06180480
>	image_sample v2, 1, 0, 0, 0, 0, 0, 0, 0, v[12:13], s[16:23], s[12:15]  ; F0800100 0064020C
>	s_waitcnt vmcnt(0)                                                     ; BF8C0770
>	v_mad_f32 v2, 2.0, v2, v18                                             ; D2820002 044A04F4
>	v_sub_f32_e64 v7, -v7, v8                                              ; D2080007 20021107
>	v_mov_b32_e32 v13, v25                                                 ; 7E1A0319
>	image_sample v8, 1, 0, 0, 0, 0, 0, 0, 0, v[12:13], s[16:23], s[12:15]  ; F0800100 0064080C
>	s_waitcnt vmcnt(0)                                                     ; BF8C0770
>	v_mad_f32 v7, -2.0, v8, v7                                             ; D2820007 041E10F5
>	v_add_f32_e32 v2, v2, v7                                               ; 06040F02
>	v_mul_f32_e32 v2, s0, v2                                               ; 10040400
>	v_mul_f32_e32 v7, v2, v2                                               ; 100E0502
>	v_mad_f32 v7, v20, v20, v7                                             ; D2820007 041E2914
>	v_max_f32_e32 v7, 0x33d6bf95, v7                                       ; 200E0EFF 33D6BF95
>	v_sqrt_f32_e32 v7, v7                                                  ; 7E0E6707
>	v_add_f32_e64 v8, 0, v7 clamp                                          ; D2060808 00020E80
>	v_mad_f32 v4, v8, v26, v4                                              ; D2820004 04123508
>	v_subrev_f32_e32 v6, v6, v14                                           ; 0A0C1D06
>	v_mul_f32_e32 v0, v0, v10                                              ; 10001500
>	v_mad_f32 v0, v5, v9, v0                                               ; D2820000 04021305
>	v_mad_f32 v9, v1, v22, v17                                             ; D2820009 04462D01
>	v_mul_f32_e32 v1, v9, v1                                               ; 10020309
>	v_mad_f32 v0, v6, v1, v0                                               ; D2820000 04020306
>	v_subrev_f32_e32 v6, v16, v15                                          ; 0A0C1F10
>	v_mul_f32_e32 v3, v3, v10                                              ; 10061503
>	v_mad_f32 v3, v5, v11, v3                                              ; D2820003 040E1705
>	v_mad_f32 v1, v6, v1, v3                                               ; D2820001 040E0306
>	v_mul_f32_e32 v1, v1, v23                                              ; 10022F01
>	v_mad_f32 v0, v21, v0, v1                                              ; D2820000 04060115
>	v_mul_f32_e32 v0, s10, v0                                              ; 1000000A
>	v_mad_f32 v1, 0.5, -v0, 0.5                                            ; D2820001 43C200F0
>	v_mad_f32 v3, 0.5, -v2, 0.5                                            ; D2820003 43C204F0
>	v_subrev_f32_e32 v3, v1, v3                                            ; 0A060701
>	v_mad_f32 v1, v8, v3, v1                                               ; D2820001 04060708
>	v_cvt_pkrtz_f16_f32_e32 v1, v1, v4                                     ; 5E020901
>	v_mul_f32_e32 v0, 0.5, v0                                              ; 100000F0
>	v_mad_f32 v3, 0, v0, 1.0                                               ; D2820003 03CA0080
>	v_subrev_f32_e32 v4, v3, v7                                            ; 0A080F03
>	v_mad_f32 v3, v8, v4, v3                                               ; D2820003 040E0908
>	v_mad_f32 v0, 0, v0, 0                                                 ; D2820000 02020080
>	v_mul_f32_e32 v2, 0.5, v2                                              ; 100404F0
>	v_mad_f32 v2, 0, v2, 0                                                 ; D2820002 02020480
>	v_subrev_f32_e32 v2, v0, v2                                            ; 0A040500
>	v_mad_f32 v0, v8, v2, v0                                               ; D2820000 04020508
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v3                                     ; 5E000700
>	exp 15, 0, 1, 1, 1, v1, v0, v1, v0                                     ; F8001C0F 00010001
>	s_endpgm                                                               ; BF810000
>
>SGPRS: 24
>VGPRS: 32
>Code Size: 1688 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 1
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>PROPERTY FS_COLOR0_WRITES_ALL_CBUFS 1
>DCL IN[0], GENERIC[0], CONSTANT
>DCL OUT[0], COLOR
>  0: MOV OUT[0], IN[0]
>  1: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = call float @llvm.SI.fs.constant(i32 0, i32 0, i32 %5)
>  %23 = call float @llvm.SI.fs.constant(i32 1, i32 0, i32 %5)
>  %24 = call float @llvm.SI.fs.constant(i32 2, i32 0, i32 %5)
>  %25 = call float @llvm.SI.fs.constant(i32 3, i32 0, i32 %5)
>  %26 = call i32 @llvm.SI.packf16(float %22, float %23)
>  %27 = bitcast i32 %26 to float
>  %28 = call i32 @llvm.SI.packf16(float %24, float %25)
>  %29 = bitcast i32 %28 to float
>  %30 = call i32 @llvm.SI.packf16(float %22, float %23)
>  %31 = bitcast i32 %30 to float
>  %32 = call i32 @llvm.SI.packf16(float %24, float %25)
>  %33 = bitcast i32 %32 to float
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 1, i32 1, float %31, float %33, float %31, float %33)
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %27, float %29, float %27, float %29)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.constant(i32, i32, i32) #1
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                ; BEFE0A7E
>	s_mov_b32 m0, s9                    ; BEFC0309
>	v_interp_mov_f32 v0, P0, 3, 0, [m0] ; C8020302
>	v_interp_mov_f32 v1, P0, 2, 0, [m0] ; C8060202
>	v_cvt_pkrtz_f16_f32_e32 v0, v1, v0  ; 5E000101
>	v_interp_mov_f32 v1, P0, 1, 0, [m0] ; C8060102
>	v_interp_mov_f32 v2, P0, 0, 0, [m0] ; C80A0002
>	v_cvt_pkrtz_f16_f32_e32 v1, v2, v1  ; 5E020302
>	exp 15, 1, 1, 0, 0, v1, v0, v1, v0  ; F800041F 00010001
>	exp 15, 0, 1, 1, 1, v1, v0, v1, v0  ; F8001C0F 00010001
>	s_endpgm                            ; BF810000
>
>SGPRS: 16
>VGPRS: 4
>Code Size: 52 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL IN[2]
>DCL IN[3]
>DCL IN[4]
>DCL IN[5]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL OUT[2], GENERIC[10]
>DCL OUT[3], GENERIC[11]
>DCL OUT[4], GENERIC[12]
>DCL OUT[5], GENERIC[13]
>DCL OUT[6], GENERIC[14]
>DCL OUT[7], GENERIC[15]
>DCL OUT[8], GENERIC[16]
>DCL CONST[0..96]
>DCL TEMP[0..9], LOCAL
>DCL ADDR[0]
>IMM[0] FLT32 {   -1.0000,    -2.0000,     0.6600,     0.3300}
>IMM[1] FLT32 {    1.0000,     0.0000,     0.0000,     0.0000}
>  0: ADD TEMP[0].xy, CONST[0], IN[0]
>  1: MOV TEMP[0].xy, TEMP[0].xyxx
>  2: MOV TEMP[1].xy, IMM[0].xyxx
>  3: F2I TEMP[2].x, IN[4].xxxx
>  4: UARL ADDR[0].x, TEMP[2].xxxx
>  5: ADD TEMP[2].zw, IMM[0].xyxy, CONST[ADDR[0].x+1].yyyy
>  6: MOV TEMP[1].zw, TEMP[2].wwzw
>  7: ABS TEMP[2].z, TEMP[1]
>  8: ABS TEMP[3].z, TEMP[1]
>  9: FSGE TEMP[2].x, -TEMP[2].zzzz, TEMP[3].zzzz
> 10: UIF TEMP[2].xxxx :0
> 11:   F2I TEMP[2].x, IN[4].xxxx
> 12:   UARL ADDR[0].x, TEMP[2].xxxx
> 13:   MUL TEMP[2].xy, IN[1].xzzw, CONST[ADDR[0].x+1].xxxx
> 14:   MOV TEMP[2].xy, TEMP[2].xyxx
> 15:   MUL TEMP[3].xy, TEMP[2], IMM[0].zzzz
> 16:   MOV TEMP[2].xy, TEMP[3].xyxx
> 17: ELSE :0
> 18:   F2I TEMP[3].x, IN[4].xxxx
> 19:   UARL ADDR[0].x, TEMP[3].xxxx
> 20:   MUL TEMP[3].xy, IN[1], CONST[ADDR[0].x+1].xxxx
> 21:   MOV TEMP[3].xy, TEMP[3].xyxx
> 22:   MUL TEMP[4].xy, TEMP[3], IMM[0].wwww
> 23:   MOV TEMP[2].xy, TEMP[4].xyxx
> 24: ENDIF
> 25: ABS TEMP[4].w, TEMP[1]
> 26: ABS TEMP[5].w, TEMP[1]
> 27: FSGE TEMP[4].x, -TEMP[4].wwww, TEMP[5].wwww
> 28: UIF TEMP[4].xxxx :0
> 29:   F2I TEMP[4].x, IN[4].xxxx
> 30:   UARL ADDR[0].x, TEMP[4].xxxx
> 31:   MUL TEMP[4].zw, IN[1].xyyz, CONST[ADDR[0].x+1].xxxx
> 32:   MUL TEMP[4].xy, TEMP[4].zwzw, IMM[0].zzzz
> 33:   MOV TEMP[2].xy, TEMP[4].xyxx
> 34: ENDIF
> 35: F2I TEMP[4].x, IN[4].yyyy
> 36: UARL ADDR[0].x, TEMP[4].xxxx
> 37: ADD TEMP[4].zw, IMM[0].xyxy, CONST[ADDR[0].x+1].yyyy
> 38: MOV TEMP[1].zw, TEMP[4].wwzw
> 39: ABS TEMP[4].z, TEMP[1]
> 40: ABS TEMP[5].z, TEMP[1]
> 41: FSGE TEMP[4].x, -TEMP[4].zzzz, TEMP[5].zzzz
> 42: UIF TEMP[4].xxxx :0
> 43:   F2I TEMP[4].x, IN[4].yyyy
> 44:   UARL ADDR[0].x, TEMP[4].xxxx
> 45:   MUL TEMP[4].xy, IN[1].xzzw, CONST[ADDR[0].x+1].xxxx
> 46:   MOV TEMP[3].xy, TEMP[4].xyxx
> 47:   MUL TEMP[4].xy, TEMP[3], IMM[0].zzzz
> 48:   MOV TEMP[3].xy, TEMP[4].xyxx
> 49: ELSE :0
> 50:   F2I TEMP[4].x, IN[4].yyyy
> 51:   UARL ADDR[0].x, TEMP[4].xxxx
> 52:   MUL TEMP[4].xy, IN[1], CONST[ADDR[0].x+1].xxxx
> 53:   MOV TEMP[4].xy, TEMP[4].xyxx
> 54:   MUL TEMP[5].xy, TEMP[4], IMM[0].wwww
> 55:   MOV TEMP[3].xy, TEMP[5].xyxx
> 56: ENDIF
> 57: ABS TEMP[5].w, TEMP[1]
> 58: ABS TEMP[6].w, TEMP[1]
> 59: FSGE TEMP[5].x, -TEMP[5].wwww, TEMP[6].wwww
> 60: UIF TEMP[5].xxxx :0
> 61:   F2I TEMP[5].x, IN[4].yyyy
> 62:   UARL ADDR[0].x, TEMP[5].xxxx
> 63:   MUL TEMP[5].zw, IN[1].xyyz, CONST[ADDR[0].x+1].xxxx
> 64:   MUL TEMP[5].xy, TEMP[5].zwzw, IMM[0].zzzz
> 65:   MOV TEMP[3].xy, TEMP[5].xyxx
> 66: ENDIF
> 67: F2I TEMP[5].x, IN[4].zzzz
> 68: UARL ADDR[0].x, TEMP[5].xxxx
> 69: ADD TEMP[5].zw, IMM[0].xyxy, CONST[ADDR[0].x+1].yyyy
> 70: MOV TEMP[1].zw, TEMP[5].wwzw
> 71: ABS TEMP[5].z, TEMP[1]
> 72: ABS TEMP[6].z, TEMP[1]
> 73: FSGE TEMP[5].x, -TEMP[5].zzzz, TEMP[6].zzzz
> 74: UIF TEMP[5].xxxx :0
> 75:   F2I TEMP[5].x, IN[4].zzzz
> 76:   UARL ADDR[0].x, TEMP[5].xxxx
> 77:   MUL TEMP[5].xy, IN[1].xzzw, CONST[ADDR[0].x+1].xxxx
> 78:   MOV TEMP[4].xy, TEMP[5].xyxx
> 79:   MUL TEMP[5].xy, TEMP[4], IMM[0].zzzz
> 80:   MOV TEMP[4].xy, TEMP[5].xyxx
> 81: ELSE :0
> 82:   F2I TEMP[5].x, IN[4].zzzz
> 83:   UARL ADDR[0].x, TEMP[5].xxxx
> 84:   MUL TEMP[5].xy, IN[1], CONST[ADDR[0].x+1].xxxx
> 85:   MOV TEMP[5].xy, TEMP[5].xyxx
> 86:   MUL TEMP[6].xy, TEMP[5], IMM[0].wwww
> 87:   MOV TEMP[4].xy, TEMP[6].xyxx
> 88: ENDIF
> 89: ABS TEMP[6].w, TEMP[1]
> 90: ABS TEMP[7].w, TEMP[1]
> 91: FSGE TEMP[6].x, -TEMP[6].wwww, TEMP[7].wwww
> 92: UIF TEMP[6].xxxx :0
> 93:   F2I TEMP[6].x, IN[4].zzzz
> 94:   UARL ADDR[0].x, TEMP[6].xxxx
> 95:   MUL TEMP[6].zw, IN[1].xyyz, CONST[ADDR[0].x+1].xxxx
> 96:   MUL TEMP[6].xy, TEMP[6].zwzw, IMM[0].zzzz
> 97:   MOV TEMP[4].xy, TEMP[6].xyxx
> 98: ENDIF
> 99: F2I TEMP[6].x, IN[4].wwww
>100: UARL ADDR[0].x, TEMP[6].xxxx
>101: ADD TEMP[6].zw, IMM[0].xyxy, CONST[ADDR[0].x+1].yyyy
>102: MOV TEMP[1].zw, TEMP[6].wwzw
>103: ABS TEMP[6].z, TEMP[1]
>104: ABS TEMP[7].z, TEMP[1]
>105: FSGE TEMP[6].x, -TEMP[6].zzzz, TEMP[7].zzzz
>106: UIF TEMP[6].xxxx :0
>107:   F2I TEMP[6].x, IN[4].wwww
>108:   UARL ADDR[0].x, TEMP[6].xxxx
>109:   MUL TEMP[6].xy, IN[1].xzzw, CONST[ADDR[0].x+1].xxxx
>110:   MOV TEMP[5].xy, TEMP[6].xyxx
>111:   MUL TEMP[6].xy, TEMP[5], IMM[0].zzzz
>112:   MOV TEMP[5].xy, TEMP[6].xyxx
>113: ELSE :0
>114:   F2I TEMP[6].x, IN[4].wwww
>115:   UARL ADDR[0].x, TEMP[6].xxxx
>116:   MUL TEMP[6].xy, IN[1], CONST[ADDR[0].x+1].xxxx
>117:   MOV TEMP[6].xy, TEMP[6].xyxx
>118:   MUL TEMP[7].xy, TEMP[6], IMM[0].wwww
>119:   MOV TEMP[5].xy, TEMP[7].xyxx
>120: ENDIF
>121: ABS TEMP[7].w, TEMP[1]
>122: ABS TEMP[8].w, TEMP[1]
>123: FSGE TEMP[7].x, -TEMP[7].wwww, TEMP[8].wwww
>124: UIF TEMP[7].xxxx :0
>125:   F2I TEMP[7].x, IN[4].wwww
>126:   UARL ADDR[0].x, TEMP[7].xxxx
>127:   MUL TEMP[7].zw, IN[1].xyyz, CONST[ADDR[0].x+1].xxxx
>128:   MUL TEMP[7].xy, TEMP[7].zwzw, IMM[0].zzzz
>129:   MOV TEMP[5].xy, TEMP[7].xyxx
>130: ENDIF
>131: F2I TEMP[7].x, IN[5].xxxx
>132: UARL ADDR[0].x, TEMP[7].xxxx
>133: ADD TEMP[7].zw, IMM[0].xyxy, CONST[ADDR[0].x+1].yyyy
>134: MOV TEMP[1].zw, TEMP[7].wwzw
>135: ABS TEMP[7].z, TEMP[1]
>136: ABS TEMP[8].z, TEMP[1]
>137: FSGE TEMP[7].x, -TEMP[7].zzzz, TEMP[8].zzzz
>138: UIF TEMP[7].xxxx :0
>139:   F2I TEMP[7].x, IN[5].xxxx
>140:   UARL ADDR[0].x, TEMP[7].xxxx
>141:   MUL TEMP[7].xy, IN[1].xzzw, CONST[ADDR[0].x+1].xxxx
>142:   MOV TEMP[6].xy, TEMP[7].xyxx
>143:   MUL TEMP[7].xy, TEMP[6], IMM[0].zzzz
>144:   MOV TEMP[6].xy, TEMP[7].xyxx
>145: ELSE :0
>146:   F2I TEMP[7].x, IN[5].xxxx
>147:   UARL ADDR[0].x, TEMP[7].xxxx
>148:   MUL TEMP[7].zw, IN[1].xyxy, CONST[ADDR[0].x+1].xxxx
>149:   MOV TEMP[6].zw, TEMP[7].wwzw
>150:   MUL TEMP[7].xy, TEMP[7].zwzw, IMM[0].wwww
>151:   MOV TEMP[6].xy, TEMP[7].xyxx
>152: ENDIF
>153: ABS TEMP[7].w, TEMP[1]
>154: ABS TEMP[8].w, TEMP[1]
>155: FSGE TEMP[7].x, -TEMP[7].wwww, TEMP[8].wwww
>156: UIF TEMP[7].xxxx :0
>157:   F2I TEMP[7].x, IN[5].xxxx
>158:   UARL ADDR[0].x, TEMP[7].xxxx
>159:   MUL TEMP[7].zw, IN[1].xyyz, CONST[ADDR[0].x+1].xxxx
>160:   MUL TEMP[7].xy, TEMP[7].zwzw, IMM[0].zzzz
>161:   MOV TEMP[6].xy, TEMP[7].xyxx
>162: ENDIF
>163: F2I TEMP[7].x, IN[5].yyyy
>164: UARL ADDR[0].x, TEMP[7].xxxx
>165: ADD TEMP[7].zw, IMM[0].xyxy, CONST[ADDR[0].x+1].yyyy
>166: MOV TEMP[1].zw, TEMP[7].wwzw
>167: ABS TEMP[7].z, TEMP[1]
>168: ABS TEMP[8].z, TEMP[1]
>169: FSGE TEMP[7].x, -TEMP[7].zzzz, TEMP[8].zzzz
>170: UIF TEMP[7].xxxx :0
>171:   F2I TEMP[7].x, IN[5].yyyy
>172:   UARL ADDR[0].x, TEMP[7].xxxx
>173:   MUL TEMP[7].zw, IN[1].xyxz, CONST[ADDR[0].x+1].xxxx
>174:   MOV TEMP[6].zw, TEMP[7].wwzw
>175:   MUL TEMP[7].zw, TEMP[6], IMM[0].zzzz
>176:   MOV TEMP[6].zw, TEMP[7].wwzw
>177: ELSE :0
>178:   F2I TEMP[7].x, IN[5].yyyy
>179:   UARL ADDR[0].x, TEMP[7].xxxx
>180:   MUL TEMP[7].xy, IN[1], CONST[ADDR[0].x+1].xxxx
>181:   MOV TEMP[8].xy, TEMP[7].xyxx
>182:   MUL TEMP[7].zw, TEMP[7].xyxy, IMM[0].wwww
>183:   MOV TEMP[6].zw, TEMP[7].wwzw
>184: ENDIF
>185: ABS TEMP[7].w, TEMP[1]
>186: ABS TEMP[9].w, TEMP[1]
>187: FSGE TEMP[7].x, -TEMP[7].wwww, TEMP[9].wwww
>188: UIF TEMP[7].xxxx :0
>189:   F2I TEMP[7].x, IN[5].yyyy
>190:   UARL ADDR[0].x, TEMP[7].xxxx
>191:   MUL TEMP[7].zw, IN[1].xyyz, CONST[ADDR[0].x+1].xxxx
>192:   MOV TEMP[1].zw, TEMP[7].wwzw
>193:   MUL TEMP[7].zw, TEMP[1], IMM[0].zzzz
>194:   MOV TEMP[6].zw, TEMP[7].wwzw
>195: ENDIF
>196: F2I TEMP[7].x, IN[5].zzzz
>197: UARL ADDR[0].x, TEMP[7].xxxx
>198: ADD TEMP[7].zw, IMM[0].xyxy, CONST[ADDR[0].x+1].yyyy
>199: MOV TEMP[1].zw, TEMP[7].wwzw
>200: ABS TEMP[7].z, TEMP[1]
>201: ABS TEMP[9].z, TEMP[1]
>202: FSGE TEMP[7].x, -TEMP[7].zzzz, TEMP[9].zzzz
>203: UIF TEMP[7].xxxx :0
>204:   F2I TEMP[7].x, IN[5].zzzz
>205:   UARL ADDR[0].x, TEMP[7].xxxx
>206:   MUL TEMP[7].xy, IN[1].xzzw, CONST[ADDR[0].x+1].xxxx
>207:   MOV TEMP[8].xy, TEMP[7].xyxx
>208:   MUL TEMP[7].xy, TEMP[8], IMM[0].zzzz
>209:   MOV TEMP[8].xy, TEMP[7].xyxx
>210: ELSE :0
>211:   F2I TEMP[7].x, IN[5].zzzz
>212:   UARL ADDR[0].x, TEMP[7].xxxx
>213:   MUL TEMP[7].zw, IN[1].xyxy, CONST[ADDR[0].x+1].xxxx
>214:   MOV TEMP[8].zw, TEMP[7].wwzw
>215:   MUL TEMP[7].xy, TEMP[7].zwzw, IMM[0].wwww
>216:   MOV TEMP[8].xy, TEMP[7].xyxx
>217: ENDIF
>218: ABS TEMP[7].w, TEMP[1]
>219: ABS TEMP[9].w, TEMP[1]
>220: FSGE TEMP[7].x, -TEMP[7].wwww, TEMP[9].wwww
>221: UIF TEMP[7].xxxx :0
>222:   F2I TEMP[7].x, IN[5].zzzz
>223:   UARL ADDR[0].x, TEMP[7].xxxx
>224:   MUL TEMP[7].zw, IN[1].xyyz, CONST[ADDR[0].x+1].xxxx
>225:   MOV TEMP[1].zw, TEMP[7].wwzw
>226:   MUL TEMP[7].xy, TEMP[7].zwzw, IMM[0].zzzz
>227:   MOV TEMP[8].xy, TEMP[7].xyxx
>228: ENDIF
>229: F2I TEMP[7].x, IN[5].wwww
>230: UARL ADDR[0].x, TEMP[7].xxxx
>231: ADD TEMP[7].xy, TEMP[1], CONST[ADDR[0].x+1].yyyy
>232: MOV TEMP[1].xy, TEMP[7].xyxx
>233: ABS TEMP[7].x, TEMP[1]
>234: ABS TEMP[9].x, TEMP[1]
>235: FSGE TEMP[7].x, -TEMP[7].xxxx, TEMP[9].xxxx
>236: UIF TEMP[7].xxxx :0
>237:   F2I TEMP[7].x, IN[5].wwww
>238:   UARL ADDR[0].x, TEMP[7].xxxx
>239:   MUL TEMP[7].xz, IN[1], CONST[ADDR[0].x+1].xxxx
>240:   MOV TEMP[1].xz, TEMP[7].xxzx
>241:   MUL TEMP[7].xz, TEMP[1], IMM[0].zzzz
>242:   MOV TEMP[1].xz, TEMP[7].xxzx
>243: ELSE :0
>244:   F2I TEMP[7].x, IN[5].wwww
>245:   UARL ADDR[0].x, TEMP[7].xxxx
>246:   MUL TEMP[7].zw, IN[1].xyxy, CONST[ADDR[0].x+1].xxxx
>247:   MOV TEMP[8].zw, TEMP[7].wwzw
>248:   MUL TEMP[7].xz, TEMP[8].zyww, IMM[0].wwww
>249:   MOV TEMP[1].xz, TEMP[7].xxzx
>250: ENDIF
>251: ABS TEMP[7].y, TEMP[1]
>252: ABS TEMP[9].y, TEMP[1]
>253: FSGE TEMP[7].x, -TEMP[7].yyyy, TEMP[9].yyyy
>254: UIF TEMP[7].xxxx :0
>255:   F2I TEMP[7].x, IN[5].wwww
>256:   UARL ADDR[0].x, TEMP[7].xxxx
>257:   MUL TEMP[7].yw, IN[1].xyzz, CONST[ADDR[0].x+1].xxxx
>258:   MUL TEMP[7].xz, TEMP[7].yyww, IMM[0].zzzz
>259:   MOV TEMP[1].xz, TEMP[7].xxzx
>260: ENDIF
>261: MAD TEMP[7].zw, IN[0].zzzz, IMM[1].xyxy, IMM[1].xyyx
>262: MOV TEMP[0].zw, TEMP[7].wwzw
>263: MOV TEMP[2].zw, IN[2].yyxy
>264: MOV TEMP[3].zw, IN[2].wwzw
>265: MOV TEMP[4].zw, IN[3].yyxy
>266: MOV TEMP[5].zw, IN[3].wwzw
>267: MOV TEMP[7].xy, TEMP[6].xyxx
>268: MOV TEMP[7].zw, IMM[1].yyyy
>269: MOV TEMP[6].xy, TEMP[6].zwzz
>270: MOV TEMP[6].zw, IMM[1].yyyy
>271: MOV TEMP[8].xy, TEMP[8].xyxx
>272: MOV TEMP[8].zw, IMM[1].yyyy
>273: MOV TEMP[1].xy, TEMP[1].xzxx
>274: MOV TEMP[1].zw, IMM[1].yyyy
>275: MOV OUT[6], TEMP[6]
>276: MOV OUT[7], TEMP[8]
>277: MOV OUT[8], TEMP[1]
>278: MOV OUT[1], TEMP[2]
>279: MOV OUT[2], TEMP[3]
>280: MOV OUT[0], TEMP[0]
>281: MOV OUT[3], TEMP[4]
>282: MOV OUT[4], TEMP[5]
>283: MOV OUT[5], TEMP[7]
>284: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 0)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 4)
>  %15 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %16 = load <16 x i8> addrspace(2)* %15, !tbaa !0
>  %17 = add i32 %5, %7
>  %18 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %16, i32 0, i32 %17)
>  %19 = extractelement <4 x float> %18, i32 0
>  %20 = extractelement <4 x float> %18, i32 1
>  %21 = extractelement <4 x float> %18, i32 2
>  %22 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %23 = load <16 x i8> addrspace(2)* %22, !tbaa !0
>  %24 = add i32 %5, %7
>  %25 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %23, i32 0, i32 %24)
>  %26 = extractelement <4 x float> %25, i32 0
>  %27 = extractelement <4 x float> %25, i32 1
>  %28 = extractelement <4 x float> %25, i32 2
>  %29 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 2
>  %30 = load <16 x i8> addrspace(2)* %29, !tbaa !0
>  %31 = add i32 %5, %7
>  %32 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %30, i32 0, i32 %31)
>  %33 = extractelement <4 x float> %32, i32 0
>  %34 = extractelement <4 x float> %32, i32 1
>  %35 = extractelement <4 x float> %32, i32 2
>  %36 = extractelement <4 x float> %32, i32 3
>  %37 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 3
>  %38 = load <16 x i8> addrspace(2)* %37, !tbaa !0
>  %39 = add i32 %5, %7
>  %40 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %38, i32 0, i32 %39)
>  %41 = extractelement <4 x float> %40, i32 0
>  %42 = extractelement <4 x float> %40, i32 1
>  %43 = extractelement <4 x float> %40, i32 2
>  %44 = extractelement <4 x float> %40, i32 3
>  %45 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 4
>  %46 = load <16 x i8> addrspace(2)* %45, !tbaa !0
>  %47 = add i32 %5, %7
>  %48 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %46, i32 0, i32 %47)
>  %49 = extractelement <4 x float> %48, i32 0
>  %50 = extractelement <4 x float> %48, i32 1
>  %51 = extractelement <4 x float> %48, i32 2
>  %52 = extractelement <4 x float> %48, i32 3
>  %53 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 5
>  %54 = load <16 x i8> addrspace(2)* %53, !tbaa !0
>  %55 = add i32 %5, %7
>  %56 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %54, i32 0, i32 %55)
>  %57 = extractelement <4 x float> %56, i32 0
>  %58 = extractelement <4 x float> %56, i32 1
>  %59 = extractelement <4 x float> %56, i32 2
>  %60 = extractelement <4 x float> %56, i32 3
>  %61 = fadd float %13, %19
>  %62 = fadd float %14, %20
>  %63 = fptosi float %49 to i32
>  %64 = bitcast i32 %63 to float
>  %65 = bitcast float %64 to i32
>  %66 = shl i32 %65, 4
>  %67 = add i32 %66, 20
>  %68 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %67)
>  %69 = fadd float -1.000000e+00, %68
>  %70 = shl i32 %65, 4
>  %71 = add i32 %70, 20
>  %72 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %71)
>  %73 = fadd float -2.000000e+00, %72
>  %74 = call float @fabs(float %69)
>  %75 = call float @fabs(float %69)
>  %76 = fsub float -0.000000e+00, %74
>  %77 = fcmp oge float %76, %75
>  %78 = sext i1 %77 to i32
>  %79 = bitcast i32 %78 to float
>  %80 = bitcast float %79 to i32
>  %81 = icmp ne i32 %80, 0
>  %82 = fptosi float %49 to i32
>  %83 = bitcast i32 %82 to float
>  %84 = bitcast float %83 to i32
>  %85 = shl i32 %84, 4
>  %86 = add i32 %85, 16
>  %87 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %86)
>  %88 = fmul float %26, %87
>  %89 = shl i32 %84, 4
>  %90 = add i32 %89, 16
>  %91 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %90)
>  br i1 %81, label %IF, label %ELSE
>
>IF:                                               ; preds = %main_body
>  %92 = fmul float %28, %91
>  %93 = fmul float %88, 0x3FE51EB860000000
>  %94 = fmul float %92, 0x3FE51EB860000000
>  br label %ENDIF
>
>ELSE:                                             ; preds = %main_body
>  %95 = fmul float %27, %91
>  %96 = fmul float %88, 0x3FD51EB860000000
>  %97 = fmul float %95, 0x3FD51EB860000000
>  br label %ENDIF
>
>ENDIF:                                            ; preds = %ELSE, %IF
>  %temp8.0 = phi float [ %93, %IF ], [ %96, %ELSE ]
>  %temp9.0 = phi float [ %94, %IF ], [ %97, %ELSE ]
>  %98 = call float @fabs(float %73)
>  %99 = call float @fabs(float %73)
>  %100 = fsub float -0.000000e+00, %98
>  %101 = fcmp oge float %100, %99
>  %102 = sext i1 %101 to i32
>  %103 = bitcast i32 %102 to float
>  %104 = bitcast float %103 to i32
>  %105 = icmp ne i32 %104, 0
>  br i1 %105, label %IF46, label %ENDIF45
>
>IF46:                                             ; preds = %ENDIF
>  %106 = fptosi float %49 to i32
>  %107 = bitcast i32 %106 to float
>  %108 = bitcast float %107 to i32
>  %109 = shl i32 %108, 4
>  %110 = add i32 %109, 16
>  %111 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %110)
>  %112 = fmul float %27, %111
>  %113 = shl i32 %108, 4
>  %114 = add i32 %113, 16
>  %115 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %114)
>  %116 = fmul float %28, %115
>  %117 = fmul float %112, 0x3FE51EB860000000
>  %118 = fmul float %116, 0x3FE51EB860000000
>  br label %ENDIF45
>
>ENDIF45:                                          ; preds = %ENDIF, %IF46
>  %temp8.1 = phi float [ %117, %IF46 ], [ %temp8.0, %ENDIF ]
>  %temp9.1 = phi float [ %118, %IF46 ], [ %temp9.0, %ENDIF ]
>  %119 = fptosi float %50 to i32
>  %120 = bitcast i32 %119 to float
>  %121 = bitcast float %120 to i32
>  %122 = shl i32 %121, 4
>  %123 = add i32 %122, 20
>  %124 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %123)
>  %125 = fadd float -1.000000e+00, %124
>  %126 = shl i32 %121, 4
>  %127 = add i32 %126, 20
>  %128 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %127)
>  %129 = fadd float -2.000000e+00, %128
>  %130 = call float @fabs(float %125)
>  %131 = call float @fabs(float %125)
>  %132 = fsub float -0.000000e+00, %130
>  %133 = fcmp oge float %132, %131
>  %134 = sext i1 %133 to i32
>  %135 = bitcast i32 %134 to float
>  %136 = bitcast float %135 to i32
>  %137 = icmp ne i32 %136, 0
>  %138 = fptosi float %50 to i32
>  %139 = bitcast i32 %138 to float
>  %140 = bitcast float %139 to i32
>  %141 = shl i32 %140, 4
>  %142 = add i32 %141, 16
>  %143 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %142)
>  %144 = fmul float %26, %143
>  %145 = shl i32 %140, 4
>  %146 = add i32 %145, 16
>  %147 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %146)
>  br i1 %137, label %IF53, label %ELSE54
>
>IF53:                                             ; preds = %ENDIF45
>  %148 = fmul float %28, %147
>  %149 = fmul float %144, 0x3FE51EB860000000
>  %150 = fmul float %148, 0x3FE51EB860000000
>  br label %ENDIF52
>
>ELSE54:                                           ; preds = %ENDIF45
>  %151 = fmul float %27, %147
>  %152 = fmul float %144, 0x3FD51EB860000000
>  %153 = fmul float %151, 0x3FD51EB860000000
>  br label %ENDIF52
>
>ENDIF52:                                          ; preds = %ELSE54, %IF53
>  %temp12.0 = phi float [ %149, %IF53 ], [ %152, %ELSE54 ]
>  %temp13.0 = phi float [ %150, %IF53 ], [ %153, %ELSE54 ]
>  %154 = call float @fabs(float %129)
>  %155 = call float @fabs(float %129)
>  %156 = fsub float -0.000000e+00, %154
>  %157 = fcmp oge float %156, %155
>  %158 = sext i1 %157 to i32
>  %159 = bitcast i32 %158 to float
>  %160 = bitcast float %159 to i32
>  %161 = icmp ne i32 %160, 0
>  br i1 %161, label %IF60, label %ENDIF59
>
>IF60:                                             ; preds = %ENDIF52
>  %162 = fptosi float %50 to i32
>  %163 = bitcast i32 %162 to float
>  %164 = bitcast float %163 to i32
>  %165 = shl i32 %164, 4
>  %166 = add i32 %165, 16
>  %167 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %166)
>  %168 = fmul float %27, %167
>  %169 = shl i32 %164, 4
>  %170 = add i32 %169, 16
>  %171 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %170)
>  %172 = fmul float %28, %171
>  %173 = fmul float %168, 0x3FE51EB860000000
>  %174 = fmul float %172, 0x3FE51EB860000000
>  br label %ENDIF59
>
>ENDIF59:                                          ; preds = %ENDIF52, %IF60
>  %temp12.1 = phi float [ %173, %IF60 ], [ %temp12.0, %ENDIF52 ]
>  %temp13.1 = phi float [ %174, %IF60 ], [ %temp13.0, %ENDIF52 ]
>  %175 = fptosi float %51 to i32
>  %176 = bitcast i32 %175 to float
>  %177 = bitcast float %176 to i32
>  %178 = shl i32 %177, 4
>  %179 = add i32 %178, 20
>  %180 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %179)
>  %181 = fadd float -1.000000e+00, %180
>  %182 = shl i32 %177, 4
>  %183 = add i32 %182, 20
>  %184 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %183)
>  %185 = fadd float -2.000000e+00, %184
>  %186 = call float @fabs(float %181)
>  %187 = call float @fabs(float %181)
>  %188 = fsub float -0.000000e+00, %186
>  %189 = fcmp oge float %188, %187
>  %190 = sext i1 %189 to i32
>  %191 = bitcast i32 %190 to float
>  %192 = bitcast float %191 to i32
>  %193 = icmp ne i32 %192, 0
>  %194 = fptosi float %51 to i32
>  %195 = bitcast i32 %194 to float
>  %196 = bitcast float %195 to i32
>  %197 = shl i32 %196, 4
>  %198 = add i32 %197, 16
>  %199 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %198)
>  %200 = fmul float %26, %199
>  %201 = shl i32 %196, 4
>  %202 = add i32 %201, 16
>  %203 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %202)
>  br i1 %193, label %IF67, label %ELSE68
>
>IF67:                                             ; preds = %ENDIF59
>  %204 = fmul float %28, %203
>  %205 = fmul float %200, 0x3FE51EB860000000
>  %206 = fmul float %204, 0x3FE51EB860000000
>  br label %ENDIF66
>
>ELSE68:                                           ; preds = %ENDIF59
>  %207 = fmul float %27, %203
>  %208 = fmul float %200, 0x3FD51EB860000000
>  %209 = fmul float %207, 0x3FD51EB860000000
>  br label %ENDIF66
>
>ENDIF66:                                          ; preds = %ELSE68, %IF67
>  %temp16.0 = phi float [ %205, %IF67 ], [ %208, %ELSE68 ]
>  %temp17.0 = phi float [ %206, %IF67 ], [ %209, %ELSE68 ]
>  %210 = call float @fabs(float %185)
>  %211 = call float @fabs(float %185)
>  %212 = fsub float -0.000000e+00, %210
>  %213 = fcmp oge float %212, %211
>  %214 = sext i1 %213 to i32
>  %215 = bitcast i32 %214 to float
>  %216 = bitcast float %215 to i32
>  %217 = icmp ne i32 %216, 0
>  br i1 %217, label %IF74, label %ENDIF73
>
>IF74:                                             ; preds = %ENDIF66
>  %218 = fptosi float %51 to i32
>  %219 = bitcast i32 %218 to float
>  %220 = bitcast float %219 to i32
>  %221 = shl i32 %220, 4
>  %222 = add i32 %221, 16
>  %223 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %222)
>  %224 = fmul float %27, %223
>  %225 = shl i32 %220, 4
>  %226 = add i32 %225, 16
>  %227 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %226)
>  %228 = fmul float %28, %227
>  %229 = fmul float %224, 0x3FE51EB860000000
>  %230 = fmul float %228, 0x3FE51EB860000000
>  br label %ENDIF73
>
>ENDIF73:                                          ; preds = %ENDIF66, %IF74
>  %temp16.1 = phi float [ %229, %IF74 ], [ %temp16.0, %ENDIF66 ]
>  %temp17.1 = phi float [ %230, %IF74 ], [ %temp17.0, %ENDIF66 ]
>  %231 = fptosi float %52 to i32
>  %232 = bitcast i32 %231 to float
>  %233 = bitcast float %232 to i32
>  %234 = shl i32 %233, 4
>  %235 = add i32 %234, 20
>  %236 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %235)
>  %237 = fadd float -1.000000e+00, %236
>  %238 = shl i32 %233, 4
>  %239 = add i32 %238, 20
>  %240 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %239)
>  %241 = fadd float -2.000000e+00, %240
>  %242 = call float @fabs(float %237)
>  %243 = call float @fabs(float %237)
>  %244 = fsub float -0.000000e+00, %242
>  %245 = fcmp oge float %244, %243
>  %246 = sext i1 %245 to i32
>  %247 = bitcast i32 %246 to float
>  %248 = bitcast float %247 to i32
>  %249 = icmp ne i32 %248, 0
>  %250 = fptosi float %52 to i32
>  %251 = bitcast i32 %250 to float
>  %252 = bitcast float %251 to i32
>  %253 = shl i32 %252, 4
>  %254 = add i32 %253, 16
>  %255 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %254)
>  %256 = fmul float %26, %255
>  %257 = shl i32 %252, 4
>  %258 = add i32 %257, 16
>  %259 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %258)
>  br i1 %249, label %IF81, label %ELSE82
>
>IF81:                                             ; preds = %ENDIF73
>  %260 = fmul float %28, %259
>  %261 = fmul float %256, 0x3FE51EB860000000
>  %262 = fmul float %260, 0x3FE51EB860000000
>  br label %ENDIF80
>
>ELSE82:                                           ; preds = %ENDIF73
>  %263 = fmul float %27, %259
>  %264 = fmul float %256, 0x3FD51EB860000000
>  %265 = fmul float %263, 0x3FD51EB860000000
>  br label %ENDIF80
>
>ENDIF80:                                          ; preds = %ELSE82, %IF81
>  %temp20.0 = phi float [ %261, %IF81 ], [ %264, %ELSE82 ]
>  %temp21.0 = phi float [ %262, %IF81 ], [ %265, %ELSE82 ]
>  %266 = call float @fabs(float %241)
>  %267 = call float @fabs(float %241)
>  %268 = fsub float -0.000000e+00, %266
>  %269 = fcmp oge float %268, %267
>  %270 = sext i1 %269 to i32
>  %271 = bitcast i32 %270 to float
>  %272 = bitcast float %271 to i32
>  %273 = icmp ne i32 %272, 0
>  br i1 %273, label %IF88, label %ENDIF87
>
>IF88:                                             ; preds = %ENDIF80
>  %274 = fptosi float %52 to i32
>  %275 = bitcast i32 %274 to float
>  %276 = bitcast float %275 to i32
>  %277 = shl i32 %276, 4
>  %278 = add i32 %277, 16
>  %279 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %278)
>  %280 = fmul float %27, %279
>  %281 = shl i32 %276, 4
>  %282 = add i32 %281, 16
>  %283 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %282)
>  %284 = fmul float %28, %283
>  %285 = fmul float %280, 0x3FE51EB860000000
>  %286 = fmul float %284, 0x3FE51EB860000000
>  br label %ENDIF87
>
>ENDIF87:                                          ; preds = %ENDIF80, %IF88
>  %temp20.1 = phi float [ %285, %IF88 ], [ %temp20.0, %ENDIF80 ]
>  %temp21.1 = phi float [ %286, %IF88 ], [ %temp21.0, %ENDIF80 ]
>  %287 = fptosi float %57 to i32
>  %288 = bitcast i32 %287 to float
>  %289 = bitcast float %288 to i32
>  %290 = shl i32 %289, 4
>  %291 = add i32 %290, 20
>  %292 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %291)
>  %293 = fadd float -1.000000e+00, %292
>  %294 = shl i32 %289, 4
>  %295 = add i32 %294, 20
>  %296 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %295)
>  %297 = fadd float -2.000000e+00, %296
>  %298 = call float @fabs(float %293)
>  %299 = call float @fabs(float %293)
>  %300 = fsub float -0.000000e+00, %298
>  %301 = fcmp oge float %300, %299
>  %302 = sext i1 %301 to i32
>  %303 = bitcast i32 %302 to float
>  %304 = bitcast float %303 to i32
>  %305 = icmp ne i32 %304, 0
>  %306 = fptosi float %57 to i32
>  %307 = bitcast i32 %306 to float
>  %308 = bitcast float %307 to i32
>  %309 = shl i32 %308, 4
>  %310 = add i32 %309, 16
>  %311 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %310)
>  %312 = fmul float %26, %311
>  %313 = shl i32 %308, 4
>  %314 = add i32 %313, 16
>  %315 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %314)
>  br i1 %305, label %IF95, label %ELSE96
>
>IF95:                                             ; preds = %ENDIF87
>  %316 = fmul float %28, %315
>  %317 = fmul float %312, 0x3FE51EB860000000
>  %318 = fmul float %316, 0x3FE51EB860000000
>  br label %ENDIF94
>
>ELSE96:                                           ; preds = %ENDIF87
>  %319 = fmul float %27, %315
>  %320 = fmul float %312, 0x3FD51EB860000000
>  %321 = fmul float %319, 0x3FD51EB860000000
>  br label %ENDIF94
>
>ENDIF94:                                          ; preds = %ELSE96, %IF95
>  %temp24.0 = phi float [ %317, %IF95 ], [ %320, %ELSE96 ]
>  %temp25.0 = phi float [ %318, %IF95 ], [ %321, %ELSE96 ]
>  %322 = call float @fabs(float %297)
>  %323 = call float @fabs(float %297)
>  %324 = fsub float -0.000000e+00, %322
>  %325 = fcmp oge float %324, %323
>  %326 = sext i1 %325 to i32
>  %327 = bitcast i32 %326 to float
>  %328 = bitcast float %327 to i32
>  %329 = icmp ne i32 %328, 0
>  br i1 %329, label %IF102, label %ENDIF101
>
>IF102:                                            ; preds = %ENDIF94
>  %330 = fptosi float %57 to i32
>  %331 = bitcast i32 %330 to float
>  %332 = bitcast float %331 to i32
>  %333 = shl i32 %332, 4
>  %334 = add i32 %333, 16
>  %335 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %334)
>  %336 = fmul float %27, %335
>  %337 = shl i32 %332, 4
>  %338 = add i32 %337, 16
>  %339 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %338)
>  %340 = fmul float %28, %339
>  %341 = fmul float %336, 0x3FE51EB860000000
>  %342 = fmul float %340, 0x3FE51EB860000000
>  br label %ENDIF101
>
>ENDIF101:                                         ; preds = %ENDIF94, %IF102
>  %temp24.1 = phi float [ %341, %IF102 ], [ %temp24.0, %ENDIF94 ]
>  %temp25.1 = phi float [ %342, %IF102 ], [ %temp25.0, %ENDIF94 ]
>  %343 = fptosi float %58 to i32
>  %344 = bitcast i32 %343 to float
>  %345 = bitcast float %344 to i32
>  %346 = shl i32 %345, 4
>  %347 = add i32 %346, 20
>  %348 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %347)
>  %349 = fadd float -1.000000e+00, %348
>  %350 = shl i32 %345, 4
>  %351 = add i32 %350, 20
>  %352 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %351)
>  %353 = fadd float -2.000000e+00, %352
>  %354 = call float @fabs(float %349)
>  %355 = call float @fabs(float %349)
>  %356 = fsub float -0.000000e+00, %354
>  %357 = fcmp oge float %356, %355
>  %358 = sext i1 %357 to i32
>  %359 = bitcast i32 %358 to float
>  %360 = bitcast float %359 to i32
>  %361 = icmp ne i32 %360, 0
>  %362 = fptosi float %58 to i32
>  %363 = bitcast i32 %362 to float
>  %364 = bitcast float %363 to i32
>  %365 = shl i32 %364, 4
>  %366 = add i32 %365, 16
>  %367 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %366)
>  %368 = fmul float %26, %367
>  %369 = shl i32 %364, 4
>  %370 = add i32 %369, 16
>  %371 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %370)
>  br i1 %361, label %IF109, label %ELSE110
>
>IF109:                                            ; preds = %ENDIF101
>  %372 = fmul float %28, %371
>  %373 = fmul float %368, 0x3FE51EB860000000
>  %374 = fmul float %372, 0x3FE51EB860000000
>  br label %ENDIF108
>
>ELSE110:                                          ; preds = %ENDIF101
>  %375 = fmul float %27, %371
>  %376 = fmul float %368, 0x3FD51EB860000000
>  %377 = fmul float %375, 0x3FD51EB860000000
>  br label %ENDIF108
>
>ENDIF108:                                         ; preds = %ELSE110, %IF109
>  %temp26.0 = phi float [ %373, %IF109 ], [ %376, %ELSE110 ]
>  %temp27.0 = phi float [ %374, %IF109 ], [ %377, %ELSE110 ]
>  %378 = call float @fabs(float %353)
>  %379 = call float @fabs(float %353)
>  %380 = fsub float -0.000000e+00, %378
>  %381 = fcmp oge float %380, %379
>  %382 = sext i1 %381 to i32
>  %383 = bitcast i32 %382 to float
>  %384 = bitcast float %383 to i32
>  %385 = icmp ne i32 %384, 0
>  br i1 %385, label %IF116, label %ENDIF115
>
>IF116:                                            ; preds = %ENDIF108
>  %386 = fptosi float %58 to i32
>  %387 = bitcast i32 %386 to float
>  %388 = bitcast float %387 to i32
>  %389 = shl i32 %388, 4
>  %390 = add i32 %389, 16
>  %391 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %390)
>  %392 = fmul float %27, %391
>  %393 = shl i32 %388, 4
>  %394 = add i32 %393, 16
>  %395 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %394)
>  %396 = fmul float %28, %395
>  %397 = fmul float %392, 0x3FE51EB860000000
>  %398 = fmul float %396, 0x3FE51EB860000000
>  br label %ENDIF115
>
>ENDIF115:                                         ; preds = %ENDIF108, %IF116
>  %temp26.1 = phi float [ %397, %IF116 ], [ %temp26.0, %ENDIF108 ]
>  %temp27.1 = phi float [ %398, %IF116 ], [ %temp27.0, %ENDIF108 ]
>  %399 = fptosi float %59 to i32
>  %400 = bitcast i32 %399 to float
>  %401 = bitcast float %400 to i32
>  %402 = shl i32 %401, 4
>  %403 = add i32 %402, 20
>  %404 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %403)
>  %405 = fadd float -1.000000e+00, %404
>  %406 = shl i32 %401, 4
>  %407 = add i32 %406, 20
>  %408 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %407)
>  %409 = fadd float -2.000000e+00, %408
>  %410 = call float @fabs(float %405)
>  %411 = call float @fabs(float %405)
>  %412 = fsub float -0.000000e+00, %410
>  %413 = fcmp oge float %412, %411
>  %414 = sext i1 %413 to i32
>  %415 = bitcast i32 %414 to float
>  %416 = bitcast float %415 to i32
>  %417 = icmp ne i32 %416, 0
>  %418 = fptosi float %59 to i32
>  %419 = bitcast i32 %418 to float
>  %420 = bitcast float %419 to i32
>  %421 = shl i32 %420, 4
>  %422 = add i32 %421, 16
>  %423 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %422)
>  %424 = fmul float %26, %423
>  %425 = shl i32 %420, 4
>  %426 = add i32 %425, 16
>  %427 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %426)
>  br i1 %417, label %IF123, label %ELSE124
>
>IF123:                                            ; preds = %ENDIF115
>  %428 = fmul float %28, %427
>  %429 = fmul float %424, 0x3FE51EB860000000
>  %430 = fmul float %428, 0x3FE51EB860000000
>  br label %ENDIF122
>
>ELSE124:                                          ; preds = %ENDIF115
>  %431 = fmul float %27, %427
>  %432 = fmul float %424, 0x3FD51EB860000000
>  %433 = fmul float %431, 0x3FD51EB860000000
>  br label %ENDIF122
>
>ENDIF122:                                         ; preds = %ELSE124, %IF123
>  %temp32.0 = phi float [ %429, %IF123 ], [ %432, %ELSE124 ]
>  %temp33.0 = phi float [ %430, %IF123 ], [ %433, %ELSE124 ]
>  %434 = call float @fabs(float %409)
>  %435 = call float @fabs(float %409)
>  %436 = fsub float -0.000000e+00, %434
>  %437 = fcmp oge float %436, %435
>  %438 = sext i1 %437 to i32
>  %439 = bitcast i32 %438 to float
>  %440 = bitcast float %439 to i32
>  %441 = icmp ne i32 %440, 0
>  br i1 %441, label %IF130, label %ENDIF129
>
>IF130:                                            ; preds = %ENDIF122
>  %442 = fptosi float %59 to i32
>  %443 = bitcast i32 %442 to float
>  %444 = bitcast float %443 to i32
>  %445 = shl i32 %444, 4
>  %446 = add i32 %445, 16
>  %447 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %446)
>  %448 = fmul float %27, %447
>  %449 = shl i32 %444, 4
>  %450 = add i32 %449, 16
>  %451 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %450)
>  %452 = fmul float %28, %451
>  %453 = fmul float %448, 0x3FE51EB860000000
>  %454 = fmul float %452, 0x3FE51EB860000000
>  br label %ENDIF129
>
>ENDIF129:                                         ; preds = %ENDIF122, %IF130
>  %temp32.1 = phi float [ %453, %IF130 ], [ %temp32.0, %ENDIF122 ]
>  %temp33.1 = phi float [ %454, %IF130 ], [ %temp33.0, %ENDIF122 ]
>  %455 = fptosi float %60 to i32
>  %456 = bitcast i32 %455 to float
>  %457 = bitcast float %456 to i32
>  %458 = shl i32 %457, 4
>  %459 = add i32 %458, 20
>  %460 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %459)
>  %461 = fadd float -1.000000e+00, %460
>  %462 = shl i32 %457, 4
>  %463 = add i32 %462, 20
>  %464 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %463)
>  %465 = fadd float -2.000000e+00, %464
>  %466 = call float @fabs(float %461)
>  %467 = call float @fabs(float %461)
>  %468 = fsub float -0.000000e+00, %466
>  %469 = fcmp oge float %468, %467
>  %470 = sext i1 %469 to i32
>  %471 = bitcast i32 %470 to float
>  %472 = bitcast float %471 to i32
>  %473 = icmp ne i32 %472, 0
>  %474 = fptosi float %60 to i32
>  %475 = bitcast i32 %474 to float
>  %476 = bitcast float %475 to i32
>  %477 = shl i32 %476, 4
>  %478 = add i32 %477, 16
>  %479 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %478)
>  %480 = fmul float %26, %479
>  %481 = shl i32 %476, 4
>  %482 = add i32 %481, 16
>  %483 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %482)
>  br i1 %473, label %IF137, label %ELSE138
>
>IF137:                                            ; preds = %ENDIF129
>  %484 = fmul float %28, %483
>  %485 = fmul float %480, 0x3FE51EB860000000
>  %486 = fmul float %484, 0x3FE51EB860000000
>  br label %ENDIF136
>
>ELSE138:                                          ; preds = %ENDIF129
>  %487 = fmul float %27, %483
>  %488 = fmul float %480, 0x3FD51EB860000000
>  %489 = fmul float %487, 0x3FD51EB860000000
>  br label %ENDIF136
>
>ENDIF136:                                         ; preds = %ELSE138, %IF137
>  %temp4.0 = phi float [ %485, %IF137 ], [ %488, %ELSE138 ]
>  %temp6.0 = phi float [ %486, %IF137 ], [ %489, %ELSE138 ]
>  %490 = call float @fabs(float %465)
>  %491 = call float @fabs(float %465)
>  %492 = fsub float -0.000000e+00, %490
>  %493 = fcmp oge float %492, %491
>  %494 = sext i1 %493 to i32
>  %495 = bitcast i32 %494 to float
>  %496 = bitcast float %495 to i32
>  %497 = icmp ne i32 %496, 0
>  br i1 %497, label %IF144, label %ENDIF143
>
>IF144:                                            ; preds = %ENDIF136
>  %498 = fptosi float %60 to i32
>  %499 = bitcast i32 %498 to float
>  %500 = bitcast float %499 to i32
>  %501 = shl i32 %500, 4
>  %502 = add i32 %501, 16
>  %503 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %502)
>  %504 = fmul float %27, %503
>  %505 = shl i32 %500, 4
>  %506 = add i32 %505, 16
>  %507 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %506)
>  %508 = fmul float %28, %507
>  %509 = fmul float %504, 0x3FE51EB860000000
>  %510 = fmul float %508, 0x3FE51EB860000000
>  br label %ENDIF143
>
>ENDIF143:                                         ; preds = %ENDIF136, %IF144
>  %temp4.1 = phi float [ %509, %IF144 ], [ %temp4.0, %ENDIF136 ]
>  %temp6.1 = phi float [ %510, %IF144 ], [ %temp6.0, %ENDIF136 ]
>  %511 = fmul float %21, 1.000000e+00
>  %512 = fadd float %511, 0.000000e+00
>  %513 = fmul float %21, 0.000000e+00
>  %514 = fadd float %513, 1.000000e+00
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %temp8.1, float %temp9.1, float %33, float %34)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %temp12.1, float %temp13.1, float %35, float %36)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 34, i32 0, float %temp16.1, float %temp17.1, float %41, float %42)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 35, i32 0, float %temp20.1, float %temp21.1, float %43, float %44)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 36, i32 0, float %temp24.1, float %temp25.1, float 0.000000e+00, float 0.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 37, i32 0, float %temp26.1, float %temp27.1, float 0.000000e+00, float 0.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 38, i32 0, float %temp32.1, float %temp33.1, float 0.000000e+00, float 0.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 39, i32 0, float %temp4.1, float %temp6.1, float 0.000000e+00, float 0.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %61, float %62, float %512, float %514)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>; Function Attrs: readnone
>declare float @fabs(float) #2
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                               ; 4A00000A
>	s_load_dwordx4 s[4:7], s[8:9], 0x0                      ; C0820900
>	s_load_dwordx4 s[12:15], s[8:9], 0x4                    ; C0860904
>	s_load_dwordx4 s[16:19], s[8:9], 0x8                    ; C0880908
>	s_load_dwordx4 s[20:23], s[8:9], 0xc                    ; C08A090C
>	s_load_dwordx4 s[24:27], s[8:9], 0x10                   ; C08C0910
>	s_load_dwordx4 s[8:11], s[8:9], 0x14                    ; C0840914
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	buffer_load_format_xyzw v[28:31], v0, s[24:27], 0 idxen ; E00C2000 80061C00
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	v_cvt_i32_f32_e32 v1, v28                               ; 7E02111C
>	v_lshlrev_b32_e32 v20, 4, v1                            ; 34280284
>	v_add_i32_e32 v1, 16, v20                               ; 4A022890
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                      ; C0800300
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	buffer_load_dword v22, v1, s[0:3], 0 offen              ; E0301000 80001601
>	buffer_load_format_xyzw v[12:15], v0, s[12:15], 0 idxen ; E00C2000 80030C00
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	v_mul_f32_e32 v24, v22, v12                             ; 10301916
>	buffer_load_format_xyzw v[16:19], v0, s[8:11], 0 idxen  ; E00C2000 80021000
>	buffer_load_format_xyzw v[4:7], v0, s[20:23], 0 idxen   ; E00C2000 80050400
>	buffer_load_format_xyzw v[8:11], v0, s[16:19], 0 idxen  ; E00C2000 80040800
>	buffer_load_format_xyzw v[0:3], v0, s[4:7], 0 idxen     ; E00C2000 80010000
>	v_add_i32_e32 v20, 20, v20                              ; 4A282894
>	buffer_load_dword v23, v20, s[0:3], 0 offen             ; E0301000 80001714
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	v_add_f32_e32 v20, -1.0, v23                            ; 06282EF3
>	v_cmp_ge_f32_e64 s[4:5], -|v20|, |v20|                  ; D00C0304 20022914
>	v_cndmask_b32_e64 v20, 0, -1, s[4:5]                    ; D2000014 00118280
>	v_cmp_eq_i32_e64 s[4:5], v20, 0                         ; D1040004 00010114
>	s_and_saveexec_b64 s[4:5], s[4:5]                       ; BE842404
>	s_xor_b64 s[4:5], exec, s[4:5]                          ; 8984047E
>	v_mul_f32_e32 v20, 0x3ea8f5c3, v24                      ; 102830FF 3EA8F5C3
>	v_mul_f32_e32 v21, v22, v13                             ; 102A1B16
>	v_mul_f32_e32 v21, 0x3ea8f5c3, v21                      ; 102A2AFF 3EA8F5C3
>	s_or_saveexec_b64 s[4:5], s[4:5]                        ; BE842504
>	s_xor_b64 exec, exec, s[4:5]                            ; 89FE047E
>	v_mul_f32_e32 v20, 0x3f28f5c3, v24                      ; 102830FF 3F28F5C3
>	v_mul_f32_e32 v21, v22, v14                             ; 102A1D16
>	v_mul_f32_e32 v21, 0x3f28f5c3, v21                      ; 102A2AFF 3F28F5C3
>	s_or_b64 exec, exec, s[4:5]                             ; 88FE047E
>	v_add_f32_e32 v22, -2.0, v23                            ; 062C2EF5
>	v_cmp_ge_f32_e64 s[4:5], -|v22|, |v22|                  ; D00C0304 20022D16
>	v_cndmask_b32_e64 v22, 0, -1, s[4:5]                    ; D2000016 00118280
>	v_cmp_ne_i32_e64 s[4:5], v22, 0                         ; D10A0004 00010116
>	s_and_saveexec_b64 s[4:5], s[4:5]                       ; BE842404
>	s_xor_b64 s[4:5], exec, s[4:5]                          ; 8984047E
>	v_cvt_i32_f32_e32 v20, v28                              ; 7E28111C
>	v_lshlrev_b32_e32 v20, 4, v20                           ; 34282884
>	v_add_i32_e32 v20, 16, v20                              ; 4A282890
>	buffer_load_dword v20, v20, s[0:3], 0 offen             ; E0301000 80001414
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	v_mul_f32_e32 v21, v20, v14                             ; 102A1D14
>	v_mul_f32_e32 v21, 0x3f28f5c3, v21                      ; 102A2AFF 3F28F5C3
>	v_mul_f32_e32 v20, v20, v13                             ; 10281B14
>	v_mul_f32_e32 v20, 0x3f28f5c3, v20                      ; 102828FF 3F28F5C3
>	s_or_b64 exec, exec, s[4:5]                             ; 88FE047E
>	v_cvt_i32_f32_e32 v22, v29                              ; 7E2C111D
>	v_lshlrev_b32_e32 v22, 4, v22                           ; 342C2C84
>	v_add_i32_e32 v23, 16, v22                              ; 4A2E2C90
>	buffer_load_dword v25, v23, s[0:3], 0 offen             ; E0301000 80001917
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	v_mul_f32_e32 v26, v25, v12                             ; 10341919
>	v_add_i32_e32 v22, 20, v22                              ; 4A2C2C94
>	buffer_load_dword v24, v22, s[0:3], 0 offen             ; E0301000 80001816
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	v_add_f32_e32 v22, -1.0, v24                            ; 062C30F3
>	v_cmp_ge_f32_e64 s[4:5], -|v22|, |v22|                  ; D00C0304 20022D16
>	v_cndmask_b32_e64 v22, 0, -1, s[4:5]                    ; D2000016 00118280
>	v_cmp_eq_i32_e64 s[4:5], v22, 0                         ; D1040004 00010116
>	s_and_saveexec_b64 s[4:5], s[4:5]                       ; BE842404
>	s_xor_b64 s[4:5], exec, s[4:5]                          ; 8984047E
>	v_mul_f32_e32 v22, 0x3ea8f5c3, v26                      ; 102C34FF 3EA8F5C3
>	v_mul_f32_e32 v23, v25, v13                             ; 102E1B19
>	v_mul_f32_e32 v23, 0x3ea8f5c3, v23                      ; 102E2EFF 3EA8F5C3
>	s_or_saveexec_b64 s[4:5], s[4:5]                        ; BE842504
>	s_xor_b64 exec, exec, s[4:5]                            ; 89FE047E
>	v_mul_f32_e32 v22, 0x3f28f5c3, v26                      ; 102C34FF 3F28F5C3
>	v_mul_f32_e32 v23, v25, v14                             ; 102E1D19
>	v_mul_f32_e32 v23, 0x3f28f5c3, v23                      ; 102E2EFF 3F28F5C3
>	s_or_b64 exec, exec, s[4:5]                             ; 88FE047E
>	v_add_f32_e32 v24, -2.0, v24                            ; 063030F5
>	v_cmp_ge_f32_e64 s[4:5], -|v24|, |v24|                  ; D00C0304 20023118
>	v_cndmask_b32_e64 v24, 0, -1, s[4:5]                    ; D2000018 00118280
>	v_cmp_ne_i32_e64 s[4:5], v24, 0                         ; D10A0004 00010118
>	s_and_saveexec_b64 s[4:5], s[4:5]                       ; BE842404
>	s_xor_b64 s[4:5], exec, s[4:5]                          ; 8984047E
>	v_cvt_i32_f32_e32 v22, v29                              ; 7E2C111D
>	v_lshlrev_b32_e32 v22, 4, v22                           ; 342C2C84
>	v_add_i32_e32 v22, 16, v22                              ; 4A2C2C90
>	buffer_load_dword v22, v22, s[0:3], 0 offen             ; E0301000 80001616
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	v_mul_f32_e32 v23, v22, v14                             ; 102E1D16
>	v_mul_f32_e32 v23, 0x3f28f5c3, v23                      ; 102E2EFF 3F28F5C3
>	v_mul_f32_e32 v22, v22, v13                             ; 102C1B16
>	v_mul_f32_e32 v22, 0x3f28f5c3, v22                      ; 102C2CFF 3F28F5C3
>	s_or_b64 exec, exec, s[4:5]                             ; 88FE047E
>	v_cvt_i32_f32_e32 v24, v30                              ; 7E30111E
>	v_lshlrev_b32_e32 v24, 4, v24                           ; 34303084
>	v_add_i32_e32 v25, 16, v24                              ; 4A323090
>	buffer_load_dword v27, v25, s[0:3], 0 offen             ; E0301000 80001B19
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	v_mul_f32_e32 v32, v27, v12                             ; 1040191B
>	v_add_i32_e32 v24, 20, v24                              ; 4A303094
>	buffer_load_dword v26, v24, s[0:3], 0 offen             ; E0301000 80001A18
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	v_add_f32_e32 v24, -1.0, v26                            ; 063034F3
>	v_cmp_ge_f32_e64 s[4:5], -|v24|, |v24|                  ; D00C0304 20023118
>	v_cndmask_b32_e64 v24, 0, -1, s[4:5]                    ; D2000018 00118280
>	v_cmp_eq_i32_e64 s[4:5], v24, 0                         ; D1040004 00010118
>	s_and_saveexec_b64 s[4:5], s[4:5]                       ; BE842404
>	s_xor_b64 s[4:5], exec, s[4:5]                          ; 8984047E
>	v_mul_f32_e32 v24, 0x3ea8f5c3, v32                      ; 103040FF 3EA8F5C3
>	v_mul_f32_e32 v25, v27, v13                             ; 10321B1B
>	v_mul_f32_e32 v25, 0x3ea8f5c3, v25                      ; 103232FF 3EA8F5C3
>	s_or_saveexec_b64 s[4:5], s[4:5]                        ; BE842504
>	s_xor_b64 exec, exec, s[4:5]                            ; 89FE047E
>	v_mul_f32_e32 v24, 0x3f28f5c3, v32                      ; 103040FF 3F28F5C3
>	v_mul_f32_e32 v25, v27, v14                             ; 10321D1B
>	v_mul_f32_e32 v25, 0x3f28f5c3, v25                      ; 103232FF 3F28F5C3
>	s_or_b64 exec, exec, s[4:5]                             ; 88FE047E
>	v_add_f32_e32 v26, -2.0, v26                            ; 063434F5
>	v_cmp_ge_f32_e64 s[4:5], -|v26|, |v26|                  ; D00C0304 2002351A
>	v_cndmask_b32_e64 v26, 0, -1, s[4:5]                    ; D200001A 00118280
>	v_cmp_ne_i32_e64 s[4:5], v26, 0                         ; D10A0004 0001011A
>	s_and_saveexec_b64 s[4:5], s[4:5]                       ; BE842404
>	s_xor_b64 s[4:5], exec, s[4:5]                          ; 8984047E
>	v_cvt_i32_f32_e32 v24, v30                              ; 7E30111E
>	v_lshlrev_b32_e32 v24, 4, v24                           ; 34303084
>	v_add_i32_e32 v24, 16, v24                              ; 4A303090
>	buffer_load_dword v24, v24, s[0:3], 0 offen             ; E0301000 80001818
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	v_mul_f32_e32 v25, v24, v14                             ; 10321D18
>	v_mul_f32_e32 v25, 0x3f28f5c3, v25                      ; 103232FF 3F28F5C3
>	v_mul_f32_e32 v24, v24, v13                             ; 10301B18
>	v_mul_f32_e32 v24, 0x3f28f5c3, v24                      ; 103030FF 3F28F5C3
>	s_or_b64 exec, exec, s[4:5]                             ; 88FE047E
>	v_cvt_i32_f32_e32 v26, v31                              ; 7E34111F
>	v_lshlrev_b32_e32 v26, 4, v26                           ; 34343484
>	v_add_i32_e32 v27, 16, v26                              ; 4A363490
>	buffer_load_dword v33, v27, s[0:3], 0 offen             ; E0301000 8000211B
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	v_mul_f32_e32 v34, v33, v12                             ; 10441921
>	v_add_i32_e32 v26, 20, v26                              ; 4A343494
>	buffer_load_dword v32, v26, s[0:3], 0 offen             ; E0301000 8000201A
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	v_add_f32_e32 v26, -1.0, v32                            ; 063440F3
>	v_cmp_ge_f32_e64 s[4:5], -|v26|, |v26|                  ; D00C0304 2002351A
>	v_cndmask_b32_e64 v26, 0, -1, s[4:5]                    ; D200001A 00118280
>	v_cmp_eq_i32_e64 s[4:5], v26, 0                         ; D1040004 0001011A
>	s_and_saveexec_b64 s[4:5], s[4:5]                       ; BE842404
>	s_xor_b64 s[4:5], exec, s[4:5]                          ; 8984047E
>	v_mul_f32_e32 v26, 0x3ea8f5c3, v34                      ; 103444FF 3EA8F5C3
>	v_mul_f32_e32 v27, v33, v13                             ; 10361B21
>	v_mul_f32_e32 v27, 0x3ea8f5c3, v27                      ; 103636FF 3EA8F5C3
>	s_or_saveexec_b64 s[4:5], s[4:5]                        ; BE842504
>	s_xor_b64 exec, exec, s[4:5]                            ; 89FE047E
>	v_mul_f32_e32 v26, 0x3f28f5c3, v34                      ; 103444FF 3F28F5C3
>	v_mul_f32_e32 v27, v33, v14                             ; 10361D21
>	v_mul_f32_e32 v27, 0x3f28f5c3, v27                      ; 103636FF 3F28F5C3
>	s_or_b64 exec, exec, s[4:5]                             ; 88FE047E
>	v_add_f32_e32 v32, -2.0, v32                            ; 064040F5
>	v_cmp_ge_f32_e64 s[4:5], -|v32|, |v32|                  ; D00C0304 20024120
>	v_cndmask_b32_e64 v32, 0, -1, s[4:5]                    ; D2000020 00118280
>	v_cmp_ne_i32_e64 s[4:5], v32, 0                         ; D10A0004 00010120
>	s_and_saveexec_b64 s[4:5], s[4:5]                       ; BE842404
>	s_xor_b64 s[4:5], exec, s[4:5]                          ; 8984047E
>	v_cvt_i32_f32_e32 v26, v31                              ; 7E34111F
>	v_lshlrev_b32_e32 v26, 4, v26                           ; 34343484
>	v_add_i32_e32 v26, 16, v26                              ; 4A343490
>	buffer_load_dword v26, v26, s[0:3], 0 offen             ; E0301000 80001A1A
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	v_mul_f32_e32 v27, v26, v14                             ; 10361D1A
>	v_mul_f32_e32 v27, 0x3f28f5c3, v27                      ; 103636FF 3F28F5C3
>	v_mul_f32_e32 v26, v26, v13                             ; 10341B1A
>	v_mul_f32_e32 v26, 0x3f28f5c3, v26                      ; 103434FF 3F28F5C3
>	s_or_b64 exec, exec, s[4:5]                             ; 88FE047E
>	v_cvt_i32_f32_e32 v28, v16                              ; 7E381110
>	v_lshlrev_b32_e32 v28, 4, v28                           ; 34383884
>	v_add_i32_e32 v29, 16, v28                              ; 4A3A3890
>	buffer_load_dword v31, v29, s[0:3], 0 offen             ; E0301000 80001F1D
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	v_mul_f32_e32 v32, v31, v12                             ; 1040191F
>	v_add_i32_e32 v28, 20, v28                              ; 4A383894
>	buffer_load_dword v30, v28, s[0:3], 0 offen             ; E0301000 80001E1C
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	v_add_f32_e32 v28, -1.0, v30                            ; 06383CF3
>	v_cmp_ge_f32_e64 s[4:5], -|v28|, |v28|                  ; D00C0304 2002391C
>	v_cndmask_b32_e64 v28, 0, -1, s[4:5]                    ; D200001C 00118280
>	v_cmp_eq_i32_e64 s[4:5], v28, 0                         ; D1040004 0001011C
>	s_and_saveexec_b64 s[4:5], s[4:5]                       ; BE842404
>	s_xor_b64 s[4:5], exec, s[4:5]                          ; 8984047E
>	v_mul_f32_e32 v28, 0x3ea8f5c3, v32                      ; 103840FF 3EA8F5C3
>	v_mul_f32_e32 v29, v31, v13                             ; 103A1B1F
>	v_mul_f32_e32 v29, 0x3ea8f5c3, v29                      ; 103A3AFF 3EA8F5C3
>	s_or_saveexec_b64 s[4:5], s[4:5]                        ; BE842504
>	s_xor_b64 exec, exec, s[4:5]                            ; 89FE047E
>	v_mul_f32_e32 v28, 0x3f28f5c3, v32                      ; 103840FF 3F28F5C3
>	v_mul_f32_e32 v29, v31, v14                             ; 103A1D1F
>	v_mul_f32_e32 v29, 0x3f28f5c3, v29                      ; 103A3AFF 3F28F5C3
>	s_or_b64 exec, exec, s[4:5]                             ; 88FE047E
>	v_add_f32_e32 v30, -2.0, v30                            ; 063C3CF5
>	v_cmp_ge_f32_e64 s[4:5], -|v30|, |v30|                  ; D00C0304 20023D1E
>	v_cndmask_b32_e64 v30, 0, -1, s[4:5]                    ; D200001E 00118280
>	v_cmp_ne_i32_e64 s[4:5], v30, 0                         ; D10A0004 0001011E
>	s_and_saveexec_b64 s[4:5], s[4:5]                       ; BE842404
>	s_xor_b64 s[4:5], exec, s[4:5]                          ; 8984047E
>	v_cvt_i32_f32_e32 v28, v16                              ; 7E381110
>	v_lshlrev_b32_e32 v28, 4, v28                           ; 34383884
>	v_add_i32_e32 v28, 16, v28                              ; 4A383890
>	buffer_load_dword v28, v28, s[0:3], 0 offen             ; E0301000 80001C1C
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	v_mul_f32_e32 v29, v28, v14                             ; 103A1D1C
>	v_mul_f32_e32 v29, 0x3f28f5c3, v29                      ; 103A3AFF 3F28F5C3
>	v_mul_f32_e32 v28, v28, v13                             ; 10381B1C
>	v_mul_f32_e32 v28, 0x3f28f5c3, v28                      ; 103838FF 3F28F5C3
>	s_or_b64 exec, exec, s[4:5]                             ; 88FE047E
>	v_cvt_i32_f32_e32 v30, v17                              ; 7E3C1111
>	v_lshlrev_b32_e32 v30, 4, v30                           ; 343C3C84
>	v_add_i32_e32 v31, 16, v30                              ; 4A3E3C90
>	buffer_load_dword v33, v31, s[0:3], 0 offen             ; E0301000 8000211F
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	v_mul_f32_e32 v34, v33, v12                             ; 10441921
>	v_add_i32_e32 v30, 20, v30                              ; 4A3C3C94
>	buffer_load_dword v32, v30, s[0:3], 0 offen             ; E0301000 8000201E
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	v_add_f32_e32 v30, -1.0, v32                            ; 063C40F3
>	v_cmp_ge_f32_e64 s[4:5], -|v30|, |v30|                  ; D00C0304 20023D1E
>	v_cndmask_b32_e64 v30, 0, -1, s[4:5]                    ; D200001E 00118280
>	v_cmp_eq_i32_e64 s[4:5], v30, 0                         ; D1040004 0001011E
>	s_and_saveexec_b64 s[4:5], s[4:5]                       ; BE842404
>	s_xor_b64 s[4:5], exec, s[4:5]                          ; 8984047E
>	v_mul_f32_e32 v30, 0x3ea8f5c3, v34                      ; 103C44FF 3EA8F5C3
>	v_mul_f32_e32 v31, v33, v13                             ; 103E1B21
>	v_mul_f32_e32 v31, 0x3ea8f5c3, v31                      ; 103E3EFF 3EA8F5C3
>	s_or_saveexec_b64 s[4:5], s[4:5]                        ; BE842504
>	s_xor_b64 exec, exec, s[4:5]                            ; 89FE047E
>	v_mul_f32_e32 v30, 0x3f28f5c3, v34                      ; 103C44FF 3F28F5C3
>	v_mul_f32_e32 v31, v33, v14                             ; 103E1D21
>	v_mul_f32_e32 v31, 0x3f28f5c3, v31                      ; 103E3EFF 3F28F5C3
>	s_or_b64 exec, exec, s[4:5]                             ; 88FE047E
>	v_add_f32_e32 v32, -2.0, v32                            ; 064040F5
>	v_cmp_ge_f32_e64 s[4:5], -|v32|, |v32|                  ; D00C0304 20024120
>	v_cndmask_b32_e64 v32, 0, -1, s[4:5]                    ; D2000020 00118280
>	v_cmp_ne_i32_e64 s[4:5], v32, 0                         ; D10A0004 00010120
>	s_and_saveexec_b64 s[4:5], s[4:5]                       ; BE842404
>	s_xor_b64 s[4:5], exec, s[4:5]                          ; 8984047E
>	v_cvt_i32_f32_e32 v30, v17                              ; 7E3C1111
>	v_lshlrev_b32_e32 v30, 4, v30                           ; 343C3C84
>	v_add_i32_e32 v30, 16, v30                              ; 4A3C3C90
>	buffer_load_dword v30, v30, s[0:3], 0 offen             ; E0301000 80001E1E
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	v_mul_f32_e32 v31, v30, v14                             ; 103E1D1E
>	v_mul_f32_e32 v31, 0x3f28f5c3, v31                      ; 103E3EFF 3F28F5C3
>	v_mul_f32_e32 v30, v30, v13                             ; 103C1B1E
>	v_mul_f32_e32 v30, 0x3f28f5c3, v30                      ; 103C3CFF 3F28F5C3
>	s_or_b64 exec, exec, s[4:5]                             ; 88FE047E
>	v_cvt_i32_f32_e32 v32, v18                              ; 7E401112
>	v_lshlrev_b32_e32 v32, 4, v32                           ; 34404084
>	v_add_i32_e32 v33, 16, v32                              ; 4A424090
>	buffer_load_dword v35, v33, s[0:3], 0 offen             ; E0301000 80002321
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	v_mul_f32_e32 v36, v35, v12                             ; 10481923
>	v_add_i32_e32 v32, 20, v32                              ; 4A404094
>	buffer_load_dword v34, v32, s[0:3], 0 offen             ; E0301000 80002220
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	v_add_f32_e32 v32, -1.0, v34                            ; 064044F3
>	v_cmp_ge_f32_e64 s[4:5], -|v32|, |v32|                  ; D00C0304 20024120
>	v_cndmask_b32_e64 v32, 0, -1, s[4:5]                    ; D2000020 00118280
>	v_cmp_eq_i32_e64 s[4:5], v32, 0                         ; D1040004 00010120
>	s_and_saveexec_b64 s[4:5], s[4:5]                       ; BE842404
>	s_xor_b64 s[4:5], exec, s[4:5]                          ; 8984047E
>	v_mul_f32_e32 v32, 0x3ea8f5c3, v36                      ; 104048FF 3EA8F5C3
>	v_mul_f32_e32 v33, v35, v13                             ; 10421B23
>	v_mul_f32_e32 v33, 0x3ea8f5c3, v33                      ; 104242FF 3EA8F5C3
>	s_or_saveexec_b64 s[4:5], s[4:5]                        ; BE842504
>	s_xor_b64 exec, exec, s[4:5]                            ; 89FE047E
>	v_mul_f32_e32 v32, 0x3f28f5c3, v36                      ; 104048FF 3F28F5C3
>	v_mul_f32_e32 v33, v35, v14                             ; 10421D23
>	v_mul_f32_e32 v33, 0x3f28f5c3, v33                      ; 104242FF 3F28F5C3
>	s_or_b64 exec, exec, s[4:5]                             ; 88FE047E
>	v_add_f32_e32 v34, -2.0, v34                            ; 064444F5
>	v_cmp_ge_f32_e64 s[4:5], -|v34|, |v34|                  ; D00C0304 20024522
>	v_cndmask_b32_e64 v34, 0, -1, s[4:5]                    ; D2000022 00118280
>	v_cmp_ne_i32_e64 s[4:5], v34, 0                         ; D10A0004 00010122
>	s_and_saveexec_b64 s[4:5], s[4:5]                       ; BE842404
>	s_xor_b64 s[4:5], exec, s[4:5]                          ; 8984047E
>	v_cvt_i32_f32_e32 v32, v18                              ; 7E401112
>	v_lshlrev_b32_e32 v32, 4, v32                           ; 34404084
>	v_add_i32_e32 v32, 16, v32                              ; 4A404090
>	buffer_load_dword v32, v32, s[0:3], 0 offen             ; E0301000 80002020
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	v_mul_f32_e32 v33, v32, v14                             ; 10421D20
>	v_mul_f32_e32 v33, 0x3f28f5c3, v33                      ; 104242FF 3F28F5C3
>	v_mul_f32_e32 v32, v32, v13                             ; 10401B20
>	v_mul_f32_e32 v32, 0x3f28f5c3, v32                      ; 104040FF 3F28F5C3
>	s_or_b64 exec, exec, s[4:5]                             ; 88FE047E
>	v_cvt_i32_f32_e32 v34, v19                              ; 7E441113
>	v_lshlrev_b32_e32 v34, 4, v34                           ; 34444484
>	v_add_i32_e32 v35, 16, v34                              ; 4A464490
>	buffer_load_dword v37, v35, s[0:3], 0 offen             ; E0301000 80002523
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	v_mul_f32_e32 v38, v37, v12                             ; 104C1925
>	v_add_i32_e32 v34, 20, v34                              ; 4A444494
>	buffer_load_dword v36, v34, s[0:3], 0 offen             ; E0301000 80002422
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	v_add_f32_e32 v34, -1.0, v36                            ; 064448F3
>	v_cmp_ge_f32_e64 s[4:5], -|v34|, |v34|                  ; D00C0304 20024522
>	v_cndmask_b32_e64 v34, 0, -1, s[4:5]                    ; D2000022 00118280
>	v_cmp_eq_i32_e64 s[4:5], v34, 0                         ; D1040004 00010122
>	s_and_saveexec_b64 s[6:7], s[4:5]                       ; BE862404
>	s_xor_b64 s[6:7], exec, s[6:7]                          ; 8986067E
>	v_mul_f32_e32 v34, 0x3ea8f5c3, v38                      ; 10444CFF 3EA8F5C3
>	v_mul_f32_e32 v35, v37, v13                             ; 10461B25
>	v_mul_f32_e32 v35, 0x3ea8f5c3, v35                      ; 104646FF 3EA8F5C3
>	s_or_saveexec_b64 s[6:7], s[6:7]                        ; BE862506
>	s_xor_b64 exec, exec, s[6:7]                            ; 89FE067E
>	v_mul_f32_e32 v34, 0x3f28f5c3, v38                      ; 10444CFF 3F28F5C3
>	v_mul_f32_e32 v35, v37, v14                             ; 10461D25
>	v_mul_f32_e32 v35, 0x3f28f5c3, v35                      ; 104646FF 3F28F5C3
>	s_or_b64 exec, exec, s[6:7]                             ; 88FE067E
>	s_buffer_load_dword s5, s[0:3], 0x1                     ; C2028101
>	s_buffer_load_dword s4, s[0:3], 0x0                     ; C2020100
>	v_add_f32_e32 v36, -2.0, v36                            ; 064848F5
>	v_cmp_ge_f32_e64 s[6:7], -|v36|, |v36|                  ; D00C0306 20024924
>	v_cndmask_b32_e64 v36, 0, -1, s[6:7]                    ; D2000024 00198280
>	v_cmp_ne_i32_e64 s[6:7], v36, 0                         ; D10A0006 00010124
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	s_and_saveexec_b64 s[6:7], s[6:7]                       ; BE862406
>	s_xor_b64 s[6:7], exec, s[6:7]                          ; 8986067E
>	v_cvt_i32_f32_e32 v16, v19                              ; 7E201113
>	v_lshlrev_b32_e32 v16, 4, v16                           ; 34202084
>	v_add_i32_e32 v16, 16, v16                              ; 4A202090
>	buffer_load_dword v16, v16, s[0:3], 0 offen             ; E0301000 80001010
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	v_mul_f32_e32 v17, v16, v14                             ; 10221D10
>	v_mul_f32_e32 v35, 0x3f28f5c3, v17                      ; 104622FF 3F28F5C3
>	v_mul_f32_e32 v12, v16, v13                             ; 10181B10
>	v_mul_f32_e32 v34, 0x3f28f5c3, v12                      ; 104418FF 3F28F5C3
>	s_or_b64 exec, exec, s[6:7]                             ; 88FE067E
>	exp 15, 32, 0, 0, 0, v20, v21, v8, v9                   ; F800020F 09081514
>	exp 15, 33, 0, 0, 0, v22, v23, v10, v11                 ; F800021F 0B0A1716
>	exp 15, 34, 0, 0, 0, v24, v25, v4, v5                   ; F800022F 05041918
>	exp 15, 35, 0, 0, 0, v26, v27, v6, v7                   ; F800023F 07061B1A
>	s_waitcnt expcnt(0)                                     ; BF8C070F
>	v_mov_b32_e32 v4, 0                                     ; 7E080280
>	exp 15, 36, 0, 0, 0, v28, v29, v4, v4                   ; F800024F 04041D1C
>	exp 15, 37, 0, 0, 0, v30, v31, v4, v4                   ; F800025F 04041F1E
>	exp 15, 38, 0, 0, 0, v32, v33, v4, v4                   ; F800026F 04042120
>	exp 15, 39, 0, 0, 0, v34, v35, v4, v4                   ; F800027F 04042322
>	s_waitcnt expcnt(0)                                     ; BF8C070F
>	v_mad_f32 v4, 0, v2, 1.0                                ; D2820004 03CA0480
>	v_add_f32_e32 v5, 0, v2                                 ; 060A0480
>	v_add_f32_e32 v6, s5, v1                                ; 060C0205
>	v_add_f32_e32 v0, s4, v0                                ; 06000004
>	exp 15, 12, 0, 1, 0, v0, v6, v5, v4                     ; F80008CF 04050600
>	s_endpgm                                                ; BF810000
>
>SGPRS: 32
>VGPRS: 40
>Code Size: 1972 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL IN[1], GENERIC[10], PERSPECTIVE
>DCL IN[2], GENERIC[11], PERSPECTIVE
>DCL IN[3], GENERIC[12], PERSPECTIVE
>DCL IN[4], GENERIC[13], PERSPECTIVE
>DCL IN[5], GENERIC[14], PERSPECTIVE
>DCL IN[6], GENERIC[15], PERSPECTIVE
>DCL IN[7], GENERIC[16], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL OUT[1], COLOR[1]
>DCL SAMP[0]
>DCL SAMP[1]
>DCL SAMP[2]
>DCL SAMP[3]
>DCL SAMP[4]
>DCL SAMP[5]
>DCL SAMP[6]
>DCL SAMP[7]
>DCL SAMP[8]
>DCL SAMP[9]
>DCL SAMP[10]
>DCL SAMP[11]
>DCL SAMP[12]
>DCL SAMP[13]
>DCL SAMP[14]
>DCL SAMP[15]
>DCL TEMP[0..7], LOCAL
>IMM[0] FLT32 {    1.0000,     0.0000,     0.0000,     0.0000}
>  0: MOV TEMP[0].xy, IN[1].xyyy
>  1: TEX TEMP[0], TEMP[0], SAMP[1], 2D
>  2: MUL TEMP[0], TEMP[0], TEMP[0]
>  3: MOV TEMP[1].y, TEMP[0].wwww
>  4: MOV TEMP[2].xy, IN[0].xyyy
>  5: TEX TEMP[2], TEMP[2], SAMP[0], 2D
>  6: MUL TEMP[2], TEMP[2], TEMP[2]
>  7: MOV TEMP[1].x, TEMP[2].wwww
>  8: MOV TEMP[3].xy, IN[2].xyyy
>  9: TEX TEMP[3], TEMP[3], SAMP[2], 2D
> 10: MUL TEMP[3], TEMP[3], TEMP[3]
> 11: MOV TEMP[1].z, TEMP[3].wwww
> 12: MOV TEMP[4].xy, IN[3].xyyy
> 13: TEX TEMP[4], TEMP[4], SAMP[3], 2D
> 14: MUL TEMP[4], TEMP[4], TEMP[4]
> 15: MOV TEMP[1].w, TEMP[4].wwww
> 16: MUL TEMP[1], TEMP[1], TEMP[1]
> 17: MUL TEMP[1], TEMP[1], TEMP[1]
> 18: MOV TEMP[5].xy, IN[0].zwzz
> 19: MOV TEMP[5].zw, IN[1].wwzw
> 20: MUL TEMP[1], TEMP[1], TEMP[5]
> 21: MUL TEMP[0], TEMP[0], TEMP[1].yyyy
> 22: MAD TEMP[0], TEMP[1].xxxx, TEMP[2], TEMP[0]
> 23: MAD TEMP[0], TEMP[1].zzzz, TEMP[3], TEMP[0]
> 24: MAD TEMP[0], TEMP[1].wwww, TEMP[4], TEMP[0]
> 25: MOV TEMP[6].xy, IN[4].xyyy
> 26: TEX TEMP[6], TEMP[6], SAMP[4], 2D
> 27: MUL TEMP[2], TEMP[6], TEMP[6]
> 28: MOV TEMP[3].x, TEMP[2].wwww
> 29: MOV TEMP[6].xy, IN[5].xyyy
> 30: TEX TEMP[6], TEMP[6], SAMP[5], 2D
> 31: MUL TEMP[4], TEMP[6], TEMP[6]
> 32: MOV TEMP[3].y, TEMP[4].wwww
> 33: MOV TEMP[6].xy, IN[6].xyyy
> 34: TEX TEMP[6], TEMP[6], SAMP[6], 2D
> 35: MUL TEMP[5], TEMP[6], TEMP[6]
> 36: MOV TEMP[3].z, TEMP[5].wwww
> 37: MOV TEMP[6].xy, IN[7].xyyy
> 38: TEX TEMP[6], TEMP[6], SAMP[7], 2D
> 39: MUL TEMP[6], TEMP[6], TEMP[6]
> 40: MOV TEMP[3].w, TEMP[6].wwww
> 41: MUL TEMP[3], TEMP[3], TEMP[3]
> 42: MUL TEMP[3], TEMP[3], TEMP[3]
> 43: MOV TEMP[7].xy, IN[2].zwzz
> 44: MOV TEMP[7].zw, IN[3].wwzw
> 45: MUL TEMP[3], TEMP[3], TEMP[7]
> 46: MAD TEMP[0], TEMP[3].xxxx, TEMP[2], TEMP[0]
> 47: MAD TEMP[0], TEMP[3].yyyy, TEMP[4], TEMP[0]
> 48: MAD TEMP[0], TEMP[3].zzzz, TEMP[5], TEMP[0]
> 49: MAD TEMP[0], TEMP[3].wwww, TEMP[6], TEMP[0]
> 50: DP4 TEMP[4].x, TEMP[3], IMM[0].xxxx
> 51: DP4 TEMP[5].x, TEMP[1], IMM[0].xxxx
> 52: ADD TEMP[4].x, TEMP[4].xxxx, TEMP[5].xxxx
> 53: RCP TEMP[2].x, TEMP[4].xxxx
> 54: MUL TEMP[0], TEMP[0], TEMP[2].xxxx
> 55: MOV TEMP[4].xy, IN[1].xyyy
> 56: TEX TEMP[4].x, TEMP[4], SAMP[9], 2D
> 57: MUL TEMP[4].x, TEMP[4].xxxx, TEMP[1].yyyy
> 58: MOV TEMP[5].xy, IN[0].xyyy
> 59: TEX TEMP[5].x, TEMP[5], SAMP[8], 2D
> 60: MAD TEMP[4].x, TEMP[1].xxxx, TEMP[5].xxxx, TEMP[4].xxxx
> 61: MOV TEMP[5].xy, IN[2].xyyy
> 62: TEX TEMP[5].x, TEMP[5], SAMP[10], 2D
> 63: MAD TEMP[4].x, TEMP[1].zzzz, TEMP[5].xxxx, TEMP[4].xxxx
> 64: MOV TEMP[5].xy, IN[3].xyyy
> 65: TEX TEMP[5].x, TEMP[5], SAMP[11], 2D
> 66: MAD TEMP[1].x, TEMP[1].wwww, TEMP[5].xxxx, TEMP[4].xxxx
> 67: MOV TEMP[4].xy, IN[4].xyyy
> 68: TEX TEMP[4].x, TEMP[4], SAMP[12], 2D
> 69: MAD TEMP[1].x, TEMP[3].xxxx, TEMP[4].xxxx, TEMP[1].xxxx
> 70: MOV TEMP[4].xy, IN[5].xyyy
> 71: TEX TEMP[4].x, TEMP[4], SAMP[13], 2D
> 72: MAD TEMP[1].x, TEMP[3].yyyy, TEMP[4].xxxx, TEMP[1].xxxx
> 73: MOV TEMP[4].xy, IN[6].xyyy
> 74: TEX TEMP[4].x, TEMP[4], SAMP[14], 2D
> 75: MAD TEMP[1].x, TEMP[3].zzzz, TEMP[4].xxxx, TEMP[1].xxxx
> 76: MOV TEMP[4].xy, IN[7].xyyy
> 77: TEX TEMP[4].x, TEMP[4], SAMP[15], 2D
> 78: MAD TEMP[1].x, TEMP[3].wwww, TEMP[4].xxxx, TEMP[1].xxxx
> 79: MUL TEMP[1], TEMP[2].xxxx, TEMP[1].xxxx
> 80: MOV OUT[1], TEMP[1]
> 81: MOV OUT[0], TEMP[0]
> 82: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %23 = load <8 x i32> addrspace(2)* %22, !tbaa !0
>  %24 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %25 = load <4 x i32> addrspace(2)* %24, !tbaa !0
>  %26 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 1
>  %27 = load <8 x i32> addrspace(2)* %26, !tbaa !0
>  %28 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 1
>  %29 = load <4 x i32> addrspace(2)* %28, !tbaa !0
>  %30 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 2
>  %31 = load <8 x i32> addrspace(2)* %30, !tbaa !0
>  %32 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 2
>  %33 = load <4 x i32> addrspace(2)* %32, !tbaa !0
>  %34 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 3
>  %35 = load <8 x i32> addrspace(2)* %34, !tbaa !0
>  %36 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 3
>  %37 = load <4 x i32> addrspace(2)* %36, !tbaa !0
>  %38 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 4
>  %39 = load <8 x i32> addrspace(2)* %38, !tbaa !0
>  %40 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 4
>  %41 = load <4 x i32> addrspace(2)* %40, !tbaa !0
>  %42 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 5
>  %43 = load <8 x i32> addrspace(2)* %42, !tbaa !0
>  %44 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 5
>  %45 = load <4 x i32> addrspace(2)* %44, !tbaa !0
>  %46 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 6
>  %47 = load <8 x i32> addrspace(2)* %46, !tbaa !0
>  %48 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 6
>  %49 = load <4 x i32> addrspace(2)* %48, !tbaa !0
>  %50 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 7
>  %51 = load <8 x i32> addrspace(2)* %50, !tbaa !0
>  %52 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 7
>  %53 = load <4 x i32> addrspace(2)* %52, !tbaa !0
>  %54 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 8
>  %55 = load <8 x i32> addrspace(2)* %54, !tbaa !0
>  %56 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 8
>  %57 = load <4 x i32> addrspace(2)* %56, !tbaa !0
>  %58 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 9
>  %59 = load <8 x i32> addrspace(2)* %58, !tbaa !0
>  %60 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 9
>  %61 = load <4 x i32> addrspace(2)* %60, !tbaa !0
>  %62 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 10
>  %63 = load <8 x i32> addrspace(2)* %62, !tbaa !0
>  %64 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 10
>  %65 = load <4 x i32> addrspace(2)* %64, !tbaa !0
>  %66 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 11
>  %67 = load <8 x i32> addrspace(2)* %66, !tbaa !0
>  %68 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 11
>  %69 = load <4 x i32> addrspace(2)* %68, !tbaa !0
>  %70 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 12
>  %71 = load <8 x i32> addrspace(2)* %70, !tbaa !0
>  %72 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 12
>  %73 = load <4 x i32> addrspace(2)* %72, !tbaa !0
>  %74 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 13
>  %75 = load <8 x i32> addrspace(2)* %74, !tbaa !0
>  %76 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 13
>  %77 = load <4 x i32> addrspace(2)* %76, !tbaa !0
>  %78 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 14
>  %79 = load <8 x i32> addrspace(2)* %78, !tbaa !0
>  %80 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 14
>  %81 = load <4 x i32> addrspace(2)* %80, !tbaa !0
>  %82 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 15
>  %83 = load <8 x i32> addrspace(2)* %82, !tbaa !0
>  %84 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 15
>  %85 = load <4 x i32> addrspace(2)* %84, !tbaa !0
>  %86 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %87 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %88 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %5, <2 x i32> %7)
>  %89 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %5, <2 x i32> %7)
>  %90 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %5, <2 x i32> %7)
>  %91 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %5, <2 x i32> %7)
>  %92 = call float @llvm.SI.fs.interp(i32 2, i32 1, i32 %5, <2 x i32> %7)
>  %93 = call float @llvm.SI.fs.interp(i32 3, i32 1, i32 %5, <2 x i32> %7)
>  %94 = call float @llvm.SI.fs.interp(i32 0, i32 2, i32 %5, <2 x i32> %7)
>  %95 = call float @llvm.SI.fs.interp(i32 1, i32 2, i32 %5, <2 x i32> %7)
>  %96 = call float @llvm.SI.fs.interp(i32 2, i32 2, i32 %5, <2 x i32> %7)
>  %97 = call float @llvm.SI.fs.interp(i32 3, i32 2, i32 %5, <2 x i32> %7)
>  %98 = call float @llvm.SI.fs.interp(i32 0, i32 3, i32 %5, <2 x i32> %7)
>  %99 = call float @llvm.SI.fs.interp(i32 1, i32 3, i32 %5, <2 x i32> %7)
>  %100 = call float @llvm.SI.fs.interp(i32 2, i32 3, i32 %5, <2 x i32> %7)
>  %101 = call float @llvm.SI.fs.interp(i32 3, i32 3, i32 %5, <2 x i32> %7)
>  %102 = call float @llvm.SI.fs.interp(i32 0, i32 4, i32 %5, <2 x i32> %7)
>  %103 = call float @llvm.SI.fs.interp(i32 1, i32 4, i32 %5, <2 x i32> %7)
>  %104 = call float @llvm.SI.fs.interp(i32 0, i32 5, i32 %5, <2 x i32> %7)
>  %105 = call float @llvm.SI.fs.interp(i32 1, i32 5, i32 %5, <2 x i32> %7)
>  %106 = call float @llvm.SI.fs.interp(i32 0, i32 6, i32 %5, <2 x i32> %7)
>  %107 = call float @llvm.SI.fs.interp(i32 1, i32 6, i32 %5, <2 x i32> %7)
>  %108 = call float @llvm.SI.fs.interp(i32 0, i32 7, i32 %5, <2 x i32> %7)
>  %109 = call float @llvm.SI.fs.interp(i32 1, i32 7, i32 %5, <2 x i32> %7)
>  %110 = bitcast float %90 to i32
>  %111 = bitcast float %91 to i32
>  %112 = insertelement <2 x i32> undef, i32 %110, i32 0
>  %113 = insertelement <2 x i32> %112, i32 %111, i32 1
>  %114 = bitcast <8 x i32> %27 to <32 x i8>
>  %115 = bitcast <4 x i32> %29 to <16 x i8>
>  %116 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %113, <32 x i8> %114, <16 x i8> %115, i32 2)
>  %117 = extractelement <4 x float> %116, i32 0
>  %118 = extractelement <4 x float> %116, i32 1
>  %119 = extractelement <4 x float> %116, i32 2
>  %120 = extractelement <4 x float> %116, i32 3
>  %121 = fmul float %117, %117
>  %122 = fmul float %118, %118
>  %123 = fmul float %119, %119
>  %124 = fmul float %120, %120
>  %125 = bitcast float %86 to i32
>  %126 = bitcast float %87 to i32
>  %127 = insertelement <2 x i32> undef, i32 %125, i32 0
>  %128 = insertelement <2 x i32> %127, i32 %126, i32 1
>  %129 = bitcast <8 x i32> %23 to <32 x i8>
>  %130 = bitcast <4 x i32> %25 to <16 x i8>
>  %131 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %128, <32 x i8> %129, <16 x i8> %130, i32 2)
>  %132 = extractelement <4 x float> %131, i32 0
>  %133 = extractelement <4 x float> %131, i32 1
>  %134 = extractelement <4 x float> %131, i32 2
>  %135 = extractelement <4 x float> %131, i32 3
>  %136 = fmul float %132, %132
>  %137 = fmul float %133, %133
>  %138 = fmul float %134, %134
>  %139 = fmul float %135, %135
>  %140 = bitcast float %94 to i32
>  %141 = bitcast float %95 to i32
>  %142 = insertelement <2 x i32> undef, i32 %140, i32 0
>  %143 = insertelement <2 x i32> %142, i32 %141, i32 1
>  %144 = bitcast <8 x i32> %31 to <32 x i8>
>  %145 = bitcast <4 x i32> %33 to <16 x i8>
>  %146 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %143, <32 x i8> %144, <16 x i8> %145, i32 2)
>  %147 = extractelement <4 x float> %146, i32 0
>  %148 = extractelement <4 x float> %146, i32 1
>  %149 = extractelement <4 x float> %146, i32 2
>  %150 = extractelement <4 x float> %146, i32 3
>  %151 = fmul float %147, %147
>  %152 = fmul float %148, %148
>  %153 = fmul float %149, %149
>  %154 = fmul float %150, %150
>  %155 = bitcast float %98 to i32
>  %156 = bitcast float %99 to i32
>  %157 = insertelement <2 x i32> undef, i32 %155, i32 0
>  %158 = insertelement <2 x i32> %157, i32 %156, i32 1
>  %159 = bitcast <8 x i32> %35 to <32 x i8>
>  %160 = bitcast <4 x i32> %37 to <16 x i8>
>  %161 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %158, <32 x i8> %159, <16 x i8> %160, i32 2)
>  %162 = extractelement <4 x float> %161, i32 0
>  %163 = extractelement <4 x float> %161, i32 1
>  %164 = extractelement <4 x float> %161, i32 2
>  %165 = extractelement <4 x float> %161, i32 3
>  %166 = fmul float %162, %162
>  %167 = fmul float %163, %163
>  %168 = fmul float %164, %164
>  %169 = fmul float %165, %165
>  %170 = fmul float %139, %139
>  %171 = fmul float %124, %124
>  %172 = fmul float %154, %154
>  %173 = fmul float %169, %169
>  %174 = fmul float %170, %170
>  %175 = fmul float %171, %171
>  %176 = fmul float %172, %172
>  %177 = fmul float %173, %173
>  %178 = fmul float %174, %88
>  %179 = fmul float %175, %89
>  %180 = fmul float %176, %92
>  %181 = fmul float %177, %93
>  %182 = fmul float %121, %179
>  %183 = fmul float %122, %179
>  %184 = fmul float %123, %179
>  %185 = fmul float %124, %179
>  %186 = fmul float %178, %136
>  %187 = fadd float %186, %182
>  %188 = fmul float %178, %137
>  %189 = fadd float %188, %183
>  %190 = fmul float %178, %138
>  %191 = fadd float %190, %184
>  %192 = fmul float %178, %139
>  %193 = fadd float %192, %185
>  %194 = fmul float %180, %151
>  %195 = fadd float %194, %187
>  %196 = fmul float %180, %152
>  %197 = fadd float %196, %189
>  %198 = fmul float %180, %153
>  %199 = fadd float %198, %191
>  %200 = fmul float %180, %154
>  %201 = fadd float %200, %193
>  %202 = fmul float %181, %166
>  %203 = fadd float %202, %195
>  %204 = fmul float %181, %167
>  %205 = fadd float %204, %197
>  %206 = fmul float %181, %168
>  %207 = fadd float %206, %199
>  %208 = fmul float %181, %169
>  %209 = fadd float %208, %201
>  %210 = bitcast float %102 to i32
>  %211 = bitcast float %103 to i32
>  %212 = insertelement <2 x i32> undef, i32 %210, i32 0
>  %213 = insertelement <2 x i32> %212, i32 %211, i32 1
>  %214 = bitcast <8 x i32> %39 to <32 x i8>
>  %215 = bitcast <4 x i32> %41 to <16 x i8>
>  %216 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %213, <32 x i8> %214, <16 x i8> %215, i32 2)
>  %217 = extractelement <4 x float> %216, i32 0
>  %218 = extractelement <4 x float> %216, i32 1
>  %219 = extractelement <4 x float> %216, i32 2
>  %220 = extractelement <4 x float> %216, i32 3
>  %221 = fmul float %217, %217
>  %222 = fmul float %218, %218
>  %223 = fmul float %219, %219
>  %224 = fmul float %220, %220
>  %225 = bitcast float %104 to i32
>  %226 = bitcast float %105 to i32
>  %227 = insertelement <2 x i32> undef, i32 %225, i32 0
>  %228 = insertelement <2 x i32> %227, i32 %226, i32 1
>  %229 = bitcast <8 x i32> %43 to <32 x i8>
>  %230 = bitcast <4 x i32> %45 to <16 x i8>
>  %231 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %228, <32 x i8> %229, <16 x i8> %230, i32 2)
>  %232 = extractelement <4 x float> %231, i32 0
>  %233 = extractelement <4 x float> %231, i32 1
>  %234 = extractelement <4 x float> %231, i32 2
>  %235 = extractelement <4 x float> %231, i32 3
>  %236 = fmul float %232, %232
>  %237 = fmul float %233, %233
>  %238 = fmul float %234, %234
>  %239 = fmul float %235, %235
>  %240 = bitcast float %106 to i32
>  %241 = bitcast float %107 to i32
>  %242 = insertelement <2 x i32> undef, i32 %240, i32 0
>  %243 = insertelement <2 x i32> %242, i32 %241, i32 1
>  %244 = bitcast <8 x i32> %47 to <32 x i8>
>  %245 = bitcast <4 x i32> %49 to <16 x i8>
>  %246 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %243, <32 x i8> %244, <16 x i8> %245, i32 2)
>  %247 = extractelement <4 x float> %246, i32 0
>  %248 = extractelement <4 x float> %246, i32 1
>  %249 = extractelement <4 x float> %246, i32 2
>  %250 = extractelement <4 x float> %246, i32 3
>  %251 = fmul float %247, %247
>  %252 = fmul float %248, %248
>  %253 = fmul float %249, %249
>  %254 = fmul float %250, %250
>  %255 = bitcast float %108 to i32
>  %256 = bitcast float %109 to i32
>  %257 = insertelement <2 x i32> undef, i32 %255, i32 0
>  %258 = insertelement <2 x i32> %257, i32 %256, i32 1
>  %259 = bitcast <8 x i32> %51 to <32 x i8>
>  %260 = bitcast <4 x i32> %53 to <16 x i8>
>  %261 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %258, <32 x i8> %259, <16 x i8> %260, i32 2)
>  %262 = extractelement <4 x float> %261, i32 0
>  %263 = extractelement <4 x float> %261, i32 1
>  %264 = extractelement <4 x float> %261, i32 2
>  %265 = extractelement <4 x float> %261, i32 3
>  %266 = fmul float %262, %262
>  %267 = fmul float %263, %263
>  %268 = fmul float %264, %264
>  %269 = fmul float %265, %265
>  %270 = fmul float %224, %224
>  %271 = fmul float %239, %239
>  %272 = fmul float %254, %254
>  %273 = fmul float %269, %269
>  %274 = fmul float %270, %270
>  %275 = fmul float %271, %271
>  %276 = fmul float %272, %272
>  %277 = fmul float %273, %273
>  %278 = fmul float %274, %96
>  %279 = fmul float %275, %97
>  %280 = fmul float %276, %100
>  %281 = fmul float %277, %101
>  %282 = fmul float %278, %221
>  %283 = fadd float %282, %203
>  %284 = fmul float %278, %222
>  %285 = fadd float %284, %205
>  %286 = fmul float %278, %223
>  %287 = fadd float %286, %207
>  %288 = fmul float %278, %224
>  %289 = fadd float %288, %209
>  %290 = fmul float %279, %236
>  %291 = fadd float %290, %283
>  %292 = fmul float %279, %237
>  %293 = fadd float %292, %285
>  %294 = fmul float %279, %238
>  %295 = fadd float %294, %287
>  %296 = fmul float %279, %239
>  %297 = fadd float %296, %289
>  %298 = fmul float %280, %251
>  %299 = fadd float %298, %291
>  %300 = fmul float %280, %252
>  %301 = fadd float %300, %293
>  %302 = fmul float %280, %253
>  %303 = fadd float %302, %295
>  %304 = fmul float %280, %254
>  %305 = fadd float %304, %297
>  %306 = fmul float %281, %266
>  %307 = fadd float %306, %299
>  %308 = fmul float %281, %267
>  %309 = fadd float %308, %301
>  %310 = fmul float %281, %268
>  %311 = fadd float %310, %303
>  %312 = fmul float %281, %269
>  %313 = fadd float %312, %305
>  %314 = fmul float %278, 1.000000e+00
>  %315 = fmul float %279, 1.000000e+00
>  %316 = fadd float %314, %315
>  %317 = fmul float %280, 1.000000e+00
>  %318 = fadd float %316, %317
>  %319 = fmul float %281, 1.000000e+00
>  %320 = fadd float %318, %319
>  %321 = fmul float %178, 1.000000e+00
>  %322 = fmul float %179, 1.000000e+00
>  %323 = fadd float %321, %322
>  %324 = fmul float %180, 1.000000e+00
>  %325 = fadd float %323, %324
>  %326 = fmul float %181, 1.000000e+00
>  %327 = fadd float %325, %326
>  %328 = fadd float %320, %327
>  %329 = fdiv float 1.000000e+00, %328
>  %330 = fmul float %307, %329
>  %331 = fmul float %309, %329
>  %332 = fmul float %311, %329
>  %333 = fmul float %313, %329
>  %334 = bitcast float %90 to i32
>  %335 = bitcast float %91 to i32
>  %336 = insertelement <2 x i32> undef, i32 %334, i32 0
>  %337 = insertelement <2 x i32> %336, i32 %335, i32 1
>  %338 = bitcast <8 x i32> %59 to <32 x i8>
>  %339 = bitcast <4 x i32> %61 to <16 x i8>
>  %340 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %337, <32 x i8> %338, <16 x i8> %339, i32 2)
>  %341 = extractelement <4 x float> %340, i32 0
>  %342 = fmul float %341, %179
>  %343 = bitcast float %86 to i32
>  %344 = bitcast float %87 to i32
>  %345 = insertelement <2 x i32> undef, i32 %343, i32 0
>  %346 = insertelement <2 x i32> %345, i32 %344, i32 1
>  %347 = bitcast <8 x i32> %55 to <32 x i8>
>  %348 = bitcast <4 x i32> %57 to <16 x i8>
>  %349 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %346, <32 x i8> %347, <16 x i8> %348, i32 2)
>  %350 = extractelement <4 x float> %349, i32 0
>  %351 = fmul float %178, %350
>  %352 = fadd float %351, %342
>  %353 = bitcast float %94 to i32
>  %354 = bitcast float %95 to i32
>  %355 = insertelement <2 x i32> undef, i32 %353, i32 0
>  %356 = insertelement <2 x i32> %355, i32 %354, i32 1
>  %357 = bitcast <8 x i32> %63 to <32 x i8>
>  %358 = bitcast <4 x i32> %65 to <16 x i8>
>  %359 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %356, <32 x i8> %357, <16 x i8> %358, i32 2)
>  %360 = extractelement <4 x float> %359, i32 0
>  %361 = fmul float %180, %360
>  %362 = fadd float %361, %352
>  %363 = bitcast float %98 to i32
>  %364 = bitcast float %99 to i32
>  %365 = insertelement <2 x i32> undef, i32 %363, i32 0
>  %366 = insertelement <2 x i32> %365, i32 %364, i32 1
>  %367 = bitcast <8 x i32> %67 to <32 x i8>
>  %368 = bitcast <4 x i32> %69 to <16 x i8>
>  %369 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %366, <32 x i8> %367, <16 x i8> %368, i32 2)
>  %370 = extractelement <4 x float> %369, i32 0
>  %371 = fmul float %181, %370
>  %372 = fadd float %371, %362
>  %373 = bitcast float %102 to i32
>  %374 = bitcast float %103 to i32
>  %375 = insertelement <2 x i32> undef, i32 %373, i32 0
>  %376 = insertelement <2 x i32> %375, i32 %374, i32 1
>  %377 = bitcast <8 x i32> %71 to <32 x i8>
>  %378 = bitcast <4 x i32> %73 to <16 x i8>
>  %379 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %376, <32 x i8> %377, <16 x i8> %378, i32 2)
>  %380 = extractelement <4 x float> %379, i32 0
>  %381 = fmul float %278, %380
>  %382 = fadd float %381, %372
>  %383 = bitcast float %104 to i32
>  %384 = bitcast float %105 to i32
>  %385 = insertelement <2 x i32> undef, i32 %383, i32 0
>  %386 = insertelement <2 x i32> %385, i32 %384, i32 1
>  %387 = bitcast <8 x i32> %75 to <32 x i8>
>  %388 = bitcast <4 x i32> %77 to <16 x i8>
>  %389 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %386, <32 x i8> %387, <16 x i8> %388, i32 2)
>  %390 = extractelement <4 x float> %389, i32 0
>  %391 = fmul float %279, %390
>  %392 = fadd float %391, %382
>  %393 = bitcast float %106 to i32
>  %394 = bitcast float %107 to i32
>  %395 = insertelement <2 x i32> undef, i32 %393, i32 0
>  %396 = insertelement <2 x i32> %395, i32 %394, i32 1
>  %397 = bitcast <8 x i32> %79 to <32 x i8>
>  %398 = bitcast <4 x i32> %81 to <16 x i8>
>  %399 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %396, <32 x i8> %397, <16 x i8> %398, i32 2)
>  %400 = extractelement <4 x float> %399, i32 0
>  %401 = fmul float %280, %400
>  %402 = fadd float %401, %392
>  %403 = bitcast float %108 to i32
>  %404 = bitcast float %109 to i32
>  %405 = insertelement <2 x i32> undef, i32 %403, i32 0
>  %406 = insertelement <2 x i32> %405, i32 %404, i32 1
>  %407 = bitcast <8 x i32> %83 to <32 x i8>
>  %408 = bitcast <4 x i32> %85 to <16 x i8>
>  %409 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %406, <32 x i8> %407, <16 x i8> %408, i32 2)
>  %410 = extractelement <4 x float> %409, i32 0
>  %411 = fmul float %281, %410
>  %412 = fadd float %411, %402
>  %413 = fmul float %329, %412
>  %414 = fmul float %329, %412
>  %415 = fmul float %329, %412
>  %416 = fmul float %329, %412
>  %417 = call i32 @llvm.SI.packf16(float %330, float %331)
>  %418 = bitcast i32 %417 to float
>  %419 = call i32 @llvm.SI.packf16(float %332, float %333)
>  %420 = bitcast i32 %419 to float
>  %421 = call i32 @llvm.SI.packf16(float %413, float %414)
>  %422 = bitcast i32 %421 to float
>  %423 = call i32 @llvm.SI.packf16(float %415, float %416)
>  %424 = bitcast i32 %423 to float
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 0, i32 1, float %418, float %420, float %418, float %420)
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 1, i32 1, float %422, float %424, float %422, float %424)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                         ; BEFE0A7E
>	s_mov_b64 s[100:101], s[6:7]                                                 ; BEE40406
>	s_mov_b32 m0, s9                                                             ; BEFC0309
>	v_interp_p1_f32 v3, v0, 1, 1, [m0]                                           ; C80C0500
>	v_interp_p2_f32 v3, [v3], v1, 1, 1, [m0]                                     ; C80D0501
>	v_interp_p1_f32 v2, v0, 0, 1, [m0]                                           ; C8080400
>	v_interp_p2_f32 v2, [v2], v1, 0, 1, [m0]                                     ; C8090401
>	s_load_dwordx4 s[24:27], s[4:5], 0x0                                         ; C08C0500
>	s_load_dwordx4 s[8:11], s[4:5], 0x4                                          ; C0840504
>	s_load_dwordx4 s[96:99], s[4:5], 0x8                                         ; C0B00508
>	s_load_dwordx4 s[84:87], s[4:5], 0xc                                         ; C0AA050C
>	s_load_dwordx4 s[60:63], s[4:5], 0x10                                        ; C09E0510
>	s_load_dwordx4 s[72:75], s[4:5], 0x14                                        ; C0A40514
>	s_load_dwordx4 s[48:51], s[4:5], 0x18                                        ; C0980518
>	s_load_dwordx4 s[36:39], s[4:5], 0x1c                                        ; C092051C
>	s_load_dwordx4 s[12:15], s[4:5], 0x20                                        ; C0860520
>	s_load_dwordx4 s[0:3], s[4:5], 0x24                                          ; C0800524
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_writelane_b32 v74, s0, 24                                                  ; 04953000
>	v_writelane_b32 v74, s1, 25                                                  ; 04953201
>	v_writelane_b32 v74, s2, 26                                                  ; 04953402
>	v_writelane_b32 v74, s3, 27                                                  ; 04953603
>	s_load_dwordx4 s[0:3], s[4:5], 0x28                                          ; C0800528
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_writelane_b32 v74, s0, 20                                                  ; 04952800
>	v_writelane_b32 v74, s1, 21                                                  ; 04952A01
>	v_writelane_b32 v74, s2, 22                                                  ; 04952C02
>	v_writelane_b32 v74, s3, 23                                                  ; 04952E03
>	s_load_dwordx4 s[0:3], s[4:5], 0x2c                                          ; C080052C
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_writelane_b32 v74, s0, 16                                                  ; 04952000
>	v_writelane_b32 v74, s1, 17                                                  ; 04952201
>	v_writelane_b32 v74, s2, 18                                                  ; 04952402
>	v_writelane_b32 v74, s3, 19                                                  ; 04952603
>	s_load_dwordx4 s[0:3], s[4:5], 0x30                                          ; C0800530
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_writelane_b32 v74, s0, 12                                                  ; 04951800
>	v_writelane_b32 v74, s1, 13                                                  ; 04951A01
>	v_writelane_b32 v74, s2, 14                                                  ; 04951C02
>	v_writelane_b32 v74, s3, 15                                                  ; 04951E03
>	s_load_dwordx4 s[0:3], s[4:5], 0x34                                          ; C0800534
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_writelane_b32 v74, s0, 8                                                   ; 04951000
>	v_writelane_b32 v74, s1, 9                                                   ; 04951201
>	v_writelane_b32 v74, s2, 10                                                  ; 04951402
>	v_writelane_b32 v74, s3, 11                                                  ; 04951603
>	s_load_dwordx4 s[0:3], s[4:5], 0x38                                          ; C0800538
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_writelane_b32 v74, s0, 4                                                   ; 04950800
>	v_writelane_b32 v74, s1, 5                                                   ; 04950A01
>	v_writelane_b32 v74, s2, 6                                                   ; 04950C02
>	v_writelane_b32 v74, s3, 7                                                   ; 04950E03
>	s_load_dwordx4 s[0:3], s[4:5], 0x3c                                          ; C080053C
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_writelane_b32 v74, s0, 0                                                   ; 04950000
>	v_writelane_b32 v74, s1, 1                                                   ; 04950201
>	v_writelane_b32 v74, s2, 2                                                   ; 04950402
>	v_writelane_b32 v74, s3, 3                                                   ; 04950603
>	s_load_dwordx8 s[28:35], s[100:101], 0x0                                     ; C0CE6500
>	s_load_dwordx8 s[0:7], s[100:101], 0x8                                       ; C0C06508
>	s_load_dwordx8 s[16:23], s[100:101], 0x10                                    ; C0C86510
>	s_load_dwordx8 s[88:95], s[100:101], 0x18                                    ; C0EC6518
>	s_load_dwordx8 s[64:71], s[100:101], 0x20                                    ; C0E06520
>	s_load_dwordx8 s[76:83], s[100:101], 0x28                                    ; C0E66528
>	s_load_dwordx8 s[52:59], s[100:101], 0x30                                    ; C0DA6530
>	s_load_dwordx8 s[40:47], s[100:101], 0x38                                    ; C0D46538
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample v[4:7], 15, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[0:7], s[8:11]        ; F0800F00 00400402
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mul_f32_e32 v8, v7, v7                                                     ; 10100F07
>	v_mul_f32_e32 v9, v8, v8                                                     ; 10121108
>	v_mul_f32_e32 v9, v9, v9                                                     ; 10121309
>	v_interp_p1_f32 v10, v0, 3, 0, [m0]                                          ; C8280300
>	v_interp_p2_f32 v10, [v10], v1, 3, 0, [m0]                                   ; C8290301
>	v_mul_f32_e32 v9, v10, v9                                                    ; 1012130A
>	v_interp_p1_f32 v11, v0, 1, 0, [m0]                                          ; C82C0100
>	v_interp_p2_f32 v11, [v11], v1, 1, 0, [m0]                                   ; C82D0101
>	v_interp_p1_f32 v10, v0, 0, 0, [m0]                                          ; C8280000
>	v_interp_p2_f32 v10, [v10], v1, 0, 0, [m0]                                   ; C8290001
>	image_sample v[12:15], 15, 0, 0, 0, 0, 0, 0, 0, v[10:11], s[28:35], s[24:27] ; F0800F00 00C70C0A
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mul_f32_e32 v16, v15, v15                                                  ; 10201F0F
>	v_mul_f32_e32 v17, v16, v16                                                  ; 10222110
>	v_mul_f32_e32 v17, v17, v17                                                  ; 10222311
>	v_interp_p1_f32 v18, v0, 2, 0, [m0]                                          ; C8480200
>	v_interp_p2_f32 v18, [v18], v1, 2, 0, [m0]                                   ; C8490201
>	v_mad_f32 v19, v17, v18, v9                                                  ; D2820013 04262511
>	v_interp_p1_f32 v21, v0, 1, 2, [m0]                                          ; C8540900
>	v_interp_p2_f32 v21, [v21], v1, 1, 2, [m0]                                   ; C8550901
>	v_interp_p1_f32 v20, v0, 0, 2, [m0]                                          ; C8500800
>	v_interp_p2_f32 v20, [v20], v1, 0, 2, [m0]                                   ; C8510801
>	image_sample v[22:25], 15, 0, 0, 0, 0, 0, 0, 0, v[20:21], s[16:23], s[96:99] ; F0800F00 03041614
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mul_f32_e32 v26, v25, v25                                                  ; 10343319
>	v_mul_f32_e32 v27, v26, v26                                                  ; 1036351A
>	v_mul_f32_e32 v27, v27, v27                                                  ; 1036371B
>	v_interp_p1_f32 v28, v0, 2, 1, [m0]                                          ; C8700600
>	v_interp_p2_f32 v28, [v28], v1, 2, 1, [m0]                                   ; C8710601
>	v_mad_f32 v19, v27, v28, v19                                                 ; D2820013 044E391B
>	v_interp_p1_f32 v30, v0, 1, 3, [m0]                                          ; C8780D00
>	v_interp_p2_f32 v30, [v30], v1, 1, 3, [m0]                                   ; C8790D01
>	v_interp_p1_f32 v29, v0, 0, 3, [m0]                                          ; C8740C00
>	v_interp_p2_f32 v29, [v29], v1, 0, 3, [m0]                                   ; C8750C01
>	image_sample v[31:34], 15, 0, 0, 0, 0, 0, 0, 0, v[29:30], s[88:95], s[84:87] ; F0800F00 02B61F1D
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mul_f32_e32 v35, v34, v34                                                  ; 10464522
>	v_mul_f32_e32 v36, v35, v35                                                  ; 10484723
>	v_mul_f32_e32 v36, v36, v36                                                  ; 10484924
>	v_interp_p1_f32 v37, v0, 3, 1, [m0]                                          ; C8940700
>	v_interp_p2_f32 v37, [v37], v1, 3, 1, [m0]                                   ; C8950701
>	v_mad_f32 v19, v36, v37, v19                                                 ; D2820013 044E4B24
>	v_interp_p1_f32 v39, v0, 1, 5, [m0]                                          ; C89C1500
>	v_interp_p2_f32 v39, [v39], v1, 1, 5, [m0]                                   ; C89D1501
>	v_interp_p1_f32 v38, v0, 0, 5, [m0]                                          ; C8981400
>	v_interp_p2_f32 v38, [v38], v1, 0, 5, [m0]                                   ; C8991401
>	image_sample v[40:43], 15, 0, 0, 0, 0, 0, 0, 0, v[38:39], s[76:83], s[72:75] ; F0800F00 02532826
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mul_f32_e32 v44, v43, v43                                                  ; 1058572B
>	v_mul_f32_e32 v45, v44, v44                                                  ; 105A592C
>	v_mul_f32_e32 v45, v45, v45                                                  ; 105A5B2D
>	v_interp_p1_f32 v46, v0, 3, 2, [m0]                                          ; C8B80B00
>	v_interp_p2_f32 v46, [v46], v1, 3, 2, [m0]                                   ; C8B90B01
>	v_mul_f32_e32 v45, v46, v45                                                  ; 105A5B2E
>	v_interp_p1_f32 v47, v0, 1, 4, [m0]                                          ; C8BC1100
>	v_interp_p2_f32 v47, [v47], v1, 1, 4, [m0]                                   ; C8BD1101
>	v_interp_p1_f32 v46, v0, 0, 4, [m0]                                          ; C8B81000
>	v_interp_p2_f32 v46, [v46], v1, 0, 4, [m0]                                   ; C8B91001
>	image_sample v[48:51], 15, 0, 0, 0, 0, 0, 0, 0, v[46:47], s[64:71], s[60:63] ; F0800F00 01F0302E
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mul_f32_e32 v52, v51, v51                                                  ; 10686733
>	v_mul_f32_e32 v53, v52, v52                                                  ; 106A6934
>	v_mul_f32_e32 v53, v53, v53                                                  ; 106A6B35
>	v_interp_p1_f32 v54, v0, 2, 2, [m0]                                          ; C8D80A00
>	v_interp_p2_f32 v54, [v54], v1, 2, 2, [m0]                                   ; C8D90A01
>	v_mad_f32 v55, v53, v54, v45                                                 ; D2820037 04B66D35
>	v_interp_p1_f32 v57, v0, 1, 6, [m0]                                          ; C8E41900
>	v_interp_p2_f32 v57, [v57], v1, 1, 6, [m0]                                   ; C8E51901
>	v_interp_p1_f32 v56, v0, 0, 6, [m0]                                          ; C8E01800
>	v_interp_p2_f32 v56, [v56], v1, 0, 6, [m0]                                   ; C8E11801
>	image_sample v[58:61], 15, 0, 0, 0, 0, 0, 0, 0, v[56:57], s[52:59], s[48:51] ; F0800F00 018D3A38
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mul_f32_e32 v62, v61, v61                                                  ; 107C7B3D
>	v_mul_f32_e32 v63, v62, v62                                                  ; 107E7D3E
>	v_mul_f32_e32 v63, v63, v63                                                  ; 107E7F3F
>	v_interp_p1_f32 v64, v0, 2, 3, [m0]                                          ; C9000E00
>	v_interp_p2_f32 v64, [v64], v1, 2, 3, [m0]                                   ; C9010E01
>	v_mad_f32 v55, v63, v64, v55                                                 ; D2820037 04DE813F
>	v_interp_p1_f32 v66, v0, 1, 7, [m0]                                          ; C9081D00
>	v_interp_p2_f32 v66, [v66], v1, 1, 7, [m0]                                   ; C9091D01
>	v_interp_p1_f32 v65, v0, 0, 7, [m0]                                          ; C9041C00
>	v_interp_p2_f32 v65, [v65], v1, 0, 7, [m0]                                   ; C9051C01
>	image_sample v[67:70], 15, 0, 0, 0, 0, 0, 0, 0, v[65:66], s[40:47], s[36:39] ; F0800F00 012A4341
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mul_f32_e32 v71, v70, v70                                                  ; 108E8D46
>	v_mul_f32_e32 v72, v71, v71                                                  ; 10908F47
>	v_mul_f32_e32 v72, v72, v72                                                  ; 10909148
>	v_interp_p1_f32 v73, v0, 3, 3, [m0]                                          ; C9240F00
>	v_interp_p2_f32 v73, [v73], v1, 3, 3, [m0]                                   ; C9250F01
>	v_mad_f32 v0, v72, v73, v55                                                  ; D2820000 04DE9348
>	v_add_f32_e32 v0, v19, v0                                                    ; 06000113
>	v_rcp_f32_e32 v0, v0                                                         ; 7E005500
>	v_mul_f32_e32 v1, v9, v8                                                     ; 10021109
>	v_mul_f32_e32 v8, v18, v17                                                   ; 10102312
>	v_mad_f32 v1, v8, v16, v1                                                    ; D2820001 04062108
>	v_mul_f32_e32 v16, v28, v27                                                  ; 1020371C
>	v_mad_f32 v1, v16, v26, v1                                                   ; D2820001 04063510
>	v_mul_f32_e32 v17, v37, v36                                                  ; 10224925
>	v_mad_f32 v1, v17, v35, v1                                                   ; D2820001 04064711
>	v_mul_f32_e32 v18, v54, v53                                                  ; 10246B36
>	v_mad_f32 v1, v18, v52, v1                                                   ; D2820001 04066912
>	v_mad_f32 v1, v45, v44, v1                                                   ; D2820001 0406592D
>	v_mul_f32_e32 v19, v64, v63                                                  ; 10267F40
>	v_mad_f32 v1, v19, v62, v1                                                   ; D2820001 04067D13
>	v_mul_f32_e32 v26, v73, v72                                                  ; 10349149
>	v_mad_f32 v1, v26, v71, v1                                                   ; D2820001 04068F1A
>	v_mul_f32_e32 v1, v0, v1                                                     ; 10020300
>	v_mul_f32_e32 v27, v6, v6                                                    ; 10360D06
>	v_mul_f32_e32 v27, v9, v27                                                   ; 10363709
>	v_mul_f32_e32 v28, v14, v14                                                  ; 10381D0E
>	v_mad_f32 v27, v8, v28, v27                                                  ; D282001B 046E3908
>	v_mul_f32_e32 v28, v24, v24                                                  ; 10383118
>	v_mad_f32 v27, v16, v28, v27                                                 ; D282001B 046E3910
>	v_mul_f32_e32 v28, v33, v33                                                  ; 10384321
>	v_mad_f32 v27, v17, v28, v27                                                 ; D282001B 046E3911
>	v_mul_f32_e32 v28, v50, v50                                                  ; 10386532
>	v_mad_f32 v27, v18, v28, v27                                                 ; D282001B 046E3912
>	v_mul_f32_e32 v28, v42, v42                                                  ; 1038552A
>	v_mad_f32 v27, v45, v28, v27                                                 ; D282001B 046E392D
>	v_mul_f32_e32 v28, v60, v60                                                  ; 1038793C
>	v_mad_f32 v27, v19, v28, v27                                                 ; D282001B 046E3913
>	v_mul_f32_e32 v28, v69, v69                                                  ; 10388B45
>	v_mad_f32 v27, v26, v28, v27                                                 ; D282001B 046E391A
>	v_mul_f32_e32 v27, v0, v27                                                   ; 10363700
>	v_cvt_pkrtz_f16_f32_e32 v1, v27, v1                                          ; 5E02031B
>	v_mul_f32_e32 v27, v5, v5                                                    ; 10360B05
>	v_mul_f32_e32 v27, v9, v27                                                   ; 10363709
>	v_mul_f32_e32 v28, v13, v13                                                  ; 10381B0D
>	v_mad_f32 v27, v8, v28, v27                                                  ; D282001B 046E3908
>	v_mul_f32_e32 v28, v23, v23                                                  ; 10382F17
>	v_mad_f32 v27, v16, v28, v27                                                 ; D282001B 046E3910
>	v_mul_f32_e32 v28, v32, v32                                                  ; 10384120
>	v_mad_f32 v27, v17, v28, v27                                                 ; D282001B 046E3911
>	v_mul_f32_e32 v28, v49, v49                                                  ; 10386331
>	v_mad_f32 v27, v18, v28, v27                                                 ; D282001B 046E3912
>	v_mul_f32_e32 v28, v41, v41                                                  ; 10385329
>	v_mad_f32 v27, v45, v28, v27                                                 ; D282001B 046E392D
>	v_mul_f32_e32 v28, v59, v59                                                  ; 1038773B
>	v_mad_f32 v27, v19, v28, v27                                                 ; D282001B 046E3913
>	v_mul_f32_e32 v28, v68, v68                                                  ; 10388944
>	v_mad_f32 v27, v26, v28, v27                                                 ; D282001B 046E391A
>	v_mul_f32_e32 v27, v0, v27                                                   ; 10363700
>	v_mul_f32_e32 v4, v4, v4                                                     ; 10080904
>	v_mul_f32_e32 v4, v9, v4                                                     ; 10080909
>	v_mul_f32_e32 v5, v12, v12                                                   ; 100A190C
>	v_mad_f32 v4, v8, v5, v4                                                     ; D2820004 04120B08
>	v_mul_f32_e32 v5, v22, v22                                                   ; 100A2D16
>	v_mad_f32 v4, v16, v5, v4                                                    ; D2820004 04120B10
>	v_mul_f32_e32 v5, v31, v31                                                   ; 100A3F1F
>	v_mad_f32 v4, v17, v5, v4                                                    ; D2820004 04120B11
>	v_mul_f32_e32 v5, v48, v48                                                   ; 100A6130
>	v_mad_f32 v4, v18, v5, v4                                                    ; D2820004 04120B12
>	v_mul_f32_e32 v5, v40, v40                                                   ; 100A5128
>	v_mad_f32 v4, v45, v5, v4                                                    ; D2820004 04120B2D
>	v_mul_f32_e32 v5, v58, v58                                                   ; 100A753A
>	v_mad_f32 v4, v19, v5, v4                                                    ; D2820004 04120B13
>	v_mul_f32_e32 v5, v67, v67                                                   ; 100A8743
>	v_mad_f32 v4, v26, v5, v4                                                    ; D2820004 04120B1A
>	v_mul_f32_e32 v4, v0, v4                                                     ; 10080900
>	v_cvt_pkrtz_f16_f32_e32 v4, v4, v27                                          ; 5E083704
>	exp 15, 0, 1, 0, 0, v4, v1, v4, v1                                           ; F800040F 01040104
>	s_load_dwordx8 s[0:7], s[100:101], 0x48                                      ; C0C06548
>	v_readlane_b32 s8, v74, 24                                                   ; 0211314A
>	v_readlane_b32 s9, v74, 25                                                   ; 0213334A
>	v_readlane_b32 s10, v74, 26                                                  ; 0215354A
>	v_readlane_b32 s11, v74, 27                                                  ; 0217374A
>	s_nop 2                                                                      ; BF800002
>	s_waitcnt expcnt(0) lgkmcnt(0)                                               ; BF8C000F
>	image_sample v1, 1, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[0:7], s[8:11]             ; F0800100 00400102
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mul_f32_e32 v1, v9, v1                                                     ; 10020309
>	s_load_dwordx8 s[0:7], s[100:101], 0x40                                      ; C0C06540
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample v2, 1, 0, 0, 0, 0, 0, 0, 0, v[10:11], s[0:7], s[12:15]          ; F0800100 0060020A
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v1, v8, v2, v1                                                     ; D2820001 04060508
>	s_load_dwordx8 s[0:7], s[100:101], 0x50                                      ; C0C06550
>	v_readlane_b32 s8, v74, 20                                                   ; 0211294A
>	v_readlane_b32 s9, v74, 21                                                   ; 02132B4A
>	v_readlane_b32 s10, v74, 22                                                  ; 02152D4A
>	v_readlane_b32 s11, v74, 23                                                  ; 02172F4A
>	s_nop 2                                                                      ; BF800002
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample v2, 1, 0, 0, 0, 0, 0, 0, 0, v[20:21], s[0:7], s[8:11]           ; F0800100 00400214
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v1, v16, v2, v1                                                    ; D2820001 04060510
>	s_load_dwordx8 s[0:7], s[100:101], 0x58                                      ; C0C06558
>	v_readlane_b32 s8, v74, 16                                                   ; 0211214A
>	v_readlane_b32 s9, v74, 17                                                   ; 0213234A
>	v_readlane_b32 s10, v74, 18                                                  ; 0215254A
>	v_readlane_b32 s11, v74, 19                                                  ; 0217274A
>	s_nop 2                                                                      ; BF800002
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample v2, 1, 0, 0, 0, 0, 0, 0, 0, v[29:30], s[0:7], s[8:11]           ; F0800100 0040021D
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v1, v17, v2, v1                                                    ; D2820001 04060511
>	s_load_dwordx8 s[0:7], s[100:101], 0x60                                      ; C0C06560
>	v_readlane_b32 s8, v74, 12                                                   ; 0211194A
>	v_readlane_b32 s9, v74, 13                                                   ; 02131B4A
>	v_readlane_b32 s10, v74, 14                                                  ; 02151D4A
>	v_readlane_b32 s11, v74, 15                                                  ; 02171F4A
>	s_nop 2                                                                      ; BF800002
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample v2, 1, 0, 0, 0, 0, 0, 0, 0, v[46:47], s[0:7], s[8:11]           ; F0800100 0040022E
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v1, v18, v2, v1                                                    ; D2820001 04060512
>	s_load_dwordx8 s[0:7], s[100:101], 0x68                                      ; C0C06568
>	v_readlane_b32 s8, v74, 8                                                    ; 0211114A
>	v_readlane_b32 s9, v74, 9                                                    ; 0213134A
>	v_readlane_b32 s10, v74, 10                                                  ; 0215154A
>	v_readlane_b32 s11, v74, 11                                                  ; 0217174A
>	s_nop 2                                                                      ; BF800002
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample v2, 1, 0, 0, 0, 0, 0, 0, 0, v[38:39], s[0:7], s[8:11]           ; F0800100 00400226
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v1, v45, v2, v1                                                    ; D2820001 0406052D
>	s_load_dwordx8 s[0:7], s[100:101], 0x70                                      ; C0C06570
>	v_readlane_b32 s8, v74, 4                                                    ; 0211094A
>	v_readlane_b32 s9, v74, 5                                                    ; 02130B4A
>	v_readlane_b32 s10, v74, 6                                                   ; 02150D4A
>	v_readlane_b32 s11, v74, 7                                                   ; 02170F4A
>	s_nop 2                                                                      ; BF800002
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample v2, 1, 0, 0, 0, 0, 0, 0, 0, v[56:57], s[0:7], s[8:11]           ; F0800100 00400238
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v1, v19, v2, v1                                                    ; D2820001 04060513
>	s_load_dwordx8 s[0:7], s[100:101], 0x78                                      ; C0C06578
>	v_readlane_b32 s8, v74, 0                                                    ; 0211014A
>	v_readlane_b32 s9, v74, 1                                                    ; 0213034A
>	v_readlane_b32 s10, v74, 2                                                   ; 0215054A
>	v_readlane_b32 s11, v74, 3                                                   ; 0217074A
>	s_nop 2                                                                      ; BF800002
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample v2, 1, 0, 0, 0, 0, 0, 0, 0, v[65:66], s[0:7], s[8:11]           ; F0800100 00400241
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v1, v26, v2, v1                                                    ; D2820001 0406051A
>	v_mul_f32_e32 v0, v1, v0                                                     ; 10000101
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v0                                           ; 5E000100
>	exp 15, 1, 1, 1, 1, v0, v0, v0, v0                                           ; F8001C1F 00000000
>	s_endpgm                                                                     ; BF810000
>
>SGPRS: 104
>VGPRS: 76
>Code Size: 1472 bytes
>LDS: 0 blocks
>Scratch: 7168 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL TEMP[0], LOCAL
>IMM[0] FLT32 {    1.0000,     0.0000,     0.0000,     0.0000}
>  0: MAD TEMP[0], IN[0].xyzx, IMM[0].xxxy, IMM[0].yyyx
>  1: MOV OUT[1], IN[1].xyxy
>  2: MOV OUT[0], TEMP[0]
>  3: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = add i32 %5, %7
>  %14 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %12, i32 0, i32 %13)
>  %15 = extractelement <4 x float> %14, i32 0
>  %16 = extractelement <4 x float> %14, i32 1
>  %17 = extractelement <4 x float> %14, i32 2
>  %18 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %19 = load <16 x i8> addrspace(2)* %18, !tbaa !0
>  %20 = add i32 %5, %7
>  %21 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %19, i32 0, i32 %20)
>  %22 = extractelement <4 x float> %21, i32 0
>  %23 = extractelement <4 x float> %21, i32 1
>  %24 = fmul float %15, 1.000000e+00
>  %25 = fadd float %24, 0.000000e+00
>  %26 = fmul float %16, 1.000000e+00
>  %27 = fadd float %26, 0.000000e+00
>  %28 = fmul float %17, 1.000000e+00
>  %29 = fadd float %28, 0.000000e+00
>  %30 = fmul float %15, 0.000000e+00
>  %31 = fadd float %30, 1.000000e+00
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %22, float %23, float %22, float %23)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %25, float %27, float %29, float %31)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                           ; 4A00000A
>	s_load_dwordx4 s[0:3], s[8:9], 0x0                  ; C0800900
>	s_load_dwordx4 s[4:7], s[8:9], 0x4                  ; C0820904
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[4:7], 0 idxen ; E00C2000 80010100
>	s_waitcnt vmcnt(0)                                  ; BF8C0770
>	exp 15, 32, 0, 0, 0, v1, v2, v1, v2                 ; F800020F 02010201
>	s_waitcnt expcnt(0)                                 ; BF8C070F
>	buffer_load_format_xyzw v[0:3], v0, s[0:3], 0 idxen ; E00C2000 80000000
>	s_waitcnt vmcnt(0)                                  ; BF8C0770
>	v_mad_f32 v4, 0, v0, 1.0                            ; D2820004 03CA0080
>	v_add_f32_e32 v5, 0, v2                             ; 060A0480
>	v_add_f32_e32 v6, 0, v1                             ; 060C0280
>	v_add_f32_e32 v0, 0, v0                             ; 06000080
>	exp 15, 12, 0, 1, 0, v0, v6, v5, v4                 ; F80008CF 04050600
>	s_endpgm                                            ; BF810000
>
>SGPRS: 16
>VGPRS: 8
>Code Size: 84 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL CONST[0..1]
>DCL TEMP[0..7], LOCAL
>IMM[0] FLT32 {    1.0000,    -1.0000,     0.0000,     2.0000}
>IMM[1] FLT32 {    1.0000,    -1.0000,    -2.0000,     0.5000}
>IMM[2] FLT32 {    2.0000,    -2.0000,     0.0000,     4.0000}
>IMM[3] FLT32 {    0.0000,     0.5000,     2.0000,     0.0000}
>  0: MOV TEMP[0].xy, IMM[0].xyxx
>  1: MAD TEMP[1].zw, CONST[0].xyxz, IMM[0].xyyx, IN[0].xyxy
>  2: MAD TEMP[1].xy, TEMP[1].zwzw, IMM[0].xyxx, IMM[0].zxzz
>  3: MOV TEMP[2].xy, TEMP[1].xyyy
>  4: TEX TEMP[2].x, TEMP[2], SAMP[0], 2D
>  5: MOV TEMP[2].z, TEMP[2].xxxx
>  6: ADD TEMP[3].zw, -CONST[0].xyxy, IN[0].xyxy
>  7: MAD TEMP[1].xy, TEMP[3].zwzw, IMM[0].xyxx, IMM[0].zxzz
>  8: MOV TEMP[3].xy, TEMP[1].xyyy
>  9: TEX TEMP[3].x, TEMP[3], SAMP[0], 2D
> 10: MOV TEMP[2].x, TEMP[3].xxxx
> 11: MAD TEMP[3].zw, CONST[0].xyxy, IMM[0].xyxy, IN[0].xyxy
> 12: MOV TEMP[0].w, TEMP[3].wwzw
> 13: MAD TEMP[1].xy, TEMP[3].zwzw, IMM[0].xyxx, IMM[0].zxzz
> 14: MOV TEMP[3].xy, TEMP[1].xyyy
> 15: TEX TEMP[3].xzw, TEMP[3], SAMP[0], 2D
> 16: MOV TEMP[4].zw, TEMP[3].wwzw
> 17: MOV TEMP[2].y, TEMP[3].xxxx
> 18: DP3 TEMP[3].x, TEMP[2].xyzz, IMM[0].xyww
> 19: MOV TEMP[0].z, TEMP[3].xxxx
> 20: ADD TEMP[5].xy, CONST[0].xzzw, IN[0]
> 21: MOV TEMP[4].xy, TEMP[5].xyxx
> 22: MAD TEMP[1].xy, TEMP[4], IMM[0].xyxx, IMM[0].zxzz
> 23: MOV TEMP[5].xy, TEMP[1].xyyy
> 24: TEX TEMP[5].xw, TEMP[5], SAMP[0], 2D
> 25: MOV TEMP[4].x, TEMP[5].xxxw
> 26: MAD TEMP[5].xy, CONST[0], TEMP[0].yxzw, IN[0]
> 27: MOV TEMP[5].xy, TEMP[5].xyxx
> 28: MAD TEMP[1].xy, TEMP[5], IMM[0].xyxx, IMM[0].zxzz
> 29: MOV TEMP[6].xy, TEMP[1].xyyy
> 30: TEX TEMP[6].xzw, TEMP[6], SAMP[0], 2D
> 31: MOV TEMP[5].zw, TEMP[6].wwzw
> 32: MOV TEMP[4].y, TEMP[6].xxxx
> 33: ADD TEMP[6].xy, CONST[0], IN[0]
> 34: MOV TEMP[5].xy, TEMP[6].xyxx
> 35: MAD TEMP[1].xy, TEMP[5], IMM[0].xyxx, IMM[0].zxzz
> 36: MOV TEMP[6].xy, TEMP[1].xyyy
> 37: TEX TEMP[6].xzw, TEMP[6], SAMP[0], 2D
> 38: MOV TEMP[5].zw, TEMP[6].wwzw
> 39: MOV TEMP[4].z, TEMP[6].xxxx
> 40: DP3 TEMP[6].x, TEMP[4].yzxx, IMM[1].xyzz
> 41: ADD TEMP[3].z, TEMP[6].xxxx, TEMP[3].xxxx
> 42: MOV TEMP[0].z, TEMP[3].zzzz
> 43: MUL TEMP[6].w, IMM[1].wwww, CONST[1].xxxx
> 44: MOV TEMP[0].w, TEMP[6].wwww
> 45: MUL TEMP[3].x, TEMP[6].wwww, TEMP[3].zzzz
> 46: MOV TEMP[5].x, TEMP[3].xxxx
> 47: MAD TEMP[3].xy, CONST[0].zyzw, TEMP[0], IN[0]
> 48: MOV TEMP[0].xy, TEMP[3].xyxx
> 49: MAD TEMP[1].xy, TEMP[0], IMM[0].xyxx, IMM[0].zxzz
> 50: MOV TEMP[3].xy, TEMP[1].xyyy
> 51: TEX TEMP[3].x, TEMP[3], SAMP[0], 2D
> 52: MOV TEMP[2].w, TEMP[3].xxxx
> 53: DP3 TEMP[3].x, TEMP[2].xyww, IMM[0].xxww
> 54: ADD TEMP[7].yz, CONST[0].xzyw, IN[0].xxyw
> 55: MOV TEMP[0].yz, TEMP[7].zyzz
> 56: MAD TEMP[1].xy, TEMP[0].yzzw, IMM[0].xyxx, IMM[0].zxzz
> 57: MOV TEMP[1].xy, TEMP[1].xyyy
> 58: TEX TEMP[1].xw, TEMP[1], SAMP[0], 2D
> 59: MOV TEMP[2].w, TEMP[1].wwww
> 60: MOV TEMP[4].w, TEMP[1].xxxx
> 61: DP3 TEMP[1].x, TEMP[4].yzww, IMM[1].yyzz
> 62: ADD TEMP[1].x, TEMP[1].xxxx, TEMP[3].xxxx
> 63: MUL TEMP[1].y, TEMP[6].wwww, TEMP[1].xxxx
> 64: MOV TEMP[5].y, TEMP[1].yyyy
> 65: MUL TEMP[1].xy, TEMP[5], IMM[2].xyxy
> 66: MOV TEMP[0].xy, TEMP[1].xyxx
> 67: MOV TEMP[0].z, IMM[0].zzzz
> 68: ADD TEMP[1].xyz, -TEMP[0], IMM[2].zzwx
> 69: MOV TEMP[0].xy, TEMP[1].xyzx
> 70: DP3 TEMP[1].x, TEMP[1].xyzz, TEMP[1].xyzz
> 71: MOV TEMP[0].z, TEMP[1].xxxx
> 72: MAX TEMP[1].x, TEMP[1].xxxx, IMM[3].xxxx
> 73: RSQ TEMP[1].x, TEMP[1].xxxx
> 74: MOV TEMP[2].z, TEMP[1].xxxx
> 75: MUL TEMP[0].xy, TEMP[0], TEMP[1].xxxx
> 76: MOV TEMP[2].xy, TEMP[0].xyxx
> 77: MAD TEMP[0].xyz, TEMP[2], IMM[3].yyzz, IMM[1].wwww
> 78: MOV TEMP[0].xyz, TEMP[0].xyzx
> 79: MOV TEMP[0].w, IMM[0].zzzz
> 80: MOV OUT[0], TEMP[0]
> 81: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %23 = load <16 x i8> addrspace(2)* %22, !tbaa !0
>  %24 = call float @llvm.SI.load.const(<16 x i8> %23, i32 0)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %23, i32 4)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %23, i32 8)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %23, i32 16)
>  %28 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %29 = load <8 x i32> addrspace(2)* %28, !tbaa !0
>  %30 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %31 = load <4 x i32> addrspace(2)* %30, !tbaa !0
>  %32 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %33 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %34 = fmul float %24, -1.000000e+00
>  %35 = fadd float %34, %32
>  %36 = fmul float %26, 1.000000e+00
>  %37 = fadd float %36, %33
>  %38 = fmul float %35, 1.000000e+00
>  %39 = fadd float %38, 0.000000e+00
>  %40 = fmul float %37, -1.000000e+00
>  %41 = fadd float %40, 1.000000e+00
>  %42 = bitcast float %39 to i32
>  %43 = bitcast float %41 to i32
>  %44 = insertelement <2 x i32> undef, i32 %42, i32 0
>  %45 = insertelement <2 x i32> %44, i32 %43, i32 1
>  %46 = bitcast <8 x i32> %29 to <32 x i8>
>  %47 = bitcast <4 x i32> %31 to <16 x i8>
>  %48 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %45, <32 x i8> %46, <16 x i8> %47, i32 2)
>  %49 = extractelement <4 x float> %48, i32 0
>  %50 = fsub float -0.000000e+00, %24
>  %51 = fadd float %50, %32
>  %52 = fsub float -0.000000e+00, %25
>  %53 = fadd float %52, %33
>  %54 = fmul float %51, 1.000000e+00
>  %55 = fadd float %54, 0.000000e+00
>  %56 = fmul float %53, -1.000000e+00
>  %57 = fadd float %56, 1.000000e+00
>  %58 = bitcast float %55 to i32
>  %59 = bitcast float %57 to i32
>  %60 = insertelement <2 x i32> undef, i32 %58, i32 0
>  %61 = insertelement <2 x i32> %60, i32 %59, i32 1
>  %62 = bitcast <8 x i32> %29 to <32 x i8>
>  %63 = bitcast <4 x i32> %31 to <16 x i8>
>  %64 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %61, <32 x i8> %62, <16 x i8> %63, i32 2)
>  %65 = extractelement <4 x float> %64, i32 0
>  %66 = fmul float %24, 1.000000e+00
>  %67 = fadd float %66, %32
>  %68 = fmul float %25, -1.000000e+00
>  %69 = fadd float %68, %33
>  %70 = fmul float %67, 1.000000e+00
>  %71 = fadd float %70, 0.000000e+00
>  %72 = fmul float %69, -1.000000e+00
>  %73 = fadd float %72, 1.000000e+00
>  %74 = bitcast float %71 to i32
>  %75 = bitcast float %73 to i32
>  %76 = insertelement <2 x i32> undef, i32 %74, i32 0
>  %77 = insertelement <2 x i32> %76, i32 %75, i32 1
>  %78 = bitcast <8 x i32> %29 to <32 x i8>
>  %79 = bitcast <4 x i32> %31 to <16 x i8>
>  %80 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %77, <32 x i8> %78, <16 x i8> %79, i32 2)
>  %81 = extractelement <4 x float> %80, i32 0
>  %82 = fmul float %65, 1.000000e+00
>  %83 = fmul float %81, -1.000000e+00
>  %84 = fadd float %83, %82
>  %85 = fmul float %49, 2.000000e+00
>  %86 = fadd float %84, %85
>  %87 = fadd float %24, %32
>  %88 = fadd float %26, %33
>  %89 = fmul float %87, 1.000000e+00
>  %90 = fadd float %89, 0.000000e+00
>  %91 = fmul float %88, -1.000000e+00
>  %92 = fadd float %91, 1.000000e+00
>  %93 = bitcast float %90 to i32
>  %94 = bitcast float %92 to i32
>  %95 = insertelement <2 x i32> undef, i32 %93, i32 0
>  %96 = insertelement <2 x i32> %95, i32 %94, i32 1
>  %97 = bitcast <8 x i32> %29 to <32 x i8>
>  %98 = bitcast <4 x i32> %31 to <16 x i8>
>  %99 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %96, <32 x i8> %97, <16 x i8> %98, i32 2)
>  %100 = extractelement <4 x float> %99, i32 0
>  %101 = fmul float %24, -1.000000e+00
>  %102 = fadd float %101, %32
>  %103 = fmul float %25, 1.000000e+00
>  %104 = fadd float %103, %33
>  %105 = fmul float %102, 1.000000e+00
>  %106 = fadd float %105, 0.000000e+00
>  %107 = fmul float %104, -1.000000e+00
>  %108 = fadd float %107, 1.000000e+00
>  %109 = bitcast float %106 to i32
>  %110 = bitcast float %108 to i32
>  %111 = insertelement <2 x i32> undef, i32 %109, i32 0
>  %112 = insertelement <2 x i32> %111, i32 %110, i32 1
>  %113 = bitcast <8 x i32> %29 to <32 x i8>
>  %114 = bitcast <4 x i32> %31 to <16 x i8>
>  %115 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %112, <32 x i8> %113, <16 x i8> %114, i32 2)
>  %116 = extractelement <4 x float> %115, i32 0
>  %117 = fadd float %24, %32
>  %118 = fadd float %25, %33
>  %119 = fmul float %117, 1.000000e+00
>  %120 = fadd float %119, 0.000000e+00
>  %121 = fmul float %118, -1.000000e+00
>  %122 = fadd float %121, 1.000000e+00
>  %123 = bitcast float %120 to i32
>  %124 = bitcast float %122 to i32
>  %125 = insertelement <2 x i32> undef, i32 %123, i32 0
>  %126 = insertelement <2 x i32> %125, i32 %124, i32 1
>  %127 = bitcast <8 x i32> %29 to <32 x i8>
>  %128 = bitcast <4 x i32> %31 to <16 x i8>
>  %129 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %126, <32 x i8> %127, <16 x i8> %128, i32 2)
>  %130 = extractelement <4 x float> %129, i32 0
>  %131 = fmul float %116, 1.000000e+00
>  %132 = fmul float %130, -1.000000e+00
>  %133 = fadd float %132, %131
>  %134 = fmul float %100, -2.000000e+00
>  %135 = fadd float %133, %134
>  %136 = fadd float %135, %86
>  %137 = fmul float 5.000000e-01, %27
>  %138 = fmul float %137, %136
>  %139 = fmul float %26, 1.000000e+00
>  %140 = fadd float %139, %32
>  %141 = fmul float %25, -1.000000e+00
>  %142 = fadd float %141, %33
>  %143 = fmul float %140, 1.000000e+00
>  %144 = fadd float %143, 0.000000e+00
>  %145 = fmul float %142, -1.000000e+00
>  %146 = fadd float %145, 1.000000e+00
>  %147 = bitcast float %144 to i32
>  %148 = bitcast float %146 to i32
>  %149 = insertelement <2 x i32> undef, i32 %147, i32 0
>  %150 = insertelement <2 x i32> %149, i32 %148, i32 1
>  %151 = bitcast <8 x i32> %29 to <32 x i8>
>  %152 = bitcast <4 x i32> %31 to <16 x i8>
>  %153 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %150, <32 x i8> %151, <16 x i8> %152, i32 2)
>  %154 = extractelement <4 x float> %153, i32 0
>  %155 = fmul float %65, 1.000000e+00
>  %156 = fmul float %81, 1.000000e+00
>  %157 = fadd float %156, %155
>  %158 = fmul float %154, 2.000000e+00
>  %159 = fadd float %157, %158
>  %160 = fadd float %26, %32
>  %161 = fadd float %25, %33
>  %162 = fmul float %160, 1.000000e+00
>  %163 = fadd float %162, 0.000000e+00
>  %164 = fmul float %161, -1.000000e+00
>  %165 = fadd float %164, 1.000000e+00
>  %166 = bitcast float %163 to i32
>  %167 = bitcast float %165 to i32
>  %168 = insertelement <2 x i32> undef, i32 %166, i32 0
>  %169 = insertelement <2 x i32> %168, i32 %167, i32 1
>  %170 = bitcast <8 x i32> %29 to <32 x i8>
>  %171 = bitcast <4 x i32> %31 to <16 x i8>
>  %172 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %169, <32 x i8> %170, <16 x i8> %171, i32 2)
>  %173 = extractelement <4 x float> %172, i32 0
>  %174 = fmul float %116, -1.000000e+00
>  %175 = fmul float %130, -1.000000e+00
>  %176 = fadd float %175, %174
>  %177 = fmul float %173, -2.000000e+00
>  %178 = fadd float %176, %177
>  %179 = fadd float %178, %159
>  %180 = fmul float %137, %179
>  %181 = fmul float %138, 2.000000e+00
>  %182 = fmul float %180, -2.000000e+00
>  %183 = fsub float -0.000000e+00, %181
>  %184 = fadd float %183, 0.000000e+00
>  %185 = fsub float -0.000000e+00, %182
>  %186 = fadd float %185, 0.000000e+00
>  %187 = fsub float -0.000000e+00, 0.000000e+00
>  %188 = fadd float %187, 4.000000e+00
>  %189 = fmul float %184, %184
>  %190 = fmul float %186, %186
>  %191 = fadd float %190, %189
>  %192 = fmul float %188, %188
>  %193 = fadd float %191, %192
>  %194 = call float @llvm.maxnum.f32(float %193, float 0x3E7AD7F2A0000000)
>  %195 = call float @llvm.AMDGPU.rsq.clamped.f32(float %194)
>  %196 = fmul float %184, %195
>  %197 = fmul float %186, %195
>  %198 = fmul float %196, 5.000000e-01
>  %199 = fadd float %198, 5.000000e-01
>  %200 = fmul float %197, 5.000000e-01
>  %201 = fadd float %200, 5.000000e-01
>  %202 = fmul float %195, 2.000000e+00
>  %203 = fadd float %202, 5.000000e-01
>  %204 = call i32 @llvm.SI.packf16(float %199, float %201)
>  %205 = bitcast i32 %204 to float
>  %206 = call i32 @llvm.SI.packf16(float %203, float 0.000000e+00)
>  %207 = bitcast i32 %206 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %205, float %207, float %205, float %207)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.maxnum.f32(float, float) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.AMDGPU.rsq.clamped.f32(float) #1
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                 ; BEFE0A7E
>	s_mov_b32 m0, s9                                                     ; BEFC0309
>	v_interp_p1_f32 v2, v0, 1, 0, [m0]                                   ; C8080100
>	v_interp_p2_f32 v2, [v2], v1, 1, 0, [m0]                             ; C8090101
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                                   ; C0800300
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	s_buffer_load_dword s8, s[0:3], 0x1                                  ; C2040101
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_subrev_f32_e32 v3, s8, v2                                          ; 0A060408
>	v_sub_f32_e32 v4, 1.0, v3                                            ; 080806F2
>	v_interp_p1_f32 v5, v0, 0, 0, [m0]                                   ; C8140000
>	v_interp_p2_f32 v5, [v5], v1, 0, 0, [m0]                             ; C8150001
>	s_buffer_load_dword s9, s[0:3], 0x0                                  ; C2048100
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_subrev_f32_e32 v0, s9, v5                                          ; 0A000A09
>	v_add_f32_e32 v3, 0, v0                                              ; 06060080
>	s_load_dwordx4 s[12:15], s[4:5], 0x0                                 ; C0860500
>	s_load_dwordx8 s[16:23], s[6:7], 0x0                                 ; C0C80700
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	image_sample v0, 1, 0, 0, 0, 0, 0, 0, 0, v[3:4], s[16:23], s[12:15]  ; F0800100 00640003
>	v_add_f32_e32 v1, s9, v5                                             ; 06020A09
>	v_add_f32_e32 v6, 0, v1                                              ; 060C0280
>	v_mov_b32_e32 v7, v4                                                 ; 7E0E0304
>	image_sample v1, 1, 0, 0, 0, 0, 0, 0, 0, v[6:7], s[16:23], s[12:15]  ; F0800100 00640106
>	s_waitcnt vmcnt(0)                                                   ; BF8C0770
>	v_add_f32_e32 v8, v0, v1                                             ; 06100300
>	s_buffer_load_dword s4, s[0:3], 0x2                                  ; C2020102
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_add_f32_e32 v5, s4, v5                                             ; 060A0A04
>	v_add_f32_e32 v9, 0, v5                                              ; 06120A80
>	v_mov_b32_e32 v10, v4                                                ; 7E140304
>	image_sample v5, 1, 0, 0, 0, 0, 0, 0, 0, v[9:10], s[16:23], s[12:15] ; F0800100 00640509
>	s_waitcnt vmcnt(0)                                                   ; BF8C0770
>	v_mad_f32 v5, 2.0, v5, v8                                            ; D2820005 04220AF4
>	v_add_f32_e32 v8, s8, v2                                             ; 06100408
>	v_sub_f32_e32 v10, 1.0, v8                                           ; 081410F2
>	v_mov_b32_e32 v4, v10                                                ; 7E08030A
>	image_sample v8, 1, 0, 0, 0, 0, 0, 0, 0, v[3:4], s[16:23], s[12:15]  ; F0800100 00640803
>	v_mov_b32_e32 v7, v10                                                ; 7E0E030A
>	image_sample v11, 1, 0, 0, 0, 0, 0, 0, 0, v[6:7], s[16:23], s[12:15] ; F0800100 00640B06
>	s_waitcnt vmcnt(0)                                                   ; BF8C0770
>	v_sub_f32_e64 v12, -v11, v8                                          ; D208000C 2002110B
>	image_sample v9, 1, 0, 0, 0, 0, 0, 0, 0, v[9:10], s[16:23], s[12:15] ; F0800100 00640909
>	s_waitcnt vmcnt(0)                                                   ; BF8C0770
>	v_mad_f32 v9, -2.0, v9, v12                                          ; D2820009 043212F5
>	v_add_f32_e32 v5, v5, v9                                             ; 060A1305
>	s_buffer_load_dword s0, s[0:3], 0x4                                  ; C2000104
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mul_f32_e64 v9, 0.5, s0                                            ; D2100009 000000F0
>	v_mul_f32_e32 v5, v5, v9                                             ; 100A1305
>	v_mad_f32 v5, -2.0, -v5, 0                                           ; D2820005 42020AF5
>	v_subrev_f32_e32 v0, v1, v0                                          ; 0A000101
>	v_add_f32_e32 v1, s4, v2                                             ; 06020404
>	v_sub_f32_e32 v4, 1.0, v1                                            ; 080802F2
>	image_sample v1, 1, 0, 0, 0, 0, 0, 0, 0, v[3:4], s[16:23], s[12:15]  ; F0800100 00640103
>	s_waitcnt vmcnt(0)                                                   ; BF8C0770
>	v_mad_f32 v0, 2.0, v1, v0                                            ; D2820000 040202F4
>	v_subrev_f32_e32 v1, v11, v8                                         ; 0A02110B
>	v_mov_b32_e32 v7, v4                                                 ; 7E0E0304
>	image_sample v2, 1, 0, 0, 0, 0, 0, 0, 0, v[6:7], s[16:23], s[12:15]  ; F0800100 00640206
>	s_waitcnt vmcnt(0)                                                   ; BF8C0770
>	v_mad_f32 v1, -2.0, v2, v1                                           ; D2820001 040604F5
>	v_add_f32_e32 v0, v0, v1                                             ; 06000300
>	v_mul_f32_e32 v0, v0, v9                                             ; 10001300
>	v_mad_f32 v0, -2.0, v0, 0                                            ; D2820000 020200F5
>	v_mul_f32_e32 v1, v0, v0                                             ; 10020100
>	v_mad_f32 v1, v5, v5, v1                                             ; D2820001 04060B05
>	v_add_f32_e32 v1, 0x41800000, v1                                     ; 060202FF 41800000
>	v_max_f32_e32 v1, 0x33d6bf95, v1                                     ; 200202FF 33D6BF95
>	v_rsq_clamp_f32_e32 v1, v1                                           ; 7E025901
>	v_mul_f32_e32 v2, v1, v5                                             ; 10040B01
>	v_mad_f32 v2, 0.5, v2, 0.5                                           ; D2820002 03C204F0
>	v_mul_f32_e32 v0, v1, v0                                             ; 10000101
>	v_mad_f32 v0, 0.5, v0, 0.5                                           ; D2820000 03C200F0
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v2                                   ; 5E000500
>	v_mad_f32 v1, 2.0, v1, 0.5                                           ; D2820001 03C202F4
>	v_cvt_pkrtz_f16_f32_e64 v1, v1, 0                                    ; D25E0001 00010101
>	exp 15, 0, 1, 1, 1, v0, v1, v0, v1                                   ; F8001C0F 01000100
>	s_endpgm                                                             ; BF810000
>
>SGPRS: 24
>VGPRS: 16
>Code Size: 412 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL CONST[0]
>DCL TEMP[0..1], LOCAL
>IMM[0] FLT32 {    1.0000,     0.0000,     0.0000,     0.0000}
>  0: ADD TEMP[0].xy, CONST[0], IN[0]
>  1: MOV TEMP[0].xy, TEMP[0].xyxx
>  2: MAD TEMP[1].zw, IN[0].zzzz, IMM[0].xyxy, IMM[0].xyyx
>  3: MOV TEMP[0].zw, TEMP[1].wwzw
>  4: MOV TEMP[1].xy, IN[1].xyxx
>  5: MOV TEMP[1].zw, IMM[0].xxyx
>  6: MOV OUT[0], TEMP[0]
>  7: MOV OUT[1], TEMP[1]
>  8: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 0)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 4)
>  %15 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %16 = load <16 x i8> addrspace(2)* %15, !tbaa !0
>  %17 = add i32 %5, %7
>  %18 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %16, i32 0, i32 %17)
>  %19 = extractelement <4 x float> %18, i32 0
>  %20 = extractelement <4 x float> %18, i32 1
>  %21 = extractelement <4 x float> %18, i32 2
>  %22 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %23 = load <16 x i8> addrspace(2)* %22, !tbaa !0
>  %24 = add i32 %5, %7
>  %25 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %23, i32 0, i32 %24)
>  %26 = extractelement <4 x float> %25, i32 0
>  %27 = extractelement <4 x float> %25, i32 1
>  %28 = fadd float %13, %19
>  %29 = fadd float %14, %20
>  %30 = fmul float %21, 1.000000e+00
>  %31 = fadd float %30, 0.000000e+00
>  %32 = fmul float %21, 0.000000e+00
>  %33 = fadd float %32, 1.000000e+00
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %26, float %27, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %28, float %29, float %31, float %33)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                            ; 4A00000A
>	s_load_dwordx4 s[4:7], s[8:9], 0x0                   ; C0820900
>	s_load_dwordx4 s[8:11], s[8:9], 0x4                  ; C0840904
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[8:11], 0 idxen ; E00C2000 80020100
>	v_mov_b32_e32 v5, 1.0                                ; 7E0A02F2
>	v_mov_b32_e32 v6, 0                                  ; 7E0C0280
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	exp 15, 32, 0, 0, 0, v1, v2, v6, v5                  ; F800020F 05060201
>	s_waitcnt expcnt(0)                                  ; BF8C070F
>	buffer_load_format_xyzw v[0:3], v0, s[4:7], 0 idxen  ; E00C2000 80010000
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_mad_f32 v4, 0, v2, 1.0                             ; D2820004 03CA0480
>	v_add_f32_e32 v5, 0, v2                              ; 060A0480
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                   ; C0800300
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	s_buffer_load_dword s4, s[0:3], 0x1                  ; C2020101
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_add_f32_e32 v6, s4, v1                             ; 060C0204
>	s_buffer_load_dword s0, s[0:3], 0x0                  ; C2000100
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_add_f32_e32 v0, s0, v0                             ; 06000000
>	exp 15, 12, 0, 1, 0, v0, v6, v5, v4                  ; F80008CF 04050600
>	s_endpgm                                             ; BF810000
>
>SGPRS: 16
>VGPRS: 8
>Code Size: 116 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL CONST[0..4]
>DCL TEMP[0..21], LOCAL
>IMM[0] FLT32 {    0.0000,   255.0000,     0.0000,     0.3145}
>IMM[1] FLT32 {    1.0000,   -64.0000,     0.0104,     0.0000}
>IMM[2] INT32 {40, 0, -1, 0}
>  0: ABS TEMP[0], CONST[1]
>  1: ADD TEMP[1], -TEMP[0].yyyy, TEMP[0].xxxx
>  2: MOV TEMP[2].z, TEMP[1].zzzz
>  3: RCP TEMP[3].x, TEMP[0].yyyy
>  4: MUL TEMP[3], TEMP[3].xxxx, CONST[1].xxzy
>  5: MOV TEMP[2].yw, TEMP[3].wyww
>  6: RCP TEMP[2].x, TEMP[0].xxxx
>  7: MUL TEMP[0], TEMP[2].xxxx, CONST[1]
>  8: MOV TEMP[4], TEMP[2].ywzw
>  9: FSGE TEMP[3].x, TEMP[1].zzzz, IMM[0].xxxx
> 10: UIF TEMP[3].xxxx :0
> 11:   MOV TEMP[3].x, TEMP[0].xxxx
> 12: ELSE :0
> 13:   MOV TEMP[3].x, TEMP[2].yyyy
> 14: ENDIF
> 15: MOV TEMP[3].x, TEMP[3].xxxx
> 16: FSGE TEMP[5].x, TEMP[1].zzzz, IMM[0].xxxx
> 17: UIF TEMP[5].xxxx :0
> 18:   MOV TEMP[5].x, TEMP[0].yyyy
> 19: ELSE :0
> 20:   MOV TEMP[5].x, TEMP[2].wwww
> 21: ENDIF
> 22: MOV TEMP[3].y, TEMP[5].xxxx
> 23: FSGE TEMP[5].x, TEMP[1].zzzz, IMM[0].xxxx
> 24: UIF TEMP[5].xxxx :0
> 25:   MOV TEMP[5].x, TEMP[0].zzzz
> 26: ELSE :0
> 27:   MOV TEMP[5].x, TEMP[2].zzzz
> 28: ENDIF
> 29: MOV TEMP[3].z, TEMP[5].xxxx
> 30: FSGE TEMP[1].x, TEMP[1].zzzz, IMM[0].xxxx
> 31: UIF TEMP[1].xxxx :0
> 32:   MOV TEMP[1].x, TEMP[0].wwww
> 33: ELSE :0
> 34:   MOV TEMP[1].x, TEMP[2].wwww
> 35: ENDIF
> 36: MOV TEMP[3].w, TEMP[1].xxxx
> 37: MOV TEMP[2].xy, TEMP[3].xyxx
> 38: ADD TEMP[1], CONST[3].xyxy, IN[0].xyxy
> 39: MOV TEMP[2].zw, TEMP[1].wwzw
> 40: MUL TEMP[1], TEMP[2], CONST[2].xyxy
> 41: MOV TEMP[2].w, TEMP[1].wwzw
> 42: MOV TEMP[1].xy, TEMP[1].zwww
> 43: TEX TEMP[1], TEMP[1], SAMP[0], 2D
> 44: MUL TEMP[1], TEMP[1].wwww, CONST[0].zzzz
> 45: MAD TEMP[1], TEMP[1].zzzz, IMM[0].yyyy, CONST[4].zzzz
> 46: MOV TEMP[2].z, TEMP[1].zzzz
> 47: ADD TEMP[5], TEMP[2], IN[0]
> 48: MOV TEMP[0].xy, TEMP[5].xyxx
> 49: MUL TEMP[3], TEMP[3].xyxy, CONST[0].xyxy
> 50: MOV TEMP[0].zw, TEMP[3].wwzw
> 51: DP2 TEMP[3].x, TEMP[3].zwww, TEMP[3].zwww
> 52: MAX TEMP[3].x, TEMP[3].xxxx, IMM[0].zzzz
> 53: SQRT TEMP[3].x, TEMP[3].xxxx
> 54: MOV TEMP[2].w, TEMP[3].xxxx
> 55: MAD TEMP[1], TEMP[3].xxxx, IMM[0].wwww, TEMP[1].zzzz
> 56: MOV TEMP[2].z, TEMP[1].zzzz
> 57: ADD TEMP[3], TEMP[0], CONST[3]
> 58: MOV TEMP[0].xy, TEMP[3].xyxx
> 59: MUL TEMP[3], TEMP[0], CONST[2]
> 60: MOV TEMP[0].xy, TEMP[3].xyxx
> 61: MOV TEMP[0].zw, TEMP[3].yyxy
> 62: MOV TEMP[3].x, IMM[1].xxxx
> 63: MOV TEMP[3].y, TEMP[1].zzzz
> 64: MOV TEMP[1].x, IMM[2].xxxx
> 65: BGNLOOP :0
> 66:   ISGE TEMP[5].x, IMM[2].yyyy, TEMP[1].xxxx
> 67:   UIF TEMP[5].xxxx :0
> 68:     BRK
> 69:   ENDIF
> 70:   MAD TEMP[6], TEMP[2].xyxy, CONST[2].xyxy, TEMP[0]
> 71:   MOV TEMP[0].zw, TEMP[6].wwzw
> 72:   MAD TEMP[7], TEMP[2].wwww, IMM[0].wwww, TEMP[3].yyyy
> 73:   MOV TEMP[3].y, TEMP[7].yyyy
> 74:   MOV TEMP[8].xy, TEMP[6].zwww
> 75:   TEX TEMP[9], TEMP[8], SAMP[0], 2D
> 76:   MOV TEMP[10], TEMP[9]
> 77:   MOV TEMP[11].yzw, TEMP[9].zyzw
> 78:   MUL TEMP[12], TEMP[9].wwww, IMM[0].yyyy
> 79:   MAD TEMP[13], TEMP[9].wwww, IMM[0].yyyy, IMM[1].yyyy
> 80:   MUL TEMP[4], TEMP[13].wwww, IMM[1].zzzz
> 81:   MOV_SAT TEMP[14], TEMP[4]
> 82:   ADD TEMP[15], -TEMP[14].wwww, IMM[1].xxxx
> 83:   MOV TEMP[3].w, TEMP[15].wwww
> 84:   MAD TEMP[16], TEMP[12].zzzz, -CONST[0].zzzz, TEMP[7].yyyy
> 85:   MOV TEMP[3].z, TEMP[16].zzzz
> 86:   MIN TEMP[17], TEMP[15].wwww, TEMP[3].xxxx
> 87:   MOV TEMP[11].x, TEMP[17].xxxx
> 88:   MOV TEMP[18].x, TEMP[3].xxxx
> 89:   MOV TEMP[19].x, TEMP[17].xxxx
> 90:   FSGE TEMP[20].x, TEMP[16].zzzz, IMM[0].xxxx
> 91:   UIF TEMP[20].xxxx :0
> 92:     MOV TEMP[21].x, TEMP[3].xxxx
> 93:   ELSE :0
> 94:     MOV TEMP[21].x, TEMP[17].xxxx
> 95:   ENDIF
> 96:   MOV TEMP[3].x, TEMP[21].xxxx
> 97:   UADD TEMP[1].x, TEMP[1].xxxx, IMM[2].zzzz
> 98: ENDLOOP :0
> 99: ADD TEMP[0].x, -TEMP[3].xxxx, IMM[1].xxxx
>100: MUL TEMP[0], TEMP[0].xxxx, CONST[4].zzzz
>101: MOV OUT[0], TEMP[0]
>102: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %23 = load <16 x i8> addrspace(2)* %22, !tbaa !0
>  %24 = call float @llvm.SI.load.const(<16 x i8> %23, i32 0)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %23, i32 4)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %23, i32 8)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %23, i32 16)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %23, i32 20)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %23, i32 24)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %23, i32 28)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %23, i32 32)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %23, i32 36)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %23, i32 48)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %23, i32 52)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %23, i32 72)
>  %36 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %37 = load <8 x i32> addrspace(2)* %36, !tbaa !0
>  %38 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %39 = load <4 x i32> addrspace(2)* %38, !tbaa !0
>  %40 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %41 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %42 = call float @fabs(float %27)
>  %43 = call float @fabs(float %28)
>  %44 = call float @fabs(float %29)
>  %45 = call float @fabs(float %30)
>  %46 = fsub float -0.000000e+00, %43
>  %47 = fadd float %46, %42
>  %48 = fdiv float 1.000000e+00, %43
>  %49 = fmul float %48, %27
>  %50 = fmul float %48, %28
>  %51 = fdiv float 1.000000e+00, %42
>  %52 = fmul float %51, %27
>  %53 = fmul float %51, %28
>  %54 = fcmp oge float %47, 0.000000e+00
>  %55 = sext i1 %54 to i32
>  %56 = bitcast i32 %55 to float
>  %57 = bitcast float %56 to i32
>  %58 = icmp ne i32 %57, 0
>  %. = select i1 %58, float %52, float %49
>  %59 = fcmp oge float %47, 0.000000e+00
>  %60 = sext i1 %59 to i32
>  %61 = bitcast i32 %60 to float
>  %62 = bitcast float %61 to i32
>  %63 = icmp ne i32 %62, 0
>  %temp20.0 = select i1 %63, float %53, float %50
>  %64 = fadd float %33, %40
>  %65 = fadd float %34, %41
>  %66 = fmul float %64, %31
>  %67 = fmul float %65, %32
>  %68 = bitcast float %66 to i32
>  %69 = bitcast float %67 to i32
>  %70 = insertelement <2 x i32> undef, i32 %68, i32 0
>  %71 = insertelement <2 x i32> %70, i32 %69, i32 1
>  %72 = bitcast <8 x i32> %37 to <32 x i8>
>  %73 = bitcast <4 x i32> %39 to <16 x i8>
>  %74 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %71, <32 x i8> %72, <16 x i8> %73, i32 2)
>  %75 = extractelement <4 x float> %74, i32 3
>  %76 = fmul float %75, %26
>  %77 = fmul float %76, 2.550000e+02
>  %78 = fadd float %77, %35
>  %79 = fadd float %., %40
>  %80 = fadd float %temp20.0, %41
>  %81 = fmul float %., %24
>  %82 = fmul float %temp20.0, %25
>  %83 = fmul float %81, %81
>  %84 = fmul float %82, %82
>  %85 = fadd float %83, %84
>  %86 = call float @llvm.maxnum.f32(float %85, float 0x3E7AD7F2A0000000)
>  %87 = call float @llvm.sqrt.f32(float %86)
>  %88 = fmul float %87, 0x3FD42085C0000000
>  %89 = fadd float %88, %78
>  %90 = fadd float %79, %33
>  %91 = fadd float %80, %34
>  %92 = fmul float %90, %31
>  %93 = fmul float %91, %32
>  %94 = fmul float %., %31
>  %95 = fmul float %temp20.0, %32
>  %96 = fmul float %87, 0x3FD42085C0000000
>  %97 = bitcast <8 x i32> %37 to <32 x i8>
>  %98 = bitcast <4 x i32> %39 to <16 x i8>
>  %99 = fsub float -0.000000e+00, %26
>  br label %LOOP
>
>LOOP:                                             ; preds = %ENDIF97, %main_body
>  %temp12.1 = phi float [ 1.000000e+00, %main_body ], [ %temp12.1., %ENDIF97 ]
>  %temp13.0 = phi float [ %89, %main_body ], [ %118, %ENDIF97 ]
>  %temp4.1 = phi float [ 0x36F4000000000000, %main_body ], [ %148, %ENDIF97 ]
>  %temp3.0 = phi float [ %93, %main_body ], [ %117, %ENDIF97 ]
>  %temp2.0 = phi float [ %92, %main_body ], [ %116, %ENDIF97 ]
>  %100 = bitcast float %temp4.1 to i32
>  %101 = icmp sge i32 0, %100
>  %102 = sext i1 %101 to i32
>  %103 = bitcast i32 %102 to float
>  %104 = bitcast float %103 to i32
>  %105 = icmp ne i32 %104, 0
>  br i1 %105, label %IF98, label %ENDIF97
>
>IF98:                                             ; preds = %LOOP
>  %temp12.1.lcssa = phi float [ %temp12.1, %LOOP ]
>  %106 = fsub float -0.000000e+00, %temp12.1.lcssa
>  %107 = fadd float %106, 1.000000e+00
>  %108 = fmul float %107, %35
>  %109 = fmul float %107, %35
>  %110 = fmul float %107, %35
>  %111 = fmul float %107, %35
>  %112 = call i32 @llvm.SI.packf16(float %108, float %109)
>  %113 = bitcast i32 %112 to float
>  %114 = call i32 @llvm.SI.packf16(float %110, float %111)
>  %115 = bitcast i32 %114 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %113, float %115, float %113, float %115)
>  ret void
>
>ENDIF97:                                          ; preds = %LOOP
>  %116 = fadd float %94, %temp2.0
>  %117 = fadd float %95, %temp3.0
>  %118 = fadd float %96, %temp13.0
>  %119 = bitcast float %116 to i32
>  %120 = bitcast float %117 to i32
>  %121 = insertelement <2 x i32> undef, i32 %119, i32 0
>  %122 = insertelement <2 x i32> %121, i32 %120, i32 1
>  %123 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %122, <32 x i8> %97, <16 x i8> %98, i32 2)
>  %124 = extractelement <4 x float> %123, i32 3
>  %125 = fmul float %124, 2.550000e+02
>  %126 = fmul float %124, 2.550000e+02
>  %127 = fadd float %126, -6.400000e+01
>  %128 = fmul float %127, 0x3F855559C0000000
>  %129 = fmul float %127, 0x3F855559C0000000
>  %130 = fmul float %127, 0x3F855559C0000000
>  %131 = fmul float %127, 0x3F855559C0000000
>  %132 = call float @llvm.AMDIL.clamp.(float %128, float 0.000000e+00, float 1.000000e+00)
>  %133 = call float @llvm.AMDIL.clamp.(float %129, float 0.000000e+00, float 1.000000e+00)
>  %134 = call float @llvm.AMDIL.clamp.(float %130, float 0.000000e+00, float 1.000000e+00)
>  %135 = call float @llvm.AMDIL.clamp.(float %131, float 0.000000e+00, float 1.000000e+00)
>  %136 = fsub float -0.000000e+00, %135
>  %137 = fadd float %136, 1.000000e+00
>  %138 = fmul float %125, %99
>  %139 = fadd float %138, %118
>  %140 = call float @llvm.minnum.f32(float %137, float %temp12.1)
>  %141 = fcmp oge float %139, 0.000000e+00
>  %142 = sext i1 %141 to i32
>  %143 = bitcast i32 %142 to float
>  %144 = bitcast float %143 to i32
>  %145 = icmp ne i32 %144, 0
>  %temp12.1. = select i1 %145, float %temp12.1, float %140
>  %146 = bitcast float %temp4.1 to i32
>  %147 = add i32 %146, -1
>  %148 = bitcast i32 %147 to float
>  br label %LOOP
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: readnone
>declare float @fabs(float) #2
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.maxnum.f32(float, float) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.sqrt.f32(float) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.clamp.(float, float, float) #2
>
>; Function Attrs: nounwind readnone
>declare float @llvm.minnum.f32(float, float) #1
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                             ; BEFE0A7E
>	s_load_dwordx4 s[12:15], s[2:3], 0x0                             ; C0860300
>	s_waitcnt lgkmcnt(0)                                             ; BF8C007F
>	s_buffer_load_dword s0, s[12:15], 0x4                            ; C2000D04
>	s_waitcnt lgkmcnt(0)                                             ; BF8C007F
>	v_rcp_f32_e64 v2, |s0|                                           ; D3540102 00000000
>	v_mul_f32_e32 v3, s0, v2                                         ; 10060400
>	s_buffer_load_dword s1, s[12:15], 0x5                            ; C2008D05
>	s_waitcnt lgkmcnt(0)                                             ; BF8C007F
>	v_rcp_f32_e64 v4, |s1|                                           ; D3540104 00000001
>	v_mul_f32_e32 v5, s0, v4                                         ; 100A0800
>	v_mov_b32_e32 v6, s1                                             ; 7E0C0201
>	v_sub_f32_e64 v6, |s0|, |v6|                                     ; D2080306 00020C00
>	v_cmp_ge_f32_e64 s[2:3], v6, 0                                   ; D00C0002 00010106
>	v_cndmask_b32_e64 v6, 0, -1, s[2:3]                              ; D2000006 00098280
>	v_cmp_ne_i32_e64 s[2:3], v6, 0                                   ; D10A0002 00010106
>	v_cndmask_b32_e64 v6, v5, v3, s[2:3]                             ; D2000006 080A0705
>	s_buffer_load_dword s0, s[12:15], 0x0                            ; C2000D00
>	s_waitcnt lgkmcnt(0)                                             ; BF8C007F
>	v_mul_f32_e32 v3, s0, v6                                         ; 10060C00
>	v_mul_f32_e32 v2, s1, v2                                         ; 10040401
>	v_mul_f32_e32 v4, s1, v4                                         ; 10080801
>	v_cndmask_b32_e64 v4, v4, v2, s[2:3]                             ; D2000004 000A0504
>	s_buffer_load_dword s0, s[12:15], 0x1                            ; C2000D01
>	s_waitcnt lgkmcnt(0)                                             ; BF8C007F
>	v_mul_f32_e32 v2, s0, v4                                         ; 10040800
>	v_mul_f32_e32 v2, v2, v2                                         ; 10040502
>	v_mad_f32 v2, v3, v3, v2                                         ; D2820002 040A0703
>	v_max_f32_e32 v2, 0x33d6bf95, v2                                 ; 200404FF 33D6BF95
>	v_sqrt_f32_e32 v2, v2                                            ; 7E046702
>	s_mov_b32 m0, s9                                                 ; BEFC0309
>	v_interp_p1_f32 v5, v0, 1, 0, [m0]                               ; C8140100
>	v_interp_p2_f32 v5, [v5], v1, 1, 0, [m0]                         ; C8150101
>	s_buffer_load_dword s16, s[12:15], 0xd                           ; C2080D0D
>	s_waitcnt lgkmcnt(0)                                             ; BF8C007F
>	v_add_f32_e32 v3, s16, v5                                        ; 06060A10
>	s_buffer_load_dword s17, s[12:15], 0x9                           ; C2088D09
>	s_waitcnt lgkmcnt(0)                                             ; BF8C007F
>	v_mul_f32_e32 v8, s17, v3                                        ; 10100611
>	v_interp_p1_f32 v9, v0, 0, 0, [m0]                               ; C8240000
>	v_interp_p2_f32 v9, [v9], v1, 0, 0, [m0]                         ; C8250001
>	s_buffer_load_dword s18, s[12:15], 0xc                           ; C2090D0C
>	s_waitcnt lgkmcnt(0)                                             ; BF8C007F
>	v_add_f32_e32 v0, s18, v9                                        ; 06001212
>	s_buffer_load_dword s19, s[12:15], 0x8                           ; C2098D08
>	s_waitcnt lgkmcnt(0)                                             ; BF8C007F
>	v_mul_f32_e32 v7, s19, v0                                        ; 100E0013
>	s_load_dwordx4 s[0:3], s[4:5], 0x0                               ; C0800500
>	s_load_dwordx8 s[4:11], s[6:7], 0x0                              ; C0C20700
>	s_waitcnt lgkmcnt(0)                                             ; BF8C007F
>	image_sample v0, 8, 0, 0, 0, 0, 0, 0, 0, v[7:8], s[4:11], s[0:3] ; F0800800 00010007
>	s_buffer_load_dword s20, s[12:15], 0x2                           ; C20A0D02
>	s_waitcnt vmcnt(0) lgkmcnt(0)                                    ; BF8C0070
>	v_mul_f32_e32 v0, s20, v0                                        ; 10000014
>	s_buffer_load_dword s12, s[12:15], 0x12                          ; C2060D12
>	v_mov_b32_e32 v1, 0x437f0000                                     ; 7E0202FF 437F0000
>	s_waitcnt lgkmcnt(0)                                             ; BF8C007F
>	v_mad_f32 v0, v1, v0, s12                                        ; D2820000 00320101
>	v_mov_b32_e32 v1, 0x3ea1042e                                     ; 7E0202FF 3EA1042E
>	v_mad_f32 v0, v2, v1, v0                                         ; D2820000 04020302
>	v_mul_f32_e32 v1, 0x3ea1042e, v2                                 ; 100204FF 3EA1042E
>	v_mul_f32_e32 v2, s17, v4                                        ; 10040811
>	v_mul_f32_e32 v3, s19, v6                                        ; 10060C13
>	v_add_f32_e32 v4, v5, v4                                         ; 06080905
>	v_add_f32_e32 v4, s16, v4                                        ; 06080810
>	v_mul_f32_e32 v5, s17, v4                                        ; 100A0811
>	v_add_f32_e32 v6, v9, v6                                         ; 060C0D09
>	v_add_f32_e32 v6, s18, v6                                        ; 060C0C12
>	v_mul_f32_e32 v4, s19, v6                                        ; 10080C13
>	v_mov_b32_e32 v6, 0x80000000                                     ; 7E0C02FF 80000000
>	v_xor_b32_e32 v6, s20, v6                                        ; 3A0C0C14
>	v_mov_b32_e32 v7, 40                                             ; 7E0E02A8
>	v_mov_b32_e32 v9, 1.0                                            ; 7E1202F2
>	s_mov_b64 s[14:15], 0                                            ; BE8E0480
>	v_mov_b32_e32 v8, v9                                             ; 7E100309
>	v_cmp_lt_i32_e64 s[16:17], v7, 1                                 ; D1020010 00010307
>	v_cndmask_b32_e64 v9, 0, -1, s[16:17]                            ; D2000809 00418280
>	v_cmp_eq_i32_e64 s[16:17], v9, 0                                 ; D1040010 00010109
>	s_and_saveexec_b64 s[16:17], s[16:17]                            ; BE902410
>	s_xor_b64 s[16:17], exec, s[16:17]                               ; 8990107E
>	s_cbranch_execz BB0_2                                            ; BF880000
>	v_add_f32_e32 v5, v5, v2                                         ; 060A0505
>	v_add_f32_e32 v4, v4, v3                                         ; 06080704
>	image_sample v9, 8, 0, 0, 0, 0, 0, 0, 0, v[4:5], s[4:11], s[0:3] ; F0800800 00010904
>	v_mov_b32_e32 v10, 0x437f0000                                    ; 7E1402FF 437F0000
>	s_waitcnt vmcnt(0) expcnt(0)                                     ; BF8C0700
>	v_mul_f32_e32 v11, 0x437f0000, v9                                ; 101612FF 437F0000
>	v_add_f32_e32 v0, v0, v1                                         ; 06000300
>	v_mad_f32 v11, v11, v6, v0                                       ; D282000B 04020D0B
>	v_cmp_ge_f32_e64 s[18:19], v11, 0                                ; D00C0012 0001010B
>	v_cndmask_b32_e64 v11, 0, -1, s[18:19]                           ; D200080B 00498280
>	v_cmp_ne_i32_e64 s[18:19], v11, 0                                ; D10A0012 0001010B
>	v_mov_b32_e32 v11, 0xc2800000                                    ; 7E1602FF C2800000
>	v_mad_f32 v9, v10, v9, v11                                       ; D2820009 042E130A
>	v_mul_f32_e32 v9, 0x3c2aaace, v9                                 ; 101212FF 3C2AAACE
>	v_add_f32_e64 v9, 0, v9 clamp                                    ; D2060809 00021280
>	v_sub_f32_e32 v9, 1.0, v9                                        ; 081212F2
>	v_min_f32_e32 v9, v8, v9                                         ; 1E121308
>	v_cndmask_b32_e64 v9, v9, v8, s[18:19]                           ; D2000809 084A1109
>	v_add_i32_e32 v7, -1, v7                                         ; 4A0E0EC1
>	s_or_b64 exec, exec, s[16:17]                                    ; 88FE107E
>	s_or_b64 s[14:15], s[16:17], s[14:15]                            ; 888E0E10
>	s_andn2_b64 exec, exec, s[14:15]                                 ; 8AFE0E7E
>	s_cbranch_execnz BB0_1                                           ; BF890000
>	s_or_b64 exec, exec, s[14:15]                                    ; 88FE0E7E
>	v_sub_f32_e32 v0, 1.0, v8                                        ; 080010F2
>	v_mul_f32_e32 v0, s12, v0                                        ; 1000000C
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v0                               ; 5E000100
>	exp 15, 0, 1, 1, 1, v0, v0, v0, v0                               ; F8001C0F 00000000
>	s_endpgm                                                         ; BF810000
>
>SGPRS: 24
>VGPRS: 12
>Code Size: 572 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL OUT[2], GENERIC[10]
>DCL TEMP[0..2], LOCAL
>IMM[0] FLT32 {    2.0000,    -2.0000,     0.0000,     1.0000}
>IMM[1] FLT32 {   -1.0000,     1.0000,     0.5000,     0.0000}
>  0: MAD TEMP[0], IN[1].xyxx, IMM[0].xyzz, IMM[1].xyzy
>  1: MOV TEMP[1].xyz, IN[0].xyzx
>  2: MUL TEMP[2], IMM[0].wwzz, IN[1].xyxx
>  3: MOV OUT[2], TEMP[2]
>  4: MOV OUT[1], TEMP[1]
>  5: MOV OUT[0], TEMP[0]
>  6: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = add i32 %5, %7
>  %14 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %12, i32 0, i32 %13)
>  %15 = extractelement <4 x float> %14, i32 0
>  %16 = extractelement <4 x float> %14, i32 1
>  %17 = extractelement <4 x float> %14, i32 2
>  %18 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %19 = load <16 x i8> addrspace(2)* %18, !tbaa !0
>  %20 = add i32 %5, %7
>  %21 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %19, i32 0, i32 %20)
>  %22 = extractelement <4 x float> %21, i32 0
>  %23 = extractelement <4 x float> %21, i32 1
>  %24 = fmul float %22, 2.000000e+00
>  %25 = fadd float %24, -1.000000e+00
>  %26 = fmul float %23, -2.000000e+00
>  %27 = fadd float %26, 1.000000e+00
>  %28 = fmul float %22, 0.000000e+00
>  %29 = fadd float %28, 5.000000e-01
>  %30 = fmul float %22, 0.000000e+00
>  %31 = fadd float %30, 1.000000e+00
>  %32 = fmul float 1.000000e+00, %22
>  %33 = fmul float 1.000000e+00, %23
>  %34 = fmul float 0.000000e+00, %22
>  %35 = fmul float 0.000000e+00, %22
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %15, float %16, float %17, float 0.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %32, float %33, float %34, float %35)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %25, float %27, float %29, float %31)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                           ; 4A00000A
>	s_load_dwordx4 s[0:3], s[8:9], 0x0                  ; C0800900
>	s_load_dwordx4 s[4:7], s[8:9], 0x4                  ; C0820904
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[0:3], 0 idxen ; E00C2000 80000100
>	v_mov_b32_e32 v5, 0                                 ; 7E0A0280
>	s_waitcnt vmcnt(0)                                  ; BF8C0770
>	exp 15, 32, 0, 0, 0, v1, v2, v3, v5                 ; F800020F 05030201
>	s_waitcnt expcnt(0)                                 ; BF8C070F
>	buffer_load_format_xyzw v[0:3], v0, s[4:7], 0 idxen ; E00C2000 80010000
>	s_waitcnt vmcnt(0)                                  ; BF8C0770
>	v_mul_f32_e32 v4, 0, v0                             ; 10080080
>	exp 15, 33, 0, 0, 0, v0, v1, v4, v4                 ; F800021F 04040100
>	s_waitcnt expcnt(0)                                 ; BF8C070F
>	v_mad_f32 v4, 0, v0, 1.0                            ; D2820004 03CA0080
>	v_mad_f32 v5, 0, v0, 0.5                            ; D2820005 03C20080
>	v_mad_f32 v6, -2.0, v1, 1.0                         ; D2820006 03CA02F5
>	v_mad_f32 v0, 2.0, v0, -1.0                         ; D2820000 03CE00F4
>	exp 15, 12, 0, 1, 0, v0, v6, v5, v4                 ; F80008CF 04050600
>	s_endpgm                                            ; BF810000
>
>SGPRS: 16
>VGPRS: 8
>Code Size: 116 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL IN[1], GENERIC[10], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SAMP[1]
>DCL SAMP[2]
>DCL SAMP[3]
>DCL SAMP[4]
>DCL SAMP[5]
>DCL TEMP[0..5], LOCAL
>IMM[0] FLT32 {    0.0000,     1.0000,    -0.0000,    -1.0000}
>IMM[1] FLT32 {    0.0000,     2.0000,     2.2000,     1.0000}
>IMM[2] FLT32 {    1.7000,     0.0010,     0.4545,     0.0000}
>  0: MUL TEMP[0].xyz, IMM[0].xyxy, IN[0].yzxw
>  1: MOV TEMP[0].xyz, TEMP[0].xyzx
>  2: MAD TEMP[1].xyz, IN[0].yzxw, IMM[0].xxyy, -TEMP[0].yzxw
>  3: MOV TEMP[2].xyz, TEMP[1].xyzx
>  4: MAD TEMP[3].xyz, IN[0].zxyw, IMM[0].zzww, TEMP[0]
>  5: MOV TEMP[0].xyz, TEMP[3].xyzx
>  6: DP2 TEMP[1].x, TEMP[1].xzzz, TEMP[1].xzzz
>  7: MAX TEMP[1].x, TEMP[1].xxxx, IMM[1].xxxx
>  8: RSQ TEMP[1].x, TEMP[1].xxxx
>  9: MUL TEMP[1].xyz, TEMP[1].xxxx, TEMP[2]
> 10: MOV TEMP[2].xyz, TEMP[1].xyzx
> 11: DP2 TEMP[1].x, TEMP[3].yzzz, TEMP[3].yzzz
> 12: MAX TEMP[1].x, TEMP[1].xxxx, IMM[1].xxxx
> 13: RSQ TEMP[1].x, TEMP[1].xxxx
> 14: MOV TEMP[0].w, TEMP[1].xxxx
> 15: MUL TEMP[1].xyz, TEMP[1].xxxx, TEMP[0]
> 16: MOV TEMP[0].xyz, TEMP[1].xyzx
> 17: MUL TEMP[1], IN[1], IMM[0].ywyy
> 18: ADD TEMP[3].xy, TEMP[1], IMM[0].xyxx
> 19: MOV TEMP[3].xy, TEMP[3].xyyy
> 20: TEX TEMP[3], TEMP[3], SAMP[0], 2D
> 21: MAD TEMP[3].xyz, TEMP[3], IMM[1].yyyy, IMM[0].wwww
> 22: MUL TEMP[4].xyz, TEMP[0], TEMP[3].yyyy
> 23: MOV TEMP[0].xyz, TEMP[4].xyzx
> 24: MAD TEMP[4].xyz, TEMP[3].xxxx, TEMP[2], TEMP[0]
> 25: MOV TEMP[0].xyz, TEMP[4].xyzx
> 26: MAD TEMP[3].xyz, TEMP[3].zzzz, IN[0], TEMP[0]
> 27: MOV TEMP[4].xyz, TEMP[3].xyzz
> 28: TEX TEMP[4], TEMP[4], SAMP[4], CUBE
> 29: POW TEMP[5].x, TEMP[4].xxxx, IMM[1].zzzz
> 30: POW TEMP[5].y, TEMP[4].yyyy, IMM[1].zzzz
> 31: POW TEMP[5].z, TEMP[4].zzzz, IMM[1].zzzz
> 32: POW TEMP[5].w, TEMP[4].wwww, IMM[1].wwww
> 33: MOV TEMP[2].w, TEMP[5].wwww
> 34: MOV TEMP[3].xyz, TEMP[3].xyzz
> 35: TEX TEMP[3], TEMP[3], SAMP[3], CUBE
> 36: POW TEMP[4].x, TEMP[3].xxxx, IMM[1].zzzz
> 37: POW TEMP[4].y, TEMP[3].yyyy, IMM[1].zzzz
> 38: POW TEMP[4].z, TEMP[3].zzzz, IMM[1].zzzz
> 39: POW TEMP[4].w, TEMP[3].wwww, IMM[1].wwww
> 40: MOV TEMP[0].w, TEMP[4].wwww
> 41: MUL TEMP[3].xyz, TEMP[4], IMM[2].xxxx
> 42: MOV TEMP[0].xyz, TEMP[3].xyzx
> 43: MAD TEMP[3].xyz, TEMP[5], IMM[2].xxxx, -TEMP[0]
> 44: MOV TEMP[2].xyz, TEMP[3].xyzx
> 45: ADD TEMP[3].xy, TEMP[1], IMM[0].xyxx
> 46: MOV TEMP[3].xy, TEMP[3].xyyy
> 47: TEX TEMP[3].x, TEMP[3], SAMP[2], 2D
> 48: MAD TEMP[3].xyz, TEMP[3].xxxx, TEMP[2], TEMP[0]
> 49: MOV TEMP[0].xyz, TEMP[3].xyzx
> 50: ADD TEMP[1].xy, TEMP[1], IMM[0].xyxx
> 51: MOV TEMP[1].xy, TEMP[1].xyyy
> 52: TEX TEMP[1], TEMP[1], SAMP[1], 2D
> 53: MOV TEMP[2].w, TEMP[1].wwww
> 54: MAD TEMP[0].xyz, TEMP[1], TEMP[0], IMM[2].yyyy
> 55: ABS TEMP[1].x, TEMP[0].xxxx
> 56: LG2 TEMP[2].x, TEMP[1].xxxx
> 57: ABS TEMP[1].x, TEMP[0].yyyy
> 58: LG2 TEMP[1].x, TEMP[1].xxxx
> 59: MOV TEMP[2].y, TEMP[1].xxxx
> 60: ABS TEMP[0].x, TEMP[0].zzzz
> 61: LG2 TEMP[0].x, TEMP[0].xxxx
> 62: MOV TEMP[2].z, TEMP[0].xxxx
> 63: MUL TEMP[0].xyz, TEMP[2], IMM[2].zzzz
> 64: EX2 TEMP[2].x, TEMP[0].xxxx
> 65: EX2 TEMP[1].x, TEMP[0].yyyy
> 66: MOV TEMP[2].y, TEMP[1].xxxx
> 67: EX2 TEMP[0].x, TEMP[0].zzzz
> 68: MOV TEMP[2].z, TEMP[0].xxxx
> 69: MOV TEMP[0].xyz, TEMP[2].xyzz
> 70: TEX TEMP[0].xyz, TEMP[0], SAMP[5], 3D
> 71: MOV TEMP[0].xyz, TEMP[0].xyzx
> 72: MOV TEMP[0].w, IMM[0].yyyy
> 73: MOV OUT[0], TEMP[0]
> 74: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %23 = load <8 x i32> addrspace(2)* %22, !tbaa !0
>  %24 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %25 = load <4 x i32> addrspace(2)* %24, !tbaa !0
>  %26 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 1
>  %27 = load <8 x i32> addrspace(2)* %26, !tbaa !0
>  %28 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 1
>  %29 = load <4 x i32> addrspace(2)* %28, !tbaa !0
>  %30 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 2
>  %31 = load <8 x i32> addrspace(2)* %30, !tbaa !0
>  %32 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 2
>  %33 = load <4 x i32> addrspace(2)* %32, !tbaa !0
>  %34 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 3
>  %35 = load <8 x i32> addrspace(2)* %34, !tbaa !0
>  %36 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 3
>  %37 = load <4 x i32> addrspace(2)* %36, !tbaa !0
>  %38 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 4
>  %39 = load <8 x i32> addrspace(2)* %38, !tbaa !0
>  %40 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 4
>  %41 = load <4 x i32> addrspace(2)* %40, !tbaa !0
>  %42 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 5
>  %43 = load <8 x i32> addrspace(2)* %42, !tbaa !0
>  %44 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 5
>  %45 = load <4 x i32> addrspace(2)* %44, !tbaa !0
>  %46 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %47 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %48 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %5, <2 x i32> %7)
>  %49 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %5, <2 x i32> %7)
>  %50 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %5, <2 x i32> %7)
>  %51 = fmul float 0.000000e+00, %47
>  %52 = fmul float 1.000000e+00, %48
>  %53 = fmul float 0.000000e+00, %46
>  %54 = fsub float -0.000000e+00, %52
>  %55 = fmul float %47, 0.000000e+00
>  %56 = fadd float %55, %54
>  %57 = fsub float -0.000000e+00, %53
>  %58 = fmul float %48, 0.000000e+00
>  %59 = fadd float %58, %57
>  %60 = fsub float -0.000000e+00, %51
>  %61 = fmul float %46, 1.000000e+00
>  %62 = fadd float %61, %60
>  %63 = fmul float %48, -0.000000e+00
>  %64 = fadd float %63, %51
>  %65 = fmul float %46, -0.000000e+00
>  %66 = fadd float %65, %52
>  %67 = fmul float %47, -1.000000e+00
>  %68 = fadd float %67, %53
>  %69 = fmul float %56, %56
>  %70 = fmul float %62, %62
>  %71 = fadd float %69, %70
>  %72 = call float @llvm.maxnum.f32(float %71, float 0x3E7AD7F2A0000000)
>  %73 = call float @llvm.AMDGPU.rsq.clamped.f32(float %72)
>  %74 = fmul float %73, %56
>  %75 = fmul float %73, %59
>  %76 = fmul float %73, %62
>  %77 = fmul float %66, %66
>  %78 = fmul float %68, %68
>  %79 = fadd float %77, %78
>  %80 = call float @llvm.maxnum.f32(float %79, float 0x3E7AD7F2A0000000)
>  %81 = call float @llvm.AMDGPU.rsq.clamped.f32(float %80)
>  %82 = fmul float %81, %64
>  %83 = fmul float %81, %66
>  %84 = fmul float %81, %68
>  %85 = fmul float %49, 1.000000e+00
>  %86 = fmul float %50, -1.000000e+00
>  %87 = fadd float %85, 0.000000e+00
>  %88 = fadd float %86, 1.000000e+00
>  %89 = bitcast float %87 to i32
>  %90 = bitcast float %88 to i32
>  %91 = insertelement <2 x i32> undef, i32 %89, i32 0
>  %92 = insertelement <2 x i32> %91, i32 %90, i32 1
>  %93 = bitcast <8 x i32> %23 to <32 x i8>
>  %94 = bitcast <4 x i32> %25 to <16 x i8>
>  %95 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %92, <32 x i8> %93, <16 x i8> %94, i32 2)
>  %96 = extractelement <4 x float> %95, i32 0
>  %97 = extractelement <4 x float> %95, i32 1
>  %98 = extractelement <4 x float> %95, i32 2
>  %99 = extractelement <4 x float> %95, i32 3
>  %100 = fmul float %96, 2.000000e+00
>  %101 = fadd float %100, -1.000000e+00
>  %102 = fmul float %97, 2.000000e+00
>  %103 = fadd float %102, -1.000000e+00
>  %104 = fmul float %98, 2.000000e+00
>  %105 = fadd float %104, -1.000000e+00
>  %106 = fmul float %82, %103
>  %107 = fmul float %83, %103
>  %108 = fmul float %84, %103
>  %109 = fmul float %101, %74
>  %110 = fadd float %109, %106
>  %111 = fmul float %101, %75
>  %112 = fadd float %111, %107
>  %113 = fmul float %101, %76
>  %114 = fadd float %113, %108
>  %115 = fmul float %105, %46
>  %116 = fadd float %115, %110
>  %117 = fmul float %105, %47
>  %118 = fadd float %117, %112
>  %119 = fmul float %105, %48
>  %120 = fadd float %119, %114
>  %121 = insertelement <4 x float> undef, float %116, i32 0
>  %122 = insertelement <4 x float> %121, float %118, i32 1
>  %123 = insertelement <4 x float> %122, float %120, i32 2
>  %124 = insertelement <4 x float> %123, float 0.000000e+00, i32 3
>  %125 = call <4 x float> @llvm.AMDGPU.cube(<4 x float> %124)
>  %126 = extractelement <4 x float> %125, i32 0
>  %127 = extractelement <4 x float> %125, i32 1
>  %128 = extractelement <4 x float> %125, i32 2
>  %129 = extractelement <4 x float> %125, i32 3
>  %130 = call float @fabs(float %128)
>  %131 = fdiv float 1.000000e+00, %130
>  %132 = fmul float %126, %131
>  %133 = fadd float %132, 1.500000e+00
>  %134 = fmul float %127, %131
>  %135 = fadd float %134, 1.500000e+00
>  %136 = bitcast float %135 to i32
>  %137 = bitcast float %133 to i32
>  %138 = bitcast float %129 to i32
>  %139 = insertelement <4 x i32> undef, i32 %136, i32 0
>  %140 = insertelement <4 x i32> %139, i32 %137, i32 1
>  %141 = insertelement <4 x i32> %140, i32 %138, i32 2
>  %142 = insertelement <4 x i32> %141, i32 undef, i32 3
>  %143 = bitcast <8 x i32> %39 to <32 x i8>
>  %144 = bitcast <4 x i32> %41 to <16 x i8>
>  %145 = call <4 x float> @llvm.SI.sample.v4i32(<4 x i32> %142, <32 x i8> %143, <16 x i8> %144, i32 4)
>  %146 = extractelement <4 x float> %145, i32 0
>  %147 = extractelement <4 x float> %145, i32 1
>  %148 = extractelement <4 x float> %145, i32 2
>  %149 = call float @llvm.pow.f32(float %146, float 0x40019999A0000000)
>  %150 = call float @llvm.pow.f32(float %147, float 0x40019999A0000000)
>  %151 = call float @llvm.pow.f32(float %148, float 0x40019999A0000000)
>  %152 = insertelement <4 x float> undef, float %116, i32 0
>  %153 = insertelement <4 x float> %152, float %118, i32 1
>  %154 = insertelement <4 x float> %153, float %120, i32 2
>  %155 = insertelement <4 x float> %154, float %99, i32 3
>  %156 = call <4 x float> @llvm.AMDGPU.cube(<4 x float> %155)
>  %157 = extractelement <4 x float> %156, i32 0
>  %158 = extractelement <4 x float> %156, i32 1
>  %159 = extractelement <4 x float> %156, i32 2
>  %160 = extractelement <4 x float> %156, i32 3
>  %161 = call float @fabs(float %159)
>  %162 = fdiv float 1.000000e+00, %161
>  %163 = fmul float %157, %162
>  %164 = fadd float %163, 1.500000e+00
>  %165 = fmul float %158, %162
>  %166 = fadd float %165, 1.500000e+00
>  %167 = bitcast float %166 to i32
>  %168 = bitcast float %164 to i32
>  %169 = bitcast float %160 to i32
>  %170 = insertelement <4 x i32> undef, i32 %167, i32 0
>  %171 = insertelement <4 x i32> %170, i32 %168, i32 1
>  %172 = insertelement <4 x i32> %171, i32 %169, i32 2
>  %173 = insertelement <4 x i32> %172, i32 undef, i32 3
>  %174 = bitcast <8 x i32> %35 to <32 x i8>
>  %175 = bitcast <4 x i32> %37 to <16 x i8>
>  %176 = call <4 x float> @llvm.SI.sample.v4i32(<4 x i32> %173, <32 x i8> %174, <16 x i8> %175, i32 4)
>  %177 = extractelement <4 x float> %176, i32 0
>  %178 = extractelement <4 x float> %176, i32 1
>  %179 = extractelement <4 x float> %176, i32 2
>  %180 = call float @llvm.pow.f32(float %177, float 0x40019999A0000000)
>  %181 = call float @llvm.pow.f32(float %178, float 0x40019999A0000000)
>  %182 = call float @llvm.pow.f32(float %179, float 0x40019999A0000000)
>  %183 = fmul float %180, 0x3FFB333340000000
>  %184 = fmul float %181, 0x3FFB333340000000
>  %185 = fmul float %182, 0x3FFB333340000000
>  %186 = fsub float -0.000000e+00, %183
>  %187 = fmul float %149, 0x3FFB333340000000
>  %188 = fadd float %187, %186
>  %189 = fsub float -0.000000e+00, %184
>  %190 = fmul float %150, 0x3FFB333340000000
>  %191 = fadd float %190, %189
>  %192 = fsub float -0.000000e+00, %185
>  %193 = fmul float %151, 0x3FFB333340000000
>  %194 = fadd float %193, %192
>  %195 = fadd float %85, 0.000000e+00
>  %196 = fadd float %86, 1.000000e+00
>  %197 = bitcast float %195 to i32
>  %198 = bitcast float %196 to i32
>  %199 = insertelement <2 x i32> undef, i32 %197, i32 0
>  %200 = insertelement <2 x i32> %199, i32 %198, i32 1
>  %201 = bitcast <8 x i32> %31 to <32 x i8>
>  %202 = bitcast <4 x i32> %33 to <16 x i8>
>  %203 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %200, <32 x i8> %201, <16 x i8> %202, i32 2)
>  %204 = extractelement <4 x float> %203, i32 0
>  %205 = fmul float %204, %188
>  %206 = fadd float %205, %183
>  %207 = fmul float %204, %191
>  %208 = fadd float %207, %184
>  %209 = fmul float %204, %194
>  %210 = fadd float %209, %185
>  %211 = fadd float %85, 0.000000e+00
>  %212 = fadd float %86, 1.000000e+00
>  %213 = bitcast float %211 to i32
>  %214 = bitcast float %212 to i32
>  %215 = insertelement <2 x i32> undef, i32 %213, i32 0
>  %216 = insertelement <2 x i32> %215, i32 %214, i32 1
>  %217 = bitcast <8 x i32> %27 to <32 x i8>
>  %218 = bitcast <4 x i32> %29 to <16 x i8>
>  %219 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %216, <32 x i8> %217, <16 x i8> %218, i32 2)
>  %220 = extractelement <4 x float> %219, i32 0
>  %221 = extractelement <4 x float> %219, i32 1
>  %222 = extractelement <4 x float> %219, i32 2
>  %223 = fmul float %220, %206
>  %224 = fadd float %223, 0x3F50624DE0000000
>  %225 = fmul float %221, %208
>  %226 = fadd float %225, 0x3F50624DE0000000
>  %227 = fmul float %222, %210
>  %228 = fadd float %227, 0x3F50624DE0000000
>  %229 = call float @fabs(float %224)
>  %230 = call float @llvm.log2.f32(float %229)
>  %231 = call float @fabs(float %226)
>  %232 = call float @llvm.log2.f32(float %231)
>  %233 = call float @fabs(float %228)
>  %234 = call float @llvm.log2.f32(float %233)
>  %235 = fmul float %230, 0x3FDD1743E0000000
>  %236 = fmul float %232, 0x3FDD1743E0000000
>  %237 = fmul float %234, 0x3FDD1743E0000000
>  %238 = call float @llvm.AMDIL.exp.(float %235)
>  %239 = call float @llvm.AMDIL.exp.(float %236)
>  %240 = call float @llvm.AMDIL.exp.(float %237)
>  %241 = bitcast float %238 to i32
>  %242 = bitcast float %239 to i32
>  %243 = bitcast float %240 to i32
>  %244 = insertelement <4 x i32> undef, i32 %241, i32 0
>  %245 = insertelement <4 x i32> %244, i32 %242, i32 1
>  %246 = insertelement <4 x i32> %245, i32 %243, i32 2
>  %247 = insertelement <4 x i32> %246, i32 undef, i32 3
>  %248 = bitcast <8 x i32> %43 to <32 x i8>
>  %249 = bitcast <4 x i32> %45 to <16 x i8>
>  %250 = call <4 x float> @llvm.SI.sample.v4i32(<4 x i32> %247, <32 x i8> %248, <16 x i8> %249, i32 3)
>  %251 = extractelement <4 x float> %250, i32 0
>  %252 = extractelement <4 x float> %250, i32 1
>  %253 = extractelement <4 x float> %250, i32 2
>  %254 = call i32 @llvm.SI.packf16(float %251, float %252)
>  %255 = bitcast i32 %254 to float
>  %256 = call i32 @llvm.SI.packf16(float %253, float 1.000000e+00)
>  %257 = bitcast i32 %256 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %255, float %257, float %255, float %257)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.maxnum.f32(float, float) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.AMDGPU.rsq.clamped.f32(float) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: readnone
>declare <4 x float> @llvm.AMDGPU.cube(<4 x float>) #2
>
>; Function Attrs: readnone
>declare float @fabs(float) #2
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v4i32(<4 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.pow.f32(float, float) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.log2.f32(float) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.exp.(float) #2
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                         ; BEFE0A7E
>	s_mov_b32 m0, s9                                                             ; BEFC0309
>	v_interp_p1_f32 v2, v0, 2, 0, [m0]                                           ; C8080200
>	v_interp_p2_f32 v2, [v2], v1, 2, 0, [m0]                                     ; C8090201
>	v_interp_p1_f32 v3, v0, 1, 0, [m0]                                           ; C80C0100
>	v_interp_p2_f32 v3, [v3], v1, 1, 0, [m0]                                     ; C80D0101
>	v_mov_b32_e32 v4, 0                                                          ; 7E080280
>	v_mad_f32 v5, 0, v3, -v2                                                     ; D2820005 840A0680
>	v_interp_p1_f32 v6, v0, 0, 0, [m0]                                           ; C8180000
>	v_interp_p2_f32 v6, [v6], v1, 0, 0, [m0]                                     ; C8190001
>	v_mad_f32 v7, 0, -v3, v6                                                     ; D2820007 441A0680
>	v_mul_f32_e32 v8, v7, v7                                                     ; 10100F07
>	v_mad_f32 v8, v5, v5, v8                                                     ; D2820008 04220B05
>	v_max_f32_e32 v8, 0x33d6bf95, v8                                             ; 201010FF 33D6BF95
>	v_rsq_clamp_f32_e32 v8, v8                                                   ; 7E105908
>	v_mul_f32_e32 v7, v7, v8                                                     ; 100E1107
>	v_mov_b32_e32 v9, 0x80000000                                                 ; 7E1202FF 80000000
>	v_mad_f32 v10, v6, v9, v2                                                    ; D282000A 040A1306
>	v_mad_f32 v11, 0, v6, -v3                                                    ; D282000B 840E0C80
>	v_mul_f32_e32 v12, v11, v11                                                  ; 1018170B
>	v_mad_f32 v12, v10, v10, v12                                                 ; D282000C 0432150A
>	v_max_f32_e32 v12, 0x33d6bf95, v12                                           ; 201818FF 33D6BF95
>	v_rsq_clamp_f32_e32 v12, v12                                                 ; 7E18590C
>	v_mul_f32_e32 v11, v11, v12                                                  ; 1016190B
>	v_interp_p1_f32 v13, v0, 1, 1, [m0]                                          ; C8340500
>	v_interp_p2_f32 v13, [v13], v1, 1, 1, [m0]                                   ; C8350501
>	v_sub_f32_e32 v14, 1.0, v13                                                  ; 081C1AF2
>	v_interp_p1_f32 v15, v0, 0, 1, [m0]                                          ; C83C0400
>	v_interp_p2_f32 v15, [v15], v1, 0, 1, [m0]                                   ; C83D0401
>	v_add_f32_e32 v13, 0, v15                                                    ; 061A1E80
>	s_load_dwordx4 s[60:63], s[4:5], 0x0                                         ; C09E0500
>	s_load_dwordx4 s[12:15], s[4:5], 0x4                                         ; C0860504
>	s_load_dwordx4 s[16:19], s[4:5], 0x8                                         ; C0880508
>	s_load_dwordx4 s[48:51], s[4:5], 0xc                                         ; C098050C
>	s_load_dwordx4 s[36:39], s[4:5], 0x10                                        ; C0920510
>	s_load_dwordx4 s[0:3], s[4:5], 0x14                                          ; C0800514
>	s_load_dwordx8 s[64:71], s[6:7], 0x0                                         ; C0E00700
>	s_load_dwordx8 s[20:27], s[6:7], 0x8                                         ; C0CA0708
>	s_load_dwordx8 s[28:35], s[6:7], 0x10                                        ; C0CE0710
>	s_load_dwordx8 s[52:59], s[6:7], 0x18                                        ; C0DA0718
>	s_load_dwordx8 s[40:47], s[6:7], 0x20                                        ; C0D40720
>	s_load_dwordx8 s[4:11], s[6:7], 0x28                                         ; C0C20728
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample v[15:18], 15, 0, 0, 0, 0, 0, 0, 0, v[13:14], s[64:71], s[60:63] ; F0800F00 01F00F0D
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v0, 2.0, v16, -1.0                                                 ; D2820000 03CE20F4
>	v_mul_f32_e32 v1, v0, v11                                                    ; 10021700
>	v_mad_f32 v11, 2.0, v15, -1.0                                                ; D282000B 03CE1EF4
>	v_mad_f32 v1, v11, v7, v1                                                    ; D2820001 04060F0B
>	v_mad_f32 v7, 2.0, v17, -1.0                                                 ; D2820007 03CE22F4
>	v_mad_f32 v17, v7, v2, v1                                                    ; D2820011 04060507
>	v_mul_f32_e32 v1, v9, v6                                                     ; 10020D09
>	v_mad_f32 v1, 0, v2, v1                                                      ; D2820001 04060480
>	v_mul_f32_e32 v1, v1, v8                                                     ; 10021101
>	v_mul_f32_e32 v10, v10, v12                                                  ; 1014190A
>	v_mul_f32_e32 v10, v0, v10                                                   ; 10141500
>	v_mad_f32 v1, v11, v1, v10                                                   ; D2820001 042A030B
>	v_mad_f32 v16, v7, v3, v1                                                    ; D2820010 04060707
>	v_mul_f32_e32 v1, v5, v8                                                     ; 10021105
>	v_mul_f32_e32 v3, 0, v3                                                      ; 10060680
>	v_mad_f32 v2, v2, v9, v3                                                     ; D2820002 040E1302
>	v_mul_f32_e32 v2, v2, v12                                                    ; 10041902
>	v_mul_f32_e32 v0, v0, v2                                                     ; 10000500
>	v_mad_f32 v0, v11, v1, v0                                                    ; D2820000 0402030B
>	v_mad_f32 v15, v7, v6, v0                                                    ; D282000F 04020D07
>	v_cubeid_f32 v8, v15, v16, v17                                               ; D2880008 0446210F
>	v_cubema_f32 v7, v15, v16, v17                                               ; D28E0007 0446210F
>	v_cubesc_f32 v6, v15, v16, v17                                               ; D28A0006 0446210F
>	v_cubetc_f32 v5, v15, v16, v17                                               ; D28C0005 0446210F
>	v_rcp_f32_e64 v0, |v7|                                                       ; D3540100 00000107
>	v_mov_b32_e32 v1, 0x3fc00000                                                 ; 7E0202FF 3FC00000
>	v_mad_f32 v7, v5, v0, v1                                                     ; D2820007 04060105
>	v_mad_f32 v6, v6, v0, v1                                                     ; D2820006 04060106
>	image_sample v[5:7], 7, 0, 0, 0, 0, 0, 0, 0, v[6:9], s[52:59], s[48:51]      ; F0800700 018D0506
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_log_f32_e32 v0, v7                                                         ; 7E004F07
>	v_mov_b32_e32 v2, 0x400ccccd                                                 ; 7E0402FF 400CCCCD
>	v_mul_legacy_f32_e32 v0, v2, v0                                              ; 0E000102
>	v_exp_f32_e32 v0, v0                                                         ; 7E004B00
>	v_mov_b32_e32 v3, 0xbfd9999a                                                 ; 7E0602FF BFD9999A
>	v_mul_f32_e32 v8, v3, v0                                                     ; 10100103
>	v_mov_b32_e32 v18, v4                                                        ; 7E240304
>	v_cubeid_f32 v22, v15, v16, v17                                              ; D2880016 0446210F
>	v_cubema_f32 v21, v15, v16, v17                                              ; D28E0015 0446210F
>	v_cubesc_f32 v20, v15, v16, v17                                              ; D28A0014 0446210F
>	v_cubetc_f32 v19, v15, v16, v17                                              ; D28C0013 0446210F
>	v_rcp_f32_e64 v4, |v21|                                                      ; D3540104 00000115
>	v_mad_f32 v21, v19, v4, v1                                                   ; D2820015 04060913
>	v_mad_f32 v20, v20, v4, v1                                                   ; D2820014 04060914
>	image_sample v[9:11], 7, 0, 0, 0, 0, 0, 0, 0, v[20:23], s[40:47], s[36:39]   ; F0800700 012A0914
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_log_f32_e32 v1, v11                                                        ; 7E024F0B
>	v_mul_legacy_f32_e32 v1, v2, v1                                              ; 0E020302
>	v_exp_f32_e32 v1, v1                                                         ; 7E024B01
>	v_mov_b32_e32 v4, 0x3fd9999a                                                 ; 7E0802FF 3FD9999A
>	v_mad_f32 v1, v1, v4, v8                                                     ; D2820001 04220901
>	v_mul_f32_e32 v0, 0x3fd9999a, v0                                             ; 100000FF 3FD9999A
>	image_sample v8, 1, 0, 0, 0, 0, 0, 0, 0, v[13:14], s[28:35], s[16:19]        ; F0800100 0087080D
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v0, v8, v1, v0                                                     ; D2820000 04020308
>	image_sample v[12:14], 7, 0, 0, 0, 0, 0, 0, 0, v[13:14], s[20:27], s[12:15]  ; F0800700 00650C0D
>	v_mov_b32_e32 v1, 0x3a83126f                                                 ; 7E0202FF 3A83126F
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v0, v14, v0, v1                                                    ; D2820000 0406010E
>	v_log_f32_e64 v0, |v0|                                                       ; D34E0100 00000100
>	v_mul_f32_e32 v0, 0x3ee8ba1f, v0                                             ; 100000FF 3EE8BA1F
>	v_exp_f32_e32 v17, v0                                                        ; 7E224B00
>	v_log_f32_e32 v0, v6                                                         ; 7E004F06
>	v_mul_legacy_f32_e32 v0, v2, v0                                              ; 0E000102
>	v_exp_f32_e32 v0, v0                                                         ; 7E004B00
>	v_mul_f32_e32 v19, v3, v0                                                    ; 10260103
>	v_log_f32_e32 v20, v10                                                       ; 7E284F0A
>	v_mul_legacy_f32_e32 v20, v2, v20                                            ; 0E282902
>	v_exp_f32_e32 v20, v20                                                       ; 7E284B14
>	v_mad_f32 v19, v20, v4, v19                                                  ; D2820013 044E0914
>	v_mul_f32_e32 v0, 0x3fd9999a, v0                                             ; 100000FF 3FD9999A
>	v_mad_f32 v0, v8, v19, v0                                                    ; D2820000 04022708
>	v_mad_f32 v0, v13, v0, v1                                                    ; D2820000 0406010D
>	v_log_f32_e64 v0, |v0|                                                       ; D34E0100 00000100
>	v_mul_f32_e32 v0, 0x3ee8ba1f, v0                                             ; 100000FF 3EE8BA1F
>	v_exp_f32_e32 v16, v0                                                        ; 7E204B00
>	v_log_f32_e32 v0, v5                                                         ; 7E004F05
>	v_mul_legacy_f32_e32 v0, v2, v0                                              ; 0E000102
>	v_exp_f32_e32 v0, v0                                                         ; 7E004B00
>	v_mul_f32_e32 v3, v3, v0                                                     ; 10060103
>	v_log_f32_e32 v5, v9                                                         ; 7E0A4F09
>	v_mul_legacy_f32_e32 v2, v2, v5                                              ; 0E040B02
>	v_exp_f32_e32 v2, v2                                                         ; 7E044B02
>	v_mad_f32 v2, v2, v4, v3                                                     ; D2820002 040E0902
>	v_mul_f32_e32 v0, 0x3fd9999a, v0                                             ; 100000FF 3FD9999A
>	v_mad_f32 v0, v8, v2, v0                                                     ; D2820000 04020508
>	v_mad_f32 v0, v12, v0, v1                                                    ; D2820000 0406010C
>	v_log_f32_e64 v0, |v0|                                                       ; D34E0100 00000100
>	v_mul_f32_e32 v0, 0x3ee8ba1f, v0                                             ; 100000FF 3EE8BA1F
>	v_exp_f32_e32 v15, v0                                                        ; 7E1E4B00
>	image_sample v[0:2], 7, 0, 0, 0, 0, 0, 0, 0, v[15:18], s[4:11], s[0:3]       ; F0800700 0001000F
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_cvt_pkrtz_f16_f32_e32 v3, v0, v1                                           ; 5E060300
>	v_cvt_pkrtz_f16_f32_e64 v0, v2, 1.0                                          ; D25E0000 0001E502
>	exp 15, 0, 1, 1, 1, v3, v0, v3, v0                                           ; F8001C0F 00030003
>	s_endpgm                                                                     ; BF810000
>
>SGPRS: 72
>VGPRS: 28
>Code Size: 824 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL CONST[0..6]
>DCL TEMP[0..2], LOCAL
>IMM[0] FLT32 {    1.0000,     0.0000,     0.0000,     0.0000}
>  0: LRP TEMP[0].xy, IN[0], CONST[5].zwzw, CONST[5]
>  1: MOV TEMP[0].xy, TEMP[0].xyxx
>  2: ADD TEMP[1].xy, TEMP[0], IMM[0].xxxx
>  3: MUL TEMP[2], TEMP[1].yyyy, CONST[1]
>  4: MAD TEMP[0], TEMP[1].xxxx, CONST[0], TEMP[2]
>  5: MAD TEMP[0], CONST[6].xxxx, CONST[2], TEMP[0]
>  6: ADD TEMP[0], TEMP[0], CONST[3]
>  7: MAD TEMP[1].xy, IN[1], CONST[4], CONST[4].zwzw
>  8: MOV TEMP[1].xy, TEMP[1].xyxx
>  9: MOV TEMP[1].zw, IMM[0].yyyy
> 10: MOV OUT[1], TEMP[1]
> 11: MOV OUT[0], TEMP[0]
> 12: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 0)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 4)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %12, i32 8)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %12, i32 12)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %12, i32 16)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %12, i32 20)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %12, i32 24)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %12, i32 28)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %12, i32 32)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %12, i32 36)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %12, i32 40)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %12, i32 44)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %12, i32 48)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %12, i32 52)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %12, i32 56)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %12, i32 60)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %12, i32 64)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %12, i32 68)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %12, i32 72)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %12, i32 76)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %12, i32 80)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %12, i32 84)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %12, i32 88)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %12, i32 92)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %12, i32 96)
>  %38 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %39 = load <16 x i8> addrspace(2)* %38, !tbaa !0
>  %40 = add i32 %5, %7
>  %41 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %39, i32 0, i32 %40)
>  %42 = extractelement <4 x float> %41, i32 0
>  %43 = extractelement <4 x float> %41, i32 1
>  %44 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %45 = load <16 x i8> addrspace(2)* %44, !tbaa !0
>  %46 = add i32 %5, %7
>  %47 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %45, i32 0, i32 %46)
>  %48 = extractelement <4 x float> %47, i32 0
>  %49 = extractelement <4 x float> %47, i32 1
>  %50 = call float @llvm.AMDGPU.lrp(float %42, float %35, float %33)
>  %51 = call float @llvm.AMDGPU.lrp(float %43, float %36, float %34)
>  %52 = fadd float %50, 1.000000e+00
>  %53 = fadd float %51, 1.000000e+00
>  %54 = fmul float %53, %17
>  %55 = fmul float %53, %18
>  %56 = fmul float %53, %19
>  %57 = fmul float %53, %20
>  %58 = fmul float %52, %13
>  %59 = fadd float %58, %54
>  %60 = fmul float %52, %14
>  %61 = fadd float %60, %55
>  %62 = fmul float %52, %15
>  %63 = fadd float %62, %56
>  %64 = fmul float %52, %16
>  %65 = fadd float %64, %57
>  %66 = fmul float %37, %21
>  %67 = fadd float %66, %59
>  %68 = fmul float %37, %22
>  %69 = fadd float %68, %61
>  %70 = fmul float %37, %23
>  %71 = fadd float %70, %63
>  %72 = fmul float %37, %24
>  %73 = fadd float %72, %65
>  %74 = fadd float %67, %25
>  %75 = fadd float %69, %26
>  %76 = fadd float %71, %27
>  %77 = fadd float %73, %28
>  %78 = fmul float %48, %29
>  %79 = fadd float %78, %31
>  %80 = fmul float %49, %30
>  %81 = fadd float %80, %32
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %79, float %81, float 0.000000e+00, float 0.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %74, float %75, float %76, float %77)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDGPU.lrp(float, float, float) #2
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                            ; 4A00000A
>	s_load_dwordx4 s[4:7], s[8:9], 0x0                   ; C0820900
>	s_load_dwordx4 s[8:11], s[8:9], 0x4                  ; C0840904
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[8:11], 0 idxen ; E00C2000 80020100
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                   ; C0800300
>	s_waitcnt vmcnt(0) lgkmcnt(0)                        ; BF8C0070
>	s_buffer_load_dword s8, s[0:3], 0x13                 ; C2040113
>	s_buffer_load_dword s9, s[0:3], 0x11                 ; C2048111
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mov_b32_e32 v5, s8                                 ; 7E0A0208
>	v_mad_f32 v5, s9, v2, v5                             ; D2820005 04160409
>	s_buffer_load_dword s8, s[0:3], 0x12                 ; C2040112
>	s_buffer_load_dword s9, s[0:3], 0x10                 ; C2048110
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mov_b32_e32 v6, s8                                 ; 7E0C0208
>	v_mad_f32 v1, s9, v1, v6                             ; D2820001 041A0209
>	v_mov_b32_e32 v2, 0                                  ; 7E040280
>	exp 15, 32, 0, 0, 0, v1, v5, v2, v2                  ; F800020F 02020501
>	s_waitcnt expcnt(0)                                  ; BF8C070F
>	buffer_load_format_xyzw v[0:3], v0, s[4:7], 0 idxen  ; E00C2000 80010000
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_sub_f32_e32 v4, 1.0, v0                            ; 080800F2
>	s_buffer_load_dword s4, s[0:3], 0x14                 ; C2020114
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v4, s4, v4                             ; 10080804
>	s_buffer_load_dword s4, s[0:3], 0x16                 ; C2020116
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v4, v0, s4, v4                             ; D2820004 04100900
>	v_add_f32_e32 v4, 1.0, v4                            ; 060808F2
>	v_sub_f32_e32 v5, 1.0, v1                            ; 080A02F2
>	s_buffer_load_dword s4, s[0:3], 0x15                 ; C2020115
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v5, s4, v5                             ; 100A0A04
>	s_buffer_load_dword s4, s[0:3], 0x17                 ; C2020117
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v0, v1, s4, v5                             ; D2820000 04140901
>	v_add_f32_e32 v0, 1.0, v0                            ; 060000F2
>	s_buffer_load_dword s4, s[0:3], 0x7                  ; C2020107
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v1, s4, v0                             ; 10020004
>	s_buffer_load_dword s4, s[0:3], 0x3                  ; C2020103
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v1, v4, s4, v1                             ; D2820001 04040904
>	s_buffer_load_dword s4, s[0:3], 0xb                  ; C202010B
>	s_buffer_load_dword s5, s[0:3], 0x18                 ; C2028118
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mov_b32_e32 v2, s4                                 ; 7E040204
>	v_mad_f32 v1, v2, s5, v1                             ; D2820001 04040B02
>	s_buffer_load_dword s4, s[0:3], 0xf                  ; C202010F
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_add_f32_e32 v1, s4, v1                             ; 06020204
>	s_buffer_load_dword s4, s[0:3], 0x6                  ; C2020106
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v2, s4, v0                             ; 10040004
>	s_buffer_load_dword s4, s[0:3], 0x2                  ; C2020102
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v2, v4, s4, v2                             ; D2820002 04080904
>	s_buffer_load_dword s4, s[0:3], 0xa                  ; C202010A
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mov_b32_e32 v3, s4                                 ; 7E060204
>	v_mad_f32 v2, v3, s5, v2                             ; D2820002 04080B03
>	s_buffer_load_dword s4, s[0:3], 0xe                  ; C202010E
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_add_f32_e32 v2, s4, v2                             ; 06040404
>	s_buffer_load_dword s4, s[0:3], 0x5                  ; C2020105
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v3, s4, v0                             ; 10060004
>	s_buffer_load_dword s4, s[0:3], 0x1                  ; C2020101
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v3, v4, s4, v3                             ; D2820003 040C0904
>	s_buffer_load_dword s4, s[0:3], 0x9                  ; C2020109
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mov_b32_e32 v5, s4                                 ; 7E0A0204
>	v_mad_f32 v3, v5, s5, v3                             ; D2820003 040C0B05
>	s_buffer_load_dword s4, s[0:3], 0xd                  ; C202010D
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_add_f32_e32 v3, s4, v3                             ; 06060604
>	s_buffer_load_dword s4, s[0:3], 0x4                  ; C2020104
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v0, s4, v0                             ; 10000004
>	s_buffer_load_dword s4, s[0:3], 0x0                  ; C2020100
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v0, v4, s4, v0                             ; D2820000 04000904
>	s_buffer_load_dword s4, s[0:3], 0x8                  ; C2020108
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mov_b32_e32 v4, s4                                 ; 7E080204
>	v_mad_f32 v0, v4, s5, v0                             ; D2820000 04000B04
>	s_buffer_load_dword s0, s[0:3], 0xc                  ; C200010C
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_add_f32_e32 v0, s0, v0                             ; 06000000
>	exp 15, 12, 0, 1, 0, v0, v3, v2, v1                  ; F80008CF 01020300
>	s_endpgm                                             ; BF810000
>
>SGPRS: 16
>VGPRS: 8
>Code Size: 436 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL OUT[1], COLOR[1]
>DCL SAMP[0]
>DCL SAMP[1]
>DCL TEMP[0..5], LOCAL
>IMM[0] FLT32 {    2.2000,     1.0000,     2.0000,    -0.9999}
>IMM[1] FLT32 {    0.0000,    -1.0000,    -2.0000,     0.0000}
>  0: MOV TEMP[0].xy, IN[0].xyyy
>  1: TEX TEMP[0], TEMP[0], SAMP[0], 2D
>  2: POW TEMP[1].x, TEMP[0].xxxx, IMM[0].xxxx
>  3: POW TEMP[1].y, TEMP[0].yyyy, IMM[0].xxxx
>  4: POW TEMP[1].z, TEMP[0].zzzz, IMM[0].xxxx
>  5: POW TEMP[1].w, TEMP[0].wwww, IMM[0].yyyy
>  6: MOV TEMP[0].xy, IN[0].xyyy
>  7: TEX TEMP[0].xy, TEMP[0], SAMP[1], 2D
>  8: MAD TEMP[2].z, TEMP[0].yyyy, IMM[0].zzzz, IMM[0].wwww
>  9: MAD TEMP[3], TEMP[0].xxxy, IMM[0].yyyz, IMM[1].xxxy
> 10: MAD TEMP[0], TEMP[0].xxxy, IMM[1].yyyz, IMM[0].yyyy
> 11: FSGE TEMP[4].x, TEMP[2].zzzz, IMM[1].xxxx
> 12: UIF TEMP[4].xxxx :0
> 13:   MOV TEMP[4].x, TEMP[3].xxxx
> 14: ELSE :0
> 15:   MOV TEMP[4].x, TEMP[0].xxxx
> 16: ENDIF
> 17: MOV TEMP[4].x, TEMP[4].xxxx
> 18: FSGE TEMP[5].x, TEMP[2].zzzz, IMM[1].xxxx
> 19: UIF TEMP[5].xxxx :0
> 20:   MOV TEMP[5].x, TEMP[3].yyyy
> 21: ELSE :0
> 22:   MOV TEMP[5].x, TEMP[0].yyyy
> 23: ENDIF
> 24: MOV TEMP[4].y, TEMP[5].xxxx
> 25: FSGE TEMP[5].x, TEMP[2].zzzz, IMM[1].xxxx
> 26: UIF TEMP[5].xxxx :0
> 27:   MOV TEMP[5].x, TEMP[3].zzzz
> 28: ELSE :0
> 29:   MOV TEMP[5].x, TEMP[0].zzzz
> 30: ENDIF
> 31: MOV TEMP[4].z, TEMP[5].xxxx
> 32: FSGE TEMP[2].x, TEMP[2].zzzz, IMM[1].xxxx
> 33: UIF TEMP[2].xxxx :0
> 34:   MOV TEMP[2].x, TEMP[3].wwww
> 35: ELSE :0
> 36:   MOV TEMP[2].x, TEMP[0].wwww
> 37: ENDIF
> 38: MOV TEMP[4].w, TEMP[2].xxxx
> 39: MOV OUT[1], TEMP[4]
> 40: MOV OUT[0], TEMP[1]
> 41: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %23 = load <8 x i32> addrspace(2)* %22, !tbaa !0
>  %24 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %25 = load <4 x i32> addrspace(2)* %24, !tbaa !0
>  %26 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 1
>  %27 = load <8 x i32> addrspace(2)* %26, !tbaa !0
>  %28 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 1
>  %29 = load <4 x i32> addrspace(2)* %28, !tbaa !0
>  %30 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %31 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %32 = bitcast float %30 to i32
>  %33 = bitcast float %31 to i32
>  %34 = insertelement <2 x i32> undef, i32 %32, i32 0
>  %35 = insertelement <2 x i32> %34, i32 %33, i32 1
>  %36 = bitcast <8 x i32> %23 to <32 x i8>
>  %37 = bitcast <4 x i32> %25 to <16 x i8>
>  %38 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %35, <32 x i8> %36, <16 x i8> %37, i32 2)
>  %39 = extractelement <4 x float> %38, i32 0
>  %40 = extractelement <4 x float> %38, i32 1
>  %41 = extractelement <4 x float> %38, i32 2
>  %42 = extractelement <4 x float> %38, i32 3
>  %43 = call float @llvm.pow.f32(float %39, float 0x40019999A0000000)
>  %44 = call float @llvm.pow.f32(float %40, float 0x40019999A0000000)
>  %45 = call float @llvm.pow.f32(float %41, float 0x40019999A0000000)
>  %46 = call float @llvm.pow.f32(float %42, float 1.000000e+00)
>  %47 = bitcast float %30 to i32
>  %48 = bitcast float %31 to i32
>  %49 = insertelement <2 x i32> undef, i32 %47, i32 0
>  %50 = insertelement <2 x i32> %49, i32 %48, i32 1
>  %51 = bitcast <8 x i32> %27 to <32 x i8>
>  %52 = bitcast <4 x i32> %29 to <16 x i8>
>  %53 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %50, <32 x i8> %51, <16 x i8> %52, i32 2)
>  %54 = extractelement <4 x float> %53, i32 0
>  %55 = extractelement <4 x float> %53, i32 1
>  %56 = fmul float %55, 2.000000e+00
>  %57 = fadd float %56, 0xBFEFFF2E40000000
>  %58 = fmul float %54, 1.000000e+00
>  %59 = fadd float %58, 0.000000e+00
>  %60 = fmul float %54, 1.000000e+00
>  %61 = fadd float %60, 0.000000e+00
>  %62 = fmul float %54, 1.000000e+00
>  %63 = fadd float %62, 0.000000e+00
>  %64 = fmul float %55, 2.000000e+00
>  %65 = fadd float %64, -1.000000e+00
>  %66 = fmul float %54, -1.000000e+00
>  %67 = fadd float %66, 1.000000e+00
>  %68 = fmul float %54, -1.000000e+00
>  %69 = fadd float %68, 1.000000e+00
>  %70 = fmul float %54, -1.000000e+00
>  %71 = fadd float %70, 1.000000e+00
>  %72 = fmul float %55, -2.000000e+00
>  %73 = fadd float %72, 1.000000e+00
>  %74 = fcmp oge float %57, 0.000000e+00
>  %75 = sext i1 %74 to i32
>  %76 = bitcast i32 %75 to float
>  %77 = bitcast float %76 to i32
>  %78 = icmp ne i32 %77, 0
>  %. = select i1 %78, float %59, float %67
>  %79 = fcmp oge float %57, 0.000000e+00
>  %80 = sext i1 %79 to i32
>  %81 = bitcast i32 %80 to float
>  %82 = bitcast float %81 to i32
>  %83 = icmp ne i32 %82, 0
>  %temp20.0 = select i1 %83, float %61, float %69
>  %84 = fcmp oge float %57, 0.000000e+00
>  %85 = sext i1 %84 to i32
>  %86 = bitcast i32 %85 to float
>  %87 = bitcast float %86 to i32
>  %88 = icmp ne i32 %87, 0
>  %.33 = select i1 %88, float %63, float %71
>  %89 = fcmp oge float %57, 0.000000e+00
>  %90 = sext i1 %89 to i32
>  %91 = bitcast i32 %90 to float
>  %92 = bitcast float %91 to i32
>  %93 = icmp ne i32 %92, 0
>  %temp8.0 = select i1 %93, float %65, float %73
>  %94 = call i32 @llvm.SI.packf16(float %43, float %44)
>  %95 = bitcast i32 %94 to float
>  %96 = call i32 @llvm.SI.packf16(float %45, float %46)
>  %97 = bitcast i32 %96 to float
>  %98 = call i32 @llvm.SI.packf16(float %., float %temp20.0)
>  %99 = bitcast i32 %98 to float
>  %100 = call i32 @llvm.SI.packf16(float %.33, float %temp8.0)
>  %101 = bitcast i32 %100 to float
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 0, i32 1, float %95, float %97, float %95, float %97)
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 1, i32 1, float %99, float %101, float %99, float %101)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.pow.f32(float, float) #1
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                   ; BEFE0A7E
>	s_mov_b32 m0, s9                                                       ; BEFC0309
>	v_interp_p1_f32 v3, v0, 1, 0, [m0]                                     ; C80C0100
>	v_interp_p2_f32 v3, [v3], v1, 1, 0, [m0]                               ; C80D0101
>	v_interp_p1_f32 v2, v0, 0, 0, [m0]                                     ; C8080000
>	v_interp_p2_f32 v2, [v2], v1, 0, 0, [m0]                               ; C8090001
>	s_load_dwordx4 s[0:3], s[4:5], 0x0                                     ; C0800500
>	s_load_dwordx4 s[8:11], s[4:5], 0x4                                    ; C0840504
>	s_load_dwordx8 s[12:19], s[6:7], 0x0                                   ; C0C60700
>	s_load_dwordx8 s[20:27], s[6:7], 0x8                                   ; C0CA0708
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	image_sample v[4:7], 15, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[12:19], s[0:3] ; F0800F00 00030402
>	s_waitcnt vmcnt(0)                                                     ; BF8C0770
>	v_log_f32_e32 v0, v7                                                   ; 7E004F07
>	v_mul_legacy_f32_e32 v0, 1.0, v0                                       ; 0E0000F2
>	v_exp_f32_e32 v0, v0                                                   ; 7E004B00
>	v_log_f32_e32 v1, v6                                                   ; 7E024F06
>	v_mov_b32_e32 v8, 0x400ccccd                                           ; 7E1002FF 400CCCCD
>	v_mul_legacy_f32_e32 v1, v8, v1                                        ; 0E020308
>	v_exp_f32_e32 v1, v1                                                   ; 7E024B01
>	v_cvt_pkrtz_f16_f32_e32 v0, v1, v0                                     ; 5E000101
>	v_log_f32_e32 v1, v5                                                   ; 7E024F05
>	v_mul_legacy_f32_e32 v1, v8, v1                                        ; 0E020308
>	v_exp_f32_e32 v1, v1                                                   ; 7E024B01
>	v_log_f32_e32 v4, v4                                                   ; 7E084F04
>	v_mul_legacy_f32_e32 v4, v8, v4                                        ; 0E080908
>	v_exp_f32_e32 v4, v4                                                   ; 7E084B04
>	v_cvt_pkrtz_f16_f32_e32 v1, v4, v1                                     ; 5E020304
>	exp 15, 0, 1, 0, 0, v1, v0, v1, v0                                     ; F800040F 00010001
>	s_waitcnt expcnt(0)                                                    ; BF8C070F
>	image_sample v[0:1], 3, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[20:27], s[8:11] ; F0800300 00450002
>	s_waitcnt vmcnt(0)                                                     ; BF8C0770
>	v_mad_f32 v2, 2.0, v1, -1.0                                            ; D2820002 03CE02F4
>	v_mad_f32 v3, -2.0, v1, 1.0                                            ; D2820003 03CA02F5
>	v_mov_b32_e32 v4, 0xbf7ff972                                           ; 7E0802FF BF7FF972
>	v_mad_f32 v4, 2.0, v1, v4                                              ; D2820004 041202F4
>	v_cmp_ge_f32_e64 s[0:1], v4, 0                                         ; D00C0000 00010104
>	v_cndmask_b32_e64 v4, 0, -1, s[0:1]                                    ; D2000004 00018280
>	v_cmp_ne_i32_e64 s[0:1], v4, 0                                         ; D10A0000 00010104
>	v_cndmask_b32_e64 v2, v3, v2, s[0:1]                                   ; D2000002 18020503
>	v_add_f32_e32 v3, 0, v0                                                ; 06060080
>	v_sub_f32_e32 v0, 1.0, v0                                              ; 080000F2
>	v_cndmask_b32_e64 v0, v0, v3, s[0:1]                                   ; D2000000 00020700
>	v_cvt_pkrtz_f16_f32_e32 v1, v0, v2                                     ; 5E020500
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v0                                     ; 5E000100
>	exp 15, 1, 1, 1, 1, v0, v1, v0, v1                                     ; F8001C1F 01000100
>	s_endpgm                                                               ; BF810000
>
>SGPRS: 32
>VGPRS: 12
>Code Size: 244 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL CONST[0..3]
>DCL TEMP[0..1], LOCAL
>IMM[0] FLT32 {    0.0000,     1.0000,     0.0000,     0.0000}
>  0: MOV TEMP[0].xy, CONST[2].xyxx
>  1: ADD TEMP[1].xy, -TEMP[0], CONST[3]
>  2: MOV TEMP[0].xy, TEMP[1].xyxx
>  3: MAD TEMP[0].xy, IN[0], TEMP[0], CONST[2]
>  4: MAD TEMP[1].zw, TEMP[0].xyxy, CONST[1].xyxy, -CONST[0].xyxy
>  5: MOV TEMP[0].xy, TEMP[0].xyxx
>  6: MOV TEMP[1].xy, TEMP[1].zwzz
>  7: MOV TEMP[1].zw, IMM[0].yyxy
>  8: MOV TEMP[0].zw, IMM[0].yyxy
>  9: MOV OUT[1], TEMP[0]
> 10: MOV OUT[0], TEMP[1]
> 11: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 0)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 4)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %12, i32 16)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %12, i32 20)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %12, i32 32)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %12, i32 36)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %12, i32 48)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %12, i32 52)
>  %21 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %22 = load <16 x i8> addrspace(2)* %21, !tbaa !0
>  %23 = add i32 %5, %7
>  %24 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %22, i32 0, i32 %23)
>  %25 = extractelement <4 x float> %24, i32 0
>  %26 = extractelement <4 x float> %24, i32 1
>  %27 = fsub float -0.000000e+00, %17
>  %28 = fadd float %27, %19
>  %29 = fsub float -0.000000e+00, %18
>  %30 = fadd float %29, %20
>  %31 = fmul float %25, %28
>  %32 = fadd float %31, %17
>  %33 = fmul float %26, %30
>  %34 = fadd float %33, %18
>  %35 = fsub float -0.000000e+00, %13
>  %36 = fmul float %32, %15
>  %37 = fadd float %36, %35
>  %38 = fsub float -0.000000e+00, %14
>  %39 = fmul float %34, %16
>  %40 = fadd float %39, %38
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %32, float %34, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %37, float %40, float 0.000000e+00, float 1.000000e+00)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                   ; C0800300
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	s_buffer_load_dword s4, s[0:3], 0x9                  ; C2020109
>	s_buffer_load_dword s5, s[0:3], 0xd                  ; C202810D
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mov_b32_e32 v1, s4                                 ; 7E020204
>	v_sub_f32_e32 v1, s5, v1                             ; 08020205
>	v_add_i32_e32 v0, s10, v0                            ; 4A00000A
>	s_load_dwordx4 s[8:11], s[8:9], 0x0                  ; C0840900
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	buffer_load_format_xyzw v[2:5], v0, s[8:11], 0 idxen ; E00C2000 80020200
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_mad_f32 v0, v3, v1, s4                             ; D2820000 00120303
>	s_buffer_load_dword s4, s[0:3], 0x8                  ; C2020108
>	s_buffer_load_dword s5, s[0:3], 0xc                  ; C202810C
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mov_b32_e32 v1, s4                                 ; 7E020204
>	v_sub_f32_e32 v1, s5, v1                             ; 08020205
>	v_mad_f32 v1, v2, v1, s4                             ; D2820001 00120302
>	v_mov_b32_e32 v2, 1.0                                ; 7E0402F2
>	v_mov_b32_e32 v3, 0                                  ; 7E060280
>	exp 15, 32, 0, 0, 0, v1, v0, v3, v2                  ; F800020F 02030001
>	s_buffer_load_dword s4, s[0:3], 0x1                  ; C2020101
>	s_buffer_load_dword s5, s[0:3], 0x5                  ; C2028105
>	s_waitcnt expcnt(0) lgkmcnt(0)                       ; BF8C000F
>	v_mov_b32_e32 v4, s4                                 ; 7E080204
>	v_mad_f32 v0, s5, v0, -v4                            ; D2820000 84120005
>	s_buffer_load_dword s4, s[0:3], 0x0                  ; C2020100
>	s_buffer_load_dword s0, s[0:3], 0x4                  ; C2000104
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mov_b32_e32 v4, s4                                 ; 7E080204
>	v_mad_f32 v1, s0, v1, -v4                            ; D2820001 84120200
>	exp 15, 12, 0, 1, 0, v1, v0, v3, v2                  ; F80008CF 02030001
>	s_endpgm                                             ; BF810000
>
>SGPRS: 16
>VGPRS: 8
>Code Size: 164 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL CONST[0]
>DCL TEMP[0..2], LOCAL
>IMM[0] FLT32 {    2.2000,     1.0000,     0.4545,     0.0000}
>  0: MAD TEMP[0].xy, IN[0], CONST[0].xxxx, CONST[0].yyyy
>  1: MOV TEMP[0].xy, TEMP[0].xyyy
>  2: TEX TEMP[0].xyz, TEMP[0], SAMP[0], 2D
>  3: POW TEMP[1].x, TEMP[0].xxxx, IMM[0].xxxx
>  4: POW TEMP[1].y, TEMP[0].yyyy, IMM[0].xxxx
>  5: POW TEMP[1].z, TEMP[0].zzzz, IMM[0].xxxx
>  6: ABS TEMP[0].x, TEMP[1].xxxx
>  7: LG2 TEMP[0].x, TEMP[0].xxxx
>  8: ABS TEMP[2].x, TEMP[1].yyyy
>  9: LG2 TEMP[2].x, TEMP[2].xxxx
> 10: MOV TEMP[0].y, TEMP[2].xxxx
> 11: ABS TEMP[1].x, TEMP[1].zzzz
> 12: LG2 TEMP[1].x, TEMP[1].xxxx
> 13: MOV TEMP[0].z, TEMP[1].xxxx
> 14: MUL TEMP[0].xyz, TEMP[0], IMM[0].zzzz
> 15: EX2 TEMP[1].x, TEMP[0].xxxx
> 16: EX2 TEMP[2].x, TEMP[0].yyyy
> 17: MOV TEMP[1].y, TEMP[2].xxxx
> 18: EX2 TEMP[0].x, TEMP[0].zzzz
> 19: MOV TEMP[1].z, TEMP[0].xxxx
> 20: MOV TEMP[1].w, IMM[0].yyyy
> 21: MOV OUT[0], TEMP[1]
> 22: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %23 = load <16 x i8> addrspace(2)* %22, !tbaa !0
>  %24 = call float @llvm.SI.load.const(<16 x i8> %23, i32 0)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %23, i32 4)
>  %26 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %27 = load <8 x i32> addrspace(2)* %26, !tbaa !0
>  %28 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %29 = load <4 x i32> addrspace(2)* %28, !tbaa !0
>  %30 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %31 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %32 = fmul float %30, %24
>  %33 = fadd float %32, %25
>  %34 = fmul float %31, %24
>  %35 = fadd float %34, %25
>  %36 = bitcast float %33 to i32
>  %37 = bitcast float %35 to i32
>  %38 = insertelement <2 x i32> undef, i32 %36, i32 0
>  %39 = insertelement <2 x i32> %38, i32 %37, i32 1
>  %40 = bitcast <8 x i32> %27 to <32 x i8>
>  %41 = bitcast <4 x i32> %29 to <16 x i8>
>  %42 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %39, <32 x i8> %40, <16 x i8> %41, i32 2)
>  %43 = extractelement <4 x float> %42, i32 0
>  %44 = extractelement <4 x float> %42, i32 1
>  %45 = extractelement <4 x float> %42, i32 2
>  %46 = call float @llvm.pow.f32(float %43, float 0x40019999A0000000)
>  %47 = call float @llvm.pow.f32(float %44, float 0x40019999A0000000)
>  %48 = call float @llvm.pow.f32(float %45, float 0x40019999A0000000)
>  %49 = call float @fabs(float %46)
>  %50 = call float @llvm.log2.f32(float %49)
>  %51 = call float @fabs(float %47)
>  %52 = call float @llvm.log2.f32(float %51)
>  %53 = call float @fabs(float %48)
>  %54 = call float @llvm.log2.f32(float %53)
>  %55 = fmul float %50, 0x3FDD1743E0000000
>  %56 = fmul float %52, 0x3FDD1743E0000000
>  %57 = fmul float %54, 0x3FDD1743E0000000
>  %58 = call float @llvm.AMDIL.exp.(float %55)
>  %59 = call float @llvm.AMDIL.exp.(float %56)
>  %60 = call float @llvm.AMDIL.exp.(float %57)
>  %61 = call i32 @llvm.SI.packf16(float %58, float %59)
>  %62 = bitcast i32 %61 to float
>  %63 = call i32 @llvm.SI.packf16(float %60, float 1.000000e+00)
>  %64 = bitcast i32 %63 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %62, float %64, float %62, float %64)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.pow.f32(float, float) #1
>
>; Function Attrs: readnone
>declare float @fabs(float) #2
>
>; Function Attrs: nounwind readnone
>declare float @llvm.log2.f32(float) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.exp.(float) #2
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                 ; BEFE0A7E
>	s_mov_b32 m0, s9                                                     ; BEFC0309
>	v_interp_p1_f32 v2, v0, 1, 0, [m0]                                   ; C8080100
>	v_interp_p2_f32 v2, [v2], v1, 1, 0, [m0]                             ; C8090101
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                                   ; C0800300
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	s_buffer_load_dword s8, s[0:3], 0x1                                  ; C2040101
>	s_buffer_load_dword s0, s[0:3], 0x0                                  ; C2000100
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mov_b32_e32 v3, s8                                                 ; 7E060208
>	v_mad_f32 v3, s0, v2, v3                                             ; D2820003 040E0400
>	v_interp_p1_f32 v4, v0, 0, 0, [m0]                                   ; C8100000
>	v_interp_p2_f32 v4, [v4], v1, 0, 0, [m0]                             ; C8110001
>	v_mov_b32_e32 v0, s8                                                 ; 7E000208
>	v_mad_f32 v2, s0, v4, v0                                             ; D2820002 04020800
>	s_load_dwordx4 s[0:3], s[4:5], 0x0                                   ; C0800500
>	s_load_dwordx8 s[4:11], s[6:7], 0x0                                  ; C0C20700
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	image_sample v[0:2], 7, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[4:11], s[0:3] ; F0800700 00010002
>	s_waitcnt vmcnt(0)                                                   ; BF8C0770
>	v_log_f32_e32 v3, v1                                                 ; 7E064F01
>	v_mov_b32_e32 v4, 0x400ccccd                                         ; 7E0802FF 400CCCCD
>	v_mul_legacy_f32_e32 v3, v4, v3                                      ; 0E060704
>	v_exp_f32_e32 v3, v3                                                 ; 7E064B03
>	v_log_f32_e64 v3, |v3|                                               ; D34E0103 00000103
>	v_mul_f32_e32 v3, 0x3ee8ba1f, v3                                     ; 100606FF 3EE8BA1F
>	v_exp_f32_e32 v3, v3                                                 ; 7E064B03
>	v_log_f32_e32 v5, v0                                                 ; 7E0A4F00
>	v_mul_legacy_f32_e32 v5, v4, v5                                      ; 0E0A0B04
>	v_exp_f32_e32 v5, v5                                                 ; 7E0A4B05
>	v_log_f32_e64 v5, |v5|                                               ; D34E0105 00000105
>	v_mul_f32_e32 v5, 0x3ee8ba1f, v5                                     ; 100A0AFF 3EE8BA1F
>	v_exp_f32_e32 v5, v5                                                 ; 7E0A4B05
>	v_cvt_pkrtz_f16_f32_e32 v3, v5, v3                                   ; 5E060705
>	v_log_f32_e32 v0, v2                                                 ; 7E004F02
>	v_mul_legacy_f32_e32 v0, v4, v0                                      ; 0E000104
>	v_exp_f32_e32 v0, v0                                                 ; 7E004B00
>	v_log_f32_e64 v0, |v0|                                               ; D34E0100 00000100
>	v_mul_f32_e32 v0, 0x3ee8ba1f, v0                                     ; 100000FF 3EE8BA1F
>	v_exp_f32_e32 v0, v0                                                 ; 7E004B00
>	v_cvt_pkrtz_f16_f32_e64 v0, v0, 1.0                                  ; D25E0000 0001E500
>	exp 15, 0, 1, 1, 1, v3, v0, v3, v0                                   ; F8001C0F 00030003
>	s_endpgm                                                             ; BF810000
>
>SGPRS: 16
>VGPRS: 8
>Code Size: 220 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL CONST[0..1]
>DCL TEMP[0..1], LOCAL
>IMM[0] FLT32 {    0.0000,     1.0000,     0.0000,     0.0000}
>  0: MOV TEMP[0].xy, CONST[1].xyxx
>  1: MAD TEMP[0].xy, IN[0], TEMP[0], -CONST[0]
>  2: MOV TEMP[0].xy, TEMP[0].xyxx
>  3: MOV TEMP[1].xy, IN[0].xyxx
>  4: MOV TEMP[0].zw, IMM[0].yyxy
>  5: MOV TEMP[1].zw, IMM[0].yyxy
>  6: MOV OUT[1], TEMP[1]
>  7: MOV OUT[0], TEMP[0]
>  8: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 0)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 4)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %12, i32 16)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %12, i32 20)
>  %17 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %18 = load <16 x i8> addrspace(2)* %17, !tbaa !0
>  %19 = add i32 %5, %7
>  %20 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %18, i32 0, i32 %19)
>  %21 = extractelement <4 x float> %20, i32 0
>  %22 = extractelement <4 x float> %20, i32 1
>  %23 = fsub float -0.000000e+00, %13
>  %24 = fmul float %21, %15
>  %25 = fadd float %24, %23
>  %26 = fsub float -0.000000e+00, %14
>  %27 = fmul float %22, %16
>  %28 = fadd float %27, %26
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %21, float %22, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %25, float %28, float 0.000000e+00, float 1.000000e+00)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                           ; 4A00000A
>	s_load_dwordx4 s[4:7], s[8:9], 0x0                  ; C0820900
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	buffer_load_format_xyzw v[0:3], v0, s[4:7], 0 idxen ; E00C2000 80010000
>	v_mov_b32_e32 v4, 1.0                               ; 7E0802F2
>	v_mov_b32_e32 v5, 0                                 ; 7E0A0280
>	s_waitcnt vmcnt(0)                                  ; BF8C0770
>	exp 15, 32, 0, 0, 0, v0, v1, v5, v4                 ; F800020F 04050100
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                  ; C0800300
>	s_waitcnt expcnt(0) lgkmcnt(0)                      ; BF8C000F
>	s_buffer_load_dword s4, s[0:3], 0x1                 ; C2020101
>	s_buffer_load_dword s5, s[0:3], 0x5                 ; C2028105
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	v_mov_b32_e32 v6, s4                                ; 7E0C0204
>	v_mad_f32 v6, s5, v1, -v6                           ; D2820006 841A0205
>	s_buffer_load_dword s4, s[0:3], 0x0                 ; C2020100
>	s_buffer_load_dword s0, s[0:3], 0x4                 ; C2000104
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	v_mov_b32_e32 v7, s4                                ; 7E0E0204
>	v_mad_f32 v0, s0, v0, -v7                           ; D2820000 841E0000
>	exp 15, 12, 0, 1, 0, v0, v6, v5, v4                 ; F80008CF 04050600
>	s_endpgm                                            ; BF810000
>
>SGPRS: 16
>VGPRS: 8
>Code Size: 108 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL CONST[0]
>DCL TEMP[0..2], LOCAL
>IMM[0] FLT32 {    2.2000,     1.0000,     0.4545,     0.0000}
>  0: MAD TEMP[0].xy, IN[0], CONST[0].xxxx, CONST[0].yyyy
>  1: MOV TEMP[0].xy, TEMP[0].xyyy
>  2: TEX TEMP[0], TEMP[0], SAMP[0], 2D
>  3: POW TEMP[1].x, TEMP[0].xxxx, IMM[0].xxxx
>  4: POW TEMP[1].y, TEMP[0].yyyy, IMM[0].xxxx
>  5: POW TEMP[1].z, TEMP[0].zzzz, IMM[0].xxxx
>  6: POW TEMP[1].w, TEMP[0].wwww, IMM[0].yyyy
>  7: ABS TEMP[0].x, TEMP[1].xxxx
>  8: LG2 TEMP[0].x, TEMP[0].xxxx
>  9: ABS TEMP[2].x, TEMP[1].yyyy
> 10: LG2 TEMP[2].x, TEMP[2].xxxx
> 11: MOV TEMP[0].y, TEMP[2].xxxx
> 12: ABS TEMP[2].x, TEMP[1].zzzz
> 13: LG2 TEMP[2].x, TEMP[2].xxxx
> 14: MOV TEMP[0].z, TEMP[2].xxxx
> 15: MOV TEMP[1].w, TEMP[1].wwww
> 16: MUL TEMP[0].xyz, TEMP[0], IMM[0].zzzz
> 17: EX2 TEMP[1].x, TEMP[0].xxxx
> 18: EX2 TEMP[2].x, TEMP[0].yyyy
> 19: MOV TEMP[1].y, TEMP[2].xxxx
> 20: EX2 TEMP[0].x, TEMP[0].zzzz
> 21: MOV TEMP[1].z, TEMP[0].xxxx
> 22: MOV OUT[0], TEMP[1]
> 23: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %23 = load <16 x i8> addrspace(2)* %22, !tbaa !0
>  %24 = call float @llvm.SI.load.const(<16 x i8> %23, i32 0)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %23, i32 4)
>  %26 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %27 = load <8 x i32> addrspace(2)* %26, !tbaa !0
>  %28 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %29 = load <4 x i32> addrspace(2)* %28, !tbaa !0
>  %30 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %31 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %32 = fmul float %30, %24
>  %33 = fadd float %32, %25
>  %34 = fmul float %31, %24
>  %35 = fadd float %34, %25
>  %36 = bitcast float %33 to i32
>  %37 = bitcast float %35 to i32
>  %38 = insertelement <2 x i32> undef, i32 %36, i32 0
>  %39 = insertelement <2 x i32> %38, i32 %37, i32 1
>  %40 = bitcast <8 x i32> %27 to <32 x i8>
>  %41 = bitcast <4 x i32> %29 to <16 x i8>
>  %42 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %39, <32 x i8> %40, <16 x i8> %41, i32 2)
>  %43 = extractelement <4 x float> %42, i32 0
>  %44 = extractelement <4 x float> %42, i32 1
>  %45 = extractelement <4 x float> %42, i32 2
>  %46 = extractelement <4 x float> %42, i32 3
>  %47 = call float @llvm.pow.f32(float %43, float 0x40019999A0000000)
>  %48 = call float @llvm.pow.f32(float %44, float 0x40019999A0000000)
>  %49 = call float @llvm.pow.f32(float %45, float 0x40019999A0000000)
>  %50 = call float @llvm.pow.f32(float %46, float 1.000000e+00)
>  %51 = call float @fabs(float %47)
>  %52 = call float @llvm.log2.f32(float %51)
>  %53 = call float @fabs(float %48)
>  %54 = call float @llvm.log2.f32(float %53)
>  %55 = call float @fabs(float %49)
>  %56 = call float @llvm.log2.f32(float %55)
>  %57 = fmul float %52, 0x3FDD1743E0000000
>  %58 = fmul float %54, 0x3FDD1743E0000000
>  %59 = fmul float %56, 0x3FDD1743E0000000
>  %60 = call float @llvm.AMDIL.exp.(float %57)
>  %61 = call float @llvm.AMDIL.exp.(float %58)
>  %62 = call float @llvm.AMDIL.exp.(float %59)
>  %63 = call i32 @llvm.SI.packf16(float %60, float %61)
>  %64 = bitcast i32 %63 to float
>  %65 = call i32 @llvm.SI.packf16(float %62, float %50)
>  %66 = bitcast i32 %65 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %64, float %66, float %64, float %66)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.pow.f32(float, float) #1
>
>; Function Attrs: readnone
>declare float @fabs(float) #2
>
>; Function Attrs: nounwind readnone
>declare float @llvm.log2.f32(float) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.exp.(float) #2
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                  ; BEFE0A7E
>	s_mov_b32 m0, s9                                                      ; BEFC0309
>	v_interp_p1_f32 v2, v0, 1, 0, [m0]                                    ; C8080100
>	v_interp_p2_f32 v2, [v2], v1, 1, 0, [m0]                              ; C8090101
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                                    ; C0800300
>	s_waitcnt lgkmcnt(0)                                                  ; BF8C007F
>	s_buffer_load_dword s8, s[0:3], 0x1                                   ; C2040101
>	s_buffer_load_dword s0, s[0:3], 0x0                                   ; C2000100
>	s_waitcnt lgkmcnt(0)                                                  ; BF8C007F
>	v_mov_b32_e32 v3, s8                                                  ; 7E060208
>	v_mad_f32 v3, s0, v2, v3                                              ; D2820003 040E0400
>	v_interp_p1_f32 v4, v0, 0, 0, [m0]                                    ; C8100000
>	v_interp_p2_f32 v4, [v4], v1, 0, 0, [m0]                              ; C8110001
>	v_mov_b32_e32 v0, s8                                                  ; 7E000208
>	v_mad_f32 v2, s0, v4, v0                                              ; D2820002 04020800
>	s_load_dwordx4 s[0:3], s[4:5], 0x0                                    ; C0800500
>	s_load_dwordx8 s[4:11], s[6:7], 0x0                                   ; C0C20700
>	s_waitcnt lgkmcnt(0)                                                  ; BF8C007F
>	image_sample v[0:3], 15, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[4:11], s[0:3] ; F0800F00 00010002
>	s_waitcnt vmcnt(0)                                                    ; BF8C0770
>	v_log_f32_e32 v4, v3                                                  ; 7E084F03
>	v_mul_legacy_f32_e32 v4, 1.0, v4                                      ; 0E0808F2
>	v_exp_f32_e32 v4, v4                                                  ; 7E084B04
>	v_log_f32_e32 v5, v2                                                  ; 7E0A4F02
>	v_mov_b32_e32 v6, 0x400ccccd                                          ; 7E0C02FF 400CCCCD
>	v_mul_legacy_f32_e32 v5, v6, v5                                       ; 0E0A0B06
>	v_exp_f32_e32 v5, v5                                                  ; 7E0A4B05
>	v_log_f32_e64 v5, |v5|                                                ; D34E0105 00000105
>	v_mul_f32_e32 v5, 0x3ee8ba1f, v5                                      ; 100A0AFF 3EE8BA1F
>	v_exp_f32_e32 v5, v5                                                  ; 7E0A4B05
>	v_cvt_pkrtz_f16_f32_e32 v4, v5, v4                                    ; 5E080905
>	v_log_f32_e32 v5, v1                                                  ; 7E0A4F01
>	v_mul_legacy_f32_e32 v5, v6, v5                                       ; 0E0A0B06
>	v_exp_f32_e32 v5, v5                                                  ; 7E0A4B05
>	v_log_f32_e64 v5, |v5|                                                ; D34E0105 00000105
>	v_mul_f32_e32 v5, 0x3ee8ba1f, v5                                      ; 100A0AFF 3EE8BA1F
>	v_exp_f32_e32 v5, v5                                                  ; 7E0A4B05
>	v_log_f32_e32 v0, v0                                                  ; 7E004F00
>	v_mul_legacy_f32_e32 v0, v6, v0                                       ; 0E000106
>	v_exp_f32_e32 v0, v0                                                  ; 7E004B00
>	v_log_f32_e64 v0, |v0|                                                ; D34E0100 00000100
>	v_mul_f32_e32 v0, 0x3ee8ba1f, v0                                      ; 100000FF 3EE8BA1F
>	v_exp_f32_e32 v0, v0                                                  ; 7E004B00
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v5                                    ; 5E000B00
>	exp 15, 0, 1, 1, 1, v0, v4, v0, v4                                    ; F8001C0F 04000400
>	s_endpgm                                                              ; BF810000
>
>SGPRS: 16
>VGPRS: 8
>Code Size: 228 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL CONST[0..1]
>DCL TEMP[0..1], LOCAL
>IMM[0] FLT32 {    0.0000,     1.0000,     0.0000,     0.0000}
>  0: MOV TEMP[0].xy, CONST[1].xyxx
>  1: MAD TEMP[0].xy, IN[0], TEMP[0], -CONST[0]
>  2: MOV TEMP[0].xy, TEMP[0].xyxx
>  3: MOV TEMP[1].xy, IN[0].xyxx
>  4: MOV TEMP[0].zw, IMM[0].yyxy
>  5: MOV TEMP[1].zw, IMM[0].yyxy
>  6: MOV OUT[1], TEMP[1]
>  7: MOV OUT[0], TEMP[0]
>  8: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 0)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 4)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %12, i32 16)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %12, i32 20)
>  %17 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %18 = load <16 x i8> addrspace(2)* %17, !tbaa !0
>  %19 = add i32 %5, %7
>  %20 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %18, i32 0, i32 %19)
>  %21 = extractelement <4 x float> %20, i32 0
>  %22 = extractelement <4 x float> %20, i32 1
>  %23 = fsub float -0.000000e+00, %13
>  %24 = fmul float %21, %15
>  %25 = fadd float %24, %23
>  %26 = fsub float -0.000000e+00, %14
>  %27 = fmul float %22, %16
>  %28 = fadd float %27, %26
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %21, float %22, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %25, float %28, float 0.000000e+00, float 1.000000e+00)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                           ; 4A00000A
>	s_load_dwordx4 s[4:7], s[8:9], 0x0                  ; C0820900
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	buffer_load_format_xyzw v[0:3], v0, s[4:7], 0 idxen ; E00C2000 80010000
>	v_mov_b32_e32 v4, 1.0                               ; 7E0802F2
>	v_mov_b32_e32 v5, 0                                 ; 7E0A0280
>	s_waitcnt vmcnt(0)                                  ; BF8C0770
>	exp 15, 32, 0, 0, 0, v0, v1, v5, v4                 ; F800020F 04050100
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                  ; C0800300
>	s_waitcnt expcnt(0) lgkmcnt(0)                      ; BF8C000F
>	s_buffer_load_dword s4, s[0:3], 0x1                 ; C2020101
>	s_buffer_load_dword s5, s[0:3], 0x5                 ; C2028105
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	v_mov_b32_e32 v6, s4                                ; 7E0C0204
>	v_mad_f32 v6, s5, v1, -v6                           ; D2820006 841A0205
>	s_buffer_load_dword s4, s[0:3], 0x0                 ; C2020100
>	s_buffer_load_dword s0, s[0:3], 0x4                 ; C2000104
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	v_mov_b32_e32 v7, s4                                ; 7E0E0204
>	v_mad_f32 v0, s0, v0, -v7                           ; D2820000 841E0000
>	exp 15, 12, 0, 1, 0, v0, v6, v5, v4                 ; F80008CF 04050600
>	s_endpgm                                            ; BF810000
>
>SGPRS: 16
>VGPRS: 8
>Code Size: 108 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL CONST[0]
>DCL TEMP[0..2], LOCAL
>IMM[0] FLT32 {    2.2000,     1.0000,     0.4545,     0.0000}
>  0: MAD TEMP[0].xy, IN[0], CONST[0].xxxx, CONST[0].yyyy
>  1: MOV TEMP[0].xy, TEMP[0].xyyy
>  2: TEX TEMP[0], TEMP[0], SAMP[0], 2D
>  3: POW TEMP[1].x, TEMP[0].xxxx, IMM[0].xxxx
>  4: POW TEMP[1].y, TEMP[0].yyyy, IMM[0].xxxx
>  5: POW TEMP[1].z, TEMP[0].zzzz, IMM[0].xxxx
>  6: POW TEMP[1].w, TEMP[0].wwww, IMM[0].yyyy
>  7: ABS TEMP[0].x, TEMP[1].xxxx
>  8: LG2 TEMP[0].x, TEMP[0].xxxx
>  9: ABS TEMP[2].x, TEMP[1].yyyy
> 10: LG2 TEMP[2].x, TEMP[2].xxxx
> 11: MOV TEMP[0].y, TEMP[2].xxxx
> 12: ABS TEMP[2].x, TEMP[1].zzzz
> 13: LG2 TEMP[2].x, TEMP[2].xxxx
> 14: MOV TEMP[0].z, TEMP[2].xxxx
> 15: MOV TEMP[1].w, TEMP[1].wwww
> 16: MUL TEMP[0].xyz, TEMP[0], IMM[0].zzzz
> 17: EX2 TEMP[1].x, TEMP[0].xxxx
> 18: EX2 TEMP[2].x, TEMP[0].yyyy
> 19: MOV TEMP[1].y, TEMP[2].xxxx
> 20: EX2 TEMP[0].x, TEMP[0].zzzz
> 21: MOV TEMP[1].z, TEMP[0].xxxx
> 22: MOV OUT[0], TEMP[1]
> 23: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %23 = load <16 x i8> addrspace(2)* %22, !tbaa !0
>  %24 = call float @llvm.SI.load.const(<16 x i8> %23, i32 0)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %23, i32 4)
>  %26 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %27 = load <8 x i32> addrspace(2)* %26, !tbaa !0
>  %28 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %29 = load <4 x i32> addrspace(2)* %28, !tbaa !0
>  %30 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %31 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %32 = fmul float %30, %24
>  %33 = fadd float %32, %25
>  %34 = fmul float %31, %24
>  %35 = fadd float %34, %25
>  %36 = bitcast float %33 to i32
>  %37 = bitcast float %35 to i32
>  %38 = insertelement <2 x i32> undef, i32 %36, i32 0
>  %39 = insertelement <2 x i32> %38, i32 %37, i32 1
>  %40 = bitcast <8 x i32> %27 to <32 x i8>
>  %41 = bitcast <4 x i32> %29 to <16 x i8>
>  %42 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %39, <32 x i8> %40, <16 x i8> %41, i32 2)
>  %43 = extractelement <4 x float> %42, i32 0
>  %44 = extractelement <4 x float> %42, i32 1
>  %45 = extractelement <4 x float> %42, i32 2
>  %46 = extractelement <4 x float> %42, i32 3
>  %47 = call float @llvm.pow.f32(float %43, float 0x40019999A0000000)
>  %48 = call float @llvm.pow.f32(float %44, float 0x40019999A0000000)
>  %49 = call float @llvm.pow.f32(float %45, float 0x40019999A0000000)
>  %50 = call float @llvm.pow.f32(float %46, float 1.000000e+00)
>  %51 = call float @fabs(float %47)
>  %52 = call float @llvm.log2.f32(float %51)
>  %53 = call float @fabs(float %48)
>  %54 = call float @llvm.log2.f32(float %53)
>  %55 = call float @fabs(float %49)
>  %56 = call float @llvm.log2.f32(float %55)
>  %57 = fmul float %52, 0x3FDD1743E0000000
>  %58 = fmul float %54, 0x3FDD1743E0000000
>  %59 = fmul float %56, 0x3FDD1743E0000000
>  %60 = call float @llvm.AMDIL.exp.(float %57)
>  %61 = call float @llvm.AMDIL.exp.(float %58)
>  %62 = call float @llvm.AMDIL.exp.(float %59)
>  %63 = call i32 @llvm.SI.packf16(float %60, float %61)
>  %64 = bitcast i32 %63 to float
>  %65 = call i32 @llvm.SI.packf16(float %62, float %50)
>  %66 = bitcast i32 %65 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %64, float %66, float %64, float %66)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.pow.f32(float, float) #1
>
>; Function Attrs: readnone
>declare float @fabs(float) #2
>
>; Function Attrs: nounwind readnone
>declare float @llvm.log2.f32(float) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.exp.(float) #2
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                  ; BEFE0A7E
>	s_mov_b32 m0, s9                                                      ; BEFC0309
>	v_interp_p1_f32 v2, v0, 1, 0, [m0]                                    ; C8080100
>	v_interp_p2_f32 v2, [v2], v1, 1, 0, [m0]                              ; C8090101
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                                    ; C0800300
>	s_waitcnt lgkmcnt(0)                                                  ; BF8C007F
>	s_buffer_load_dword s8, s[0:3], 0x1                                   ; C2040101
>	s_buffer_load_dword s0, s[0:3], 0x0                                   ; C2000100
>	s_waitcnt lgkmcnt(0)                                                  ; BF8C007F
>	v_mov_b32_e32 v3, s8                                                  ; 7E060208
>	v_mad_f32 v3, s0, v2, v3                                              ; D2820003 040E0400
>	v_interp_p1_f32 v4, v0, 0, 0, [m0]                                    ; C8100000
>	v_interp_p2_f32 v4, [v4], v1, 0, 0, [m0]                              ; C8110001
>	v_mov_b32_e32 v0, s8                                                  ; 7E000208
>	v_mad_f32 v2, s0, v4, v0                                              ; D2820002 04020800
>	s_load_dwordx4 s[0:3], s[4:5], 0x0                                    ; C0800500
>	s_load_dwordx8 s[4:11], s[6:7], 0x0                                   ; C0C20700
>	s_waitcnt lgkmcnt(0)                                                  ; BF8C007F
>	image_sample v[0:3], 15, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[4:11], s[0:3] ; F0800F00 00010002
>	s_waitcnt vmcnt(0)                                                    ; BF8C0770
>	v_log_f32_e32 v4, v3                                                  ; 7E084F03
>	v_mul_legacy_f32_e32 v4, 1.0, v4                                      ; 0E0808F2
>	v_exp_f32_e32 v4, v4                                                  ; 7E084B04
>	v_log_f32_e32 v5, v2                                                  ; 7E0A4F02
>	v_mov_b32_e32 v6, 0x400ccccd                                          ; 7E0C02FF 400CCCCD
>	v_mul_legacy_f32_e32 v5, v6, v5                                       ; 0E0A0B06
>	v_exp_f32_e32 v5, v5                                                  ; 7E0A4B05
>	v_log_f32_e64 v5, |v5|                                                ; D34E0105 00000105
>	v_mul_f32_e32 v5, 0x3ee8ba1f, v5                                      ; 100A0AFF 3EE8BA1F
>	v_exp_f32_e32 v5, v5                                                  ; 7E0A4B05
>	v_cvt_pkrtz_f16_f32_e32 v4, v5, v4                                    ; 5E080905
>	v_log_f32_e32 v5, v1                                                  ; 7E0A4F01
>	v_mul_legacy_f32_e32 v5, v6, v5                                       ; 0E0A0B06
>	v_exp_f32_e32 v5, v5                                                  ; 7E0A4B05
>	v_log_f32_e64 v5, |v5|                                                ; D34E0105 00000105
>	v_mul_f32_e32 v5, 0x3ee8ba1f, v5                                      ; 100A0AFF 3EE8BA1F
>	v_exp_f32_e32 v5, v5                                                  ; 7E0A4B05
>	v_log_f32_e32 v0, v0                                                  ; 7E004F00
>	v_mul_legacy_f32_e32 v0, v6, v0                                       ; 0E000106
>	v_exp_f32_e32 v0, v0                                                  ; 7E004B00
>	v_log_f32_e64 v0, |v0|                                                ; D34E0100 00000100
>	v_mul_f32_e32 v0, 0x3ee8ba1f, v0                                      ; 100000FF 3EE8BA1F
>	v_exp_f32_e32 v0, v0                                                  ; 7E004B00
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v5                                    ; 5E000B00
>	exp 15, 0, 1, 1, 1, v0, v4, v0, v4                                    ; F8001C0F 04000400
>	s_endpgm                                                              ; BF810000
>
>SGPRS: 16
>VGPRS: 8
>Code Size: 228 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL CONST[0..1]
>DCL TEMP[0..1], LOCAL
>IMM[0] FLT32 {    0.0000,     1.0000,     0.0000,     0.0000}
>  0: MOV TEMP[0].xy, CONST[1].xyxx
>  1: MAD TEMP[0].xy, IN[0], TEMP[0], -CONST[0]
>  2: MOV TEMP[0].xy, TEMP[0].xyxx
>  3: MOV TEMP[1].xy, IN[0].xyxx
>  4: MOV TEMP[0].zw, IMM[0].yyxy
>  5: MOV TEMP[1].zw, IMM[0].yyxy
>  6: MOV OUT[1], TEMP[1]
>  7: MOV OUT[0], TEMP[0]
>  8: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 0)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 4)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %12, i32 16)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %12, i32 20)
>  %17 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %18 = load <16 x i8> addrspace(2)* %17, !tbaa !0
>  %19 = add i32 %5, %7
>  %20 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %18, i32 0, i32 %19)
>  %21 = extractelement <4 x float> %20, i32 0
>  %22 = extractelement <4 x float> %20, i32 1
>  %23 = fsub float -0.000000e+00, %13
>  %24 = fmul float %21, %15
>  %25 = fadd float %24, %23
>  %26 = fsub float -0.000000e+00, %14
>  %27 = fmul float %22, %16
>  %28 = fadd float %27, %26
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %21, float %22, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %25, float %28, float 0.000000e+00, float 1.000000e+00)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                           ; 4A00000A
>	s_load_dwordx4 s[4:7], s[8:9], 0x0                  ; C0820900
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	buffer_load_format_xyzw v[0:3], v0, s[4:7], 0 idxen ; E00C2000 80010000
>	v_mov_b32_e32 v4, 1.0                               ; 7E0802F2
>	v_mov_b32_e32 v5, 0                                 ; 7E0A0280
>	s_waitcnt vmcnt(0)                                  ; BF8C0770
>	exp 15, 32, 0, 0, 0, v0, v1, v5, v4                 ; F800020F 04050100
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                  ; C0800300
>	s_waitcnt expcnt(0) lgkmcnt(0)                      ; BF8C000F
>	s_buffer_load_dword s4, s[0:3], 0x1                 ; C2020101
>	s_buffer_load_dword s5, s[0:3], 0x5                 ; C2028105
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	v_mov_b32_e32 v6, s4                                ; 7E0C0204
>	v_mad_f32 v6, s5, v1, -v6                           ; D2820006 841A0205
>	s_buffer_load_dword s4, s[0:3], 0x0                 ; C2020100
>	s_buffer_load_dword s0, s[0:3], 0x4                 ; C2000104
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	v_mov_b32_e32 v7, s4                                ; 7E0E0204
>	v_mad_f32 v0, s0, v0, -v7                           ; D2820000 841E0000
>	exp 15, 12, 0, 1, 0, v0, v6, v5, v4                 ; F80008CF 04050600
>	s_endpgm                                            ; BF810000
>
>SGPRS: 16
>VGPRS: 8
>Code Size: 108 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL CONST[0..1]
>DCL TEMP[0..2], LOCAL
>IMM[0] FLT32 {    2.2000,     1.0000,     0.4545,     0.0000}
>  0: MAD TEMP[0].xy, IN[0], CONST[1].xxxx, CONST[1].yyyy
>  1: MOV TEMP[0].xy, TEMP[0].xyyy
>  2: TEX TEMP[0], TEMP[0], SAMP[0], 2D
>  3: POW TEMP[1].x, TEMP[0].xxxx, IMM[0].xxxx
>  4: POW TEMP[1].y, TEMP[0].yyyy, IMM[0].xxxx
>  5: POW TEMP[1].z, TEMP[0].zzzz, IMM[0].xxxx
>  6: POW TEMP[1].w, TEMP[0].wwww, IMM[0].yyyy
>  7: MUL TEMP[0].xyz, TEMP[1], CONST[0]
>  8: ABS TEMP[1].x, TEMP[0].xxxx
>  9: LG2 TEMP[1].x, TEMP[1].xxxx
> 10: ABS TEMP[2].x, TEMP[0].yyyy
> 11: LG2 TEMP[2].x, TEMP[2].xxxx
> 12: MOV TEMP[1].y, TEMP[2].xxxx
> 13: ABS TEMP[0].x, TEMP[0].zzzz
> 14: LG2 TEMP[0].x, TEMP[0].xxxx
> 15: MOV TEMP[1].z, TEMP[0].xxxx
> 16: MUL TEMP[0].xyz, TEMP[1], IMM[0].zzzz
> 17: EX2 TEMP[1].x, TEMP[0].xxxx
> 18: EX2 TEMP[2].x, TEMP[0].yyyy
> 19: MOV TEMP[1].y, TEMP[2].xxxx
> 20: EX2 TEMP[0].x, TEMP[0].zzzz
> 21: MOV TEMP[1].z, TEMP[0].xxxx
> 22: MOV TEMP[1].w, IMM[0].yyyy
> 23: MOV OUT[0], TEMP[1]
> 24: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %23 = load <16 x i8> addrspace(2)* %22, !tbaa !0
>  %24 = call float @llvm.SI.load.const(<16 x i8> %23, i32 0)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %23, i32 4)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %23, i32 8)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %23, i32 16)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %23, i32 20)
>  %29 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %30 = load <8 x i32> addrspace(2)* %29, !tbaa !0
>  %31 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %32 = load <4 x i32> addrspace(2)* %31, !tbaa !0
>  %33 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %34 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %35 = fmul float %33, %27
>  %36 = fadd float %35, %28
>  %37 = fmul float %34, %27
>  %38 = fadd float %37, %28
>  %39 = bitcast float %36 to i32
>  %40 = bitcast float %38 to i32
>  %41 = insertelement <2 x i32> undef, i32 %39, i32 0
>  %42 = insertelement <2 x i32> %41, i32 %40, i32 1
>  %43 = bitcast <8 x i32> %30 to <32 x i8>
>  %44 = bitcast <4 x i32> %32 to <16 x i8>
>  %45 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %42, <32 x i8> %43, <16 x i8> %44, i32 2)
>  %46 = extractelement <4 x float> %45, i32 0
>  %47 = extractelement <4 x float> %45, i32 1
>  %48 = extractelement <4 x float> %45, i32 2
>  %49 = call float @llvm.pow.f32(float %46, float 0x40019999A0000000)
>  %50 = call float @llvm.pow.f32(float %47, float 0x40019999A0000000)
>  %51 = call float @llvm.pow.f32(float %48, float 0x40019999A0000000)
>  %52 = fmul float %49, %24
>  %53 = fmul float %50, %25
>  %54 = fmul float %51, %26
>  %55 = call float @fabs(float %52)
>  %56 = call float @llvm.log2.f32(float %55)
>  %57 = call float @fabs(float %53)
>  %58 = call float @llvm.log2.f32(float %57)
>  %59 = call float @fabs(float %54)
>  %60 = call float @llvm.log2.f32(float %59)
>  %61 = fmul float %56, 0x3FDD1743E0000000
>  %62 = fmul float %58, 0x3FDD1743E0000000
>  %63 = fmul float %60, 0x3FDD1743E0000000
>  %64 = call float @llvm.AMDIL.exp.(float %61)
>  %65 = call float @llvm.AMDIL.exp.(float %62)
>  %66 = call float @llvm.AMDIL.exp.(float %63)
>  %67 = call i32 @llvm.SI.packf16(float %64, float %65)
>  %68 = bitcast i32 %67 to float
>  %69 = call i32 @llvm.SI.packf16(float %66, float 1.000000e+00)
>  %70 = bitcast i32 %69 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %68, float %70, float %68, float %70)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.pow.f32(float, float) #1
>
>; Function Attrs: readnone
>declare float @fabs(float) #2
>
>; Function Attrs: nounwind readnone
>declare float @llvm.log2.f32(float) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.exp.(float) #2
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                   ; BEFE0A7E
>	s_mov_b32 m0, s9                                                       ; BEFC0309
>	v_interp_p1_f32 v2, v0, 1, 0, [m0]                                     ; C8080100
>	v_interp_p2_f32 v2, [v2], v1, 1, 0, [m0]                               ; C8090101
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                                     ; C0800300
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	s_buffer_load_dword s8, s[0:3], 0x5                                    ; C2040105
>	s_buffer_load_dword s9, s[0:3], 0x4                                    ; C2048104
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mov_b32_e32 v3, s8                                                   ; 7E060208
>	v_mad_f32 v3, s9, v2, v3                                               ; D2820003 040E0409
>	v_interp_p1_f32 v4, v0, 0, 0, [m0]                                     ; C8100000
>	v_interp_p2_f32 v4, [v4], v1, 0, 0, [m0]                               ; C8110001
>	v_mov_b32_e32 v0, s8                                                   ; 7E000208
>	v_mad_f32 v2, s9, v4, v0                                               ; D2820002 04020809
>	s_load_dwordx4 s[8:11], s[4:5], 0x0                                    ; C0840500
>	s_load_dwordx8 s[12:19], s[6:7], 0x0                                   ; C0C60700
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	image_sample v[0:2], 7, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[12:19], s[8:11] ; F0800700 00430002
>	s_waitcnt vmcnt(0)                                                     ; BF8C0770
>	v_log_f32_e32 v3, v1                                                   ; 7E064F01
>	v_mov_b32_e32 v4, 0x400ccccd                                           ; 7E0802FF 400CCCCD
>	v_mul_legacy_f32_e32 v3, v4, v3                                        ; 0E060704
>	v_exp_f32_e32 v3, v3                                                   ; 7E064B03
>	s_buffer_load_dword s4, s[0:3], 0x1                                    ; C2020101
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mul_f32_e32 v3, s4, v3                                               ; 10060604
>	v_log_f32_e64 v3, |v3|                                                 ; D34E0103 00000103
>	v_mul_f32_e32 v3, 0x3ee8ba1f, v3                                       ; 100606FF 3EE8BA1F
>	v_exp_f32_e32 v3, v3                                                   ; 7E064B03
>	v_log_f32_e32 v5, v0                                                   ; 7E0A4F00
>	v_mul_legacy_f32_e32 v5, v4, v5                                        ; 0E0A0B04
>	v_exp_f32_e32 v5, v5                                                   ; 7E0A4B05
>	s_buffer_load_dword s4, s[0:3], 0x0                                    ; C2020100
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mul_f32_e32 v5, s4, v5                                               ; 100A0A04
>	v_log_f32_e64 v5, |v5|                                                 ; D34E0105 00000105
>	v_mul_f32_e32 v5, 0x3ee8ba1f, v5                                       ; 100A0AFF 3EE8BA1F
>	v_exp_f32_e32 v5, v5                                                   ; 7E0A4B05
>	v_cvt_pkrtz_f16_f32_e32 v3, v5, v3                                     ; 5E060705
>	v_log_f32_e32 v0, v2                                                   ; 7E004F02
>	v_mul_legacy_f32_e32 v0, v4, v0                                        ; 0E000104
>	v_exp_f32_e32 v0, v0                                                   ; 7E004B00
>	s_buffer_load_dword s0, s[0:3], 0x2                                    ; C2000102
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mul_f32_e32 v0, s0, v0                                               ; 10000000
>	v_log_f32_e64 v0, |v0|                                                 ; D34E0100 00000100
>	v_mul_f32_e32 v0, 0x3ee8ba1f, v0                                       ; 100000FF 3EE8BA1F
>	v_exp_f32_e32 v0, v0                                                   ; 7E004B00
>	v_cvt_pkrtz_f16_f32_e64 v0, v0, 1.0                                    ; D25E0000 0001E500
>	exp 15, 0, 1, 1, 1, v3, v0, v3, v0                                     ; F8001C0F 00030003
>	s_endpgm                                                               ; BF810000
>
>SGPRS: 24
>VGPRS: 8
>Code Size: 256 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL CONST[0..4]
>DCL TEMP[0..2], LOCAL
>IMM[0] FLT32 {    0.5000,     0.0000,     1.0000,     0.0000}
>  0: MAD TEMP[0].x, IN[0].xxxx, IMM[0].xxxx, IMM[0].xxxx
>  1: MOV TEMP[1].xy, CONST[3].xyxx
>  2: ADD TEMP[1].yz, -TEMP[1].xxyw, CONST[4].xxyw
>  3: MAD TEMP[2].x, TEMP[0].xxxx, TEMP[1].yyyy, CONST[3].xxxx
>  4: MOV TEMP[0].x, TEMP[2].xxxx
>  5: MAD TEMP[2].w, IN[0].yyyy, IMM[0].xxxx, IMM[0].xxxx
>  6: MOV TEMP[0].w, TEMP[2].wwww
>  7: MAD TEMP[1].z, TEMP[2].wwww, TEMP[1].zzzz, CONST[3].yyyy
>  8: MOV TEMP[0].z, TEMP[1].zzzz
>  9: ADD TEMP[0].xy, TEMP[0].xzzw, CONST[2]
> 10: MAD TEMP[1].zw, TEMP[0].xyxy, CONST[1].xyxy, -CONST[0].xyxy
> 11: MOV TEMP[0].xy, TEMP[0].xyxx
> 12: MOV TEMP[1].xy, TEMP[1].zwzz
> 13: MOV TEMP[1].zw, IMM[0].zzyz
> 14: MOV TEMP[0].zw, IMM[0].zzyz
> 15: MOV OUT[1], TEMP[0]
> 16: MOV OUT[0], TEMP[1]
> 17: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 0)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 4)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %12, i32 16)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %12, i32 20)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %12, i32 32)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %12, i32 36)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %12, i32 48)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %12, i32 52)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %12, i32 64)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %12, i32 68)
>  %23 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %24 = load <16 x i8> addrspace(2)* %23, !tbaa !0
>  %25 = add i32 %5, %7
>  %26 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %24, i32 0, i32 %25)
>  %27 = extractelement <4 x float> %26, i32 0
>  %28 = extractelement <4 x float> %26, i32 1
>  %29 = fmul float %27, 5.000000e-01
>  %30 = fadd float %29, 5.000000e-01
>  %31 = fsub float -0.000000e+00, %19
>  %32 = fadd float %31, %21
>  %33 = fsub float -0.000000e+00, %20
>  %34 = fadd float %33, %22
>  %35 = fmul float %30, %32
>  %36 = fadd float %35, %19
>  %37 = fmul float %28, 5.000000e-01
>  %38 = fadd float %37, 5.000000e-01
>  %39 = fmul float %38, %34
>  %40 = fadd float %39, %20
>  %41 = fadd float %36, %17
>  %42 = fadd float %40, %18
>  %43 = fsub float -0.000000e+00, %13
>  %44 = fmul float %41, %15
>  %45 = fadd float %44, %43
>  %46 = fsub float -0.000000e+00, %14
>  %47 = fmul float %42, %16
>  %48 = fadd float %47, %46
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %41, float %42, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %45, float %48, float 0.000000e+00, float 1.000000e+00)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                   ; C0800300
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	s_buffer_load_dword s4, s[0:3], 0xd                  ; C202010D
>	s_buffer_load_dword s5, s[0:3], 0x11                 ; C2028111
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mov_b32_e32 v1, s4                                 ; 7E020204
>	v_sub_f32_e32 v1, s5, v1                             ; 08020205
>	v_add_i32_e32 v0, s10, v0                            ; 4A00000A
>	s_load_dwordx4 s[8:11], s[8:9], 0x0                  ; C0840900
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	buffer_load_format_xyzw v[2:5], v0, s[8:11], 0 idxen ; E00C2000 80020200
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_mad_f32 v0, 0.5, v3, 0.5                           ; D2820000 03C206F0
>	v_mad_f32 v0, v0, v1, s4                             ; D2820000 00120300
>	s_buffer_load_dword s4, s[0:3], 0x9                  ; C2020109
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_add_f32_e32 v0, s4, v0                             ; 06000004
>	s_buffer_load_dword s4, s[0:3], 0xc                  ; C202010C
>	s_buffer_load_dword s5, s[0:3], 0x10                 ; C2028110
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mov_b32_e32 v1, s4                                 ; 7E020204
>	v_sub_f32_e32 v1, s5, v1                             ; 08020205
>	v_mad_f32 v2, 0.5, v2, 0.5                           ; D2820002 03C204F0
>	v_mad_f32 v1, v2, v1, s4                             ; D2820001 00120302
>	s_buffer_load_dword s4, s[0:3], 0x8                  ; C2020108
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_add_f32_e32 v1, s4, v1                             ; 06020204
>	v_mov_b32_e32 v2, 1.0                                ; 7E0402F2
>	v_mov_b32_e32 v3, 0                                  ; 7E060280
>	exp 15, 32, 0, 0, 0, v1, v0, v3, v2                  ; F800020F 02030001
>	s_buffer_load_dword s4, s[0:3], 0x1                  ; C2020101
>	s_buffer_load_dword s5, s[0:3], 0x5                  ; C2028105
>	s_waitcnt expcnt(0) lgkmcnt(0)                       ; BF8C000F
>	v_mov_b32_e32 v4, s4                                 ; 7E080204
>	v_mad_f32 v0, s5, v0, -v4                            ; D2820000 84120005
>	s_buffer_load_dword s4, s[0:3], 0x0                  ; C2020100
>	s_buffer_load_dword s0, s[0:3], 0x4                  ; C2000104
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mov_b32_e32 v4, s4                                 ; 7E080204
>	v_mad_f32 v1, s0, v1, -v4                            ; D2820001 84120200
>	exp 15, 12, 0, 1, 0, v1, v0, v3, v2                  ; F80008CF 02030001
>	s_endpgm                                             ; BF810000
>
>SGPRS: 16
>VGPRS: 8
>Code Size: 204 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL CONST[0]
>DCL TEMP[0..2], LOCAL
>IMM[0] FLT32 {    2.2000,     1.0000,     0.4545,     0.0000}
>  0: MAD TEMP[0].xy, IN[0], CONST[0].xxxx, CONST[0].yyyy
>  1: MOV TEMP[0].xy, TEMP[0].xyyy
>  2: TEX TEMP[0].xyz, TEMP[0], SAMP[0], 2D
>  3: POW TEMP[1].x, TEMP[0].xxxx, IMM[0].xxxx
>  4: POW TEMP[1].y, TEMP[0].yyyy, IMM[0].xxxx
>  5: POW TEMP[1].z, TEMP[0].zzzz, IMM[0].xxxx
>  6: ABS TEMP[0].x, TEMP[1].xxxx
>  7: LG2 TEMP[0].x, TEMP[0].xxxx
>  8: ABS TEMP[2].x, TEMP[1].yyyy
>  9: LG2 TEMP[2].x, TEMP[2].xxxx
> 10: MOV TEMP[0].y, TEMP[2].xxxx
> 11: ABS TEMP[1].x, TEMP[1].zzzz
> 12: LG2 TEMP[1].x, TEMP[1].xxxx
> 13: MOV TEMP[0].z, TEMP[1].xxxx
> 14: MUL TEMP[0].xyz, TEMP[0], IMM[0].zzzz
> 15: EX2 TEMP[1].x, TEMP[0].xxxx
> 16: EX2 TEMP[2].x, TEMP[0].yyyy
> 17: MOV TEMP[1].y, TEMP[2].xxxx
> 18: EX2 TEMP[0].x, TEMP[0].zzzz
> 19: MOV TEMP[1].z, TEMP[0].xxxx
> 20: MOV TEMP[1].w, IMM[0].yyyy
> 21: MOV OUT[0], TEMP[1]
> 22: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %23 = load <16 x i8> addrspace(2)* %22, !tbaa !0
>  %24 = call float @llvm.SI.load.const(<16 x i8> %23, i32 0)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %23, i32 4)
>  %26 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %27 = load <8 x i32> addrspace(2)* %26, !tbaa !0
>  %28 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %29 = load <4 x i32> addrspace(2)* %28, !tbaa !0
>  %30 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %31 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %32 = fmul float %30, %24
>  %33 = fadd float %32, %25
>  %34 = fmul float %31, %24
>  %35 = fadd float %34, %25
>  %36 = bitcast float %33 to i32
>  %37 = bitcast float %35 to i32
>  %38 = insertelement <2 x i32> undef, i32 %36, i32 0
>  %39 = insertelement <2 x i32> %38, i32 %37, i32 1
>  %40 = bitcast <8 x i32> %27 to <32 x i8>
>  %41 = bitcast <4 x i32> %29 to <16 x i8>
>  %42 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %39, <32 x i8> %40, <16 x i8> %41, i32 2)
>  %43 = extractelement <4 x float> %42, i32 0
>  %44 = extractelement <4 x float> %42, i32 1
>  %45 = extractelement <4 x float> %42, i32 2
>  %46 = call float @llvm.pow.f32(float %43, float 0x40019999A0000000)
>  %47 = call float @llvm.pow.f32(float %44, float 0x40019999A0000000)
>  %48 = call float @llvm.pow.f32(float %45, float 0x40019999A0000000)
>  %49 = call float @fabs(float %46)
>  %50 = call float @llvm.log2.f32(float %49)
>  %51 = call float @fabs(float %47)
>  %52 = call float @llvm.log2.f32(float %51)
>  %53 = call float @fabs(float %48)
>  %54 = call float @llvm.log2.f32(float %53)
>  %55 = fmul float %50, 0x3FDD1743E0000000
>  %56 = fmul float %52, 0x3FDD1743E0000000
>  %57 = fmul float %54, 0x3FDD1743E0000000
>  %58 = call float @llvm.AMDIL.exp.(float %55)
>  %59 = call float @llvm.AMDIL.exp.(float %56)
>  %60 = call float @llvm.AMDIL.exp.(float %57)
>  %61 = call i32 @llvm.SI.packf16(float %58, float %59)
>  %62 = bitcast i32 %61 to float
>  %63 = call i32 @llvm.SI.packf16(float %60, float 1.000000e+00)
>  %64 = bitcast i32 %63 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %62, float %64, float %62, float %64)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.pow.f32(float, float) #1
>
>; Function Attrs: readnone
>declare float @fabs(float) #2
>
>; Function Attrs: nounwind readnone
>declare float @llvm.log2.f32(float) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.exp.(float) #2
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                 ; BEFE0A7E
>	s_mov_b32 m0, s9                                                     ; BEFC0309
>	v_interp_p1_f32 v2, v0, 1, 0, [m0]                                   ; C8080100
>	v_interp_p2_f32 v2, [v2], v1, 1, 0, [m0]                             ; C8090101
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                                   ; C0800300
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	s_buffer_load_dword s8, s[0:3], 0x1                                  ; C2040101
>	s_buffer_load_dword s0, s[0:3], 0x0                                  ; C2000100
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mov_b32_e32 v3, s8                                                 ; 7E060208
>	v_mad_f32 v3, s0, v2, v3                                             ; D2820003 040E0400
>	v_interp_p1_f32 v4, v0, 0, 0, [m0]                                   ; C8100000
>	v_interp_p2_f32 v4, [v4], v1, 0, 0, [m0]                             ; C8110001
>	v_mov_b32_e32 v0, s8                                                 ; 7E000208
>	v_mad_f32 v2, s0, v4, v0                                             ; D2820002 04020800
>	s_load_dwordx4 s[0:3], s[4:5], 0x0                                   ; C0800500
>	s_load_dwordx8 s[4:11], s[6:7], 0x0                                  ; C0C20700
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	image_sample v[0:2], 7, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[4:11], s[0:3] ; F0800700 00010002
>	s_waitcnt vmcnt(0)                                                   ; BF8C0770
>	v_log_f32_e32 v3, v1                                                 ; 7E064F01
>	v_mov_b32_e32 v4, 0x400ccccd                                         ; 7E0802FF 400CCCCD
>	v_mul_legacy_f32_e32 v3, v4, v3                                      ; 0E060704
>	v_exp_f32_e32 v3, v3                                                 ; 7E064B03
>	v_log_f32_e64 v3, |v3|                                               ; D34E0103 00000103
>	v_mul_f32_e32 v3, 0x3ee8ba1f, v3                                     ; 100606FF 3EE8BA1F
>	v_exp_f32_e32 v3, v3                                                 ; 7E064B03
>	v_log_f32_e32 v5, v0                                                 ; 7E0A4F00
>	v_mul_legacy_f32_e32 v5, v4, v5                                      ; 0E0A0B04
>	v_exp_f32_e32 v5, v5                                                 ; 7E0A4B05
>	v_log_f32_e64 v5, |v5|                                               ; D34E0105 00000105
>	v_mul_f32_e32 v5, 0x3ee8ba1f, v5                                     ; 100A0AFF 3EE8BA1F
>	v_exp_f32_e32 v5, v5                                                 ; 7E0A4B05
>	v_cvt_pkrtz_f16_f32_e32 v3, v5, v3                                   ; 5E060705
>	v_log_f32_e32 v0, v2                                                 ; 7E004F02
>	v_mul_legacy_f32_e32 v0, v4, v0                                      ; 0E000104
>	v_exp_f32_e32 v0, v0                                                 ; 7E004B00
>	v_log_f32_e64 v0, |v0|                                               ; D34E0100 00000100
>	v_mul_f32_e32 v0, 0x3ee8ba1f, v0                                     ; 100000FF 3EE8BA1F
>	v_exp_f32_e32 v0, v0                                                 ; 7E004B00
>	v_cvt_pkrtz_f16_f32_e64 v0, v0, 1.0                                  ; D25E0000 0001E500
>	exp 15, 0, 1, 1, 1, v3, v0, v3, v0                                   ; F8001C0F 00030003
>	s_endpgm                                                             ; BF810000
>
>SGPRS: 16
>VGPRS: 8
>Code Size: 220 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL CONST[0..1]
>DCL TEMP[0..1], LOCAL
>IMM[0] FLT32 {    0.0000,     1.0000,     0.0000,     0.0000}
>  0: MOV TEMP[0].xy, CONST[1].xyxx
>  1: MAD TEMP[0].xy, IN[0], TEMP[0], -CONST[0]
>  2: MOV TEMP[0].xy, TEMP[0].xyxx
>  3: MOV TEMP[1].xy, IN[0].xyxx
>  4: MOV TEMP[0].zw, IMM[0].yyxy
>  5: MOV TEMP[1].zw, IMM[0].yyxy
>  6: MOV OUT[1], TEMP[1]
>  7: MOV OUT[0], TEMP[0]
>  8: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 0)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 4)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %12, i32 16)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %12, i32 20)
>  %17 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %18 = load <16 x i8> addrspace(2)* %17, !tbaa !0
>  %19 = add i32 %5, %7
>  %20 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %18, i32 0, i32 %19)
>  %21 = extractelement <4 x float> %20, i32 0
>  %22 = extractelement <4 x float> %20, i32 1
>  %23 = fsub float -0.000000e+00, %13
>  %24 = fmul float %21, %15
>  %25 = fadd float %24, %23
>  %26 = fsub float -0.000000e+00, %14
>  %27 = fmul float %22, %16
>  %28 = fadd float %27, %26
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %21, float %22, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %25, float %28, float 0.000000e+00, float 1.000000e+00)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                           ; 4A00000A
>	s_load_dwordx4 s[4:7], s[8:9], 0x0                  ; C0820900
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	buffer_load_format_xyzw v[0:3], v0, s[4:7], 0 idxen ; E00C2000 80010000
>	v_mov_b32_e32 v4, 1.0                               ; 7E0802F2
>	v_mov_b32_e32 v5, 0                                 ; 7E0A0280
>	s_waitcnt vmcnt(0)                                  ; BF8C0770
>	exp 15, 32, 0, 0, 0, v0, v1, v5, v4                 ; F800020F 04050100
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                  ; C0800300
>	s_waitcnt expcnt(0) lgkmcnt(0)                      ; BF8C000F
>	s_buffer_load_dword s4, s[0:3], 0x1                 ; C2020101
>	s_buffer_load_dword s5, s[0:3], 0x5                 ; C2028105
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	v_mov_b32_e32 v6, s4                                ; 7E0C0204
>	v_mad_f32 v6, s5, v1, -v6                           ; D2820006 841A0205
>	s_buffer_load_dword s4, s[0:3], 0x0                 ; C2020100
>	s_buffer_load_dword s0, s[0:3], 0x4                 ; C2000104
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	v_mov_b32_e32 v7, s4                                ; 7E0E0204
>	v_mad_f32 v0, s0, v0, -v7                           ; D2820000 841E0000
>	exp 15, 12, 0, 1, 0, v0, v6, v5, v4                 ; F80008CF 04050600
>	s_endpgm                                            ; BF810000
>
>SGPRS: 16
>VGPRS: 8
>Code Size: 108 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL CONST[0]
>DCL TEMP[0..2], LOCAL
>IMM[0] FLT32 {    2.2000,     1.0000,     0.4545,     0.0000}
>  0: MAD TEMP[0].xy, IN[0], CONST[0].xxxx, CONST[0].yyyy
>  1: MOV TEMP[0].xy, TEMP[0].xyyy
>  2: TEX TEMP[0], TEMP[0], SAMP[0], 2D
>  3: POW TEMP[1].x, TEMP[0].xxxx, IMM[0].xxxx
>  4: POW TEMP[1].y, TEMP[0].yyyy, IMM[0].xxxx
>  5: POW TEMP[1].z, TEMP[0].zzzz, IMM[0].xxxx
>  6: POW TEMP[1].w, TEMP[0].wwww, IMM[0].yyyy
>  7: ABS TEMP[0].x, TEMP[1].xxxx
>  8: LG2 TEMP[0].x, TEMP[0].xxxx
>  9: ABS TEMP[2].x, TEMP[1].yyyy
> 10: LG2 TEMP[2].x, TEMP[2].xxxx
> 11: MOV TEMP[0].y, TEMP[2].xxxx
> 12: ABS TEMP[2].x, TEMP[1].zzzz
> 13: LG2 TEMP[2].x, TEMP[2].xxxx
> 14: MOV TEMP[0].z, TEMP[2].xxxx
> 15: MOV TEMP[1].w, TEMP[1].wwww
> 16: MUL TEMP[0].xyz, TEMP[0], IMM[0].zzzz
> 17: EX2 TEMP[1].x, TEMP[0].xxxx
> 18: EX2 TEMP[2].x, TEMP[0].yyyy
> 19: MOV TEMP[1].y, TEMP[2].xxxx
> 20: EX2 TEMP[0].x, TEMP[0].zzzz
> 21: MOV TEMP[1].z, TEMP[0].xxxx
> 22: MOV OUT[0], TEMP[1]
> 23: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %23 = load <16 x i8> addrspace(2)* %22, !tbaa !0
>  %24 = call float @llvm.SI.load.const(<16 x i8> %23, i32 0)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %23, i32 4)
>  %26 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %27 = load <8 x i32> addrspace(2)* %26, !tbaa !0
>  %28 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %29 = load <4 x i32> addrspace(2)* %28, !tbaa !0
>  %30 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %31 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %32 = fmul float %30, %24
>  %33 = fadd float %32, %25
>  %34 = fmul float %31, %24
>  %35 = fadd float %34, %25
>  %36 = bitcast float %33 to i32
>  %37 = bitcast float %35 to i32
>  %38 = insertelement <2 x i32> undef, i32 %36, i32 0
>  %39 = insertelement <2 x i32> %38, i32 %37, i32 1
>  %40 = bitcast <8 x i32> %27 to <32 x i8>
>  %41 = bitcast <4 x i32> %29 to <16 x i8>
>  %42 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %39, <32 x i8> %40, <16 x i8> %41, i32 2)
>  %43 = extractelement <4 x float> %42, i32 0
>  %44 = extractelement <4 x float> %42, i32 1
>  %45 = extractelement <4 x float> %42, i32 2
>  %46 = extractelement <4 x float> %42, i32 3
>  %47 = call float @llvm.pow.f32(float %43, float 0x40019999A0000000)
>  %48 = call float @llvm.pow.f32(float %44, float 0x40019999A0000000)
>  %49 = call float @llvm.pow.f32(float %45, float 0x40019999A0000000)
>  %50 = call float @llvm.pow.f32(float %46, float 1.000000e+00)
>  %51 = call float @fabs(float %47)
>  %52 = call float @llvm.log2.f32(float %51)
>  %53 = call float @fabs(float %48)
>  %54 = call float @llvm.log2.f32(float %53)
>  %55 = call float @fabs(float %49)
>  %56 = call float @llvm.log2.f32(float %55)
>  %57 = fmul float %52, 0x3FDD1743E0000000
>  %58 = fmul float %54, 0x3FDD1743E0000000
>  %59 = fmul float %56, 0x3FDD1743E0000000
>  %60 = call float @llvm.AMDIL.exp.(float %57)
>  %61 = call float @llvm.AMDIL.exp.(float %58)
>  %62 = call float @llvm.AMDIL.exp.(float %59)
>  %63 = call i32 @llvm.SI.packf16(float %60, float %61)
>  %64 = bitcast i32 %63 to float
>  %65 = call i32 @llvm.SI.packf16(float %62, float %50)
>  %66 = bitcast i32 %65 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %64, float %66, float %64, float %66)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.pow.f32(float, float) #1
>
>; Function Attrs: readnone
>declare float @fabs(float) #2
>
>; Function Attrs: nounwind readnone
>declare float @llvm.log2.f32(float) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.exp.(float) #2
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                  ; BEFE0A7E
>	s_mov_b32 m0, s9                                                      ; BEFC0309
>	v_interp_p1_f32 v2, v0, 1, 0, [m0]                                    ; C8080100
>	v_interp_p2_f32 v2, [v2], v1, 1, 0, [m0]                              ; C8090101
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                                    ; C0800300
>	s_waitcnt lgkmcnt(0)                                                  ; BF8C007F
>	s_buffer_load_dword s8, s[0:3], 0x1                                   ; C2040101
>	s_buffer_load_dword s0, s[0:3], 0x0                                   ; C2000100
>	s_waitcnt lgkmcnt(0)                                                  ; BF8C007F
>	v_mov_b32_e32 v3, s8                                                  ; 7E060208
>	v_mad_f32 v3, s0, v2, v3                                              ; D2820003 040E0400
>	v_interp_p1_f32 v4, v0, 0, 0, [m0]                                    ; C8100000
>	v_interp_p2_f32 v4, [v4], v1, 0, 0, [m0]                              ; C8110001
>	v_mov_b32_e32 v0, s8                                                  ; 7E000208
>	v_mad_f32 v2, s0, v4, v0                                              ; D2820002 04020800
>	s_load_dwordx4 s[0:3], s[4:5], 0x0                                    ; C0800500
>	s_load_dwordx8 s[4:11], s[6:7], 0x0                                   ; C0C20700
>	s_waitcnt lgkmcnt(0)                                                  ; BF8C007F
>	image_sample v[0:3], 15, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[4:11], s[0:3] ; F0800F00 00010002
>	s_waitcnt vmcnt(0)                                                    ; BF8C0770
>	v_log_f32_e32 v4, v3                                                  ; 7E084F03
>	v_mul_legacy_f32_e32 v4, 1.0, v4                                      ; 0E0808F2
>	v_exp_f32_e32 v4, v4                                                  ; 7E084B04
>	v_log_f32_e32 v5, v2                                                  ; 7E0A4F02
>	v_mov_b32_e32 v6, 0x400ccccd                                          ; 7E0C02FF 400CCCCD
>	v_mul_legacy_f32_e32 v5, v6, v5                                       ; 0E0A0B06
>	v_exp_f32_e32 v5, v5                                                  ; 7E0A4B05
>	v_log_f32_e64 v5, |v5|                                                ; D34E0105 00000105
>	v_mul_f32_e32 v5, 0x3ee8ba1f, v5                                      ; 100A0AFF 3EE8BA1F
>	v_exp_f32_e32 v5, v5                                                  ; 7E0A4B05
>	v_cvt_pkrtz_f16_f32_e32 v4, v5, v4                                    ; 5E080905
>	v_log_f32_e32 v5, v1                                                  ; 7E0A4F01
>	v_mul_legacy_f32_e32 v5, v6, v5                                       ; 0E0A0B06
>	v_exp_f32_e32 v5, v5                                                  ; 7E0A4B05
>	v_log_f32_e64 v5, |v5|                                                ; D34E0105 00000105
>	v_mul_f32_e32 v5, 0x3ee8ba1f, v5                                      ; 100A0AFF 3EE8BA1F
>	v_exp_f32_e32 v5, v5                                                  ; 7E0A4B05
>	v_log_f32_e32 v0, v0                                                  ; 7E004F00
>	v_mul_legacy_f32_e32 v0, v6, v0                                       ; 0E000106
>	v_exp_f32_e32 v0, v0                                                  ; 7E004B00
>	v_log_f32_e64 v0, |v0|                                                ; D34E0100 00000100
>	v_mul_f32_e32 v0, 0x3ee8ba1f, v0                                      ; 100000FF 3EE8BA1F
>	v_exp_f32_e32 v0, v0                                                  ; 7E004B00
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v5                                    ; 5E000B00
>	exp 15, 0, 1, 1, 1, v0, v4, v0, v4                                    ; F8001C0F 04000400
>	s_endpgm                                                              ; BF810000
>
>SGPRS: 16
>VGPRS: 8
>Code Size: 228 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL IN[2]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL OUT[2], GENERIC[10]
>DCL CONST[0..1]
>DCL TEMP[0..3], LOCAL
>IMM[0] FLT32 {    2.2000,     0.0000,     1.0000,     0.0000}
>  0: ABS TEMP[0].x, IN[2].xxxx
>  1: LG2 TEMP[0].x, TEMP[0].xxxx
>  2: ABS TEMP[1].x, IN[2].yyyy
>  3: LG2 TEMP[1].x, TEMP[1].xxxx
>  4: MOV TEMP[0].y, TEMP[1].xxxx
>  5: ABS TEMP[1].x, IN[2].zzzz
>  6: LG2 TEMP[1].x, TEMP[1].xxxx
>  7: MOV TEMP[0].z, TEMP[1].xxxx
>  8: MUL TEMP[1].xyz, TEMP[0], IMM[0].xxxx
>  9: MOV TEMP[0].z, TEMP[1].xyzx
> 10: EX2 TEMP[2].x, TEMP[1].xxxx
> 11: EX2 TEMP[3].x, TEMP[1].yyyy
> 12: MOV TEMP[2].y, TEMP[3].xxxx
> 13: EX2 TEMP[1].x, TEMP[1].zzzz
> 14: MOV TEMP[2].z, TEMP[1].xxxx
> 15: MOV TEMP[0].xy, CONST[1].xyxx
> 16: MAD TEMP[0].xy, IN[0], TEMP[0], -CONST[0]
> 17: MOV TEMP[0].xy, TEMP[0].xyxx
> 18: MOV TEMP[1].xy, IN[1].xyxx
> 19: MOV TEMP[2].w, IN[2].wwww
> 20: MOV TEMP[0].zw, IMM[0].zzyz
> 21: MOV TEMP[1].zw, IMM[0].zzyz
> 22: MOV OUT[1], TEMP[2]
> 23: MOV OUT[2], TEMP[1]
> 24: MOV OUT[0], TEMP[0]
> 25: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 0)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 4)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %12, i32 16)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %12, i32 20)
>  %17 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %18 = load <16 x i8> addrspace(2)* %17, !tbaa !0
>  %19 = add i32 %5, %7
>  %20 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %18, i32 0, i32 %19)
>  %21 = extractelement <4 x float> %20, i32 0
>  %22 = extractelement <4 x float> %20, i32 1
>  %23 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %24 = load <16 x i8> addrspace(2)* %23, !tbaa !0
>  %25 = add i32 %5, %7
>  %26 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %24, i32 0, i32 %25)
>  %27 = extractelement <4 x float> %26, i32 0
>  %28 = extractelement <4 x float> %26, i32 1
>  %29 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 2
>  %30 = load <16 x i8> addrspace(2)* %29, !tbaa !0
>  %31 = add i32 %5, %7
>  %32 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %30, i32 0, i32 %31)
>  %33 = extractelement <4 x float> %32, i32 0
>  %34 = extractelement <4 x float> %32, i32 1
>  %35 = extractelement <4 x float> %32, i32 2
>  %36 = extractelement <4 x float> %32, i32 3
>  %37 = call float @fabs(float %33)
>  %38 = call float @llvm.log2.f32(float %37)
>  %39 = call float @fabs(float %34)
>  %40 = call float @llvm.log2.f32(float %39)
>  %41 = call float @fabs(float %35)
>  %42 = call float @llvm.log2.f32(float %41)
>  %43 = fmul float %38, 0x40019999A0000000
>  %44 = fmul float %40, 0x40019999A0000000
>  %45 = fmul float %42, 0x40019999A0000000
>  %46 = call float @llvm.AMDIL.exp.(float %43)
>  %47 = call float @llvm.AMDIL.exp.(float %44)
>  %48 = call float @llvm.AMDIL.exp.(float %45)
>  %49 = fsub float -0.000000e+00, %13
>  %50 = fmul float %21, %15
>  %51 = fadd float %50, %49
>  %52 = fsub float -0.000000e+00, %14
>  %53 = fmul float %22, %16
>  %54 = fadd float %53, %52
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %46, float %47, float %48, float %36)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %27, float %28, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %51, float %54, float 0.000000e+00, float 1.000000e+00)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>; Function Attrs: readnone
>declare float @fabs(float) #2
>
>; Function Attrs: nounwind readnone
>declare float @llvm.log2.f32(float) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.exp.(float) #2
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                             ; 4A00000A
>	s_load_dwordx4 s[4:7], s[8:9], 0x0                    ; C0820900
>	s_load_dwordx4 s[12:15], s[8:9], 0x4                  ; C0860904
>	s_load_dwordx4 s[8:11], s[8:9], 0x8                   ; C0840908
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[8:11], 0 idxen  ; E00C2000 80020100
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_log_f32_e64 v5, |v3|                                ; D34E0105 00000103
>	v_mul_f32_e32 v5, 0x400ccccd, v5                      ; 100A0AFF 400CCCCD
>	v_exp_f32_e32 v5, v5                                  ; 7E0A4B05
>	v_log_f32_e64 v6, |v2|                                ; D34E0106 00000102
>	v_mul_f32_e32 v6, 0x400ccccd, v6                      ; 100C0CFF 400CCCCD
>	v_exp_f32_e32 v6, v6                                  ; 7E0C4B06
>	v_log_f32_e64 v7, |v1|                                ; D34E0107 00000101
>	v_mul_f32_e32 v7, 0x400ccccd, v7                      ; 100E0EFF 400CCCCD
>	v_exp_f32_e32 v7, v7                                  ; 7E0E4B07
>	exp 15, 32, 0, 0, 0, v7, v6, v5, v4                   ; F800020F 04050607
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	buffer_load_format_xyzw v[1:4], v0, s[12:15], 0 idxen ; E00C2000 80030100
>	v_mov_b32_e32 v5, 1.0                                 ; 7E0A02F2
>	v_mov_b32_e32 v6, 0                                   ; 7E0C0280
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	exp 15, 33, 0, 0, 0, v1, v2, v6, v5                   ; F800021F 05060201
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	buffer_load_format_xyzw v[0:3], v0, s[4:7], 0 idxen   ; E00C2000 80010000
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                    ; C0800300
>	s_waitcnt vmcnt(0) lgkmcnt(0)                         ; BF8C0070
>	s_buffer_load_dword s4, s[0:3], 0x1                   ; C2020101
>	s_buffer_load_dword s5, s[0:3], 0x5                   ; C2028105
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mov_b32_e32 v4, s4                                  ; 7E080204
>	v_mad_f32 v4, s5, v1, -v4                             ; D2820004 84120205
>	s_buffer_load_dword s4, s[0:3], 0x0                   ; C2020100
>	s_buffer_load_dword s0, s[0:3], 0x4                   ; C2000104
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mov_b32_e32 v7, s4                                  ; 7E0E0204
>	v_mad_f32 v0, s0, v0, -v7                             ; D2820000 841E0000
>	exp 15, 12, 0, 1, 0, v0, v4, v6, v5                   ; F80008CF 05060400
>	s_endpgm                                              ; BF810000
>
>SGPRS: 24
>VGPRS: 8
>Code Size: 212 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL IN[1], GENERIC[10], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL CONST[0]
>DCL TEMP[0..2], LOCAL
>IMM[0] FLT32 {   -0.5000,     0.0000,     2.0000,    -2.0000}
>IMM[1] FLT32 {    3.0000,     1.0000,     0.4545,     0.0000}
>  0: ADD TEMP[0].xy, IMM[0].xxxx, IN[1]
>  1: DP2 TEMP[0].x, TEMP[0].xyyy, TEMP[0].xyyy
>  2: MAX TEMP[1].x, TEMP[0].xxxx, IMM[0].yyyy
>  3: SQRT TEMP[0].x, TEMP[1].xxxx
>  4: MOV TEMP[1].x, -CONST[0]
>  5: MAD TEMP[2].x, TEMP[0].xxxx, IMM[0].zzzz, TEMP[1].xxxx
>  6: ADD TEMP[1].y, TEMP[1].xxxx, CONST[0].yyyy
>  7: RCP TEMP[1].x, TEMP[1].yyyy
>  8: MUL TEMP[1], TEMP[1].xxxx, TEMP[2].xxxx
>  9: MOV_SAT TEMP[1], TEMP[1]
> 10: MAD TEMP[2].y, TEMP[1].xxxx, IMM[0].wwww, IMM[1].xxxx
> 11: MUL TEMP[1].x, TEMP[1].xxxx, TEMP[1].xxxx
> 12: MAD TEMP[1].x, TEMP[2].yyyy, -TEMP[1].xxxx, IMM[1].yyyy
> 13: MUL TEMP[1].w, TEMP[1].xxxx, IN[0].wwww
> 14: MOV TEMP[1].w, TEMP[1].wwww
> 15: ABS TEMP[2].x, IN[0].xxxx
> 16: LG2 TEMP[0].x, TEMP[2].xxxx
> 17: ABS TEMP[2].x, IN[0].yyyy
> 18: LG2 TEMP[2].x, TEMP[2].xxxx
> 19: MOV TEMP[0].y, TEMP[2].xxxx
> 20: ABS TEMP[2].x, IN[0].zzzz
> 21: LG2 TEMP[2].x, TEMP[2].xxxx
> 22: MOV TEMP[0].z, TEMP[2].xxxx
> 23: MUL TEMP[0].xyz, TEMP[0], IMM[1].zzzz
> 24: EX2 TEMP[1].x, TEMP[0].xxxx
> 25: EX2 TEMP[2].x, TEMP[0].yyyy
> 26: MOV TEMP[1].y, TEMP[2].xxxx
> 27: EX2 TEMP[0].x, TEMP[0].zzzz
> 28: MOV TEMP[1].z, TEMP[0].xxxx
> 29: MOV OUT[0], TEMP[1]
> 30: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %23 = load <16 x i8> addrspace(2)* %22, !tbaa !0
>  %24 = call float @llvm.SI.load.const(<16 x i8> %23, i32 0)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %23, i32 4)
>  %26 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %27 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %28 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %5, <2 x i32> %7)
>  %29 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %5, <2 x i32> %7)
>  %30 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %5, <2 x i32> %7)
>  %31 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %5, <2 x i32> %7)
>  %32 = fadd float -5.000000e-01, %30
>  %33 = fadd float -5.000000e-01, %31
>  %34 = fmul float %32, %32
>  %35 = fmul float %33, %33
>  %36 = fadd float %34, %35
>  %37 = call float @llvm.maxnum.f32(float %36, float 0x3E7AD7F2A0000000)
>  %38 = call float @llvm.sqrt.f32(float %37)
>  %39 = fsub float -0.000000e+00, %24
>  %40 = fmul float %38, 2.000000e+00
>  %41 = fadd float %40, %39
>  %42 = fadd float %39, %25
>  %43 = fdiv float 1.000000e+00, %42
>  %44 = fmul float %43, %41
>  %45 = fmul float %43, %41
>  %46 = fmul float %43, %41
>  %47 = fmul float %43, %41
>  %48 = call float @llvm.AMDIL.clamp.(float %44, float 0.000000e+00, float 1.000000e+00)
>  %49 = call float @llvm.AMDIL.clamp.(float %45, float 0.000000e+00, float 1.000000e+00)
>  %50 = call float @llvm.AMDIL.clamp.(float %46, float 0.000000e+00, float 1.000000e+00)
>  %51 = call float @llvm.AMDIL.clamp.(float %47, float 0.000000e+00, float 1.000000e+00)
>  %52 = fmul float %48, -2.000000e+00
>  %53 = fadd float %52, 3.000000e+00
>  %54 = fmul float %48, %48
>  %55 = fsub float -0.000000e+00, %54
>  %56 = fmul float %53, %55
>  %57 = fadd float %56, 1.000000e+00
>  %58 = fmul float %57, %29
>  %59 = call float @fabs(float %26)
>  %60 = call float @llvm.log2.f32(float %59)
>  %61 = call float @fabs(float %27)
>  %62 = call float @llvm.log2.f32(float %61)
>  %63 = call float @fabs(float %28)
>  %64 = call float @llvm.log2.f32(float %63)
>  %65 = fmul float %60, 0x3FDD1743E0000000
>  %66 = fmul float %62, 0x3FDD1743E0000000
>  %67 = fmul float %64, 0x3FDD1743E0000000
>  %68 = call float @llvm.AMDIL.exp.(float %65)
>  %69 = call float @llvm.AMDIL.exp.(float %66)
>  %70 = call float @llvm.AMDIL.exp.(float %67)
>  %71 = call i32 @llvm.SI.packf16(float %68, float %69)
>  %72 = bitcast i32 %71 to float
>  %73 = call i32 @llvm.SI.packf16(float %70, float %58)
>  %74 = bitcast i32 %73 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %72, float %74, float %72, float %74)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.maxnum.f32(float, float) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.sqrt.f32(float) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.clamp.(float, float, float) #2
>
>; Function Attrs: readnone
>declare float @fabs(float) #2
>
>; Function Attrs: nounwind readnone
>declare float @llvm.log2.f32(float) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.exp.(float) #2
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                     ; BEFE0A7E
>	s_mov_b32 m0, s9                         ; BEFC0309
>	v_interp_p1_f32 v2, v0, 0, 1, [m0]       ; C8080400
>	v_interp_p2_f32 v2, [v2], v1, 0, 1, [m0] ; C8090401
>	v_add_f32_e32 v2, -0.5, v2               ; 060404F1
>	v_interp_p1_f32 v3, v0, 1, 1, [m0]       ; C80C0500
>	v_interp_p2_f32 v3, [v3], v1, 1, 1, [m0] ; C80D0501
>	v_add_f32_e32 v3, -0.5, v3               ; 060606F1
>	v_mul_f32_e32 v3, v3, v3                 ; 10060703
>	v_mad_f32 v2, v2, v2, v3                 ; D2820002 040E0502
>	v_max_f32_e32 v2, 0x33d6bf95, v2         ; 200404FF 33D6BF95
>	v_sqrt_f32_e32 v2, v2                    ; 7E046702
>	s_load_dwordx4 s[0:3], s[2:3], 0x0       ; C0800300
>	s_waitcnt lgkmcnt(0)                     ; BF8C007F
>	s_buffer_load_dword s4, s[0:3], 0x0      ; C2020100
>	s_waitcnt lgkmcnt(0)                     ; BF8C007F
>	v_mad_f32 v2, 2.0, v2, -s4               ; D2820002 801204F4
>	s_buffer_load_dword s0, s[0:3], 0x1      ; C2000101
>	v_mov_b32_e32 v3, s4                     ; 7E060204
>	s_waitcnt lgkmcnt(0)                     ; BF8C007F
>	v_sub_f32_e32 v3, s0, v3                 ; 08060600
>	v_rcp_f32_e32 v3, v3                     ; 7E065503
>	v_mul_f32_e32 v2, v2, v3                 ; 10040702
>	v_add_f32_e64 v2, 0, v2 clamp            ; D2060802 00020480
>	v_mov_b32_e32 v3, 0x40400000             ; 7E0602FF 40400000
>	v_mad_f32 v3, -2.0, v2, v3               ; D2820003 040E04F5
>	v_mul_f32_e32 v2, v2, v2                 ; 10040502
>	v_mad_f32 v2, -v3, v2, 1.0               ; D2820002 23CA0503
>	v_interp_p1_f32 v3, v0, 3, 0, [m0]       ; C80C0300
>	v_interp_p2_f32 v3, [v3], v1, 3, 0, [m0] ; C80D0301
>	v_mul_f32_e32 v2, v3, v2                 ; 10040503
>	v_interp_p1_f32 v3, v0, 2, 0, [m0]       ; C80C0200
>	v_interp_p2_f32 v3, [v3], v1, 2, 0, [m0] ; C80D0201
>	v_log_f32_e64 v3, |v3|                   ; D34E0103 00000103
>	v_mul_f32_e32 v3, 0x3ee8ba1f, v3         ; 100606FF 3EE8BA1F
>	v_exp_f32_e32 v3, v3                     ; 7E064B03
>	v_cvt_pkrtz_f16_f32_e32 v2, v3, v2       ; 5E040503
>	v_interp_p1_f32 v3, v0, 1, 0, [m0]       ; C80C0100
>	v_interp_p2_f32 v3, [v3], v1, 1, 0, [m0] ; C80D0101
>	v_log_f32_e64 v3, |v3|                   ; D34E0103 00000103
>	v_mul_f32_e32 v3, 0x3ee8ba1f, v3         ; 100606FF 3EE8BA1F
>	v_exp_f32_e32 v3, v3                     ; 7E064B03
>	v_interp_p1_f32 v4, v0, 0, 0, [m0]       ; C8100000
>	v_interp_p2_f32 v4, [v4], v1, 0, 0, [m0] ; C8110001
>	v_log_f32_e64 v0, |v4|                   ; D34E0100 00000104
>	v_mul_f32_e32 v0, 0x3ee8ba1f, v0         ; 100000FF 3EE8BA1F
>	v_exp_f32_e32 v0, v0                     ; 7E004B00
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v3       ; 5E000700
>	exp 15, 0, 1, 1, 1, v0, v2, v0, v2       ; F8001C0F 02000200
>	s_endpgm                                 ; BF810000
>
>SGPRS: 16
>VGPRS: 8
>Code Size: 256 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL IN[2]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL OUT[2], GENERIC[10]
>DCL CONST[0..1]
>DCL TEMP[0..3], LOCAL
>IMM[0] FLT32 {    2.2000,     0.0000,     1.0000,     0.0000}
>  0: ABS TEMP[0].x, IN[2].xxxx
>  1: LG2 TEMP[0].x, TEMP[0].xxxx
>  2: ABS TEMP[1].x, IN[2].yyyy
>  3: LG2 TEMP[1].x, TEMP[1].xxxx
>  4: MOV TEMP[0].y, TEMP[1].xxxx
>  5: ABS TEMP[1].x, IN[2].zzzz
>  6: LG2 TEMP[1].x, TEMP[1].xxxx
>  7: MOV TEMP[0].z, TEMP[1].xxxx
>  8: MUL TEMP[1].xyz, TEMP[0], IMM[0].xxxx
>  9: MOV TEMP[0].z, TEMP[1].xyzx
> 10: EX2 TEMP[2].x, TEMP[1].xxxx
> 11: EX2 TEMP[3].x, TEMP[1].yyyy
> 12: MOV TEMP[2].y, TEMP[3].xxxx
> 13: EX2 TEMP[1].x, TEMP[1].zzzz
> 14: MOV TEMP[2].z, TEMP[1].xxxx
> 15: MOV TEMP[0].xy, CONST[1].xyxx
> 16: MAD TEMP[0].xy, IN[0], TEMP[0], -CONST[0]
> 17: MOV TEMP[0].xy, TEMP[0].xyxx
> 18: MOV TEMP[1].xy, IN[1].xyxx
> 19: MOV TEMP[2].w, IN[2].wwww
> 20: MOV TEMP[0].zw, IMM[0].zzyz
> 21: MOV TEMP[1].zw, IMM[0].zzyz
> 22: MOV OUT[1], TEMP[2]
> 23: MOV OUT[2], TEMP[1]
> 24: MOV OUT[0], TEMP[0]
> 25: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 0)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 4)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %12, i32 16)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %12, i32 20)
>  %17 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %18 = load <16 x i8> addrspace(2)* %17, !tbaa !0
>  %19 = add i32 %5, %7
>  %20 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %18, i32 0, i32 %19)
>  %21 = extractelement <4 x float> %20, i32 0
>  %22 = extractelement <4 x float> %20, i32 1
>  %23 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %24 = load <16 x i8> addrspace(2)* %23, !tbaa !0
>  %25 = add i32 %5, %7
>  %26 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %24, i32 0, i32 %25)
>  %27 = extractelement <4 x float> %26, i32 0
>  %28 = extractelement <4 x float> %26, i32 1
>  %29 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 2
>  %30 = load <16 x i8> addrspace(2)* %29, !tbaa !0
>  %31 = add i32 %5, %7
>  %32 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %30, i32 0, i32 %31)
>  %33 = extractelement <4 x float> %32, i32 0
>  %34 = extractelement <4 x float> %32, i32 1
>  %35 = extractelement <4 x float> %32, i32 2
>  %36 = extractelement <4 x float> %32, i32 3
>  %37 = call float @fabs(float %33)
>  %38 = call float @llvm.log2.f32(float %37)
>  %39 = call float @fabs(float %34)
>  %40 = call float @llvm.log2.f32(float %39)
>  %41 = call float @fabs(float %35)
>  %42 = call float @llvm.log2.f32(float %41)
>  %43 = fmul float %38, 0x40019999A0000000
>  %44 = fmul float %40, 0x40019999A0000000
>  %45 = fmul float %42, 0x40019999A0000000
>  %46 = call float @llvm.AMDIL.exp.(float %43)
>  %47 = call float @llvm.AMDIL.exp.(float %44)
>  %48 = call float @llvm.AMDIL.exp.(float %45)
>  %49 = fsub float -0.000000e+00, %13
>  %50 = fmul float %21, %15
>  %51 = fadd float %50, %49
>  %52 = fsub float -0.000000e+00, %14
>  %53 = fmul float %22, %16
>  %54 = fadd float %53, %52
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %46, float %47, float %48, float %36)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %27, float %28, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %51, float %54, float 0.000000e+00, float 1.000000e+00)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>; Function Attrs: readnone
>declare float @fabs(float) #2
>
>; Function Attrs: nounwind readnone
>declare float @llvm.log2.f32(float) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.exp.(float) #2
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                             ; 4A00000A
>	s_load_dwordx4 s[4:7], s[8:9], 0x0                    ; C0820900
>	s_load_dwordx4 s[12:15], s[8:9], 0x4                  ; C0860904
>	s_load_dwordx4 s[8:11], s[8:9], 0x8                   ; C0840908
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[8:11], 0 idxen  ; E00C2000 80020100
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_log_f32_e64 v5, |v3|                                ; D34E0105 00000103
>	v_mul_f32_e32 v5, 0x400ccccd, v5                      ; 100A0AFF 400CCCCD
>	v_exp_f32_e32 v5, v5                                  ; 7E0A4B05
>	v_log_f32_e64 v6, |v2|                                ; D34E0106 00000102
>	v_mul_f32_e32 v6, 0x400ccccd, v6                      ; 100C0CFF 400CCCCD
>	v_exp_f32_e32 v6, v6                                  ; 7E0C4B06
>	v_log_f32_e64 v7, |v1|                                ; D34E0107 00000101
>	v_mul_f32_e32 v7, 0x400ccccd, v7                      ; 100E0EFF 400CCCCD
>	v_exp_f32_e32 v7, v7                                  ; 7E0E4B07
>	exp 15, 32, 0, 0, 0, v7, v6, v5, v4                   ; F800020F 04050607
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	buffer_load_format_xyzw v[1:4], v0, s[12:15], 0 idxen ; E00C2000 80030100
>	v_mov_b32_e32 v5, 1.0                                 ; 7E0A02F2
>	v_mov_b32_e32 v6, 0                                   ; 7E0C0280
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	exp 15, 33, 0, 0, 0, v1, v2, v6, v5                   ; F800021F 05060201
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	buffer_load_format_xyzw v[0:3], v0, s[4:7], 0 idxen   ; E00C2000 80010000
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                    ; C0800300
>	s_waitcnt vmcnt(0) lgkmcnt(0)                         ; BF8C0070
>	s_buffer_load_dword s4, s[0:3], 0x1                   ; C2020101
>	s_buffer_load_dword s5, s[0:3], 0x5                   ; C2028105
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mov_b32_e32 v4, s4                                  ; 7E080204
>	v_mad_f32 v4, s5, v1, -v4                             ; D2820004 84120205
>	s_buffer_load_dword s4, s[0:3], 0x0                   ; C2020100
>	s_buffer_load_dword s0, s[0:3], 0x4                   ; C2000104
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mov_b32_e32 v7, s4                                  ; 7E0E0204
>	v_mad_f32 v0, s0, v0, -v7                             ; D2820000 841E0000
>	exp 15, 12, 0, 1, 0, v0, v4, v6, v5                   ; F80008CF 05060400
>	s_endpgm                                              ; BF810000
>
>SGPRS: 24
>VGPRS: 8
>Code Size: 212 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL IN[1], GENERIC[10], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL TEMP[0..3], LOCAL
>IMM[0] FLT32 {    2.2000,     1.0000,     0.4545,     0.0000}
>  0: MOV TEMP[0].xy, IN[1].xyyy
>  1: TEX TEMP[0], TEMP[0], SAMP[0], 2D
>  2: POW TEMP[1].x, TEMP[0].xxxx, IMM[0].xxxx
>  3: POW TEMP[1].y, TEMP[0].yyyy, IMM[0].xxxx
>  4: POW TEMP[1].z, TEMP[0].zzzz, IMM[0].xxxx
>  5: POW TEMP[1].w, TEMP[0].wwww, IMM[0].yyyy
>  6: MUL TEMP[0].w, TEMP[1].wwww, IN[0].wwww
>  7: MOV TEMP[0].w, TEMP[0].wwww
>  8: MUL TEMP[1].xyz, TEMP[1], IN[0]
>  9: ABS TEMP[2].x, TEMP[1].xxxx
> 10: LG2 TEMP[2].x, TEMP[2].xxxx
> 11: ABS TEMP[3].x, TEMP[1].yyyy
> 12: LG2 TEMP[3].x, TEMP[3].xxxx
> 13: MOV TEMP[2].y, TEMP[3].xxxx
> 14: ABS TEMP[1].x, TEMP[1].zzzz
> 15: LG2 TEMP[1].x, TEMP[1].xxxx
> 16: MOV TEMP[2].z, TEMP[1].xxxx
> 17: MUL TEMP[1].xyz, TEMP[2], IMM[0].zzzz
> 18: EX2 TEMP[0].x, TEMP[1].xxxx
> 19: EX2 TEMP[2].x, TEMP[1].yyyy
> 20: MOV TEMP[0].y, TEMP[2].xxxx
> 21: EX2 TEMP[1].x, TEMP[1].zzzz
> 22: MOV TEMP[0].z, TEMP[1].xxxx
> 23: MOV OUT[0], TEMP[0]
> 24: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %23 = load <8 x i32> addrspace(2)* %22, !tbaa !0
>  %24 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %25 = load <4 x i32> addrspace(2)* %24, !tbaa !0
>  %26 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %27 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %28 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %5, <2 x i32> %7)
>  %29 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %5, <2 x i32> %7)
>  %30 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %5, <2 x i32> %7)
>  %31 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %5, <2 x i32> %7)
>  %32 = bitcast float %30 to i32
>  %33 = bitcast float %31 to i32
>  %34 = insertelement <2 x i32> undef, i32 %32, i32 0
>  %35 = insertelement <2 x i32> %34, i32 %33, i32 1
>  %36 = bitcast <8 x i32> %23 to <32 x i8>
>  %37 = bitcast <4 x i32> %25 to <16 x i8>
>  %38 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %35, <32 x i8> %36, <16 x i8> %37, i32 2)
>  %39 = extractelement <4 x float> %38, i32 0
>  %40 = extractelement <4 x float> %38, i32 1
>  %41 = extractelement <4 x float> %38, i32 2
>  %42 = extractelement <4 x float> %38, i32 3
>  %43 = call float @llvm.pow.f32(float %39, float 0x40019999A0000000)
>  %44 = call float @llvm.pow.f32(float %40, float 0x40019999A0000000)
>  %45 = call float @llvm.pow.f32(float %41, float 0x40019999A0000000)
>  %46 = call float @llvm.pow.f32(float %42, float 1.000000e+00)
>  %47 = fmul float %46, %29
>  %48 = fmul float %43, %26
>  %49 = fmul float %44, %27
>  %50 = fmul float %45, %28
>  %51 = call float @fabs(float %48)
>  %52 = call float @llvm.log2.f32(float %51)
>  %53 = call float @fabs(float %49)
>  %54 = call float @llvm.log2.f32(float %53)
>  %55 = call float @fabs(float %50)
>  %56 = call float @llvm.log2.f32(float %55)
>  %57 = fmul float %52, 0x3FDD1743E0000000
>  %58 = fmul float %54, 0x3FDD1743E0000000
>  %59 = fmul float %56, 0x3FDD1743E0000000
>  %60 = call float @llvm.AMDIL.exp.(float %57)
>  %61 = call float @llvm.AMDIL.exp.(float %58)
>  %62 = call float @llvm.AMDIL.exp.(float %59)
>  %63 = call i32 @llvm.SI.packf16(float %60, float %61)
>  %64 = bitcast i32 %63 to float
>  %65 = call i32 @llvm.SI.packf16(float %62, float %47)
>  %66 = bitcast i32 %65 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %64, float %66, float %64, float %66)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.pow.f32(float, float) #1
>
>; Function Attrs: readnone
>declare float @fabs(float) #2
>
>; Function Attrs: nounwind readnone
>declare float @llvm.log2.f32(float) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.exp.(float) #2
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                  ; BEFE0A7E
>	s_mov_b32 m0, s9                                                      ; BEFC0309
>	v_interp_p1_f32 v3, v0, 1, 1, [m0]                                    ; C80C0500
>	v_interp_p2_f32 v3, [v3], v1, 1, 1, [m0]                              ; C80D0501
>	v_interp_p1_f32 v2, v0, 0, 1, [m0]                                    ; C8080400
>	v_interp_p2_f32 v2, [v2], v1, 0, 1, [m0]                              ; C8090401
>	s_load_dwordx4 s[0:3], s[4:5], 0x0                                    ; C0800500
>	s_load_dwordx8 s[4:11], s[6:7], 0x0                                   ; C0C20700
>	s_waitcnt lgkmcnt(0)                                                  ; BF8C007F
>	image_sample v[2:5], 15, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[4:11], s[0:3] ; F0800F00 00010202
>	s_waitcnt vmcnt(0)                                                    ; BF8C0770
>	v_log_f32_e32 v6, v5                                                  ; 7E0C4F05
>	v_mul_legacy_f32_e32 v6, 1.0, v6                                      ; 0E0C0CF2
>	v_exp_f32_e32 v6, v6                                                  ; 7E0C4B06
>	v_interp_p1_f32 v7, v0, 3, 0, [m0]                                    ; C81C0300
>	v_interp_p2_f32 v7, [v7], v1, 3, 0, [m0]                              ; C81D0301
>	v_mul_f32_e32 v6, v7, v6                                              ; 100C0D07
>	v_log_f32_e32 v7, v4                                                  ; 7E0E4F04
>	v_mov_b32_e32 v8, 0x400ccccd                                          ; 7E1002FF 400CCCCD
>	v_mul_legacy_f32_e32 v7, v8, v7                                       ; 0E0E0F08
>	v_exp_f32_e32 v7, v7                                                  ; 7E0E4B07
>	v_interp_p1_f32 v9, v0, 2, 0, [m0]                                    ; C8240200
>	v_interp_p2_f32 v9, [v9], v1, 2, 0, [m0]                              ; C8250201
>	v_mul_f32_e32 v7, v9, v7                                              ; 100E0F09
>	v_log_f32_e64 v7, |v7|                                                ; D34E0107 00000107
>	v_mul_f32_e32 v7, 0x3ee8ba1f, v7                                      ; 100E0EFF 3EE8BA1F
>	v_exp_f32_e32 v7, v7                                                  ; 7E0E4B07
>	v_cvt_pkrtz_f16_f32_e32 v6, v7, v6                                    ; 5E0C0D07
>	v_log_f32_e32 v7, v3                                                  ; 7E0E4F03
>	v_mul_legacy_f32_e32 v7, v8, v7                                       ; 0E0E0F08
>	v_exp_f32_e32 v7, v7                                                  ; 7E0E4B07
>	v_interp_p1_f32 v9, v0, 1, 0, [m0]                                    ; C8240100
>	v_interp_p2_f32 v9, [v9], v1, 1, 0, [m0]                              ; C8250101
>	v_mul_f32_e32 v7, v9, v7                                              ; 100E0F09
>	v_log_f32_e64 v7, |v7|                                                ; D34E0107 00000107
>	v_mul_f32_e32 v7, 0x3ee8ba1f, v7                                      ; 100E0EFF 3EE8BA1F
>	v_exp_f32_e32 v7, v7                                                  ; 7E0E4B07
>	v_log_f32_e32 v2, v2                                                  ; 7E044F02
>	v_mul_legacy_f32_e32 v2, v8, v2                                       ; 0E040508
>	v_exp_f32_e32 v2, v2                                                  ; 7E044B02
>	v_interp_p1_f32 v3, v0, 0, 0, [m0]                                    ; C80C0000
>	v_interp_p2_f32 v3, [v3], v1, 0, 0, [m0]                              ; C80D0001
>	v_mul_f32_e32 v0, v3, v2                                              ; 10000503
>	v_log_f32_e64 v0, |v0|                                                ; D34E0100 00000100
>	v_mul_f32_e32 v0, 0x3ee8ba1f, v0                                      ; 100000FF 3EE8BA1F
>	v_exp_f32_e32 v0, v0                                                  ; 7E004B00
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v7                                    ; 5E000F00
>	exp 15, 0, 1, 1, 1, v0, v6, v0, v6                                    ; F8001C0F 06000600
>	s_endpgm                                                              ; BF810000
>
>SGPRS: 16
>VGPRS: 12
>Code Size: 232 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL CONST[0..1]
>DCL TEMP[0..1], LOCAL
>IMM[0] FLT32 {    0.5000,    -0.5000,     0.0000,     1.0000}
>  0: MOV TEMP[0].xy, CONST[1].xyxx
>  1: MAD TEMP[1].xy, IN[0], TEMP[0], -CONST[0]
>  2: MOV TEMP[0].xy, TEMP[1].xyxx
>  3: MAD TEMP[0].xy, TEMP[0], IMM[0], IMM[0].xxxx
>  4: MOV TEMP[0].xy, TEMP[0].xyxx
>  5: MOV TEMP[1].xy, TEMP[1].xyxx
>  6: MOV TEMP[1].zw, IMM[0].wwzw
>  7: MOV TEMP[0].zw, IMM[0].wwzw
>  8: MOV OUT[0], TEMP[1]
>  9: MOV OUT[1], TEMP[0]
> 10: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 0)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 4)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %12, i32 16)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %12, i32 20)
>  %17 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %18 = load <16 x i8> addrspace(2)* %17, !tbaa !0
>  %19 = add i32 %5, %7
>  %20 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %18, i32 0, i32 %19)
>  %21 = extractelement <4 x float> %20, i32 0
>  %22 = extractelement <4 x float> %20, i32 1
>  %23 = fsub float -0.000000e+00, %13
>  %24 = fmul float %21, %15
>  %25 = fadd float %24, %23
>  %26 = fsub float -0.000000e+00, %14
>  %27 = fmul float %22, %16
>  %28 = fadd float %27, %26
>  %29 = fmul float %25, 5.000000e-01
>  %30 = fadd float %29, 5.000000e-01
>  %31 = fmul float %28, -5.000000e-01
>  %32 = fadd float %31, 5.000000e-01
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %30, float %32, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %25, float %28, float 0.000000e+00, float 1.000000e+00)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                           ; 4A00000A
>	s_load_dwordx4 s[4:7], s[8:9], 0x0                  ; C0820900
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	buffer_load_format_xyzw v[0:3], v0, s[4:7], 0 idxen ; E00C2000 80010000
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                  ; C0800300
>	s_waitcnt vmcnt(0) lgkmcnt(0)                       ; BF8C0070
>	s_buffer_load_dword s4, s[0:3], 0x1                 ; C2020101
>	s_buffer_load_dword s5, s[0:3], 0x5                 ; C2028105
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	v_mov_b32_e32 v4, s4                                ; 7E080204
>	v_mad_f32 v4, s5, v1, -v4                           ; D2820004 84120205
>	v_mad_f32 v5, -0.5, v4, 0.5                         ; D2820005 03C208F1
>	s_buffer_load_dword s4, s[0:3], 0x0                 ; C2020100
>	s_buffer_load_dword s0, s[0:3], 0x4                 ; C2000104
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	v_mov_b32_e32 v6, s4                                ; 7E0C0204
>	v_mad_f32 v0, s0, v0, -v6                           ; D2820000 841A0000
>	v_mad_f32 v1, 0.5, v0, 0.5                          ; D2820001 03C200F0
>	v_mov_b32_e32 v2, 1.0                               ; 7E0402F2
>	v_mov_b32_e32 v3, 0                                 ; 7E060280
>	exp 15, 32, 0, 0, 0, v1, v5, v3, v2                 ; F800020F 02030501
>	exp 15, 12, 0, 1, 0, v0, v4, v3, v2                 ; F80008CF 02030400
>	s_endpgm                                            ; BF810000
>
>SGPRS: 16
>VGPRS: 8
>Code Size: 120 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL CONST[0]
>DCL TEMP[0..2], LOCAL
>IMM[0] FLT32 {    2.2000,     1.0000,     0.4545,     0.0000}
>  0: MOV TEMP[0].xy, IN[0].xyyy
>  1: TEX TEMP[0].xyz, TEMP[0], SAMP[0], 2D
>  2: POW TEMP[1].x, TEMP[0].xxxx, IMM[0].xxxx
>  3: POW TEMP[1].y, TEMP[0].yyyy, IMM[0].xxxx
>  4: POW TEMP[1].z, TEMP[0].zzzz, IMM[0].xxxx
>  5: ABS TEMP[0].x, TEMP[1].xxxx
>  6: LG2 TEMP[0].x, TEMP[0].xxxx
>  7: ABS TEMP[2].x, TEMP[1].yyyy
>  8: LG2 TEMP[2].x, TEMP[2].xxxx
>  9: MOV TEMP[0].y, TEMP[2].xxxx
> 10: ABS TEMP[1].x, TEMP[1].zzzz
> 11: LG2 TEMP[1].x, TEMP[1].xxxx
> 12: MOV TEMP[0].z, TEMP[1].xxxx
> 13: MUL TEMP[0].xyz, TEMP[0], IMM[0].zzzz
> 14: EX2 TEMP[1].x, TEMP[0].xxxx
> 15: EX2 TEMP[2].x, TEMP[0].yyyy
> 16: MOV TEMP[1].y, TEMP[2].xxxx
> 17: EX2 TEMP[0].x, TEMP[0].zzzz
> 18: MOV TEMP[1].z, TEMP[0].xxxx
> 19: MOV TEMP[1].w, CONST[0].xxxx
> 20: MOV OUT[0], TEMP[1]
> 21: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %23 = load <16 x i8> addrspace(2)* %22, !tbaa !0
>  %24 = call float @llvm.SI.load.const(<16 x i8> %23, i32 0)
>  %25 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %26 = load <8 x i32> addrspace(2)* %25, !tbaa !0
>  %27 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %28 = load <4 x i32> addrspace(2)* %27, !tbaa !0
>  %29 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %30 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %31 = bitcast float %29 to i32
>  %32 = bitcast float %30 to i32
>  %33 = insertelement <2 x i32> undef, i32 %31, i32 0
>  %34 = insertelement <2 x i32> %33, i32 %32, i32 1
>  %35 = bitcast <8 x i32> %26 to <32 x i8>
>  %36 = bitcast <4 x i32> %28 to <16 x i8>
>  %37 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %34, <32 x i8> %35, <16 x i8> %36, i32 2)
>  %38 = extractelement <4 x float> %37, i32 0
>  %39 = extractelement <4 x float> %37, i32 1
>  %40 = extractelement <4 x float> %37, i32 2
>  %41 = call float @llvm.pow.f32(float %38, float 0x40019999A0000000)
>  %42 = call float @llvm.pow.f32(float %39, float 0x40019999A0000000)
>  %43 = call float @llvm.pow.f32(float %40, float 0x40019999A0000000)
>  %44 = call float @fabs(float %41)
>  %45 = call float @llvm.log2.f32(float %44)
>  %46 = call float @fabs(float %42)
>  %47 = call float @llvm.log2.f32(float %46)
>  %48 = call float @fabs(float %43)
>  %49 = call float @llvm.log2.f32(float %48)
>  %50 = fmul float %45, 0x3FDD1743E0000000
>  %51 = fmul float %47, 0x3FDD1743E0000000
>  %52 = fmul float %49, 0x3FDD1743E0000000
>  %53 = call float @llvm.AMDIL.exp.(float %50)
>  %54 = call float @llvm.AMDIL.exp.(float %51)
>  %55 = call float @llvm.AMDIL.exp.(float %52)
>  %56 = call i32 @llvm.SI.packf16(float %53, float %54)
>  %57 = bitcast i32 %56 to float
>  %58 = call i32 @llvm.SI.packf16(float %55, float %24)
>  %59 = bitcast i32 %58 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %57, float %59, float %57, float %59)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.pow.f32(float, float) #1
>
>; Function Attrs: readnone
>declare float @fabs(float) #2
>
>; Function Attrs: nounwind readnone
>declare float @llvm.log2.f32(float) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.exp.(float) #2
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                   ; BEFE0A7E
>	s_mov_b32 m0, s9                                                       ; BEFC0309
>	v_interp_p1_f32 v3, v0, 1, 0, [m0]                                     ; C80C0100
>	v_interp_p2_f32 v3, [v3], v1, 1, 0, [m0]                               ; C80D0101
>	v_interp_p1_f32 v2, v0, 0, 0, [m0]                                     ; C8080000
>	v_interp_p2_f32 v2, [v2], v1, 0, 0, [m0]                               ; C8090001
>	s_load_dwordx4 s[8:11], s[4:5], 0x0                                    ; C0840500
>	s_load_dwordx8 s[12:19], s[6:7], 0x0                                   ; C0C60700
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	image_sample v[0:2], 7, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[12:19], s[8:11] ; F0800700 00430002
>	s_waitcnt vmcnt(0)                                                     ; BF8C0770
>	v_log_f32_e32 v3, v1                                                   ; 7E064F01
>	v_mov_b32_e32 v4, 0x400ccccd                                           ; 7E0802FF 400CCCCD
>	v_mul_legacy_f32_e32 v3, v4, v3                                        ; 0E060704
>	v_exp_f32_e32 v3, v3                                                   ; 7E064B03
>	v_log_f32_e64 v3, |v3|                                                 ; D34E0103 00000103
>	v_mul_f32_e32 v3, 0x3ee8ba1f, v3                                       ; 100606FF 3EE8BA1F
>	v_exp_f32_e32 v3, v3                                                   ; 7E064B03
>	v_log_f32_e32 v5, v0                                                   ; 7E0A4F00
>	v_mul_legacy_f32_e32 v5, v4, v5                                        ; 0E0A0B04
>	v_exp_f32_e32 v5, v5                                                   ; 7E0A4B05
>	v_log_f32_e64 v5, |v5|                                                 ; D34E0105 00000105
>	v_mul_f32_e32 v5, 0x3ee8ba1f, v5                                       ; 100A0AFF 3EE8BA1F
>	v_exp_f32_e32 v5, v5                                                   ; 7E0A4B05
>	v_cvt_pkrtz_f16_f32_e32 v3, v5, v3                                     ; 5E060705
>	v_log_f32_e32 v0, v2                                                   ; 7E004F02
>	v_mul_legacy_f32_e32 v0, v4, v0                                        ; 0E000104
>	v_exp_f32_e32 v0, v0                                                   ; 7E004B00
>	v_log_f32_e64 v0, |v0|                                                 ; D34E0100 00000100
>	v_mul_f32_e32 v0, 0x3ee8ba1f, v0                                       ; 100000FF 3EE8BA1F
>	v_exp_f32_e32 v0, v0                                                   ; 7E004B00
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                                     ; C0800300
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	s_buffer_load_dword s0, s[0:3], 0x0                                    ; C2000100
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_cvt_pkrtz_f16_f32_e64 v0, v0, s0                                     ; D25E0000 00000100
>	exp 15, 0, 1, 1, 1, v3, v0, v3, v0                                     ; F8001C0F 00030003
>	s_endpgm                                                               ; BF810000
>
>SGPRS: 24
>VGPRS: 8
>Code Size: 192 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL CONST[0..1]
>DCL TEMP[0..1], LOCAL
>IMM[0] FLT32 {    0.0000,     1.0000,     0.0000,     0.0000}
>  0: MOV TEMP[0].xy, CONST[1].xyxx
>  1: MAD TEMP[0].xy, IN[0], TEMP[0], -CONST[0]
>  2: MOV TEMP[0].xy, TEMP[0].xyxx
>  3: MOV TEMP[1].x, IN[1].xxxx
>  4: MOV TEMP[0].zw, IMM[0].yyxy
>  5: MOV TEMP[1].zw, IMM[0].yyxy
>  6: MOV OUT[1], TEMP[1]
>  7: MOV OUT[0], TEMP[0]
>  8: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 0)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 4)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %12, i32 16)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %12, i32 20)
>  %17 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %18 = load <16 x i8> addrspace(2)* %17, !tbaa !0
>  %19 = add i32 %5, %7
>  %20 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %18, i32 0, i32 %19)
>  %21 = extractelement <4 x float> %20, i32 0
>  %22 = extractelement <4 x float> %20, i32 1
>  %23 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %24 = load <16 x i8> addrspace(2)* %23, !tbaa !0
>  %25 = add i32 %5, %7
>  %26 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %24, i32 0, i32 %25)
>  %27 = extractelement <4 x float> %26, i32 0
>  %28 = fsub float -0.000000e+00, %13
>  %29 = fmul float %21, %15
>  %30 = fadd float %29, %28
>  %31 = fsub float -0.000000e+00, %14
>  %32 = fmul float %22, %16
>  %33 = fadd float %32, %31
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %27, float 0.000000e+00, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %30, float %33, float 0.000000e+00, float 1.000000e+00)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                            ; 4A00000A
>	s_load_dwordx4 s[4:7], s[8:9], 0x0                   ; C0820900
>	s_load_dwordx4 s[8:11], s[8:9], 0x4                  ; C0840904
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[8:11], 0 idxen ; E00C2000 80020100
>	v_mov_b32_e32 v5, 1.0                                ; 7E0A02F2
>	v_mov_b32_e32 v6, 0                                  ; 7E0C0280
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	exp 15, 32, 0, 0, 0, v1, v6, v6, v5                  ; F800020F 05060601
>	s_waitcnt expcnt(0)                                  ; BF8C070F
>	buffer_load_format_xyzw v[0:3], v0, s[4:7], 0 idxen  ; E00C2000 80010000
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                   ; C0800300
>	s_waitcnt vmcnt(0) lgkmcnt(0)                        ; BF8C0070
>	s_buffer_load_dword s4, s[0:3], 0x1                  ; C2020101
>	s_buffer_load_dword s5, s[0:3], 0x5                  ; C2028105
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mov_b32_e32 v4, s4                                 ; 7E080204
>	v_mad_f32 v4, s5, v1, -v4                            ; D2820004 84120205
>	s_buffer_load_dword s4, s[0:3], 0x0                  ; C2020100
>	s_buffer_load_dword s0, s[0:3], 0x4                  ; C2000104
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mov_b32_e32 v7, s4                                 ; 7E0E0204
>	v_mad_f32 v0, s0, v0, -v7                            ; D2820000 841E0000
>	exp 15, 12, 0, 1, 0, v0, v4, v6, v5                  ; F80008CF 05060400
>	s_endpgm                                             ; BF810000
>
>SGPRS: 16
>VGPRS: 8
>Code Size: 124 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL TEMP[0], LOCAL
>IMM[0] FLT32 {    0.0000,     0.0000,     0.0000,     0.0000}
>  0: MAX TEMP[0].x, IN[0].xxxx, IMM[0].xxxx
>  1: SQRT TEMP[0].x, TEMP[0].xxxx
>  2: MOV TEMP[0].w, TEMP[0].xxxx
>  3: MOV TEMP[0].xyz, IN[0].xxxx
>  4: MOV OUT[0], TEMP[0]
>  5: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %23 = call float @llvm.maxnum.f32(float %22, float 0x3E7AD7F2A0000000)
>  %24 = call float @llvm.sqrt.f32(float %23)
>  %25 = call i32 @llvm.SI.packf16(float %22, float %22)
>  %26 = bitcast i32 %25 to float
>  %27 = call i32 @llvm.SI.packf16(float %22, float %24)
>  %28 = bitcast i32 %27 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %26, float %28, float %26, float %28)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.maxnum.f32(float, float) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.sqrt.f32(float) #1
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                     ; BEFE0A7E
>	s_mov_b32 m0, s9                         ; BEFC0309
>	v_interp_p1_f32 v2, v0, 0, 0, [m0]       ; C8080000
>	v_interp_p2_f32 v2, [v2], v1, 0, 0, [m0] ; C8090001
>	v_max_f32_e32 v0, 0x33d6bf95, v2         ; 200004FF 33D6BF95
>	v_sqrt_f32_e32 v0, v0                    ; 7E006700
>	v_cvt_pkrtz_f16_f32_e32 v0, v2, v0       ; 5E000102
>	v_cvt_pkrtz_f16_f32_e32 v1, v2, v2       ; 5E020502
>	exp 15, 0, 1, 1, 1, v1, v0, v1, v0       ; F8001C0F 00010001
>	s_endpgm                                 ; BF810000
>
>SGPRS: 16
>VGPRS: 4
>Code Size: 48 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL IN[2]
>DCL IN[3]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL CONST[0..99]
>DCL TEMP[0..4], LOCAL
>DCL ADDR[0]
>IMM[0] FLT32 {    3.0000,     1.0000,     0.0000,     0.0000}
>IMM[1] INT32 {1, 2, 0, 0}
>  0: MUL TEMP[0].x, IMM[0].xxxx, IN[2].xxxx
>  1: MAD TEMP[1], IN[0].xyzx, IMM[0].yyyz, IMM[0].zzzy
>  2: F2I TEMP[2].x, TEMP[0].xxxx
>  3: UARL ADDR[0].x, TEMP[2].xxxx
>  4: UARL ADDR[0].x, TEMP[2].xxxx
>  5: DP4 TEMP[2].x, TEMP[1], CONST[ADDR[0].x+4]
>  6: F2I TEMP[3].x, TEMP[0].xxxx
>  7: UADD TEMP[3].x, IMM[1].xxxx, TEMP[3].xxxx
>  8: UARL ADDR[0].x, TEMP[3].xxxx
>  9: DP4 TEMP[3].x, TEMP[1], CONST[ADDR[0].x+4]
> 10: MOV TEMP[2].y, TEMP[3].xxxx
> 11: F2I TEMP[0].x, TEMP[0].xxxx
> 12: UADD TEMP[0].x, IMM[1].yyyy, TEMP[0].xxxx
> 13: UARL ADDR[0].x, TEMP[0].xxxx
> 14: DP4 TEMP[0].x, TEMP[1], CONST[ADDR[0].x+4]
> 15: MOV TEMP[2].z, TEMP[0].xxxx
> 16: MUL TEMP[0].xyz, TEMP[2], IN[1].xxxx
> 17: MOV TEMP[2].xyz, TEMP[0].xyzx
> 18: FSLT TEMP[0].x, IMM[0].zzzz, IN[1].yyyy
> 19: UIF TEMP[0].xxxx :0
> 20:   MUL TEMP[0].w, IMM[0].xxxx, IN[2].yyyy
> 21:   MOV TEMP[2].w, TEMP[0].wwww
> 22:   F2I TEMP[3].x, TEMP[0].wwww
> 23:   UARL ADDR[0].x, TEMP[3].xxxx
> 24:   UARL ADDR[0].x, TEMP[3].xxxx
> 25:   DP4 TEMP[3].x, TEMP[1], CONST[ADDR[0].x+4]
> 26:   F2I TEMP[4].x, TEMP[0].wwww
> 27:   UADD TEMP[4].x, IMM[1].xxxx, TEMP[4].xxxx
> 28:   UARL ADDR[0].x, TEMP[4].xxxx
> 29:   DP4 TEMP[4].x, TEMP[1], CONST[ADDR[0].x+4]
> 30:   MOV TEMP[3].y, TEMP[4].xxxx
> 31:   F2I TEMP[0].x, TEMP[0].wwww
> 32:   UADD TEMP[0].x, IMM[1].yyyy, TEMP[0].xxxx
> 33:   UARL ADDR[0].x, TEMP[0].xxxx
> 34:   DP4 TEMP[0].x, TEMP[1], CONST[ADDR[0].x+4]
> 35:   MOV TEMP[3].z, TEMP[0].xxxx
> 36:   MAD TEMP[0].xyz, IN[1].yyyy, TEMP[3], TEMP[2]
> 37:   MOV TEMP[2].xyz, TEMP[0].xyzx
> 38:   FSLT TEMP[0].x, IMM[0].zzzz, IN[1].zzzz
> 39:   UIF TEMP[0].xxxx :0
> 40:     MUL TEMP[0].w, IMM[0].xxxx, IN[2].zzzz
> 41:     MOV TEMP[2].w, TEMP[0].wwww
> 42:     F2I TEMP[4].x, TEMP[0].wwww
> 43:     UARL ADDR[0].x, TEMP[4].xxxx
> 44:     UARL ADDR[0].x, TEMP[4].xxxx
> 45:     DP4 TEMP[3].x, TEMP[1], CONST[ADDR[0].x+4]
> 46:     F2I TEMP[4].x, TEMP[0].wwww
> 47:     UADD TEMP[4].x, IMM[1].xxxx, TEMP[4].xxxx
> 48:     UARL ADDR[0].x, TEMP[4].xxxx
> 49:     DP4 TEMP[4].x, TEMP[1], CONST[ADDR[0].x+4]
> 50:     MOV TEMP[3].y, TEMP[4].xxxx
> 51:     F2I TEMP[0].x, TEMP[0].wwww
> 52:     UADD TEMP[0].x, IMM[1].yyyy, TEMP[0].xxxx
> 53:     UARL ADDR[0].x, TEMP[0].xxxx
> 54:     DP4 TEMP[0].x, TEMP[1], CONST[ADDR[0].x+4]
> 55:     MOV TEMP[3].z, TEMP[0].xxxx
> 56:     MAD TEMP[0].xyz, IN[1].zzzz, TEMP[3], TEMP[2]
> 57:     MOV TEMP[2].xyz, TEMP[0].xyzx
> 58:   ENDIF
> 59: ENDIF
> 60: MUL TEMP[1], TEMP[2].yyyy, CONST[1]
> 61: MAD TEMP[1], TEMP[2].xxxx, CONST[0], TEMP[1]
> 62: MAD TEMP[1], TEMP[2].zzzz, CONST[2], TEMP[1]
> 63: ADD TEMP[0], TEMP[1], CONST[3]
> 64: MOV TEMP[2].xy, IN[3].xyxx
> 65: MOV TEMP[1].xyz, TEMP[2].xyzx
> 66: MOV TEMP[1].w, IMM[0].yyyy
> 67: MOV OUT[1], TEMP[1]
> 68: MOV OUT[0], TEMP[0]
> 69: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 0)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 4)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %12, i32 8)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %12, i32 12)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %12, i32 16)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %12, i32 20)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %12, i32 24)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %12, i32 28)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %12, i32 32)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %12, i32 36)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %12, i32 40)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %12, i32 44)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %12, i32 48)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %12, i32 52)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %12, i32 56)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %12, i32 60)
>  %29 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %30 = load <16 x i8> addrspace(2)* %29, !tbaa !0
>  %31 = add i32 %5, %7
>  %32 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %30, i32 0, i32 %31)
>  %33 = extractelement <4 x float> %32, i32 0
>  %34 = extractelement <4 x float> %32, i32 1
>  %35 = extractelement <4 x float> %32, i32 2
>  %36 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %37 = load <16 x i8> addrspace(2)* %36, !tbaa !0
>  %38 = add i32 %5, %7
>  %39 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %37, i32 0, i32 %38)
>  %40 = extractelement <4 x float> %39, i32 0
>  %41 = extractelement <4 x float> %39, i32 1
>  %42 = extractelement <4 x float> %39, i32 2
>  %43 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 2
>  %44 = load <16 x i8> addrspace(2)* %43, !tbaa !0
>  %45 = add i32 %5, %7
>  %46 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %44, i32 0, i32 %45)
>  %47 = extractelement <4 x float> %46, i32 0
>  %48 = extractelement <4 x float> %46, i32 1
>  %49 = extractelement <4 x float> %46, i32 2
>  %50 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 3
>  %51 = load <16 x i8> addrspace(2)* %50, !tbaa !0
>  %52 = add i32 %5, %7
>  %53 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %51, i32 0, i32 %52)
>  %54 = extractelement <4 x float> %53, i32 0
>  %55 = extractelement <4 x float> %53, i32 1
>  %56 = fmul float 3.000000e+00, %47
>  %57 = fmul float %33, 1.000000e+00
>  %58 = fadd float %57, 0.000000e+00
>  %59 = fmul float %34, 1.000000e+00
>  %60 = fadd float %59, 0.000000e+00
>  %61 = fmul float %35, 1.000000e+00
>  %62 = fadd float %61, 0.000000e+00
>  %63 = fmul float %33, 0.000000e+00
>  %64 = fadd float %63, 1.000000e+00
>  %65 = fptosi float %56 to i32
>  %66 = bitcast i32 %65 to float
>  %67 = bitcast float %66 to i32
>  %68 = shl i32 %67, 4
>  %69 = add i32 %68, 64
>  %70 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %69)
>  %71 = shl i32 %67, 4
>  %72 = add i32 %71, 68
>  %73 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %72)
>  %74 = shl i32 %67, 4
>  %75 = add i32 %74, 72
>  %76 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %75)
>  %77 = shl i32 %67, 4
>  %78 = add i32 %77, 76
>  %79 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %78)
>  %80 = fmul float %58, %70
>  %81 = fmul float %60, %73
>  %82 = fadd float %80, %81
>  %83 = fmul float %62, %76
>  %84 = fadd float %82, %83
>  %85 = fmul float %64, %79
>  %86 = fadd float %84, %85
>  %87 = fptosi float %56 to i32
>  %88 = bitcast i32 %87 to float
>  %89 = bitcast float %88 to i32
>  %90 = add i32 1, %89
>  %91 = bitcast i32 %90 to float
>  %92 = bitcast float %91 to i32
>  %93 = shl i32 %92, 4
>  %94 = add i32 %93, 64
>  %95 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %94)
>  %96 = shl i32 %92, 4
>  %97 = add i32 %96, 68
>  %98 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %97)
>  %99 = shl i32 %92, 4
>  %100 = add i32 %99, 72
>  %101 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %100)
>  %102 = shl i32 %92, 4
>  %103 = add i32 %102, 76
>  %104 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %103)
>  %105 = fmul float %58, %95
>  %106 = fmul float %60, %98
>  %107 = fadd float %105, %106
>  %108 = fmul float %62, %101
>  %109 = fadd float %107, %108
>  %110 = fmul float %64, %104
>  %111 = fadd float %109, %110
>  %112 = fptosi float %56 to i32
>  %113 = bitcast i32 %112 to float
>  %114 = bitcast float %113 to i32
>  %115 = add i32 2, %114
>  %116 = bitcast i32 %115 to float
>  %117 = bitcast float %116 to i32
>  %118 = shl i32 %117, 4
>  %119 = add i32 %118, 64
>  %120 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %119)
>  %121 = shl i32 %117, 4
>  %122 = add i32 %121, 68
>  %123 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %122)
>  %124 = shl i32 %117, 4
>  %125 = add i32 %124, 72
>  %126 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %125)
>  %127 = shl i32 %117, 4
>  %128 = add i32 %127, 76
>  %129 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %128)
>  %130 = fmul float %58, %120
>  %131 = fmul float %60, %123
>  %132 = fadd float %130, %131
>  %133 = fmul float %62, %126
>  %134 = fadd float %132, %133
>  %135 = fmul float %64, %129
>  %136 = fadd float %134, %135
>  %137 = fmul float %86, %40
>  %138 = fmul float %111, %40
>  %139 = fmul float %136, %40
>  %140 = fcmp olt float 0.000000e+00, %41
>  %141 = sext i1 %140 to i32
>  %142 = bitcast i32 %141 to float
>  %143 = bitcast float %142 to i32
>  %144 = icmp ne i32 %143, 0
>  br i1 %144, label %IF, label %ENDIF
>
>IF:                                               ; preds = %main_body
>  %145 = fmul float 3.000000e+00, %48
>  %146 = fptosi float %145 to i32
>  %147 = bitcast i32 %146 to float
>  %148 = bitcast float %147 to i32
>  %149 = shl i32 %148, 4
>  %150 = add i32 %149, 64
>  %151 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %150)
>  %152 = shl i32 %148, 4
>  %153 = add i32 %152, 68
>  %154 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %153)
>  %155 = shl i32 %148, 4
>  %156 = add i32 %155, 72
>  %157 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %156)
>  %158 = shl i32 %148, 4
>  %159 = add i32 %158, 76
>  %160 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %159)
>  %161 = fmul float %58, %151
>  %162 = fmul float %60, %154
>  %163 = fadd float %161, %162
>  %164 = fmul float %62, %157
>  %165 = fadd float %163, %164
>  %166 = fmul float %64, %160
>  %167 = fadd float %165, %166
>  %168 = fptosi float %145 to i32
>  %169 = bitcast i32 %168 to float
>  %170 = bitcast float %169 to i32
>  %171 = add i32 1, %170
>  %172 = bitcast i32 %171 to float
>  %173 = bitcast float %172 to i32
>  %174 = shl i32 %173, 4
>  %175 = add i32 %174, 64
>  %176 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %175)
>  %177 = shl i32 %173, 4
>  %178 = add i32 %177, 68
>  %179 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %178)
>  %180 = shl i32 %173, 4
>  %181 = add i32 %180, 72
>  %182 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %181)
>  %183 = shl i32 %173, 4
>  %184 = add i32 %183, 76
>  %185 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %184)
>  %186 = fmul float %58, %176
>  %187 = fmul float %60, %179
>  %188 = fadd float %186, %187
>  %189 = fmul float %62, %182
>  %190 = fadd float %188, %189
>  %191 = fmul float %64, %185
>  %192 = fadd float %190, %191
>  %193 = fptosi float %145 to i32
>  %194 = bitcast i32 %193 to float
>  %195 = bitcast float %194 to i32
>  %196 = add i32 2, %195
>  %197 = bitcast i32 %196 to float
>  %198 = bitcast float %197 to i32
>  %199 = shl i32 %198, 4
>  %200 = add i32 %199, 64
>  %201 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %200)
>  %202 = shl i32 %198, 4
>  %203 = add i32 %202, 68
>  %204 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %203)
>  %205 = shl i32 %198, 4
>  %206 = add i32 %205, 72
>  %207 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %206)
>  %208 = shl i32 %198, 4
>  %209 = add i32 %208, 76
>  %210 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %209)
>  %211 = fmul float %58, %201
>  %212 = fmul float %60, %204
>  %213 = fadd float %211, %212
>  %214 = fmul float %62, %207
>  %215 = fadd float %213, %214
>  %216 = fmul float %64, %210
>  %217 = fadd float %215, %216
>  %218 = fmul float %41, %167
>  %219 = fadd float %218, %137
>  %220 = fmul float %41, %192
>  %221 = fadd float %220, %138
>  %222 = fmul float %41, %217
>  %223 = fadd float %222, %139
>  %224 = fcmp olt float 0.000000e+00, %42
>  %225 = sext i1 %224 to i32
>  %226 = bitcast i32 %225 to float
>  %227 = bitcast float %226 to i32
>  %228 = icmp ne i32 %227, 0
>  br i1 %228, label %IF44, label %ENDIF
>
>ENDIF:                                            ; preds = %IF44, %IF, %main_body
>  %temp8.0 = phi float [ %137, %main_body ], [ %327, %IF44 ], [ %219, %IF ]
>  %temp9.0 = phi float [ %138, %main_body ], [ %329, %IF44 ], [ %221, %IF ]
>  %temp10.0 = phi float [ %139, %main_body ], [ %331, %IF44 ], [ %223, %IF ]
>  %229 = fmul float %temp9.0, %17
>  %230 = fmul float %temp9.0, %18
>  %231 = fmul float %temp9.0, %19
>  %232 = fmul float %temp9.0, %20
>  %233 = fmul float %temp8.0, %13
>  %234 = fadd float %233, %229
>  %235 = fmul float %temp8.0, %14
>  %236 = fadd float %235, %230
>  %237 = fmul float %temp8.0, %15
>  %238 = fadd float %237, %231
>  %239 = fmul float %temp8.0, %16
>  %240 = fadd float %239, %232
>  %241 = fmul float %temp10.0, %21
>  %242 = fadd float %241, %234
>  %243 = fmul float %temp10.0, %22
>  %244 = fadd float %243, %236
>  %245 = fmul float %temp10.0, %23
>  %246 = fadd float %245, %238
>  %247 = fmul float %temp10.0, %24
>  %248 = fadd float %247, %240
>  %249 = fadd float %242, %25
>  %250 = fadd float %244, %26
>  %251 = fadd float %246, %27
>  %252 = fadd float %248, %28
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %54, float %55, float %temp10.0, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %249, float %250, float %251, float %252)
>  ret void
>
>IF44:                                             ; preds = %IF
>  %253 = fmul float 3.000000e+00, %49
>  %254 = fptosi float %253 to i32
>  %255 = bitcast i32 %254 to float
>  %256 = bitcast float %255 to i32
>  %257 = shl i32 %256, 4
>  %258 = add i32 %257, 64
>  %259 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %258)
>  %260 = shl i32 %256, 4
>  %261 = add i32 %260, 68
>  %262 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %261)
>  %263 = shl i32 %256, 4
>  %264 = add i32 %263, 72
>  %265 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %264)
>  %266 = shl i32 %256, 4
>  %267 = add i32 %266, 76
>  %268 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %267)
>  %269 = fmul float %58, %259
>  %270 = fmul float %60, %262
>  %271 = fadd float %269, %270
>  %272 = fmul float %62, %265
>  %273 = fadd float %271, %272
>  %274 = fmul float %64, %268
>  %275 = fadd float %273, %274
>  %276 = fptosi float %253 to i32
>  %277 = bitcast i32 %276 to float
>  %278 = bitcast float %277 to i32
>  %279 = add i32 1, %278
>  %280 = bitcast i32 %279 to float
>  %281 = bitcast float %280 to i32
>  %282 = shl i32 %281, 4
>  %283 = add i32 %282, 64
>  %284 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %283)
>  %285 = shl i32 %281, 4
>  %286 = add i32 %285, 68
>  %287 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %286)
>  %288 = shl i32 %281, 4
>  %289 = add i32 %288, 72
>  %290 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %289)
>  %291 = shl i32 %281, 4
>  %292 = add i32 %291, 76
>  %293 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %292)
>  %294 = fmul float %58, %284
>  %295 = fmul float %60, %287
>  %296 = fadd float %294, %295
>  %297 = fmul float %62, %290
>  %298 = fadd float %296, %297
>  %299 = fmul float %64, %293
>  %300 = fadd float %298, %299
>  %301 = fptosi float %253 to i32
>  %302 = bitcast i32 %301 to float
>  %303 = bitcast float %302 to i32
>  %304 = add i32 2, %303
>  %305 = bitcast i32 %304 to float
>  %306 = bitcast float %305 to i32
>  %307 = shl i32 %306, 4
>  %308 = add i32 %307, 64
>  %309 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %308)
>  %310 = shl i32 %306, 4
>  %311 = add i32 %310, 68
>  %312 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %311)
>  %313 = shl i32 %306, 4
>  %314 = add i32 %313, 72
>  %315 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %314)
>  %316 = shl i32 %306, 4
>  %317 = add i32 %316, 76
>  %318 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %317)
>  %319 = fmul float %58, %309
>  %320 = fmul float %60, %312
>  %321 = fadd float %319, %320
>  %322 = fmul float %62, %315
>  %323 = fadd float %321, %322
>  %324 = fmul float %64, %318
>  %325 = fadd float %323, %324
>  %326 = fmul float %42, %275
>  %327 = fadd float %326, %219
>  %328 = fmul float %42, %300
>  %329 = fadd float %328, %221
>  %330 = fmul float %42, %325
>  %331 = fadd float %330, %223
>  br label %ENDIF
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v11, s10, v0                               ; 4A16000A
>	s_load_dwordx4 s[12:15], s[8:9], 0x0                     ; C0860900
>	s_load_dwordx4 s[16:19], s[8:9], 0x4                     ; C0880904
>	s_load_dwordx4 s[20:23], s[8:9], 0x8                     ; C08A0908
>	s_load_dwordx4 s[4:7], s[8:9], 0xc                       ; C082090C
>	s_waitcnt lgkmcnt(0)                                     ; BF8C007F
>	buffer_load_format_xyzw v[6:9], v11, s[12:15], 0 idxen   ; E00C2000 8003060B
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_add_f32_e32 v2, 0, v7                                  ; 06040E80
>	buffer_load_format_xyzw v[15:18], v11, s[20:23], 0 idxen ; E00C2000 80050F0B
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v0, 0x40400000, v15                        ; 10001EFF 40400000
>	v_cvt_i32_f32_e32 v0, v0                                 ; 7E001100
>	v_lshlrev_b32_e32 v3, 4, v0                              ; 34060084
>	v_add_i32_e32 v0, 0x64, v3                               ; 4A0006FF 00000064
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                       ; C0800300
>	s_waitcnt lgkmcnt(0)                                     ; BF8C007F
>	buffer_load_dword v0, v0, s[0:3], 0 offen                ; E0301000 80000000
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v0, v0, v2                                 ; 10000500
>	v_add_f32_e32 v4, 0, v6                                  ; 06080C80
>	v_add_i32_e32 v1, 0x60, v3                               ; 4A0206FF 00000060
>	buffer_load_dword v1, v1, s[0:3], 0 offen                ; E0301000 80000101
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v0, v4, v1, v0                                 ; D2820000 04020304
>	v_add_f32_e32 v5, 0, v8                                  ; 060A1080
>	v_add_i32_e32 v1, 0x68, v3                               ; 4A0206FF 00000068
>	buffer_load_dword v1, v1, s[0:3], 0 offen                ; E0301000 80000101
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v0, v5, v1, v0                                 ; D2820000 04020305
>	v_mad_f32 v6, 0, v6, 1.0                                 ; D2820006 03CA0C80
>	v_add_i32_e32 v1, 0x6c, v3                               ; 4A0206FF 0000006C
>	buffer_load_dword v1, v1, s[0:3], 0 offen                ; E0301000 80000101
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v0, v6, v1, v0                                 ; D2820000 04020306
>	buffer_load_format_xyzw v[7:10], v11, s[16:19], 0 idxen  ; E00C2000 8004070B
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v0, v7, v0                                 ; 10000107
>	v_add_i32_e32 v1, 0x54, v3                               ; 4A0206FF 00000054
>	buffer_load_dword v1, v1, s[0:3], 0 offen                ; E0301000 80000101
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v1, v1, v2                                 ; 10020501
>	v_add_i32_e32 v12, 0x50, v3                              ; 4A1806FF 00000050
>	buffer_load_dword v12, v12, s[0:3], 0 offen              ; E0301000 80000C0C
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v1, v4, v12, v1                                ; D2820001 04061904
>	v_add_i32_e32 v12, 0x58, v3                              ; 4A1806FF 00000058
>	buffer_load_dword v12, v12, s[0:3], 0 offen              ; E0301000 80000C0C
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v1, v5, v12, v1                                ; D2820001 04061905
>	v_add_i32_e32 v12, 0x5c, v3                              ; 4A1806FF 0000005C
>	buffer_load_dword v12, v12, s[0:3], 0 offen              ; E0301000 80000C0C
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v1, v6, v12, v1                                ; D2820001 04061906
>	v_mul_f32_e32 v1, v7, v1                                 ; 10020307
>	v_add_i32_e32 v12, 0x44, v3                              ; 4A1806FF 00000044
>	buffer_load_dword v12, v12, s[0:3], 0 offen              ; E0301000 80000C0C
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v12, v12, v2                               ; 1018050C
>	v_add_i32_e32 v13, 64, v3                                ; 4A1A06C0
>	buffer_load_dword v13, v13, s[0:3], 0 offen              ; E0301000 80000D0D
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v12, v4, v13, v12                              ; D282000C 04321B04
>	v_add_i32_e32 v13, 0x48, v3                              ; 4A1A06FF 00000048
>	buffer_load_dword v13, v13, s[0:3], 0 offen              ; E0301000 80000D0D
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v12, v5, v13, v12                              ; D282000C 04321B05
>	v_add_i32_e32 v3, 0x4c, v3                               ; 4A0606FF 0000004C
>	buffer_load_dword v3, v3, s[0:3], 0 offen                ; E0301000 80000303
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v3, v6, v3, v12                                ; D2820003 04320706
>	v_mul_f32_e32 v3, v7, v3                                 ; 10060707
>	buffer_load_format_xyzw v[11:14], v11, s[4:7], 0 idxen   ; E00C2000 80010B0B
>	v_cmp_gt_f32_e64 s[4:5], v8, 0                           ; D0080004 00010108
>	v_cndmask_b32_e64 v19, 0, -1, s[4:5]                     ; D2000813 00118280
>	v_cmp_ne_i32_e64 s[20:21], v19, 0                        ; D10A0014 00010113
>	s_buffer_load_dword s9, s[0:3], 0xf                      ; C204810F
>	s_buffer_load_dword s7, s[0:3], 0xe                      ; C203810E
>	s_buffer_load_dword s5, s[0:3], 0xd                      ; C202810D
>	s_buffer_load_dword s4, s[0:3], 0xc                      ; C202010C
>	s_buffer_load_dword s12, s[0:3], 0xb                     ; C206010B
>	s_buffer_load_dword s10, s[0:3], 0xa                     ; C205010A
>	s_buffer_load_dword s8, s[0:3], 0x9                      ; C2040109
>	s_buffer_load_dword s6, s[0:3], 0x8                      ; C2030108
>	s_buffer_load_dword s17, s[0:3], 0x7                     ; C2088107
>	s_buffer_load_dword s15, s[0:3], 0x6                     ; C2078106
>	s_buffer_load_dword s13, s[0:3], 0x5                     ; C2068105
>	s_buffer_load_dword s11, s[0:3], 0x4                     ; C2058104
>	s_buffer_load_dword s19, s[0:3], 0x3                     ; C2098103
>	s_buffer_load_dword s18, s[0:3], 0x2                     ; C2090102
>	s_buffer_load_dword s16, s[0:3], 0x1                     ; C2080101
>	s_buffer_load_dword s14, s[0:3], 0x0                     ; C2070100
>	s_waitcnt vmcnt(0) lgkmcnt(0)                            ; BF8C0070
>	s_and_saveexec_b64 s[20:21], s[20:21]                    ; BE942414
>	s_xor_b64 s[20:21], exec, s[20:21]                       ; 8994147E
>	s_cbranch_execz BB0_3                                    ; BF880000
>	v_mul_f32_e32 v19, 0x40400000, v16                       ; 102620FF 40400000
>	v_cvt_i32_f32_e32 v19, v19                               ; 7E261113
>	v_lshlrev_b32_e32 v19, 4, v19                            ; 34262684
>	v_add_i32_e32 v20, 0x60, v19                             ; 4A2826FF 00000060
>	buffer_load_dword v20, v20, s[0:3], 0 offen              ; E0301000 80001414
>	v_add_i32_e32 v21, 0x64, v19                             ; 4A2A26FF 00000064
>	buffer_load_dword v21, v21, s[0:3], 0 offen              ; E0301000 80001515
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v21, v21, v2                               ; 102A0515
>	v_mad_f32 v20, v4, v20, v21                              ; D2820014 04562904
>	v_add_i32_e32 v21, 0x68, v19                             ; 4A2A26FF 00000068
>	buffer_load_dword v21, v21, s[0:3], 0 offen              ; E0301000 80001515
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v20, v5, v21, v20                              ; D2820014 04522B05
>	v_add_i32_e32 v21, 0x6c, v19                             ; 4A2A26FF 0000006C
>	buffer_load_dword v21, v21, s[0:3], 0 offen              ; E0301000 80001515
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v20, v6, v21, v20                              ; D2820014 04522B06
>	v_mad_f32 v0, v8, v20, v0                                ; D2820000 04022908
>	v_add_i32_e32 v20, 0x50, v19                             ; 4A2826FF 00000050
>	buffer_load_dword v20, v20, s[0:3], 0 offen              ; E0301000 80001414
>	v_add_i32_e32 v21, 0x54, v19                             ; 4A2A26FF 00000054
>	buffer_load_dword v21, v21, s[0:3], 0 offen              ; E0301000 80001515
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v21, v21, v2                               ; 102A0515
>	v_mad_f32 v20, v4, v20, v21                              ; D2820014 04562904
>	v_add_i32_e32 v21, 0x58, v19                             ; 4A2A26FF 00000058
>	buffer_load_dword v21, v21, s[0:3], 0 offen              ; E0301000 80001515
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v20, v5, v21, v20                              ; D2820014 04522B05
>	v_add_i32_e32 v21, 0x5c, v19                             ; 4A2A26FF 0000005C
>	buffer_load_dword v21, v21, s[0:3], 0 offen              ; E0301000 80001515
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v20, v6, v21, v20                              ; D2820014 04522B06
>	v_mad_f32 v1, v8, v20, v1                                ; D2820001 04062908
>	v_add_i32_e32 v20, 64, v19                               ; 4A2826C0
>	buffer_load_dword v20, v20, s[0:3], 0 offen              ; E0301000 80001414
>	v_add_i32_e32 v21, 0x44, v19                             ; 4A2A26FF 00000044
>	buffer_load_dword v21, v21, s[0:3], 0 offen              ; E0301000 80001515
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v21, v21, v2                               ; 102A0515
>	v_mad_f32 v20, v4, v20, v21                              ; D2820014 04562904
>	v_add_i32_e32 v21, 0x48, v19                             ; 4A2A26FF 00000048
>	buffer_load_dword v21, v21, s[0:3], 0 offen              ; E0301000 80001515
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v20, v5, v21, v20                              ; D2820014 04522B05
>	v_add_i32_e32 v19, 0x4c, v19                             ; 4A2626FF 0000004C
>	buffer_load_dword v19, v19, s[0:3], 0 offen              ; E0301000 80001313
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v19, v6, v19, v20                              ; D2820013 04522706
>	v_mad_f32 v3, v8, v19, v3                                ; D2820003 040E2708
>	v_cmp_gt_f32_e64 s[22:23], v9, 0                         ; D0080016 00010109
>	v_cndmask_b32_e64 v19, 0, -1, s[22:23]                   ; D2000813 00598280
>	v_cmp_ne_i32_e64 s[22:23], v19, 0                        ; D10A0016 00010113
>	s_and_saveexec_b64 s[22:23], s[22:23]                    ; BE962416
>	s_xor_b64 s[22:23], exec, s[22:23]                       ; 8996167E
>	s_cbranch_execz BB0_2                                    ; BF880000
>	v_mul_f32_e32 v15, 0x40400000, v17                       ; 101E22FF 40400000
>	v_cvt_i32_f32_e32 v15, v15                               ; 7E1E110F
>	v_lshlrev_b32_e32 v15, 4, v15                            ; 341E1E84
>	v_add_i32_e32 v16, 0x60, v15                             ; 4A201EFF 00000060
>	buffer_load_dword v16, v16, s[0:3], 0 offen              ; E0301000 80001010
>	v_add_i32_e32 v17, 0x64, v15                             ; 4A221EFF 00000064
>	buffer_load_dword v17, v17, s[0:3], 0 offen              ; E0301000 80001111
>	s_waitcnt vmcnt(0) expcnt(0)                             ; BF8C0700
>	v_mul_f32_e32 v17, v17, v2                               ; 10220511
>	v_mad_f32 v16, v4, v16, v17                              ; D2820010 04462104
>	v_add_i32_e32 v17, 0x68, v15                             ; 4A221EFF 00000068
>	buffer_load_dword v17, v17, s[0:3], 0 offen              ; E0301000 80001111
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v16, v5, v17, v16                              ; D2820010 04422305
>	v_add_i32_e32 v17, 0x6c, v15                             ; 4A221EFF 0000006C
>	buffer_load_dword v17, v17, s[0:3], 0 offen              ; E0301000 80001111
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v16, v6, v17, v16                              ; D2820010 04422306
>	v_mad_f32 v0, v9, v16, v0                                ; D2820000 04022109
>	v_add_i32_e32 v16, 0x50, v15                             ; 4A201EFF 00000050
>	buffer_load_dword v16, v16, s[0:3], 0 offen              ; E0301000 80001010
>	v_add_i32_e32 v17, 0x54, v15                             ; 4A221EFF 00000054
>	buffer_load_dword v17, v17, s[0:3], 0 offen              ; E0301000 80001111
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v17, v17, v2                               ; 10220511
>	v_mad_f32 v16, v4, v16, v17                              ; D2820010 04462104
>	v_add_i32_e32 v17, 0x58, v15                             ; 4A221EFF 00000058
>	buffer_load_dword v17, v17, s[0:3], 0 offen              ; E0301000 80001111
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v16, v5, v17, v16                              ; D2820010 04422305
>	v_add_i32_e32 v17, 0x5c, v15                             ; 4A221EFF 0000005C
>	buffer_load_dword v17, v17, s[0:3], 0 offen              ; E0301000 80001111
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v16, v6, v17, v16                              ; D2820010 04422306
>	v_mad_f32 v1, v9, v16, v1                                ; D2820001 04062109
>	v_add_i32_e32 v16, 64, v15                               ; 4A201EC0
>	buffer_load_dword v16, v16, s[0:3], 0 offen              ; E0301000 80001010
>	v_add_i32_e32 v17, 0x44, v15                             ; 4A221EFF 00000044
>	buffer_load_dword v17, v17, s[0:3], 0 offen              ; E0301000 80001111
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v2, v17, v2                                ; 10040511
>	v_mad_f32 v2, v4, v16, v2                                ; D2820002 040A2104
>	v_add_i32_e32 v4, 0x48, v15                              ; 4A081EFF 00000048
>	buffer_load_dword v4, v4, s[0:3], 0 offen                ; E0301000 80000404
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v2, v5, v4, v2                                 ; D2820002 040A0905
>	v_add_i32_e32 v4, 0x4c, v15                              ; 4A081EFF 0000004C
>	buffer_load_dword v4, v4, s[0:3], 0 offen                ; E0301000 80000404
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v2, v6, v4, v2                                 ; D2820002 040A0906
>	v_mad_f32 v3, v9, v2, v3                                 ; D2820003 040E0509
>	s_or_b64 exec, exec, s[22:23]                            ; 88FE167E
>	s_or_b64 exec, exec, s[20:21]                            ; 88FE147E
>	v_mov_b32_e32 v2, 1.0                                    ; 7E0402F2
>	exp 15, 32, 0, 0, 0, v11, v12, v0, v2                    ; F800020F 02000C0B
>	s_waitcnt expcnt(0)                                      ; BF8C070F
>	v_mul_f32_e32 v2, s17, v1                                ; 10040211
>	v_mad_f32 v2, v3, s19, v2                                ; D2820002 04082703
>	v_mad_f32 v2, v0, s12, v2                                ; D2820002 04081900
>	v_add_f32_e32 v2, s9, v2                                 ; 06040409
>	v_mul_f32_e32 v4, s15, v1                                ; 1008020F
>	v_mad_f32 v4, v3, s18, v4                                ; D2820004 04102503
>	v_mad_f32 v4, v0, s10, v4                                ; D2820004 04101500
>	v_add_f32_e32 v4, s7, v4                                 ; 06080807
>	v_mul_f32_e32 v5, s13, v1                                ; 100A020D
>	v_mad_f32 v5, v3, s16, v5                                ; D2820005 04142103
>	v_mad_f32 v5, v0, s8, v5                                 ; D2820005 04141100
>	v_add_f32_e32 v5, s5, v5                                 ; 060A0A05
>	v_mul_f32_e32 v1, s11, v1                                ; 1002020B
>	v_mad_f32 v1, v3, s14, v1                                ; D2820001 04041D03
>	v_mad_f32 v0, v0, s6, v1                                 ; D2820000 04040D00
>	v_add_f32_e32 v0, s4, v0                                 ; 06000004
>	exp 15, 12, 0, 1, 0, v0, v5, v4, v2                      ; F80008CF 02040500
>	s_endpgm                                                 ; BF810000
>
>SGPRS: 32
>VGPRS: 24
>Code Size: 1412 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL CONST[0..1]
>DCL TEMP[0..1], LOCAL
>IMM[0] FLT32 {    0.0000,     1.0000,     0.3300,     0.0000}
>  0: ADD TEMP[0].x, -CONST[0].xxxx, IN[0].zzzz
>  1: FSGE TEMP[0].x, TEMP[0].xxxx, IMM[0].xxxx
>  2: UIF TEMP[0].xxxx :0
>  3:   MOV TEMP[0].x, IMM[0].yyyy
>  4: ELSE :0
>  5:   MOV TEMP[0].x, IMM[0].zzzz
>  6: ENDIF
>  7: ADD TEMP[1].y, -CONST[1].xxxx, CONST[1].yyyy
>  8: MAD TEMP[1].y, CONST[1].zzzz, TEMP[1].yyyy, CONST[1].xxxx
>  9: MUL TEMP[0].w, TEMP[1].yyyy, TEMP[0].xxxx
> 10: MOV TEMP[0].w, TEMP[0].wwww
> 11: MOV TEMP[0].xyz, IMM[0].yyyy
> 12: MOV OUT[0], TEMP[0]
> 13: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %23 = load <16 x i8> addrspace(2)* %22, !tbaa !0
>  %24 = call float @llvm.SI.load.const(<16 x i8> %23, i32 0)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %23, i32 16)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %23, i32 20)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %23, i32 24)
>  %28 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %5, <2 x i32> %7)
>  %29 = fsub float -0.000000e+00, %24
>  %30 = fadd float %29, %28
>  %31 = fcmp oge float %30, 0.000000e+00
>  %32 = sext i1 %31 to i32
>  %33 = bitcast i32 %32 to float
>  %34 = bitcast float %33 to i32
>  %35 = icmp ne i32 %34, 0
>  %. = select i1 %35, float 1.000000e+00, float 0x3FD51EB860000000
>  %36 = fsub float -0.000000e+00, %25
>  %37 = fadd float %36, %26
>  %38 = fmul float %27, %37
>  %39 = fadd float %38, %25
>  %40 = fmul float %39, %.
>  %41 = call i32 @llvm.SI.packf16(float 1.000000e+00, float 1.000000e+00)
>  %42 = bitcast i32 %41 to float
>  %43 = call i32 @llvm.SI.packf16(float 1.000000e+00, float %40)
>  %44 = bitcast i32 %43 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %42, float %44, float %42, float %44)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                     ; BEFE0A7E
>	s_mov_b32 m0, s9                         ; BEFC0309
>	v_interp_p1_f32 v2, v0, 2, 0, [m0]       ; C8080200
>	v_interp_p2_f32 v2, [v2], v1, 2, 0, [m0] ; C8090201
>	s_load_dwordx4 s[0:3], s[2:3], 0x0       ; C0800300
>	s_waitcnt lgkmcnt(0)                     ; BF8C007F
>	s_buffer_load_dword s4, s[0:3], 0x0      ; C2020100
>	s_waitcnt lgkmcnt(0)                     ; BF8C007F
>	v_subrev_f32_e32 v0, s4, v2              ; 0A000404
>	v_cmp_ge_f32_e64 s[4:5], v0, 0           ; D00C0004 00010100
>	v_cndmask_b32_e64 v0, 0, -1, s[4:5]      ; D2000000 00118280
>	v_cmp_ne_i32_e64 s[4:5], v0, 0           ; D10A0004 00010100
>	v_mov_b32_e32 v0, 0x3ea8f5c3             ; 7E0002FF 3EA8F5C3
>	v_cndmask_b32_e64 v0, v0, 1.0, s[4:5]    ; D2000000 0011E500
>	s_buffer_load_dword s4, s[0:3], 0x4      ; C2020104
>	s_buffer_load_dword s5, s[0:3], 0x5      ; C2028105
>	s_waitcnt lgkmcnt(0)                     ; BF8C007F
>	v_mov_b32_e32 v1, s4                     ; 7E020204
>	v_sub_f32_e32 v1, s5, v1                 ; 08020205
>	s_buffer_load_dword s0, s[0:3], 0x6      ; C2000106
>	v_mov_b32_e32 v2, s4                     ; 7E040204
>	s_waitcnt lgkmcnt(0)                     ; BF8C007F
>	v_mad_f32 v1, v1, s0, v2                 ; D2820001 04080101
>	v_mul_f32_e32 v0, v0, v1                 ; 10000300
>	v_cvt_pkrtz_f16_f32_e32 v0, 1.0, v0      ; 5E0000F2
>	v_cvt_pkrtz_f16_f32_e64 v1, 1.0, 1.0     ; D25E0001 0001E4F2
>	exp 15, 0, 1, 1, 1, v1, v0, v1, v0       ; F8001C0F 00010001
>	s_endpgm                                 ; BF810000
>
>SGPRS: 16
>VGPRS: 4
>Code Size: 144 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL CONST[0..7]
>DCL TEMP[0..1], LOCAL
>IMM[0] FLT32 {    1.0000,     0.0000,     0.0000,     0.0000}
>  0: MUL TEMP[0], CONST[5], IN[0].yyyy
>  1: MAD TEMP[0], IN[0].xxxx, CONST[4], TEMP[0]
>  2: MAD TEMP[0], IN[0].zzzz, CONST[6], TEMP[0]
>  3: ADD TEMP[0], TEMP[0], CONST[7]
>  4: MUL TEMP[1], TEMP[0].yyyy, CONST[1]
>  5: MAD TEMP[1], TEMP[0].xxxx, CONST[0], TEMP[1]
>  6: MAD TEMP[1], TEMP[0].zzzz, CONST[2], TEMP[1]
>  7: MAD TEMP[1], TEMP[0].wwww, CONST[3], TEMP[1]
>  8: MOV TEMP[0].z, TEMP[0].zzzz
>  9: MOV TEMP[0].xy, IN[1].xyxx
> 10: MOV TEMP[0].w, IMM[0].xxxx
> 11: MOV OUT[1], TEMP[0]
> 12: MOV OUT[0], TEMP[1]
> 13: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 0)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 4)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %12, i32 8)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %12, i32 12)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %12, i32 16)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %12, i32 20)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %12, i32 24)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %12, i32 28)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %12, i32 32)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %12, i32 36)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %12, i32 40)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %12, i32 44)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %12, i32 48)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %12, i32 52)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %12, i32 56)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %12, i32 60)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %12, i32 64)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %12, i32 68)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %12, i32 72)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %12, i32 76)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %12, i32 80)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %12, i32 84)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %12, i32 88)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %12, i32 92)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %12, i32 96)
>  %38 = call float @llvm.SI.load.const(<16 x i8> %12, i32 100)
>  %39 = call float @llvm.SI.load.const(<16 x i8> %12, i32 104)
>  %40 = call float @llvm.SI.load.const(<16 x i8> %12, i32 108)
>  %41 = call float @llvm.SI.load.const(<16 x i8> %12, i32 112)
>  %42 = call float @llvm.SI.load.const(<16 x i8> %12, i32 116)
>  %43 = call float @llvm.SI.load.const(<16 x i8> %12, i32 120)
>  %44 = call float @llvm.SI.load.const(<16 x i8> %12, i32 124)
>  %45 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %46 = load <16 x i8> addrspace(2)* %45, !tbaa !0
>  %47 = add i32 %5, %7
>  %48 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %46, i32 0, i32 %47)
>  %49 = extractelement <4 x float> %48, i32 0
>  %50 = extractelement <4 x float> %48, i32 1
>  %51 = extractelement <4 x float> %48, i32 2
>  %52 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %53 = load <16 x i8> addrspace(2)* %52, !tbaa !0
>  %54 = add i32 %5, %7
>  %55 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %53, i32 0, i32 %54)
>  %56 = extractelement <4 x float> %55, i32 0
>  %57 = extractelement <4 x float> %55, i32 1
>  %58 = fmul float %33, %50
>  %59 = fmul float %34, %50
>  %60 = fmul float %35, %50
>  %61 = fmul float %36, %50
>  %62 = fmul float %49, %29
>  %63 = fadd float %62, %58
>  %64 = fmul float %49, %30
>  %65 = fadd float %64, %59
>  %66 = fmul float %49, %31
>  %67 = fadd float %66, %60
>  %68 = fmul float %49, %32
>  %69 = fadd float %68, %61
>  %70 = fmul float %51, %37
>  %71 = fadd float %70, %63
>  %72 = fmul float %51, %38
>  %73 = fadd float %72, %65
>  %74 = fmul float %51, %39
>  %75 = fadd float %74, %67
>  %76 = fmul float %51, %40
>  %77 = fadd float %76, %69
>  %78 = fadd float %71, %41
>  %79 = fadd float %73, %42
>  %80 = fadd float %75, %43
>  %81 = fadd float %77, %44
>  %82 = fmul float %79, %17
>  %83 = fmul float %79, %18
>  %84 = fmul float %79, %19
>  %85 = fmul float %79, %20
>  %86 = fmul float %78, %13
>  %87 = fadd float %86, %82
>  %88 = fmul float %78, %14
>  %89 = fadd float %88, %83
>  %90 = fmul float %78, %15
>  %91 = fadd float %90, %84
>  %92 = fmul float %78, %16
>  %93 = fadd float %92, %85
>  %94 = fmul float %80, %21
>  %95 = fadd float %94, %87
>  %96 = fmul float %80, %22
>  %97 = fadd float %96, %89
>  %98 = fmul float %80, %23
>  %99 = fadd float %98, %91
>  %100 = fmul float %80, %24
>  %101 = fadd float %100, %93
>  %102 = fmul float %81, %25
>  %103 = fadd float %102, %95
>  %104 = fmul float %81, %26
>  %105 = fadd float %104, %97
>  %106 = fmul float %81, %27
>  %107 = fadd float %106, %99
>  %108 = fmul float %81, %28
>  %109 = fadd float %108, %101
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %56, float %57, float %80, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %103, float %105, float %107, float %109)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                            ; 4A00000A
>	s_load_dwordx4 s[4:7], s[8:9], 0x0                   ; C0820900
>	s_load_dwordx4 s[8:11], s[8:9], 0x4                  ; C0840904
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[4:7], 0 idxen  ; E00C2000 80010100
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                   ; C0800300
>	s_waitcnt vmcnt(0) lgkmcnt(0)                        ; BF8C0070
>	s_buffer_load_dword s4, s[0:3], 0x16                 ; C2020116
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v5, s4, v2                             ; 100A0404
>	s_buffer_load_dword s4, s[0:3], 0x12                 ; C2020112
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v5, v1, s4, v5                             ; D2820005 04140901
>	s_buffer_load_dword s4, s[0:3], 0x1a                 ; C202011A
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v5, v3, s4, v5                             ; D2820005 04140903
>	s_buffer_load_dword s4, s[0:3], 0x1e                 ; C202011E
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_add_f32_e32 v5, s4, v5                             ; 060A0A04
>	buffer_load_format_xyzw v[6:9], v0, s[8:11], 0 idxen ; E00C2000 80020600
>	v_mov_b32_e32 v0, 1.0                                ; 7E0002F2
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	exp 15, 32, 0, 0, 0, v6, v7, v5, v0                  ; F800020F 00050706
>	s_buffer_load_dword s4, s[0:3], 0x14                 ; C2020114
>	s_waitcnt expcnt(0) lgkmcnt(0)                       ; BF8C000F
>	v_mul_f32_e32 v0, s4, v2                             ; 10000404
>	s_buffer_load_dword s4, s[0:3], 0x10                 ; C2020110
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v0, v1, s4, v0                             ; D2820000 04000901
>	s_buffer_load_dword s4, s[0:3], 0x18                 ; C2020118
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v0, v3, s4, v0                             ; D2820000 04000903
>	s_buffer_load_dword s4, s[0:3], 0x1c                 ; C202011C
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_add_f32_e32 v0, s4, v0                             ; 06000004
>	s_buffer_load_dword s4, s[0:3], 0x15                 ; C2020115
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v6, s4, v2                             ; 100C0404
>	s_buffer_load_dword s4, s[0:3], 0x11                 ; C2020111
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v6, v1, s4, v6                             ; D2820006 04180901
>	s_buffer_load_dword s4, s[0:3], 0x19                 ; C2020119
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v6, v3, s4, v6                             ; D2820006 04180903
>	s_buffer_load_dword s4, s[0:3], 0x1d                 ; C202011D
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_add_f32_e32 v6, s4, v6                             ; 060C0C04
>	s_buffer_load_dword s4, s[0:3], 0x7                  ; C2020107
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v7, s4, v6                             ; 100E0C04
>	s_buffer_load_dword s4, s[0:3], 0x3                  ; C2020103
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v7, v0, s4, v7                             ; D2820007 041C0900
>	s_buffer_load_dword s4, s[0:3], 0xb                  ; C202010B
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v7, v5, s4, v7                             ; D2820007 041C0905
>	s_buffer_load_dword s4, s[0:3], 0x17                 ; C2020117
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v8, s4, v2                             ; 10100404
>	s_buffer_load_dword s4, s[0:3], 0x13                 ; C2020113
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v8, v1, s4, v8                             ; D2820008 04200901
>	s_buffer_load_dword s4, s[0:3], 0x1b                 ; C202011B
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v1, v3, s4, v8                             ; D2820001 04200903
>	s_buffer_load_dword s4, s[0:3], 0x1f                 ; C202011F
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_add_f32_e32 v1, s4, v1                             ; 06020204
>	s_buffer_load_dword s4, s[0:3], 0xf                  ; C202010F
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v2, v1, s4, v7                             ; D2820002 041C0901
>	s_buffer_load_dword s4, s[0:3], 0x6                  ; C2020106
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v3, s4, v6                             ; 10060C04
>	s_buffer_load_dword s4, s[0:3], 0x2                  ; C2020102
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v3, v0, s4, v3                             ; D2820003 040C0900
>	s_buffer_load_dword s4, s[0:3], 0xa                  ; C202010A
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v3, v5, s4, v3                             ; D2820003 040C0905
>	s_buffer_load_dword s4, s[0:3], 0xe                  ; C202010E
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v3, v1, s4, v3                             ; D2820003 040C0901
>	s_buffer_load_dword s4, s[0:3], 0x5                  ; C2020105
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v4, s4, v6                             ; 10080C04
>	s_buffer_load_dword s4, s[0:3], 0x1                  ; C2020101
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v4, v0, s4, v4                             ; D2820004 04100900
>	s_buffer_load_dword s4, s[0:3], 0x9                  ; C2020109
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v4, v5, s4, v4                             ; D2820004 04100905
>	s_buffer_load_dword s4, s[0:3], 0xd                  ; C202010D
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v4, v1, s4, v4                             ; D2820004 04100901
>	s_buffer_load_dword s4, s[0:3], 0x4                  ; C2020104
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v6, s4, v6                             ; 100C0C04
>	s_buffer_load_dword s4, s[0:3], 0x0                  ; C2020100
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v0, v0, s4, v6                             ; D2820000 04180900
>	s_buffer_load_dword s4, s[0:3], 0x8                  ; C2020108
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v0, v5, s4, v0                             ; D2820000 04000905
>	s_buffer_load_dword s0, s[0:3], 0xc                  ; C200010C
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v0, v1, s0, v0                             ; D2820000 04000101
>	exp 15, 12, 0, 1, 0, v0, v4, v3, v2                  ; F80008CF 02030400
>	s_endpgm                                             ; BF810000
>
>SGPRS: 16
>VGPRS: 12
>Code Size: 532 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL CONST[0]
>DCL TEMP[0..2], LOCAL
>IMM[0] FLT32 {    2.2000,     1.0000,     0.5000,     0.0000}
>IMM[1] FLT32 {    0.1000,     0.0000,     0.0000,     0.0000}
>  0: ADD TEMP[0].x, -CONST[0].xxxx, IN[0].zzzz
>  1: MOV TEMP[1].xy, IN[0].xyyy
>  2: TEX TEMP[1].w, TEMP[1], SAMP[0], 2D
>  3: POW TEMP[1].w, TEMP[1].wwww, IMM[0].yyyy
>  4: MUL TEMP[2].y, TEMP[1].wwww, IMM[0].zzzz
>  5: FSGE TEMP[0].x, TEMP[0].xxxx, IMM[0].wwww
>  6: UIF TEMP[0].xxxx :0
>  7:   MOV TEMP[0].x, TEMP[1].wwww
>  8: ELSE :0
>  9:   MOV TEMP[0].x, TEMP[2].yyyy
> 10: ENDIF
> 11: ADD TEMP[1].y, -TEMP[1].wwww, IMM[1].xxxx
> 12: FSGE TEMP[1].x, TEMP[1].yyyy, IMM[0].wwww
> 13: UIF TEMP[1].xxxx :0
> 14:   MOV TEMP[1].x, IMM[0].wwww
> 15: ELSE :0
> 16:   MOV TEMP[1].x, TEMP[0].xxxx
> 17: ENDIF
> 18: MOV TEMP[0].w, TEMP[1].xxxx
> 19: MOV TEMP[0].xyz, IMM[0].yyyy
> 20: MOV OUT[0], TEMP[0]
> 21: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %23 = load <16 x i8> addrspace(2)* %22, !tbaa !0
>  %24 = call float @llvm.SI.load.const(<16 x i8> %23, i32 0)
>  %25 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %26 = load <8 x i32> addrspace(2)* %25, !tbaa !0
>  %27 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %28 = load <4 x i32> addrspace(2)* %27, !tbaa !0
>  %29 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %30 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %31 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %5, <2 x i32> %7)
>  %32 = fsub float -0.000000e+00, %24
>  %33 = fadd float %32, %31
>  %34 = bitcast float %29 to i32
>  %35 = bitcast float %30 to i32
>  %36 = insertelement <2 x i32> undef, i32 %34, i32 0
>  %37 = insertelement <2 x i32> %36, i32 %35, i32 1
>  %38 = bitcast <8 x i32> %26 to <32 x i8>
>  %39 = bitcast <4 x i32> %28 to <16 x i8>
>  %40 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %37, <32 x i8> %38, <16 x i8> %39, i32 2)
>  %41 = extractelement <4 x float> %40, i32 3
>  %42 = call float @llvm.pow.f32(float %41, float 1.000000e+00)
>  %43 = fmul float %42, 5.000000e-01
>  %44 = fcmp oge float %33, 0.000000e+00
>  %45 = sext i1 %44 to i32
>  %46 = bitcast i32 %45 to float
>  %47 = bitcast float %46 to i32
>  %48 = icmp ne i32 %47, 0
>  %. = select i1 %48, float %42, float %43
>  %49 = fsub float -0.000000e+00, %42
>  %50 = fadd float %49, 0x3FB99999A0000000
>  %51 = fcmp oge float %50, 0.000000e+00
>  %52 = sext i1 %51 to i32
>  %53 = bitcast i32 %52 to float
>  %54 = bitcast float %53 to i32
>  %55 = icmp ne i32 %54, 0
>  %temp4.0 = select i1 %55, float 0.000000e+00, float %.
>  %56 = call i32 @llvm.SI.packf16(float 1.000000e+00, float 1.000000e+00)
>  %57 = bitcast i32 %56 to float
>  %58 = call i32 @llvm.SI.packf16(float 1.000000e+00, float %temp4.0)
>  %59 = bitcast i32 %58 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %57, float %59, float %57, float %59)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.pow.f32(float, float) #1
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                               ; BEFE0A7E
>	s_mov_b32 m0, s9                                                   ; BEFC0309
>	v_interp_p1_f32 v3, v0, 1, 0, [m0]                                 ; C80C0100
>	v_interp_p2_f32 v3, [v3], v1, 1, 0, [m0]                           ; C80D0101
>	v_interp_p1_f32 v2, v0, 0, 0, [m0]                                 ; C8080000
>	v_interp_p2_f32 v2, [v2], v1, 0, 0, [m0]                           ; C8090001
>	s_load_dwordx4 s[8:11], s[4:5], 0x0                                ; C0840500
>	s_load_dwordx8 s[12:19], s[6:7], 0x0                               ; C0C60700
>	s_waitcnt lgkmcnt(0)                                               ; BF8C007F
>	image_sample v2, 8, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[12:19], s[8:11] ; F0800800 00430202
>	s_waitcnt vmcnt(0)                                                 ; BF8C0770
>	v_log_f32_e32 v2, v2                                               ; 7E044F02
>	v_mul_legacy_f32_e32 v2, 1.0, v2                                   ; 0E0404F2
>	v_exp_f32_e32 v2, v2                                               ; 7E044B02
>	v_mul_f32_e32 v3, 0.5, v2                                          ; 100604F0
>	v_interp_p1_f32 v4, v0, 2, 0, [m0]                                 ; C8100200
>	v_interp_p2_f32 v4, [v4], v1, 2, 0, [m0]                           ; C8110201
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                                 ; C0800300
>	s_waitcnt lgkmcnt(0)                                               ; BF8C007F
>	s_buffer_load_dword s0, s[0:3], 0x0                                ; C2000100
>	s_waitcnt lgkmcnt(0)                                               ; BF8C007F
>	v_subrev_f32_e32 v0, s0, v4                                        ; 0A000800
>	v_cmp_ge_f32_e64 s[0:1], v0, 0                                     ; D00C0000 00010100
>	v_cndmask_b32_e64 v0, 0, -1, s[0:1]                                ; D2000000 00018280
>	v_cmp_ne_i32_e64 s[0:1], v0, 0                                     ; D10A0000 00010100
>	v_cndmask_b32_e64 v0, v3, v2, s[0:1]                               ; D2000000 18020503
>	v_sub_f32_e32 v1, 0x3dcccccd, v2                                   ; 080204FF 3DCCCCCD
>	v_cmp_ge_f32_e64 s[0:1], v1, 0                                     ; D00C0000 00010101
>	v_cndmask_b32_e64 v1, 0, -1, s[0:1]                                ; D2000801 00018280
>	v_cmp_ne_i32_e64 s[0:1], v1, 0                                     ; D10A0000 00010101
>	v_cndmask_b32_e64 v0, v0, 0, s[0:1]                                ; D2000000 00010100
>	v_cvt_pkrtz_f16_f32_e32 v0, 1.0, v0                                ; 5E0000F2
>	v_cvt_pkrtz_f16_f32_e64 v1, 1.0, 1.0                               ; D25E0001 0001E4F2
>	exp 15, 0, 1, 1, 1, v1, v0, v1, v0                                 ; F8001C0F 00010001
>	s_endpgm                                                           ; BF810000
>
>SGPRS: 24
>VGPRS: 8
>Code Size: 188 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL IN[2]
>DCL IN[3]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL CONST[0..99]
>DCL TEMP[0..4], LOCAL
>DCL ADDR[0]
>IMM[0] FLT32 {    3.0000,     1.0000,     0.0000,     0.0000}
>IMM[1] INT32 {1, 2, 0, 0}
>  0: MUL TEMP[0].x, IMM[0].xxxx, IN[2].xxxx
>  1: MAD TEMP[1], IN[0].xyzx, IMM[0].yyyz, IMM[0].zzzy
>  2: F2I TEMP[2].x, TEMP[0].xxxx
>  3: UARL ADDR[0].x, TEMP[2].xxxx
>  4: UARL ADDR[0].x, TEMP[2].xxxx
>  5: DP4 TEMP[2].x, TEMP[1], CONST[ADDR[0].x+4]
>  6: F2I TEMP[3].x, TEMP[0].xxxx
>  7: UADD TEMP[3].x, IMM[1].xxxx, TEMP[3].xxxx
>  8: UARL ADDR[0].x, TEMP[3].xxxx
>  9: DP4 TEMP[3].x, TEMP[1], CONST[ADDR[0].x+4]
> 10: MOV TEMP[2].y, TEMP[3].xxxx
> 11: F2I TEMP[0].x, TEMP[0].xxxx
> 12: UADD TEMP[0].x, IMM[1].yyyy, TEMP[0].xxxx
> 13: UARL ADDR[0].x, TEMP[0].xxxx
> 14: DP4 TEMP[0].x, TEMP[1], CONST[ADDR[0].x+4]
> 15: MOV TEMP[2].z, TEMP[0].xxxx
> 16: MUL TEMP[0].xyz, TEMP[2], IN[1].xxxx
> 17: MOV TEMP[2].xyz, TEMP[0].xyzx
> 18: FSLT TEMP[0].x, IMM[0].zzzz, IN[1].yyyy
> 19: UIF TEMP[0].xxxx :0
> 20:   MUL TEMP[0].w, IMM[0].xxxx, IN[2].yyyy
> 21:   MOV TEMP[2].w, TEMP[0].wwww
> 22:   F2I TEMP[3].x, TEMP[0].wwww
> 23:   UARL ADDR[0].x, TEMP[3].xxxx
> 24:   UARL ADDR[0].x, TEMP[3].xxxx
> 25:   DP4 TEMP[3].x, TEMP[1], CONST[ADDR[0].x+4]
> 26:   F2I TEMP[4].x, TEMP[0].wwww
> 27:   UADD TEMP[4].x, IMM[1].xxxx, TEMP[4].xxxx
> 28:   UARL ADDR[0].x, TEMP[4].xxxx
> 29:   DP4 TEMP[4].x, TEMP[1], CONST[ADDR[0].x+4]
> 30:   MOV TEMP[3].y, TEMP[4].xxxx
> 31:   F2I TEMP[0].x, TEMP[0].wwww
> 32:   UADD TEMP[0].x, IMM[1].yyyy, TEMP[0].xxxx
> 33:   UARL ADDR[0].x, TEMP[0].xxxx
> 34:   DP4 TEMP[0].x, TEMP[1], CONST[ADDR[0].x+4]
> 35:   MOV TEMP[3].z, TEMP[0].xxxx
> 36:   MAD TEMP[0].xyz, IN[1].yyyy, TEMP[3], TEMP[2]
> 37:   MOV TEMP[2].xyz, TEMP[0].xyzx
> 38:   FSLT TEMP[0].x, IMM[0].zzzz, IN[1].zzzz
> 39:   UIF TEMP[0].xxxx :0
> 40:     MUL TEMP[0].w, IMM[0].xxxx, IN[2].zzzz
> 41:     MOV TEMP[2].w, TEMP[0].wwww
> 42:     F2I TEMP[4].x, TEMP[0].wwww
> 43:     UARL ADDR[0].x, TEMP[4].xxxx
> 44:     UARL ADDR[0].x, TEMP[4].xxxx
> 45:     DP4 TEMP[3].x, TEMP[1], CONST[ADDR[0].x+4]
> 46:     F2I TEMP[4].x, TEMP[0].wwww
> 47:     UADD TEMP[4].x, IMM[1].xxxx, TEMP[4].xxxx
> 48:     UARL ADDR[0].x, TEMP[4].xxxx
> 49:     DP4 TEMP[4].x, TEMP[1], CONST[ADDR[0].x+4]
> 50:     MOV TEMP[3].y, TEMP[4].xxxx
> 51:     F2I TEMP[0].x, TEMP[0].wwww
> 52:     UADD TEMP[0].x, IMM[1].yyyy, TEMP[0].xxxx
> 53:     UARL ADDR[0].x, TEMP[0].xxxx
> 54:     DP4 TEMP[0].x, TEMP[1], CONST[ADDR[0].x+4]
> 55:     MOV TEMP[3].z, TEMP[0].xxxx
> 56:     MAD TEMP[0].xyz, IN[1].zzzz, TEMP[3], TEMP[2]
> 57:     MOV TEMP[2].xyz, TEMP[0].xyzx
> 58:   ENDIF
> 59: ENDIF
> 60: MUL TEMP[1], TEMP[2].yyyy, CONST[1]
> 61: MAD TEMP[1], TEMP[2].xxxx, CONST[0], TEMP[1]
> 62: MAD TEMP[1], TEMP[2].zzzz, CONST[2], TEMP[1]
> 63: ADD TEMP[0], TEMP[1], CONST[3]
> 64: MOV TEMP[2].xy, IN[3].xyxx
> 65: MOV TEMP[1].xyz, TEMP[2].xyzx
> 66: MOV TEMP[1].w, IMM[0].yyyy
> 67: MOV OUT[1], TEMP[1]
> 68: MOV OUT[0], TEMP[0]
> 69: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 0)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 4)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %12, i32 8)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %12, i32 12)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %12, i32 16)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %12, i32 20)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %12, i32 24)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %12, i32 28)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %12, i32 32)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %12, i32 36)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %12, i32 40)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %12, i32 44)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %12, i32 48)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %12, i32 52)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %12, i32 56)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %12, i32 60)
>  %29 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %30 = load <16 x i8> addrspace(2)* %29, !tbaa !0
>  %31 = add i32 %5, %7
>  %32 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %30, i32 0, i32 %31)
>  %33 = extractelement <4 x float> %32, i32 0
>  %34 = extractelement <4 x float> %32, i32 1
>  %35 = extractelement <4 x float> %32, i32 2
>  %36 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %37 = load <16 x i8> addrspace(2)* %36, !tbaa !0
>  %38 = add i32 %5, %7
>  %39 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %37, i32 0, i32 %38)
>  %40 = extractelement <4 x float> %39, i32 0
>  %41 = extractelement <4 x float> %39, i32 1
>  %42 = extractelement <4 x float> %39, i32 2
>  %43 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 2
>  %44 = load <16 x i8> addrspace(2)* %43, !tbaa !0
>  %45 = add i32 %5, %7
>  %46 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %44, i32 0, i32 %45)
>  %47 = extractelement <4 x float> %46, i32 0
>  %48 = extractelement <4 x float> %46, i32 1
>  %49 = extractelement <4 x float> %46, i32 2
>  %50 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 3
>  %51 = load <16 x i8> addrspace(2)* %50, !tbaa !0
>  %52 = add i32 %5, %7
>  %53 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %51, i32 0, i32 %52)
>  %54 = extractelement <4 x float> %53, i32 0
>  %55 = extractelement <4 x float> %53, i32 1
>  %56 = fmul float 3.000000e+00, %47
>  %57 = fmul float %33, 1.000000e+00
>  %58 = fadd float %57, 0.000000e+00
>  %59 = fmul float %34, 1.000000e+00
>  %60 = fadd float %59, 0.000000e+00
>  %61 = fmul float %35, 1.000000e+00
>  %62 = fadd float %61, 0.000000e+00
>  %63 = fmul float %33, 0.000000e+00
>  %64 = fadd float %63, 1.000000e+00
>  %65 = fptosi float %56 to i32
>  %66 = bitcast i32 %65 to float
>  %67 = bitcast float %66 to i32
>  %68 = shl i32 %67, 4
>  %69 = add i32 %68, 64
>  %70 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %69)
>  %71 = shl i32 %67, 4
>  %72 = add i32 %71, 68
>  %73 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %72)
>  %74 = shl i32 %67, 4
>  %75 = add i32 %74, 72
>  %76 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %75)
>  %77 = shl i32 %67, 4
>  %78 = add i32 %77, 76
>  %79 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %78)
>  %80 = fmul float %58, %70
>  %81 = fmul float %60, %73
>  %82 = fadd float %80, %81
>  %83 = fmul float %62, %76
>  %84 = fadd float %82, %83
>  %85 = fmul float %64, %79
>  %86 = fadd float %84, %85
>  %87 = fptosi float %56 to i32
>  %88 = bitcast i32 %87 to float
>  %89 = bitcast float %88 to i32
>  %90 = add i32 1, %89
>  %91 = bitcast i32 %90 to float
>  %92 = bitcast float %91 to i32
>  %93 = shl i32 %92, 4
>  %94 = add i32 %93, 64
>  %95 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %94)
>  %96 = shl i32 %92, 4
>  %97 = add i32 %96, 68
>  %98 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %97)
>  %99 = shl i32 %92, 4
>  %100 = add i32 %99, 72
>  %101 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %100)
>  %102 = shl i32 %92, 4
>  %103 = add i32 %102, 76
>  %104 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %103)
>  %105 = fmul float %58, %95
>  %106 = fmul float %60, %98
>  %107 = fadd float %105, %106
>  %108 = fmul float %62, %101
>  %109 = fadd float %107, %108
>  %110 = fmul float %64, %104
>  %111 = fadd float %109, %110
>  %112 = fptosi float %56 to i32
>  %113 = bitcast i32 %112 to float
>  %114 = bitcast float %113 to i32
>  %115 = add i32 2, %114
>  %116 = bitcast i32 %115 to float
>  %117 = bitcast float %116 to i32
>  %118 = shl i32 %117, 4
>  %119 = add i32 %118, 64
>  %120 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %119)
>  %121 = shl i32 %117, 4
>  %122 = add i32 %121, 68
>  %123 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %122)
>  %124 = shl i32 %117, 4
>  %125 = add i32 %124, 72
>  %126 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %125)
>  %127 = shl i32 %117, 4
>  %128 = add i32 %127, 76
>  %129 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %128)
>  %130 = fmul float %58, %120
>  %131 = fmul float %60, %123
>  %132 = fadd float %130, %131
>  %133 = fmul float %62, %126
>  %134 = fadd float %132, %133
>  %135 = fmul float %64, %129
>  %136 = fadd float %134, %135
>  %137 = fmul float %86, %40
>  %138 = fmul float %111, %40
>  %139 = fmul float %136, %40
>  %140 = fcmp olt float 0.000000e+00, %41
>  %141 = sext i1 %140 to i32
>  %142 = bitcast i32 %141 to float
>  %143 = bitcast float %142 to i32
>  %144 = icmp ne i32 %143, 0
>  br i1 %144, label %IF, label %ENDIF
>
>IF:                                               ; preds = %main_body
>  %145 = fmul float 3.000000e+00, %48
>  %146 = fptosi float %145 to i32
>  %147 = bitcast i32 %146 to float
>  %148 = bitcast float %147 to i32
>  %149 = shl i32 %148, 4
>  %150 = add i32 %149, 64
>  %151 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %150)
>  %152 = shl i32 %148, 4
>  %153 = add i32 %152, 68
>  %154 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %153)
>  %155 = shl i32 %148, 4
>  %156 = add i32 %155, 72
>  %157 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %156)
>  %158 = shl i32 %148, 4
>  %159 = add i32 %158, 76
>  %160 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %159)
>  %161 = fmul float %58, %151
>  %162 = fmul float %60, %154
>  %163 = fadd float %161, %162
>  %164 = fmul float %62, %157
>  %165 = fadd float %163, %164
>  %166 = fmul float %64, %160
>  %167 = fadd float %165, %166
>  %168 = fptosi float %145 to i32
>  %169 = bitcast i32 %168 to float
>  %170 = bitcast float %169 to i32
>  %171 = add i32 1, %170
>  %172 = bitcast i32 %171 to float
>  %173 = bitcast float %172 to i32
>  %174 = shl i32 %173, 4
>  %175 = add i32 %174, 64
>  %176 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %175)
>  %177 = shl i32 %173, 4
>  %178 = add i32 %177, 68
>  %179 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %178)
>  %180 = shl i32 %173, 4
>  %181 = add i32 %180, 72
>  %182 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %181)
>  %183 = shl i32 %173, 4
>  %184 = add i32 %183, 76
>  %185 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %184)
>  %186 = fmul float %58, %176
>  %187 = fmul float %60, %179
>  %188 = fadd float %186, %187
>  %189 = fmul float %62, %182
>  %190 = fadd float %188, %189
>  %191 = fmul float %64, %185
>  %192 = fadd float %190, %191
>  %193 = fptosi float %145 to i32
>  %194 = bitcast i32 %193 to float
>  %195 = bitcast float %194 to i32
>  %196 = add i32 2, %195
>  %197 = bitcast i32 %196 to float
>  %198 = bitcast float %197 to i32
>  %199 = shl i32 %198, 4
>  %200 = add i32 %199, 64
>  %201 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %200)
>  %202 = shl i32 %198, 4
>  %203 = add i32 %202, 68
>  %204 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %203)
>  %205 = shl i32 %198, 4
>  %206 = add i32 %205, 72
>  %207 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %206)
>  %208 = shl i32 %198, 4
>  %209 = add i32 %208, 76
>  %210 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %209)
>  %211 = fmul float %58, %201
>  %212 = fmul float %60, %204
>  %213 = fadd float %211, %212
>  %214 = fmul float %62, %207
>  %215 = fadd float %213, %214
>  %216 = fmul float %64, %210
>  %217 = fadd float %215, %216
>  %218 = fmul float %41, %167
>  %219 = fadd float %218, %137
>  %220 = fmul float %41, %192
>  %221 = fadd float %220, %138
>  %222 = fmul float %41, %217
>  %223 = fadd float %222, %139
>  %224 = fcmp olt float 0.000000e+00, %42
>  %225 = sext i1 %224 to i32
>  %226 = bitcast i32 %225 to float
>  %227 = bitcast float %226 to i32
>  %228 = icmp ne i32 %227, 0
>  br i1 %228, label %IF44, label %ENDIF
>
>ENDIF:                                            ; preds = %IF44, %IF, %main_body
>  %temp8.0 = phi float [ %137, %main_body ], [ %327, %IF44 ], [ %219, %IF ]
>  %temp9.0 = phi float [ %138, %main_body ], [ %329, %IF44 ], [ %221, %IF ]
>  %temp10.0 = phi float [ %139, %main_body ], [ %331, %IF44 ], [ %223, %IF ]
>  %229 = fmul float %temp9.0, %17
>  %230 = fmul float %temp9.0, %18
>  %231 = fmul float %temp9.0, %19
>  %232 = fmul float %temp9.0, %20
>  %233 = fmul float %temp8.0, %13
>  %234 = fadd float %233, %229
>  %235 = fmul float %temp8.0, %14
>  %236 = fadd float %235, %230
>  %237 = fmul float %temp8.0, %15
>  %238 = fadd float %237, %231
>  %239 = fmul float %temp8.0, %16
>  %240 = fadd float %239, %232
>  %241 = fmul float %temp10.0, %21
>  %242 = fadd float %241, %234
>  %243 = fmul float %temp10.0, %22
>  %244 = fadd float %243, %236
>  %245 = fmul float %temp10.0, %23
>  %246 = fadd float %245, %238
>  %247 = fmul float %temp10.0, %24
>  %248 = fadd float %247, %240
>  %249 = fadd float %242, %25
>  %250 = fadd float %244, %26
>  %251 = fadd float %246, %27
>  %252 = fadd float %248, %28
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %54, float %55, float %temp10.0, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %249, float %250, float %251, float %252)
>  ret void
>
>IF44:                                             ; preds = %IF
>  %253 = fmul float 3.000000e+00, %49
>  %254 = fptosi float %253 to i32
>  %255 = bitcast i32 %254 to float
>  %256 = bitcast float %255 to i32
>  %257 = shl i32 %256, 4
>  %258 = add i32 %257, 64
>  %259 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %258)
>  %260 = shl i32 %256, 4
>  %261 = add i32 %260, 68
>  %262 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %261)
>  %263 = shl i32 %256, 4
>  %264 = add i32 %263, 72
>  %265 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %264)
>  %266 = shl i32 %256, 4
>  %267 = add i32 %266, 76
>  %268 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %267)
>  %269 = fmul float %58, %259
>  %270 = fmul float %60, %262
>  %271 = fadd float %269, %270
>  %272 = fmul float %62, %265
>  %273 = fadd float %271, %272
>  %274 = fmul float %64, %268
>  %275 = fadd float %273, %274
>  %276 = fptosi float %253 to i32
>  %277 = bitcast i32 %276 to float
>  %278 = bitcast float %277 to i32
>  %279 = add i32 1, %278
>  %280 = bitcast i32 %279 to float
>  %281 = bitcast float %280 to i32
>  %282 = shl i32 %281, 4
>  %283 = add i32 %282, 64
>  %284 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %283)
>  %285 = shl i32 %281, 4
>  %286 = add i32 %285, 68
>  %287 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %286)
>  %288 = shl i32 %281, 4
>  %289 = add i32 %288, 72
>  %290 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %289)
>  %291 = shl i32 %281, 4
>  %292 = add i32 %291, 76
>  %293 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %292)
>  %294 = fmul float %58, %284
>  %295 = fmul float %60, %287
>  %296 = fadd float %294, %295
>  %297 = fmul float %62, %290
>  %298 = fadd float %296, %297
>  %299 = fmul float %64, %293
>  %300 = fadd float %298, %299
>  %301 = fptosi float %253 to i32
>  %302 = bitcast i32 %301 to float
>  %303 = bitcast float %302 to i32
>  %304 = add i32 2, %303
>  %305 = bitcast i32 %304 to float
>  %306 = bitcast float %305 to i32
>  %307 = shl i32 %306, 4
>  %308 = add i32 %307, 64
>  %309 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %308)
>  %310 = shl i32 %306, 4
>  %311 = add i32 %310, 68
>  %312 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %311)
>  %313 = shl i32 %306, 4
>  %314 = add i32 %313, 72
>  %315 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %314)
>  %316 = shl i32 %306, 4
>  %317 = add i32 %316, 76
>  %318 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %317)
>  %319 = fmul float %58, %309
>  %320 = fmul float %60, %312
>  %321 = fadd float %319, %320
>  %322 = fmul float %62, %315
>  %323 = fadd float %321, %322
>  %324 = fmul float %64, %318
>  %325 = fadd float %323, %324
>  %326 = fmul float %42, %275
>  %327 = fadd float %326, %219
>  %328 = fmul float %42, %300
>  %329 = fadd float %328, %221
>  %330 = fmul float %42, %325
>  %331 = fadd float %330, %223
>  br label %ENDIF
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v11, s10, v0                               ; 4A16000A
>	s_load_dwordx4 s[12:15], s[8:9], 0x0                     ; C0860900
>	s_load_dwordx4 s[16:19], s[8:9], 0x4                     ; C0880904
>	s_load_dwordx4 s[20:23], s[8:9], 0x8                     ; C08A0908
>	s_load_dwordx4 s[4:7], s[8:9], 0xc                       ; C082090C
>	s_waitcnt lgkmcnt(0)                                     ; BF8C007F
>	buffer_load_format_xyzw v[6:9], v11, s[12:15], 0 idxen   ; E00C2000 8003060B
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_add_f32_e32 v2, 0, v7                                  ; 06040E80
>	buffer_load_format_xyzw v[15:18], v11, s[20:23], 0 idxen ; E00C2000 80050F0B
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v0, 0x40400000, v15                        ; 10001EFF 40400000
>	v_cvt_i32_f32_e32 v0, v0                                 ; 7E001100
>	v_lshlrev_b32_e32 v3, 4, v0                              ; 34060084
>	v_add_i32_e32 v0, 0x64, v3                               ; 4A0006FF 00000064
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                       ; C0800300
>	s_waitcnt lgkmcnt(0)                                     ; BF8C007F
>	buffer_load_dword v0, v0, s[0:3], 0 offen                ; E0301000 80000000
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v0, v0, v2                                 ; 10000500
>	v_add_f32_e32 v4, 0, v6                                  ; 06080C80
>	v_add_i32_e32 v1, 0x60, v3                               ; 4A0206FF 00000060
>	buffer_load_dword v1, v1, s[0:3], 0 offen                ; E0301000 80000101
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v0, v4, v1, v0                                 ; D2820000 04020304
>	v_add_f32_e32 v5, 0, v8                                  ; 060A1080
>	v_add_i32_e32 v1, 0x68, v3                               ; 4A0206FF 00000068
>	buffer_load_dword v1, v1, s[0:3], 0 offen                ; E0301000 80000101
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v0, v5, v1, v0                                 ; D2820000 04020305
>	v_mad_f32 v6, 0, v6, 1.0                                 ; D2820006 03CA0C80
>	v_add_i32_e32 v1, 0x6c, v3                               ; 4A0206FF 0000006C
>	buffer_load_dword v1, v1, s[0:3], 0 offen                ; E0301000 80000101
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v0, v6, v1, v0                                 ; D2820000 04020306
>	buffer_load_format_xyzw v[7:10], v11, s[16:19], 0 idxen  ; E00C2000 8004070B
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v0, v7, v0                                 ; 10000107
>	v_add_i32_e32 v1, 0x54, v3                               ; 4A0206FF 00000054
>	buffer_load_dword v1, v1, s[0:3], 0 offen                ; E0301000 80000101
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v1, v1, v2                                 ; 10020501
>	v_add_i32_e32 v12, 0x50, v3                              ; 4A1806FF 00000050
>	buffer_load_dword v12, v12, s[0:3], 0 offen              ; E0301000 80000C0C
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v1, v4, v12, v1                                ; D2820001 04061904
>	v_add_i32_e32 v12, 0x58, v3                              ; 4A1806FF 00000058
>	buffer_load_dword v12, v12, s[0:3], 0 offen              ; E0301000 80000C0C
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v1, v5, v12, v1                                ; D2820001 04061905
>	v_add_i32_e32 v12, 0x5c, v3                              ; 4A1806FF 0000005C
>	buffer_load_dword v12, v12, s[0:3], 0 offen              ; E0301000 80000C0C
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v1, v6, v12, v1                                ; D2820001 04061906
>	v_mul_f32_e32 v1, v7, v1                                 ; 10020307
>	v_add_i32_e32 v12, 0x44, v3                              ; 4A1806FF 00000044
>	buffer_load_dword v12, v12, s[0:3], 0 offen              ; E0301000 80000C0C
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v12, v12, v2                               ; 1018050C
>	v_add_i32_e32 v13, 64, v3                                ; 4A1A06C0
>	buffer_load_dword v13, v13, s[0:3], 0 offen              ; E0301000 80000D0D
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v12, v4, v13, v12                              ; D282000C 04321B04
>	v_add_i32_e32 v13, 0x48, v3                              ; 4A1A06FF 00000048
>	buffer_load_dword v13, v13, s[0:3], 0 offen              ; E0301000 80000D0D
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v12, v5, v13, v12                              ; D282000C 04321B05
>	v_add_i32_e32 v3, 0x4c, v3                               ; 4A0606FF 0000004C
>	buffer_load_dword v3, v3, s[0:3], 0 offen                ; E0301000 80000303
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v3, v6, v3, v12                                ; D2820003 04320706
>	v_mul_f32_e32 v3, v7, v3                                 ; 10060707
>	buffer_load_format_xyzw v[11:14], v11, s[4:7], 0 idxen   ; E00C2000 80010B0B
>	v_cmp_gt_f32_e64 s[4:5], v8, 0                           ; D0080004 00010108
>	v_cndmask_b32_e64 v19, 0, -1, s[4:5]                     ; D2000813 00118280
>	v_cmp_ne_i32_e64 s[20:21], v19, 0                        ; D10A0014 00010113
>	s_buffer_load_dword s9, s[0:3], 0xf                      ; C204810F
>	s_buffer_load_dword s7, s[0:3], 0xe                      ; C203810E
>	s_buffer_load_dword s5, s[0:3], 0xd                      ; C202810D
>	s_buffer_load_dword s4, s[0:3], 0xc                      ; C202010C
>	s_buffer_load_dword s12, s[0:3], 0xb                     ; C206010B
>	s_buffer_load_dword s10, s[0:3], 0xa                     ; C205010A
>	s_buffer_load_dword s8, s[0:3], 0x9                      ; C2040109
>	s_buffer_load_dword s6, s[0:3], 0x8                      ; C2030108
>	s_buffer_load_dword s17, s[0:3], 0x7                     ; C2088107
>	s_buffer_load_dword s15, s[0:3], 0x6                     ; C2078106
>	s_buffer_load_dword s13, s[0:3], 0x5                     ; C2068105
>	s_buffer_load_dword s11, s[0:3], 0x4                     ; C2058104
>	s_buffer_load_dword s19, s[0:3], 0x3                     ; C2098103
>	s_buffer_load_dword s18, s[0:3], 0x2                     ; C2090102
>	s_buffer_load_dword s16, s[0:3], 0x1                     ; C2080101
>	s_buffer_load_dword s14, s[0:3], 0x0                     ; C2070100
>	s_waitcnt vmcnt(0) lgkmcnt(0)                            ; BF8C0070
>	s_and_saveexec_b64 s[20:21], s[20:21]                    ; BE942414
>	s_xor_b64 s[20:21], exec, s[20:21]                       ; 8994147E
>	s_cbranch_execz BB0_3                                    ; BF880000
>	v_mul_f32_e32 v19, 0x40400000, v16                       ; 102620FF 40400000
>	v_cvt_i32_f32_e32 v19, v19                               ; 7E261113
>	v_lshlrev_b32_e32 v19, 4, v19                            ; 34262684
>	v_add_i32_e32 v20, 0x60, v19                             ; 4A2826FF 00000060
>	buffer_load_dword v20, v20, s[0:3], 0 offen              ; E0301000 80001414
>	v_add_i32_e32 v21, 0x64, v19                             ; 4A2A26FF 00000064
>	buffer_load_dword v21, v21, s[0:3], 0 offen              ; E0301000 80001515
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v21, v21, v2                               ; 102A0515
>	v_mad_f32 v20, v4, v20, v21                              ; D2820014 04562904
>	v_add_i32_e32 v21, 0x68, v19                             ; 4A2A26FF 00000068
>	buffer_load_dword v21, v21, s[0:3], 0 offen              ; E0301000 80001515
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v20, v5, v21, v20                              ; D2820014 04522B05
>	v_add_i32_e32 v21, 0x6c, v19                             ; 4A2A26FF 0000006C
>	buffer_load_dword v21, v21, s[0:3], 0 offen              ; E0301000 80001515
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v20, v6, v21, v20                              ; D2820014 04522B06
>	v_mad_f32 v0, v8, v20, v0                                ; D2820000 04022908
>	v_add_i32_e32 v20, 0x50, v19                             ; 4A2826FF 00000050
>	buffer_load_dword v20, v20, s[0:3], 0 offen              ; E0301000 80001414
>	v_add_i32_e32 v21, 0x54, v19                             ; 4A2A26FF 00000054
>	buffer_load_dword v21, v21, s[0:3], 0 offen              ; E0301000 80001515
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v21, v21, v2                               ; 102A0515
>	v_mad_f32 v20, v4, v20, v21                              ; D2820014 04562904
>	v_add_i32_e32 v21, 0x58, v19                             ; 4A2A26FF 00000058
>	buffer_load_dword v21, v21, s[0:3], 0 offen              ; E0301000 80001515
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v20, v5, v21, v20                              ; D2820014 04522B05
>	v_add_i32_e32 v21, 0x5c, v19                             ; 4A2A26FF 0000005C
>	buffer_load_dword v21, v21, s[0:3], 0 offen              ; E0301000 80001515
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v20, v6, v21, v20                              ; D2820014 04522B06
>	v_mad_f32 v1, v8, v20, v1                                ; D2820001 04062908
>	v_add_i32_e32 v20, 64, v19                               ; 4A2826C0
>	buffer_load_dword v20, v20, s[0:3], 0 offen              ; E0301000 80001414
>	v_add_i32_e32 v21, 0x44, v19                             ; 4A2A26FF 00000044
>	buffer_load_dword v21, v21, s[0:3], 0 offen              ; E0301000 80001515
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v21, v21, v2                               ; 102A0515
>	v_mad_f32 v20, v4, v20, v21                              ; D2820014 04562904
>	v_add_i32_e32 v21, 0x48, v19                             ; 4A2A26FF 00000048
>	buffer_load_dword v21, v21, s[0:3], 0 offen              ; E0301000 80001515
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v20, v5, v21, v20                              ; D2820014 04522B05
>	v_add_i32_e32 v19, 0x4c, v19                             ; 4A2626FF 0000004C
>	buffer_load_dword v19, v19, s[0:3], 0 offen              ; E0301000 80001313
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v19, v6, v19, v20                              ; D2820013 04522706
>	v_mad_f32 v3, v8, v19, v3                                ; D2820003 040E2708
>	v_cmp_gt_f32_e64 s[22:23], v9, 0                         ; D0080016 00010109
>	v_cndmask_b32_e64 v19, 0, -1, s[22:23]                   ; D2000813 00598280
>	v_cmp_ne_i32_e64 s[22:23], v19, 0                        ; D10A0016 00010113
>	s_and_saveexec_b64 s[22:23], s[22:23]                    ; BE962416
>	s_xor_b64 s[22:23], exec, s[22:23]                       ; 8996167E
>	s_cbranch_execz BB0_2                                    ; BF880000
>	v_mul_f32_e32 v15, 0x40400000, v17                       ; 101E22FF 40400000
>	v_cvt_i32_f32_e32 v15, v15                               ; 7E1E110F
>	v_lshlrev_b32_e32 v15, 4, v15                            ; 341E1E84
>	v_add_i32_e32 v16, 0x60, v15                             ; 4A201EFF 00000060
>	buffer_load_dword v16, v16, s[0:3], 0 offen              ; E0301000 80001010
>	v_add_i32_e32 v17, 0x64, v15                             ; 4A221EFF 00000064
>	buffer_load_dword v17, v17, s[0:3], 0 offen              ; E0301000 80001111
>	s_waitcnt vmcnt(0) expcnt(0)                             ; BF8C0700
>	v_mul_f32_e32 v17, v17, v2                               ; 10220511
>	v_mad_f32 v16, v4, v16, v17                              ; D2820010 04462104
>	v_add_i32_e32 v17, 0x68, v15                             ; 4A221EFF 00000068
>	buffer_load_dword v17, v17, s[0:3], 0 offen              ; E0301000 80001111
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v16, v5, v17, v16                              ; D2820010 04422305
>	v_add_i32_e32 v17, 0x6c, v15                             ; 4A221EFF 0000006C
>	buffer_load_dword v17, v17, s[0:3], 0 offen              ; E0301000 80001111
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v16, v6, v17, v16                              ; D2820010 04422306
>	v_mad_f32 v0, v9, v16, v0                                ; D2820000 04022109
>	v_add_i32_e32 v16, 0x50, v15                             ; 4A201EFF 00000050
>	buffer_load_dword v16, v16, s[0:3], 0 offen              ; E0301000 80001010
>	v_add_i32_e32 v17, 0x54, v15                             ; 4A221EFF 00000054
>	buffer_load_dword v17, v17, s[0:3], 0 offen              ; E0301000 80001111
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v17, v17, v2                               ; 10220511
>	v_mad_f32 v16, v4, v16, v17                              ; D2820010 04462104
>	v_add_i32_e32 v17, 0x58, v15                             ; 4A221EFF 00000058
>	buffer_load_dword v17, v17, s[0:3], 0 offen              ; E0301000 80001111
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v16, v5, v17, v16                              ; D2820010 04422305
>	v_add_i32_e32 v17, 0x5c, v15                             ; 4A221EFF 0000005C
>	buffer_load_dword v17, v17, s[0:3], 0 offen              ; E0301000 80001111
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v16, v6, v17, v16                              ; D2820010 04422306
>	v_mad_f32 v1, v9, v16, v1                                ; D2820001 04062109
>	v_add_i32_e32 v16, 64, v15                               ; 4A201EC0
>	buffer_load_dword v16, v16, s[0:3], 0 offen              ; E0301000 80001010
>	v_add_i32_e32 v17, 0x44, v15                             ; 4A221EFF 00000044
>	buffer_load_dword v17, v17, s[0:3], 0 offen              ; E0301000 80001111
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v2, v17, v2                                ; 10040511
>	v_mad_f32 v2, v4, v16, v2                                ; D2820002 040A2104
>	v_add_i32_e32 v4, 0x48, v15                              ; 4A081EFF 00000048
>	buffer_load_dword v4, v4, s[0:3], 0 offen                ; E0301000 80000404
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v2, v5, v4, v2                                 ; D2820002 040A0905
>	v_add_i32_e32 v4, 0x4c, v15                              ; 4A081EFF 0000004C
>	buffer_load_dword v4, v4, s[0:3], 0 offen                ; E0301000 80000404
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v2, v6, v4, v2                                 ; D2820002 040A0906
>	v_mad_f32 v3, v9, v2, v3                                 ; D2820003 040E0509
>	s_or_b64 exec, exec, s[22:23]                            ; 88FE167E
>	s_or_b64 exec, exec, s[20:21]                            ; 88FE147E
>	v_mov_b32_e32 v2, 1.0                                    ; 7E0402F2
>	exp 15, 32, 0, 0, 0, v11, v12, v0, v2                    ; F800020F 02000C0B
>	s_waitcnt expcnt(0)                                      ; BF8C070F
>	v_mul_f32_e32 v2, s17, v1                                ; 10040211
>	v_mad_f32 v2, v3, s19, v2                                ; D2820002 04082703
>	v_mad_f32 v2, v0, s12, v2                                ; D2820002 04081900
>	v_add_f32_e32 v2, s9, v2                                 ; 06040409
>	v_mul_f32_e32 v4, s15, v1                                ; 1008020F
>	v_mad_f32 v4, v3, s18, v4                                ; D2820004 04102503
>	v_mad_f32 v4, v0, s10, v4                                ; D2820004 04101500
>	v_add_f32_e32 v4, s7, v4                                 ; 06080807
>	v_mul_f32_e32 v5, s13, v1                                ; 100A020D
>	v_mad_f32 v5, v3, s16, v5                                ; D2820005 04142103
>	v_mad_f32 v5, v0, s8, v5                                 ; D2820005 04141100
>	v_add_f32_e32 v5, s5, v5                                 ; 060A0A05
>	v_mul_f32_e32 v1, s11, v1                                ; 1002020B
>	v_mad_f32 v1, v3, s14, v1                                ; D2820001 04041D03
>	v_mad_f32 v0, v0, s6, v1                                 ; D2820000 04040D00
>	v_add_f32_e32 v0, s4, v0                                 ; 06000004
>	exp 15, 12, 0, 1, 0, v0, v5, v4, v2                      ; F80008CF 02040500
>	s_endpgm                                                 ; BF810000
>
>SGPRS: 32
>VGPRS: 24
>Code Size: 1412 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL CONST[0]
>DCL TEMP[0..2], LOCAL
>IMM[0] FLT32 {    2.2000,     1.0000,     0.5000,     0.0000}
>IMM[1] FLT32 {    0.1000,     0.0000,     0.0000,     0.0000}
>  0: ADD TEMP[0].x, -CONST[0].xxxx, IN[0].zzzz
>  1: MOV TEMP[1].xy, IN[0].xyyy
>  2: TEX TEMP[1].w, TEMP[1], SAMP[0], 2D
>  3: POW TEMP[1].w, TEMP[1].wwww, IMM[0].yyyy
>  4: MUL TEMP[2].y, TEMP[1].wwww, IMM[0].zzzz
>  5: FSGE TEMP[0].x, TEMP[0].xxxx, IMM[0].wwww
>  6: UIF TEMP[0].xxxx :0
>  7:   MOV TEMP[0].x, TEMP[1].wwww
>  8: ELSE :0
>  9:   MOV TEMP[0].x, TEMP[2].yyyy
> 10: ENDIF
> 11: ADD TEMP[1].y, -TEMP[1].wwww, IMM[1].xxxx
> 12: FSGE TEMP[1].x, TEMP[1].yyyy, IMM[0].wwww
> 13: UIF TEMP[1].xxxx :0
> 14:   MOV TEMP[1].x, IMM[0].wwww
> 15: ELSE :0
> 16:   MOV TEMP[1].x, TEMP[0].xxxx
> 17: ENDIF
> 18: MOV TEMP[0].w, TEMP[1].xxxx
> 19: MOV TEMP[0].xyz, IMM[0].yyyy
> 20: MOV OUT[0], TEMP[0]
> 21: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %23 = load <16 x i8> addrspace(2)* %22, !tbaa !0
>  %24 = call float @llvm.SI.load.const(<16 x i8> %23, i32 0)
>  %25 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %26 = load <8 x i32> addrspace(2)* %25, !tbaa !0
>  %27 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %28 = load <4 x i32> addrspace(2)* %27, !tbaa !0
>  %29 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %30 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %31 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %5, <2 x i32> %7)
>  %32 = fsub float -0.000000e+00, %24
>  %33 = fadd float %32, %31
>  %34 = bitcast float %29 to i32
>  %35 = bitcast float %30 to i32
>  %36 = insertelement <2 x i32> undef, i32 %34, i32 0
>  %37 = insertelement <2 x i32> %36, i32 %35, i32 1
>  %38 = bitcast <8 x i32> %26 to <32 x i8>
>  %39 = bitcast <4 x i32> %28 to <16 x i8>
>  %40 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %37, <32 x i8> %38, <16 x i8> %39, i32 2)
>  %41 = extractelement <4 x float> %40, i32 3
>  %42 = call float @llvm.pow.f32(float %41, float 1.000000e+00)
>  %43 = fmul float %42, 5.000000e-01
>  %44 = fcmp oge float %33, 0.000000e+00
>  %45 = sext i1 %44 to i32
>  %46 = bitcast i32 %45 to float
>  %47 = bitcast float %46 to i32
>  %48 = icmp ne i32 %47, 0
>  %. = select i1 %48, float %42, float %43
>  %49 = fsub float -0.000000e+00, %42
>  %50 = fadd float %49, 0x3FB99999A0000000
>  %51 = fcmp oge float %50, 0.000000e+00
>  %52 = sext i1 %51 to i32
>  %53 = bitcast i32 %52 to float
>  %54 = bitcast float %53 to i32
>  %55 = icmp ne i32 %54, 0
>  %temp4.0 = select i1 %55, float 0.000000e+00, float %.
>  %56 = call i32 @llvm.SI.packf16(float 1.000000e+00, float 1.000000e+00)
>  %57 = bitcast i32 %56 to float
>  %58 = call i32 @llvm.SI.packf16(float 1.000000e+00, float %temp4.0)
>  %59 = bitcast i32 %58 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %57, float %59, float %57, float %59)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.pow.f32(float, float) #1
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                               ; BEFE0A7E
>	s_mov_b32 m0, s9                                                   ; BEFC0309
>	v_interp_p1_f32 v3, v0, 1, 0, [m0]                                 ; C80C0100
>	v_interp_p2_f32 v3, [v3], v1, 1, 0, [m0]                           ; C80D0101
>	v_interp_p1_f32 v2, v0, 0, 0, [m0]                                 ; C8080000
>	v_interp_p2_f32 v2, [v2], v1, 0, 0, [m0]                           ; C8090001
>	s_load_dwordx4 s[8:11], s[4:5], 0x0                                ; C0840500
>	s_load_dwordx8 s[12:19], s[6:7], 0x0                               ; C0C60700
>	s_waitcnt lgkmcnt(0)                                               ; BF8C007F
>	image_sample v2, 8, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[12:19], s[8:11] ; F0800800 00430202
>	s_waitcnt vmcnt(0)                                                 ; BF8C0770
>	v_log_f32_e32 v2, v2                                               ; 7E044F02
>	v_mul_legacy_f32_e32 v2, 1.0, v2                                   ; 0E0404F2
>	v_exp_f32_e32 v2, v2                                               ; 7E044B02
>	v_mul_f32_e32 v3, 0.5, v2                                          ; 100604F0
>	v_interp_p1_f32 v4, v0, 2, 0, [m0]                                 ; C8100200
>	v_interp_p2_f32 v4, [v4], v1, 2, 0, [m0]                           ; C8110201
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                                 ; C0800300
>	s_waitcnt lgkmcnt(0)                                               ; BF8C007F
>	s_buffer_load_dword s0, s[0:3], 0x0                                ; C2000100
>	s_waitcnt lgkmcnt(0)                                               ; BF8C007F
>	v_subrev_f32_e32 v0, s0, v4                                        ; 0A000800
>	v_cmp_ge_f32_e64 s[0:1], v0, 0                                     ; D00C0000 00010100
>	v_cndmask_b32_e64 v0, 0, -1, s[0:1]                                ; D2000000 00018280
>	v_cmp_ne_i32_e64 s[0:1], v0, 0                                     ; D10A0000 00010100
>	v_cndmask_b32_e64 v0, v3, v2, s[0:1]                               ; D2000000 18020503
>	v_sub_f32_e32 v1, 0x3dcccccd, v2                                   ; 080204FF 3DCCCCCD
>	v_cmp_ge_f32_e64 s[0:1], v1, 0                                     ; D00C0000 00010101
>	v_cndmask_b32_e64 v1, 0, -1, s[0:1]                                ; D2000801 00018280
>	v_cmp_ne_i32_e64 s[0:1], v1, 0                                     ; D10A0000 00010101
>	v_cndmask_b32_e64 v0, v0, 0, s[0:1]                                ; D2000000 00010100
>	v_cvt_pkrtz_f16_f32_e32 v0, 1.0, v0                                ; 5E0000F2
>	v_cvt_pkrtz_f16_f32_e64 v1, 1.0, 1.0                               ; D25E0001 0001E4F2
>	exp 15, 0, 1, 1, 1, v1, v0, v1, v0                                 ; F8001C0F 00010001
>	s_endpgm                                                           ; BF810000
>
>SGPRS: 24
>VGPRS: 8
>Code Size: 188 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL TEMP[0..1], LOCAL
>IMM[0] FLT32 {    1.0000,     0.0000,     0.0000,     0.0000}
>  0: MAD TEMP[0], IN[0].xyzx, IMM[0].xxxy, IMM[0].yyyx
>  1: MOV TEMP[1].xy, IN[1].xyxx
>  2: MOV TEMP[1].zw, IMM[0].xxyx
>  3: MOV OUT[1], TEMP[1]
>  4: MOV OUT[0], TEMP[0]
>  5: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = add i32 %5, %7
>  %14 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %12, i32 0, i32 %13)
>  %15 = extractelement <4 x float> %14, i32 0
>  %16 = extractelement <4 x float> %14, i32 1
>  %17 = extractelement <4 x float> %14, i32 2
>  %18 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %19 = load <16 x i8> addrspace(2)* %18, !tbaa !0
>  %20 = add i32 %5, %7
>  %21 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %19, i32 0, i32 %20)
>  %22 = extractelement <4 x float> %21, i32 0
>  %23 = extractelement <4 x float> %21, i32 1
>  %24 = fmul float %15, 1.000000e+00
>  %25 = fadd float %24, 0.000000e+00
>  %26 = fmul float %16, 1.000000e+00
>  %27 = fadd float %26, 0.000000e+00
>  %28 = fmul float %17, 1.000000e+00
>  %29 = fadd float %28, 0.000000e+00
>  %30 = fmul float %15, 0.000000e+00
>  %31 = fadd float %30, 1.000000e+00
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %22, float %23, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %25, float %27, float %29, float %31)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                           ; 4A00000A
>	s_load_dwordx4 s[0:3], s[8:9], 0x0                  ; C0800900
>	s_load_dwordx4 s[4:7], s[8:9], 0x4                  ; C0820904
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[4:7], 0 idxen ; E00C2000 80010100
>	v_mov_b32_e32 v5, 1.0                               ; 7E0A02F2
>	v_mov_b32_e32 v6, 0                                 ; 7E0C0280
>	s_waitcnt vmcnt(0)                                  ; BF8C0770
>	exp 15, 32, 0, 0, 0, v1, v2, v6, v5                 ; F800020F 05060201
>	s_waitcnt expcnt(0)                                 ; BF8C070F
>	buffer_load_format_xyzw v[0:3], v0, s[0:3], 0 idxen ; E00C2000 80000000
>	s_waitcnt vmcnt(0)                                  ; BF8C0770
>	v_mad_f32 v4, 0, v0, 1.0                            ; D2820004 03CA0080
>	v_add_f32_e32 v5, 0, v2                             ; 060A0480
>	v_add_f32_e32 v6, 0, v1                             ; 060C0280
>	v_add_f32_e32 v0, 0, v0                             ; 06000080
>	exp 15, 12, 0, 1, 0, v0, v6, v5, v4                 ; F80008CF 04050600
>	s_endpgm                                            ; BF810000
>
>SGPRS: 16
>VGPRS: 8
>Code Size: 92 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL CONST[0]
>DCL TEMP[0..3], LOCAL
>IMM[0] FLT32 {    1.0000,    -1.0000,     0.0000,     0.5000}
>IMM[1] FLT32 {    0.1250,     0.0000,     0.0000,     0.0000}
>  0: ADD TEMP[0].xy, CONST[0], IN[0]
>  1: MOV TEMP[0].xy, TEMP[0].xyxx
>  2: MAD TEMP[1].xy, TEMP[0], IMM[0].xyxx, IMM[0].zxzz
>  3: MOV TEMP[2].xy, TEMP[1].xyyy
>  4: TEX TEMP[2].zw, TEMP[2], SAMP[0], 2D
>  5: MOV TEMP[0].zw, TEMP[2].wwzw
>  6: ADD TEMP[3].xy, CONST[0].xwzw, IN[0]
>  7: MOV TEMP[0].xy, TEMP[3].xyxx
>  8: MAD TEMP[1].xy, TEMP[0], IMM[0].xyxx, IMM[0].zxzz
>  9: MOV TEMP[3].xy, TEMP[1].xyyy
> 10: TEX TEMP[3].w, TEMP[3], SAMP[0], 2D
> 11: ADD TEMP[2].x, TEMP[2].wwww, TEMP[3].wwww
> 12: ADD TEMP[3].yz, CONST[0].xzyw, IN[0].xxyw
> 13: MOV TEMP[0].yz, TEMP[3].zyzz
> 14: MAD TEMP[1].xy, TEMP[0].yzzw, IMM[0].xyxx, IMM[0].zxzz
> 15: MOV TEMP[3].xy, TEMP[1].xyyy
> 16: TEX TEMP[3].w, TEMP[3], SAMP[0], 2D
> 17: ADD TEMP[2].x, TEMP[2].xxxx, TEMP[3].wwww
> 18: ADD TEMP[3].yz, CONST[0].xzww, IN[0].xxyw
> 19: MOV TEMP[0].yz, TEMP[3].zyzz
> 20: MAD TEMP[1].xy, TEMP[0].yzzw, IMM[0].xyxx, IMM[0].zxzz
> 21: MOV TEMP[3].xy, TEMP[1].xyyy
> 22: TEX TEMP[3].w, TEMP[3], SAMP[0], 2D
> 23: ADD TEMP[2].x, TEMP[2].xxxx, TEMP[3].wwww
> 24: MAD TEMP[3].yz, CONST[0].xxyw, IMM[0].wwww, IN[0].xxyw
> 25: MOV TEMP[0].yz, TEMP[3].zyzz
> 26: MAD TEMP[1].xy, TEMP[0].yzzw, IMM[0].xyxx, IMM[0].zxzz
> 27: MOV TEMP[3].xy, TEMP[1].xyyy
> 28: TEX TEMP[3].w, TEMP[3], SAMP[0], 2D
> 29: ADD TEMP[2].x, TEMP[2].xxxx, TEMP[3].wwww
> 30: MAD TEMP[3].yz, CONST[0].xxww, IMM[0].wwww, IN[0].xxyw
> 31: MOV TEMP[0].yz, TEMP[3].zyzz
> 32: MAD TEMP[1].xy, TEMP[0].yzzw, IMM[0].xyxx, IMM[0].zxzz
> 33: MOV TEMP[3].xy, TEMP[1].xyyy
> 34: TEX TEMP[3].w, TEMP[3], SAMP[0], 2D
> 35: ADD TEMP[2].x, TEMP[2].xxxx, TEMP[3].wwww
> 36: MAD TEMP[3].yz, CONST[0].xzyw, IMM[0].wwww, IN[0].xxyw
> 37: MOV TEMP[0].yz, TEMP[3].zyzz
> 38: MAD TEMP[1].xy, TEMP[0].yzzw, IMM[0].xyxx, IMM[0].zxzz
> 39: MOV TEMP[3].xy, TEMP[1].xyyy
> 40: TEX TEMP[3].w, TEMP[3], SAMP[0], 2D
> 41: ADD TEMP[2].x, TEMP[2].xxxx, TEMP[3].wwww
> 42: MAD TEMP[3].yz, CONST[0].xzww, IMM[0].wwww, IN[0].xxyw
> 43: MOV TEMP[0].yz, TEMP[3].zyzz
> 44: MAD TEMP[1].xy, TEMP[0].yzzw, IMM[0].xyxx, IMM[0].zxzz
> 45: MOV TEMP[0].xy, TEMP[1].xyyy
> 46: TEX TEMP[0].w, TEMP[0], SAMP[0], 2D
> 47: ADD TEMP[0].x, TEMP[2].xxxx, TEMP[0].wwww
> 48: MUL TEMP[0], TEMP[0].xxxx, IMM[1].xxxx
> 49: MOV OUT[0], TEMP[0]
> 50: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %23 = load <16 x i8> addrspace(2)* %22, !tbaa !0
>  %24 = call float @llvm.SI.load.const(<16 x i8> %23, i32 0)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %23, i32 4)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %23, i32 8)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %23, i32 12)
>  %28 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %29 = load <8 x i32> addrspace(2)* %28, !tbaa !0
>  %30 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %31 = load <4 x i32> addrspace(2)* %30, !tbaa !0
>  %32 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %33 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %34 = fadd float %24, %32
>  %35 = fadd float %25, %33
>  %36 = fmul float %34, 1.000000e+00
>  %37 = fadd float %36, 0.000000e+00
>  %38 = fmul float %35, -1.000000e+00
>  %39 = fadd float %38, 1.000000e+00
>  %40 = bitcast float %37 to i32
>  %41 = bitcast float %39 to i32
>  %42 = insertelement <2 x i32> undef, i32 %40, i32 0
>  %43 = insertelement <2 x i32> %42, i32 %41, i32 1
>  %44 = bitcast <8 x i32> %29 to <32 x i8>
>  %45 = bitcast <4 x i32> %31 to <16 x i8>
>  %46 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %43, <32 x i8> %44, <16 x i8> %45, i32 2)
>  %47 = extractelement <4 x float> %46, i32 3
>  %48 = fadd float %24, %32
>  %49 = fadd float %27, %33
>  %50 = fmul float %48, 1.000000e+00
>  %51 = fadd float %50, 0.000000e+00
>  %52 = fmul float %49, -1.000000e+00
>  %53 = fadd float %52, 1.000000e+00
>  %54 = bitcast float %51 to i32
>  %55 = bitcast float %53 to i32
>  %56 = insertelement <2 x i32> undef, i32 %54, i32 0
>  %57 = insertelement <2 x i32> %56, i32 %55, i32 1
>  %58 = bitcast <8 x i32> %29 to <32 x i8>
>  %59 = bitcast <4 x i32> %31 to <16 x i8>
>  %60 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %57, <32 x i8> %58, <16 x i8> %59, i32 2)
>  %61 = extractelement <4 x float> %60, i32 3
>  %62 = fadd float %47, %61
>  %63 = fadd float %26, %32
>  %64 = fadd float %25, %33
>  %65 = fmul float %63, 1.000000e+00
>  %66 = fadd float %65, 0.000000e+00
>  %67 = fmul float %64, -1.000000e+00
>  %68 = fadd float %67, 1.000000e+00
>  %69 = bitcast float %66 to i32
>  %70 = bitcast float %68 to i32
>  %71 = insertelement <2 x i32> undef, i32 %69, i32 0
>  %72 = insertelement <2 x i32> %71, i32 %70, i32 1
>  %73 = bitcast <8 x i32> %29 to <32 x i8>
>  %74 = bitcast <4 x i32> %31 to <16 x i8>
>  %75 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %72, <32 x i8> %73, <16 x i8> %74, i32 2)
>  %76 = extractelement <4 x float> %75, i32 3
>  %77 = fadd float %62, %76
>  %78 = fadd float %26, %32
>  %79 = fadd float %27, %33
>  %80 = fmul float %78, 1.000000e+00
>  %81 = fadd float %80, 0.000000e+00
>  %82 = fmul float %79, -1.000000e+00
>  %83 = fadd float %82, 1.000000e+00
>  %84 = bitcast float %81 to i32
>  %85 = bitcast float %83 to i32
>  %86 = insertelement <2 x i32> undef, i32 %84, i32 0
>  %87 = insertelement <2 x i32> %86, i32 %85, i32 1
>  %88 = bitcast <8 x i32> %29 to <32 x i8>
>  %89 = bitcast <4 x i32> %31 to <16 x i8>
>  %90 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %87, <32 x i8> %88, <16 x i8> %89, i32 2)
>  %91 = extractelement <4 x float> %90, i32 3
>  %92 = fadd float %77, %91
>  %93 = fmul float %24, 5.000000e-01
>  %94 = fadd float %93, %32
>  %95 = fmul float %25, 5.000000e-01
>  %96 = fadd float %95, %33
>  %97 = fmul float %94, 1.000000e+00
>  %98 = fadd float %97, 0.000000e+00
>  %99 = fmul float %96, -1.000000e+00
>  %100 = fadd float %99, 1.000000e+00
>  %101 = bitcast float %98 to i32
>  %102 = bitcast float %100 to i32
>  %103 = insertelement <2 x i32> undef, i32 %101, i32 0
>  %104 = insertelement <2 x i32> %103, i32 %102, i32 1
>  %105 = bitcast <8 x i32> %29 to <32 x i8>
>  %106 = bitcast <4 x i32> %31 to <16 x i8>
>  %107 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %104, <32 x i8> %105, <16 x i8> %106, i32 2)
>  %108 = extractelement <4 x float> %107, i32 3
>  %109 = fadd float %92, %108
>  %110 = fmul float %24, 5.000000e-01
>  %111 = fadd float %110, %32
>  %112 = fmul float %27, 5.000000e-01
>  %113 = fadd float %112, %33
>  %114 = fmul float %111, 1.000000e+00
>  %115 = fadd float %114, 0.000000e+00
>  %116 = fmul float %113, -1.000000e+00
>  %117 = fadd float %116, 1.000000e+00
>  %118 = bitcast float %115 to i32
>  %119 = bitcast float %117 to i32
>  %120 = insertelement <2 x i32> undef, i32 %118, i32 0
>  %121 = insertelement <2 x i32> %120, i32 %119, i32 1
>  %122 = bitcast <8 x i32> %29 to <32 x i8>
>  %123 = bitcast <4 x i32> %31 to <16 x i8>
>  %124 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %121, <32 x i8> %122, <16 x i8> %123, i32 2)
>  %125 = extractelement <4 x float> %124, i32 3
>  %126 = fadd float %109, %125
>  %127 = fmul float %26, 5.000000e-01
>  %128 = fadd float %127, %32
>  %129 = fmul float %25, 5.000000e-01
>  %130 = fadd float %129, %33
>  %131 = fmul float %128, 1.000000e+00
>  %132 = fadd float %131, 0.000000e+00
>  %133 = fmul float %130, -1.000000e+00
>  %134 = fadd float %133, 1.000000e+00
>  %135 = bitcast float %132 to i32
>  %136 = bitcast float %134 to i32
>  %137 = insertelement <2 x i32> undef, i32 %135, i32 0
>  %138 = insertelement <2 x i32> %137, i32 %136, i32 1
>  %139 = bitcast <8 x i32> %29 to <32 x i8>
>  %140 = bitcast <4 x i32> %31 to <16 x i8>
>  %141 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %138, <32 x i8> %139, <16 x i8> %140, i32 2)
>  %142 = extractelement <4 x float> %141, i32 3
>  %143 = fadd float %126, %142
>  %144 = fmul float %26, 5.000000e-01
>  %145 = fadd float %144, %32
>  %146 = fmul float %27, 5.000000e-01
>  %147 = fadd float %146, %33
>  %148 = fmul float %145, 1.000000e+00
>  %149 = fadd float %148, 0.000000e+00
>  %150 = fmul float %147, -1.000000e+00
>  %151 = fadd float %150, 1.000000e+00
>  %152 = bitcast float %149 to i32
>  %153 = bitcast float %151 to i32
>  %154 = insertelement <2 x i32> undef, i32 %152, i32 0
>  %155 = insertelement <2 x i32> %154, i32 %153, i32 1
>  %156 = bitcast <8 x i32> %29 to <32 x i8>
>  %157 = bitcast <4 x i32> %31 to <16 x i8>
>  %158 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %155, <32 x i8> %156, <16 x i8> %157, i32 2)
>  %159 = extractelement <4 x float> %158, i32 3
>  %160 = fadd float %143, %159
>  %161 = fmul float %160, 1.250000e-01
>  %162 = fmul float %160, 1.250000e-01
>  %163 = fmul float %160, 1.250000e-01
>  %164 = fmul float %160, 1.250000e-01
>  %165 = call i32 @llvm.SI.packf16(float %161, float %162)
>  %166 = bitcast i32 %165 to float
>  %167 = call i32 @llvm.SI.packf16(float %163, float %164)
>  %168 = bitcast i32 %167 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %166, float %168, float %166, float %168)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                ; BEFE0A7E
>	s_mov_b32 m0, s9                                                    ; BEFC0309
>	v_interp_p1_f32 v2, v0, 1, 0, [m0]                                  ; C8080100
>	v_interp_p2_f32 v2, [v2], v1, 1, 0, [m0]                            ; C8090101
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                                  ; C0800300
>	s_waitcnt lgkmcnt(0)                                                ; BF8C007F
>	s_buffer_load_dword s8, s[0:3], 0x3                                 ; C2040103
>	s_waitcnt lgkmcnt(0)                                                ; BF8C007F
>	v_add_f32_e32 v3, s8, v2                                            ; 06060408
>	v_sub_f32_e32 v4, 1.0, v3                                           ; 080806F2
>	v_interp_p1_f32 v5, v0, 0, 0, [m0]                                  ; C8140000
>	v_interp_p2_f32 v5, [v5], v1, 0, 0, [m0]                            ; C8150001
>	s_buffer_load_dword s9, s[0:3], 0x0                                 ; C2048100
>	s_waitcnt lgkmcnt(0)                                                ; BF8C007F
>	v_add_f32_e32 v0, s9, v5                                            ; 06000A09
>	v_add_f32_e32 v3, 0, v0                                             ; 06060080
>	s_load_dwordx4 s[12:15], s[4:5], 0x0                                ; C0860500
>	s_load_dwordx8 s[16:23], s[6:7], 0x0                                ; C0C80700
>	s_waitcnt lgkmcnt(0)                                                ; BF8C007F
>	image_sample v0, 8, 0, 0, 0, 0, 0, 0, 0, v[3:4], s[16:23], s[12:15] ; F0800800 00640003
>	s_buffer_load_dword s4, s[0:3], 0x1                                 ; C2020101
>	s_waitcnt vmcnt(0) lgkmcnt(0)                                       ; BF8C0070
>	v_add_f32_e32 v1, s4, v2                                            ; 06020404
>	v_sub_f32_e32 v7, 1.0, v1                                           ; 080E02F2
>	v_mov_b32_e32 v8, v3                                                ; 7E100303
>	v_mov_b32_e32 v9, v4                                                ; 7E120304
>	v_mov_b32_e32 v9, v7                                                ; 7E120307
>	image_sample v1, 8, 0, 0, 0, 0, 0, 0, 0, v[8:9], s[16:23], s[12:15] ; F0800800 00640108
>	s_waitcnt vmcnt(0)                                                  ; BF8C0770
>	v_add_f32_e32 v0, v0, v1                                            ; 06000300
>	s_buffer_load_dword s0, s[0:3], 0x2                                 ; C2000102
>	s_waitcnt lgkmcnt(0)                                                ; BF8C007F
>	v_add_f32_e32 v1, s0, v5                                            ; 06020A00
>	v_add_f32_e32 v6, 0, v1                                             ; 060C0280
>	image_sample v1, 8, 0, 0, 0, 0, 0, 0, 0, v[6:7], s[16:23], s[12:15] ; F0800800 00640106
>	s_waitcnt vmcnt(0)                                                  ; BF8C0770
>	v_add_f32_e32 v0, v1, v0                                            ; 06000101
>	v_mov_b32_e32 v7, v4                                                ; 7E0E0304
>	image_sample v1, 8, 0, 0, 0, 0, 0, 0, 0, v[6:7], s[16:23], s[12:15] ; F0800800 00640106
>	s_waitcnt vmcnt(0)                                                  ; BF8C0770
>	v_add_f32_e32 v0, v1, v0                                            ; 06000101
>	v_mad_f32 v1, 0.5, s4, v2                                           ; D2820001 040808F0
>	v_sub_f32_e32 v4, 1.0, v1                                           ; 080802F2
>	v_mad_f32 v1, 0.5, s9, v5                                           ; D2820001 041412F0
>	v_add_f32_e32 v3, 0, v1                                             ; 06060280
>	image_sample v1, 8, 0, 0, 0, 0, 0, 0, 0, v[3:4], s[16:23], s[12:15] ; F0800800 00640103
>	s_waitcnt vmcnt(0)                                                  ; BF8C0770
>	v_add_f32_e32 v0, v1, v0                                            ; 06000101
>	v_mad_f32 v1, 0.5, s8, v2                                           ; D2820001 040810F0
>	v_sub_f32_e32 v1, 1.0, v1                                           ; 080202F2
>	v_mov_b32_e32 v6, v3                                                ; 7E0C0303
>	v_mov_b32_e32 v7, v4                                                ; 7E0E0304
>	v_mov_b32_e32 v7, v1                                                ; 7E0E0301
>	image_sample v2, 8, 0, 0, 0, 0, 0, 0, 0, v[6:7], s[16:23], s[12:15] ; F0800800 00640206
>	s_waitcnt vmcnt(0)                                                  ; BF8C0770
>	v_add_f32_e32 v0, v2, v0                                            ; 06000102
>	v_mad_f32 v2, 0.5, s0, v5                                           ; D2820002 041400F0
>	v_add_f32_e32 v3, 0, v2                                             ; 06060480
>	image_sample v2, 8, 0, 0, 0, 0, 0, 0, 0, v[3:4], s[16:23], s[12:15] ; F0800800 00640203
>	s_waitcnt vmcnt(0)                                                  ; BF8C0770
>	v_add_f32_e32 v0, v2, v0                                            ; 06000102
>	v_mov_b32_e32 v4, v1                                                ; 7E080301
>	image_sample v1, 8, 0, 0, 0, 0, 0, 0, 0, v[3:4], s[16:23], s[12:15] ; F0800800 00640103
>	s_waitcnt vmcnt(0)                                                  ; BF8C0770
>	v_add_f32_e32 v0, v1, v0                                            ; 06000101
>	v_mul_f32_e32 v0, 0x3e000000, v0                                    ; 100000FF 3E000000
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v0                                  ; 5E000100
>	exp 15, 0, 1, 1, 1, v0, v0, v0, v0                                  ; F8001C0F 00000000
>	s_endpgm                                                            ; BF810000
>
>SGPRS: 24
>VGPRS: 12
>Code Size: 332 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL CONST[0..7]
>DCL TEMP[0..1], LOCAL
>  0: MUL TEMP[0], CONST[5], IN[0].yyyy
>  1: MAD TEMP[0], IN[0].xxxx, CONST[4], TEMP[0]
>  2: ADD TEMP[0], TEMP[0], CONST[7]
>  3: MUL TEMP[1], TEMP[0].yyyy, CONST[1]
>  4: MAD TEMP[1], TEMP[0].xxxx, CONST[0], TEMP[1]
>  5: MAD TEMP[1], TEMP[0].zzzz, CONST[2], TEMP[1]
>  6: MAD TEMP[0], TEMP[0].wwww, CONST[3], TEMP[1]
>  7: MOV OUT[1], IN[1]
>  8: MOV OUT[0], TEMP[0]
>  9: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 0)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 4)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %12, i32 8)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %12, i32 12)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %12, i32 16)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %12, i32 20)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %12, i32 24)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %12, i32 28)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %12, i32 32)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %12, i32 36)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %12, i32 40)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %12, i32 44)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %12, i32 48)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %12, i32 52)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %12, i32 56)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %12, i32 60)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %12, i32 64)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %12, i32 68)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %12, i32 72)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %12, i32 76)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %12, i32 80)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %12, i32 84)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %12, i32 88)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %12, i32 92)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %12, i32 112)
>  %38 = call float @llvm.SI.load.const(<16 x i8> %12, i32 116)
>  %39 = call float @llvm.SI.load.const(<16 x i8> %12, i32 120)
>  %40 = call float @llvm.SI.load.const(<16 x i8> %12, i32 124)
>  %41 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %42 = load <16 x i8> addrspace(2)* %41, !tbaa !0
>  %43 = add i32 %5, %7
>  %44 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %42, i32 0, i32 %43)
>  %45 = extractelement <4 x float> %44, i32 0
>  %46 = extractelement <4 x float> %44, i32 1
>  %47 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %48 = load <16 x i8> addrspace(2)* %47, !tbaa !0
>  %49 = add i32 %5, %7
>  %50 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %48, i32 0, i32 %49)
>  %51 = extractelement <4 x float> %50, i32 0
>  %52 = extractelement <4 x float> %50, i32 1
>  %53 = extractelement <4 x float> %50, i32 2
>  %54 = extractelement <4 x float> %50, i32 3
>  %55 = fmul float %33, %46
>  %56 = fmul float %34, %46
>  %57 = fmul float %35, %46
>  %58 = fmul float %36, %46
>  %59 = fmul float %45, %29
>  %60 = fadd float %59, %55
>  %61 = fmul float %45, %30
>  %62 = fadd float %61, %56
>  %63 = fmul float %45, %31
>  %64 = fadd float %63, %57
>  %65 = fmul float %45, %32
>  %66 = fadd float %65, %58
>  %67 = fadd float %60, %37
>  %68 = fadd float %62, %38
>  %69 = fadd float %64, %39
>  %70 = fadd float %66, %40
>  %71 = fmul float %68, %17
>  %72 = fmul float %68, %18
>  %73 = fmul float %68, %19
>  %74 = fmul float %68, %20
>  %75 = fmul float %67, %13
>  %76 = fadd float %75, %71
>  %77 = fmul float %67, %14
>  %78 = fadd float %77, %72
>  %79 = fmul float %67, %15
>  %80 = fadd float %79, %73
>  %81 = fmul float %67, %16
>  %82 = fadd float %81, %74
>  %83 = fmul float %69, %21
>  %84 = fadd float %83, %76
>  %85 = fmul float %69, %22
>  %86 = fadd float %85, %78
>  %87 = fmul float %69, %23
>  %88 = fadd float %87, %80
>  %89 = fmul float %69, %24
>  %90 = fadd float %89, %82
>  %91 = fmul float %70, %25
>  %92 = fadd float %91, %84
>  %93 = fmul float %70, %26
>  %94 = fadd float %93, %86
>  %95 = fmul float %70, %27
>  %96 = fadd float %95, %88
>  %97 = fmul float %70, %28
>  %98 = fadd float %97, %90
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %51, float %52, float %53, float %54)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %92, float %94, float %96, float %98)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                            ; 4A00000A
>	s_load_dwordx4 s[4:7], s[8:9], 0x0                   ; C0820900
>	s_load_dwordx4 s[8:11], s[8:9], 0x4                  ; C0840904
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[8:11], 0 idxen ; E00C2000 80020100
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	exp 15, 32, 0, 0, 0, v1, v2, v3, v4                  ; F800020F 04030201
>	s_waitcnt expcnt(0)                                  ; BF8C070F
>	buffer_load_format_xyzw v[0:3], v0, s[4:7], 0 idxen  ; E00C2000 80010000
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                   ; C0800300
>	s_waitcnt vmcnt(0) lgkmcnt(0)                        ; BF8C0070
>	s_buffer_load_dword s4, s[0:3], 0x14                 ; C2020114
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v4, s4, v1                             ; 10080204
>	s_buffer_load_dword s4, s[0:3], 0x10                 ; C2020110
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v4, v0, s4, v4                             ; D2820004 04100900
>	s_buffer_load_dword s4, s[0:3], 0x1c                 ; C202011C
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_add_f32_e32 v4, s4, v4                             ; 06080804
>	s_buffer_load_dword s4, s[0:3], 0x15                 ; C2020115
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v5, s4, v1                             ; 100A0204
>	s_buffer_load_dword s4, s[0:3], 0x11                 ; C2020111
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v5, v0, s4, v5                             ; D2820005 04140900
>	s_buffer_load_dword s4, s[0:3], 0x1d                 ; C202011D
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_add_f32_e32 v5, s4, v5                             ; 060A0A04
>	s_buffer_load_dword s4, s[0:3], 0x7                  ; C2020107
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v6, s4, v5                             ; 100C0A04
>	s_buffer_load_dword s4, s[0:3], 0x3                  ; C2020103
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v6, v4, s4, v6                             ; D2820006 04180904
>	s_buffer_load_dword s4, s[0:3], 0x16                 ; C2020116
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v7, s4, v1                             ; 100E0204
>	s_buffer_load_dword s4, s[0:3], 0x12                 ; C2020112
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v7, v0, s4, v7                             ; D2820007 041C0900
>	s_buffer_load_dword s4, s[0:3], 0x1e                 ; C202011E
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_add_f32_e32 v7, s4, v7                             ; 060E0E04
>	s_buffer_load_dword s4, s[0:3], 0xb                  ; C202010B
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v6, v7, s4, v6                             ; D2820006 04180907
>	s_buffer_load_dword s4, s[0:3], 0x17                 ; C2020117
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v8, s4, v1                             ; 10100204
>	s_buffer_load_dword s4, s[0:3], 0x13                 ; C2020113
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v0, v0, s4, v8                             ; D2820000 04200900
>	s_buffer_load_dword s4, s[0:3], 0x1f                 ; C202011F
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_add_f32_e32 v0, s4, v0                             ; 06000004
>	s_buffer_load_dword s4, s[0:3], 0xf                  ; C202010F
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v1, v0, s4, v6                             ; D2820001 04180900
>	s_buffer_load_dword s4, s[0:3], 0x6                  ; C2020106
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v2, s4, v5                             ; 10040A04
>	s_buffer_load_dword s4, s[0:3], 0x2                  ; C2020102
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v2, v4, s4, v2                             ; D2820002 04080904
>	s_buffer_load_dword s4, s[0:3], 0xa                  ; C202010A
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v2, v7, s4, v2                             ; D2820002 04080907
>	s_buffer_load_dword s4, s[0:3], 0xe                  ; C202010E
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v2, v0, s4, v2                             ; D2820002 04080900
>	s_buffer_load_dword s4, s[0:3], 0x5                  ; C2020105
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v3, s4, v5                             ; 10060A04
>	s_buffer_load_dword s4, s[0:3], 0x1                  ; C2020101
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v3, v4, s4, v3                             ; D2820003 040C0904
>	s_buffer_load_dword s4, s[0:3], 0x9                  ; C2020109
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v3, v7, s4, v3                             ; D2820003 040C0907
>	s_buffer_load_dword s4, s[0:3], 0xd                  ; C202010D
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v3, v0, s4, v3                             ; D2820003 040C0900
>	s_buffer_load_dword s4, s[0:3], 0x4                  ; C2020104
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v5, s4, v5                             ; 100A0A04
>	s_buffer_load_dword s4, s[0:3], 0x0                  ; C2020100
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v4, v4, s4, v5                             ; D2820004 04140904
>	s_buffer_load_dword s4, s[0:3], 0x8                  ; C2020108
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v4, v7, s4, v4                             ; D2820004 04100907
>	s_buffer_load_dword s0, s[0:3], 0xc                  ; C200010C
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v0, v0, s0, v4                             ; D2820000 04100100
>	exp 15, 12, 0, 1, 0, v0, v3, v2, v1                  ; F80008CF 01020300
>	s_endpgm                                             ; BF810000
>
>SGPRS: 16
>VGPRS: 12
>Code Size: 468 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL CONST[0]
>DCL TEMP[0..1], LOCAL
>IMM[0] FLT32 {    1.0000,     0.0000,     0.0000,     0.0000}
>  0: MUL TEMP[0], CONST[0], IN[0]
>  1: MAD TEMP[1], TEMP[0].wwww, IMM[0].xxxy, IMM[0].yyyx
>  2: MUL TEMP[0], TEMP[0], TEMP[1]
>  3: MOV OUT[0], TEMP[0]
>  4: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %23 = load <16 x i8> addrspace(2)* %22, !tbaa !0
>  %24 = call float @llvm.SI.load.const(<16 x i8> %23, i32 0)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %23, i32 4)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %23, i32 8)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %23, i32 12)
>  %28 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %29 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %30 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %5, <2 x i32> %7)
>  %31 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %5, <2 x i32> %7)
>  %32 = fmul float %24, %28
>  %33 = fmul float %25, %29
>  %34 = fmul float %26, %30
>  %35 = fmul float %27, %31
>  %36 = fmul float %35, 1.000000e+00
>  %37 = fadd float %36, 0.000000e+00
>  %38 = fmul float %35, 1.000000e+00
>  %39 = fadd float %38, 0.000000e+00
>  %40 = fmul float %35, 1.000000e+00
>  %41 = fadd float %40, 0.000000e+00
>  %42 = fmul float %35, 0.000000e+00
>  %43 = fadd float %42, 1.000000e+00
>  %44 = fmul float %32, %37
>  %45 = fmul float %33, %39
>  %46 = fmul float %34, %41
>  %47 = fmul float %35, %43
>  %48 = call i32 @llvm.SI.packf16(float %44, float %45)
>  %49 = bitcast i32 %48 to float
>  %50 = call i32 @llvm.SI.packf16(float %46, float %47)
>  %51 = bitcast i32 %50 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %49, float %51, float %49, float %51)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                     ; BEFE0A7E
>	s_mov_b32 m0, s9                         ; BEFC0309
>	v_interp_p1_f32 v2, v0, 3, 0, [m0]       ; C8080300
>	v_interp_p2_f32 v2, [v2], v1, 3, 0, [m0] ; C8090301
>	s_load_dwordx4 s[0:3], s[2:3], 0x0       ; C0800300
>	s_waitcnt lgkmcnt(0)                     ; BF8C007F
>	s_buffer_load_dword s4, s[0:3], 0x3      ; C2020103
>	s_waitcnt lgkmcnt(0)                     ; BF8C007F
>	v_mad_f32 v3, s4, v2, 0                  ; D2820003 02020404
>	v_interp_p1_f32 v4, v0, 1, 0, [m0]       ; C8100100
>	v_interp_p2_f32 v4, [v4], v1, 1, 0, [m0] ; C8110101
>	s_buffer_load_dword s5, s[0:3], 0x1      ; C2028101
>	s_waitcnt lgkmcnt(0)                     ; BF8C007F
>	v_mul_f32_e32 v4, s5, v4                 ; 10080805
>	v_mul_f32_e32 v4, v3, v4                 ; 10080903
>	v_interp_p1_f32 v5, v0, 0, 0, [m0]       ; C8140000
>	v_interp_p2_f32 v5, [v5], v1, 0, 0, [m0] ; C8150001
>	s_buffer_load_dword s5, s[0:3], 0x0      ; C2028100
>	s_waitcnt lgkmcnt(0)                     ; BF8C007F
>	v_mul_f32_e32 v5, s5, v5                 ; 100A0A05
>	v_mul_f32_e32 v5, v3, v5                 ; 100A0B03
>	v_cvt_pkrtz_f16_f32_e32 v4, v5, v4       ; 5E080905
>	v_interp_p1_f32 v5, v0, 2, 0, [m0]       ; C8140200
>	v_interp_p2_f32 v5, [v5], v1, 2, 0, [m0] ; C8150201
>	s_buffer_load_dword s0, s[0:3], 0x2      ; C2000102
>	s_waitcnt lgkmcnt(0)                     ; BF8C007F
>	v_mul_f32_e32 v0, s0, v5                 ; 10000A00
>	v_mul_f32_e32 v0, v3, v0                 ; 10000103
>	v_mul_f32_e32 v1, s4, v2                 ; 10020404
>	v_mad_f32 v2, 0, v1, 1.0                 ; D2820002 03CA0280
>	v_mul_f32_e32 v1, v2, v1                 ; 10020302
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v1       ; 5E000300
>	exp 15, 0, 1, 1, 1, v4, v0, v4, v0       ; F8001C0F 00040004
>	s_endpgm                                 ; BF810000
>
>SGPRS: 16
>VGPRS: 8
>Code Size: 148 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL CONST[0..15]
>DCL TEMP[0..3], LOCAL
>  0: MUL TEMP[0], CONST[5], IN[0].yyyy
>  1: MAD TEMP[0], IN[0].xxxx, CONST[4], TEMP[0]
>  2: MAD TEMP[0], IN[0].zzzz, CONST[6], TEMP[0]
>  3: MAD TEMP[0], IN[0].wwww, CONST[7], TEMP[0]
>  4: MUL TEMP[1], TEMP[0].yyyy, CONST[1]
>  5: MAD TEMP[1], TEMP[0].xxxx, CONST[0], TEMP[1]
>  6: MAD TEMP[1], TEMP[0].zzzz, CONST[2], TEMP[1]
>  7: MAD TEMP[1], TEMP[0].wwww, CONST[3], TEMP[1]
>  8: MUL TEMP[2].xy, CONST[13], IN[1].yyyy
>  9: MOV TEMP[0].xy, TEMP[2].xyxx
> 10: MAD TEMP[2].xy, IN[1].xxxx, CONST[12], TEMP[0]
> 11: MOV TEMP[0].xy, TEMP[2].xyxx
> 12: ADD TEMP[2].xy, TEMP[0], CONST[15]
> 13: MOV TEMP[2].xy, TEMP[2].xyxx
> 14: MUL TEMP[3].xy, CONST[9], IN[1].yyyy
> 15: MOV TEMP[0].xy, TEMP[3].xyxx
> 16: MAD TEMP[0].xy, IN[1].xxxx, CONST[8], TEMP[0]
> 17: ADD TEMP[0].zw, TEMP[0].xyxy, CONST[11].xyxy
> 18: MOV TEMP[2].zw, TEMP[0].wwzw
> 19: MOV OUT[1], TEMP[2]
> 20: MOV OUT[0], TEMP[1]
> 21: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 0)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 4)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %12, i32 8)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %12, i32 12)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %12, i32 16)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %12, i32 20)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %12, i32 24)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %12, i32 28)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %12, i32 32)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %12, i32 36)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %12, i32 40)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %12, i32 44)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %12, i32 48)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %12, i32 52)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %12, i32 56)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %12, i32 60)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %12, i32 64)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %12, i32 68)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %12, i32 72)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %12, i32 76)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %12, i32 80)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %12, i32 84)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %12, i32 88)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %12, i32 92)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %12, i32 96)
>  %38 = call float @llvm.SI.load.const(<16 x i8> %12, i32 100)
>  %39 = call float @llvm.SI.load.const(<16 x i8> %12, i32 104)
>  %40 = call float @llvm.SI.load.const(<16 x i8> %12, i32 108)
>  %41 = call float @llvm.SI.load.const(<16 x i8> %12, i32 112)
>  %42 = call float @llvm.SI.load.const(<16 x i8> %12, i32 116)
>  %43 = call float @llvm.SI.load.const(<16 x i8> %12, i32 120)
>  %44 = call float @llvm.SI.load.const(<16 x i8> %12, i32 124)
>  %45 = call float @llvm.SI.load.const(<16 x i8> %12, i32 128)
>  %46 = call float @llvm.SI.load.const(<16 x i8> %12, i32 132)
>  %47 = call float @llvm.SI.load.const(<16 x i8> %12, i32 144)
>  %48 = call float @llvm.SI.load.const(<16 x i8> %12, i32 148)
>  %49 = call float @llvm.SI.load.const(<16 x i8> %12, i32 176)
>  %50 = call float @llvm.SI.load.const(<16 x i8> %12, i32 180)
>  %51 = call float @llvm.SI.load.const(<16 x i8> %12, i32 192)
>  %52 = call float @llvm.SI.load.const(<16 x i8> %12, i32 196)
>  %53 = call float @llvm.SI.load.const(<16 x i8> %12, i32 208)
>  %54 = call float @llvm.SI.load.const(<16 x i8> %12, i32 212)
>  %55 = call float @llvm.SI.load.const(<16 x i8> %12, i32 240)
>  %56 = call float @llvm.SI.load.const(<16 x i8> %12, i32 244)
>  %57 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %58 = load <16 x i8> addrspace(2)* %57, !tbaa !0
>  %59 = add i32 %5, %7
>  %60 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %58, i32 0, i32 %59)
>  %61 = extractelement <4 x float> %60, i32 0
>  %62 = extractelement <4 x float> %60, i32 1
>  %63 = extractelement <4 x float> %60, i32 2
>  %64 = extractelement <4 x float> %60, i32 3
>  %65 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %66 = load <16 x i8> addrspace(2)* %65, !tbaa !0
>  %67 = add i32 %5, %7
>  %68 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %66, i32 0, i32 %67)
>  %69 = extractelement <4 x float> %68, i32 0
>  %70 = extractelement <4 x float> %68, i32 1
>  %71 = fmul float %33, %62
>  %72 = fmul float %34, %62
>  %73 = fmul float %35, %62
>  %74 = fmul float %36, %62
>  %75 = fmul float %61, %29
>  %76 = fadd float %75, %71
>  %77 = fmul float %61, %30
>  %78 = fadd float %77, %72
>  %79 = fmul float %61, %31
>  %80 = fadd float %79, %73
>  %81 = fmul float %61, %32
>  %82 = fadd float %81, %74
>  %83 = fmul float %63, %37
>  %84 = fadd float %83, %76
>  %85 = fmul float %63, %38
>  %86 = fadd float %85, %78
>  %87 = fmul float %63, %39
>  %88 = fadd float %87, %80
>  %89 = fmul float %63, %40
>  %90 = fadd float %89, %82
>  %91 = fmul float %64, %41
>  %92 = fadd float %91, %84
>  %93 = fmul float %64, %42
>  %94 = fadd float %93, %86
>  %95 = fmul float %64, %43
>  %96 = fadd float %95, %88
>  %97 = fmul float %64, %44
>  %98 = fadd float %97, %90
>  %99 = fmul float %94, %17
>  %100 = fmul float %94, %18
>  %101 = fmul float %94, %19
>  %102 = fmul float %94, %20
>  %103 = fmul float %92, %13
>  %104 = fadd float %103, %99
>  %105 = fmul float %92, %14
>  %106 = fadd float %105, %100
>  %107 = fmul float %92, %15
>  %108 = fadd float %107, %101
>  %109 = fmul float %92, %16
>  %110 = fadd float %109, %102
>  %111 = fmul float %96, %21
>  %112 = fadd float %111, %104
>  %113 = fmul float %96, %22
>  %114 = fadd float %113, %106
>  %115 = fmul float %96, %23
>  %116 = fadd float %115, %108
>  %117 = fmul float %96, %24
>  %118 = fadd float %117, %110
>  %119 = fmul float %98, %25
>  %120 = fadd float %119, %112
>  %121 = fmul float %98, %26
>  %122 = fadd float %121, %114
>  %123 = fmul float %98, %27
>  %124 = fadd float %123, %116
>  %125 = fmul float %98, %28
>  %126 = fadd float %125, %118
>  %127 = fmul float %53, %70
>  %128 = fmul float %54, %70
>  %129 = fmul float %69, %51
>  %130 = fadd float %129, %127
>  %131 = fmul float %69, %52
>  %132 = fadd float %131, %128
>  %133 = fadd float %130, %55
>  %134 = fadd float %132, %56
>  %135 = fmul float %47, %70
>  %136 = fmul float %48, %70
>  %137 = fmul float %69, %45
>  %138 = fadd float %137, %135
>  %139 = fmul float %69, %46
>  %140 = fadd float %139, %136
>  %141 = fadd float %138, %49
>  %142 = fadd float %140, %50
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %133, float %134, float %141, float %142)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %120, float %122, float %124, float %126)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                            ; 4A00000A
>	s_load_dwordx4 s[4:7], s[8:9], 0x0                   ; C0820900
>	s_load_dwordx4 s[8:11], s[8:9], 0x4                  ; C0840904
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[8:11], 0 idxen ; E00C2000 80020100
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                   ; C0800300
>	s_waitcnt vmcnt(0) lgkmcnt(0)                        ; BF8C0070
>	s_buffer_load_dword s8, s[0:3], 0x25                 ; C2040125
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v5, s8, v2                             ; 100A0408
>	s_buffer_load_dword s8, s[0:3], 0x21                 ; C2040121
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v5, v1, s8, v5                             ; D2820005 04141101
>	s_buffer_load_dword s8, s[0:3], 0x2d                 ; C204012D
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_add_f32_e32 v5, s8, v5                             ; 060A0A08
>	s_buffer_load_dword s8, s[0:3], 0x24                 ; C2040124
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v6, s8, v2                             ; 100C0408
>	s_buffer_load_dword s8, s[0:3], 0x20                 ; C2040120
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v6, v1, s8, v6                             ; D2820006 04181101
>	s_buffer_load_dword s8, s[0:3], 0x2c                 ; C204012C
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_add_f32_e32 v6, s8, v6                             ; 060C0C08
>	s_buffer_load_dword s8, s[0:3], 0x35                 ; C2040135
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v7, s8, v2                             ; 100E0408
>	s_buffer_load_dword s8, s[0:3], 0x31                 ; C2040131
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v7, v1, s8, v7                             ; D2820007 041C1101
>	s_buffer_load_dword s8, s[0:3], 0x3d                 ; C204013D
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_add_f32_e32 v7, s8, v7                             ; 060E0E08
>	s_buffer_load_dword s8, s[0:3], 0x34                 ; C2040134
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v8, s8, v2                             ; 10100408
>	s_buffer_load_dword s8, s[0:3], 0x30                 ; C2040130
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v1, v1, s8, v8                             ; D2820001 04201101
>	s_buffer_load_dword s8, s[0:3], 0x3c                 ; C204013C
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_add_f32_e32 v1, s8, v1                             ; 06020208
>	exp 15, 32, 0, 0, 0, v1, v7, v6, v5                  ; F800020F 05060701
>	s_waitcnt expcnt(0)                                  ; BF8C070F
>	buffer_load_format_xyzw v[0:3], v0, s[4:7], 0 idxen  ; E00C2000 80010000
>	s_buffer_load_dword s4, s[0:3], 0x14                 ; C2020114
>	s_waitcnt vmcnt(0) lgkmcnt(0)                        ; BF8C0070
>	v_mul_f32_e32 v4, s4, v1                             ; 10080204
>	s_buffer_load_dword s4, s[0:3], 0x10                 ; C2020110
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v4, v0, s4, v4                             ; D2820004 04100900
>	s_buffer_load_dword s4, s[0:3], 0x18                 ; C2020118
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v4, v2, s4, v4                             ; D2820004 04100902
>	s_buffer_load_dword s4, s[0:3], 0x1c                 ; C202011C
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v4, v3, s4, v4                             ; D2820004 04100903
>	s_buffer_load_dword s4, s[0:3], 0x15                 ; C2020115
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v5, s4, v1                             ; 100A0204
>	s_buffer_load_dword s4, s[0:3], 0x11                 ; C2020111
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v5, v0, s4, v5                             ; D2820005 04140900
>	s_buffer_load_dword s4, s[0:3], 0x19                 ; C2020119
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v5, v2, s4, v5                             ; D2820005 04140902
>	s_buffer_load_dword s4, s[0:3], 0x1d                 ; C202011D
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v5, v3, s4, v5                             ; D2820005 04140903
>	s_buffer_load_dword s4, s[0:3], 0x7                  ; C2020107
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v6, s4, v5                             ; 100C0A04
>	s_buffer_load_dword s4, s[0:3], 0x3                  ; C2020103
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v6, v4, s4, v6                             ; D2820006 04180904
>	s_buffer_load_dword s4, s[0:3], 0x16                 ; C2020116
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v7, s4, v1                             ; 100E0204
>	s_buffer_load_dword s4, s[0:3], 0x12                 ; C2020112
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v7, v0, s4, v7                             ; D2820007 041C0900
>	s_buffer_load_dword s4, s[0:3], 0x1a                 ; C202011A
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v7, v2, s4, v7                             ; D2820007 041C0902
>	s_buffer_load_dword s4, s[0:3], 0x1e                 ; C202011E
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v7, v3, s4, v7                             ; D2820007 041C0903
>	s_buffer_load_dword s4, s[0:3], 0xb                  ; C202010B
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v6, v7, s4, v6                             ; D2820006 04180907
>	s_buffer_load_dword s4, s[0:3], 0x17                 ; C2020117
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v8, s4, v1                             ; 10100204
>	s_buffer_load_dword s4, s[0:3], 0x13                 ; C2020113
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v8, v0, s4, v8                             ; D2820008 04200900
>	s_buffer_load_dword s4, s[0:3], 0x1b                 ; C202011B
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v8, v2, s4, v8                             ; D2820008 04200902
>	s_buffer_load_dword s4, s[0:3], 0x1f                 ; C202011F
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v0, v3, s4, v8                             ; D2820000 04200903
>	s_buffer_load_dword s4, s[0:3], 0xf                  ; C202010F
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v1, v0, s4, v6                             ; D2820001 04180900
>	s_buffer_load_dword s4, s[0:3], 0x6                  ; C2020106
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v2, s4, v5                             ; 10040A04
>	s_buffer_load_dword s4, s[0:3], 0x2                  ; C2020102
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v2, v4, s4, v2                             ; D2820002 04080904
>	s_buffer_load_dword s4, s[0:3], 0xa                  ; C202010A
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v2, v7, s4, v2                             ; D2820002 04080907
>	s_buffer_load_dword s4, s[0:3], 0xe                  ; C202010E
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v2, v0, s4, v2                             ; D2820002 04080900
>	s_buffer_load_dword s4, s[0:3], 0x5                  ; C2020105
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v3, s4, v5                             ; 10060A04
>	s_buffer_load_dword s4, s[0:3], 0x1                  ; C2020101
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v3, v4, s4, v3                             ; D2820003 040C0904
>	s_buffer_load_dword s4, s[0:3], 0x9                  ; C2020109
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v3, v7, s4, v3                             ; D2820003 040C0907
>	s_buffer_load_dword s4, s[0:3], 0xd                  ; C202010D
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v3, v0, s4, v3                             ; D2820003 040C0900
>	s_buffer_load_dword s4, s[0:3], 0x4                  ; C2020104
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v5, s4, v5                             ; 100A0A04
>	s_buffer_load_dword s4, s[0:3], 0x0                  ; C2020100
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v4, v4, s4, v5                             ; D2820004 04140904
>	s_buffer_load_dword s4, s[0:3], 0x8                  ; C2020108
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v4, v7, s4, v4                             ; D2820004 04100907
>	s_buffer_load_dword s0, s[0:3], 0xc                  ; C200010C
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v0, v0, s0, v4                             ; D2820000 04100100
>	exp 15, 12, 0, 1, 0, v0, v3, v2, v1                  ; F80008CF 01020300
>	s_endpgm                                             ; BF810000
>
>SGPRS: 16
>VGPRS: 12
>Code Size: 704 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL TEMP[0..3], LOCAL
>IMM[0] FLT32 {   -0.0010,     0.0000,    -1.0000,     1.0000}
>IMM[1] FLT32 {    1.0000,    -0.0000,     0.0000,     0.0000}
>  0: MOV TEMP[0].xy, IN[0].xyyy
>  1: TEX TEMP[0], TEMP[0], SAMP[0], 2D
>  2: ADD TEMP[1].x, TEMP[0].wwww, IMM[0].xxxx
>  3: FSGE TEMP[2].x, TEMP[1].xxxx, IMM[0].yyyy
>  4: UIF TEMP[2].xxxx :0
>  5:   MOV TEMP[2].x, IMM[0].yyyy
>  6: ELSE :0
>  7:   MOV TEMP[2].x, IMM[0].zzzz
>  8: ENDIF
>  9: MOV TEMP[2].x, TEMP[2].xxxx
> 10: FSGE TEMP[3].x, TEMP[1].xxxx, IMM[0].yyyy
> 11: UIF TEMP[3].xxxx :0
> 12:   MOV TEMP[3].x, IMM[0].yyyy
> 13: ELSE :0
> 14:   MOV TEMP[3].x, IMM[0].zzzz
> 15: ENDIF
> 16: MOV TEMP[2].y, TEMP[3].xxxx
> 17: FSGE TEMP[3].x, TEMP[1].xxxx, IMM[0].yyyy
> 18: UIF TEMP[3].xxxx :0
> 19:   MOV TEMP[3].x, IMM[0].yyyy
> 20: ELSE :0
> 21:   MOV TEMP[3].x, IMM[0].zzzz
> 22: ENDIF
> 23: MOV TEMP[2].z, TEMP[3].xxxx
> 24: FSGE TEMP[3].x, TEMP[1].xxxx, IMM[0].yyyy
> 25: UIF TEMP[3].xxxx :0
> 26: ELSE :0
> 27: ENDIF
> 28: FSLT TEMP[2].xyz, TEMP[2].xyzz, IMM[0].yyyy
> 29: OR TEMP[3].x, TEMP[2].xxxx, TEMP[2].zzzz
> 30: OR TEMP[2].x, TEMP[3].xxxx, TEMP[2].yyyy
> 31: AND TEMP[2].x, TEMP[2].xxxx, IMM[0].wwww
> 32: KILL_IF -TEMP[2].xxxx
> 33: MAD TEMP[1], TEMP[0].wwww, IMM[1].xxxy, IMM[1].yyyx
> 34: MUL TEMP[0], TEMP[0], TEMP[1]
> 35: MOV OUT[0], TEMP[0]
> 36: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %23 = load <8 x i32> addrspace(2)* %22, !tbaa !0
>  %24 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %25 = load <4 x i32> addrspace(2)* %24, !tbaa !0
>  %26 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %27 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %28 = bitcast float %26 to i32
>  %29 = bitcast float %27 to i32
>  %30 = insertelement <2 x i32> undef, i32 %28, i32 0
>  %31 = insertelement <2 x i32> %30, i32 %29, i32 1
>  %32 = bitcast <8 x i32> %23 to <32 x i8>
>  %33 = bitcast <4 x i32> %25 to <16 x i8>
>  %34 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %31, <32 x i8> %32, <16 x i8> %33, i32 2)
>  %35 = extractelement <4 x float> %34, i32 0
>  %36 = extractelement <4 x float> %34, i32 1
>  %37 = extractelement <4 x float> %34, i32 2
>  %38 = extractelement <4 x float> %34, i32 3
>  %39 = fadd float %38, 0xBF50624DE0000000
>  %40 = fcmp oge float %39, 0.000000e+00
>  %41 = sext i1 %40 to i32
>  %42 = bitcast i32 %41 to float
>  %43 = bitcast float %42 to i32
>  %44 = icmp ne i32 %43, 0
>  %. = select i1 %44, float 0.000000e+00, float -1.000000e+00
>  %45 = fcmp oge float %39, 0.000000e+00
>  %46 = sext i1 %45 to i32
>  %47 = bitcast i32 %46 to float
>  %48 = bitcast float %47 to i32
>  %49 = icmp ne i32 %48, 0
>  %temp12.0 = select i1 %49, float 0.000000e+00, float -1.000000e+00
>  %50 = fcmp oge float %39, 0.000000e+00
>  %51 = sext i1 %50 to i32
>  %52 = bitcast i32 %51 to float
>  %53 = bitcast float %52 to i32
>  %54 = icmp ne i32 %53, 0
>  %.25 = select i1 %54, float 0.000000e+00, float -1.000000e+00
>  %55 = fcmp oge float %39, 0.000000e+00
>  %56 = sext i1 %55 to i32
>  %57 = bitcast i32 %56 to float
>  %58 = bitcast float %57 to i32
>  %59 = icmp ne i32 %58, 0
>  %60 = fcmp olt float %., 0.000000e+00
>  %61 = sext i1 %60 to i32
>  %62 = fcmp olt float %temp12.0, 0.000000e+00
>  %63 = sext i1 %62 to i32
>  %64 = fcmp olt float %.25, 0.000000e+00
>  %65 = sext i1 %64 to i32
>  %66 = bitcast i32 %61 to float
>  %67 = bitcast i32 %63 to float
>  %68 = bitcast i32 %65 to float
>  %69 = bitcast float %66 to i32
>  %70 = bitcast float %68 to i32
>  %71 = or i32 %69, %70
>  %72 = bitcast i32 %71 to float
>  %73 = bitcast float %72 to i32
>  %74 = bitcast float %67 to i32
>  %75 = or i32 %73, %74
>  %76 = bitcast i32 %75 to float
>  %77 = bitcast float %76 to i32
>  %78 = and i32 %77, 1065353216
>  %79 = bitcast i32 %78 to float
>  %80 = fsub float -0.000000e+00, %79
>  %81 = fsub float -0.000000e+00, %79
>  %82 = fsub float -0.000000e+00, %79
>  %83 = fsub float -0.000000e+00, %79
>  call void @llvm.AMDGPU.kill(float %80)
>  call void @llvm.AMDGPU.kill(float %81)
>  call void @llvm.AMDGPU.kill(float %82)
>  call void @llvm.AMDGPU.kill(float %83)
>  %84 = fmul float %38, 1.000000e+00
>  %85 = fadd float %84, -0.000000e+00
>  %86 = fmul float %38, 1.000000e+00
>  %87 = fadd float %86, -0.000000e+00
>  %88 = fmul float %38, 1.000000e+00
>  %89 = fadd float %88, -0.000000e+00
>  %90 = fmul float %38, -0.000000e+00
>  %91 = fadd float %90, 1.000000e+00
>  %92 = fmul float %35, %85
>  %93 = fmul float %36, %87
>  %94 = fmul float %37, %89
>  %95 = fmul float %38, %91
>  %96 = call i32 @llvm.SI.packf16(float %92, float %93)
>  %97 = bitcast i32 %96 to float
>  %98 = call i32 @llvm.SI.packf16(float %94, float %95)
>  %99 = bitcast i32 %98 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %97, float %99, float %97, float %99)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>declare void @llvm.AMDGPU.kill(float)
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                  ; BEFE0A7E
>	s_mov_b32 m0, s9                                                      ; BEFC0309
>	v_interp_p1_f32 v3, v0, 1, 0, [m0]                                    ; C80C0100
>	v_interp_p2_f32 v3, [v3], v1, 1, 0, [m0]                              ; C80D0101
>	v_interp_p1_f32 v2, v0, 0, 0, [m0]                                    ; C8080000
>	v_interp_p2_f32 v2, [v2], v1, 0, 0, [m0]                              ; C8090001
>	s_load_dwordx4 s[0:3], s[4:5], 0x0                                    ; C0800500
>	s_load_dwordx8 s[4:11], s[6:7], 0x0                                   ; C0C20700
>	s_waitcnt lgkmcnt(0)                                                  ; BF8C007F
>	image_sample v[0:3], 15, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[4:11], s[0:3] ; F0800F00 00010002
>	v_mov_b32_e32 v4, 0xba83126f                                          ; 7E0802FF BA83126F
>	s_waitcnt vmcnt(0)                                                    ; BF8C0770
>	v_add_f32_e32 v4, v3, v4                                              ; 06080903
>	v_cmp_ge_f32_e64 s[0:1], v4, 0                                        ; D00C0000 00010104
>	v_cndmask_b32_e64 v4, 0, -1, s[0:1]                                   ; D2000004 00018280
>	v_cmp_ne_i32_e64 s[0:1], v4, 0                                        ; D10A0000 00010104
>	v_cndmask_b32_e64 v4, -1.0, 0, s[0:1]                                 ; D2000004 180100F3
>	v_cmp_lt_f32_e64 s[0:1], v4, 0                                        ; D0020000 00010104
>	v_cndmask_b32_e64 v4, 0, -1, s[0:1]                                   ; D2000004 00018280
>	v_and_b32_e32 v4, 1.0, v4                                             ; 360808F2
>	v_xor_b32_e32 v4, 0x80000000, v4                                      ; 3A0808FF 80000000
>	v_cmpx_le_f32_e32 vcc, 0, v4                                          ; 7C260880
>	v_cmpx_le_f32_e32 vcc, 0, v4                                          ; 7C260880
>	v_cmpx_le_f32_e32 vcc, 0, v4                                          ; 7C260880
>	v_cmpx_le_f32_e32 vcc, 0, v4                                          ; 7C260880
>	v_mov_b32_e32 v4, 0x80000000                                          ; 7E0802FF 80000000
>	v_mad_f32 v5, v3, v4, 1.0                                             ; D2820005 03CA0903
>	v_mul_f32_e32 v5, v5, v3                                              ; 100A0705
>	v_add_f32_e32 v4, v4, v3                                              ; 06080704
>	v_mul_f32_e32 v6, v4, v2                                              ; 100C0504
>	v_cvt_pkrtz_f16_f32_e32 v5, v6, v5                                    ; 5E0A0B06
>	v_mul_f32_e32 v6, v4, v1                                              ; 100C0304
>	v_mul_f32_e32 v0, v4, v0                                              ; 10000104
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v6                                    ; 5E000D00
>	exp 15, 0, 1, 1, 1, v0, v5, v0, v5                                    ; F8001C0F 05000500
>	s_endpgm                                                              ; BF810000
>
>SGPRS: 16
>VGPRS: 8
>Code Size: 192 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL IN[2]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL CONST[0..95]
>DCL TEMP[0..2], LOCAL
>DCL ADDR[0]
>IMM[0] FLT32 {    4.0000,     0.0000,     0.0000,     0.0000}
>IMM[1] INT32 {1, 2, 3, 0}
>  0: FRC TEMP[0].x, IN[1].xxxx
>  1: ADD TEMP[1].x, -TEMP[0].xxxx, IN[1].xxxx
>  2: MUL TEMP[1].x, TEMP[1].xxxx, IMM[0].xxxx
>  3: F2I TEMP[2].x, TEMP[1].xxxx
>  4: UADD TEMP[2].x, IMM[1].xxxx, TEMP[2].xxxx
>  5: UARL ADDR[0].x, TEMP[2].xxxx
>  6: UARL ADDR[0].x, TEMP[2].xxxx
>  7: MUL TEMP[0], IN[0].yyyy, CONST[ADDR[0].x]
>  8: F2I TEMP[2].x, TEMP[1].xxxx
>  9: UARL ADDR[0].x, TEMP[2].xxxx
> 10: UARL ADDR[0].x, TEMP[2].xxxx
> 11: MAD TEMP[0], IN[0].xxxx, CONST[ADDR[0].x], TEMP[0]
> 12: F2I TEMP[2].x, TEMP[1].xxxx
> 13: UADD TEMP[2].x, IMM[1].yyyy, TEMP[2].xxxx
> 14: UARL ADDR[0].x, TEMP[2].xxxx
> 15: UARL ADDR[0].x, TEMP[2].xxxx
> 16: MAD TEMP[0], IN[0].zzzz, CONST[ADDR[0].x], TEMP[0]
> 17: F2I TEMP[1].x, TEMP[1].xxxx
> 18: UADD TEMP[1].x, IMM[1].zzzz, TEMP[1].xxxx
> 19: UARL ADDR[0].x, TEMP[1].xxxx
> 20: UARL ADDR[0].x, TEMP[1].xxxx
> 21: MAD TEMP[0], IN[0].wwww, CONST[ADDR[0].x], TEMP[0]
> 22: MOV TEMP[1].xyz, IN[2].xxxx
> 23: MOV OUT[1], TEMP[1]
> 24: MOV OUT[0], TEMP[0]
> 25: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %14 = load <16 x i8> addrspace(2)* %13, !tbaa !0
>  %15 = add i32 %5, %7
>  %16 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %14, i32 0, i32 %15)
>  %17 = extractelement <4 x float> %16, i32 0
>  %18 = extractelement <4 x float> %16, i32 1
>  %19 = extractelement <4 x float> %16, i32 2
>  %20 = extractelement <4 x float> %16, i32 3
>  %21 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %22 = load <16 x i8> addrspace(2)* %21, !tbaa !0
>  %23 = add i32 %5, %7
>  %24 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %22, i32 0, i32 %23)
>  %25 = extractelement <4 x float> %24, i32 0
>  %26 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 2
>  %27 = load <16 x i8> addrspace(2)* %26, !tbaa !0
>  %28 = add i32 %5, %7
>  %29 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %27, i32 0, i32 %28)
>  %30 = extractelement <4 x float> %29, i32 0
>  %31 = call float @llvm.AMDIL.fraction.(float %25)
>  %32 = fsub float -0.000000e+00, %31
>  %33 = fadd float %32, %25
>  %34 = fmul float %33, 4.000000e+00
>  %35 = fptosi float %34 to i32
>  %36 = bitcast i32 %35 to float
>  %37 = bitcast float %36 to i32
>  %38 = add i32 1, %37
>  %39 = bitcast i32 %38 to float
>  %40 = bitcast float %39 to i32
>  %41 = shl i32 %40, 4
>  %42 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %41)
>  %43 = fmul float %18, %42
>  %44 = shl i32 %40, 4
>  %45 = add i32 %44, 4
>  %46 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %45)
>  %47 = fmul float %18, %46
>  %48 = shl i32 %40, 4
>  %49 = add i32 %48, 8
>  %50 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %49)
>  %51 = fmul float %18, %50
>  %52 = shl i32 %40, 4
>  %53 = add i32 %52, 12
>  %54 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %53)
>  %55 = fmul float %18, %54
>  %56 = fptosi float %34 to i32
>  %57 = bitcast i32 %56 to float
>  %58 = bitcast float %57 to i32
>  %59 = shl i32 %58, 4
>  %60 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %59)
>  %61 = fmul float %17, %60
>  %62 = fadd float %61, %43
>  %63 = shl i32 %58, 4
>  %64 = add i32 %63, 4
>  %65 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %64)
>  %66 = fmul float %17, %65
>  %67 = fadd float %66, %47
>  %68 = shl i32 %58, 4
>  %69 = add i32 %68, 8
>  %70 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %69)
>  %71 = fmul float %17, %70
>  %72 = fadd float %71, %51
>  %73 = shl i32 %58, 4
>  %74 = add i32 %73, 12
>  %75 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %74)
>  %76 = fmul float %17, %75
>  %77 = fadd float %76, %55
>  %78 = fptosi float %34 to i32
>  %79 = bitcast i32 %78 to float
>  %80 = bitcast float %79 to i32
>  %81 = add i32 2, %80
>  %82 = bitcast i32 %81 to float
>  %83 = bitcast float %82 to i32
>  %84 = shl i32 %83, 4
>  %85 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %84)
>  %86 = fmul float %19, %85
>  %87 = fadd float %86, %62
>  %88 = shl i32 %83, 4
>  %89 = add i32 %88, 4
>  %90 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %89)
>  %91 = fmul float %19, %90
>  %92 = fadd float %91, %67
>  %93 = shl i32 %83, 4
>  %94 = add i32 %93, 8
>  %95 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %94)
>  %96 = fmul float %19, %95
>  %97 = fadd float %96, %72
>  %98 = shl i32 %83, 4
>  %99 = add i32 %98, 12
>  %100 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %99)
>  %101 = fmul float %19, %100
>  %102 = fadd float %101, %77
>  %103 = fptosi float %34 to i32
>  %104 = bitcast i32 %103 to float
>  %105 = bitcast float %104 to i32
>  %106 = add i32 3, %105
>  %107 = bitcast i32 %106 to float
>  %108 = bitcast float %107 to i32
>  %109 = shl i32 %108, 4
>  %110 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %109)
>  %111 = fmul float %20, %110
>  %112 = fadd float %111, %87
>  %113 = shl i32 %108, 4
>  %114 = add i32 %113, 4
>  %115 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %114)
>  %116 = fmul float %20, %115
>  %117 = fadd float %116, %92
>  %118 = shl i32 %108, 4
>  %119 = add i32 %118, 8
>  %120 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %119)
>  %121 = fmul float %20, %120
>  %122 = fadd float %121, %97
>  %123 = shl i32 %108, 4
>  %124 = add i32 %123, 12
>  %125 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %124)
>  %126 = fmul float %20, %125
>  %127 = fadd float %126, %102
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %30, float %30, float %30, float 0.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %112, float %117, float %122, float %127)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.fraction.(float) #2
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                             ; 4A00000A
>	s_load_dwordx4 s[4:7], s[8:9], 0x0                    ; C0820900
>	s_load_dwordx4 s[12:15], s[8:9], 0x4                  ; C0860904
>	s_load_dwordx4 s[8:11], s[8:9], 0x8                   ; C0840908
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[8:11], 0 idxen  ; E00C2000 80020100
>	v_mov_b32_e32 v5, 0                                   ; 7E0A0280
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	exp 15, 32, 0, 0, 0, v1, v1, v1, v5                   ; F800020F 05010101
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	buffer_load_format_xyzw v[1:4], v0, s[12:15], 0 idxen ; E00C2000 80030100
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_fract_f32_e32 v5, v1                                ; 7E0A4101
>	v_subrev_f32_e32 v1, v5, v1                           ; 0A020305
>	v_mul_f32_e32 v1, 4.0, v1                             ; 100202F6
>	v_cvt_i32_f32_e32 v1, v1                              ; 7E021101
>	v_lshlrev_b32_e32 v1, 4, v1                           ; 34020284
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                    ; C0800300
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	buffer_load_dword v2, v1, s[0:3], 0 offen             ; E0301000 80000201
>	v_add_i32_e32 v3, 16, v1                              ; 4A060290
>	buffer_load_dword v4, v3, s[0:3], 0 offen             ; E0301000 80000403
>	buffer_load_format_xyzw v[5:8], v0, s[4:7], 0 idxen   ; E00C2000 80010500
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mul_f32_e32 v0, v4, v6                              ; 10000D04
>	v_mad_f32 v0, v5, v2, v0                              ; D2820000 04020505
>	v_add_i32_e32 v2, 32, v1                              ; 4A0402A0
>	buffer_load_dword v4, v2, s[0:3], 0 offen             ; E0301000 80000402
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mad_f32 v0, v7, v4, v0                              ; D2820000 04020907
>	v_add_i32_e32 v4, 48, v1                              ; 4A0802B0
>	buffer_load_dword v9, v4, s[0:3], 0 offen             ; E0301000 80000904
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mad_f32 v0, v8, v9, v0                              ; D2820000 04021308
>	v_or_b32_e32 v9, 12, v1                               ; 3812028C
>	buffer_load_dword v9, v9, s[0:3], 0 offen             ; E0301000 80000909
>	v_or_b32_e32 v10, 12, v3                              ; 3814068C
>	buffer_load_dword v10, v10, s[0:3], 0 offen           ; E0301000 80000A0A
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mul_f32_e32 v10, v10, v6                            ; 10140D0A
>	v_mad_f32 v9, v5, v9, v10                             ; D2820009 042A1305
>	v_or_b32_e32 v10, 12, v2                              ; 3814048C
>	buffer_load_dword v10, v10, s[0:3], 0 offen           ; E0301000 80000A0A
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mad_f32 v9, v7, v10, v9                             ; D2820009 04261507
>	v_or_b32_e32 v10, 12, v4                              ; 3814088C
>	buffer_load_dword v10, v10, s[0:3], 0 offen           ; E0301000 80000A0A
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mad_f32 v9, v8, v10, v9                             ; D2820009 04261508
>	v_or_b32_e32 v10, 8, v1                               ; 38140288
>	buffer_load_dword v10, v10, s[0:3], 0 offen           ; E0301000 80000A0A
>	v_or_b32_e32 v11, 8, v3                               ; 38160688
>	buffer_load_dword v11, v11, s[0:3], 0 offen           ; E0301000 80000B0B
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mul_f32_e32 v11, v11, v6                            ; 10160D0B
>	v_mad_f32 v10, v5, v10, v11                           ; D282000A 042E1505
>	v_or_b32_e32 v11, 8, v2                               ; 38160488
>	buffer_load_dword v11, v11, s[0:3], 0 offen           ; E0301000 80000B0B
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mad_f32 v10, v7, v11, v10                           ; D282000A 042A1707
>	v_or_b32_e32 v11, 8, v4                               ; 38160888
>	buffer_load_dword v11, v11, s[0:3], 0 offen           ; E0301000 80000B0B
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mad_f32 v10, v8, v11, v10                           ; D282000A 042A1708
>	v_or_b32_e32 v1, 4, v1                                ; 38020284
>	buffer_load_dword v1, v1, s[0:3], 0 offen             ; E0301000 80000101
>	v_or_b32_e32 v3, 4, v3                                ; 38060684
>	buffer_load_dword v3, v3, s[0:3], 0 offen             ; E0301000 80000303
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mul_f32_e32 v3, v3, v6                              ; 10060D03
>	v_mad_f32 v1, v5, v1, v3                              ; D2820001 040E0305
>	v_or_b32_e32 v2, 4, v2                                ; 38040484
>	buffer_load_dword v2, v2, s[0:3], 0 offen             ; E0301000 80000202
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mad_f32 v1, v7, v2, v1                              ; D2820001 04060507
>	v_or_b32_e32 v2, 4, v4                                ; 38040884
>	buffer_load_dword v2, v2, s[0:3], 0 offen             ; E0301000 80000202
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mad_f32 v1, v8, v2, v1                              ; D2820001 04060508
>	exp 15, 12, 0, 1, 0, v0, v1, v10, v9                  ; F80008CF 090A0100
>	s_endpgm                                              ; BF810000
>
>SGPRS: 24
>VGPRS: 12
>Code Size: 456 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL TEMP[0], LOCAL
>IMM[0] FLT32 {    1.0000,     0.0000,     0.0000,     0.0000}
>  0: MAD TEMP[0], IN[0].xxxx, IMM[0].xyyy, IMM[0].yyyx
>  1: MOV OUT[0], TEMP[0]
>  2: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %23 = fmul float %22, 1.000000e+00
>  %24 = fadd float %23, 0.000000e+00
>  %25 = fmul float %22, 0.000000e+00
>  %26 = fadd float %25, 0.000000e+00
>  %27 = fmul float %22, 0.000000e+00
>  %28 = fadd float %27, 0.000000e+00
>  %29 = fmul float %22, 0.000000e+00
>  %30 = fadd float %29, 1.000000e+00
>  %31 = call i32 @llvm.SI.packf16(float %24, float %26)
>  %32 = bitcast i32 %31 to float
>  %33 = call i32 @llvm.SI.packf16(float %28, float %30)
>  %34 = bitcast i32 %33 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %32, float %34, float %32, float %34)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                     ; BEFE0A7E
>	s_mov_b32 m0, s9                         ; BEFC0309
>	v_interp_p1_f32 v2, v0, 0, 0, [m0]       ; C8080000
>	v_interp_p2_f32 v2, [v2], v1, 0, 0, [m0] ; C8090001
>	v_mad_f32 v0, 0, v2, 1.0                 ; D2820000 03CA0480
>	v_mad_f32 v1, 0, v2, 0                   ; D2820001 02020480
>	v_cvt_pkrtz_f16_f32_e32 v0, v1, v0       ; 5E000101
>	v_add_f32_e32 v2, 0, v2                  ; 06040480
>	v_cvt_pkrtz_f16_f32_e32 v1, v2, v1       ; 5E020302
>	exp 15, 0, 1, 1, 1, v1, v0, v1, v0       ; F8001C0F 00010001
>	s_endpgm                                 ; BF810000
>
>SGPRS: 16
>VGPRS: 4
>Code Size: 56 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL IN[2]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL CONST[0..95]
>DCL TEMP[0..2], LOCAL
>DCL ADDR[0]
>IMM[0] FLT32 {    4.0000,     0.0000,     0.0000,     0.0000}
>IMM[1] INT32 {1, 2, 3, 0}
>  0: FRC TEMP[0].x, IN[1].xxxx
>  1: ADD TEMP[1].x, -TEMP[0].xxxx, IN[1].xxxx
>  2: MUL TEMP[1].x, TEMP[1].xxxx, IMM[0].xxxx
>  3: F2I TEMP[2].x, TEMP[1].xxxx
>  4: UADD TEMP[2].x, IMM[1].xxxx, TEMP[2].xxxx
>  5: UARL ADDR[0].x, TEMP[2].xxxx
>  6: UARL ADDR[0].x, TEMP[2].xxxx
>  7: MUL TEMP[0], IN[0].yyyy, CONST[ADDR[0].x]
>  8: F2I TEMP[2].x, TEMP[1].xxxx
>  9: UARL ADDR[0].x, TEMP[2].xxxx
> 10: UARL ADDR[0].x, TEMP[2].xxxx
> 11: MAD TEMP[0], IN[0].xxxx, CONST[ADDR[0].x], TEMP[0]
> 12: F2I TEMP[2].x, TEMP[1].xxxx
> 13: UADD TEMP[2].x, IMM[1].yyyy, TEMP[2].xxxx
> 14: UARL ADDR[0].x, TEMP[2].xxxx
> 15: UARL ADDR[0].x, TEMP[2].xxxx
> 16: MAD TEMP[0], IN[0].zzzz, CONST[ADDR[0].x], TEMP[0]
> 17: F2I TEMP[1].x, TEMP[1].xxxx
> 18: UADD TEMP[1].x, IMM[1].zzzz, TEMP[1].xxxx
> 19: UARL ADDR[0].x, TEMP[1].xxxx
> 20: UARL ADDR[0].x, TEMP[1].xxxx
> 21: MAD TEMP[0], IN[0].wwww, CONST[ADDR[0].x], TEMP[0]
> 22: MOV TEMP[1].xyz, IN[2].xxxx
> 23: MOV OUT[1], TEMP[1]
> 24: MOV OUT[0], TEMP[0]
> 25: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %14 = load <16 x i8> addrspace(2)* %13, !tbaa !0
>  %15 = add i32 %5, %7
>  %16 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %14, i32 0, i32 %15)
>  %17 = extractelement <4 x float> %16, i32 0
>  %18 = extractelement <4 x float> %16, i32 1
>  %19 = extractelement <4 x float> %16, i32 2
>  %20 = extractelement <4 x float> %16, i32 3
>  %21 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %22 = load <16 x i8> addrspace(2)* %21, !tbaa !0
>  %23 = add i32 %5, %7
>  %24 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %22, i32 0, i32 %23)
>  %25 = extractelement <4 x float> %24, i32 0
>  %26 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 2
>  %27 = load <16 x i8> addrspace(2)* %26, !tbaa !0
>  %28 = add i32 %5, %7
>  %29 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %27, i32 0, i32 %28)
>  %30 = extractelement <4 x float> %29, i32 0
>  %31 = call float @llvm.AMDIL.fraction.(float %25)
>  %32 = fsub float -0.000000e+00, %31
>  %33 = fadd float %32, %25
>  %34 = fmul float %33, 4.000000e+00
>  %35 = fptosi float %34 to i32
>  %36 = bitcast i32 %35 to float
>  %37 = bitcast float %36 to i32
>  %38 = add i32 1, %37
>  %39 = bitcast i32 %38 to float
>  %40 = bitcast float %39 to i32
>  %41 = shl i32 %40, 4
>  %42 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %41)
>  %43 = fmul float %18, %42
>  %44 = shl i32 %40, 4
>  %45 = add i32 %44, 4
>  %46 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %45)
>  %47 = fmul float %18, %46
>  %48 = shl i32 %40, 4
>  %49 = add i32 %48, 8
>  %50 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %49)
>  %51 = fmul float %18, %50
>  %52 = shl i32 %40, 4
>  %53 = add i32 %52, 12
>  %54 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %53)
>  %55 = fmul float %18, %54
>  %56 = fptosi float %34 to i32
>  %57 = bitcast i32 %56 to float
>  %58 = bitcast float %57 to i32
>  %59 = shl i32 %58, 4
>  %60 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %59)
>  %61 = fmul float %17, %60
>  %62 = fadd float %61, %43
>  %63 = shl i32 %58, 4
>  %64 = add i32 %63, 4
>  %65 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %64)
>  %66 = fmul float %17, %65
>  %67 = fadd float %66, %47
>  %68 = shl i32 %58, 4
>  %69 = add i32 %68, 8
>  %70 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %69)
>  %71 = fmul float %17, %70
>  %72 = fadd float %71, %51
>  %73 = shl i32 %58, 4
>  %74 = add i32 %73, 12
>  %75 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %74)
>  %76 = fmul float %17, %75
>  %77 = fadd float %76, %55
>  %78 = fptosi float %34 to i32
>  %79 = bitcast i32 %78 to float
>  %80 = bitcast float %79 to i32
>  %81 = add i32 2, %80
>  %82 = bitcast i32 %81 to float
>  %83 = bitcast float %82 to i32
>  %84 = shl i32 %83, 4
>  %85 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %84)
>  %86 = fmul float %19, %85
>  %87 = fadd float %86, %62
>  %88 = shl i32 %83, 4
>  %89 = add i32 %88, 4
>  %90 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %89)
>  %91 = fmul float %19, %90
>  %92 = fadd float %91, %67
>  %93 = shl i32 %83, 4
>  %94 = add i32 %93, 8
>  %95 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %94)
>  %96 = fmul float %19, %95
>  %97 = fadd float %96, %72
>  %98 = shl i32 %83, 4
>  %99 = add i32 %98, 12
>  %100 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %99)
>  %101 = fmul float %19, %100
>  %102 = fadd float %101, %77
>  %103 = fptosi float %34 to i32
>  %104 = bitcast i32 %103 to float
>  %105 = bitcast float %104 to i32
>  %106 = add i32 3, %105
>  %107 = bitcast i32 %106 to float
>  %108 = bitcast float %107 to i32
>  %109 = shl i32 %108, 4
>  %110 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %109)
>  %111 = fmul float %20, %110
>  %112 = fadd float %111, %87
>  %113 = shl i32 %108, 4
>  %114 = add i32 %113, 4
>  %115 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %114)
>  %116 = fmul float %20, %115
>  %117 = fadd float %116, %92
>  %118 = shl i32 %108, 4
>  %119 = add i32 %118, 8
>  %120 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %119)
>  %121 = fmul float %20, %120
>  %122 = fadd float %121, %97
>  %123 = shl i32 %108, 4
>  %124 = add i32 %123, 12
>  %125 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %124)
>  %126 = fmul float %20, %125
>  %127 = fadd float %126, %102
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %30, float %30, float %30, float 0.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %112, float %117, float %122, float %127)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.fraction.(float) #2
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                             ; 4A00000A
>	s_load_dwordx4 s[4:7], s[8:9], 0x0                    ; C0820900
>	s_load_dwordx4 s[12:15], s[8:9], 0x4                  ; C0860904
>	s_load_dwordx4 s[8:11], s[8:9], 0x8                   ; C0840908
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[8:11], 0 idxen  ; E00C2000 80020100
>	v_mov_b32_e32 v5, 0                                   ; 7E0A0280
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	exp 15, 32, 0, 0, 0, v1, v1, v1, v5                   ; F800020F 05010101
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	buffer_load_format_xyzw v[1:4], v0, s[12:15], 0 idxen ; E00C2000 80030100
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_fract_f32_e32 v5, v1                                ; 7E0A4101
>	v_subrev_f32_e32 v1, v5, v1                           ; 0A020305
>	v_mul_f32_e32 v1, 4.0, v1                             ; 100202F6
>	v_cvt_i32_f32_e32 v1, v1                              ; 7E021101
>	v_lshlrev_b32_e32 v1, 4, v1                           ; 34020284
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                    ; C0800300
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	buffer_load_dword v2, v1, s[0:3], 0 offen             ; E0301000 80000201
>	v_add_i32_e32 v3, 16, v1                              ; 4A060290
>	buffer_load_dword v4, v3, s[0:3], 0 offen             ; E0301000 80000403
>	buffer_load_format_xyzw v[5:8], v0, s[4:7], 0 idxen   ; E00C2000 80010500
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mul_f32_e32 v0, v4, v6                              ; 10000D04
>	v_mad_f32 v0, v5, v2, v0                              ; D2820000 04020505
>	v_add_i32_e32 v2, 32, v1                              ; 4A0402A0
>	buffer_load_dword v4, v2, s[0:3], 0 offen             ; E0301000 80000402
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mad_f32 v0, v7, v4, v0                              ; D2820000 04020907
>	v_add_i32_e32 v4, 48, v1                              ; 4A0802B0
>	buffer_load_dword v9, v4, s[0:3], 0 offen             ; E0301000 80000904
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mad_f32 v0, v8, v9, v0                              ; D2820000 04021308
>	v_or_b32_e32 v9, 12, v1                               ; 3812028C
>	buffer_load_dword v9, v9, s[0:3], 0 offen             ; E0301000 80000909
>	v_or_b32_e32 v10, 12, v3                              ; 3814068C
>	buffer_load_dword v10, v10, s[0:3], 0 offen           ; E0301000 80000A0A
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mul_f32_e32 v10, v10, v6                            ; 10140D0A
>	v_mad_f32 v9, v5, v9, v10                             ; D2820009 042A1305
>	v_or_b32_e32 v10, 12, v2                              ; 3814048C
>	buffer_load_dword v10, v10, s[0:3], 0 offen           ; E0301000 80000A0A
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mad_f32 v9, v7, v10, v9                             ; D2820009 04261507
>	v_or_b32_e32 v10, 12, v4                              ; 3814088C
>	buffer_load_dword v10, v10, s[0:3], 0 offen           ; E0301000 80000A0A
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mad_f32 v9, v8, v10, v9                             ; D2820009 04261508
>	v_or_b32_e32 v10, 8, v1                               ; 38140288
>	buffer_load_dword v10, v10, s[0:3], 0 offen           ; E0301000 80000A0A
>	v_or_b32_e32 v11, 8, v3                               ; 38160688
>	buffer_load_dword v11, v11, s[0:3], 0 offen           ; E0301000 80000B0B
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mul_f32_e32 v11, v11, v6                            ; 10160D0B
>	v_mad_f32 v10, v5, v10, v11                           ; D282000A 042E1505
>	v_or_b32_e32 v11, 8, v2                               ; 38160488
>	buffer_load_dword v11, v11, s[0:3], 0 offen           ; E0301000 80000B0B
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mad_f32 v10, v7, v11, v10                           ; D282000A 042A1707
>	v_or_b32_e32 v11, 8, v4                               ; 38160888
>	buffer_load_dword v11, v11, s[0:3], 0 offen           ; E0301000 80000B0B
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mad_f32 v10, v8, v11, v10                           ; D282000A 042A1708
>	v_or_b32_e32 v1, 4, v1                                ; 38020284
>	buffer_load_dword v1, v1, s[0:3], 0 offen             ; E0301000 80000101
>	v_or_b32_e32 v3, 4, v3                                ; 38060684
>	buffer_load_dword v3, v3, s[0:3], 0 offen             ; E0301000 80000303
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mul_f32_e32 v3, v3, v6                              ; 10060D03
>	v_mad_f32 v1, v5, v1, v3                              ; D2820001 040E0305
>	v_or_b32_e32 v2, 4, v2                                ; 38040484
>	buffer_load_dword v2, v2, s[0:3], 0 offen             ; E0301000 80000202
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mad_f32 v1, v7, v2, v1                              ; D2820001 04060507
>	v_or_b32_e32 v2, 4, v4                                ; 38040884
>	buffer_load_dword v2, v2, s[0:3], 0 offen             ; E0301000 80000202
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mad_f32 v1, v8, v2, v1                              ; D2820001 04060508
>	exp 15, 12, 0, 1, 0, v0, v1, v10, v9                  ; F80008CF 090A0100
>	s_endpgm                                              ; BF810000
>
>SGPRS: 24
>VGPRS: 12
>Code Size: 456 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL TEMP[0], LOCAL
>IMM[0] FLT32 {    1.0000,     0.0000,     0.0000,     0.0000}
>  0: MAD TEMP[0], IN[0].xxxx, IMM[0].xyxy, IMM[0].yyyx
>  1: MOV OUT[0], TEMP[0]
>  2: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %23 = fmul float %22, 1.000000e+00
>  %24 = fadd float %23, 0.000000e+00
>  %25 = fmul float %22, 0.000000e+00
>  %26 = fadd float %25, 0.000000e+00
>  %27 = fmul float %22, 1.000000e+00
>  %28 = fadd float %27, 0.000000e+00
>  %29 = fmul float %22, 0.000000e+00
>  %30 = fadd float %29, 1.000000e+00
>  %31 = call i32 @llvm.SI.packf16(float %24, float %26)
>  %32 = bitcast i32 %31 to float
>  %33 = call i32 @llvm.SI.packf16(float %28, float %30)
>  %34 = bitcast i32 %33 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %32, float %34, float %32, float %34)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                     ; BEFE0A7E
>	s_mov_b32 m0, s9                         ; BEFC0309
>	v_interp_p1_f32 v2, v0, 0, 0, [m0]       ; C8080000
>	v_interp_p2_f32 v2, [v2], v1, 0, 0, [m0] ; C8090001
>	v_mad_f32 v0, 0, v2, 1.0                 ; D2820000 03CA0480
>	v_add_f32_e32 v1, 0, v2                  ; 06020480
>	v_cvt_pkrtz_f16_f32_e32 v0, v1, v0       ; 5E000101
>	v_mad_f32 v2, 0, v2, 0                   ; D2820002 02020480
>	v_cvt_pkrtz_f16_f32_e32 v1, v1, v2       ; 5E020501
>	exp 15, 0, 1, 1, 1, v1, v0, v1, v0       ; F8001C0F 00010001
>	s_endpgm                                 ; BF810000
>
>SGPRS: 16
>VGPRS: 4
>Code Size: 56 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL OUT[2], GENERIC[10]
>DCL OUT[3], GENERIC[11]
>DCL OUT[4], GENERIC[12]
>DCL SAMP[0]
>DCL CONST[0..15]
>DCL TEMP[0..4], LOCAL
>IMM[0] FLT32 {    0.5000,     0.0000,   255.0000,     1.0000}
>IMM[1] FLT32 {    0.5000,    -0.5000,     1.0000,     0.0000}
>  0: ADD TEMP[0].xy, CONST[15], IN[0]
>  1: MOV TEMP[1].xy, TEMP[0].xyxx
>  2: ADD TEMP[0].zw, TEMP[0].xyxy, CONST[15]
>  3: MOV TEMP[1].zw, TEMP[0].wwzw
>  4: MUL TEMP[0].xy, TEMP[1], CONST[9]
>  5: MOV TEMP[1].xy, TEMP[0].xyxx
>  6: ADD TEMP[2].zw, TEMP[1], IMM[0].xxxx
>  7: MUL TEMP[2].xy, TEMP[2].zwzw, CONST[10]
>  8: MOV TEMP[3].xy, TEMP[2].xyyy
>  9: MOV TEMP[3].w, IMM[0].yyyy
> 10: TXL TEMP[3].w, TEMP[3], SAMP[0], 2D
> 11: MUL TEMP[3].z, TEMP[3].wwww, CONST[9].zzzz
> 12: MUL TEMP[3].z, TEMP[3].zzzz, IMM[0].zzzz
> 13: MUL TEMP[2], TEMP[0].yyyy, CONST[12]
> 14: MAD TEMP[2], TEMP[0].xxxx, CONST[11], TEMP[2]
> 15: MAD TEMP[1], TEMP[3].zzzz, CONST[13], TEMP[2]
> 16: ADD TEMP[1], TEMP[1], CONST[14]
> 17: ADD TEMP[0].xy, TEMP[1], CONST[8].zwzw
> 18: MUL TEMP[0].zw, TEMP[0].xyxy, CONST[8].xyxy
> 19: MOV TEMP[0].zw, TEMP[0].wwzw
> 20: MUL TEMP[2], TEMP[1].yyyy, CONST[1]
> 21: MAD TEMP[2], TEMP[1].xxxx, CONST[0], TEMP[2]
> 22: MAD TEMP[2], TEMP[1].zzzz, CONST[2], TEMP[2]
> 23: MAD TEMP[2], TEMP[1].wwww, CONST[3], TEMP[2]
> 24: RCP TEMP[3].x, TEMP[2].wwww
> 25: MUL TEMP[3].xy, TEMP[2], TEMP[3].xxxx
> 26: MOV TEMP[3].xy, TEMP[3].xyxx
> 27: MOV TEMP[4], TEMP[2]
> 28: MOV TEMP[0].xy, IN[1].xyxx
> 29: MUL TEMP[2], TEMP[1].yyyy, CONST[5]
> 30: MAD TEMP[2], TEMP[1].xxxx, CONST[4], TEMP[2]
> 31: MAD TEMP[2], TEMP[1].zzzz, CONST[6], TEMP[2]
> 32: MAD TEMP[2], TEMP[1].wwww, CONST[7], TEMP[2]
> 33: MOV TEMP[1].xyz, TEMP[1].xyzx
> 34: MAD TEMP[2], TEMP[2], IMM[1].xyzz, IMM[0].xxyy
> 35: MOV TEMP[1].w, IMM[0].wwww
> 36: MOV TEMP[3].zw, IMM[0].wwyw
> 37: MOV OUT[1], TEMP[0]
> 38: MOV OUT[2], TEMP[2]
> 39: MOV OUT[0], TEMP[4]
> 40: MOV OUT[3], TEMP[1]
> 41: MOV OUT[4], TEMP[3]
> 42: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 0)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 4)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %12, i32 8)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %12, i32 12)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %12, i32 16)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %12, i32 20)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %12, i32 24)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %12, i32 28)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %12, i32 32)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %12, i32 36)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %12, i32 40)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %12, i32 44)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %12, i32 48)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %12, i32 52)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %12, i32 56)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %12, i32 60)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %12, i32 64)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %12, i32 68)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %12, i32 72)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %12, i32 76)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %12, i32 80)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %12, i32 84)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %12, i32 88)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %12, i32 92)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %12, i32 96)
>  %38 = call float @llvm.SI.load.const(<16 x i8> %12, i32 100)
>  %39 = call float @llvm.SI.load.const(<16 x i8> %12, i32 104)
>  %40 = call float @llvm.SI.load.const(<16 x i8> %12, i32 108)
>  %41 = call float @llvm.SI.load.const(<16 x i8> %12, i32 112)
>  %42 = call float @llvm.SI.load.const(<16 x i8> %12, i32 116)
>  %43 = call float @llvm.SI.load.const(<16 x i8> %12, i32 120)
>  %44 = call float @llvm.SI.load.const(<16 x i8> %12, i32 124)
>  %45 = call float @llvm.SI.load.const(<16 x i8> %12, i32 128)
>  %46 = call float @llvm.SI.load.const(<16 x i8> %12, i32 132)
>  %47 = call float @llvm.SI.load.const(<16 x i8> %12, i32 136)
>  %48 = call float @llvm.SI.load.const(<16 x i8> %12, i32 140)
>  %49 = call float @llvm.SI.load.const(<16 x i8> %12, i32 144)
>  %50 = call float @llvm.SI.load.const(<16 x i8> %12, i32 148)
>  %51 = call float @llvm.SI.load.const(<16 x i8> %12, i32 152)
>  %52 = call float @llvm.SI.load.const(<16 x i8> %12, i32 160)
>  %53 = call float @llvm.SI.load.const(<16 x i8> %12, i32 164)
>  %54 = call float @llvm.SI.load.const(<16 x i8> %12, i32 176)
>  %55 = call float @llvm.SI.load.const(<16 x i8> %12, i32 180)
>  %56 = call float @llvm.SI.load.const(<16 x i8> %12, i32 184)
>  %57 = call float @llvm.SI.load.const(<16 x i8> %12, i32 188)
>  %58 = call float @llvm.SI.load.const(<16 x i8> %12, i32 192)
>  %59 = call float @llvm.SI.load.const(<16 x i8> %12, i32 196)
>  %60 = call float @llvm.SI.load.const(<16 x i8> %12, i32 200)
>  %61 = call float @llvm.SI.load.const(<16 x i8> %12, i32 204)
>  %62 = call float @llvm.SI.load.const(<16 x i8> %12, i32 208)
>  %63 = call float @llvm.SI.load.const(<16 x i8> %12, i32 212)
>  %64 = call float @llvm.SI.load.const(<16 x i8> %12, i32 216)
>  %65 = call float @llvm.SI.load.const(<16 x i8> %12, i32 220)
>  %66 = call float @llvm.SI.load.const(<16 x i8> %12, i32 224)
>  %67 = call float @llvm.SI.load.const(<16 x i8> %12, i32 228)
>  %68 = call float @llvm.SI.load.const(<16 x i8> %12, i32 232)
>  %69 = call float @llvm.SI.load.const(<16 x i8> %12, i32 236)
>  %70 = call float @llvm.SI.load.const(<16 x i8> %12, i32 240)
>  %71 = call float @llvm.SI.load.const(<16 x i8> %12, i32 244)
>  %72 = call float @llvm.SI.load.const(<16 x i8> %12, i32 248)
>  %73 = call float @llvm.SI.load.const(<16 x i8> %12, i32 252)
>  %74 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %75 = load <8 x i32> addrspace(2)* %74, !tbaa !0
>  %76 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %77 = load <4 x i32> addrspace(2)* %76, !tbaa !0
>  %78 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %79 = load <16 x i8> addrspace(2)* %78, !tbaa !0
>  %80 = add i32 %5, %7
>  %81 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %79, i32 0, i32 %80)
>  %82 = extractelement <4 x float> %81, i32 0
>  %83 = extractelement <4 x float> %81, i32 1
>  %84 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %85 = load <16 x i8> addrspace(2)* %84, !tbaa !0
>  %86 = add i32 %5, %7
>  %87 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %85, i32 0, i32 %86)
>  %88 = extractelement <4 x float> %87, i32 0
>  %89 = extractelement <4 x float> %87, i32 1
>  %90 = fadd float %70, %82
>  %91 = fadd float %71, %83
>  %92 = fadd float %90, %72
>  %93 = fadd float %91, %73
>  %94 = fmul float %90, %49
>  %95 = fmul float %91, %50
>  %96 = fadd float %92, 5.000000e-01
>  %97 = fadd float %93, 5.000000e-01
>  %98 = fmul float %96, %52
>  %99 = fmul float %97, %53
>  %100 = bitcast float %98 to i32
>  %101 = bitcast float %99 to i32
>  %102 = bitcast float 0.000000e+00 to i32
>  %103 = insertelement <4 x i32> undef, i32 %100, i32 0
>  %104 = insertelement <4 x i32> %103, i32 %101, i32 1
>  %105 = insertelement <4 x i32> %104, i32 %102, i32 2
>  %106 = insertelement <4 x i32> %105, i32 undef, i32 3
>  %107 = bitcast <8 x i32> %75 to <32 x i8>
>  %108 = bitcast <4 x i32> %77 to <16 x i8>
>  %109 = call <4 x float> @llvm.SI.samplel.v4i32(<4 x i32> %106, <32 x i8> %107, <16 x i8> %108, i32 2)
>  %110 = extractelement <4 x float> %109, i32 3
>  %111 = fmul float %110, %51
>  %112 = fmul float %111, 2.550000e+02
>  %113 = fmul float %95, %58
>  %114 = fmul float %95, %59
>  %115 = fmul float %95, %60
>  %116 = fmul float %95, %61
>  %117 = fmul float %94, %54
>  %118 = fadd float %117, %113
>  %119 = fmul float %94, %55
>  %120 = fadd float %119, %114
>  %121 = fmul float %94, %56
>  %122 = fadd float %121, %115
>  %123 = fmul float %94, %57
>  %124 = fadd float %123, %116
>  %125 = fmul float %112, %62
>  %126 = fadd float %125, %118
>  %127 = fmul float %112, %63
>  %128 = fadd float %127, %120
>  %129 = fmul float %112, %64
>  %130 = fadd float %129, %122
>  %131 = fmul float %112, %65
>  %132 = fadd float %131, %124
>  %133 = fadd float %126, %66
>  %134 = fadd float %128, %67
>  %135 = fadd float %130, %68
>  %136 = fadd float %132, %69
>  %137 = fadd float %133, %47
>  %138 = fadd float %134, %48
>  %139 = fmul float %137, %45
>  %140 = fmul float %138, %46
>  %141 = fmul float %134, %17
>  %142 = fmul float %134, %18
>  %143 = fmul float %134, %19
>  %144 = fmul float %134, %20
>  %145 = fmul float %133, %13
>  %146 = fadd float %145, %141
>  %147 = fmul float %133, %14
>  %148 = fadd float %147, %142
>  %149 = fmul float %133, %15
>  %150 = fadd float %149, %143
>  %151 = fmul float %133, %16
>  %152 = fadd float %151, %144
>  %153 = fmul float %135, %21
>  %154 = fadd float %153, %146
>  %155 = fmul float %135, %22
>  %156 = fadd float %155, %148
>  %157 = fmul float %135, %23
>  %158 = fadd float %157, %150
>  %159 = fmul float %135, %24
>  %160 = fadd float %159, %152
>  %161 = fmul float %136, %25
>  %162 = fadd float %161, %154
>  %163 = fmul float %136, %26
>  %164 = fadd float %163, %156
>  %165 = fmul float %136, %27
>  %166 = fadd float %165, %158
>  %167 = fmul float %136, %28
>  %168 = fadd float %167, %160
>  %169 = fdiv float 1.000000e+00, %168
>  %170 = fmul float %162, %169
>  %171 = fmul float %164, %169
>  %172 = fmul float %134, %33
>  %173 = fmul float %134, %34
>  %174 = fmul float %134, %35
>  %175 = fmul float %134, %36
>  %176 = fmul float %133, %29
>  %177 = fadd float %176, %172
>  %178 = fmul float %133, %30
>  %179 = fadd float %178, %173
>  %180 = fmul float %133, %31
>  %181 = fadd float %180, %174
>  %182 = fmul float %133, %32
>  %183 = fadd float %182, %175
>  %184 = fmul float %135, %37
>  %185 = fadd float %184, %177
>  %186 = fmul float %135, %38
>  %187 = fadd float %186, %179
>  %188 = fmul float %135, %39
>  %189 = fadd float %188, %181
>  %190 = fmul float %135, %40
>  %191 = fadd float %190, %183
>  %192 = fmul float %136, %41
>  %193 = fadd float %192, %185
>  %194 = fmul float %136, %42
>  %195 = fadd float %194, %187
>  %196 = fmul float %136, %43
>  %197 = fadd float %196, %189
>  %198 = fmul float %136, %44
>  %199 = fadd float %198, %191
>  %200 = fmul float %193, 5.000000e-01
>  %201 = fadd float %200, 5.000000e-01
>  %202 = fmul float %195, -5.000000e-01
>  %203 = fadd float %202, 5.000000e-01
>  %204 = fmul float %197, 1.000000e+00
>  %205 = fadd float %204, 0.000000e+00
>  %206 = fmul float %199, 1.000000e+00
>  %207 = fadd float %206, 0.000000e+00
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %88, float %89, float %139, float %140)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %201, float %203, float %205, float %207)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 34, i32 0, float %133, float %134, float %135, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 35, i32 0, float %170, float %171, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %162, float %164, float %166, float %168)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.samplel.v4i32(<4 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                                              ; 4A00000A
>	s_load_dwordx4 s[12:15], s[8:9], 0x0                                   ; C0860900
>	s_load_dwordx4 s[8:11], s[8:9], 0x4                                    ; C0840904
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[12:15], 0 idxen                  ; E00C2000 80030100
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                                     ; C0800300
>	s_waitcnt vmcnt(0) lgkmcnt(0)                                          ; BF8C0070
>	s_buffer_load_dword s12, s[0:3], 0x3c                                  ; C206013C
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_add_f32_e32 v5, s12, v1                                              ; 060A020C
>	s_buffer_load_dword s12, s[0:3], 0x24                                  ; C2060124
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mul_f32_e32 v6, s12, v5                                              ; 100C0A0C
>	s_buffer_load_dword s12, s[0:3], 0x3d                                  ; C206013D
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_add_f32_e32 v1, s12, v2                                              ; 0602040C
>	s_buffer_load_dword s12, s[0:3], 0x25                                  ; C2060125
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mul_f32_e32 v2, s12, v1                                              ; 1004020C
>	s_buffer_load_dword s12, s[0:3], 0x31                                  ; C2060131
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mul_f32_e32 v3, s12, v2                                              ; 1006040C
>	s_buffer_load_dword s12, s[0:3], 0x2d                                  ; C206012D
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mad_f32 v3, v6, s12, v3                                              ; D2820003 040C1906
>	s_buffer_load_dword s12, s[0:3], 0x3f                                  ; C206013F
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_add_f32_e32 v1, s12, v1                                              ; 0602020C
>	v_add_f32_e32 v1, 0.5, v1                                              ; 060202F0
>	s_buffer_load_dword s12, s[0:3], 0x29                                  ; C2060129
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mul_f32_e32 v8, s12, v1                                              ; 1010020C
>	s_buffer_load_dword s12, s[0:3], 0x3e                                  ; C206013E
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_add_f32_e32 v1, s12, v5                                              ; 06020A0C
>	v_add_f32_e32 v1, 0.5, v1                                              ; 060202F0
>	s_buffer_load_dword s12, s[0:3], 0x28                                  ; C2060128
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mul_f32_e32 v7, s12, v1                                              ; 100E020C
>	v_mov_b32_e32 v9, 0                                                    ; 7E120280
>	s_load_dwordx4 s[12:15], s[4:5], 0x0                                   ; C0860500
>	s_load_dwordx8 s[16:23], s[6:7], 0x0                                   ; C0C80700
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	image_sample_l v1, 8, 0, 0, 0, 0, 0, 0, 0, v[7:10], s[16:23], s[12:15] ; F0900800 00640107
>	s_buffer_load_dword s4, s[0:3], 0x26                                   ; C2020126
>	s_waitcnt vmcnt(0) lgkmcnt(0)                                          ; BF8C0070
>	v_mul_f32_e32 v1, s4, v1                                               ; 10020204
>	v_mul_f32_e32 v1, 0x437f0000, v1                                       ; 100202FF 437F0000
>	s_buffer_load_dword s4, s[0:3], 0x35                                   ; C2020135
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mad_f32 v3, v1, s4, v3                                               ; D2820003 040C0901
>	s_buffer_load_dword s4, s[0:3], 0x39                                   ; C2020139
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_add_f32_e32 v3, s4, v3                                               ; 06060604
>	s_buffer_load_dword s4, s[0:3], 0x23                                   ; C2020123
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_add_f32_e32 v4, s4, v3                                               ; 06080604
>	s_buffer_load_dword s4, s[0:3], 0x21                                   ; C2020121
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mul_f32_e32 v4, s4, v4                                               ; 10080804
>	s_buffer_load_dword s4, s[0:3], 0x30                                   ; C2020130
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mul_f32_e32 v5, s4, v2                                               ; 100A0404
>	s_buffer_load_dword s4, s[0:3], 0x2c                                   ; C202012C
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mad_f32 v5, v6, s4, v5                                               ; D2820005 04140906
>	s_buffer_load_dword s4, s[0:3], 0x34                                   ; C2020134
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mad_f32 v5, v1, s4, v5                                               ; D2820005 04140901
>	s_buffer_load_dword s4, s[0:3], 0x38                                   ; C2020138
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_add_f32_e32 v5, s4, v5                                               ; 060A0A04
>	s_buffer_load_dword s4, s[0:3], 0x22                                   ; C2020122
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_add_f32_e32 v11, s4, v5                                              ; 06160A04
>	s_buffer_load_dword s4, s[0:3], 0x20                                   ; C2020120
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mul_f32_e32 v11, s4, v11                                             ; 10161604
>	buffer_load_format_xyzw v[12:15], v0, s[8:11], 0 idxen                 ; E00C2000 80020C00
>	s_waitcnt vmcnt(0)                                                     ; BF8C0770
>	exp 15, 32, 0, 0, 0, v12, v13, v11, v4                                 ; F800020F 040B0D0C
>	s_buffer_load_dword s4, s[0:3], 0x15                                   ; C2020115
>	s_waitcnt expcnt(0) lgkmcnt(0)                                         ; BF8C000F
>	v_mul_f32_e32 v0, s4, v3                                               ; 10000604
>	s_buffer_load_dword s4, s[0:3], 0x11                                   ; C2020111
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mad_f32 v0, v5, s4, v0                                               ; D2820000 04000905
>	s_buffer_load_dword s4, s[0:3], 0x32                                   ; C2020132
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mul_f32_e32 v4, s4, v2                                               ; 10080404
>	s_buffer_load_dword s4, s[0:3], 0x2e                                   ; C202012E
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mad_f32 v4, v6, s4, v4                                               ; D2820004 04100906
>	s_buffer_load_dword s4, s[0:3], 0x36                                   ; C2020136
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mad_f32 v4, v1, s4, v4                                               ; D2820004 04100901
>	s_buffer_load_dword s4, s[0:3], 0x3a                                   ; C202013A
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_add_f32_e32 v4, s4, v4                                               ; 06080804
>	s_buffer_load_dword s4, s[0:3], 0x19                                   ; C2020119
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mad_f32 v0, v4, s4, v0                                               ; D2820000 04000904
>	s_buffer_load_dword s4, s[0:3], 0x33                                   ; C2020133
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mul_f32_e32 v2, s4, v2                                               ; 10040404
>	s_buffer_load_dword s4, s[0:3], 0x2f                                   ; C202012F
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mad_f32 v2, v6, s4, v2                                               ; D2820002 04080906
>	s_buffer_load_dword s4, s[0:3], 0x37                                   ; C2020137
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mad_f32 v1, v1, s4, v2                                               ; D2820001 04080901
>	s_buffer_load_dword s4, s[0:3], 0x3b                                   ; C202013B
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_add_f32_e32 v1, s4, v1                                               ; 06020204
>	s_buffer_load_dword s4, s[0:3], 0x1d                                   ; C202011D
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mad_f32 v0, v1, s4, v0                                               ; D2820000 04000901
>	v_mad_f32 v0, -0.5, v0, 0.5                                            ; D2820000 03C200F1
>	s_buffer_load_dword s4, s[0:3], 0x17                                   ; C2020117
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mul_f32_e32 v2, s4, v3                                               ; 10040604
>	s_buffer_load_dword s4, s[0:3], 0x13                                   ; C2020113
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mad_f32 v2, v5, s4, v2                                               ; D2820002 04080905
>	s_buffer_load_dword s4, s[0:3], 0x1b                                   ; C202011B
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mad_f32 v2, v4, s4, v2                                               ; D2820002 04080904
>	s_buffer_load_dword s4, s[0:3], 0x1f                                   ; C202011F
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mad_f32 v2, v1, s4, v2                                               ; D2820002 04080901
>	v_add_f32_e32 v2, 0, v2                                                ; 06040480
>	s_buffer_load_dword s4, s[0:3], 0x16                                   ; C2020116
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mul_f32_e32 v6, s4, v3                                               ; 100C0604
>	s_buffer_load_dword s4, s[0:3], 0x12                                   ; C2020112
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mad_f32 v6, v5, s4, v6                                               ; D2820006 04180905
>	s_buffer_load_dword s4, s[0:3], 0x1a                                   ; C202011A
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mad_f32 v6, v4, s4, v6                                               ; D2820006 04180904
>	s_buffer_load_dword s4, s[0:3], 0x1e                                   ; C202011E
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mad_f32 v6, v1, s4, v6                                               ; D2820006 04180901
>	v_add_f32_e32 v6, 0, v6                                                ; 060C0C80
>	s_buffer_load_dword s4, s[0:3], 0x14                                   ; C2020114
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mul_f32_e32 v11, s4, v3                                              ; 10160604
>	s_buffer_load_dword s4, s[0:3], 0x10                                   ; C2020110
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mad_f32 v11, v5, s4, v11                                             ; D282000B 042C0905
>	s_buffer_load_dword s4, s[0:3], 0x18                                   ; C2020118
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mad_f32 v11, v4, s4, v11                                             ; D282000B 042C0904
>	s_buffer_load_dword s4, s[0:3], 0x1c                                   ; C202011C
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mad_f32 v11, v1, s4, v11                                             ; D282000B 042C0901
>	v_mad_f32 v11, 0.5, v11, 0.5                                           ; D282000B 03C216F0
>	exp 15, 33, 0, 0, 0, v11, v0, v6, v2                                   ; F800021F 0206000B
>	s_waitcnt expcnt(0)                                                    ; BF8C070F
>	v_mov_b32_e32 v0, 1.0                                                  ; 7E0002F2
>	exp 15, 34, 0, 0, 0, v5, v3, v4, v0                                    ; F800022F 00040305
>	s_buffer_load_dword s4, s[0:3], 0x5                                    ; C2020105
>	s_waitcnt expcnt(0) lgkmcnt(0)                                         ; BF8C000F
>	v_mul_f32_e32 v2, s4, v3                                               ; 10040604
>	s_buffer_load_dword s4, s[0:3], 0x1                                    ; C2020101
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mad_f32 v2, v5, s4, v2                                               ; D2820002 04080905
>	s_buffer_load_dword s4, s[0:3], 0x9                                    ; C2020109
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mad_f32 v2, v4, s4, v2                                               ; D2820002 04080904
>	s_buffer_load_dword s4, s[0:3], 0xd                                    ; C202010D
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mad_f32 v2, v1, s4, v2                                               ; D2820002 04080901
>	s_buffer_load_dword s4, s[0:3], 0x7                                    ; C2020107
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mul_f32_e32 v6, s4, v3                                               ; 100C0604
>	s_buffer_load_dword s4, s[0:3], 0x3                                    ; C2020103
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mad_f32 v6, v5, s4, v6                                               ; D2820006 04180905
>	s_buffer_load_dword s4, s[0:3], 0xb                                    ; C202010B
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mad_f32 v6, v4, s4, v6                                               ; D2820006 04180904
>	s_buffer_load_dword s4, s[0:3], 0xf                                    ; C202010F
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mad_f32 v6, v1, s4, v6                                               ; D2820006 04180901
>	v_rcp_f32_e32 v11, v6                                                  ; 7E165506
>	v_mul_f32_e32 v12, v11, v2                                             ; 1018050B
>	s_buffer_load_dword s4, s[0:3], 0x4                                    ; C2020104
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mul_f32_e32 v13, s4, v3                                              ; 101A0604
>	s_buffer_load_dword s4, s[0:3], 0x0                                    ; C2020100
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mad_f32 v13, v5, s4, v13                                             ; D282000D 04340905
>	s_buffer_load_dword s4, s[0:3], 0x8                                    ; C2020108
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mad_f32 v13, v4, s4, v13                                             ; D282000D 04340904
>	s_buffer_load_dword s4, s[0:3], 0xc                                    ; C202010C
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mad_f32 v13, v1, s4, v13                                             ; D282000D 04340901
>	v_mul_f32_e32 v11, v11, v13                                            ; 10161B0B
>	exp 15, 35, 0, 0, 0, v11, v12, v9, v0                                  ; F800023F 00090C0B
>	s_buffer_load_dword s4, s[0:3], 0x6                                    ; C2020106
>	s_waitcnt expcnt(0) lgkmcnt(0)                                         ; BF8C000F
>	v_mul_f32_e32 v0, s4, v3                                               ; 10000604
>	s_buffer_load_dword s4, s[0:3], 0x2                                    ; C2020102
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mad_f32 v0, v5, s4, v0                                               ; D2820000 04000905
>	s_buffer_load_dword s4, s[0:3], 0xa                                    ; C202010A
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mad_f32 v0, v4, s4, v0                                               ; D2820000 04000904
>	s_buffer_load_dword s0, s[0:3], 0xe                                    ; C200010E
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mad_f32 v0, v1, s0, v0                                               ; D2820000 04000101
>	exp 15, 12, 0, 1, 0, v13, v2, v0, v6                                   ; F80008CF 0600020D
>	s_endpgm                                                               ; BF810000
>
>SGPRS: 32
>VGPRS: 16
>Code Size: 1032 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL IN[1], GENERIC[10], PERSPECTIVE
>DCL IN[2], GENERIC[11], PERSPECTIVE
>DCL IN[3], GENERIC[12], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SAMP[1]
>DCL SAMP[2]
>DCL SAMP[3]
>DCL SAMP[4]
>DCL SAMP[5]
>DCL SAMP[6]
>DCL SAMP[7]
>DCL CONST[0..8]
>DCL TEMP[0..6], LOCAL
>IMM[0] FLT32 {    1.4427,     1.0000,     0.5000,    -0.5000}
>IMM[1] FLT32 {    1.0000,    -1.0000,     0.0000,     0.0000}
>  0: ADD TEMP[0].xyz, -CONST[0], IN[2]
>  1: MUL TEMP[1].w, TEMP[0].zzzz, CONST[8].xxxx
>  2: MUL TEMP[1].w, TEMP[1].wwww, IMM[0].xxxx
>  3: EX2 TEMP[1].x, TEMP[1].wwww
>  4: ADD TEMP[1].w, -TEMP[1].xxxx, IMM[0].yyyy
>  5: DP3 TEMP[2].x, TEMP[0].xyzz, TEMP[0].xyzz
>  6: RCP TEMP[3].x, TEMP[0].zzzz
>  7: MUL TEMP[2].x, TEMP[2].xxxx, CONST[8].yyyy
>  8: MUL TEMP[1].x, TEMP[1].wwww, TEMP[2].xxxx
>  9: MUL TEMP[1].x, TEMP[3].xxxx, TEMP[1].xxxx
> 10: MUL TEMP[1].x, TEMP[1].xxxx, IMM[0].xxxx
> 11: MOV TEMP[0].x, TEMP[1].xxxx
> 12: EX2 TEMP[1].x, TEMP[1].xxxx
> 13: MOV_SAT TEMP[0].x, TEMP[1].xxxx
> 14: ADD TEMP[1].x, -TEMP[0].xxxx, IMM[0].yyyy
> 15: MAD TEMP[2].x, IN[3].yyyy, IMM[0].zzzz, IMM[0].zzzz
> 16: MOV TEMP[2].x, TEMP[2].xxxx
> 17: MOV TEMP[2].y, CONST[8].wwww
> 18: MOV TEMP[3].xy, TEMP[2].xyyy
> 19: TEX TEMP[3].x, TEMP[3], SAMP[2], 2D
> 20: MUL TEMP[1].x, TEMP[1].xxxx, TEMP[3].xxxx
> 21: MOV TEMP[0].x, TEMP[1].xxxx
> 22: MUL TEMP[3].yzw, CONST[4].xxyw, IN[2].yyyy
> 23: MOV TEMP[0].yzw, TEMP[3].zyzw
> 24: MAD TEMP[3].yzw, IN[2].xxxx, CONST[3].xxyw, TEMP[0]
> 25: MOV TEMP[0].yzw, TEMP[3].zyzw
> 26: MAD TEMP[3].yzw, IN[2].zzzz, CONST[5].xxyw, TEMP[0]
> 27: MOV TEMP[0].yzw, TEMP[3].zyzw
> 28: ADD TEMP[3].yzw, TEMP[0], CONST[6].xxyw
> 29: MOV TEMP[0].yz, TEMP[3].zyzw
> 30: RCP TEMP[3].x, TEMP[3].wwww
> 31: MOV TEMP[0].w, TEMP[3].xxxx
> 32: MUL TEMP[3].yz, TEMP[3].xxxx, TEMP[0]
> 33: MOV TEMP[0].yz, TEMP[3].zyzz
> 34: MAD TEMP[3].yz, TEMP[0], IMM[0].zzwy, IMM[0].wwww
> 35: MOV TEMP[0].yz, TEMP[3].zyzz
> 36: MAD TEMP[0].xy, TEMP[0].yzzw, IMM[1].xyxx, IMM[1].zxzz
> 37: MOV TEMP[0].xy, TEMP[0].xyyy
> 38: TEX TEMP[0].x, TEMP[0], SAMP[3], 2D
> 39: MOV_SAT TEMP[0].x, TEMP[0].xxxx
> 40: MUL TEMP[0].y, TEMP[0].xxxx, CONST[1].xxxx
> 41: MAD TEMP[3].xy, IN[0], IMM[1].xyxx, IMM[1].zxzz
> 42: MOV TEMP[3].xy, TEMP[3].xyyy
> 43: TEX TEMP[3], TEMP[3], SAMP[0], 2D
> 44: MOV TEMP[2].w, TEMP[3]
> 45: MOV TEMP[4], -CONST[2].xxxx
> 46: FSGE TEMP[5].x, TEMP[4].xxxx, IMM[1].zzzz
> 47: UIF TEMP[5].xxxx :0
> 48:   MOV TEMP[5].x, IMM[1].zzzz
> 49: ELSE :0
> 50:   MOV TEMP[5].x, TEMP[3].xxxx
> 51: ENDIF
> 52: MOV TEMP[5].x, TEMP[5].xxxx
> 53: FSGE TEMP[6].x, TEMP[4].yyyy, IMM[1].zzzz
> 54: UIF TEMP[6].xxxx :0
> 55:   MOV TEMP[6].x, IMM[1].zzzz
> 56: ELSE :0
> 57:   MOV TEMP[6].x, TEMP[3].yyyy
> 58: ENDIF
> 59: MOV TEMP[5].y, TEMP[6].xxxx
> 60: FSGE TEMP[6].x, TEMP[4].zzzz, IMM[1].zzzz
> 61: UIF TEMP[6].xxxx :0
> 62:   MOV TEMP[6].x, IMM[1].zzzz
> 63: ELSE :0
> 64:   MOV TEMP[6].x, TEMP[3].zzzz
> 65: ENDIF
> 66: MOV TEMP[5].z, TEMP[6].xxxx
> 67: FSGE TEMP[3].x, TEMP[4].wwww, IMM[1].zzzz
> 68: UIF TEMP[3].xxxx :0
> 69: ELSE :0
> 70: ENDIF
> 71: MOV TEMP[2].xyz, TEMP[5].xyzx
> 72: MOV TEMP[3].xyz, TEMP[5].xyzz
> 73: TEX TEMP[3], TEMP[3], SAMP[7], 3D
> 74: LRP TEMP[0].xyz, TEMP[0].yyyy, TEMP[3], TEMP[2]
> 75: MUL TEMP[2], IN[1], IMM[1].xyxx
> 76: ADD TEMP[3].xy, TEMP[2], IMM[1].zxzz
> 77: MOV TEMP[3].xy, TEMP[3].xyyy
> 78: TEX TEMP[3], TEMP[3], SAMP[1], 2D
> 79: ADD TEMP[4].y, -TEMP[3].wwww, IMM[0].yyyy
> 80: MAD TEMP[0].yzw, TEMP[0].xxyz, TEMP[4].yyyy, TEMP[3].xxyz
> 81: ADD TEMP[2].xy, TEMP[2], IMM[1].zxzz
> 82: MOV TEMP[2].xy, TEMP[2].xyyy
> 83: TEX TEMP[2].xzw, TEMP[2], SAMP[4], 2D
> 84: MOV TEMP[3].w, TEMP[2].wwww
> 85: MOV TEMP[4].xy, IN[0].zwww
> 86: TEX TEMP[4], TEMP[4], SAMP[5], 2D
> 87: LRP TEMP[0].xyz, TEMP[2].xxxx, TEMP[0].yzww, TEMP[4]
> 88: MOV TEMP[4].xyz, TEMP[0].xyzx
> 89: MAD TEMP[2].y, TEMP[2].zzzz, -TEMP[2].xxxx, TEMP[2].xxxx
> 90: MOV TEMP[0].xyz, TEMP[0].xyzz
> 91: TEX TEMP[0], TEMP[0], SAMP[6], 3D
> 92: LRP TEMP[0].xyz, TEMP[2].yyyy, TEMP[0], TEMP[4]
> 93: MOV TEMP[3].xyz, TEMP[0].xyzx
> 94: ADD TEMP[0].yzw, -TEMP[0].xxyz, CONST[7].xxyz
> 95: MAD TEMP[0].xyz, TEMP[1].xxxx, TEMP[0].yzww, TEMP[3]
> 96: MOV TEMP[0].xyz, TEMP[0].xyzx
> 97: MOV TEMP[0].w, IMM[0].yyyy
> 98: MOV OUT[0], TEMP[0]
> 99: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %23 = load <16 x i8> addrspace(2)* %22, !tbaa !0
>  %24 = call float @llvm.SI.load.const(<16 x i8> %23, i32 0)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %23, i32 4)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %23, i32 8)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %23, i32 16)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %23, i32 32)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %23, i32 48)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %23, i32 52)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %23, i32 60)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %23, i32 64)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %23, i32 68)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %23, i32 76)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %23, i32 80)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %23, i32 84)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %23, i32 92)
>  %38 = call float @llvm.SI.load.const(<16 x i8> %23, i32 96)
>  %39 = call float @llvm.SI.load.const(<16 x i8> %23, i32 100)
>  %40 = call float @llvm.SI.load.const(<16 x i8> %23, i32 108)
>  %41 = call float @llvm.SI.load.const(<16 x i8> %23, i32 112)
>  %42 = call float @llvm.SI.load.const(<16 x i8> %23, i32 116)
>  %43 = call float @llvm.SI.load.const(<16 x i8> %23, i32 120)
>  %44 = call float @llvm.SI.load.const(<16 x i8> %23, i32 128)
>  %45 = call float @llvm.SI.load.const(<16 x i8> %23, i32 132)
>  %46 = call float @llvm.SI.load.const(<16 x i8> %23, i32 140)
>  %47 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %48 = load <8 x i32> addrspace(2)* %47, !tbaa !0
>  %49 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %50 = load <4 x i32> addrspace(2)* %49, !tbaa !0
>  %51 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 1
>  %52 = load <8 x i32> addrspace(2)* %51, !tbaa !0
>  %53 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 1
>  %54 = load <4 x i32> addrspace(2)* %53, !tbaa !0
>  %55 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 2
>  %56 = load <8 x i32> addrspace(2)* %55, !tbaa !0
>  %57 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 2
>  %58 = load <4 x i32> addrspace(2)* %57, !tbaa !0
>  %59 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 3
>  %60 = load <8 x i32> addrspace(2)* %59, !tbaa !0
>  %61 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 3
>  %62 = load <4 x i32> addrspace(2)* %61, !tbaa !0
>  %63 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 4
>  %64 = load <8 x i32> addrspace(2)* %63, !tbaa !0
>  %65 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 4
>  %66 = load <4 x i32> addrspace(2)* %65, !tbaa !0
>  %67 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 5
>  %68 = load <8 x i32> addrspace(2)* %67, !tbaa !0
>  %69 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 5
>  %70 = load <4 x i32> addrspace(2)* %69, !tbaa !0
>  %71 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 6
>  %72 = load <8 x i32> addrspace(2)* %71, !tbaa !0
>  %73 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 6
>  %74 = load <4 x i32> addrspace(2)* %73, !tbaa !0
>  %75 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 7
>  %76 = load <8 x i32> addrspace(2)* %75, !tbaa !0
>  %77 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 7
>  %78 = load <4 x i32> addrspace(2)* %77, !tbaa !0
>  %79 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %80 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %81 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %5, <2 x i32> %7)
>  %82 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %5, <2 x i32> %7)
>  %83 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %5, <2 x i32> %7)
>  %84 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %5, <2 x i32> %7)
>  %85 = call float @llvm.SI.fs.interp(i32 0, i32 2, i32 %5, <2 x i32> %7)
>  %86 = call float @llvm.SI.fs.interp(i32 1, i32 2, i32 %5, <2 x i32> %7)
>  %87 = call float @llvm.SI.fs.interp(i32 2, i32 2, i32 %5, <2 x i32> %7)
>  %88 = call float @llvm.SI.fs.interp(i32 1, i32 3, i32 %5, <2 x i32> %7)
>  %89 = fsub float -0.000000e+00, %24
>  %90 = fadd float %89, %85
>  %91 = fsub float -0.000000e+00, %25
>  %92 = fadd float %91, %86
>  %93 = fsub float -0.000000e+00, %26
>  %94 = fadd float %93, %87
>  %95 = fmul float %94, %44
>  %96 = fmul float %95, 0x3FF7154CA0000000
>  %97 = call float @llvm.AMDIL.exp.(float %96)
>  %98 = fsub float -0.000000e+00, %97
>  %99 = fadd float %98, 1.000000e+00
>  %100 = fmul float %90, %90
>  %101 = fmul float %92, %92
>  %102 = fadd float %101, %100
>  %103 = fmul float %94, %94
>  %104 = fadd float %102, %103
>  %105 = fdiv float 1.000000e+00, %94
>  %106 = fmul float %104, %45
>  %107 = fmul float %99, %106
>  %108 = fmul float %105, %107
>  %109 = fmul float %108, 0x3FF7154CA0000000
>  %110 = call float @llvm.AMDIL.exp.(float %109)
>  %111 = call float @llvm.AMDIL.clamp.(float %110, float 0.000000e+00, float 1.000000e+00)
>  %112 = fsub float -0.000000e+00, %111
>  %113 = fadd float %112, 1.000000e+00
>  %114 = fmul float %88, 5.000000e-01
>  %115 = fadd float %114, 5.000000e-01
>  %116 = bitcast float %115 to i32
>  %117 = bitcast float %46 to i32
>  %118 = insertelement <2 x i32> undef, i32 %116, i32 0
>  %119 = insertelement <2 x i32> %118, i32 %117, i32 1
>  %120 = bitcast <8 x i32> %56 to <32 x i8>
>  %121 = bitcast <4 x i32> %58 to <16 x i8>
>  %122 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %119, <32 x i8> %120, <16 x i8> %121, i32 2)
>  %123 = extractelement <4 x float> %122, i32 0
>  %124 = fmul float %113, %123
>  %125 = fmul float %32, %86
>  %126 = fmul float %33, %86
>  %127 = fmul float %34, %86
>  %128 = fmul float %85, %29
>  %129 = fadd float %128, %125
>  %130 = fmul float %85, %30
>  %131 = fadd float %130, %126
>  %132 = fmul float %85, %31
>  %133 = fadd float %132, %127
>  %134 = fmul float %87, %35
>  %135 = fadd float %134, %129
>  %136 = fmul float %87, %36
>  %137 = fadd float %136, %131
>  %138 = fmul float %87, %37
>  %139 = fadd float %138, %133
>  %140 = fadd float %135, %38
>  %141 = fadd float %137, %39
>  %142 = fadd float %139, %40
>  %143 = fdiv float 1.000000e+00, %142
>  %144 = fmul float %143, %140
>  %145 = fmul float %143, %141
>  %146 = fmul float %144, 5.000000e-01
>  %147 = fadd float %146, -5.000000e-01
>  %148 = fmul float %145, -5.000000e-01
>  %149 = fadd float %148, -5.000000e-01
>  %150 = fmul float %147, 1.000000e+00
>  %151 = fadd float %150, 0.000000e+00
>  %152 = fmul float %149, -1.000000e+00
>  %153 = fadd float %152, 1.000000e+00
>  %154 = bitcast float %151 to i32
>  %155 = bitcast float %153 to i32
>  %156 = insertelement <2 x i32> undef, i32 %154, i32 0
>  %157 = insertelement <2 x i32> %156, i32 %155, i32 1
>  %158 = bitcast <8 x i32> %60 to <32 x i8>
>  %159 = bitcast <4 x i32> %62 to <16 x i8>
>  %160 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %157, <32 x i8> %158, <16 x i8> %159, i32 2)
>  %161 = extractelement <4 x float> %160, i32 0
>  %162 = call float @llvm.AMDIL.clamp.(float %161, float 0.000000e+00, float 1.000000e+00)
>  %163 = fmul float %162, %27
>  %164 = fmul float %79, 1.000000e+00
>  %165 = fadd float %164, 0.000000e+00
>  %166 = fmul float %80, -1.000000e+00
>  %167 = fadd float %166, 1.000000e+00
>  %168 = bitcast float %165 to i32
>  %169 = bitcast float %167 to i32
>  %170 = insertelement <2 x i32> undef, i32 %168, i32 0
>  %171 = insertelement <2 x i32> %170, i32 %169, i32 1
>  %172 = bitcast <8 x i32> %48 to <32 x i8>
>  %173 = bitcast <4 x i32> %50 to <16 x i8>
>  %174 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %171, <32 x i8> %172, <16 x i8> %173, i32 2)
>  %175 = extractelement <4 x float> %174, i32 0
>  %176 = extractelement <4 x float> %174, i32 1
>  %177 = extractelement <4 x float> %174, i32 2
>  %178 = fsub float -0.000000e+00, %28
>  %179 = fsub float -0.000000e+00, %28
>  %180 = fsub float -0.000000e+00, %28
>  %181 = fsub float -0.000000e+00, %28
>  %182 = fcmp oge float %178, 0.000000e+00
>  %183 = sext i1 %182 to i32
>  %184 = bitcast i32 %183 to float
>  %185 = bitcast float %184 to i32
>  %186 = icmp ne i32 %185, 0
>  %. = select i1 %186, float 0.000000e+00, float %175
>  %187 = fcmp oge float %179, 0.000000e+00
>  %188 = sext i1 %187 to i32
>  %189 = bitcast i32 %188 to float
>  %190 = bitcast float %189 to i32
>  %191 = icmp ne i32 %190, 0
>  %temp24.0 = select i1 %191, float 0.000000e+00, float %176
>  %192 = fcmp oge float %180, 0.000000e+00
>  %193 = sext i1 %192 to i32
>  %194 = bitcast i32 %193 to float
>  %195 = bitcast float %194 to i32
>  %196 = icmp ne i32 %195, 0
>  %.37 = select i1 %196, float 0.000000e+00, float %177
>  %197 = fcmp oge float %181, 0.000000e+00
>  %198 = sext i1 %197 to i32
>  %199 = bitcast i32 %198 to float
>  %200 = bitcast float %199 to i32
>  %201 = icmp ne i32 %200, 0
>  %202 = bitcast float %. to i32
>  %203 = bitcast float %temp24.0 to i32
>  %204 = bitcast float %.37 to i32
>  %205 = insertelement <4 x i32> undef, i32 %202, i32 0
>  %206 = insertelement <4 x i32> %205, i32 %203, i32 1
>  %207 = insertelement <4 x i32> %206, i32 %204, i32 2
>  %208 = insertelement <4 x i32> %207, i32 undef, i32 3
>  %209 = bitcast <8 x i32> %76 to <32 x i8>
>  %210 = bitcast <4 x i32> %78 to <16 x i8>
>  %211 = call <4 x float> @llvm.SI.sample.v4i32(<4 x i32> %208, <32 x i8> %209, <16 x i8> %210, i32 3)
>  %212 = extractelement <4 x float> %211, i32 0
>  %213 = extractelement <4 x float> %211, i32 1
>  %214 = extractelement <4 x float> %211, i32 2
>  %215 = call float @llvm.AMDGPU.lrp(float %163, float %212, float %.)
>  %216 = call float @llvm.AMDGPU.lrp(float %163, float %213, float %temp24.0)
>  %217 = call float @llvm.AMDGPU.lrp(float %163, float %214, float %.37)
>  %218 = fmul float %83, 1.000000e+00
>  %219 = fmul float %84, -1.000000e+00
>  %220 = fadd float %218, 0.000000e+00
>  %221 = fadd float %219, 1.000000e+00
>  %222 = bitcast float %220 to i32
>  %223 = bitcast float %221 to i32
>  %224 = insertelement <2 x i32> undef, i32 %222, i32 0
>  %225 = insertelement <2 x i32> %224, i32 %223, i32 1
>  %226 = bitcast <8 x i32> %52 to <32 x i8>
>  %227 = bitcast <4 x i32> %54 to <16 x i8>
>  %228 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %225, <32 x i8> %226, <16 x i8> %227, i32 2)
>  %229 = extractelement <4 x float> %228, i32 0
>  %230 = extractelement <4 x float> %228, i32 1
>  %231 = extractelement <4 x float> %228, i32 2
>  %232 = extractelement <4 x float> %228, i32 3
>  %233 = fsub float -0.000000e+00, %232
>  %234 = fadd float %233, 1.000000e+00
>  %235 = fmul float %215, %234
>  %236 = fadd float %235, %229
>  %237 = fmul float %216, %234
>  %238 = fadd float %237, %230
>  %239 = fmul float %217, %234
>  %240 = fadd float %239, %231
>  %241 = fadd float %218, 0.000000e+00
>  %242 = fadd float %219, 1.000000e+00
>  %243 = bitcast float %241 to i32
>  %244 = bitcast float %242 to i32
>  %245 = insertelement <2 x i32> undef, i32 %243, i32 0
>  %246 = insertelement <2 x i32> %245, i32 %244, i32 1
>  %247 = bitcast <8 x i32> %64 to <32 x i8>
>  %248 = bitcast <4 x i32> %66 to <16 x i8>
>  %249 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %246, <32 x i8> %247, <16 x i8> %248, i32 2)
>  %250 = extractelement <4 x float> %249, i32 0
>  %251 = extractelement <4 x float> %249, i32 2
>  %252 = bitcast float %81 to i32
>  %253 = bitcast float %82 to i32
>  %254 = insertelement <2 x i32> undef, i32 %252, i32 0
>  %255 = insertelement <2 x i32> %254, i32 %253, i32 1
>  %256 = bitcast <8 x i32> %68 to <32 x i8>
>  %257 = bitcast <4 x i32> %70 to <16 x i8>
>  %258 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %255, <32 x i8> %256, <16 x i8> %257, i32 2)
>  %259 = extractelement <4 x float> %258, i32 0
>  %260 = extractelement <4 x float> %258, i32 1
>  %261 = extractelement <4 x float> %258, i32 2
>  %262 = call float @llvm.AMDGPU.lrp(float %250, float %236, float %259)
>  %263 = call float @llvm.AMDGPU.lrp(float %250, float %238, float %260)
>  %264 = call float @llvm.AMDGPU.lrp(float %250, float %240, float %261)
>  %265 = fsub float -0.000000e+00, %250
>  %266 = fmul float %251, %265
>  %267 = fadd float %266, %250
>  %268 = bitcast float %262 to i32
>  %269 = bitcast float %263 to i32
>  %270 = bitcast float %264 to i32
>  %271 = insertelement <4 x i32> undef, i32 %268, i32 0
>  %272 = insertelement <4 x i32> %271, i32 %269, i32 1
>  %273 = insertelement <4 x i32> %272, i32 %270, i32 2
>  %274 = insertelement <4 x i32> %273, i32 undef, i32 3
>  %275 = bitcast <8 x i32> %72 to <32 x i8>
>  %276 = bitcast <4 x i32> %74 to <16 x i8>
>  %277 = call <4 x float> @llvm.SI.sample.v4i32(<4 x i32> %274, <32 x i8> %275, <16 x i8> %276, i32 3)
>  %278 = extractelement <4 x float> %277, i32 0
>  %279 = extractelement <4 x float> %277, i32 1
>  %280 = extractelement <4 x float> %277, i32 2
>  %281 = call float @llvm.AMDGPU.lrp(float %267, float %278, float %262)
>  %282 = call float @llvm.AMDGPU.lrp(float %267, float %279, float %263)
>  %283 = call float @llvm.AMDGPU.lrp(float %267, float %280, float %264)
>  %284 = fsub float -0.000000e+00, %281
>  %285 = fadd float %284, %41
>  %286 = fsub float -0.000000e+00, %282
>  %287 = fadd float %286, %42
>  %288 = fsub float -0.000000e+00, %283
>  %289 = fadd float %288, %43
>  %290 = fmul float %124, %285
>  %291 = fadd float %290, %281
>  %292 = fmul float %124, %287
>  %293 = fadd float %292, %282
>  %294 = fmul float %124, %289
>  %295 = fadd float %294, %283
>  %296 = call i32 @llvm.SI.packf16(float %291, float %293)
>  %297 = bitcast i32 %296 to float
>  %298 = call i32 @llvm.SI.packf16(float %295, float 1.000000e+00)
>  %299 = bitcast i32 %298 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %297, float %299, float %297, float %299)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.exp.(float) #2
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.clamp.(float, float, float) #2
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v4i32(<4 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDGPU.lrp(float, float, float) #2
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                         ; BEFE0A7E
>	s_mov_b32 m0, s9                                                             ; BEFC0309
>	v_interp_p1_f32 v2, v0, 0, 2, [m0]                                           ; C8080800
>	v_interp_p2_f32 v2, [v2], v1, 0, 2, [m0]                                     ; C8090801
>	v_interp_p1_f32 v3, v0, 1, 2, [m0]                                           ; C80C0900
>	v_interp_p2_f32 v3, [v3], v1, 1, 2, [m0]                                     ; C80D0901
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                                           ; C0800300
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	s_buffer_load_dword s8, s[0:3], 0x11                                         ; C2040111
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v4, s8, v3                                                     ; 10080608
>	s_buffer_load_dword s8, s[0:3], 0xd                                          ; C204010D
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v4, v2, s8, v4                                                     ; D2820004 04101102
>	v_interp_p1_f32 v5, v0, 2, 2, [m0]                                           ; C8140A00
>	v_interp_p2_f32 v5, [v5], v1, 2, 2, [m0]                                     ; C8150A01
>	s_buffer_load_dword s8, s[0:3], 0x15                                         ; C2040115
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v4, v5, s8, v4                                                     ; D2820004 04101105
>	s_buffer_load_dword s8, s[0:3], 0x19                                         ; C2040119
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_add_f32_e32 v4, s8, v4                                                     ; 06080808
>	s_buffer_load_dword s8, s[0:3], 0x13                                         ; C2040113
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v6, s8, v3                                                     ; 100C0608
>	s_buffer_load_dword s8, s[0:3], 0xf                                          ; C204010F
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v6, v2, s8, v6                                                     ; D2820006 04181102
>	s_buffer_load_dword s8, s[0:3], 0x17                                         ; C2040117
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v6, v5, s8, v6                                                     ; D2820006 04181105
>	s_buffer_load_dword s8, s[0:3], 0x1b                                         ; C204011B
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_add_f32_e32 v6, s8, v6                                                     ; 060C0C08
>	v_rcp_f32_e32 v6, v6                                                         ; 7E0C5506
>	v_mul_f32_e32 v4, v4, v6                                                     ; 10080D04
>	v_mad_f32 v4, -0.5, v4, -0.5                                                 ; D2820004 03C608F1
>	v_sub_f32_e32 v8, 1.0, v4                                                    ; 081008F2
>	s_buffer_load_dword s8, s[0:3], 0x10                                         ; C2040110
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v4, s8, v3                                                     ; 10080608
>	s_buffer_load_dword s8, s[0:3], 0xc                                          ; C204010C
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v4, v2, s8, v4                                                     ; D2820004 04101102
>	s_buffer_load_dword s8, s[0:3], 0x14                                         ; C2040114
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v4, v5, s8, v4                                                     ; D2820004 04101105
>	s_buffer_load_dword s8, s[0:3], 0x18                                         ; C2040118
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_add_f32_e32 v4, s8, v4                                                     ; 06080808
>	v_mul_f32_e32 v4, v4, v6                                                     ; 10080D04
>	v_mad_f32 v4, 0.5, v4, -0.5                                                  ; D2820004 03C608F0
>	v_add_f32_e32 v7, 0, v4                                                      ; 060E0880
>	s_load_dwordx4 s[68:71], s[4:5], 0x0                                         ; C0A20500
>	s_load_dwordx4 s[48:51], s[4:5], 0x4                                         ; C0980504
>	s_load_dwordx4 s[96:99], s[4:5], 0x8                                         ; C0B00508
>	s_load_dwordx4 s[72:75], s[4:5], 0xc                                         ; C0A4050C
>	s_load_dwordx4 s[44:47], s[4:5], 0x10                                        ; C0960510
>	s_load_dwordx4 s[32:35], s[4:5], 0x14                                        ; C0900514
>	s_load_dwordx4 s[20:23], s[4:5], 0x18                                        ; C08A0518
>	s_load_dwordx4 s[76:79], s[4:5], 0x1c                                        ; C0A6051C
>	s_load_dwordx8 s[80:87], s[6:7], 0x0                                         ; C0E80700
>	s_load_dwordx8 s[60:67], s[6:7], 0x8                                         ; C0DE0708
>	s_load_dwordx8 s[12:19], s[6:7], 0x10                                        ; C0C60710
>	s_load_dwordx8 s[88:95], s[6:7], 0x18                                        ; C0EC0718
>	s_load_dwordx8 s[52:59], s[6:7], 0x20                                        ; C0DA0720
>	s_load_dwordx8 s[36:43], s[6:7], 0x28                                        ; C0D20728
>	s_load_dwordx8 s[24:31], s[6:7], 0x30                                        ; C0CC0730
>	s_load_dwordx8 s[4:11], s[6:7], 0x38                                         ; C0C20738
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample v4, 1, 0, 0, 0, 0, 0, 0, 0, v[7:8], s[88:95], s[72:75]          ; F0800100 02560407
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_add_f32_e64 v4, 0, v4 clamp                                                ; D2060804 00020880
>	s_buffer_load_dword s72, s[0:3], 0x4                                         ; C2240104
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v6, s72, v4                                                    ; 100C0848
>	v_mad_f32 v4, -v4, s72, 1.0                                                  ; D2820004 23C89104
>	v_interp_p1_f32 v7, v0, 1, 0, [m0]                                           ; C81C0100
>	v_interp_p2_f32 v7, [v7], v1, 1, 0, [m0]                                     ; C81D0101
>	v_sub_f32_e32 v8, 1.0, v7                                                    ; 08100EF2
>	v_interp_p1_f32 v9, v0, 0, 0, [m0]                                           ; C8240000
>	v_interp_p2_f32 v9, [v9], v1, 0, 0, [m0]                                     ; C8250001
>	v_add_f32_e32 v7, 0, v9                                                      ; 060E1280
>	image_sample v[7:9], 7, 0, 0, 0, 0, 0, 0, 0, v[7:8], s[80:87], s[68:71]      ; F0800700 02340707
>	s_buffer_load_dword s68, s[0:3], 0x8                                         ; C2220108
>	s_waitcnt vmcnt(0) lgkmcnt(0)                                                ; BF8C0070
>	v_cmp_ge_f32_e64 s[68:69], -s68, 0                                           ; D00C0044 20010044
>	v_cndmask_b32_e64 v10, 0, -1, s[68:69]                                       ; D200000A 01118280
>	v_cmp_ne_i32_e64 s[68:69], v10, 0                                            ; D10A0044 0001010A
>	v_cndmask_b32_e64 v12, v9, 0, s[68:69]                                       ; D200000C 09110109
>	v_mul_f32_e32 v14, v12, v4                                                   ; 101C090C
>	v_cndmask_b32_e64 v11, v8, 0, s[68:69]                                       ; D200080B 01110108
>	v_cndmask_b32_e64 v10, v7, 0, s[68:69]                                       ; D200000A 19110107
>	image_sample v[7:9], 7, 0, 0, 0, 0, 0, 0, 0, v[10:13], s[4:11], s[76:79]     ; F0800700 0261070A
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v14, v6, v9, v14                                                   ; D282000E 043A1306
>	v_interp_p1_f32 v15, v0, 1, 1, [m0]                                          ; C83C0500
>	v_interp_p2_f32 v15, [v15], v1, 1, 1, [m0]                                   ; C83D0501
>	v_sub_f32_e32 v16, 1.0, v15                                                  ; 08201EF2
>	v_interp_p1_f32 v17, v0, 0, 1, [m0]                                          ; C8440400
>	v_interp_p2_f32 v17, [v17], v1, 0, 1, [m0]                                   ; C8450401
>	v_add_f32_e32 v15, 0, v17                                                    ; 061E2280
>	image_sample v[17:20], 15, 0, 0, 0, 0, 0, 0, 0, v[15:16], s[60:67], s[48:51] ; F0800F00 018F110F
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_sub_f32_e32 v21, 1.0, v20                                                  ; 082A28F2
>	v_mad_f32 v14, v14, v21, v19                                                 ; D282000E 044E2B0E
>	image_sample v[15:16], 5, 0, 0, 0, 0, 0, 0, 0, v[15:16], s[52:59], s[44:47]  ; F0800500 016D0F0F
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_sub_f32_e32 v22, 1.0, v15                                                  ; 082C1EF2
>	v_interp_p1_f32 v24, v0, 3, 0, [m0]                                          ; C8600300
>	v_interp_p2_f32 v24, [v24], v1, 3, 0, [m0]                                   ; C8610301
>	v_interp_p1_f32 v23, v0, 2, 0, [m0]                                          ; C85C0200
>	v_interp_p2_f32 v23, [v23], v1, 2, 0, [m0]                                   ; C85D0201
>	image_sample v[23:25], 7, 0, 0, 0, 0, 0, 0, 0, v[23:24], s[36:43], s[32:35]  ; F0800700 01091717
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mul_f32_e32 v26, v25, v22                                                  ; 10342D19
>	v_mad_f32 v28, v15, v14, v26                                                 ; D282001C 046A1D0F
>	v_mul_f32_e32 v14, v11, v4                                                   ; 101C090B
>	v_mad_f32 v14, v6, v8, v14                                                   ; D282000E 043A1106
>	v_mad_f32 v14, v14, v21, v18                                                 ; D282000E 044A2B0E
>	v_mul_f32_e32 v30, v24, v22                                                  ; 103C2D18
>	v_mad_f32 v27, v15, v14, v30                                                 ; D282001B 047A1D0F
>	v_mul_f32_e32 v4, v10, v4                                                    ; 1008090A
>	v_mad_f32 v4, v6, v7, v4                                                     ; D2820004 04120F06
>	v_mad_f32 v4, v4, v21, v17                                                   ; D2820004 04462B04
>	v_mul_f32_e32 v6, v23, v22                                                   ; 100C2D17
>	v_mad_f32 v26, v15, v4, v6                                                   ; D282001A 041A090F
>	image_sample v[6:8], 7, 0, 0, 0, 0, 0, 0, 0, v[26:29], s[24:31], s[20:23]    ; F0800700 00A6061A
>	v_mad_f32 v4, -v16, v15, v15                                                 ; D2820004 243E1F10
>	v_sub_f32_e32 v9, 1.0, v4                                                    ; 081208F2
>	v_mul_f32_e32 v10, v27, v9                                                   ; 1014131B
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v10, v4, v7, v10                                                   ; D282000A 042A0F04
>	s_buffer_load_dword s4, s[0:3], 0x1d                                         ; C202011D
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v11, s4, v10                                                   ; 08161404
>	s_buffer_load_dword s4, s[0:3], 0x1                                          ; C2020101
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_subrev_f32_e32 v3, s4, v3                                                  ; 0A060604
>	s_buffer_load_dword s4, s[0:3], 0x0                                          ; C2020100
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_subrev_f32_e32 v2, s4, v2                                                  ; 0A040404
>	v_mul_f32_e32 v2, v2, v2                                                     ; 10040502
>	v_mad_f32 v2, v3, v3, v2                                                     ; D2820002 040A0703
>	s_buffer_load_dword s4, s[0:3], 0x2                                          ; C2020102
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_subrev_f32_e32 v3, s4, v5                                                  ; 0A060A04
>	v_mad_f32 v2, v3, v3, v2                                                     ; D2820002 040A0703
>	s_buffer_load_dword s4, s[0:3], 0x21                                         ; C2020121
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v2, s4, v2                                                     ; 10040404
>	s_buffer_load_dword s4, s[0:3], 0x20                                         ; C2020120
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v5, s4, v3                                                     ; 100A0604
>	v_mul_f32_e32 v5, 0x3fb8aa65, v5                                             ; 100A0AFF 3FB8AA65
>	v_exp_f32_e32 v5, v5                                                         ; 7E0A4B05
>	v_sub_f32_e32 v5, 1.0, v5                                                    ; 080A0AF2
>	v_mul_f32_e32 v2, v2, v5                                                     ; 10040B02
>	v_rcp_f32_e32 v3, v3                                                         ; 7E065503
>	v_mul_f32_e32 v2, v2, v3                                                     ; 10040702
>	v_mul_f32_e32 v2, 0x3fb8aa65, v2                                             ; 100404FF 3FB8AA65
>	v_exp_f32_e32 v2, v2                                                         ; 7E044B02
>	v_add_f32_e64 v2, 0, v2 clamp                                                ; D2060802 00020480
>	v_sub_f32_e32 v2, 1.0, v2                                                    ; 080404F2
>	v_interp_p1_f32 v3, v0, 1, 3, [m0]                                           ; C80C0D00
>	v_interp_p2_f32 v3, [v3], v1, 1, 3, [m0]                                     ; C80D0D01
>	v_mad_f32 v0, 0.5, v3, 0.5                                                   ; D2820000 03C206F0
>	s_buffer_load_dword s4, s[0:3], 0x23                                         ; C2020123
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mov_b32_e32 v1, s4                                                         ; 7E020204
>	image_sample v0, 1, 0, 0, 0, 0, 0, 0, 0, v[0:1], s[12:19], s[96:99]          ; F0800100 03030000
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mul_f32_e32 v0, v0, v2                                                     ; 10000500
>	v_mad_f32 v1, v0, v11, v10                                                   ; D2820001 042A1700
>	v_mul_f32_e32 v2, v26, v9                                                    ; 1004131A
>	v_mad_f32 v2, v4, v6, v2                                                     ; D2820002 040A0D04
>	s_buffer_load_dword s4, s[0:3], 0x1c                                         ; C202011C
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v3, s4, v2                                                     ; 08060404
>	v_mad_f32 v2, v0, v3, v2                                                     ; D2820002 040A0700
>	v_cvt_pkrtz_f16_f32_e32 v1, v2, v1                                           ; 5E020302
>	v_mul_f32_e32 v2, v28, v9                                                    ; 1004131C
>	v_mad_f32 v2, v4, v8, v2                                                     ; D2820002 040A1104
>	s_buffer_load_dword s0, s[0:3], 0x1e                                         ; C200011E
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v3, s0, v2                                                     ; 08060400
>	v_mad_f32 v0, v0, v3, v2                                                     ; D2820000 040A0700
>	v_cvt_pkrtz_f16_f32_e64 v0, v0, 1.0                                          ; D25E0000 0001E500
>	exp 15, 0, 1, 1, 1, v1, v0, v1, v0                                           ; F8001C0F 00010001
>	s_endpgm                                                                     ; BF810000
>
>SGPRS: 104
>VGPRS: 32
>Code Size: 952 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL OUT[2], GENERIC[10]
>DCL OUT[3], GENERIC[11]
>DCL OUT[4], GENERIC[12]
>DCL OUT[5], GENERIC[13]
>DCL OUT[6], GENERIC[14]
>DCL OUT[7], GENERIC[15]
>DCL CONST[0..10]
>DCL TEMP[0..7], LOCAL
>IMM[0] FLT32 {    1.0000,     0.0000,    -1.0000,    -0.5000}
>IMM[1] FLT32 {    0.5000,    -0.5000,     0.0000,     0.0000}
>  0: MAD TEMP[0].xy, IN[0], CONST[8].xxxx, CONST[9].xxxx
>  1: MOV TEMP[0].xy, TEMP[0].xyxx
>  2: MAD TEMP[1], CONST[8].yyyy, IMM[0].xyyx, IMM[0].yxzy
>  3: MUL TEMP[2].xy, TEMP[1], IN[0]
>  4: MOV TEMP[1].xy, TEMP[2].xyxx
>  5: MAD TEMP[1].xy, TEMP[1], TEMP[1].zwzw, CONST[9].xxxx
>  6: MOV TEMP[1].xy, TEMP[1].xyxx
>  7: MAD TEMP[2].xy, IN[0], -CONST[8].zzzz, CONST[9].xxxx
>  8: MOV TEMP[2].xy, TEMP[2].xyxx
>  9: MAD TEMP[3], CONST[8].ywwy, IMM[0].yxxy, IMM[0].xyyz
> 10: MUL TEMP[4].xy, TEMP[3], IN[0]
> 11: MOV TEMP[3].xy, TEMP[4].xyxx
> 12: MAD TEMP[4].xy, TEMP[3], TEMP[3].zwzw, CONST[9].xxxx
> 13: MOV TEMP[4].xy, TEMP[4].xyxx
> 14: MUL TEMP[5].xy, CONST[5], IN[0].yyyy
> 15: MOV TEMP[3].xy, TEMP[5].xyxx
> 16: MAD TEMP[5].xy, IN[0].xxxx, CONST[4], TEMP[3]
> 17: MOV TEMP[3].xy, TEMP[5].xyxx
> 18: MAD TEMP[5].xy, CONST[10].xxxx, CONST[6], TEMP[3]
> 19: MOV TEMP[3].xy, TEMP[5].xyxx
> 20: ADD TEMP[5].xy, TEMP[3], CONST[7]
> 21: MOV TEMP[3].xy, TEMP[5].xyxx
> 22: MAD TEMP[5].xy, TEMP[3], IMM[1].xyzz, IMM[0].wwww
> 23: MOV TEMP[5].xy, TEMP[5].xyxx
> 24: MUL TEMP[3], CONST[1], IN[0].yyyy
> 25: MAD TEMP[3], IN[0].xxxx, CONST[0], TEMP[3]
> 26: MAD TEMP[3], CONST[10].xxxx, CONST[2], TEMP[3]
> 27: ADD TEMP[3], TEMP[3], CONST[3]
> 28: MOV TEMP[6].xyz, TEMP[3].xywx
> 29: MOV TEMP[7].xy, IN[0].xyxx
> 30: MOV TEMP[7].z, CONST[10].xxxx
> 31: MOV TEMP[5].zw, IN[0].wwzw
> 32: MOV TEMP[0].zw, IMM[0].xxyx
> 33: MOV TEMP[1].zw, IMM[0].xxyx
> 34: MOV TEMP[2].zw, IMM[0].xxyx
> 35: MOV TEMP[4].zw, IMM[0].xxyx
> 36: MOV TEMP[7].w, IMM[0].xxxx
> 37: MOV TEMP[6].w, IMM[0].xxxx
> 38: MOV OUT[6], TEMP[6]
> 39: MOV OUT[7], TEMP[5]
> 40: MOV OUT[1], TEMP[0]
> 41: MOV OUT[0], TEMP[3]
> 42: MOV OUT[2], TEMP[1]
> 43: MOV OUT[3], TEMP[2]
> 44: MOV OUT[4], TEMP[4]
> 45: MOV OUT[5], TEMP[7]
> 46: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 0)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 4)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %12, i32 8)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %12, i32 12)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %12, i32 16)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %12, i32 20)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %12, i32 24)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %12, i32 28)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %12, i32 32)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %12, i32 36)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %12, i32 40)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %12, i32 44)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %12, i32 48)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %12, i32 52)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %12, i32 56)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %12, i32 60)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %12, i32 64)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %12, i32 68)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %12, i32 80)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %12, i32 84)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %12, i32 96)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %12, i32 100)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %12, i32 112)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %12, i32 116)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %12, i32 128)
>  %38 = call float @llvm.SI.load.const(<16 x i8> %12, i32 132)
>  %39 = call float @llvm.SI.load.const(<16 x i8> %12, i32 136)
>  %40 = call float @llvm.SI.load.const(<16 x i8> %12, i32 140)
>  %41 = call float @llvm.SI.load.const(<16 x i8> %12, i32 144)
>  %42 = call float @llvm.SI.load.const(<16 x i8> %12, i32 160)
>  %43 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %44 = load <16 x i8> addrspace(2)* %43, !tbaa !0
>  %45 = add i32 %5, %7
>  %46 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %44, i32 0, i32 %45)
>  %47 = extractelement <4 x float> %46, i32 0
>  %48 = extractelement <4 x float> %46, i32 1
>  %49 = extractelement <4 x float> %46, i32 2
>  %50 = extractelement <4 x float> %46, i32 3
>  %51 = fmul float %47, %37
>  %52 = fadd float %51, %41
>  %53 = fmul float %48, %37
>  %54 = fadd float %53, %41
>  %55 = fmul float %38, 1.000000e+00
>  %56 = fadd float %55, 0.000000e+00
>  %57 = fmul float %38, 0.000000e+00
>  %58 = fadd float %57, 1.000000e+00
>  %59 = fmul float %38, 0.000000e+00
>  %60 = fadd float %59, -1.000000e+00
>  %61 = fmul float %38, 1.000000e+00
>  %62 = fadd float %61, 0.000000e+00
>  %63 = fmul float %56, %47
>  %64 = fmul float %58, %48
>  %65 = fmul float %63, %60
>  %66 = fadd float %65, %41
>  %67 = fmul float %64, %62
>  %68 = fadd float %67, %41
>  %69 = fsub float -0.000000e+00, %39
>  %70 = fmul float %47, %69
>  %71 = fadd float %70, %41
>  %72 = fsub float -0.000000e+00, %39
>  %73 = fmul float %48, %72
>  %74 = fadd float %73, %41
>  %75 = fmul float %38, 0.000000e+00
>  %76 = fadd float %75, 1.000000e+00
>  %77 = fmul float %40, 1.000000e+00
>  %78 = fadd float %77, 0.000000e+00
>  %79 = fmul float %40, 1.000000e+00
>  %80 = fadd float %79, 0.000000e+00
>  %81 = fmul float %38, 0.000000e+00
>  %82 = fadd float %81, -1.000000e+00
>  %83 = fmul float %76, %47
>  %84 = fmul float %78, %48
>  %85 = fmul float %83, %80
>  %86 = fadd float %85, %41
>  %87 = fmul float %84, %82
>  %88 = fadd float %87, %41
>  %89 = fmul float %31, %48
>  %90 = fmul float %32, %48
>  %91 = fmul float %47, %29
>  %92 = fadd float %91, %89
>  %93 = fmul float %47, %30
>  %94 = fadd float %93, %90
>  %95 = fmul float %42, %33
>  %96 = fadd float %95, %92
>  %97 = fmul float %42, %34
>  %98 = fadd float %97, %94
>  %99 = fadd float %96, %35
>  %100 = fadd float %98, %36
>  %101 = fmul float %99, 5.000000e-01
>  %102 = fadd float %101, -5.000000e-01
>  %103 = fmul float %100, -5.000000e-01
>  %104 = fadd float %103, -5.000000e-01
>  %105 = fmul float %17, %48
>  %106 = fmul float %18, %48
>  %107 = fmul float %19, %48
>  %108 = fmul float %20, %48
>  %109 = fmul float %47, %13
>  %110 = fadd float %109, %105
>  %111 = fmul float %47, %14
>  %112 = fadd float %111, %106
>  %113 = fmul float %47, %15
>  %114 = fadd float %113, %107
>  %115 = fmul float %47, %16
>  %116 = fadd float %115, %108
>  %117 = fmul float %42, %21
>  %118 = fadd float %117, %110
>  %119 = fmul float %42, %22
>  %120 = fadd float %119, %112
>  %121 = fmul float %42, %23
>  %122 = fadd float %121, %114
>  %123 = fmul float %42, %24
>  %124 = fadd float %123, %116
>  %125 = fadd float %118, %25
>  %126 = fadd float %120, %26
>  %127 = fadd float %122, %27
>  %128 = fadd float %124, %28
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %52, float %54, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %66, float %68, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 34, i32 0, float %71, float %74, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 35, i32 0, float %86, float %88, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 36, i32 0, float %47, float %48, float %42, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 37, i32 0, float %125, float %126, float %128, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 38, i32 0, float %102, float %104, float %49, float %50)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %125, float %126, float %127, float %128)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                           ; 4A00000A
>	s_load_dwordx4 s[4:7], s[8:9], 0x0                  ; C0820900
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	buffer_load_format_xyzw v[0:3], v0, s[4:7], 0 idxen ; E00C2000 80010000
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                  ; C0800300
>	s_waitcnt vmcnt(0) lgkmcnt(0)                       ; BF8C0070
>	s_buffer_load_dword s4, s[0:3], 0x24                ; C2020124
>	s_buffer_load_dword s5, s[0:3], 0x20                ; C2028120
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	v_mov_b32_e32 v4, s4                                ; 7E080204
>	v_mad_f32 v4, s5, v1, v4                            ; D2820004 04120205
>	v_mov_b32_e32 v5, s4                                ; 7E0A0204
>	v_mad_f32 v5, s5, v0, v5                            ; D2820005 04160005
>	v_mov_b32_e32 v6, 1.0                               ; 7E0C02F2
>	v_mov_b32_e32 v7, 0                                 ; 7E0E0280
>	exp 15, 32, 0, 0, 0, v5, v4, v7, v6                 ; F800020F 06070405
>	s_buffer_load_dword s5, s[0:3], 0x21                ; C2028121
>	s_waitcnt expcnt(0) lgkmcnt(0)                      ; BF8C000F
>	v_mad_f32 v4, 0, s5, -1.0                           ; D2820004 03CC0A80
>	v_add_f32_e64 v5, 0, s5                             ; D2060005 00000A80
>	v_mul_f32_e32 v8, v0, v5                            ; 10100B00
>	v_mad_f32 v8, v8, v4, s4                            ; D2820008 00120908
>	v_mad_f32 v9, 0, s5, 1.0                            ; D2820009 03C80A80
>	v_mul_f32_e32 v10, v1, v9                           ; 10141301
>	v_mad_f32 v5, v10, v5, s4                           ; D2820005 00120B0A
>	exp 15, 33, 0, 0, 0, v8, v5, v7, v6                 ; F800021F 06070508
>	s_buffer_load_dword s5, s[0:3], 0x22                ; C2028122
>	s_waitcnt expcnt(0)                                 ; BF8C070F
>	v_mov_b32_e32 v5, s4                                ; 7E0A0204
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	v_mad_f32 v5, -v1, s5, v5                           ; D2820005 24140B01
>	v_mov_b32_e32 v8, s4                                ; 7E100204
>	v_mad_f32 v8, -v0, s5, v8                           ; D2820008 24200B00
>	exp 15, 34, 0, 0, 0, v8, v5, v7, v6                 ; F800022F 06070508
>	s_buffer_load_dword s5, s[0:3], 0x23                ; C2028123
>	s_waitcnt expcnt(0) lgkmcnt(0)                      ; BF8C000F
>	v_add_f32_e64 v5, 0, s5                             ; D2060005 00000A80
>	v_mul_f32_e32 v8, v1, v5                            ; 10100B01
>	v_mad_f32 v4, v8, v4, s4                            ; D2820004 00120908
>	v_mul_f32_e32 v8, v0, v9                            ; 10101300
>	v_mad_f32 v5, v8, v5, s4                            ; D2820005 00120B08
>	exp 15, 35, 0, 0, 0, v5, v4, v7, v6                 ; F800023F 06070405
>	s_buffer_load_dword s4, s[0:3], 0x28                ; C2020128
>	s_waitcnt expcnt(0) lgkmcnt(0)                      ; BF8C000F
>	v_mov_b32_e32 v4, s4                                ; 7E080204
>	exp 15, 36, 0, 0, 0, v0, v1, v4, v6                 ; F800024F 06040100
>	s_buffer_load_dword s5, s[0:3], 0x7                 ; C2028107
>	s_waitcnt expcnt(0) lgkmcnt(0)                      ; BF8C000F
>	v_mul_f32_e32 v4, s5, v1                            ; 10080205
>	s_buffer_load_dword s5, s[0:3], 0x3                 ; C2028103
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	v_mad_f32 v4, v0, s5, v4                            ; D2820004 04100B00
>	s_buffer_load_dword s5, s[0:3], 0xb                 ; C202810B
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	v_mov_b32_e32 v5, s5                                ; 7E0A0205
>	v_mad_f32 v4, v5, s4, v4                            ; D2820004 04100905
>	s_buffer_load_dword s5, s[0:3], 0xf                 ; C202810F
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	v_add_f32_e32 v4, s5, v4                            ; 06080805
>	s_buffer_load_dword s5, s[0:3], 0x5                 ; C2028105
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	v_mul_f32_e32 v5, s5, v1                            ; 100A0205
>	s_buffer_load_dword s5, s[0:3], 0x1                 ; C2028101
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	v_mad_f32 v5, v0, s5, v5                            ; D2820005 04140B00
>	s_buffer_load_dword s5, s[0:3], 0x9                 ; C2028109
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	v_mov_b32_e32 v7, s5                                ; 7E0E0205
>	v_mad_f32 v5, v7, s4, v5                            ; D2820005 04140907
>	s_buffer_load_dword s5, s[0:3], 0xd                 ; C202810D
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	v_add_f32_e32 v5, s5, v5                            ; 060A0A05
>	s_buffer_load_dword s5, s[0:3], 0x4                 ; C2028104
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	v_mul_f32_e32 v7, s5, v1                            ; 100E0205
>	s_buffer_load_dword s5, s[0:3], 0x0                 ; C2028100
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	v_mad_f32 v7, v0, s5, v7                            ; D2820007 041C0B00
>	s_buffer_load_dword s5, s[0:3], 0x8                 ; C2028108
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	v_mov_b32_e32 v8, s5                                ; 7E100205
>	v_mad_f32 v7, v8, s4, v7                            ; D2820007 041C0908
>	s_buffer_load_dword s5, s[0:3], 0xc                 ; C202810C
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	v_add_f32_e32 v7, s5, v7                            ; 060E0E05
>	exp 15, 37, 0, 0, 0, v7, v5, v4, v6                 ; F800025F 06040507
>	s_buffer_load_dword s5, s[0:3], 0x14                ; C2028114
>	s_waitcnt expcnt(0) lgkmcnt(0)                      ; BF8C000F
>	v_mul_f32_e32 v6, s5, v1                            ; 100C0205
>	s_buffer_load_dword s5, s[0:3], 0x10                ; C2028110
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	v_mad_f32 v6, v0, s5, v6                            ; D2820006 04180B00
>	s_buffer_load_dword s5, s[0:3], 0x18                ; C2028118
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	v_mov_b32_e32 v8, s5                                ; 7E100205
>	v_mad_f32 v6, v8, s4, v6                            ; D2820006 04180908
>	s_buffer_load_dword s5, s[0:3], 0x1c                ; C202811C
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	v_add_f32_e32 v6, s5, v6                            ; 060C0C05
>	v_mad_f32 v6, 0.5, v6, -0.5                         ; D2820006 03C60CF0
>	s_buffer_load_dword s5, s[0:3], 0x15                ; C2028115
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	v_mul_f32_e32 v8, s5, v1                            ; 10100205
>	s_buffer_load_dword s5, s[0:3], 0x11                ; C2028111
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	v_mad_f32 v8, v0, s5, v8                            ; D2820008 04200B00
>	s_buffer_load_dword s5, s[0:3], 0x19                ; C2028119
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	v_mov_b32_e32 v9, s5                                ; 7E120205
>	v_mad_f32 v8, v9, s4, v8                            ; D2820008 04200909
>	s_buffer_load_dword s5, s[0:3], 0x1d                ; C202811D
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	v_add_f32_e32 v8, s5, v8                            ; 06101005
>	v_mad_f32 v8, -0.5, v8, -0.5                        ; D2820008 03C610F1
>	exp 15, 38, 0, 0, 0, v6, v8, v2, v3                 ; F800026F 03020806
>	s_buffer_load_dword s5, s[0:3], 0x6                 ; C2028106
>	s_waitcnt expcnt(0) lgkmcnt(0)                      ; BF8C000F
>	v_mul_f32_e32 v6, s5, v1                            ; 100C0205
>	s_buffer_load_dword s5, s[0:3], 0x2                 ; C2028102
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	v_mad_f32 v0, v0, s5, v6                            ; D2820000 04180B00
>	s_buffer_load_dword s5, s[0:3], 0xa                 ; C202810A
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	v_mov_b32_e32 v1, s5                                ; 7E020205
>	v_mad_f32 v0, v1, s4, v0                            ; D2820000 04000901
>	s_buffer_load_dword s0, s[0:3], 0xe                 ; C200010E
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	v_add_f32_e32 v0, s0, v0                            ; 06000000
>	exp 15, 12, 0, 1, 0, v7, v5, v0, v4                 ; F80008CF 04000507
>	s_endpgm                                            ; BF810000
>
>SGPRS: 16
>VGPRS: 12
>Code Size: 660 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL IN[1], GENERIC[10], PERSPECTIVE
>DCL IN[2], GENERIC[11], PERSPECTIVE
>DCL IN[3], GENERIC[12], PERSPECTIVE
>DCL IN[4], GENERIC[13], PERSPECTIVE
>DCL IN[5], GENERIC[14], PERSPECTIVE
>DCL IN[6], GENERIC[15], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SAMP[1]
>DCL SAMP[2]
>DCL SAMP[3]
>DCL SAMP[4]
>DCL SAMP[5]
>DCL SAMP[6]
>DCL SAMP[7]
>DCL SAMP[8]
>DCL SAMP[9]
>DCL SAMP[10]
>DCL CONST[0..23]
>DCL TEMP[0..24], LOCAL
>IMM[0] FLT32 {    1.0000,    -1.0000,     0.0000,    -0.0000}
>IMM[1] FLT32 {    4.0000,     2.2000,     1.0000,     2.0000}
>IMM[2] FLT32 {   -1.0000,     1.0000,     4.0000,    -2.0000}
>IMM[3] FLT32 {    0.0000,     0.5000,   -16.0000,    -1.4427}
>IMM[4] FLT32 {   16.0000,    -1.4427,     0.0050,    -0.5000}
>IMM[5] FLT32 {    0.5000,    -0.5000,     6.0000,     0.2500}
>IMM[6] FLT32 {    2.0000,    -2.0000,     6.0000,     2.2000}
>IMM[7] FLT32 {   -0.6700,     0.4545,    -0.4000,     0.4000}
>IMM[8] FLT32 {    1.4427,     0.0000,     0.0000,     0.0000}
>  0: MAD TEMP[0].xy, IN[6], IMM[0].xyxx, IMM[0].zxzz
>  1: MOV TEMP[0].xy, TEMP[0].xyyy
>  2: TEX TEMP[0], TEMP[0], SAMP[8], 2D
>  3: MOV TEMP[1].xz, TEMP[0]
>  4: ABS TEMP[2].x, TEMP[0]
>  5: MOV TEMP[2], -TEMP[2].xxxx
>  6: FSGE TEMP[3].x, TEMP[2].xxxx, IMM[0].zzzz
>  7: UIF TEMP[3].xxxx :0
>  8:   MOV TEMP[3].x, IMM[0].yyyy
>  9: ELSE :0
> 10:   MOV TEMP[3].x, IMM[0].wwww
> 11: ENDIF
> 12: MOV TEMP[3].x, TEMP[3].xxxx
> 13: FSGE TEMP[4].x, TEMP[2].yyyy, IMM[0].zzzz
> 14: UIF TEMP[4].xxxx :0
> 15:   MOV TEMP[4].x, IMM[0].yyyy
> 16: ELSE :0
> 17:   MOV TEMP[4].x, IMM[0].wwww
> 18: ENDIF
> 19: MOV TEMP[3].y, TEMP[4].xxxx
> 20: FSGE TEMP[4].x, TEMP[2].zzzz, IMM[0].zzzz
> 21: UIF TEMP[4].xxxx :0
> 22:   MOV TEMP[4].x, IMM[0].yyyy
> 23: ELSE :0
> 24:   MOV TEMP[4].x, IMM[0].wwww
> 25: ENDIF
> 26: MOV TEMP[3].z, TEMP[4].xxxx
> 27: FSGE TEMP[2].x, TEMP[2].wwww, IMM[0].zzzz
> 28: UIF TEMP[2].xxxx :0
> 29: ELSE :0
> 30: ENDIF
> 31: FSLT TEMP[3].xyz, TEMP[3].xyzz, IMM[0].zzzz
> 32: OR TEMP[4].x, TEMP[3].xxxx, TEMP[3].zzzz
> 33: OR TEMP[3].x, TEMP[4].xxxx, TEMP[3].yyyy
> 34: AND TEMP[3].x, TEMP[3].xxxx, IMM[0].xxxx
> 35: KILL_IF -TEMP[3].xxxx
> 36: MUL TEMP[3].yw, CONST[18].xxzy, IN[4].yyyy
> 37: MOV TEMP[1].yw, TEMP[3].wyww
> 38: MAD TEMP[3].yw, IN[4].xxxx, CONST[17].xxzy, TEMP[1]
> 39: MOV TEMP[1].yw, TEMP[3].wyww
> 40: MAD TEMP[3].yw, IN[4].zzzz, CONST[19].xxzy, TEMP[1]
> 41: MOV TEMP[1].yw, TEMP[3].wyww
> 42: ADD TEMP[3].yw, TEMP[1], CONST[20].xxzy
> 43: MOV TEMP[1].yw, TEMP[3].wyww
> 44: MOV TEMP[3].xy, IN[6].zwww
> 45: TEX TEMP[3].yzw, TEMP[3], SAMP[5], 2D
> 46: MOV TEMP[2].w, TEMP[3].zyzw
> 47: MUL TEMP[3], TEMP[3].wwww, IMM[1].xxxx
> 48: MOV_SAT TEMP[3], TEMP[3]
> 49: MOV TEMP[2].x, TEMP[3].xxxx
> 50: ADD TEMP[4].x, -TEMP[3].xxxx, IMM[0].xxxx
> 51: MOV TEMP[4].x, TEMP[4].xxxx
> 52: MOV TEMP[4].y, IMM[0].zzzz
> 53: MOV TEMP[5].xy, TEMP[4].xyyy
> 54: MOV TEMP[5].w, IMM[0].zzzz
> 55: TXL TEMP[5], TEMP[5], SAMP[4], 2D
> 56: POW TEMP[6].x, TEMP[5].xxxx, IMM[1].yyyy
> 57: POW TEMP[6].y, TEMP[5].yyyy, IMM[1].yyyy
> 58: POW TEMP[6].z, TEMP[5].zzzz, IMM[1].yyyy
> 59: POW TEMP[6].w, TEMP[5].wwww, IMM[1].zzzz
> 60: MOV TEMP[4].w, TEMP[6]
> 61: MAD TEMP[5].xy, TEMP[1].ywzw, IMM[0].xyxx, IMM[0].zxzz
> 62: MOV TEMP[5].xy, TEMP[5].xyyy
> 63: TEX TEMP[5].xyz, TEMP[5], SAMP[3], 2D
> 64: MAD TEMP[7].yw, TEMP[5].xxzy, IMM[1].wwww, IMM[0].yyyy
> 65: DP2 TEMP[5].x, IMM[0].xyyy, TEMP[7].ywww
> 66: DP2 TEMP[7].x, IMM[0].yyyy, TEMP[7].ywww
> 67: MOV TEMP[5].y, TEMP[7].xxxx
> 68: MOV TEMP[5].zw, -TEMP[5].yyxy
> 69: ADD TEMP[5], TEMP[5], IMM[1].wwww
> 70: MUL TEMP[5], TEMP[5], TEMP[5]
> 71: MUL TEMP[5], TEMP[5], TEMP[5]
> 72: DP4 TEMP[7].x, IMM[0].xxxx, TEMP[5]
> 73: RCP TEMP[7].x, TEMP[7].xxxx
> 74: MUL TEMP[5], TEMP[7].xxxx, TEMP[5]
> 75: RCP TEMP[7].x, CONST[0].xxxx
> 76: MUL TEMP[5], TEMP[7].xxxx, TEMP[5]
> 77: MUL TEMP[7], TEMP[5], TEMP[5]
> 78: MOV TEMP[8].xy, IN[0].xyyy
> 79: TEX TEMP[8], TEMP[8], SAMP[0], 2D
> 80: MOV TEMP[9].xy, IN[0].xyyy
> 81: TEX TEMP[9].xw, TEMP[9], SAMP[1], 2D
> 82: MOV TEMP[10].w, TEMP[9].wwww
> 83: MAD TEMP[11].yw, TEMP[8].xxzy, IMM[1].wwww, IMM[0].yyyy
> 84: MOV TEMP[1].yw, TEMP[11].wyww
> 85: MUL TEMP[11].yz, TEMP[5].xxxx, TEMP[1].xyww
> 86: MOV TEMP[2].yz, TEMP[11].zyzz
> 87: MAD TEMP[9].z, TEMP[9].xxxx, IMM[1].wwww, IMM[0].yyyy
> 88: MOV TEMP[10].z, TEMP[9].zzzz
> 89: MOV TEMP[10].xy, TEMP[8].zwzz
> 90: MOV TEMP[8].xy, IN[1].xyyy
> 91: TEX TEMP[8], TEMP[8], SAMP[0], 2D
> 92: MOV TEMP[9].zw, TEMP[8].wwzw
> 93: MOV TEMP[12].xy, IN[1].xyyy
> 94: TEX TEMP[12].xw, TEMP[12], SAMP[1], 2D
> 95: MOV TEMP[13].w, TEMP[12].wwww
> 96: ADD TEMP[14].xy, TEMP[8], TEMP[8]
> 97: MOV TEMP[9].xy, TEMP[14].xyxx
> 98: MAD TEMP[14].xy, TEMP[9], IMM[2].xyzy, IMM[2].yxzy
> 99: MOV TEMP[9].xy, TEMP[14].xyxx
>100: MUL TEMP[14].xy, TEMP[5].yyyy, TEMP[9]
>101: MOV TEMP[9].xy, TEMP[14].xyxx
>102: MAD TEMP[15].yw, TEMP[1], TEMP[5].xxxx, TEMP[9].xxzy
>103: MOV TEMP[1].yw, TEMP[15].wyww
>104: MAD TEMP[12].z, TEMP[12].xxxx, IMM[2].wwww, IMM[0].xxxx
>105: MOV TEMP[13].z, TEMP[12].zzzz
>106: MOV TEMP[13].xy, TEMP[8].zwzz
>107: MUL TEMP[8].xyz, TEMP[7].yyyy, TEMP[13]
>108: MOV TEMP[13].xyz, TEMP[8].xyzx
>109: MAD TEMP[8].xyz, TEMP[10], TEMP[7].xxxx, TEMP[13]
>110: MOV TEMP[10].xyz, TEMP[8].xyzx
>111: ADD TEMP[8].xy, TEMP[2].yzzw, TEMP[2].yzzw
>112: MOV TEMP[5].xy, TEMP[8].xyxx
>113: MUL TEMP[8].xy, TEMP[9], TEMP[5]
>114: MOV TEMP[13].xy, TEMP[8].xyxx
>115: DP2 TEMP[8].x, TEMP[11].yzzz, TEMP[14].yxxx
>116: MOV TEMP[13].z, TEMP[8].xxxx
>117: ADD TEMP[8].xyz, TEMP[10], TEMP[13]
>118: MOV TEMP[10].xyz, TEMP[8].xyzx
>119: MOV TEMP[8].xy, IN[2].xyyy
>120: TEX TEMP[8], TEMP[8], SAMP[0], 2D
>121: MOV TEMP[13].w, TEMP[8].wwww
>122: MOV TEMP[12].xy, IN[2].xyyy
>123: TEX TEMP[12].xw, TEMP[12], SAMP[1], 2D
>124: MOV TEMP[15].w, TEMP[12].wwww
>125: MAD TEMP[16].xy, TEMP[8], IMM[2].wwww, IMM[0].xxxx
>126: MOV TEMP[7].xy, TEMP[16].xyxx
>127: MUL TEMP[16].zw, TEMP[5].zzzz, TEMP[16].xyxy
>128: MOV TEMP[9].zw, TEMP[16].wwzw
>129: MAD TEMP[17].yw, TEMP[7].xxzy, TEMP[5].zzzz, TEMP[1]
>130: MOV TEMP[1].yw, TEMP[17].wyww
>131: MAD TEMP[12].z, TEMP[12].xxxx, IMM[1].wwww, IMM[0].yyyy
>132: MOV TEMP[15].z, TEMP[12].zzzz
>133: MOV TEMP[15].xy, TEMP[8].zwzz
>134: MAD TEMP[8].xyz, TEMP[15], TEMP[7].zzzz, TEMP[10]
>135: MOV TEMP[7].xyz, TEMP[8].xyzx
>136: MUL TEMP[8].xy, TEMP[5], TEMP[16].zwzw
>137: MOV TEMP[10].xy, TEMP[8].xyxx
>138: DP2 TEMP[8].x, TEMP[11].yzzz, TEMP[16].wzzz
>139: MOV TEMP[10].z, TEMP[8].xxxx
>140: ADD TEMP[8].xyz, TEMP[7], TEMP[10]
>141: MOV TEMP[7].xyz, TEMP[8].xyzx
>142: ADD TEMP[8].xy, TEMP[9], TEMP[9]
>143: MOV TEMP[10].xy, TEMP[8].xyxx
>144: MUL TEMP[8].xy, TEMP[16].zwzw, TEMP[10]
>145: MOV TEMP[13].xy, TEMP[8].xyxx
>146: DP2 TEMP[8].x, TEMP[14].xyyy, TEMP[16].wzzz
>147: MOV TEMP[13].z, TEMP[8].xxxx
>148: ADD TEMP[8].xyz, TEMP[7], TEMP[13]
>149: MOV TEMP[7].xyz, TEMP[8].xyzx
>150: MOV TEMP[8].xy, IN[3].xyyy
>151: TEX TEMP[8], TEMP[8], SAMP[0], 2D
>152: MOV TEMP[13].zw, TEMP[8].wwzw
>153: MOV TEMP[12].xy, IN[3].xyyy
>154: TEX TEMP[12].xw, TEMP[12], SAMP[1], 2D
>155: MOV TEMP[15].w, TEMP[12].wwww
>156: ADD TEMP[17].zw, TEMP[8].xyyx, TEMP[8].xyyx
>157: MOV TEMP[10].zw, TEMP[17].wwzw
>158: MAD TEMP[17].zw, TEMP[10], IMM[0].ywyx, IMM[0].ywxy
>159: MOV TEMP[10].zw, TEMP[17].wwzw
>160: MUL TEMP[18].xy, TEMP[5].wwww, TEMP[17].zwzw
>161: MOV TEMP[13].xy, TEMP[18].xyxx
>162: MAD TEMP[17].xy, TEMP[17].wzzw, TEMP[5].wwww, TEMP[1].ywzw
>163: MOV TEMP[19].xy, TEMP[17].xyxx
>164: MAD TEMP[12].z, TEMP[12].xxxx, IMM[2].wwww, IMM[0].xxxx
>165: MOV TEMP[15].z, TEMP[12].zzzz
>166: MOV TEMP[15].xy, TEMP[8].zwzz
>167: MAD TEMP[7].xyz, TEMP[15], TEMP[7].wwww, TEMP[7]
>168: MUL TEMP[8].xy, TEMP[5], TEMP[13].yxzw
>169: MOV TEMP[5].xy, TEMP[8].xyxx
>170: DP2 TEMP[8].x, TEMP[11].yzzz, TEMP[18].xyyy
>171: MOV TEMP[5].z, TEMP[8].xxxx
>172: ADD TEMP[7].yzw, TEMP[5].xxyz, TEMP[7].xxyz
>173: MOV TEMP[2].yzw, TEMP[7].zyzw
>174: MUL TEMP[7].xy, TEMP[10], TEMP[13].yxzw
>175: MOV TEMP[5].xy, TEMP[7].xyxx
>176: DP2 TEMP[7].x, TEMP[14].xyyy, TEMP[18].xyyy
>177: MOV TEMP[5].z, TEMP[7].xxxx
>178: ADD TEMP[7].yzw, TEMP[2], TEMP[5].xxyz
>179: MOV TEMP[2].yzw, TEMP[7].zyzw
>180: MUL TEMP[7].yw, TEMP[9].xzzw, TEMP[13].xyzx
>181: MOV TEMP[1].yw, TEMP[7].wyww
>182: ADD TEMP[7].xy, TEMP[1].ywzw, TEMP[1].ywzw
>183: MOV TEMP[5].xy, TEMP[7].xyxx
>184: DP2 TEMP[7].x, TEMP[16].zwww, TEMP[18].xyyy
>185: MOV TEMP[5].z, TEMP[7].xxxx
>186: ADD TEMP[7].yzw, TEMP[2], TEMP[5].xxyz
>187: MOV TEMP[2].yzw, TEMP[7].zyzw
>188: MAD TEMP[7].yzw, TEMP[17].xxyx, -TEMP[17].xxyy, TEMP[2]
>189: MOV TEMP[2].yz, TEMP[7].zyzw
>190: MOV TEMP[19].z, IMM[0].xxxx
>191: DP3 TEMP[8].x, CONST[1].xyzz, TEMP[19].xyzz
>192: MOV_SAT TEMP[8].x, TEMP[8].xxxx
>193: ADD TEMP[11].xyz, CONST[8], -IN[4]
>194: MOV TEMP[5].xyz, TEMP[11].xyzx
>195: DP3 TEMP[11].x, TEMP[11].xyzz, TEMP[11].xyzz
>196: MAX TEMP[11].x, TEMP[11].xxxx, IMM[3].xxxx
>197: RSQ TEMP[11].x, TEMP[11].xxxx
>198: MUL TEMP[12].xyz, TEMP[11].xxxx, TEMP[5]
>199: MAD TEMP[14].xyz, TEMP[5], TEMP[11].xxxx, -CONST[5]
>200: MOV TEMP[9].xyz, TEMP[14].xyzx
>201: MAD TEMP[14].xyz, CONST[5].wwww, TEMP[9], CONST[5]
>202: MOV TEMP[9].xyz, TEMP[14].xyzx
>203: DP3 TEMP[14].x, TEMP[14].xyzz, TEMP[14].xyzz
>204: MAX TEMP[14].x, TEMP[14].xxxx, IMM[3].xxxx
>205: RSQ TEMP[14].x, TEMP[14].xxxx
>206: MAD TEMP[14].xyz, TEMP[9], TEMP[14].xxxx, CONST[1]
>207: MOV TEMP[9].xyz, TEMP[14].xyzx
>208: RCP TEMP[14].x, TEMP[14].zzzz
>209: MAD TEMP[14].xy, TEMP[9], TEMP[14].xxxx, -TEMP[19]
>210: RCP TEMP[15].x, CONST[3].wwww
>211: ADD TEMP[16].zw, TEMP[2].xyyz, TEMP[15].xxxx
>212: MUL TEMP[17].w, TEMP[7].wwww, TEMP[7].wwww
>213: MAD TEMP[18].x, TEMP[16].zzzz, TEMP[16].wwww, -TEMP[17].wwww
>214: MUL TEMP[20].y, TEMP[14].xxxx, TEMP[14].xxxx
>215: ADD TEMP[7].w, TEMP[7].wwww, TEMP[7].wwww
>216: MOV TEMP[2].w, TEMP[7].wwww
>217: MUL TEMP[21].x, TEMP[14].xxxx, TEMP[7].wwww
>218: MAD TEMP[22].z, TEMP[16].zzzz, TEMP[14].yyyy, -TEMP[21].xxxx
>219: MUL TEMP[22].z, TEMP[14].yyyy, TEMP[22].zzzz
>220: MAD TEMP[16].z, TEMP[20].yyyy, TEMP[16].wwww, TEMP[22].zzzz
>221: MUL TEMP[16].z, TEMP[16].zzzz, IMM[3].yyyy
>222: RCP TEMP[22].x, TEMP[18].xxxx
>223: MUL TEMP[23].z, TEMP[22].xxxx, TEMP[16].zzzz
>224: MOV TEMP[24].x, -TEMP[18].xxxx
>225: FSGE TEMP[24].x, TEMP[24].xxxx, IMM[0].zzzz
>226: UIF TEMP[24].xxxx :0
>227:   MOV TEMP[24].x, IMM[0].xxxx
>228: ELSE :0
>229:   MOV TEMP[24].x, IMM[0].zzzz
>230: ENDIF
>231: MOV TEMP[10].w, TEMP[24].xxxx
>232: MAD TEMP[16].z, TEMP[16].zzzz, TEMP[22].xxxx, IMM[3].zzzz
>233: FSGE TEMP[16].x, TEMP[16].zzzz, IMM[0].zzzz
>234: UIF TEMP[16].xxxx :0
>235:   MOV TEMP[16].x, IMM[0].xxxx
>236: ELSE :0
>237:   MOV TEMP[16].x, IMM[0].zzzz
>238: ENDIF
>239: MOV TEMP[9].z, TEMP[16].xxxx
>240: MUL TEMP[16].w, TEMP[23].zzzz, IMM[3].wwww
>241: EX2 TEMP[16].x, TEMP[16].wwww
>242: MAX TEMP[18].x, TEMP[18].xxxx, IMM[3].xxxx
>243: RSQ TEMP[18].x, TEMP[18].xxxx
>244: MUL TEMP[16].w, TEMP[16].xxxx, TEMP[18].xxxx
>245: MOV TEMP[9].w, TEMP[16].wwww
>246: MAD TEMP[15].xz, TEMP[15].xxxx, IMM[4].xxxx, TEMP[2].yyzw
>247: MAD TEMP[18].w, TEMP[15].xxxx, TEMP[15].zzzz, -TEMP[17].wwww
>248: MAD TEMP[21].x, TEMP[15].xxxx, TEMP[14].yyyy, -TEMP[21].xxxx
>249: MUL TEMP[14].x, TEMP[14].yyyy, TEMP[21].xxxx
>250: MAD TEMP[14].x, TEMP[20].yyyy, TEMP[15].zzzz, TEMP[14].xxxx
>251: MUL TEMP[14].x, TEMP[14].xxxx, IMM[3].yyyy
>252: RCP TEMP[15].x, TEMP[18].wwww
>253: MOV TEMP[9].y, TEMP[15].xxxx
>254: MUL TEMP[20].x, TEMP[15].xxxx, TEMP[14].xxxx
>255: MOV TEMP[21].x, -TEMP[18].wwww
>256: FSGE TEMP[21].x, TEMP[21].xxxx, IMM[0].zzzz
>257: UIF TEMP[21].xxxx :0
>258:   MOV TEMP[21].x, IMM[0].xxxx
>259: ELSE :0
>260:   MOV TEMP[21].x, IMM[0].zzzz
>261: ENDIF
>262: MOV TEMP[10].y, TEMP[21].xxxx
>263: MAD TEMP[14].x, TEMP[14].xxxx, TEMP[15].xxxx, IMM[3].zzzz
>264: FSGE TEMP[14].x, TEMP[14].xxxx, IMM[0].zzzz
>265: UIF TEMP[14].xxxx :0
>266:   MOV TEMP[14].x, IMM[0].xxxx
>267: ELSE :0
>268:   MOV TEMP[14].x, IMM[0].zzzz
>269: ENDIF
>270: MOV TEMP[9].x, TEMP[14].xxxx
>271: ADD TEMP[9].xz, TEMP[9], TEMP[10].yyww
>272: MUL TEMP[14].y, TEMP[20].xxxx, IMM[3].wwww
>273: EX2 TEMP[14].x, TEMP[14].yyyy
>274: MAX TEMP[15].x, TEMP[18].wwww, IMM[3].xxxx
>275: RSQ TEMP[15].x, TEMP[15].xxxx
>276: MUL TEMP[14].w, TEMP[15].xxxx, TEMP[14].xxxx
>277: MOV TEMP[5].w, TEMP[14].wwww
>278: MAD TEMP[11].xyz, TEMP[5], TEMP[11].xxxx, -CONST[6]
>279: MOV TEMP[5].xyz, TEMP[11].xyzx
>280: MAD TEMP[11].xyz, CONST[6].wwww, TEMP[5], CONST[6]
>281: MOV TEMP[5].xyz, TEMP[11].xyzx
>282: DP3 TEMP[11].x, TEMP[11].xyzz, TEMP[11].xyzz
>283: MAX TEMP[11].x, TEMP[11].xxxx, IMM[3].xxxx
>284: RSQ TEMP[11].x, TEMP[11].xxxx
>285: MAD TEMP[11].xyz, TEMP[5], TEMP[11].xxxx, CONST[2]
>286: MOV TEMP[5].xyz, TEMP[11].xyzx
>287: RCP TEMP[11].x, TEMP[11].zzzz
>288: MAD TEMP[11].xy, TEMP[5], TEMP[11].xxxx, -TEMP[19]
>289: RCP TEMP[14].x, CONST[4].wwww
>290: ADD TEMP[14].xy, TEMP[14].xxxx, TEMP[2].yzzw
>291: MAD TEMP[15].w, TEMP[14].xxxx, TEMP[14].yyyy, -TEMP[17].wwww
>292: MUL TEMP[17].z, TEMP[11].xxxx, TEMP[11].xxxx
>293: MUL TEMP[7].w, TEMP[7].wwww, TEMP[11].xxxx
>294: MAD TEMP[7].w, TEMP[14].xxxx, TEMP[11].yyyy, -TEMP[7].wwww
>295: MUL TEMP[7].w, TEMP[11].yyyy, TEMP[7].wwww
>296: MAD TEMP[7].w, TEMP[17].zzzz, TEMP[14].yyyy, TEMP[7].wwww
>297: MUL TEMP[7].w, TEMP[7].wwww, IMM[3].yyyy
>298: RCP TEMP[5].x, TEMP[15].wwww
>299: MUL TEMP[11].y, TEMP[7].wwww, TEMP[5].xxxx
>300: MOV TEMP[5].y, TEMP[11].yyyy
>301: MOV TEMP[14].x, -TEMP[15].wwww
>302: FSGE TEMP[14].x, TEMP[14].xxxx, IMM[0].zzzz
>303: UIF TEMP[14].xxxx :0
>304:   MOV TEMP[14].x, IMM[0].xxxx
>305: ELSE :0
>306:   MOV TEMP[14].x, IMM[0].zzzz
>307: ENDIF
>308: MOV TEMP[5].z, TEMP[14].xxxx
>309: MAD TEMP[7].w, TEMP[7].wwww, TEMP[5].xxxx, IMM[3].zzzz
>310: FSGE TEMP[7].x, TEMP[7].wwww, IMM[0].zzzz
>311: UIF TEMP[7].xxxx :0
>312:   MOV TEMP[7].x, IMM[0].xxxx
>313: ELSE :0
>314:   MOV TEMP[7].x, IMM[0].zzzz
>315: ENDIF
>316: ADD TEMP[7].w, TEMP[7].xxxx, TEMP[14].xxxx
>317: MOV TEMP[2].w, TEMP[7].wwww
>318: MUL TEMP[14].xw, TEMP[5].yyzw, IMM[4].yyxz
>319: MAX TEMP[17].x, TEMP[15].wwww, IMM[3].xxxx
>320: RSQ TEMP[17].x, TEMP[17].xxxx
>321: EX2 TEMP[18].x, TEMP[14].xxxx
>322: MUL TEMP[17].x, TEMP[17].xxxx, TEMP[18].xxxx
>323: MUL TEMP[18].xyz, CONST[10].xyww, IN[4].yyyy
>324: MOV TEMP[10].xyz, TEMP[18].xyzx
>325: MAD TEMP[18].xyz, IN[4].xxxx, CONST[9].xyww, TEMP[10]
>326: MOV TEMP[10].xyz, TEMP[18].xyzx
>327: MAD TEMP[18].xyz, IN[4].zzzz, CONST[11].xyww, TEMP[10]
>328: MOV TEMP[10].xyz, TEMP[18].xyzx
>329: ADD TEMP[18].xyz, TEMP[10], CONST[12].xyww
>330: MOV TEMP[10].xyz, TEMP[18].xyzx
>331: RCP TEMP[18].x, TEMP[18].zzzz
>332: MUL TEMP[18].xy, TEMP[18].xxxx, TEMP[10]
>333: MOV TEMP[10].xy, TEMP[18].xyxx
>334: MAD TEMP[18].xy, TEMP[10], IMM[5], IMM[4].wwww
>335: MOV TEMP[10].xy, TEMP[18].xyxx
>336: MAD TEMP[18].yz, TEMP[2], IMM[5].zzzz, TEMP[10].xxyw
>337: MOV TEMP[2].yz, TEMP[18].zyzz
>338: MAD TEMP[18].xy, TEMP[15].wwww, IMM[1].wwww, TEMP[2].yzzw
>339: MAD TEMP[20].xy, TEMP[15].wwww, IMM[6], TEMP[2].yzzw
>340: MOV TEMP[20].xy, TEMP[20].xyyy
>341: TEX TEMP[20], TEMP[20], SAMP[2], 2D
>342: MUL TEMP[13], TEMP[20], IMM[5].wwww
>343: MOV TEMP[18].xy, TEMP[18].xyyy
>344: TEX TEMP[18], TEMP[18], SAMP[2], 2D
>345: MAD TEMP[10], TEMP[18], IMM[5].wwww, TEMP[13]
>346: MAD TEMP[18].xy, TEMP[15].wwww, IMM[6].yzzw, TEMP[2].yzzw
>347: MOV TEMP[18].xy, TEMP[18].xyyy
>348: TEX TEMP[18], TEMP[18], SAMP[2], 2D
>349: MAD TEMP[10], TEMP[18], IMM[5].wwww, TEMP[10]
>350: MAD TEMP[15].yz, TEMP[15].wwww, IMM[2].wwww, TEMP[2]
>351: MOV TEMP[15].xy, TEMP[15].yzzz
>352: TEX TEMP[15], TEMP[15], SAMP[2], 2D
>353: MOV TEMP[13].w, TEMP[15].wwww
>354: MAD TEMP[10], TEMP[15], IMM[5].wwww, TEMP[10]
>355: ABS TEMP[15].x, TEMP[10].xxxx
>356: LG2 TEMP[13].x, TEMP[15].xxxx
>357: ABS TEMP[15].x, TEMP[10].yyyy
>358: LG2 TEMP[15].x, TEMP[15].xxxx
>359: MOV TEMP[13].y, TEMP[15].xxxx
>360: ABS TEMP[15].x, TEMP[10].zzzz
>361: LG2 TEMP[15].x, TEMP[15].xxxx
>362: MOV TEMP[13].z, TEMP[15].xxxx
>363: MUL TEMP[15].xyz, TEMP[13], IMM[1].yyyy
>364: EX2 TEMP[13].x, TEMP[15].xxxx
>365: EX2 TEMP[18].x, TEMP[15].yyyy
>366: MOV TEMP[13].y, TEMP[18].xxxx
>367: EX2 TEMP[15].x, TEMP[15].zzzz
>368: MOV TEMP[13].z, TEMP[15].xxxx
>369: DP3 TEMP[12].x, TEMP[19].xyzz, TEMP[12].xyzz
>370: ADD TEMP[12].w, -TEMP[12].xxxx, IMM[0].xxxx
>371: MUL TEMP[15].y, TEMP[12].wwww, TEMP[12].wwww
>372: MUL TEMP[15].y, TEMP[15].yyyy, TEMP[15].yyyy
>373: MUL TEMP[12].w, TEMP[12].wwww, TEMP[15].yyyy
>374: MUL TEMP[15].xyz, TEMP[16].wwww, CONST[3]
>375: MOV TEMP[16], -TEMP[9].zzzz
>376: FSGE TEMP[18].x, TEMP[16].xxxx, IMM[0].zzzz
>377: UIF TEMP[18].xxxx :0
>378:   MOV TEMP[18].x, TEMP[15].xxxx
>379: ELSE :0
>380:   MOV TEMP[18].x, IMM[0].zzzz
>381: ENDIF
>382: MOV TEMP[18].x, TEMP[18].xxxx
>383: FSGE TEMP[19].x, TEMP[16].yyyy, IMM[0].zzzz
>384: UIF TEMP[19].xxxx :0
>385:   MOV TEMP[19].x, TEMP[15].yyyy
>386: ELSE :0
>387:   MOV TEMP[19].x, IMM[0].zzzz
>388: ENDIF
>389: MOV TEMP[18].y, TEMP[19].xxxx
>390: FSGE TEMP[19].x, TEMP[16].zzzz, IMM[0].zzzz
>391: UIF TEMP[19].xxxx :0
>392:   MOV TEMP[15].x, TEMP[15].zzzz
>393: ELSE :0
>394:   MOV TEMP[15].x, IMM[0].zzzz
>395: ENDIF
>396: MOV TEMP[18].z, TEMP[15].xxxx
>397: FSGE TEMP[15].x, TEMP[16].wwww, IMM[0].zzzz
>398: UIF TEMP[15].xxxx :0
>399: ELSE :0
>400: ENDIF
>401: MUL TEMP[15].yzw, TEMP[17].xxxx, CONST[4].xxyz
>402: MOV TEMP[7], -TEMP[7].wwww
>403: FSGE TEMP[16].x, TEMP[7].xxxx, IMM[0].zzzz
>404: UIF TEMP[16].xxxx :0
>405: ELSE :0
>406: ENDIF
>407: FSGE TEMP[17].x, TEMP[7].yyyy, IMM[0].zzzz
>408: UIF TEMP[17].xxxx :0
>409:   MOV TEMP[17].x, TEMP[15].yyyy
>410: ELSE :0
>411:   MOV TEMP[17].x, IMM[0].zzzz
>412: ENDIF
>413: MOV TEMP[16].y, TEMP[17].xxxx
>414: FSGE TEMP[17].x, TEMP[7].zzzz, IMM[0].zzzz
>415: UIF TEMP[17].xxxx :0
>416:   MOV TEMP[17].x, TEMP[15].zzzz
>417: ELSE :0
>418:   MOV TEMP[17].x, IMM[0].zzzz
>419: ENDIF
>420: MOV TEMP[16].z, TEMP[17].xxxx
>421: FSGE TEMP[7].x, TEMP[7].wwww, IMM[0].zzzz
>422: UIF TEMP[7].xxxx :0
>423:   MOV TEMP[7].x, TEMP[15].wwww
>424: ELSE :0
>425:   MOV TEMP[7].x, IMM[0].zzzz
>426: ENDIF
>427: MOV TEMP[16].w, TEMP[7].xxxx
>428: MOV TEMP[2].yzw, TEMP[16].zyzw
>429: ADD TEMP[7].yzw, TEMP[2], TEMP[18].xxyz
>430: MOV TEMP[2].yzw, TEMP[7].zyzw
>431: MAD TEMP[7].yzw, TEMP[2], TEMP[12].wwww, TEMP[6].xxyz
>432: MOV TEMP[2].yzw, TEMP[7].zyzw
>433: MUL TEMP[7].xyz, TEMP[6], CONST[3]
>434: MUL TEMP[7].xzw, TEMP[14].wwww, TEMP[7].xyyz
>435: MOV TEMP[9], -TEMP[9].xxxx
>436: FSGE TEMP[12].x, TEMP[9].xxxx, IMM[0].zzzz
>437: UIF TEMP[12].xxxx :0
>438:   MOV TEMP[12].x, TEMP[7].xxxx
>439: ELSE :0
>440:   MOV TEMP[12].x, IMM[0].zzzz
>441: ENDIF
>442: MOV TEMP[12].x, TEMP[12].xxxx
>443: FSGE TEMP[14].x, TEMP[9].yyyy, IMM[0].zzzz
>444: UIF TEMP[14].xxxx :0
>445: ELSE :0
>446: ENDIF
>447: FSGE TEMP[14].x, TEMP[9].zzzz, IMM[0].zzzz
>448: UIF TEMP[14].xxxx :0
>449:   MOV TEMP[14].x, TEMP[7].zzzz
>450: ELSE :0
>451:   MOV TEMP[14].x, IMM[0].zzzz
>452: ENDIF
>453: MOV TEMP[12].z, TEMP[14].xxxx
>454: FSGE TEMP[9].x, TEMP[9].wwww, IMM[0].zzzz
>455: UIF TEMP[9].xxxx :0
>456:   MOV TEMP[7].x, TEMP[7].wwww
>457: ELSE :0
>458:   MOV TEMP[7].x, IMM[0].zzzz
>459: ENDIF
>460: MOV TEMP[12].w, TEMP[7].xxxx
>461: MAD TEMP[7].yzw, TEMP[2], TEMP[8].xxxx, TEMP[12].xxzw
>462: MOV TEMP[2].yzw, TEMP[7].zyzw
>463: MAD TEMP[7].yzw, TEMP[13].xxyz, CONST[23].xxxx, TEMP[2]
>464: MOV TEMP[2].w, TEMP[7].zyzw
>465: MUL TEMP[8].y, TEMP[11].yyyy, TEMP[11].yyyy
>466: MUL TEMP[8].y, TEMP[11].yyyy, TEMP[8].yyyy
>467: MUL TEMP[3].y, TEMP[3].xxxx, TEMP[8].yyyy
>468: MAD TEMP[3].xyz, TEMP[3].yyyy, TEMP[6], TEMP[7].yzww
>469: MOV TEMP[2].xyz, TEMP[3].xyzx
>470: MUL TEMP[3].xyz, CONST[14].xyww, IN[4].yyyy
>471: MOV TEMP[4].xyz, TEMP[3].xyzx
>472: MAD TEMP[3].xyz, IN[4].xxxx, CONST[13].xyww, TEMP[4]
>473: MOV TEMP[4].xyz, TEMP[3].xyzx
>474: MAD TEMP[3].xyz, IN[4].zzzz, CONST[15].xyww, TEMP[4]
>475: MOV TEMP[4].xyz, TEMP[3].xyzx
>476: ADD TEMP[3].xyz, TEMP[4], CONST[16].xyww
>477: RCP TEMP[7].x, TEMP[3].zzzz
>478: MUL TEMP[3].yw, TEMP[7].xxxx, TEMP[3].xxzy
>479: MOV TEMP[1].yw, TEMP[3].wyww
>480: MAD TEMP[3].yw, TEMP[1], IMM[5].xxzy, IMM[4].wwww
>481: MOV TEMP[1].yw, TEMP[3].wyww
>482: MAD TEMP[3].xy, TEMP[1].ywzw, IMM[0].xyxx, IMM[0].zxzz
>483: MOV TEMP[3].xy, TEMP[3].xyyy
>484: TEX TEMP[3].xw, TEMP[3], SAMP[10], 2D
>485: MOV TEMP[5].w, TEMP[3].wwww
>486: MOV_SAT TEMP[3].x, TEMP[3].xxxx
>487: MUL TEMP[3].y, TEMP[3].xxxx, CONST[21].xxxx
>488: MUL TEMP[3].xyz, TEMP[3].yyyy, TEMP[2]
>489: MOV TEMP[4].xyz, TEMP[3].xyzx
>490: MAD TEMP[3].xyz, TEMP[4], IMM[7].xxxx, TEMP[2]
>491: MAD TEMP[6], TEMP[10].wwww, TEMP[6].wwww, TEMP[6].wwww
>492: MOV_SAT TEMP[6], TEMP[6]
>493: MOV TEMP[6].w, TEMP[6].wwww
>494: ABS TEMP[7].x, TEMP[3].xxxx
>495: LG2 TEMP[4].x, TEMP[7].xxxx
>496: ABS TEMP[7].x, TEMP[3].yyyy
>497: LG2 TEMP[7].x, TEMP[7].xxxx
>498: MOV TEMP[4].y, TEMP[7].xxxx
>499: ABS TEMP[3].x, TEMP[3].zzzz
>500: LG2 TEMP[3].x, TEMP[3].xxxx
>501: MOV TEMP[4].z, TEMP[3].xxxx
>502: MUL TEMP[3].xyz, TEMP[4], IMM[7].yyyy
>503: EX2 TEMP[4].x, TEMP[3].xxxx
>504: EX2 TEMP[7].x, TEMP[3].yyyy
>505: MOV TEMP[4].y, TEMP[7].xxxx
>506: EX2 TEMP[3].x, TEMP[3].zzzz
>507: MOV TEMP[4].z, TEMP[3].xxxx
>508: MAD TEMP[3].xy, IN[6], IMM[0].xyxx, IMM[0].zxzz
>509: MOV TEMP[3].xy, TEMP[3].xyyy
>510: TEX TEMP[3], TEMP[3], SAMP[7], 2D
>511: MOV TEMP[2].w, TEMP[3].wwww
>512: ADD TEMP[7].y, -TEMP[3].wwww, IMM[0].xxxx
>513: MAD TEMP[3].xyz, TEMP[4], TEMP[7].yyyy, TEMP[3]
>514: MOV TEMP[2].xyz, TEMP[3].xyzx
>515: ADD TEMP[3].xyz, TEMP[2], IMM[7].zzzz
>516: MOV TEMP[2].xyz, TEMP[3].xyzx
>517: MAD TEMP[3].xyz, TEMP[0].xxxx, TEMP[2], IMM[7].wwww
>518: MOV TEMP[2].xyz, TEMP[3].xyzx
>519: MOV TEMP[3].xyz, TEMP[3].xyzz
>520: TEX TEMP[3], TEMP[3], SAMP[9], 3D
>521: MAD TEMP[0].x, TEMP[0].zzzz, -TEMP[0].xxxx, TEMP[0].xxxx
>522: LRP TEMP[0].xyz, TEMP[0].xxxx, TEMP[3], TEMP[2]
>523: MOV TEMP[5].xyz, TEMP[0].xyzx
>524: RCP TEMP[1].x, IN[5].zzzz
>525: MUL TEMP[2].x, TEMP[1].xxxx, IN[5].yyyy
>526: MAD TEMP[2].x, TEMP[2].xxxx, IMM[3].yyyy, IMM[3].yyyy
>527: MOV TEMP[1].x, TEMP[2].xxxx
>528: MOV TEMP[1].y, CONST[7].wwww
>529: MOV TEMP[2].xy, TEMP[1].xyyy
>530: TEX TEMP[2].x, TEMP[2], SAMP[6], 2D
>531: MOV TEMP[1].x, TEMP[2].xxxx
>532: ADD TEMP[3].yzw, -CONST[8].xxyz, IN[4].xxyz
>533: MOV TEMP[1].w, TEMP[3].zyzw
>534: DP3 TEMP[4].x, TEMP[3].yzww, TEMP[3].yzww
>535: MOV TEMP[1].y, TEMP[4].xxxx
>536: MUL TEMP[1].yz, TEMP[1].xyww, CONST[7].xyxw
>537: MUL TEMP[4].z, TEMP[1].zzzz, IMM[8].xxxx
>538: EX2 TEMP[4].x, TEMP[4].zzzz
>539: ADD TEMP[4].z, -TEMP[4].xxxx, IMM[0].xxxx
>540: MUL TEMP[1].y, TEMP[4].zzzz, TEMP[1].yyyy
>541: RCP TEMP[3].x, TEMP[3].wwww
>542: MUL TEMP[1].y, TEMP[3].xxxx, TEMP[1].yyyy
>543: MUL TEMP[1].y, TEMP[1].yyyy, IMM[8].xxxx
>544: EX2 TEMP[1].x, TEMP[1].yyyy
>545: MOV_SAT TEMP[1].x, TEMP[1].xxxx
>546: ADD TEMP[1].y, -TEMP[1].xxxx, IMM[0].xxxx
>547: MUL TEMP[1].x, TEMP[1].yyyy, TEMP[2].xxxx
>548: ADD TEMP[0].yzw, -TEMP[0].xxyz, CONST[22].xxyz
>549: MAD TEMP[0].xyz, TEMP[1].xxxx, TEMP[0].yzww, TEMP[5]
>550: MOV TEMP[6].xyz, TEMP[0].xyzx
>551: MOV OUT[0], TEMP[6]
>552: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %23 = load <16 x i8> addrspace(2)* %22, !tbaa !0
>  %24 = call float @llvm.SI.load.const(<16 x i8> %23, i32 0)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %23, i32 16)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %23, i32 20)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %23, i32 24)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %23, i32 32)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %23, i32 36)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %23, i32 40)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %23, i32 48)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %23, i32 52)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %23, i32 56)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %23, i32 60)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %23, i32 64)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %23, i32 68)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %23, i32 72)
>  %38 = call float @llvm.SI.load.const(<16 x i8> %23, i32 76)
>  %39 = call float @llvm.SI.load.const(<16 x i8> %23, i32 80)
>  %40 = call float @llvm.SI.load.const(<16 x i8> %23, i32 84)
>  %41 = call float @llvm.SI.load.const(<16 x i8> %23, i32 88)
>  %42 = call float @llvm.SI.load.const(<16 x i8> %23, i32 92)
>  %43 = call float @llvm.SI.load.const(<16 x i8> %23, i32 96)
>  %44 = call float @llvm.SI.load.const(<16 x i8> %23, i32 100)
>  %45 = call float @llvm.SI.load.const(<16 x i8> %23, i32 104)
>  %46 = call float @llvm.SI.load.const(<16 x i8> %23, i32 108)
>  %47 = call float @llvm.SI.load.const(<16 x i8> %23, i32 112)
>  %48 = call float @llvm.SI.load.const(<16 x i8> %23, i32 116)
>  %49 = call float @llvm.SI.load.const(<16 x i8> %23, i32 124)
>  %50 = call float @llvm.SI.load.const(<16 x i8> %23, i32 128)
>  %51 = call float @llvm.SI.load.const(<16 x i8> %23, i32 132)
>  %52 = call float @llvm.SI.load.const(<16 x i8> %23, i32 136)
>  %53 = call float @llvm.SI.load.const(<16 x i8> %23, i32 144)
>  %54 = call float @llvm.SI.load.const(<16 x i8> %23, i32 148)
>  %55 = call float @llvm.SI.load.const(<16 x i8> %23, i32 156)
>  %56 = call float @llvm.SI.load.const(<16 x i8> %23, i32 160)
>  %57 = call float @llvm.SI.load.const(<16 x i8> %23, i32 164)
>  %58 = call float @llvm.SI.load.const(<16 x i8> %23, i32 172)
>  %59 = call float @llvm.SI.load.const(<16 x i8> %23, i32 176)
>  %60 = call float @llvm.SI.load.const(<16 x i8> %23, i32 180)
>  %61 = call float @llvm.SI.load.const(<16 x i8> %23, i32 188)
>  %62 = call float @llvm.SI.load.const(<16 x i8> %23, i32 192)
>  %63 = call float @llvm.SI.load.const(<16 x i8> %23, i32 196)
>  %64 = call float @llvm.SI.load.const(<16 x i8> %23, i32 204)
>  %65 = call float @llvm.SI.load.const(<16 x i8> %23, i32 208)
>  %66 = call float @llvm.SI.load.const(<16 x i8> %23, i32 212)
>  %67 = call float @llvm.SI.load.const(<16 x i8> %23, i32 220)
>  %68 = call float @llvm.SI.load.const(<16 x i8> %23, i32 224)
>  %69 = call float @llvm.SI.load.const(<16 x i8> %23, i32 228)
>  %70 = call float @llvm.SI.load.const(<16 x i8> %23, i32 236)
>  %71 = call float @llvm.SI.load.const(<16 x i8> %23, i32 240)
>  %72 = call float @llvm.SI.load.const(<16 x i8> %23, i32 244)
>  %73 = call float @llvm.SI.load.const(<16 x i8> %23, i32 252)
>  %74 = call float @llvm.SI.load.const(<16 x i8> %23, i32 256)
>  %75 = call float @llvm.SI.load.const(<16 x i8> %23, i32 260)
>  %76 = call float @llvm.SI.load.const(<16 x i8> %23, i32 268)
>  %77 = call float @llvm.SI.load.const(<16 x i8> %23, i32 272)
>  %78 = call float @llvm.SI.load.const(<16 x i8> %23, i32 276)
>  %79 = call float @llvm.SI.load.const(<16 x i8> %23, i32 288)
>  %80 = call float @llvm.SI.load.const(<16 x i8> %23, i32 292)
>  %81 = call float @llvm.SI.load.const(<16 x i8> %23, i32 304)
>  %82 = call float @llvm.SI.load.const(<16 x i8> %23, i32 308)
>  %83 = call float @llvm.SI.load.const(<16 x i8> %23, i32 320)
>  %84 = call float @llvm.SI.load.const(<16 x i8> %23, i32 324)
>  %85 = call float @llvm.SI.load.const(<16 x i8> %23, i32 336)
>  %86 = call float @llvm.SI.load.const(<16 x i8> %23, i32 352)
>  %87 = call float @llvm.SI.load.const(<16 x i8> %23, i32 356)
>  %88 = call float @llvm.SI.load.const(<16 x i8> %23, i32 360)
>  %89 = call float @llvm.SI.load.const(<16 x i8> %23, i32 368)
>  %90 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %91 = load <8 x i32> addrspace(2)* %90, !tbaa !0
>  %92 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %93 = load <4 x i32> addrspace(2)* %92, !tbaa !0
>  %94 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 1
>  %95 = load <8 x i32> addrspace(2)* %94, !tbaa !0
>  %96 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 1
>  %97 = load <4 x i32> addrspace(2)* %96, !tbaa !0
>  %98 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 2
>  %99 = load <8 x i32> addrspace(2)* %98, !tbaa !0
>  %100 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 2
>  %101 = load <4 x i32> addrspace(2)* %100, !tbaa !0
>  %102 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 3
>  %103 = load <8 x i32> addrspace(2)* %102, !tbaa !0
>  %104 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 3
>  %105 = load <4 x i32> addrspace(2)* %104, !tbaa !0
>  %106 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 4
>  %107 = load <8 x i32> addrspace(2)* %106, !tbaa !0
>  %108 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 4
>  %109 = load <4 x i32> addrspace(2)* %108, !tbaa !0
>  %110 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 5
>  %111 = load <8 x i32> addrspace(2)* %110, !tbaa !0
>  %112 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 5
>  %113 = load <4 x i32> addrspace(2)* %112, !tbaa !0
>  %114 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 6
>  %115 = load <8 x i32> addrspace(2)* %114, !tbaa !0
>  %116 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 6
>  %117 = load <4 x i32> addrspace(2)* %116, !tbaa !0
>  %118 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 7
>  %119 = load <8 x i32> addrspace(2)* %118, !tbaa !0
>  %120 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 7
>  %121 = load <4 x i32> addrspace(2)* %120, !tbaa !0
>  %122 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 8
>  %123 = load <8 x i32> addrspace(2)* %122, !tbaa !0
>  %124 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 8
>  %125 = load <4 x i32> addrspace(2)* %124, !tbaa !0
>  %126 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 9
>  %127 = load <8 x i32> addrspace(2)* %126, !tbaa !0
>  %128 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 9
>  %129 = load <4 x i32> addrspace(2)* %128, !tbaa !0
>  %130 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 10
>  %131 = load <8 x i32> addrspace(2)* %130, !tbaa !0
>  %132 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 10
>  %133 = load <4 x i32> addrspace(2)* %132, !tbaa !0
>  %134 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %135 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %136 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %5, <2 x i32> %7)
>  %137 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %5, <2 x i32> %7)
>  %138 = call float @llvm.SI.fs.interp(i32 0, i32 2, i32 %5, <2 x i32> %7)
>  %139 = call float @llvm.SI.fs.interp(i32 1, i32 2, i32 %5, <2 x i32> %7)
>  %140 = call float @llvm.SI.fs.interp(i32 0, i32 3, i32 %5, <2 x i32> %7)
>  %141 = call float @llvm.SI.fs.interp(i32 1, i32 3, i32 %5, <2 x i32> %7)
>  %142 = call float @llvm.SI.fs.interp(i32 0, i32 4, i32 %5, <2 x i32> %7)
>  %143 = call float @llvm.SI.fs.interp(i32 1, i32 4, i32 %5, <2 x i32> %7)
>  %144 = call float @llvm.SI.fs.interp(i32 2, i32 4, i32 %5, <2 x i32> %7)
>  %145 = call float @llvm.SI.fs.interp(i32 1, i32 5, i32 %5, <2 x i32> %7)
>  %146 = call float @llvm.SI.fs.interp(i32 2, i32 5, i32 %5, <2 x i32> %7)
>  %147 = call float @llvm.SI.fs.interp(i32 0, i32 6, i32 %5, <2 x i32> %7)
>  %148 = call float @llvm.SI.fs.interp(i32 1, i32 6, i32 %5, <2 x i32> %7)
>  %149 = call float @llvm.SI.fs.interp(i32 2, i32 6, i32 %5, <2 x i32> %7)
>  %150 = call float @llvm.SI.fs.interp(i32 3, i32 6, i32 %5, <2 x i32> %7)
>  %151 = fmul float %147, 1.000000e+00
>  %152 = fadd float %151, 0.000000e+00
>  %153 = fmul float %148, -1.000000e+00
>  %154 = fadd float %153, 1.000000e+00
>  %155 = bitcast float %152 to i32
>  %156 = bitcast float %154 to i32
>  %157 = insertelement <2 x i32> undef, i32 %155, i32 0
>  %158 = insertelement <2 x i32> %157, i32 %156, i32 1
>  %159 = bitcast <8 x i32> %123 to <32 x i8>
>  %160 = bitcast <4 x i32> %125 to <16 x i8>
>  %161 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %158, <32 x i8> %159, <16 x i8> %160, i32 2)
>  %162 = extractelement <4 x float> %161, i32 0
>  %163 = extractelement <4 x float> %161, i32 2
>  %164 = call float @fabs(float %162)
>  %165 = fsub float -0.000000e+00, %164
>  %166 = fsub float -0.000000e+00, %164
>  %167 = fsub float -0.000000e+00, %164
>  %168 = fsub float -0.000000e+00, %164
>  %169 = fcmp oge float %165, 0.000000e+00
>  %170 = sext i1 %169 to i32
>  %171 = bitcast i32 %170 to float
>  %172 = bitcast float %171 to i32
>  %173 = icmp ne i32 %172, 0
>  %. = select i1 %173, float -1.000000e+00, float -0.000000e+00
>  %174 = fcmp oge float %166, 0.000000e+00
>  %175 = sext i1 %174 to i32
>  %176 = bitcast i32 %175 to float
>  %177 = bitcast float %176 to i32
>  %178 = icmp ne i32 %177, 0
>  %temp16.0 = select i1 %178, float -1.000000e+00, float -0.000000e+00
>  %179 = fcmp oge float %167, 0.000000e+00
>  %180 = sext i1 %179 to i32
>  %181 = bitcast i32 %180 to float
>  %182 = bitcast float %181 to i32
>  %183 = icmp ne i32 %182, 0
>  %.163 = select i1 %183, float -1.000000e+00, float -0.000000e+00
>  %184 = fcmp oge float %168, 0.000000e+00
>  %185 = sext i1 %184 to i32
>  %186 = bitcast i32 %185 to float
>  %187 = bitcast float %186 to i32
>  %188 = icmp ne i32 %187, 0
>  %189 = fcmp olt float %., 0.000000e+00
>  %190 = sext i1 %189 to i32
>  %191 = fcmp olt float %temp16.0, 0.000000e+00
>  %192 = sext i1 %191 to i32
>  %193 = fcmp olt float %.163, 0.000000e+00
>  %194 = sext i1 %193 to i32
>  %195 = bitcast i32 %190 to float
>  %196 = bitcast i32 %192 to float
>  %197 = bitcast i32 %194 to float
>  %198 = bitcast float %195 to i32
>  %199 = bitcast float %197 to i32
>  %200 = or i32 %198, %199
>  %201 = bitcast i32 %200 to float
>  %202 = bitcast float %201 to i32
>  %203 = bitcast float %196 to i32
>  %204 = or i32 %202, %203
>  %205 = bitcast i32 %204 to float
>  %206 = bitcast float %205 to i32
>  %207 = and i32 %206, 1065353216
>  %208 = bitcast i32 %207 to float
>  %209 = fsub float -0.000000e+00, %208
>  %210 = fsub float -0.000000e+00, %208
>  %211 = fsub float -0.000000e+00, %208
>  %212 = fsub float -0.000000e+00, %208
>  call void @llvm.AMDGPU.kill(float %209)
>  call void @llvm.AMDGPU.kill(float %210)
>  call void @llvm.AMDGPU.kill(float %211)
>  call void @llvm.AMDGPU.kill(float %212)
>  %213 = fmul float %79, %143
>  %214 = fmul float %80, %143
>  %215 = fmul float %142, %77
>  %216 = fadd float %215, %213
>  %217 = fmul float %142, %78
>  %218 = fadd float %217, %214
>  %219 = fmul float %144, %81
>  %220 = fadd float %219, %216
>  %221 = fmul float %144, %82
>  %222 = fadd float %221, %218
>  %223 = fadd float %220, %83
>  %224 = fadd float %222, %84
>  %225 = bitcast float %149 to i32
>  %226 = bitcast float %150 to i32
>  %227 = insertelement <2 x i32> undef, i32 %225, i32 0
>  %228 = insertelement <2 x i32> %227, i32 %226, i32 1
>  %229 = bitcast <8 x i32> %111 to <32 x i8>
>  %230 = bitcast <4 x i32> %113 to <16 x i8>
>  %231 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %228, <32 x i8> %229, <16 x i8> %230, i32 2)
>  %232 = extractelement <4 x float> %231, i32 3
>  %233 = fmul float %232, 4.000000e+00
>  %234 = fmul float %232, 4.000000e+00
>  %235 = fmul float %232, 4.000000e+00
>  %236 = fmul float %232, 4.000000e+00
>  %237 = call float @llvm.AMDIL.clamp.(float %233, float 0.000000e+00, float 1.000000e+00)
>  %238 = call float @llvm.AMDIL.clamp.(float %234, float 0.000000e+00, float 1.000000e+00)
>  %239 = call float @llvm.AMDIL.clamp.(float %235, float 0.000000e+00, float 1.000000e+00)
>  %240 = call float @llvm.AMDIL.clamp.(float %236, float 0.000000e+00, float 1.000000e+00)
>  %241 = fsub float -0.000000e+00, %237
>  %242 = fadd float %241, 1.000000e+00
>  %243 = bitcast float %242 to i32
>  %244 = bitcast float 0.000000e+00 to i32
>  %245 = bitcast float 0.000000e+00 to i32
>  %246 = insertelement <4 x i32> undef, i32 %243, i32 0
>  %247 = insertelement <4 x i32> %246, i32 %244, i32 1
>  %248 = insertelement <4 x i32> %247, i32 %245, i32 2
>  %249 = insertelement <4 x i32> %248, i32 undef, i32 3
>  %250 = bitcast <8 x i32> %107 to <32 x i8>
>  %251 = bitcast <4 x i32> %109 to <16 x i8>
>  %252 = call <4 x float> @llvm.SI.samplel.v4i32(<4 x i32> %249, <32 x i8> %250, <16 x i8> %251, i32 2)
>  %253 = extractelement <4 x float> %252, i32 0
>  %254 = extractelement <4 x float> %252, i32 1
>  %255 = extractelement <4 x float> %252, i32 2
>  %256 = extractelement <4 x float> %252, i32 3
>  %257 = call float @llvm.pow.f32(float %253, float 0x40019999A0000000)
>  %258 = call float @llvm.pow.f32(float %254, float 0x40019999A0000000)
>  %259 = call float @llvm.pow.f32(float %255, float 0x40019999A0000000)
>  %260 = call float @llvm.pow.f32(float %256, float 1.000000e+00)
>  %261 = fmul float %223, 1.000000e+00
>  %262 = fadd float %261, 0.000000e+00
>  %263 = fmul float %224, -1.000000e+00
>  %264 = fadd float %263, 1.000000e+00
>  %265 = bitcast float %262 to i32
>  %266 = bitcast float %264 to i32
>  %267 = insertelement <2 x i32> undef, i32 %265, i32 0
>  %268 = insertelement <2 x i32> %267, i32 %266, i32 1
>  %269 = bitcast <8 x i32> %103 to <32 x i8>
>  %270 = bitcast <4 x i32> %105 to <16 x i8>
>  %271 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %268, <32 x i8> %269, <16 x i8> %270, i32 2)
>  %272 = extractelement <4 x float> %271, i32 0
>  %273 = extractelement <4 x float> %271, i32 1
>  %274 = fmul float %272, 2.000000e+00
>  %275 = fadd float %274, -1.000000e+00
>  %276 = fmul float %273, 2.000000e+00
>  %277 = fadd float %276, -1.000000e+00
>  %278 = fmul float 1.000000e+00, %275
>  %279 = fmul float -1.000000e+00, %277
>  %280 = fadd float %278, %279
>  %281 = fmul float -1.000000e+00, %275
>  %282 = fmul float -1.000000e+00, %277
>  %283 = fadd float %281, %282
>  %284 = fsub float -0.000000e+00, %280
>  %285 = fsub float -0.000000e+00, %283
>  %286 = fadd float %280, 2.000000e+00
>  %287 = fadd float %283, 2.000000e+00
>  %288 = fadd float %284, 2.000000e+00
>  %289 = fadd float %285, 2.000000e+00
>  %290 = fmul float %286, %286
>  %291 = fmul float %287, %287
>  %292 = fmul float %288, %288
>  %293 = fmul float %289, %289
>  %294 = fmul float %290, %290
>  %295 = fmul float %291, %291
>  %296 = fmul float %292, %292
>  %297 = fmul float %293, %293
>  %298 = fmul float 1.000000e+00, %294
>  %299 = fmul float 1.000000e+00, %295
>  %300 = fadd float %298, %299
>  %301 = fmul float 1.000000e+00, %296
>  %302 = fadd float %300, %301
>  %303 = fmul float 1.000000e+00, %297
>  %304 = fadd float %302, %303
>  %305 = fdiv float 1.000000e+00, %304
>  %306 = fmul float %305, %294
>  %307 = fmul float %305, %295
>  %308 = fmul float %305, %296
>  %309 = fmul float %305, %297
>  %310 = fdiv float 1.000000e+00, %24
>  %311 = fmul float %310, %306
>  %312 = fmul float %310, %307
>  %313 = fmul float %310, %308
>  %314 = fmul float %310, %309
>  %315 = fmul float %311, %311
>  %316 = fmul float %312, %312
>  %317 = fmul float %313, %313
>  %318 = fmul float %314, %314
>  %319 = bitcast float %134 to i32
>  %320 = bitcast float %135 to i32
>  %321 = insertelement <2 x i32> undef, i32 %319, i32 0
>  %322 = insertelement <2 x i32> %321, i32 %320, i32 1
>  %323 = bitcast <8 x i32> %91 to <32 x i8>
>  %324 = bitcast <4 x i32> %93 to <16 x i8>
>  %325 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %322, <32 x i8> %323, <16 x i8> %324, i32 2)
>  %326 = extractelement <4 x float> %325, i32 0
>  %327 = extractelement <4 x float> %325, i32 1
>  %328 = extractelement <4 x float> %325, i32 2
>  %329 = extractelement <4 x float> %325, i32 3
>  %330 = bitcast float %134 to i32
>  %331 = bitcast float %135 to i32
>  %332 = insertelement <2 x i32> undef, i32 %330, i32 0
>  %333 = insertelement <2 x i32> %332, i32 %331, i32 1
>  %334 = bitcast <8 x i32> %95 to <32 x i8>
>  %335 = bitcast <4 x i32> %97 to <16 x i8>
>  %336 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %333, <32 x i8> %334, <16 x i8> %335, i32 2)
>  %337 = extractelement <4 x float> %336, i32 0
>  %338 = fmul float %326, 2.000000e+00
>  %339 = fadd float %338, -1.000000e+00
>  %340 = fmul float %327, 2.000000e+00
>  %341 = fadd float %340, -1.000000e+00
>  %342 = fmul float %311, %339
>  %343 = fmul float %311, %341
>  %344 = fmul float %337, 2.000000e+00
>  %345 = fadd float %344, -1.000000e+00
>  %346 = bitcast float %136 to i32
>  %347 = bitcast float %137 to i32
>  %348 = insertelement <2 x i32> undef, i32 %346, i32 0
>  %349 = insertelement <2 x i32> %348, i32 %347, i32 1
>  %350 = bitcast <8 x i32> %91 to <32 x i8>
>  %351 = bitcast <4 x i32> %93 to <16 x i8>
>  %352 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %349, <32 x i8> %350, <16 x i8> %351, i32 2)
>  %353 = extractelement <4 x float> %352, i32 0
>  %354 = extractelement <4 x float> %352, i32 1
>  %355 = extractelement <4 x float> %352, i32 2
>  %356 = extractelement <4 x float> %352, i32 3
>  %357 = bitcast float %136 to i32
>  %358 = bitcast float %137 to i32
>  %359 = insertelement <2 x i32> undef, i32 %357, i32 0
>  %360 = insertelement <2 x i32> %359, i32 %358, i32 1
>  %361 = bitcast <8 x i32> %95 to <32 x i8>
>  %362 = bitcast <4 x i32> %97 to <16 x i8>
>  %363 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %360, <32 x i8> %361, <16 x i8> %362, i32 2)
>  %364 = extractelement <4 x float> %363, i32 0
>  %365 = fadd float %353, %353
>  %366 = fadd float %354, %354
>  %367 = fmul float %365, -1.000000e+00
>  %368 = fadd float %367, 1.000000e+00
>  %369 = fmul float %366, 1.000000e+00
>  %370 = fadd float %369, -1.000000e+00
>  %371 = fmul float %312, %368
>  %372 = fmul float %312, %370
>  %373 = fmul float %339, %311
>  %374 = fadd float %373, %371
>  %375 = fmul float %341, %311
>  %376 = fadd float %375, %372
>  %377 = fmul float %364, -2.000000e+00
>  %378 = fadd float %377, 1.000000e+00
>  %379 = fmul float %316, %355
>  %380 = fmul float %316, %356
>  %381 = fmul float %316, %378
>  %382 = fmul float %328, %315
>  %383 = fadd float %382, %379
>  %384 = fmul float %329, %315
>  %385 = fadd float %384, %380
>  %386 = fmul float %345, %315
>  %387 = fadd float %386, %381
>  %388 = fadd float %342, %342
>  %389 = fadd float %343, %343
>  %390 = fmul float %371, %388
>  %391 = fmul float %372, %389
>  %392 = fmul float %342, %372
>  %393 = fmul float %343, %371
>  %394 = fadd float %392, %393
>  %395 = fadd float %383, %390
>  %396 = fadd float %385, %391
>  %397 = fadd float %387, %394
>  %398 = bitcast float %138 to i32
>  %399 = bitcast float %139 to i32
>  %400 = insertelement <2 x i32> undef, i32 %398, i32 0
>  %401 = insertelement <2 x i32> %400, i32 %399, i32 1
>  %402 = bitcast <8 x i32> %91 to <32 x i8>
>  %403 = bitcast <4 x i32> %93 to <16 x i8>
>  %404 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %401, <32 x i8> %402, <16 x i8> %403, i32 2)
>  %405 = extractelement <4 x float> %404, i32 0
>  %406 = extractelement <4 x float> %404, i32 1
>  %407 = extractelement <4 x float> %404, i32 2
>  %408 = extractelement <4 x float> %404, i32 3
>  %409 = bitcast float %138 to i32
>  %410 = bitcast float %139 to i32
>  %411 = insertelement <2 x i32> undef, i32 %409, i32 0
>  %412 = insertelement <2 x i32> %411, i32 %410, i32 1
>  %413 = bitcast <8 x i32> %95 to <32 x i8>
>  %414 = bitcast <4 x i32> %97 to <16 x i8>
>  %415 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %412, <32 x i8> %413, <16 x i8> %414, i32 2)
>  %416 = extractelement <4 x float> %415, i32 0
>  %417 = fmul float %405, -2.000000e+00
>  %418 = fadd float %417, 1.000000e+00
>  %419 = fmul float %406, -2.000000e+00
>  %420 = fadd float %419, 1.000000e+00
>  %421 = fmul float %313, %418
>  %422 = fmul float %313, %420
>  %423 = fmul float %418, %313
>  %424 = fadd float %423, %374
>  %425 = fmul float %420, %313
>  %426 = fadd float %425, %376
>  %427 = fmul float %416, 2.000000e+00
>  %428 = fadd float %427, -1.000000e+00
>  %429 = fmul float %407, %317
>  %430 = fadd float %429, %395
>  %431 = fmul float %408, %317
>  %432 = fadd float %431, %396
>  %433 = fmul float %428, %317
>  %434 = fadd float %433, %397
>  %435 = fmul float %388, %421
>  %436 = fmul float %389, %422
>  %437 = fmul float %342, %422
>  %438 = fmul float %343, %421
>  %439 = fadd float %437, %438
>  %440 = fadd float %430, %435
>  %441 = fadd float %432, %436
>  %442 = fadd float %434, %439
>  %443 = fadd float %371, %371
>  %444 = fadd float %372, %372
>  %445 = fmul float %421, %443
>  %446 = fmul float %422, %444
>  %447 = fmul float %371, %422
>  %448 = fmul float %372, %421
>  %449 = fadd float %447, %448
>  %450 = fadd float %440, %445
>  %451 = fadd float %441, %446
>  %452 = fadd float %442, %449
>  %453 = bitcast float %140 to i32
>  %454 = bitcast float %141 to i32
>  %455 = insertelement <2 x i32> undef, i32 %453, i32 0
>  %456 = insertelement <2 x i32> %455, i32 %454, i32 1
>  %457 = bitcast <8 x i32> %91 to <32 x i8>
>  %458 = bitcast <4 x i32> %93 to <16 x i8>
>  %459 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %456, <32 x i8> %457, <16 x i8> %458, i32 2)
>  %460 = extractelement <4 x float> %459, i32 0
>  %461 = extractelement <4 x float> %459, i32 1
>  %462 = extractelement <4 x float> %459, i32 2
>  %463 = extractelement <4 x float> %459, i32 3
>  %464 = bitcast float %140 to i32
>  %465 = bitcast float %141 to i32
>  %466 = insertelement <2 x i32> undef, i32 %464, i32 0
>  %467 = insertelement <2 x i32> %466, i32 %465, i32 1
>  %468 = bitcast <8 x i32> %95 to <32 x i8>
>  %469 = bitcast <4 x i32> %97 to <16 x i8>
>  %470 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %467, <32 x i8> %468, <16 x i8> %469, i32 2)
>  %471 = extractelement <4 x float> %470, i32 0
>  %472 = fadd float %461, %461
>  %473 = fadd float %460, %460
>  %474 = fmul float %472, -1.000000e+00
>  %475 = fadd float %474, 1.000000e+00
>  %476 = fmul float %473, 1.000000e+00
>  %477 = fadd float %476, -1.000000e+00
>  %478 = fmul float %314, %475
>  %479 = fmul float %314, %477
>  %480 = fmul float %477, %314
>  %481 = fadd float %480, %424
>  %482 = fmul float %475, %314
>  %483 = fadd float %482, %426
>  %484 = fmul float %471, -2.000000e+00
>  %485 = fadd float %484, 1.000000e+00
>  %486 = fmul float %462, %318
>  %487 = fadd float %486, %450
>  %488 = fmul float %463, %318
>  %489 = fadd float %488, %451
>  %490 = fmul float %485, %318
>  %491 = fadd float %490, %452
>  %492 = fmul float %388, %479
>  %493 = fmul float %389, %478
>  %494 = fmul float %342, %478
>  %495 = fmul float %343, %479
>  %496 = fadd float %494, %495
>  %497 = fadd float %492, %487
>  %498 = fadd float %493, %489
>  %499 = fadd float %496, %491
>  %500 = fmul float %443, %479
>  %501 = fmul float %444, %478
>  %502 = fmul float %371, %478
>  %503 = fmul float %372, %479
>  %504 = fadd float %502, %503
>  %505 = fadd float %497, %500
>  %506 = fadd float %498, %501
>  %507 = fadd float %499, %504
>  %508 = fmul float %421, %479
>  %509 = fmul float %422, %478
>  %510 = fadd float %508, %508
>  %511 = fadd float %509, %509
>  %512 = fmul float %421, %478
>  %513 = fmul float %422, %479
>  %514 = fadd float %512, %513
>  %515 = fadd float %505, %510
>  %516 = fadd float %506, %511
>  %517 = fadd float %507, %514
>  %518 = fsub float -0.000000e+00, %481
>  %519 = fmul float %481, %518
>  %520 = fadd float %519, %515
>  %521 = fsub float -0.000000e+00, %483
>  %522 = fmul float %483, %521
>  %523 = fadd float %522, %516
>  %524 = fsub float -0.000000e+00, %483
>  %525 = fmul float %481, %524
>  %526 = fadd float %525, %517
>  %527 = fmul float %25, %481
>  %528 = fmul float %26, %483
>  %529 = fadd float %528, %527
>  %530 = fmul float %27, 1.000000e+00
>  %531 = fadd float %529, %530
>  %532 = call float @llvm.AMDIL.clamp.(float %531, float 0.000000e+00, float 1.000000e+00)
>  %533 = fsub float -0.000000e+00, %142
>  %534 = fadd float %50, %533
>  %535 = fsub float -0.000000e+00, %143
>  %536 = fadd float %51, %535
>  %537 = fsub float -0.000000e+00, %144
>  %538 = fadd float %52, %537
>  %539 = fmul float %534, %534
>  %540 = fmul float %536, %536
>  %541 = fadd float %540, %539
>  %542 = fmul float %538, %538
>  %543 = fadd float %541, %542
>  %544 = call float @llvm.maxnum.f32(float %543, float 0x3E7AD7F2A0000000)
>  %545 = call float @llvm.AMDGPU.rsq.clamped.f32(float %544)
>  %546 = fmul float %545, %534
>  %547 = fmul float %545, %536
>  %548 = fmul float %545, %538
>  %549 = fsub float -0.000000e+00, %39
>  %550 = fmul float %534, %545
>  %551 = fadd float %550, %549
>  %552 = fsub float -0.000000e+00, %40
>  %553 = fmul float %536, %545
>  %554 = fadd float %553, %552
>  %555 = fsub float -0.000000e+00, %41
>  %556 = fmul float %538, %545
>  %557 = fadd float %556, %555
>  %558 = fmul float %42, %551
>  %559 = fadd float %558, %39
>  %560 = fmul float %42, %554
>  %561 = fadd float %560, %40
>  %562 = fmul float %42, %557
>  %563 = fadd float %562, %41
>  %564 = fmul float %559, %559
>  %565 = fmul float %561, %561
>  %566 = fadd float %565, %564
>  %567 = fmul float %563, %563
>  %568 = fadd float %566, %567
>  %569 = call float @llvm.maxnum.f32(float %568, float 0x3E7AD7F2A0000000)
>  %570 = call float @llvm.AMDGPU.rsq.clamped.f32(float %569)
>  %571 = fmul float %559, %570
>  %572 = fadd float %571, %25
>  %573 = fmul float %561, %570
>  %574 = fadd float %573, %26
>  %575 = fmul float %563, %570
>  %576 = fadd float %575, %27
>  %577 = fdiv float 1.000000e+00, %576
>  %578 = fsub float -0.000000e+00, %481
>  %579 = fmul float %572, %577
>  %580 = fadd float %579, %578
>  %581 = fsub float -0.000000e+00, %483
>  %582 = fmul float %574, %577
>  %583 = fadd float %582, %581
>  %584 = fdiv float 1.000000e+00, %34
>  %585 = fadd float %520, %584
>  %586 = fadd float %523, %584
>  %587 = fmul float %526, %526
>  %588 = fsub float -0.000000e+00, %587
>  %589 = fmul float %585, %586
>  %590 = fadd float %589, %588
>  %591 = fmul float %580, %580
>  %592 = fadd float %526, %526
>  %593 = fmul float %580, %592
>  %594 = fsub float -0.000000e+00, %593
>  %595 = fmul float %585, %583
>  %596 = fadd float %595, %594
>  %597 = fmul float %583, %596
>  %598 = fmul float %591, %586
>  %599 = fadd float %598, %597
>  %600 = fmul float %599, 5.000000e-01
>  %601 = fdiv float 1.000000e+00, %590
>  %602 = fmul float %601, %600
>  %603 = fsub float -0.000000e+00, %590
>  %604 = fcmp oge float %603, 0.000000e+00
>  %605 = sext i1 %604 to i32
>  %606 = bitcast i32 %605 to float
>  %607 = bitcast float %606 to i32
>  %608 = icmp ne i32 %607, 0
>  %temp96.0 = select i1 %608, float 1.000000e+00, float 0.000000e+00
>  %609 = fmul float %600, %601
>  %610 = fadd float %609, -1.600000e+01
>  %611 = fcmp oge float %610, 0.000000e+00
>  %612 = sext i1 %611 to i32
>  %613 = bitcast i32 %612 to float
>  %614 = bitcast float %613 to i32
>  %615 = icmp ne i32 %614, 0
>  %.164 = select i1 %615, float 1.000000e+00, float 0.000000e+00
>  %616 = fmul float %602, 0xBFF7154CA0000000
>  %617 = call float @llvm.AMDIL.exp.(float %616)
>  %618 = call float @llvm.maxnum.f32(float %590, float 0x3E7AD7F2A0000000)
>  %619 = call float @llvm.AMDGPU.rsq.clamped.f32(float %618)
>  %620 = fmul float %617, %619
>  %621 = fmul float %584, 1.600000e+01
>  %622 = fadd float %621, %520
>  %623 = fmul float %584, 1.600000e+01
>  %624 = fadd float %623, %523
>  %625 = fsub float -0.000000e+00, %587
>  %626 = fmul float %622, %624
>  %627 = fadd float %626, %625
>  %628 = fsub float -0.000000e+00, %593
>  %629 = fmul float %622, %583
>  %630 = fadd float %629, %628
>  %631 = fmul float %583, %630
>  %632 = fmul float %591, %624
>  %633 = fadd float %632, %631
>  %634 = fmul float %633, 5.000000e-01
>  %635 = fdiv float 1.000000e+00, %627
>  %636 = fmul float %635, %634
>  %637 = fsub float -0.000000e+00, %627
>  %638 = fcmp oge float %637, 0.000000e+00
>  %639 = sext i1 %638 to i32
>  %640 = bitcast i32 %639 to float
>  %641 = bitcast float %640 to i32
>  %642 = icmp ne i32 %641, 0
>  %temp84.0 = select i1 %642, float 1.000000e+00, float 0.000000e+00
>  %643 = fmul float %634, %635
>  %644 = fadd float %643, -1.600000e+01
>  %645 = fcmp oge float %644, 0.000000e+00
>  %646 = sext i1 %645 to i32
>  %647 = bitcast i32 %646 to float
>  %648 = bitcast float %647 to i32
>  %649 = icmp ne i32 %648, 0
>  %.165 = select i1 %649, float 1.000000e+00, float 0.000000e+00
>  %650 = fadd float %.165, %temp84.0
>  %651 = fadd float %.164, %temp96.0
>  %652 = fmul float %636, 0xBFF7154CA0000000
>  %653 = call float @llvm.AMDIL.exp.(float %652)
>  %654 = call float @llvm.maxnum.f32(float %627, float 0x3E7AD7F2A0000000)
>  %655 = call float @llvm.AMDGPU.rsq.clamped.f32(float %654)
>  %656 = fmul float %655, %653
>  %657 = fsub float -0.000000e+00, %43
>  %658 = fmul float %534, %545
>  %659 = fadd float %658, %657
>  %660 = fsub float -0.000000e+00, %44
>  %661 = fmul float %536, %545
>  %662 = fadd float %661, %660
>  %663 = fsub float -0.000000e+00, %45
>  %664 = fmul float %538, %545
>  %665 = fadd float %664, %663
>  %666 = fmul float %46, %659
>  %667 = fadd float %666, %43
>  %668 = fmul float %46, %662
>  %669 = fadd float %668, %44
>  %670 = fmul float %46, %665
>  %671 = fadd float %670, %45
>  %672 = fmul float %667, %667
>  %673 = fmul float %669, %669
>  %674 = fadd float %673, %672
>  %675 = fmul float %671, %671
>  %676 = fadd float %674, %675
>  %677 = call float @llvm.maxnum.f32(float %676, float 0x3E7AD7F2A0000000)
>  %678 = call float @llvm.AMDGPU.rsq.clamped.f32(float %677)
>  %679 = fmul float %667, %678
>  %680 = fadd float %679, %28
>  %681 = fmul float %669, %678
>  %682 = fadd float %681, %29
>  %683 = fmul float %671, %678
>  %684 = fadd float %683, %30
>  %685 = fdiv float 1.000000e+00, %684
>  %686 = fsub float -0.000000e+00, %481
>  %687 = fmul float %680, %685
>  %688 = fadd float %687, %686
>  %689 = fsub float -0.000000e+00, %483
>  %690 = fmul float %682, %685
>  %691 = fadd float %690, %689
>  %692 = fdiv float 1.000000e+00, %38
>  %693 = fadd float %692, %520
>  %694 = fadd float %692, %523
>  %695 = fsub float -0.000000e+00, %587
>  %696 = fmul float %693, %694
>  %697 = fadd float %696, %695
>  %698 = fmul float %688, %688
>  %699 = fmul float %592, %688
>  %700 = fsub float -0.000000e+00, %699
>  %701 = fmul float %693, %691
>  %702 = fadd float %701, %700
>  %703 = fmul float %691, %702
>  %704 = fmul float %698, %694
>  %705 = fadd float %704, %703
>  %706 = fmul float %705, 5.000000e-01
>  %707 = fdiv float 1.000000e+00, %697
>  %708 = fmul float %706, %707
>  %709 = fsub float -0.000000e+00, %697
>  %710 = fcmp oge float %709, 0.000000e+00
>  %711 = sext i1 %710 to i32
>  %712 = bitcast i32 %711 to float
>  %713 = bitcast float %712 to i32
>  %714 = icmp ne i32 %713, 0
>  %temp56.1 = select i1 %714, float 1.000000e+00, float 0.000000e+00
>  %715 = fmul float %706, %707
>  %716 = fadd float %715, -1.600000e+01
>  %717 = fcmp oge float %716, 0.000000e+00
>  %718 = sext i1 %717 to i32
>  %719 = bitcast i32 %718 to float
>  %720 = bitcast float %719 to i32
>  %721 = icmp ne i32 %720, 0
>  %.166 = select i1 %721, float 1.000000e+00, float 0.000000e+00
>  %722 = fadd float %.166, %temp56.1
>  %723 = fmul float %708, 0xBFF7154CA0000000
>  %724 = fmul float %656, 0x3F747AE140000000
>  %725 = call float @llvm.maxnum.f32(float %697, float 0x3E7AD7F2A0000000)
>  %726 = call float @llvm.AMDGPU.rsq.clamped.f32(float %725)
>  %727 = call float @llvm.AMDIL.exp.(float %723)
>  %728 = fmul float %726, %727
>  %729 = fmul float %56, %143
>  %730 = fmul float %57, %143
>  %731 = fmul float %58, %143
>  %732 = fmul float %142, %53
>  %733 = fadd float %732, %729
>  %734 = fmul float %142, %54
>  %735 = fadd float %734, %730
>  %736 = fmul float %142, %55
>  %737 = fadd float %736, %731
>  %738 = fmul float %144, %59
>  %739 = fadd float %738, %733
>  %740 = fmul float %144, %60
>  %741 = fadd float %740, %735
>  %742 = fmul float %144, %61
>  %743 = fadd float %742, %737
>  %744 = fadd float %739, %62
>  %745 = fadd float %741, %63
>  %746 = fadd float %743, %64
>  %747 = fdiv float 1.000000e+00, %746
>  %748 = fmul float %747, %744
>  %749 = fmul float %747, %745
>  %750 = fmul float %748, 5.000000e-01
>  %751 = fadd float %750, -5.000000e-01
>  %752 = fmul float %749, -5.000000e-01
>  %753 = fadd float %752, -5.000000e-01
>  %754 = fmul float %520, 6.000000e+00
>  %755 = fadd float %754, %751
>  %756 = fmul float %523, 6.000000e+00
>  %757 = fadd float %756, %753
>  %758 = fmul float %697, 2.000000e+00
>  %759 = fadd float %758, %755
>  %760 = fmul float %697, 2.000000e+00
>  %761 = fadd float %760, %757
>  %762 = fmul float %697, 2.000000e+00
>  %763 = fadd float %762, %755
>  %764 = fmul float %697, -2.000000e+00
>  %765 = fadd float %764, %757
>  %766 = bitcast float %763 to i32
>  %767 = bitcast float %765 to i32
>  %768 = insertelement <2 x i32> undef, i32 %766, i32 0
>  %769 = insertelement <2 x i32> %768, i32 %767, i32 1
>  %770 = bitcast <8 x i32> %99 to <32 x i8>
>  %771 = bitcast <4 x i32> %101 to <16 x i8>
>  %772 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %769, <32 x i8> %770, <16 x i8> %771, i32 2)
>  %773 = extractelement <4 x float> %772, i32 0
>  %774 = extractelement <4 x float> %772, i32 1
>  %775 = extractelement <4 x float> %772, i32 2
>  %776 = extractelement <4 x float> %772, i32 3
>  %777 = fmul float %773, 2.500000e-01
>  %778 = fmul float %774, 2.500000e-01
>  %779 = fmul float %775, 2.500000e-01
>  %780 = fmul float %776, 2.500000e-01
>  %781 = bitcast float %759 to i32
>  %782 = bitcast float %761 to i32
>  %783 = insertelement <2 x i32> undef, i32 %781, i32 0
>  %784 = insertelement <2 x i32> %783, i32 %782, i32 1
>  %785 = bitcast <8 x i32> %99 to <32 x i8>
>  %786 = bitcast <4 x i32> %101 to <16 x i8>
>  %787 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %784, <32 x i8> %785, <16 x i8> %786, i32 2)
>  %788 = extractelement <4 x float> %787, i32 0
>  %789 = extractelement <4 x float> %787, i32 1
>  %790 = extractelement <4 x float> %787, i32 2
>  %791 = extractelement <4 x float> %787, i32 3
>  %792 = fmul float %788, 2.500000e-01
>  %793 = fadd float %792, %777
>  %794 = fmul float %789, 2.500000e-01
>  %795 = fadd float %794, %778
>  %796 = fmul float %790, 2.500000e-01
>  %797 = fadd float %796, %779
>  %798 = fmul float %791, 2.500000e-01
>  %799 = fadd float %798, %780
>  %800 = fmul float %697, -2.000000e+00
>  %801 = fadd float %800, %755
>  %802 = fmul float %697, 6.000000e+00
>  %803 = fadd float %802, %757
>  %804 = bitcast float %801 to i32
>  %805 = bitcast float %803 to i32
>  %806 = insertelement <2 x i32> undef, i32 %804, i32 0
>  %807 = insertelement <2 x i32> %806, i32 %805, i32 1
>  %808 = bitcast <8 x i32> %99 to <32 x i8>
>  %809 = bitcast <4 x i32> %101 to <16 x i8>
>  %810 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %807, <32 x i8> %808, <16 x i8> %809, i32 2)
>  %811 = extractelement <4 x float> %810, i32 0
>  %812 = extractelement <4 x float> %810, i32 1
>  %813 = extractelement <4 x float> %810, i32 2
>  %814 = extractelement <4 x float> %810, i32 3
>  %815 = fmul float %811, 2.500000e-01
>  %816 = fadd float %815, %793
>  %817 = fmul float %812, 2.500000e-01
>  %818 = fadd float %817, %795
>  %819 = fmul float %813, 2.500000e-01
>  %820 = fadd float %819, %797
>  %821 = fmul float %814, 2.500000e-01
>  %822 = fadd float %821, %799
>  %823 = fmul float %697, -2.000000e+00
>  %824 = fadd float %823, %755
>  %825 = fmul float %697, -2.000000e+00
>  %826 = fadd float %825, %757
>  %827 = bitcast float %824 to i32
>  %828 = bitcast float %826 to i32
>  %829 = insertelement <2 x i32> undef, i32 %827, i32 0
>  %830 = insertelement <2 x i32> %829, i32 %828, i32 1
>  %831 = bitcast <8 x i32> %99 to <32 x i8>
>  %832 = bitcast <4 x i32> %101 to <16 x i8>
>  %833 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %830, <32 x i8> %831, <16 x i8> %832, i32 2)
>  %834 = extractelement <4 x float> %833, i32 0
>  %835 = extractelement <4 x float> %833, i32 1
>  %836 = extractelement <4 x float> %833, i32 2
>  %837 = extractelement <4 x float> %833, i32 3
>  %838 = fmul float %834, 2.500000e-01
>  %839 = fadd float %838, %816
>  %840 = fmul float %835, 2.500000e-01
>  %841 = fadd float %840, %818
>  %842 = fmul float %836, 2.500000e-01
>  %843 = fadd float %842, %820
>  %844 = fmul float %837, 2.500000e-01
>  %845 = fadd float %844, %822
>  %846 = call float @fabs(float %839)
>  %847 = call float @llvm.log2.f32(float %846)
>  %848 = call float @fabs(float %841)
>  %849 = call float @llvm.log2.f32(float %848)
>  %850 = call float @fabs(float %843)
>  %851 = call float @llvm.log2.f32(float %850)
>  %852 = fmul float %847, 0x40019999A0000000
>  %853 = fmul float %849, 0x40019999A0000000
>  %854 = fmul float %851, 0x40019999A0000000
>  %855 = call float @llvm.AMDIL.exp.(float %852)
>  %856 = call float @llvm.AMDIL.exp.(float %853)
>  %857 = call float @llvm.AMDIL.exp.(float %854)
>  %858 = fmul float %481, %546
>  %859 = fmul float %483, %547
>  %860 = fadd float %859, %858
>  %861 = fmul float 1.000000e+00, %548
>  %862 = fadd float %860, %861
>  %863 = fsub float -0.000000e+00, %862
>  %864 = fadd float %863, 1.000000e+00
>  %865 = fmul float %864, %864
>  %866 = fmul float %865, %865
>  %867 = fmul float %864, %866
>  %868 = fmul float %620, %31
>  %869 = fmul float %620, %32
>  %870 = fmul float %620, %33
>  %871 = fsub float -0.000000e+00, %651
>  %872 = fsub float -0.000000e+00, %651
>  %873 = fsub float -0.000000e+00, %651
>  %874 = fcmp oge float %871, 0.000000e+00
>  %875 = sext i1 %874 to i32
>  %876 = bitcast i32 %875 to float
>  %877 = bitcast float %876 to i32
>  %878 = icmp ne i32 %877, 0
>  %temp72.0 = select i1 %878, float %868, float 0.000000e+00
>  %879 = fcmp oge float %872, 0.000000e+00
>  %880 = sext i1 %879 to i32
>  %881 = bitcast i32 %880 to float
>  %882 = bitcast float %881 to i32
>  %883 = icmp ne i32 %882, 0
>  %.167 = select i1 %883, float %869, float 0.000000e+00
>  %884 = fcmp oge float %873, 0.000000e+00
>  %885 = sext i1 %884 to i32
>  %886 = bitcast i32 %885 to float
>  %887 = bitcast float %886 to i32
>  %888 = icmp ne i32 %887, 0
>  %temp60.0 = select i1 %888, float %870, float 0.000000e+00
>  %889 = fmul float %728, %35
>  %890 = fmul float %728, %36
>  %891 = fmul float %728, %37
>  %892 = fsub float -0.000000e+00, %722
>  %893 = fsub float -0.000000e+00, %722
>  %894 = fsub float -0.000000e+00, %722
>  %895 = fcmp oge float %892, 0.000000e+00
>  %896 = sext i1 %895 to i32
>  %897 = bitcast i32 %896 to float
>  %898 = bitcast float %897 to i32
>  %899 = icmp ne i32 %898, 0
>  %temp68.0 = select i1 %899, float %889, float 0.000000e+00
>  %900 = fcmp oge float %893, 0.000000e+00
>  %901 = sext i1 %900 to i32
>  %902 = bitcast i32 %901 to float
>  %903 = bitcast float %902 to i32
>  %904 = icmp ne i32 %903, 0
>  %.168 = select i1 %904, float %890, float 0.000000e+00
>  %905 = fcmp oge float %894, 0.000000e+00
>  %906 = sext i1 %905 to i32
>  %907 = bitcast i32 %906 to float
>  %908 = bitcast float %907 to i32
>  %909 = icmp ne i32 %908, 0
>  %temp28.1 = select i1 %909, float %891, float 0.000000e+00
>  %910 = fadd float %temp68.0, %temp72.0
>  %911 = fadd float %.168, %.167
>  %912 = fadd float %temp28.1, %temp60.0
>  %913 = fmul float %910, %867
>  %914 = fadd float %913, %257
>  %915 = fmul float %911, %867
>  %916 = fadd float %915, %258
>  %917 = fmul float %912, %867
>  %918 = fadd float %917, %259
>  %919 = fmul float %257, %31
>  %920 = fmul float %258, %32
>  %921 = fmul float %259, %33
>  %922 = fmul float %724, %919
>  %923 = fmul float %724, %920
>  %924 = fmul float %724, %921
>  %925 = fsub float -0.000000e+00, %650
>  %926 = fsub float -0.000000e+00, %650
>  %927 = fsub float -0.000000e+00, %650
>  %928 = fsub float -0.000000e+00, %650
>  %929 = fcmp oge float %925, 0.000000e+00
>  %930 = sext i1 %929 to i32
>  %931 = bitcast i32 %930 to float
>  %932 = bitcast float %931 to i32
>  %933 = icmp ne i32 %932, 0
>  %.169 = select i1 %933, float %922, float 0.000000e+00
>  %934 = fcmp oge float %926, 0.000000e+00
>  %935 = sext i1 %934 to i32
>  %936 = bitcast i32 %935 to float
>  %937 = bitcast float %936 to i32
>  %938 = icmp ne i32 %937, 0
>  %939 = fcmp oge float %927, 0.000000e+00
>  %940 = sext i1 %939 to i32
>  %941 = bitcast i32 %940 to float
>  %942 = bitcast float %941 to i32
>  %943 = icmp ne i32 %942, 0
>  %temp56.2 = select i1 %943, float %923, float 0.000000e+00
>  %944 = fcmp oge float %928, 0.000000e+00
>  %945 = sext i1 %944 to i32
>  %946 = bitcast i32 %945 to float
>  %947 = bitcast float %946 to i32
>  %948 = icmp ne i32 %947, 0
>  %.170 = select i1 %948, float %924, float 0.000000e+00
>  %949 = fmul float %914, %532
>  %950 = fadd float %949, %.169
>  %951 = fmul float %916, %532
>  %952 = fadd float %951, %temp56.2
>  %953 = fmul float %918, %532
>  %954 = fadd float %953, %.170
>  %955 = fmul float %855, %89
>  %956 = fadd float %955, %950
>  %957 = fmul float %856, %89
>  %958 = fadd float %957, %952
>  %959 = fmul float %857, %89
>  %960 = fadd float %959, %954
>  %961 = fmul float %708, %708
>  %962 = fmul float %708, %961
>  %963 = fmul float %237, %962
>  %964 = fmul float %963, %257
>  %965 = fadd float %964, %956
>  %966 = fmul float %963, %258
>  %967 = fadd float %966, %958
>  %968 = fmul float %963, %259
>  %969 = fadd float %968, %960
>  %970 = fmul float %68, %143
>  %971 = fmul float %69, %143
>  %972 = fmul float %70, %143
>  %973 = fmul float %142, %65
>  %974 = fadd float %973, %970
>  %975 = fmul float %142, %66
>  %976 = fadd float %975, %971
>  %977 = fmul float %142, %67
>  %978 = fadd float %977, %972
>  %979 = fmul float %144, %71
>  %980 = fadd float %979, %974
>  %981 = fmul float %144, %72
>  %982 = fadd float %981, %976
>  %983 = fmul float %144, %73
>  %984 = fadd float %983, %978
>  %985 = fadd float %980, %74
>  %986 = fadd float %982, %75
>  %987 = fadd float %984, %76
>  %988 = fdiv float 1.000000e+00, %987
>  %989 = fmul float %988, %985
>  %990 = fmul float %988, %986
>  %991 = fmul float %989, 5.000000e-01
>  %992 = fadd float %991, -5.000000e-01
>  %993 = fmul float %990, -5.000000e-01
>  %994 = fadd float %993, -5.000000e-01
>  %995 = fmul float %992, 1.000000e+00
>  %996 = fadd float %995, 0.000000e+00
>  %997 = fmul float %994, -1.000000e+00
>  %998 = fadd float %997, 1.000000e+00
>  %999 = bitcast float %996 to i32
>  %1000 = bitcast float %998 to i32
>  %1001 = insertelement <2 x i32> undef, i32 %999, i32 0
>  %1002 = insertelement <2 x i32> %1001, i32 %1000, i32 1
>  %1003 = bitcast <8 x i32> %131 to <32 x i8>
>  %1004 = bitcast <4 x i32> %133 to <16 x i8>
>  %1005 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %1002, <32 x i8> %1003, <16 x i8> %1004, i32 2)
>  %1006 = extractelement <4 x float> %1005, i32 0
>  %1007 = call float @llvm.AMDIL.clamp.(float %1006, float 0.000000e+00, float 1.000000e+00)
>  %1008 = fmul float %1007, %85
>  %1009 = fmul float %1008, %965
>  %1010 = fmul float %1008, %967
>  %1011 = fmul float %1008, %969
>  %1012 = fmul float %1009, 0xBFE570A3E0000000
>  %1013 = fadd float %1012, %965
>  %1014 = fmul float %1010, 0xBFE570A3E0000000
>  %1015 = fadd float %1014, %967
>  %1016 = fmul float %1011, 0xBFE570A3E0000000
>  %1017 = fadd float %1016, %969
>  %1018 = fmul float %845, %260
>  %1019 = fadd float %1018, %260
>  %1020 = fmul float %845, %260
>  %1021 = fadd float %1020, %260
>  %1022 = fmul float %845, %260
>  %1023 = fadd float %1022, %260
>  %1024 = fmul float %845, %260
>  %1025 = fadd float %1024, %260
>  %1026 = call float @llvm.AMDIL.clamp.(float %1019, float 0.000000e+00, float 1.000000e+00)
>  %1027 = call float @llvm.AMDIL.clamp.(float %1021, float 0.000000e+00, float 1.000000e+00)
>  %1028 = call float @llvm.AMDIL.clamp.(float %1023, float 0.000000e+00, float 1.000000e+00)
>  %1029 = call float @llvm.AMDIL.clamp.(float %1025, float 0.000000e+00, float 1.000000e+00)
>  %1030 = call float @fabs(float %1013)
>  %1031 = call float @llvm.log2.f32(float %1030)
>  %1032 = call float @fabs(float %1015)
>  %1033 = call float @llvm.log2.f32(float %1032)
>  %1034 = call float @fabs(float %1017)
>  %1035 = call float @llvm.log2.f32(float %1034)
>  %1036 = fmul float %1031, 0x3FDD1743E0000000
>  %1037 = fmul float %1033, 0x3FDD1743E0000000
>  %1038 = fmul float %1035, 0x3FDD1743E0000000
>  %1039 = call float @llvm.AMDIL.exp.(float %1036)
>  %1040 = call float @llvm.AMDIL.exp.(float %1037)
>  %1041 = call float @llvm.AMDIL.exp.(float %1038)
>  %1042 = fmul float %147, 1.000000e+00
>  %1043 = fadd float %1042, 0.000000e+00
>  %1044 = fmul float %148, -1.000000e+00
>  %1045 = fadd float %1044, 1.000000e+00
>  %1046 = bitcast float %1043 to i32
>  %1047 = bitcast float %1045 to i32
>  %1048 = insertelement <2 x i32> undef, i32 %1046, i32 0
>  %1049 = insertelement <2 x i32> %1048, i32 %1047, i32 1
>  %1050 = bitcast <8 x i32> %119 to <32 x i8>
>  %1051 = bitcast <4 x i32> %121 to <16 x i8>
>  %1052 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %1049, <32 x i8> %1050, <16 x i8> %1051, i32 2)
>  %1053 = extractelement <4 x float> %1052, i32 0
>  %1054 = extractelement <4 x float> %1052, i32 1
>  %1055 = extractelement <4 x float> %1052, i32 2
>  %1056 = extractelement <4 x float> %1052, i32 3
>  %1057 = fsub float -0.000000e+00, %1056
>  %1058 = fadd float %1057, 1.000000e+00
>  %1059 = fmul float %1039, %1058
>  %1060 = fadd float %1059, %1053
>  %1061 = fmul float %1040, %1058
>  %1062 = fadd float %1061, %1054
>  %1063 = fmul float %1041, %1058
>  %1064 = fadd float %1063, %1055
>  %1065 = fadd float %1060, 0xBFD99999A0000000
>  %1066 = fadd float %1062, 0xBFD99999A0000000
>  %1067 = fadd float %1064, 0xBFD99999A0000000
>  %1068 = fmul float %162, %1065
>  %1069 = fadd float %1068, 0x3FD99999A0000000
>  %1070 = fmul float %162, %1066
>  %1071 = fadd float %1070, 0x3FD99999A0000000
>  %1072 = fmul float %162, %1067
>  %1073 = fadd float %1072, 0x3FD99999A0000000
>  %1074 = bitcast float %1069 to i32
>  %1075 = bitcast float %1071 to i32
>  %1076 = bitcast float %1073 to i32
>  %1077 = insertelement <4 x i32> undef, i32 %1074, i32 0
>  %1078 = insertelement <4 x i32> %1077, i32 %1075, i32 1
>  %1079 = insertelement <4 x i32> %1078, i32 %1076, i32 2
>  %1080 = insertelement <4 x i32> %1079, i32 undef, i32 3
>  %1081 = bitcast <8 x i32> %127 to <32 x i8>
>  %1082 = bitcast <4 x i32> %129 to <16 x i8>
>  %1083 = call <4 x float> @llvm.SI.sample.v4i32(<4 x i32> %1080, <32 x i8> %1081, <16 x i8> %1082, i32 3)
>  %1084 = extractelement <4 x float> %1083, i32 0
>  %1085 = extractelement <4 x float> %1083, i32 1
>  %1086 = extractelement <4 x float> %1083, i32 2
>  %1087 = fsub float -0.000000e+00, %162
>  %1088 = fmul float %163, %1087
>  %1089 = fadd float %1088, %162
>  %1090 = call float @llvm.AMDGPU.lrp(float %1089, float %1084, float %1069)
>  %1091 = call float @llvm.AMDGPU.lrp(float %1089, float %1085, float %1071)
>  %1092 = call float @llvm.AMDGPU.lrp(float %1089, float %1086, float %1073)
>  %1093 = fdiv float 1.000000e+00, %146
>  %1094 = fmul float %1093, %145
>  %1095 = fmul float %1094, 5.000000e-01
>  %1096 = fadd float %1095, 5.000000e-01
>  %1097 = bitcast float %1096 to i32
>  %1098 = bitcast float %49 to i32
>  %1099 = insertelement <2 x i32> undef, i32 %1097, i32 0
>  %1100 = insertelement <2 x i32> %1099, i32 %1098, i32 1
>  %1101 = bitcast <8 x i32> %115 to <32 x i8>
>  %1102 = bitcast <4 x i32> %117 to <16 x i8>
>  %1103 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %1100, <32 x i8> %1101, <16 x i8> %1102, i32 2)
>  %1104 = extractelement <4 x float> %1103, i32 0
>  %1105 = fsub float -0.000000e+00, %50
>  %1106 = fadd float %1105, %142
>  %1107 = fsub float -0.000000e+00, %51
>  %1108 = fadd float %1107, %143
>  %1109 = fsub float -0.000000e+00, %52
>  %1110 = fadd float %1109, %144
>  %1111 = fmul float %1106, %1106
>  %1112 = fmul float %1108, %1108
>  %1113 = fadd float %1112, %1111
>  %1114 = fmul float %1110, %1110
>  %1115 = fadd float %1113, %1114
>  %1116 = fmul float %1115, %48
>  %1117 = fmul float %1110, %47
>  %1118 = fmul float %1117, 0x3FF7154CA0000000
>  %1119 = call float @llvm.AMDIL.exp.(float %1118)
>  %1120 = fsub float -0.000000e+00, %1119
>  %1121 = fadd float %1120, 1.000000e+00
>  %1122 = fmul float %1121, %1116
>  %1123 = fdiv float 1.000000e+00, %1110
>  %1124 = fmul float %1123, %1122
>  %1125 = fmul float %1124, 0x3FF7154CA0000000
>  %1126 = call float @llvm.AMDIL.exp.(float %1125)
>  %1127 = call float @llvm.AMDIL.clamp.(float %1126, float 0.000000e+00, float 1.000000e+00)
>  %1128 = fsub float -0.000000e+00, %1127
>  %1129 = fadd float %1128, 1.000000e+00
>  %1130 = fmul float %1129, %1104
>  %1131 = fsub float -0.000000e+00, %1090
>  %1132 = fadd float %1131, %86
>  %1133 = fsub float -0.000000e+00, %1091
>  %1134 = fadd float %1133, %87
>  %1135 = fsub float -0.000000e+00, %1092
>  %1136 = fadd float %1135, %88
>  %1137 = fmul float %1130, %1132
>  %1138 = fadd float %1137, %1090
>  %1139 = fmul float %1130, %1134
>  %1140 = fadd float %1139, %1091
>  %1141 = fmul float %1130, %1136
>  %1142 = fadd float %1141, %1092
>  %1143 = call i32 @llvm.SI.packf16(float %1138, float %1140)
>  %1144 = bitcast i32 %1143 to float
>  %1145 = call i32 @llvm.SI.packf16(float %1142, float %1029)
>  %1146 = bitcast i32 %1145 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %1144, float %1146, float %1144, float %1146)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: readnone
>declare float @fabs(float) #2
>
>declare void @llvm.AMDGPU.kill(float)
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.clamp.(float, float, float) #2
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.samplel.v4i32(<4 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.pow.f32(float, float) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.maxnum.f32(float, float) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.AMDGPU.rsq.clamped.f32(float) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.exp.(float) #2
>
>; Function Attrs: nounwind readnone
>declare float @llvm.log2.f32(float) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v4i32(<4 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDGPU.lrp(float, float, float) #2
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                         ; BEFE0A7E
>	s_mov_b64 s[100:101], s[6:7]                                                 ; BEE40406
>	s_mov_b32 m0, s9                                                             ; BEFC0309
>	v_interp_p1_f32 v2, v0, 1, 6, [m0]                                           ; C8081900
>	v_interp_p2_f32 v2, [v2], v1, 1, 6, [m0]                                     ; C8091901
>	v_sub_f32_e32 v8, 1.0, v2                                                    ; 081004F2
>	v_interp_p1_f32 v2, v0, 0, 6, [m0]                                           ; C8081800
>	v_interp_p2_f32 v2, [v2], v1, 0, 6, [m0]                                     ; C8091801
>	v_add_f32_e32 v7, 0, v2                                                      ; 060E0480
>	s_load_dwordx4 s[88:91], s[4:5], 0x0                                         ; C0AC0500
>	s_load_dwordx4 s[60:63], s[4:5], 0x4                                         ; C09E0504
>	s_load_dwordx4 s[32:35], s[4:5], 0x8                                         ; C0900508
>	s_load_dwordx4 s[12:15], s[4:5], 0xc                                         ; C086050C
>	s_load_dwordx4 s[52:55], s[4:5], 0x10                                        ; C09A0510
>	s_load_dwordx4 s[56:59], s[4:5], 0x14                                        ; C09C0514
>	s_load_dwordx4 s[8:11], s[4:5], 0x18                                         ; C0840518
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_writelane_b32 v57, s8, 0                                                   ; 04730008
>	v_writelane_b32 v57, s9, 1                                                   ; 04730209
>	v_writelane_b32 v57, s10, 2                                                  ; 0473040A
>	v_writelane_b32 v57, s11, 3                                                  ; 0473060B
>	s_load_dwordx4 s[24:27], s[4:5], 0x1c                                        ; C08C051C
>	s_load_dwordx4 s[16:19], s[4:5], 0x20                                        ; C0880520
>	s_load_dwordx4 s[8:11], s[4:5], 0x24                                         ; C0840524
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_writelane_b32 v57, s8, 4                                                   ; 04730808
>	v_writelane_b32 v57, s9, 5                                                   ; 04730A09
>	v_writelane_b32 v57, s10, 6                                                  ; 04730C0A
>	v_writelane_b32 v57, s11, 7                                                  ; 04730E0B
>	s_load_dwordx4 s[28:31], s[4:5], 0x28                                        ; C08E0528
>	s_load_dwordx8 s[36:43], s[100:101], 0x40                                    ; C0D26540
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample v[2:3], 5, 0, 0, 0, 0, 0, 0, 0, v[7:8], s[36:43], s[16:19]      ; F0800500 00890207
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_cmp_ge_f32_e64 s[0:1], -|v2|, 0                                            ; D00C0100 20010102
>	v_mov_b32_e32 v10, 0                                                         ; 7E140280
>	v_cndmask_b32_e64 v4, 0, -1, s[0:1]                                          ; D2000004 00018280
>	v_cmp_ne_i32_e64 s[0:1], v4, 0                                               ; D10A0000 00010104
>	v_mov_b32_e32 v4, 0x80000000                                                 ; 7E0802FF 80000000
>	v_cndmask_b32_e64 v4, v4, -1.0, s[0:1]                                       ; D2000004 0001E704
>	v_cmp_lt_f32_e64 s[0:1], v4, 0                                               ; D0020000 00010104
>	v_cndmask_b32_e64 v4, 0, -1, s[0:1]                                          ; D2000004 00018280
>	v_and_b32_e32 v4, 1.0, v4                                                    ; 360808F2
>	v_xor_b32_e32 v4, 0x80000000, v4                                             ; 3A0808FF 80000000
>	v_cmpx_le_f32_e32 vcc, 0, v4                                                 ; 7C260880
>	v_cmpx_le_f32_e32 vcc, 0, v4                                                 ; 7C260880
>	v_cmpx_le_f32_e32 vcc, 0, v4                                                 ; 7C260880
>	v_cmpx_le_f32_e32 vcc, 0, v4                                                 ; 7C260880
>	v_interp_p1_f32 v4, v0, 0, 4, [m0]                                           ; C8101000
>	v_interp_p2_f32 v4, [v4], v1, 0, 4, [m0]                                     ; C8111001
>	v_interp_p1_f32 v5, v0, 1, 4, [m0]                                           ; C8141100
>	v_interp_p2_f32 v5, [v5], v1, 1, 4, [m0]                                     ; C8151101
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                                           ; C0800300
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	s_buffer_load_dword s4, s[0:3], 0x49                                         ; C2020149
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v6, s4, v5                                                     ; 100C0A04
>	s_buffer_load_dword s4, s[0:3], 0x45                                         ; C2020145
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v13, v4, s4, v6                                                    ; D282000D 04180904
>	v_interp_p1_f32 v6, v0, 2, 4, [m0]                                           ; C8181200
>	v_interp_p2_f32 v6, [v6], v1, 2, 4, [m0]                                     ; C8191201
>	s_buffer_load_dword s4, s[0:3], 0x4d                                         ; C202014D
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v13, v6, s4, v13                                                   ; D282000D 04340906
>	s_buffer_load_dword s4, s[0:3], 0x51                                         ; C2020151
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_add_f32_e32 v13, s4, v13                                                   ; 061A1A04
>	v_sub_f32_e32 v14, 1.0, v13                                                  ; 081C1AF2
>	s_buffer_load_dword s4, s[0:3], 0x48                                         ; C2020148
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v15, s4, v5                                                    ; 101E0A04
>	s_buffer_load_dword s4, s[0:3], 0x44                                         ; C2020144
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v15, v4, s4, v15                                                   ; D282000F 043C0904
>	s_buffer_load_dword s4, s[0:3], 0x4c                                         ; C202014C
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v15, v6, s4, v15                                                   ; D282000F 043C0906
>	s_buffer_load_dword s4, s[0:3], 0x50                                         ; C2020150
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_add_f32_e32 v15, s4, v15                                                   ; 061E1E04
>	v_add_f32_e32 v13, 0, v15                                                    ; 061A1E80
>	s_load_dwordx8 s[92:99], s[100:101], 0x0                                     ; C0EE6500
>	s_load_dwordx8 s[80:87], s[100:101], 0x8                                     ; C0E86508
>	s_load_dwordx8 s[44:51], s[100:101], 0x10                                    ; C0D66510
>	s_load_dwordx8 s[4:11], s[100:101], 0x18                                     ; C0C26518
>	s_load_dwordx8 s[64:71], s[100:101], 0x20                                    ; C0E06520
>	s_load_dwordx8 s[72:79], s[100:101], 0x28                                    ; C0E46528
>	s_load_dwordx8 s[16:23], s[100:101], 0x30                                    ; C0C86530
>	s_load_dwordx8 s[36:43], s[100:101], 0x38                                    ; C0D26538
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample v[13:14], 3, 0, 0, 0, 0, 0, 0, 0, v[13:14], s[4:11], s[12:15]   ; F0800300 00610D0D
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v15, 2.0, v14, -1.0                                                ; D282000F 03CE1CF4
>	v_mad_f32 v13, 2.0, v13, -1.0                                                ; D282000D 03CE1AF4
>	v_subrev_f32_e32 v14, v15, v13                                               ; 0A1C1B0F
>	v_add_f32_e32 v16, 2.0, v14                                                  ; 06201CF4
>	v_mul_f32_e32 v16, v16, v16                                                  ; 10202110
>	v_sub_f32_e64 v13, -v15, v13                                                 ; D208000D 20021B0F
>	v_add_f32_e32 v15, 2.0, v13                                                  ; 061E1AF4
>	v_mul_f32_e32 v15, v15, v15                                                  ; 101E1F0F
>	v_mul_f32_e32 v15, v15, v15                                                  ; 101E1F0F
>	v_mad_f32 v17, v16, v16, v15                                                 ; D2820011 043E2110
>	v_sub_f32_e32 v14, 2.0, v14                                                  ; 081C1CF4
>	v_mul_f32_e32 v14, v14, v14                                                  ; 101C1D0E
>	v_mad_f32 v17, v14, v14, v17                                                 ; D2820011 04461D0E
>	v_sub_f32_e32 v13, 2.0, v13                                                  ; 081A1AF4
>	v_mul_f32_e32 v13, v13, v13                                                  ; 101A1B0D
>	v_mad_f32 v17, v13, v13, v17                                                 ; D2820011 04461B0D
>	v_rcp_f32_e32 v17, v17                                                       ; 7E225511
>	v_mul_f32_e32 v16, v16, v16                                                  ; 10202110
>	v_mul_f32_e32 v16, v16, v17                                                  ; 10202310
>	s_buffer_load_dword s4, s[0:3], 0x0                                          ; C2020100
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_rcp_f32_e32 v18, s4                                                        ; 7E245404
>	v_mul_f32_e32 v16, v16, v18                                                  ; 10202510
>	v_interp_p1_f32 v20, v0, 1, 0, [m0]                                          ; C8500100
>	v_interp_p2_f32 v20, [v20], v1, 1, 0, [m0]                                   ; C8510101
>	v_interp_p1_f32 v19, v0, 0, 0, [m0]                                          ; C84C0000
>	v_interp_p2_f32 v19, [v19], v1, 0, 0, [m0]                                   ; C84D0001
>	image_sample v[21:24], 15, 0, 0, 0, 0, 0, 0, 0, v[19:20], s[92:99], s[88:91] ; F0800F00 02D71513
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v25, 2.0, v22, -1.0                                                ; D2820019 03CE2CF4
>	v_mul_f32_e32 v26, v25, v16                                                  ; 10342119
>	v_mad_f32 v27, v16, v25, v26                                                 ; D282001B 046A3310
>	v_mul_f32_e32 v15, v15, v17                                                  ; 101E230F
>	v_mul_f32_e32 v15, v15, v18                                                  ; 101E250F
>	v_mul_f32_e32 v28, v15, v15                                                  ; 10381F0F
>	v_interp_p1_f32 v30, v0, 1, 1, [m0]                                          ; C8780500
>	v_interp_p2_f32 v30, [v30], v1, 1, 1, [m0]                                   ; C8790501
>	v_interp_p1_f32 v29, v0, 0, 1, [m0]                                          ; C8740400
>	v_interp_p2_f32 v29, [v29], v1, 0, 1, [m0]                                   ; C8750401
>	image_sample v[31:34], 15, 0, 0, 0, 0, 0, 0, 0, v[29:30], s[92:99], s[88:91] ; F0800F00 02D71F1D
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mul_f32_e32 v35, v34, v28                                                  ; 10463922
>	v_mul_f32_e32 v36, v16, v16                                                  ; 10482110
>	v_mad_f32 v35, v24, v36, v35                                                 ; D2820023 048E4918
>	v_mad_f32 v37, 2.0, v32, -1.0                                                ; D2820025 03CE40F4
>	v_mul_f32_e32 v38, v37, v15                                                  ; 104C1F25
>	v_mad_f32 v35, v38, v27, v35                                                 ; D2820023 048E3726
>	v_mul_f32_e32 v14, v14, v14                                                  ; 101C1D0E
>	v_mul_f32_e32 v14, v14, v17                                                  ; 101C230E
>	v_mul_f32_e32 v14, v14, v18                                                  ; 101C250E
>	v_mul_f32_e32 v39, v14, v14                                                  ; 104E1D0E
>	v_interp_p1_f32 v41, v0, 1, 2, [m0]                                          ; C8A40900
>	v_interp_p2_f32 v41, [v41], v1, 1, 2, [m0]                                   ; C8A50901
>	v_interp_p1_f32 v40, v0, 0, 2, [m0]                                          ; C8A00800
>	v_interp_p2_f32 v40, [v40], v1, 0, 2, [m0]                                   ; C8A10801
>	image_sample v[42:45], 15, 0, 0, 0, 0, 0, 0, 0, v[40:41], s[92:99], s[88:91] ; F0800F00 02D72A28
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v35, v45, v39, v35                                                 ; D2820023 048E4F2D
>	v_mad_f32 v46, -2.0, v43, 1.0                                                ; D282002E 03CA56F5
>	v_mul_f32_e32 v47, v46, v14                                                  ; 105E1D2E
>	v_mad_f32 v35, v27, v47, v35                                                 ; D2820023 048E5F1B
>	v_mad_f32 v37, v15, v37, v38                                                 ; D2820025 049A4B0F
>	v_mad_f32 v35, v47, v37, v35                                                 ; D2820023 048E4B2F
>	v_mul_f32_e32 v13, v13, v13                                                  ; 101A1B0D
>	v_mul_f32_e32 v13, v13, v17                                                  ; 101A230D
>	v_mul_f32_e32 v13, v13, v18                                                  ; 101A250D
>	v_mul_f32_e32 v17, v13, v13                                                  ; 10221B0D
>	v_interp_p1_f32 v49, v0, 1, 3, [m0]                                          ; C8C40D00
>	v_interp_p2_f32 v49, [v49], v1, 1, 3, [m0]                                   ; C8C50D01
>	v_interp_p1_f32 v48, v0, 0, 3, [m0]                                          ; C8C00C00
>	v_interp_p2_f32 v48, [v48], v1, 0, 3, [m0]                                   ; C8C10C01
>	image_sample v[50:53], 15, 0, 0, 0, 0, 0, 0, 0, v[48:49], s[92:99], s[88:91] ; F0800F00 02D73230
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v18, v53, v17, v35                                                 ; D2820012 048E2335
>	v_mad_f32 v35, -2.0, v51, 1.0                                                ; D2820023 03CA66F5
>	v_mul_f32_e32 v54, v35, v13                                                  ; 106C1B23
>	v_mad_f32 v18, v27, v54, v18                                                 ; D2820012 044A6D1B
>	v_mad_f32 v18, v37, v54, v18                                                 ; D2820012 044A6D25
>	v_mul_f32_e32 v27, v54, v47                                                  ; 10365F36
>	v_mad_f32 v18, 2.0, v27, v18                                                 ; D2820012 044A36F4
>	v_mad_f32 v25, v16, v25, v38                                                 ; D2820019 049A3310
>	v_mad_f32 v25, v14, v46, v25                                                 ; D2820019 04665D0E
>	v_mad_f32 v25, v13, v35, v25                                                 ; D2820019 0466470D
>	v_mad_f32 v18, -v25, v25, v18                                                ; D2820012 244A3319
>	s_buffer_load_dword s4, s[0:3], 0xf                                          ; C202010F
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_rcp_f32_e32 v27, s4                                                        ; 7E365404
>	v_add_f32_e32 v35, v27, v18                                                  ; 0646251B
>	v_mad_f32 v37, 2.0, v21, -1.0                                                ; D2820025 03CE2AF4
>	v_mul_f32_e32 v46, v37, v16                                                  ; 105C2125
>	v_mad_f32 v55, v16, v37, v46                                                 ; D2820037 04BA4B10
>	v_mul_f32_e32 v56, v33, v28                                                  ; 10703921
>	v_mad_f32 v21, v23, v36, v56                                                 ; D2820015 04E24917
>	v_mad_f32 v22, -2.0, v31, 1.0                                                ; D2820016 03CA3EF5
>	v_mul_f32_e32 v23, v22, v15                                                  ; 102E1F16
>	v_mad_f32 v21, v23, v55, v21                                                 ; D2820015 04566F17
>	v_mad_f32 v21, v44, v39, v21                                                 ; D2820015 04564F2C
>	v_mad_f32 v24, -2.0, v42, 1.0                                                ; D2820018 03CA54F5
>	v_mul_f32_e32 v31, v24, v14                                                  ; 103E1D18
>	v_mad_f32 v21, v55, v31, v21                                                 ; D2820015 04563F37
>	v_mad_f32 v15, v15, v22, v23                                                 ; D282000F 045E2D0F
>	v_mad_f32 v21, v31, v15, v21                                                 ; D2820015 04561F1F
>	v_mad_f32 v21, v52, v17, v21                                                 ; D2820015 04562334
>	v_mad_f32 v22, 2.0, v50, -1.0                                                ; D2820016 03CE64F4
>	v_mul_f32_e32 v32, v22, v13                                                  ; 10401B16
>	v_mad_f32 v21, v55, v32, v21                                                 ; D2820015 04564137
>	v_mad_f32 v15, v15, v32, v21                                                 ; D282000F 0456410F
>	v_mul_f32_e32 v21, v32, v31                                                  ; 102A3F20
>	v_mad_f32 v15, 2.0, v21, v15                                                 ; D282000F 043E2AF4
>	v_mad_f32 v16, v16, v37, v23                                                 ; D2820010 045E4B10
>	v_mad_f32 v14, v14, v24, v16                                                 ; D282000E 0442310E
>	v_mad_f32 v13, v13, v22, v14                                                 ; D282000D 043A2D0D
>	v_mad_f32 v14, -v13, v13, v15                                                ; D282000E 243E1B0D
>	v_add_f32_e32 v15, v27, v14                                                  ; 061E1D1B
>	v_mul_f32_e32 v16, v23, v26                                                  ; 10203517
>	v_mad_f32 v16, v46, v38, v16                                                 ; D2820010 04424D2E
>	image_sample v21, 1, 0, 0, 0, 0, 0, 0, 0, v[29:30], s[80:87], s[60:63]       ; F0800100 01F4151D
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v21, -2.0, v21, 1.0                                                ; D2820015 03CA2AF5
>	v_mul_f32_e32 v21, v21, v28                                                  ; 102A3915
>	image_sample v19, 1, 0, 0, 0, 0, 0, 0, 0, v[19:20], s[80:87], s[60:63]       ; F0800100 01F41313
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v19, 2.0, v19, -1.0                                                ; D2820013 03CE26F4
>	v_mad_f32 v19, v19, v36, v21                                                 ; D2820013 04564913
>	v_add_f32_e32 v16, v16, v19                                                  ; 06202710
>	image_sample v19, 1, 0, 0, 0, 0, 0, 0, 0, v[40:41], s[80:87], s[60:63]       ; F0800100 01F41328
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v19, 2.0, v19, -1.0                                                ; D2820013 03CE26F4
>	v_mad_f32 v16, v19, v39, v16                                                 ; D2820010 04424F13
>	v_mul_f32_e32 v19, v31, v26                                                  ; 1026351F
>	v_mad_f32 v19, v46, v47, v19                                                 ; D2820013 044E5F2E
>	v_add_f32_e32 v16, v19, v16                                                  ; 06202113
>	v_mul_f32_e32 v19, v31, v38                                                  ; 10264D1F
>	v_mad_f32 v19, v23, v47, v19                                                 ; D2820013 044E5F17
>	v_add_f32_e32 v16, v19, v16                                                  ; 06202113
>	image_sample v19, 1, 0, 0, 0, 0, 0, 0, 0, v[48:49], s[80:87], s[60:63]       ; F0800100 01F41330
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v19, -2.0, v19, 1.0                                                ; D2820013 03CA26F5
>	v_mad_f32 v16, v19, v17, v16                                                 ; D2820010 04422313
>	v_mul_f32_e32 v17, v32, v26                                                  ; 10223520
>	v_mad_f32 v17, v46, v54, v17                                                 ; D2820011 04466D2E
>	v_add_f32_e32 v16, v16, v17                                                  ; 06202310
>	v_mul_f32_e32 v17, v32, v38                                                  ; 10224D20
>	v_mad_f32 v17, v23, v54, v17                                                 ; D2820011 04466D17
>	v_add_f32_e32 v16, v17, v16                                                  ; 06202111
>	v_mul_f32_e32 v17, v32, v47                                                  ; 10225F20
>	v_mad_f32 v17, v31, v54, v17                                                 ; D2820011 04466D1F
>	v_add_f32_e32 v16, v17, v16                                                  ; 06202111
>	v_mad_f32 v16, -v13, v25, v16                                                ; D2820010 2442330D
>	v_mul_f32_e32 v17, v16, v16                                                  ; 10222110
>	v_mad_f32 v19, v15, v35, -v17                                                ; D2820013 8446470F
>	v_max_f32_e32 v20, 0x33d6bf95, v19                                           ; 202826FF 33D6BF95
>	v_rsq_clamp_f32_e32 v20, v20                                                 ; 7E285914
>	v_rcp_f32_e32 v21, v19                                                       ; 7E2A5513
>	v_add_f32_e32 v16, v16, v16                                                  ; 06202110
>	s_buffer_load_dword s4, s[0:3], 0x21                                         ; C2020121
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v22, s4, v5                                                    ; 082C0A04
>	s_buffer_load_dword s5, s[0:3], 0x20                                         ; C2028120
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v23, s5, v4                                                    ; 082E0805
>	v_mul_f32_e32 v24, v23, v23                                                  ; 10302F17
>	v_mad_f32 v24, v22, v22, v24                                                 ; D2820018 04622D16
>	s_buffer_load_dword s60, s[0:3], 0x22                                        ; C21E0122
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v26, s60, v6                                                   ; 08340C3C
>	v_mad_f32 v24, v26, v26, v24                                                 ; D2820018 0462351A
>	v_max_f32_e32 v24, 0x33d6bf95, v24                                           ; 203030FF 33D6BF95
>	v_rsq_clamp_f32_e32 v24, v24                                                 ; 7E305918
>	s_buffer_load_dword s6, s[0:3], 0x15                                         ; C2030115
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v28, v24, v22, -s6                                                 ; D282001C 801A2D18
>	s_buffer_load_dword s7, s[0:3], 0x17                                         ; C2038117
>	v_mov_b32_e32 v29, s6                                                        ; 7E3A0206
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v28, v28, s7, v29                                                  ; D282001C 04740F1C
>	s_buffer_load_dword s6, s[0:3], 0x14                                         ; C2030114
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v29, v24, v23, -s6                                                 ; D282001D 801A2F18
>	v_mov_b32_e32 v30, s6                                                        ; 7E3C0206
>	v_mad_f32 v29, v29, s7, v30                                                  ; D282001D 04780F1D
>	v_mul_f32_e32 v30, v29, v29                                                  ; 103C3B1D
>	v_mad_f32 v30, v28, v28, v30                                                 ; D282001E 047A391C
>	s_buffer_load_dword s6, s[0:3], 0x16                                         ; C2030116
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v31, v24, v26, -s6                                                 ; D282001F 801A3518
>	v_mov_b32_e32 v32, s6                                                        ; 7E400206
>	v_mad_f32 v31, v31, s7, v32                                                  ; D282001F 04800F1F
>	v_mad_f32 v30, v31, v31, v30                                                 ; D282001E 047A3F1F
>	v_max_f32_e32 v30, 0x33d6bf95, v30                                           ; 203C3CFF 33D6BF95
>	v_rsq_clamp_f32_e32 v30, v30                                                 ; 7E3C591E
>	s_buffer_load_dword s6, s[0:3], 0x4                                          ; C2030104
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v29, v29, v30, s6                                                  ; D282001D 001A3D1D
>	s_buffer_load_dword s7, s[0:3], 0x6                                          ; C2038106
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v31, v31, v30, s7                                                  ; D282001F 001E3D1F
>	v_rcp_f32_e32 v31, v31                                                       ; 7E3E551F
>	v_mad_f32 v29, v29, v31, -v13                                                ; D282001D 84363F1D
>	v_mul_f32_e32 v32, v16, v29                                                  ; 10403B10
>	s_buffer_load_dword s8, s[0:3], 0x5                                          ; C2040105
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v28, v28, v30, s8                                                  ; D282001C 00223D1C
>	v_mad_f32 v28, v28, v31, -v25                                                ; D282001C 84663F1C
>	v_mad_f32 v15, v15, v28, -v32                                                ; D282000F 8482390F
>	v_mul_f32_e32 v15, v15, v28                                                  ; 101E390F
>	v_mul_f32_e32 v29, v29, v29                                                  ; 103A3B1D
>	v_mad_f32 v15, v29, v35, v15                                                 ; D282000F 043E471D
>	v_mul_f32_e32 v15, 0.5, v15                                                  ; 101E1EF0
>	v_mul_f32_e32 v30, v15, v21                                                  ; 103C2B0F
>	v_mov_b32_e32 v31, 0xbfb8aa65                                                ; 7E3E02FF BFB8AA65
>	v_mul_f32_e32 v30, v31, v30                                                  ; 103C3D1F
>	v_exp_f32_e32 v30, v30                                                       ; 7E3C4B1E
>	v_mul_f32_e32 v20, v20, v30                                                  ; 10283D14
>	s_buffer_load_dword s9, s[0:3], 0xe                                          ; C204810E
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v30, s9, v20                                                   ; 103C2809
>	v_cmp_ge_f32_e64 s[10:11], -v19, 0                                           ; D00C000A 20010113
>	v_cndmask_b32_e64 v19, 0, -1, s[10:11]                                       ; D2000813 00298280
>	v_cmp_ne_i32_e64 s[10:11], v19, 0                                            ; D10A000A 00010113
>	v_cndmask_b32_e64 v19, 0, 1.0, s[10:11]                                      ; D2000813 0029E480
>	v_mov_b32_e32 v33, 0xc1800000                                                ; 7E4202FF C1800000
>	v_mad_f32 v15, v21, v15, v33                                                 ; D282000F 04861F15
>	v_cmp_ge_f32_e64 s[10:11], v15, 0                                            ; D00C000A 0001010F
>	v_cndmask_b32_e64 v15, 0, -1, s[10:11]                                       ; D200080F 00298280
>	v_cmp_ne_i32_e64 s[10:11], v15, 0                                            ; D10A000A 0001010F
>	v_cndmask_b32_e64 v15, 0, 1.0, s[10:11]                                      ; D200080F 0029E480
>	v_add_f32_e32 v15, v19, v15                                                  ; 061E1F13
>	v_cmp_ge_f32_e64 s[10:11], -v15, 0                                           ; D00C000A 2001010F
>	v_cndmask_b32_e64 v15, 0, -1, s[10:11]                                       ; D200080F 00298280
>	v_cmp_ne_i32_e64 s[62:63], v15, 0                                            ; D10A003E 0001010F
>	v_cndmask_b32_e64 v15, 0, v30, s[62:63]                                      ; D200080F 00FA3C80
>	s_buffer_load_dword s10, s[0:3], 0x13                                        ; C2050113
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_rcp_f32_e32 v19, s10                                                       ; 7E26540A
>	v_add_f32_e32 v21, v18, v19                                                  ; 062A2712
>	v_add_f32_e32 v19, v14, v19                                                  ; 0626270E
>	v_mad_f32 v30, v19, v21, -v17                                                ; D282001E 84462B13
>	v_max_f32_e32 v34, 0x33d6bf95, v30                                           ; 20443CFF 33D6BF95
>	v_rsq_clamp_f32_e32 v34, v34                                                 ; 7E445922
>	v_rcp_f32_e32 v35, v30                                                       ; 7E46551E
>	s_buffer_load_dword s10, s[0:3], 0x19                                        ; C2050119
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v36, v24, v22, -s10                                                ; D2820024 802A2D18
>	s_buffer_load_dword s11, s[0:3], 0x1b                                        ; C205811B
>	v_mov_b32_e32 v37, s10                                                       ; 7E4A020A
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v36, v36, s11, v37                                                 ; D2820024 04941724
>	s_buffer_load_dword s10, s[0:3], 0x18                                        ; C2050118
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v37, v24, v23, -s10                                                ; D2820025 802A2F18
>	v_mov_b32_e32 v38, s10                                                       ; 7E4C020A
>	v_mad_f32 v37, v37, s11, v38                                                 ; D2820025 04981725
>	v_mul_f32_e32 v38, v37, v37                                                  ; 104C4B25
>	v_mad_f32 v38, v36, v36, v38                                                 ; D2820026 049A4924
>	s_buffer_load_dword s10, s[0:3], 0x1a                                        ; C205011A
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v39, v24, v26, -s10                                                ; D2820027 802A3518
>	v_mov_b32_e32 v40, s10                                                       ; 7E50020A
>	v_mad_f32 v39, v39, s11, v40                                                 ; D2820027 04A01727
>	v_mad_f32 v38, v39, v39, v38                                                 ; D2820026 049A4F27
>	v_max_f32_e32 v38, 0x33d6bf95, v38                                           ; 204C4CFF 33D6BF95
>	v_rsq_clamp_f32_e32 v38, v38                                                 ; 7E4C5926
>	s_buffer_load_dword s10, s[0:3], 0x8                                         ; C2050108
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v37, v37, v38, s10                                                 ; D2820025 002A4D25
>	s_buffer_load_dword s10, s[0:3], 0xa                                         ; C205010A
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v39, v39, v38, s10                                                 ; D2820027 002A4D27
>	v_rcp_f32_e32 v39, v39                                                       ; 7E4E5527
>	v_mad_f32 v37, v37, v39, -v13                                                ; D2820025 84364F25
>	v_mul_f32_e32 v16, v37, v16                                                  ; 10202125
>	s_buffer_load_dword s10, s[0:3], 0x9                                         ; C2050109
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v36, v36, v38, s10                                                 ; D2820024 002A4D24
>	v_mad_f32 v36, v36, v39, -v25                                                ; D2820024 84664F24
>	v_mad_f32 v16, v19, v36, -v16                                                ; D2820010 84424913
>	v_mul_f32_e32 v16, v16, v36                                                  ; 10204910
>	v_mul_f32_e32 v19, v37, v37                                                  ; 10264B25
>	v_mad_f32 v16, v19, v21, v16                                                 ; D2820010 04422B13
>	v_mul_f32_e32 v16, 0.5, v16                                                  ; 102020F0
>	v_mul_f32_e32 v19, v35, v16                                                  ; 10262123
>	v_mul_f32_e32 v21, v31, v19                                                  ; 102A271F
>	v_exp_f32_e32 v21, v21                                                       ; 7E2A4B15
>	v_mul_f32_e32 v21, v21, v34                                                  ; 102A4515
>	s_buffer_load_dword s10, s[0:3], 0x12                                        ; C2050112
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v34, s10, v21                                                  ; 10442A0A
>	v_cmp_ge_f32_e64 s[10:11], -v30, 0                                           ; D00C000A 2001011E
>	v_cndmask_b32_e64 v36, 0, -1, s[10:11]                                       ; D2000024 00298280
>	v_cmp_ne_i32_e64 s[10:11], v36, 0                                            ; D10A000A 00010124
>	v_cndmask_b32_e64 v36, 0, 1.0, s[10:11]                                      ; D2000024 0029E480
>	v_mad_f32 v16, v16, v35, v33                                                 ; D2820010 04864710
>	v_cmp_ge_f32_e64 s[10:11], v16, 0                                            ; D00C000A 00010110
>	v_cndmask_b32_e64 v16, 0, -1, s[10:11]                                       ; D2000010 00298280
>	v_cmp_ne_i32_e64 s[10:11], v16, 0                                            ; D10A000A 00010110
>	v_cndmask_b32_e64 v16, 0, 1.0, s[10:11]                                      ; D2000010 0029E480
>	v_add_f32_e32 v16, v36, v16                                                  ; 06202124
>	v_cmp_ge_f32_e64 s[10:11], -v16, 0                                           ; D00C000A 20010110
>	v_cndmask_b32_e64 v16, 0, -1, s[10:11]                                       ; D2000010 00298280
>	v_cmp_ne_i32_e64 s[80:81], v16, 0                                            ; D10A0050 00010110
>	v_cndmask_b32_e64 v16, 0, v34, s[80:81]                                      ; D2000010 01424480
>	v_add_f32_e32 v15, v15, v16                                                  ; 061E210F
>	v_mul_f32_e32 v16, v23, v24                                                  ; 10203117
>	v_mul_f32_e32 v16, v16, v13                                                  ; 10201B10
>	v_mul_f32_e32 v22, v22, v24                                                  ; 102C3116
>	v_mad_f32 v16, v25, v22, v16                                                 ; D2820010 04422D19
>	v_mad_f32 v16, v24, v26, v16                                                 ; D2820010 04423518
>	v_sub_f32_e32 v16, 1.0, v16                                                  ; 082020F2
>	v_mul_f32_e32 v22, v16, v16                                                  ; 102C2110
>	v_mul_f32_e32 v22, v22, v22                                                  ; 102C2D16
>	v_mul_f32_e32 v16, v22, v16                                                  ; 10202116
>	v_interp_p1_f32 v23, v0, 3, 6, [m0]                                          ; C85C1B00
>	v_interp_p2_f32 v23, [v23], v1, 3, 6, [m0]                                   ; C85D1B01
>	v_interp_p1_f32 v22, v0, 2, 6, [m0]                                          ; C8581A00
>	v_interp_p2_f32 v22, [v22], v1, 2, 6, [m0]                                   ; C8591A01
>	image_sample v22, 8, 0, 0, 0, 0, 0, 0, 0, v[22:23], s[72:79], s[56:59]       ; F0800800 01D21616
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mul_f32_e32 v22, 4.0, v22                                                  ; 102C2CF6
>	v_add_f32_e64 v22, 0, v22 clamp                                              ; D2060816 00022C80
>	v_sub_f32_e32 v9, 1.0, v22                                                   ; 08122CF2
>	v_mov_b32_e32 v11, v10                                                       ; 7E16030A
>	image_sample_l v[9:12], 15, 0, 0, 0, 0, 0, 0, 0, v[9:12], s[64:71], s[52:55] ; F0900F00 01B00909
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_log_f32_e32 v23, v11                                                       ; 7E2E4F0B
>	v_mov_b32_e32 v24, 0x400ccccd                                                ; 7E3002FF 400CCCCD
>	v_mul_legacy_f32_e32 v23, v24, v23                                           ; 0E2E2F18
>	v_exp_f32_e32 v23, v23                                                       ; 7E2E4B17
>	v_mad_f32 v15, v15, v16, v23                                                 ; D282000F 045E210F
>	v_mov_b32_e32 v26, 0x41800000                                                ; 7E3402FF 41800000
>	v_mad_f32 v34, v27, v26, v18                                                 ; D2820022 044A351B
>	v_mad_f32 v26, v27, v26, v14                                                 ; D282001A 043A351B
>	v_mad_f32 v17, v26, v34, -v17                                                ; D2820011 8446451A
>	v_max_f32_e32 v27, 0x33d6bf95, v17                                           ; 203622FF 33D6BF95
>	v_rsq_clamp_f32_e32 v27, v27                                                 ; 7E36591B
>	v_rcp_f32_e32 v35, v17                                                       ; 7E465511
>	v_mad_f32 v26, v26, v28, -v32                                                ; D282001A 8482391A
>	v_mul_f32_e32 v26, v26, v28                                                  ; 1034391A
>	v_mad_f32 v26, v29, v34, v26                                                 ; D282001A 046A451D
>	v_mul_f32_e32 v26, 0.5, v26                                                  ; 103434F0
>	v_mul_f32_e32 v28, v26, v35                                                  ; 1038471A
>	v_mul_f32_e32 v28, v31, v28                                                  ; 1038391F
>	v_exp_f32_e32 v28, v28                                                       ; 7E384B1C
>	v_mul_f32_e32 v27, v28, v27                                                  ; 1036371C
>	v_mul_f32_e32 v27, 0x3ba3d70a, v27                                           ; 103636FF 3BA3D70A
>	v_mul_f32_e32 v28, s9, v23                                                   ; 10382E09
>	v_mul_f32_e32 v28, v28, v27                                                  ; 1038371C
>	v_cmp_ge_f32_e64 s[10:11], -v17, 0                                           ; D00C000A 20010111
>	v_cndmask_b32_e64 v17, 0, -1, s[10:11]                                       ; D2000811 00298280
>	v_cmp_ne_i32_e64 s[10:11], v17, 0                                            ; D10A000A 00010111
>	v_cndmask_b32_e64 v17, 0, 1.0, s[10:11]                                      ; D2000811 0029E480
>	v_mad_f32 v26, v35, v26, v33                                                 ; D282001A 04863523
>	v_cmp_ge_f32_e64 s[10:11], v26, 0                                            ; D00C000A 0001011A
>	v_cndmask_b32_e64 v26, 0, -1, s[10:11]                                       ; D200001A 00298280
>	v_cmp_ne_i32_e64 s[10:11], v26, 0                                            ; D10A000A 0001011A
>	v_cndmask_b32_e64 v26, 0, 1.0, s[10:11]                                      ; D200001A 0029E480
>	v_add_f32_e32 v17, v17, v26                                                  ; 06223511
>	v_cmp_ge_f32_e64 s[10:11], -v17, 0                                           ; D00C000A 20010111
>	v_cndmask_b32_e64 v17, 0, -1, s[10:11]                                       ; D2000811 00298280
>	v_cmp_ne_i32_e64 s[52:53], v17, 0                                            ; D10A0034 00010111
>	v_cndmask_b32_e64 v17, 0, v28, s[52:53]                                      ; D2000811 00D23880
>	v_mul_f32_e32 v13, s6, v13                                                   ; 101A1A06
>	v_mad_f32 v13, s8, v25, v13                                                  ; D282000D 04363208
>	v_add_f32_e32 v13, s7, v13                                                   ; 061A1A07
>	v_add_f32_e64 v13, 0, v13 clamp                                              ; D206080D 00021A80
>	v_mad_f32 v15, v15, v13, v17                                                 ; D282000F 04461B0F
>	s_buffer_load_dword s6, s[0:3], 0x29                                         ; C2030129
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v17, s6, v5                                                    ; 10220A06
>	s_buffer_load_dword s6, s[0:3], 0x25                                         ; C2030125
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v17, v4, s6, v17                                                   ; D2820011 04440D04
>	s_buffer_load_dword s6, s[0:3], 0x2d                                         ; C203012D
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v17, v6, s6, v17                                                   ; D2820011 04440D06
>	s_buffer_load_dword s6, s[0:3], 0x31                                         ; C2030131
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_add_f32_e32 v17, s6, v17                                                   ; 06222206
>	s_buffer_load_dword s6, s[0:3], 0x2b                                         ; C203012B
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v25, s6, v5                                                    ; 10320A06
>	s_buffer_load_dword s6, s[0:3], 0x27                                         ; C2030127
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v25, v4, s6, v25                                                   ; D2820019 04640D04
>	s_buffer_load_dword s6, s[0:3], 0x2f                                         ; C203012F
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v25, v6, s6, v25                                                   ; D2820019 04640D06
>	s_buffer_load_dword s6, s[0:3], 0x33                                         ; C2030133
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_add_f32_e32 v25, s6, v25                                                   ; 06323206
>	v_rcp_f32_e32 v25, v25                                                       ; 7E325519
>	v_mul_f32_e32 v17, v17, v25                                                  ; 10223311
>	v_mad_f32 v17, -0.5, v17, -0.5                                               ; D2820011 03C622F1
>	v_mov_b32_e32 v26, 0x40c00000                                                ; 7E3402FF 40C00000
>	v_mad_f32 v17, v18, v26, v17                                                 ; D2820011 04463512
>	v_mad_f32 v29, -2.0, v30, v17                                                ; D282001D 04463CF5
>	s_buffer_load_dword s6, s[0:3], 0x28                                         ; C2030128
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v18, s6, v5                                                    ; 10240A06
>	s_buffer_load_dword s6, s[0:3], 0x24                                         ; C2030124
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v18, v4, s6, v18                                                   ; D2820012 04480D04
>	s_buffer_load_dword s6, s[0:3], 0x2c                                         ; C203012C
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v18, v6, s6, v18                                                   ; D2820012 04480D06
>	s_buffer_load_dword s6, s[0:3], 0x30                                         ; C2030130
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_add_f32_e32 v18, s6, v18                                                   ; 06242406
>	v_mul_f32_e32 v18, v18, v25                                                  ; 10243312
>	v_mad_f32 v18, 0.5, v18, -0.5                                                ; D2820012 03C624F0
>	v_mad_f32 v14, v14, v26, v18                                                 ; D282000E 044A350E
>	v_mad_f32 v28, 2.0, v30, v14                                                 ; D282001C 043A3CF4
>	image_sample v[31:34], 15, 0, 0, 0, 0, 0, 0, 0, v[28:29], s[44:51], s[32:35] ; F0800F00 010B1F1C
>	v_mov_b32_e32 v18, 0x3e800000                                                ; 7E2402FF 3E800000
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mul_f32_e32 v25, 0x3e800000, v33                                           ; 103242FF 3E800000
>	v_mad_f32 v35, 2.0, v30, v17                                                 ; D2820023 04463CF4
>	v_mov_b32_e32 v36, v28                                                       ; 7E48031C
>	v_mov_b32_e32 v37, v29                                                       ; 7E4A031D
>	v_mov_b32_e32 v37, v35                                                       ; 7E4A0323
>	image_sample v[35:38], 15, 0, 0, 0, 0, 0, 0, 0, v[36:37], s[44:51], s[32:35] ; F0800F00 010B2324
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v25, v37, v18, v25                                                 ; D2820019 04662525
>	v_mad_f32 v40, v30, v26, v17                                                 ; D2820028 0446351E
>	v_mad_f32 v39, -2.0, v30, v14                                                ; D2820027 043A3CF5
>	image_sample v[41:44], 15, 0, 0, 0, 0, 0, 0, 0, v[39:40], s[44:51], s[32:35] ; F0800F00 010B2927
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v14, v43, v18, v25                                                 ; D282000E 0466252B
>	v_mov_b32_e32 v40, v29                                                       ; 7E50031D
>	image_sample v[45:48], 15, 0, 0, 0, 0, 0, 0, 0, v[39:40], s[44:51], s[32:35] ; F0800F00 010B2D27
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v14, v47, v18, v14                                                 ; D282000E 043A252F
>	v_log_f32_e64 v14, |v14|                                                     ; D34E010E 0000010E
>	v_mul_f32_e32 v14, 0x400ccccd, v14                                           ; 101C1CFF 400CCCCD
>	v_exp_f32_e32 v14, v14                                                       ; 7E1C4B0E
>	s_buffer_load_dword s6, s[0:3], 0x5c                                         ; C203015C
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v14, v14, s6, v15                                                  ; D282000E 043C0D0E
>	v_mul_f32_e32 v15, v19, v19                                                  ; 101E2713
>	v_mul_f32_e32 v15, v15, v19                                                  ; 101E270F
>	v_mul_f32_e32 v15, v15, v22                                                  ; 101E2D0F
>	v_mad_f32 v14, v15, v23, v14                                                 ; D282000E 043A2F0F
>	s_buffer_load_dword s7, s[0:3], 0x39                                         ; C2038139
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v17, s7, v5                                                    ; 10220A07
>	s_buffer_load_dword s7, s[0:3], 0x35                                         ; C2038135
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v17, v4, s7, v17                                                   ; D2820011 04440F04
>	s_buffer_load_dword s7, s[0:3], 0x3d                                         ; C203813D
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v17, v6, s7, v17                                                   ; D2820011 04440F06
>	s_buffer_load_dword s7, s[0:3], 0x41                                         ; C2038141
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_add_f32_e32 v17, s7, v17                                                   ; 06222207
>	s_buffer_load_dword s7, s[0:3], 0x3b                                         ; C203813B
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v19, s7, v5                                                    ; 10260A07
>	s_buffer_load_dword s7, s[0:3], 0x37                                         ; C2038137
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v19, v4, s7, v19                                                   ; D2820013 044C0F04
>	s_buffer_load_dword s7, s[0:3], 0x3f                                         ; C203813F
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v19, v6, s7, v19                                                   ; D2820013 044C0F06
>	s_buffer_load_dword s7, s[0:3], 0x43                                         ; C2038143
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_add_f32_e32 v19, s7, v19                                                   ; 06262607
>	v_rcp_f32_e32 v19, v19                                                       ; 7E265513
>	v_mul_f32_e32 v17, v17, v19                                                  ; 10222711
>	v_mad_f32 v17, -0.5, v17, -0.5                                               ; D2820011 03C622F1
>	v_sub_f32_e32 v23, 1.0, v17                                                  ; 082E22F2
>	s_buffer_load_dword s7, s[0:3], 0x38                                         ; C2038138
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v17, s7, v5                                                    ; 10220A07
>	s_buffer_load_dword s7, s[0:3], 0x34                                         ; C2038134
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v17, v4, s7, v17                                                   ; D2820011 04440F04
>	s_buffer_load_dword s7, s[0:3], 0x3c                                         ; C203813C
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v17, v6, s7, v17                                                   ; D2820011 04440F06
>	s_buffer_load_dword s7, s[0:3], 0x40                                         ; C2038140
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_add_f32_e32 v17, s7, v17                                                   ; 06222207
>	v_mul_f32_e32 v17, v17, v19                                                  ; 10222711
>	v_mad_f32 v17, 0.5, v17, -0.5                                                ; D2820011 03C622F0
>	v_add_f32_e32 v22, 0, v17                                                    ; 062C2280
>	s_load_dwordx8 s[8:15], s[100:101], 0x50                                     ; C0C46550
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample v17, 1, 0, 0, 0, 0, 0, 0, 0, v[22:23], s[8:15], s[28:31]        ; F0800100 00E21116
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_add_f32_e64 v17, 0, v17 clamp                                              ; D2060811 00022280
>	s_buffer_load_dword s7, s[0:3], 0x54                                         ; C2038154
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v17, s7, v17                                                   ; 10222207
>	v_mul_f32_e32 v19, v14, v17                                                  ; 1026230E
>	v_mov_b32_e32 v22, 0xbf2b851f                                                ; 7E2C02FF BF2B851F
>	v_mad_f32 v14, v19, v22, v14                                                 ; D282000E 043A2D13
>	v_log_f32_e64 v14, |v14|                                                     ; D34E010E 0000010E
>	v_mul_f32_e32 v14, 0x3ee8ba1f, v14                                           ; 101C1CFF 3EE8BA1F
>	v_exp_f32_e32 v14, v14                                                       ; 7E1C4B0E
>	image_sample v[49:52], 15, 0, 0, 0, 0, 0, 0, 0, v[7:8], s[36:43], s[24:27]   ; F0800F00 00C93107
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_sub_f32_e32 v7, 1.0, v52                                                   ; 080E68F2
>	v_mad_f32 v8, v14, v7, v51                                                   ; D2820008 04CE0F0E
>	v_mov_b32_e32 v14, 0xbecccccd                                                ; 7E1C02FF BECCCCCD
>	v_add_f32_e32 v8, v14, v8                                                    ; 0610110E
>	v_mov_b32_e32 v19, 0x3ecccccd                                                ; 7E2602FF 3ECCCCCD
>	v_mad_f32 v55, v2, v8, v19                                                   ; D2820037 044E1102
>	s_buffer_load_dword s7, s[0:3], 0xd                                          ; C203810D
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v8, s7, v20                                                    ; 10102807
>	v_cndmask_b32_e64 v8, 0, v8, s[62:63]                                        ; D2000008 00FA1080
>	s_buffer_load_dword s8, s[0:3], 0x11                                         ; C2040111
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v23, s8, v21                                                   ; 102E2A08
>	v_cndmask_b32_e64 v23, 0, v23, s[80:81]                                      ; D2000817 01422E80
>	v_add_f32_e32 v8, v8, v23                                                    ; 06102F08
>	v_log_f32_e32 v23, v10                                                       ; 7E2E4F0A
>	v_mul_legacy_f32_e32 v23, v24, v23                                           ; 0E2E2F18
>	v_exp_f32_e32 v23, v23                                                       ; 7E2E4B17
>	v_mad_f32 v8, v8, v16, v23                                                   ; D2820008 045E2108
>	v_mul_f32_e32 v25, s7, v23                                                   ; 10322E07
>	v_mul_f32_e32 v25, v25, v27                                                  ; 10323719
>	v_cndmask_b32_e64 v25, 0, v25, s[52:53]                                      ; D2000819 00D23280
>	v_mad_f32 v8, v8, v13, v25                                                   ; D2820008 04661B08
>	v_mul_f32_e32 v25, 0x3e800000, v32                                           ; 103240FF 3E800000
>	v_mad_f32 v25, v36, v18, v25                                                 ; D2820019 04662524
>	v_mad_f32 v25, v42, v18, v25                                                 ; D2820019 0466252A
>	v_mad_f32 v25, v46, v18, v25                                                 ; D2820019 0466252E
>	v_log_f32_e64 v25, |v25|                                                     ; D34E0119 00000119
>	v_mul_f32_e32 v25, 0x400ccccd, v25                                           ; 103232FF 400CCCCD
>	v_exp_f32_e32 v25, v25                                                       ; 7E324B19
>	v_mad_f32 v8, v25, s6, v8                                                    ; D2820008 04200D19
>	v_mad_f32 v8, v15, v23, v8                                                   ; D2820008 04222F0F
>	v_mul_f32_e32 v23, v8, v17                                                   ; 102E2308
>	v_mad_f32 v8, v23, v22, v8                                                   ; D2820008 04222D17
>	v_log_f32_e64 v8, |v8|                                                       ; D34E0108 00000108
>	v_mul_f32_e32 v8, 0x3ee8ba1f, v8                                             ; 101010FF 3EE8BA1F
>	v_exp_f32_e32 v8, v8                                                         ; 7E104B08
>	v_mad_f32 v8, v8, v7, v50                                                    ; D2820008 04CA0F08
>	v_add_f32_e32 v8, v14, v8                                                    ; 0610110E
>	v_mad_f32 v54, v2, v8, v19                                                   ; D2820036 044E1102
>	s_buffer_load_dword s7, s[0:3], 0xc                                          ; C203810C
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v8, s7, v20                                                    ; 10102807
>	v_cndmask_b32_e64 v8, 0, v8, s[62:63]                                        ; D2000008 00FA1080
>	s_buffer_load_dword s8, s[0:3], 0x10                                         ; C2040110
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v20, s8, v21                                                   ; 10282A08
>	v_cndmask_b32_e64 v20, 0, v20, s[80:81]                                      ; D2000014 01422880
>	v_add_f32_e32 v8, v8, v20                                                    ; 06102908
>	v_log_f32_e32 v20, v9                                                        ; 7E284F09
>	v_mul_legacy_f32_e32 v20, v24, v20                                           ; 0E282918
>	v_exp_f32_e32 v20, v20                                                       ; 7E284B14
>	v_mad_f32 v8, v8, v16, v20                                                   ; D2820008 04522108
>	v_mul_f32_e32 v16, s7, v20                                                   ; 10202807
>	v_mul_f32_e32 v16, v16, v27                                                  ; 10203710
>	v_cndmask_b32_e64 v16, 0, v16, s[52:53]                                      ; D2000010 00D22080
>	v_mad_f32 v8, v8, v13, v16                                                   ; D2820008 04421B08
>	v_mul_f32_e32 v13, 0x3e800000, v31                                           ; 101A3EFF 3E800000
>	v_mad_f32 v13, v35, v18, v13                                                 ; D282000D 04362523
>	v_mad_f32 v13, v41, v18, v13                                                 ; D282000D 04362529
>	v_mad_f32 v13, v45, v18, v13                                                 ; D282000D 0436252D
>	v_log_f32_e64 v13, |v13|                                                     ; D34E010D 0000010D
>	v_mul_f32_e32 v13, 0x400ccccd, v13                                           ; 101A1AFF 400CCCCD
>	v_exp_f32_e32 v13, v13                                                       ; 7E1A4B0D
>	v_mad_f32 v8, v13, s6, v8                                                    ; D2820008 04200D0D
>	v_mad_f32 v8, v15, v20, v8                                                   ; D2820008 0422290F
>	v_mul_f32_e32 v13, v8, v17                                                   ; 101A2308
>	v_mad_f32 v8, v13, v22, v8                                                   ; D2820008 04222D0D
>	v_log_f32_e64 v8, |v8|                                                       ; D34E0108 00000108
>	v_mul_f32_e32 v8, 0x3ee8ba1f, v8                                             ; 101010FF 3EE8BA1F
>	v_exp_f32_e32 v8, v8                                                         ; 7E104B08
>	v_mad_f32 v7, v8, v7, v49                                                    ; D2820007 04C60F08
>	v_add_f32_e32 v7, v14, v7                                                    ; 060E0F0E
>	v_mad_f32 v53, v2, v7, v19                                                   ; D2820035 044E0F02
>	s_load_dwordx8 s[8:15], s[100:101], 0x48                                     ; C0C46548
>	v_readlane_b32 s24, v57, 4                                                   ; 02310939
>	v_readlane_b32 s25, v57, 5                                                   ; 02330B39
>	v_readlane_b32 s26, v57, 6                                                   ; 02350D39
>	v_readlane_b32 s27, v57, 7                                                   ; 02370F39
>	s_nop 2                                                                      ; BF800002
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample v[13:15], 7, 0, 0, 0, 0, 0, 0, 0, v[53:56], s[8:15], s[24:27]   ; F0800700 00C20D35
>	v_mad_f32 v2, -v3, v2, v2                                                    ; D2820002 240A0503
>	v_sub_f32_e32 v3, 1.0, v2                                                    ; 080604F2
>	v_mul_f32_e32 v7, v54, v3                                                    ; 100E0736
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v7, v2, v14, v7                                                    ; D2820007 041E1D02
>	s_buffer_load_dword s6, s[0:3], 0x59                                         ; C2030159
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v8, s6, v7                                                     ; 08100E06
>	v_subrev_f32_e32 v5, s4, v5                                                  ; 0A0A0A04
>	v_subrev_f32_e32 v4, s5, v4                                                  ; 0A080805
>	v_mul_f32_e32 v4, v4, v4                                                     ; 10080904
>	v_mad_f32 v4, v5, v5, v4                                                     ; D2820004 04120B05
>	v_subrev_f32_e32 v5, s60, v6                                                 ; 0A0A0C3C
>	v_mad_f32 v4, v5, v5, v4                                                     ; D2820004 04120B05
>	s_buffer_load_dword s4, s[0:3], 0x1d                                         ; C202011D
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v4, s4, v4                                                     ; 10080804
>	s_buffer_load_dword s4, s[0:3], 0x1c                                         ; C202011C
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v6, s4, v5                                                     ; 100C0A04
>	v_mul_f32_e32 v6, 0x3fb8aa65, v6                                             ; 100C0CFF 3FB8AA65
>	v_exp_f32_e32 v6, v6                                                         ; 7E0C4B06
>	v_sub_f32_e32 v6, 1.0, v6                                                    ; 080C0CF2
>	v_mul_f32_e32 v4, v4, v6                                                     ; 10080D04
>	v_rcp_f32_e32 v5, v5                                                         ; 7E0A5505
>	v_mul_f32_e32 v4, v4, v5                                                     ; 10080B04
>	v_mul_f32_e32 v4, 0x3fb8aa65, v4                                             ; 100808FF 3FB8AA65
>	v_exp_f32_e32 v4, v4                                                         ; 7E084B04
>	v_add_f32_e64 v4, 0, v4 clamp                                                ; D2060804 00020880
>	v_sub_f32_e32 v4, 1.0, v4                                                    ; 080808F2
>	v_interp_p1_f32 v5, v0, 1, 5, [m0]                                           ; C8141500
>	v_interp_p2_f32 v5, [v5], v1, 1, 5, [m0]                                     ; C8151501
>	v_interp_p1_f32 v6, v0, 2, 5, [m0]                                           ; C8181600
>	v_interp_p2_f32 v6, [v6], v1, 2, 5, [m0]                                     ; C8191601
>	v_rcp_f32_e32 v0, v6                                                         ; 7E005506
>	v_mul_f32_e32 v0, v5, v0                                                     ; 10000105
>	v_mad_f32 v0, 0.5, v0, 0.5                                                   ; D2820000 03C200F0
>	s_buffer_load_dword s4, s[0:3], 0x1f                                         ; C202011F
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mov_b32_e32 v1, s4                                                         ; 7E020204
>	v_readlane_b32 s4, v57, 0                                                    ; 02090139
>	v_readlane_b32 s5, v57, 1                                                    ; 020B0339
>	v_readlane_b32 s6, v57, 2                                                    ; 020D0539
>	v_readlane_b32 s7, v57, 3                                                    ; 020F0739
>	s_nop 2                                                                      ; BF800002
>	image_sample v0, 1, 0, 0, 0, 0, 0, 0, 0, v[0:1], s[16:23], s[4:7]            ; F0800100 00240000
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mul_f32_e32 v0, v0, v4                                                     ; 10000900
>	v_mad_f32 v1, v0, v8, v7                                                     ; D2820001 041E1100
>	v_mul_f32_e32 v4, v53, v3                                                    ; 10080735
>	v_mad_f32 v4, v2, v13, v4                                                    ; D2820004 04121B02
>	s_buffer_load_dword s4, s[0:3], 0x58                                         ; C2020158
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v5, s4, v4                                                     ; 080A0804
>	v_mad_f32 v4, v0, v5, v4                                                     ; D2820004 04120B00
>	v_cvt_pkrtz_f16_f32_e32 v1, v4, v1                                           ; 5E020304
>	v_mul_f32_e32 v3, v55, v3                                                    ; 10060737
>	v_mad_f32 v2, v2, v15, v3                                                    ; D2820002 040E1F02
>	s_buffer_load_dword s0, s[0:3], 0x5a                                         ; C200015A
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v3, s0, v2                                                     ; 08060400
>	v_mad_f32 v0, v0, v3, v2                                                     ; D2820000 040A0700
>	v_mul_f32_e32 v2, 0x3e800000, v34                                            ; 100444FF 3E800000
>	v_mad_f32 v2, v38, v18, v2                                                   ; D2820002 040A2526
>	v_mad_f32 v2, v44, v18, v2                                                   ; D2820002 040A252C
>	v_mad_f32 v2, v48, v18, v2                                                   ; D2820002 040A2530
>	v_log_f32_e32 v3, v12                                                        ; 7E064F0C
>	v_mul_legacy_f32_e32 v3, 1.0, v3                                             ; 0E0606F2
>	v_exp_f32_e32 v3, v3                                                         ; 7E064B03
>	v_mad_f32 v2, v2, v3, v3                                                     ; D2820002 040E0702
>	v_add_f32_e64 v2, 0, v2 clamp                                                ; D2060802 00020480
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v2                                           ; 5E000500
>	exp 15, 0, 1, 1, 1, v1, v0, v1, v0                                           ; F8001C0F 00010001
>	s_endpgm                                                                     ; BF810000
>
>SGPRS: 104
>VGPRS: 60
>Code Size: 4136 bytes
>LDS: 0 blocks
>Scratch: 2048 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL IN[2]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL OUT[2], GENERIC[10]
>DCL OUT[3], GENERIC[11]
>DCL OUT[4], GENERIC[12]
>DCL OUT[5], GENERIC[13]
>DCL OUT[6], GENERIC[14]
>DCL OUT[7], GENERIC[15]
>DCL CONST[0..13]
>DCL TEMP[0..7], LOCAL
>IMM[0] FLT32 {    0.5000,    -0.5000,     0.0000,     1.0000}
>  0: MUL TEMP[0].xy, CONST[13], IN[1]
>  1: MOV TEMP[0].xy, TEMP[0].xyxx
>  2: ADD TEMP[1], CONST[12], IN[1].xyxy
>  3: MUL TEMP[1], TEMP[1], CONST[13].xyxy
>  4: MUL TEMP[2].xy, TEMP[1], CONST[13].zzzz
>  5: MOV TEMP[2].xy, TEMP[2].xyxx
>  6: MUL TEMP[3].xy, TEMP[1].zwzw, CONST[13].wwww
>  7: MOV TEMP[3].xy, TEMP[3].xyxx
>  8: MUL TEMP[4].xyz, CONST[1], IN[0].yyyy
>  9: MOV TEMP[1].xyz, TEMP[4].xyzx
> 10: MAD TEMP[4].xyz, IN[0].xxxx, CONST[0], TEMP[1]
> 11: MOV TEMP[1].xyz, TEMP[4].xyzx
> 12: MAD TEMP[4].xyz, IN[0].zzzz, CONST[2], TEMP[1]
> 13: MOV TEMP[1].xyz, TEMP[4].xyzx
> 14: ADD TEMP[4].xyz, TEMP[1], CONST[3]
> 15: MUL TEMP[5].xy, TEMP[4].yyyy, CONST[9]
> 16: MOV TEMP[5].xy, TEMP[5].xyxx
> 17: MAD TEMP[6].xy, TEMP[4].xxxx, CONST[8], TEMP[5]
> 18: MOV TEMP[5].xy, TEMP[6].xyxx
> 19: MAD TEMP[6].xy, TEMP[4].zzzz, CONST[10], TEMP[5]
> 20: MOV TEMP[5].xy, TEMP[6].xyxx
> 21: ADD TEMP[6].xy, TEMP[5], CONST[11]
> 22: MOV TEMP[5].xy, TEMP[6].xyxx
> 23: MAD TEMP[6].xy, TEMP[5], IMM[0].xyzz, IMM[0].yyyy
> 24: MOV TEMP[6].xy, TEMP[6].xyxx
> 25: MUL TEMP[5], TEMP[4].yyyy, CONST[5]
> 26: MAD TEMP[5], TEMP[4].xxxx, CONST[4], TEMP[5]
> 27: MAD TEMP[5], TEMP[4].zzzz, CONST[6], TEMP[5]
> 28: MOV TEMP[4].xyz, TEMP[4].xyzx
> 29: ADD TEMP[1], TEMP[5], CONST[7]
> 30: MOV TEMP[5].xyz, TEMP[1].xywx
> 31: MOV TEMP[7].xy, IN[2].xyxx
> 32: MOV TEMP[6].zw, IMM[0].zzzz
> 33: MOV TEMP[0].zw, IMM[0].wwzw
> 34: MOV TEMP[2].zw, IMM[0].wwzw
> 35: MOV TEMP[3].zw, IMM[0].wwzw
> 36: MOV TEMP[7].zw, IMM[0].wwzw
> 37: MOV TEMP[4].w, IMM[0].wwww
> 38: MOV TEMP[5].w, IMM[0].wwww
> 39: MOV OUT[6], TEMP[5]
> 40: MOV OUT[7], TEMP[6]
> 41: MOV OUT[1], TEMP[0]
> 42: MOV OUT[0], TEMP[1]
> 43: MOV OUT[2], TEMP[2]
> 44: MOV OUT[3], TEMP[3]
> 45: MOV OUT[4], TEMP[7]
> 46: MOV OUT[5], TEMP[4]
> 47: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 0)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 4)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %12, i32 8)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %12, i32 16)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %12, i32 20)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %12, i32 24)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %12, i32 32)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %12, i32 36)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %12, i32 40)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %12, i32 48)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %12, i32 52)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %12, i32 56)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %12, i32 64)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %12, i32 68)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %12, i32 72)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %12, i32 76)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %12, i32 80)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %12, i32 84)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %12, i32 88)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %12, i32 92)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %12, i32 96)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %12, i32 100)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %12, i32 104)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %12, i32 108)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %12, i32 112)
>  %38 = call float @llvm.SI.load.const(<16 x i8> %12, i32 116)
>  %39 = call float @llvm.SI.load.const(<16 x i8> %12, i32 120)
>  %40 = call float @llvm.SI.load.const(<16 x i8> %12, i32 124)
>  %41 = call float @llvm.SI.load.const(<16 x i8> %12, i32 128)
>  %42 = call float @llvm.SI.load.const(<16 x i8> %12, i32 132)
>  %43 = call float @llvm.SI.load.const(<16 x i8> %12, i32 144)
>  %44 = call float @llvm.SI.load.const(<16 x i8> %12, i32 148)
>  %45 = call float @llvm.SI.load.const(<16 x i8> %12, i32 160)
>  %46 = call float @llvm.SI.load.const(<16 x i8> %12, i32 164)
>  %47 = call float @llvm.SI.load.const(<16 x i8> %12, i32 176)
>  %48 = call float @llvm.SI.load.const(<16 x i8> %12, i32 180)
>  %49 = call float @llvm.SI.load.const(<16 x i8> %12, i32 192)
>  %50 = call float @llvm.SI.load.const(<16 x i8> %12, i32 196)
>  %51 = call float @llvm.SI.load.const(<16 x i8> %12, i32 200)
>  %52 = call float @llvm.SI.load.const(<16 x i8> %12, i32 204)
>  %53 = call float @llvm.SI.load.const(<16 x i8> %12, i32 208)
>  %54 = call float @llvm.SI.load.const(<16 x i8> %12, i32 212)
>  %55 = call float @llvm.SI.load.const(<16 x i8> %12, i32 216)
>  %56 = call float @llvm.SI.load.const(<16 x i8> %12, i32 220)
>  %57 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %58 = load <16 x i8> addrspace(2)* %57, !tbaa !0
>  %59 = add i32 %5, %7
>  %60 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %58, i32 0, i32 %59)
>  %61 = extractelement <4 x float> %60, i32 0
>  %62 = extractelement <4 x float> %60, i32 1
>  %63 = extractelement <4 x float> %60, i32 2
>  %64 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %65 = load <16 x i8> addrspace(2)* %64, !tbaa !0
>  %66 = add i32 %5, %7
>  %67 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %65, i32 0, i32 %66)
>  %68 = extractelement <4 x float> %67, i32 0
>  %69 = extractelement <4 x float> %67, i32 1
>  %70 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 2
>  %71 = load <16 x i8> addrspace(2)* %70, !tbaa !0
>  %72 = add i32 %5, %7
>  %73 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %71, i32 0, i32 %72)
>  %74 = extractelement <4 x float> %73, i32 0
>  %75 = extractelement <4 x float> %73, i32 1
>  %76 = fmul float %53, %68
>  %77 = fmul float %54, %69
>  %78 = fadd float %49, %68
>  %79 = fadd float %50, %69
>  %80 = fadd float %51, %68
>  %81 = fadd float %52, %69
>  %82 = fmul float %78, %53
>  %83 = fmul float %79, %54
>  %84 = fmul float %80, %53
>  %85 = fmul float %81, %54
>  %86 = fmul float %82, %55
>  %87 = fmul float %83, %55
>  %88 = fmul float %84, %56
>  %89 = fmul float %85, %56
>  %90 = fmul float %16, %62
>  %91 = fmul float %17, %62
>  %92 = fmul float %18, %62
>  %93 = fmul float %61, %13
>  %94 = fadd float %93, %90
>  %95 = fmul float %61, %14
>  %96 = fadd float %95, %91
>  %97 = fmul float %61, %15
>  %98 = fadd float %97, %92
>  %99 = fmul float %63, %19
>  %100 = fadd float %99, %94
>  %101 = fmul float %63, %20
>  %102 = fadd float %101, %96
>  %103 = fmul float %63, %21
>  %104 = fadd float %103, %98
>  %105 = fadd float %100, %22
>  %106 = fadd float %102, %23
>  %107 = fadd float %104, %24
>  %108 = fmul float %106, %43
>  %109 = fmul float %106, %44
>  %110 = fmul float %105, %41
>  %111 = fadd float %110, %108
>  %112 = fmul float %105, %42
>  %113 = fadd float %112, %109
>  %114 = fmul float %107, %45
>  %115 = fadd float %114, %111
>  %116 = fmul float %107, %46
>  %117 = fadd float %116, %113
>  %118 = fadd float %115, %47
>  %119 = fadd float %117, %48
>  %120 = fmul float %118, 5.000000e-01
>  %121 = fadd float %120, -5.000000e-01
>  %122 = fmul float %119, -5.000000e-01
>  %123 = fadd float %122, -5.000000e-01
>  %124 = fmul float %106, %29
>  %125 = fmul float %106, %30
>  %126 = fmul float %106, %31
>  %127 = fmul float %106, %32
>  %128 = fmul float %105, %25
>  %129 = fadd float %128, %124
>  %130 = fmul float %105, %26
>  %131 = fadd float %130, %125
>  %132 = fmul float %105, %27
>  %133 = fadd float %132, %126
>  %134 = fmul float %105, %28
>  %135 = fadd float %134, %127
>  %136 = fmul float %107, %33
>  %137 = fadd float %136, %129
>  %138 = fmul float %107, %34
>  %139 = fadd float %138, %131
>  %140 = fmul float %107, %35
>  %141 = fadd float %140, %133
>  %142 = fmul float %107, %36
>  %143 = fadd float %142, %135
>  %144 = fadd float %137, %37
>  %145 = fadd float %139, %38
>  %146 = fadd float %141, %39
>  %147 = fadd float %143, %40
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %76, float %77, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %86, float %87, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 34, i32 0, float %88, float %89, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 35, i32 0, float %74, float %75, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 36, i32 0, float %105, float %106, float %107, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 37, i32 0, float %144, float %145, float %147, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 38, i32 0, float %121, float %123, float 0.000000e+00, float 0.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %144, float %145, float %146, float %147)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                             ; 4A00000A
>	s_load_dwordx4 s[4:7], s[8:9], 0x0                    ; C0820900
>	s_load_dwordx4 s[12:15], s[8:9], 0x4                  ; C0860904
>	s_load_dwordx4 s[8:11], s[8:9], 0x8                   ; C0840908
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[12:15], 0 idxen ; E00C2000 80030100
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                    ; C0800300
>	s_waitcnt vmcnt(0) lgkmcnt(0)                         ; BF8C0070
>	s_buffer_load_dword s12, s[0:3], 0x35                 ; C2060135
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v5, s12, v2                             ; 100A040C
>	s_buffer_load_dword s13, s[0:3], 0x34                 ; C2068134
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v6, s13, v1                             ; 100C020D
>	v_mov_b32_e32 v7, 1.0                                 ; 7E0E02F2
>	v_mov_b32_e32 v8, 0                                   ; 7E100280
>	exp 15, 32, 0, 0, 0, v6, v5, v8, v7                   ; F800020F 07080506
>	s_buffer_load_dword s14, s[0:3], 0x31                 ; C2070131
>	s_waitcnt expcnt(0) lgkmcnt(0)                        ; BF8C000F
>	v_add_f32_e32 v5, s14, v2                             ; 060A040E
>	v_mul_f32_e32 v5, s12, v5                             ; 100A0A0C
>	s_buffer_load_dword s14, s[0:3], 0x36                 ; C2070136
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v5, s14, v5                             ; 100A0A0E
>	s_buffer_load_dword s15, s[0:3], 0x30                 ; C2078130
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v6, s15, v1                             ; 060C020F
>	v_mul_f32_e32 v6, s13, v6                             ; 100C0C0D
>	v_mul_f32_e32 v6, s14, v6                             ; 100C0C0E
>	exp 15, 33, 0, 0, 0, v6, v5, v8, v7                   ; F800021F 07080506
>	s_buffer_load_dword s14, s[0:3], 0x33                 ; C2070133
>	s_waitcnt expcnt(0) lgkmcnt(0)                        ; BF8C000F
>	v_add_f32_e32 v5, s14, v2                             ; 060A040E
>	v_mul_f32_e32 v5, s12, v5                             ; 100A0A0C
>	s_buffer_load_dword s12, s[0:3], 0x37                 ; C2060137
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v5, s12, v5                             ; 100A0A0C
>	s_buffer_load_dword s14, s[0:3], 0x32                 ; C2070132
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v1, s14, v1                             ; 0602020E
>	v_mul_f32_e32 v1, s13, v1                             ; 1002020D
>	v_mul_f32_e32 v1, s12, v1                             ; 1002020C
>	exp 15, 34, 0, 0, 0, v1, v5, v8, v7                   ; F800022F 07080501
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	buffer_load_format_xyzw v[1:4], v0, s[8:11], 0 idxen  ; E00C2000 80020100
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	exp 15, 35, 0, 0, 0, v1, v2, v8, v7                   ; F800023F 07080201
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	buffer_load_format_xyzw v[0:3], v0, s[4:7], 0 idxen   ; E00C2000 80010000
>	s_buffer_load_dword s4, s[0:3], 0x6                   ; C2020106
>	s_waitcnt vmcnt(0) lgkmcnt(0)                         ; BF8C0070
>	v_mul_f32_e32 v4, s4, v1                              ; 10080204
>	s_buffer_load_dword s4, s[0:3], 0x2                   ; C2020102
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v4, v0, s4, v4                              ; D2820004 04100900
>	s_buffer_load_dword s4, s[0:3], 0xa                   ; C202010A
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v4, v2, s4, v4                              ; D2820004 04100902
>	s_buffer_load_dword s4, s[0:3], 0xe                   ; C202010E
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v4, s4, v4                              ; 06080804
>	s_buffer_load_dword s4, s[0:3], 0x5                   ; C2020105
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v5, s4, v1                              ; 100A0204
>	s_buffer_load_dword s4, s[0:3], 0x1                   ; C2020101
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v5, v0, s4, v5                              ; D2820005 04140900
>	s_buffer_load_dword s4, s[0:3], 0x9                   ; C2020109
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v5, v2, s4, v5                              ; D2820005 04140902
>	s_buffer_load_dword s4, s[0:3], 0xd                   ; C202010D
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v5, s4, v5                              ; 060A0A04
>	s_buffer_load_dword s4, s[0:3], 0x4                   ; C2020104
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v6, s4, v1                              ; 100C0204
>	s_buffer_load_dword s4, s[0:3], 0x0                   ; C2020100
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v6, v0, s4, v6                              ; D2820006 04180900
>	s_buffer_load_dword s4, s[0:3], 0x8                   ; C2020108
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v0, v2, s4, v6                              ; D2820000 04180902
>	s_buffer_load_dword s4, s[0:3], 0xc                   ; C202010C
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v0, s4, v0                              ; 06000004
>	exp 15, 36, 0, 0, 0, v0, v5, v4, v7                   ; F800024F 07040500
>	s_buffer_load_dword s4, s[0:3], 0x17                  ; C2020117
>	s_waitcnt expcnt(0) lgkmcnt(0)                        ; BF8C000F
>	v_mul_f32_e32 v1, s4, v5                              ; 10020A04
>	s_buffer_load_dword s4, s[0:3], 0x13                  ; C2020113
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v1, v0, s4, v1                              ; D2820001 04040900
>	s_buffer_load_dword s4, s[0:3], 0x1b                  ; C202011B
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v1, v4, s4, v1                              ; D2820001 04040904
>	s_buffer_load_dword s4, s[0:3], 0x1f                  ; C202011F
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v1, s4, v1                              ; 06020204
>	s_buffer_load_dword s4, s[0:3], 0x15                  ; C2020115
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v2, s4, v5                              ; 10040A04
>	s_buffer_load_dword s4, s[0:3], 0x11                  ; C2020111
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v2, v0, s4, v2                              ; D2820002 04080900
>	s_buffer_load_dword s4, s[0:3], 0x19                  ; C2020119
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v2, v4, s4, v2                              ; D2820002 04080904
>	s_buffer_load_dword s4, s[0:3], 0x1d                  ; C202011D
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v2, s4, v2                              ; 06040404
>	s_buffer_load_dword s4, s[0:3], 0x14                  ; C2020114
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v3, s4, v5                              ; 10060A04
>	s_buffer_load_dword s4, s[0:3], 0x10                  ; C2020110
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v3, v0, s4, v3                              ; D2820003 040C0900
>	s_buffer_load_dword s4, s[0:3], 0x18                  ; C2020118
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v3, v4, s4, v3                              ; D2820003 040C0904
>	s_buffer_load_dword s4, s[0:3], 0x1c                  ; C202011C
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v3, s4, v3                              ; 06060604
>	exp 15, 37, 0, 0, 0, v3, v2, v1, v7                   ; F800025F 07010203
>	s_buffer_load_dword s4, s[0:3], 0x24                  ; C2020124
>	s_waitcnt expcnt(0) lgkmcnt(0)                        ; BF8C000F
>	v_mul_f32_e32 v6, s4, v5                              ; 100C0A04
>	s_buffer_load_dword s4, s[0:3], 0x20                  ; C2020120
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v6, v0, s4, v6                              ; D2820006 04180900
>	s_buffer_load_dword s4, s[0:3], 0x28                  ; C2020128
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v6, v4, s4, v6                              ; D2820006 04180904
>	s_buffer_load_dword s4, s[0:3], 0x2c                  ; C202012C
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v6, s4, v6                              ; 060C0C04
>	v_mad_f32 v6, 0.5, v6, -0.5                           ; D2820006 03C60CF0
>	s_buffer_load_dword s4, s[0:3], 0x25                  ; C2020125
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v7, s4, v5                              ; 100E0A04
>	s_buffer_load_dword s4, s[0:3], 0x21                  ; C2020121
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v7, v0, s4, v7                              ; D2820007 041C0900
>	s_buffer_load_dword s4, s[0:3], 0x29                  ; C2020129
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v7, v4, s4, v7                              ; D2820007 041C0904
>	s_buffer_load_dword s4, s[0:3], 0x2d                  ; C202012D
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v7, s4, v7                              ; 060E0E04
>	v_mad_f32 v7, -0.5, v7, -0.5                          ; D2820007 03C60EF1
>	exp 15, 38, 0, 0, 0, v6, v7, v8, v8                   ; F800026F 08080706
>	s_buffer_load_dword s4, s[0:3], 0x16                  ; C2020116
>	s_waitcnt expcnt(0) lgkmcnt(0)                        ; BF8C000F
>	v_mul_f32_e32 v5, s4, v5                              ; 100A0A04
>	s_buffer_load_dword s4, s[0:3], 0x12                  ; C2020112
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v0, v0, s4, v5                              ; D2820000 04140900
>	s_buffer_load_dword s4, s[0:3], 0x1a                  ; C202011A
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v0, v4, s4, v0                              ; D2820000 04000904
>	s_buffer_load_dword s0, s[0:3], 0x1e                  ; C200011E
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v0, s0, v0                              ; 06000000
>	exp 15, 12, 0, 1, 0, v3, v2, v0, v1                   ; F80008CF 01000203
>	s_endpgm                                              ; BF810000
>
>SGPRS: 24
>VGPRS: 12
>Code Size: 780 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL IN[1], GENERIC[10], PERSPECTIVE
>DCL IN[2], GENERIC[11], PERSPECTIVE
>DCL IN[3], GENERIC[12], PERSPECTIVE
>DCL IN[4], GENERIC[13], PERSPECTIVE
>DCL IN[5], GENERIC[14], PERSPECTIVE
>DCL IN[6], GENERIC[15], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SAMP[1]
>DCL SAMP[2]
>DCL SAMP[3]
>DCL SAMP[4]
>DCL SAMP[5]
>DCL SAMP[6]
>DCL SAMP[7]
>DCL SAMP[8]
>DCL SAMP[9]
>DCL SAMP[10]
>DCL SAMP[11]
>DCL CONST[0..14]
>DCL TEMP[0..24], LOCAL
>IMM[0] FLT32 {    1.0000,    -1.0000,     0.0000,     2.2000}
>IMM[1] FLT32 {    0.3330,     2.0000,    -2.0000,     0.0000}
>IMM[2] FLT32 {    0.5000,    -0.0000,   -16.0000,    -1.4427}
>IMM[3] FLT32 {   16.0000,     0.0050,     0.5000,    -0.5000}
>IMM[4] FLT32 {    6.0000,     2.0000,    -2.0000,     0.2500}
>IMM[5] FLT32 {    0.8000,     0.4545,     0.4000,     1.4427}
>  0: MAD TEMP[0].xy, IN[6], IMM[0].xyxx, IMM[0].zxzz
>  1: MOV TEMP[0].xy, TEMP[0].xyyy
>  2: TEX TEMP[0], TEMP[0], SAMP[9], 2D
>  3: MOV TEMP[1].z, TEMP[0]
>  4: ABS TEMP[2].x, TEMP[0]
>  5: MOV TEMP[2], -TEMP[2].xxxx
>  6: FSGE TEMP[3].x, TEMP[2].xxxx, IMM[0].zzzz
>  7: UIF TEMP[3].xxxx :0
>  8:   MOV TEMP[3].x, IMM[0].yyyy
>  9: ELSE :0
> 10:   MOV TEMP[3].x, IMM[0].zzzz
> 11: ENDIF
> 12: MOV TEMP[3].x, TEMP[3].xxxx
> 13: FSGE TEMP[4].x, TEMP[2].yyyy, IMM[0].zzzz
> 14: UIF TEMP[4].xxxx :0
> 15:   MOV TEMP[4].x, IMM[0].yyyy
> 16: ELSE :0
> 17:   MOV TEMP[4].x, IMM[0].zzzz
> 18: ENDIF
> 19: MOV TEMP[3].y, TEMP[4].xxxx
> 20: FSGE TEMP[4].x, TEMP[2].zzzz, IMM[0].zzzz
> 21: UIF TEMP[4].xxxx :0
> 22:   MOV TEMP[4].x, IMM[0].yyyy
> 23: ELSE :0
> 24:   MOV TEMP[4].x, IMM[0].zzzz
> 25: ENDIF
> 26: MOV TEMP[3].z, TEMP[4].xxxx
> 27: FSGE TEMP[2].x, TEMP[2].wwww, IMM[0].zzzz
> 28: UIF TEMP[2].xxxx :0
> 29: ELSE :0
> 30: ENDIF
> 31: FSLT TEMP[3].xyz, TEMP[3].xyzz, IMM[0].zzzz
> 32: OR TEMP[4].x, TEMP[3].xxxx, TEMP[3].zzzz
> 33: OR TEMP[3].x, TEMP[4].xxxx, TEMP[3].yyyy
> 34: AND TEMP[3].x, TEMP[3].xxxx, IMM[0].xxxx
> 35: KILL_IF -TEMP[3].xxxx
> 36: MOV TEMP[3].xy, IN[3].xyyy
> 37: TEX TEMP[3].xyz, TEMP[3], SAMP[6], 2D
> 38: POW TEMP[4].x, TEMP[3].xxxx, IMM[0].wwww
> 39: POW TEMP[4].y, TEMP[3].yyyy, IMM[0].wwww
> 40: POW TEMP[4].z, TEMP[3].zzzz, IMM[0].wwww
> 41: MOV TEMP[2].xyz, TEMP[4].xyzx
> 42: RCP TEMP[3].x, CONST[0].xxxx
> 43: MUL TEMP[3].y, TEMP[3].xxxx, IMM[1].xxxx
> 44: MUL TEMP[5].w, TEMP[3].yyyy, TEMP[3].yyyy
> 45: MOV TEMP[6].xy, IN[0].xyyy
> 46: TEX TEMP[6], TEMP[6], SAMP[0], 2D
> 47: MOV TEMP[7].w, TEMP[6].wwzw
> 48: MOV TEMP[8].xy, IN[0].xyyy
> 49: TEX TEMP[8].x, TEMP[8], SAMP[1], 2D
> 50: MAD TEMP[9].xy, TEMP[6], IMM[1].yyyy, IMM[0].yyyy
> 51: MOV TEMP[7].xy, TEMP[9].xyxx
> 52: MUL TEMP[10].yz, TEMP[3].yyyy, TEMP[7].xxyw
> 53: MOV TEMP[9].z, TEMP[10].zyzz
> 54: MAD TEMP[8].z, TEMP[8].xxxx, IMM[1].yyyy, IMM[0].yyyy
> 55: MOV TEMP[8].z, TEMP[8].zzzz
> 56: MOV TEMP[8].xy, TEMP[6].zwzz
> 57: MOV TEMP[6].xy, IN[1].xyyy
> 58: TEX TEMP[6], TEMP[6], SAMP[2], 2D
> 59: MOV TEMP[11].w, TEMP[6].wwww
> 60: MOV TEMP[12].xy, IN[1].xyyy
> 61: TEX TEMP[12].xw, TEMP[12], SAMP[3], 2D
> 62: MOV TEMP[13].w, TEMP[12].wwww
> 63: ADD TEMP[14].zw, TEMP[6].xyxy, TEMP[6].xyxy
> 64: MOV TEMP[7].zw, TEMP[14].wwzw
> 65: MAD TEMP[14].zw, TEMP[7], IMM[0].yxyx, IMM[0].yxxy
> 66: MOV TEMP[7].zw, TEMP[14].wwzw
> 67: MUL TEMP[14].zw, TEMP[3].yyyy, TEMP[7]
> 68: MOV TEMP[7].zw, TEMP[14].wwzw
> 69: MAD TEMP[15].xy, TEMP[7], TEMP[3].yyyy, TEMP[14].zwzw
> 70: MOV TEMP[7].xy, TEMP[15].xyxx
> 71: MAD TEMP[12].z, TEMP[12].xxxx, IMM[1].zzzz, IMM[0].xxxx
> 72: MOV TEMP[13].z, TEMP[12].zzzz
> 73: MOV TEMP[13].xy, TEMP[6].zwzz
> 74: MUL TEMP[6].xyz, TEMP[5].wwww, TEMP[13]
> 75: MOV TEMP[11].xyz, TEMP[6].xyzx
> 76: MAD TEMP[6].xyz, TEMP[8], TEMP[5].wwww, TEMP[11]
> 77: MOV TEMP[8].xyz, TEMP[6].xyzx
> 78: ADD TEMP[6].xw, TEMP[10].yyzz, TEMP[10].yyzz
> 79: MOV TEMP[9].xw, TEMP[6].xxxw
> 80: MUL TEMP[6].xy, TEMP[14].zwzw, TEMP[9].xwzw
> 81: MOV TEMP[11].xy, TEMP[6].xyxx
> 82: DP2 TEMP[6].x, TEMP[10].yzzz, TEMP[14].wzzz
> 83: MOV TEMP[11].z, TEMP[6].xxxx
> 84: ADD TEMP[6].xyz, TEMP[8], TEMP[11]
> 85: MOV TEMP[8].xyz, TEMP[6].xyzx
> 86: MOV TEMP[6].xy, IN[2].xyyy
> 87: TEX TEMP[6], TEMP[6], SAMP[2], 2D
> 88: MOV TEMP[11].zw, TEMP[6].wwzw
> 89: MOV TEMP[12].xy, IN[2].xyyy
> 90: TEX TEMP[12].xw, TEMP[12], SAMP[3], 2D
> 91: MOV TEMP[13].xw, TEMP[12].xxxw
> 92: MAD TEMP[15].xy, TEMP[6].yxzw, IMM[1].yyyy, IMM[0].yyyy
> 93: MOV TEMP[11].xy, TEMP[15].xyxx
> 94: MUL TEMP[15].yz, TEMP[3].yyyy, TEMP[11].xxyw
> 95: MOV TEMP[13].yz, TEMP[15].zyzz
> 96: MAD TEMP[3].xy, TEMP[11].yxzw, TEMP[3].yyyy, TEMP[7]
> 97: MOV TEMP[3].xy, TEMP[3].xyxx
> 98: MAD TEMP[12].z, TEMP[12].xxxx, IMM[1].yyyy, IMM[0].yyyy
> 99: MOV TEMP[12].z, TEMP[12].zzzz
>100: MOV TEMP[12].xy, TEMP[6].zwzz
>101: MAD TEMP[5].xyz, TEMP[12], TEMP[5].wwww, TEMP[8]
>102: MOV TEMP[8].xyz, TEMP[5].xyzx
>103: MUL TEMP[5].xy, TEMP[9].xwzw, TEMP[13].zyzw
>104: MOV TEMP[11].xy, TEMP[5].xyxx
>105: DP2 TEMP[5].x, TEMP[10].yzzz, TEMP[15].yzzz
>106: MOV TEMP[11].z, TEMP[5].xxxx
>107: ADD TEMP[5].xyz, TEMP[8], TEMP[11]
>108: MOV TEMP[9].xyz, TEMP[5].xyzx
>109: MUL TEMP[5].yw, TEMP[7].xzzw, TEMP[13].xzzy
>110: MOV TEMP[1].yw, TEMP[5].wyww
>111: ADD TEMP[5].xy, TEMP[1].ywzw, TEMP[1].ywzw
>112: MOV TEMP[8].xy, TEMP[5].xyxx
>113: DP2 TEMP[5].x, TEMP[14].zwww, TEMP[15].yzzz
>114: MOV TEMP[8].z, TEMP[5].xxxx
>115: ADD TEMP[5].xyz, TEMP[9], TEMP[8]
>116: MOV TEMP[7].xyz, TEMP[5].xyzx
>117: MAD TEMP[5].xyz, TEMP[3].xyxw, -TEMP[3].xyyw, TEMP[7]
>118: MOV TEMP[7].xy, TEMP[5].xyzx
>119: MOV TEMP[3].z, IMM[0].xxxx
>120: DP3 TEMP[6].x, CONST[1].xyzz, TEMP[3].xyzz
>121: MOV_SAT TEMP[6].x, TEMP[6].xxxx
>122: ADD TEMP[10].xyz, CONST[8], -IN[4]
>123: MOV TEMP[9].xyz, TEMP[10].xyzx
>124: DP3 TEMP[10].x, TEMP[10].xyzz, TEMP[10].xyzz
>125: MAX TEMP[10].x, TEMP[10].xxxx, IMM[1].wwww
>126: RSQ TEMP[10].x, TEMP[10].xxxx
>127: MUL TEMP[14].xyz, TEMP[10].xxxx, TEMP[9]
>128: MAD TEMP[15].xyz, TEMP[9], TEMP[10].xxxx, -CONST[5]
>129: MOV TEMP[11].xyz, TEMP[15].xyzx
>130: MAD TEMP[15].xyz, CONST[5].wwww, TEMP[11], CONST[5]
>131: MOV TEMP[11].xyz, TEMP[15].xyzx
>132: DP3 TEMP[15].x, TEMP[15].xyzz, TEMP[15].xyzz
>133: MAX TEMP[15].x, TEMP[15].xxxx, IMM[1].wwww
>134: RSQ TEMP[15].x, TEMP[15].xxxx
>135: MAD TEMP[15].xyz, TEMP[11], TEMP[15].xxxx, CONST[1]
>136: MOV TEMP[11].xyz, TEMP[15].xyzx
>137: RCP TEMP[15].x, TEMP[15].zzzz
>138: MAD TEMP[15].xy, TEMP[11], TEMP[15].xxxx, -TEMP[3]
>139: RCP TEMP[16].x, CONST[3].wwww
>140: ADD TEMP[17].zw, TEMP[16].xxxx, TEMP[5].xyxy
>141: MUL TEMP[18].w, TEMP[5].zzzz, TEMP[5].zzzz
>142: MOV TEMP[7].w, TEMP[18].wwww
>143: MAD TEMP[19].w, TEMP[17].zzzz, TEMP[17].wwww, -TEMP[18].wwww
>144: MUL TEMP[20].w, TEMP[15].xxxx, TEMP[15].xxxx
>145: ADD TEMP[5].z, TEMP[5].zzzz, TEMP[5].zzzz
>146: MOV TEMP[7].z, TEMP[5].zzzz
>147: MUL TEMP[21].x, TEMP[15].xxxx, TEMP[5].zzzz
>148: MAD TEMP[22].z, TEMP[17].zzzz, TEMP[15].yyyy, -TEMP[21].xxxx
>149: MUL TEMP[22].z, TEMP[15].yyyy, TEMP[22].zzzz
>150: MAD TEMP[17].z, TEMP[20].wwww, TEMP[17].wwww, TEMP[22].zzzz
>151: MUL TEMP[17].z, TEMP[17].zzzz, IMM[2].xxxx
>152: RCP TEMP[22].x, TEMP[19].wwww
>153: MUL TEMP[23].x, TEMP[22].xxxx, TEMP[17].zzzz
>154: MOV TEMP[24].x, -TEMP[19].wwww
>155: FSGE TEMP[24].x, TEMP[24].xxxx, IMM[0].zzzz
>156: UIF TEMP[24].xxxx :0
>157:   MOV TEMP[24].x, IMM[0].xxxx
>158: ELSE :0
>159:   MOV TEMP[24].x, IMM[2].yyyy
>160: ENDIF
>161: MAD TEMP[17].z, TEMP[17].zzzz, TEMP[22].xxxx, IMM[2].zzzz
>162: FSGE TEMP[17].x, TEMP[17].zzzz, IMM[0].zzzz
>163: UIF TEMP[17].xxxx :0
>164:   MOV TEMP[17].x, IMM[0].xxxx
>165: ELSE :0
>166:   MOV TEMP[17].x, IMM[2].yyyy
>167: ENDIF
>168: ADD TEMP[17].z, TEMP[17].xxxx, TEMP[24].xxxx
>169: MUL TEMP[22].w, TEMP[23].xxxx, IMM[2].wwww
>170: EX2 TEMP[22].x, TEMP[22].wwww
>171: MAX TEMP[19].x, TEMP[19].wwww, IMM[1].wwww
>172: RSQ TEMP[19].x, TEMP[19].xxxx
>173: MUL TEMP[19].w, TEMP[19].xxxx, TEMP[22].xxxx
>174: MOV TEMP[9].w, TEMP[19].wwww
>175: MAD TEMP[16].xy, TEMP[16].xxxx, IMM[3].xxxx, TEMP[7]
>176: MAD TEMP[22].w, TEMP[16].xxxx, TEMP[16].yyyy, -TEMP[18].wwww
>177: MAD TEMP[21].x, TEMP[16].xxxx, TEMP[15].yyyy, -TEMP[21].xxxx
>178: MUL TEMP[15].x, TEMP[15].yyyy, TEMP[21].xxxx
>179: MAD TEMP[15].w, TEMP[20].wwww, TEMP[16].yyyy, TEMP[15].xxxx
>180: MUL TEMP[15].w, TEMP[15].wwww, IMM[2].xxxx
>181: RCP TEMP[11].x, TEMP[22].wwww
>182: MUL TEMP[16].y, TEMP[15].wwww, TEMP[11].xxxx
>183: MOV TEMP[20].x, -TEMP[22].wwww
>184: FSGE TEMP[20].x, TEMP[20].xxxx, IMM[0].zzzz
>185: UIF TEMP[20].xxxx :0
>186:   MOV TEMP[20].x, IMM[0].xxxx
>187: ELSE :0
>188:   MOV TEMP[20].x, IMM[2].yyyy
>189: ENDIF
>190: MAD TEMP[11].w, TEMP[15].wwww, TEMP[11].xxxx, IMM[2].zzzz
>191: FSGE TEMP[11].x, TEMP[11].wwww, IMM[0].zzzz
>192: UIF TEMP[11].xxxx :0
>193:   MOV TEMP[11].x, IMM[0].xxxx
>194: ELSE :0
>195:   MOV TEMP[11].x, IMM[2].yyyy
>196: ENDIF
>197: ADD TEMP[11].w, TEMP[11].xxxx, TEMP[20].xxxx
>198: MOV TEMP[8].w, TEMP[11].wwww
>199: MUL TEMP[15].x, TEMP[16].yyyy, IMM[2].wwww
>200: MAX TEMP[16].x, TEMP[22].wwww, IMM[1].wwww
>201: RSQ TEMP[16].x, TEMP[16].xxxx
>202: EX2 TEMP[15].x, TEMP[15].xxxx
>203: MUL TEMP[15].w, TEMP[16].xxxx, TEMP[15].xxxx
>204: MUL TEMP[15].w, TEMP[15].wwww, IMM[3].yyyy
>205: MOV TEMP[2].w, TEMP[15].wwww
>206: MAD TEMP[10].xyz, TEMP[9], TEMP[10].xxxx, -CONST[6]
>207: MOV TEMP[9].xyz, TEMP[10].xyzx
>208: MAD TEMP[10].xyz, CONST[6].wwww, TEMP[9], CONST[6]
>209: MOV TEMP[9].xyz, TEMP[10].xyzx
>210: DP3 TEMP[10].x, TEMP[10].xyzz, TEMP[10].xyzz
>211: MAX TEMP[10].x, TEMP[10].xxxx, IMM[1].wwww
>212: RSQ TEMP[10].x, TEMP[10].xxxx
>213: MAD TEMP[10].xyz, TEMP[9], TEMP[10].xxxx, CONST[2]
>214: MOV TEMP[9].xyz, TEMP[10].xyzx
>215: RCP TEMP[10].x, TEMP[10].zzzz
>216: MAD TEMP[10].xy, TEMP[9], TEMP[10].xxxx, -TEMP[3]
>217: RCP TEMP[16].x, CONST[4].wwww
>218: ADD TEMP[16].xy, TEMP[16].xxxx, TEMP[7]
>219: MAD TEMP[18].w, TEMP[16].xxxx, TEMP[16].yyyy, -TEMP[18].wwww
>220: MUL TEMP[20].w, TEMP[10].xxxx, TEMP[10].xxxx
>221: MUL TEMP[5].z, TEMP[5].zzzz, TEMP[10].xxxx
>222: MAD TEMP[5].z, TEMP[16].xxxx, TEMP[10].yyyy, -TEMP[5].zzzz
>223: MUL TEMP[5].z, TEMP[10].yyyy, TEMP[5].zzzz
>224: MAD TEMP[5].z, TEMP[20].wwww, TEMP[16].yyyy, TEMP[5].zzzz
>225: MUL TEMP[5].z, TEMP[5].zzzz, IMM[2].xxxx
>226: RCP TEMP[10].x, TEMP[18].wwww
>227: MUL TEMP[16].x, TEMP[10].xxxx, TEMP[5].zzzz
>228: MOV TEMP[20].x, -TEMP[18].wwww
>229: FSGE TEMP[20].x, TEMP[20].xxxx, IMM[0].zzzz
>230: UIF TEMP[20].xxxx :0
>231:   MOV TEMP[20].x, IMM[0].xxxx
>232: ELSE :0
>233:   MOV TEMP[20].x, IMM[2].yyyy
>234: ENDIF
>235: MAD TEMP[5].z, TEMP[5].zzzz, TEMP[10].xxxx, IMM[2].zzzz
>236: FSGE TEMP[5].x, TEMP[5].zzzz, IMM[0].zzzz
>237: UIF TEMP[5].xxxx :0
>238:   MOV TEMP[5].x, IMM[0].xxxx
>239: ELSE :0
>240:   MOV TEMP[5].x, IMM[2].yyyy
>241: ENDIF
>242: ADD TEMP[5].z, TEMP[5].xxxx, TEMP[20].xxxx
>243: MOV TEMP[7].z, TEMP[5].zzzz
>244: MUL TEMP[10].w, TEMP[16].xxxx, IMM[2].wwww
>245: EX2 TEMP[10].x, TEMP[10].wwww
>246: MAX TEMP[16].x, TEMP[18].wwww, IMM[1].wwww
>247: RSQ TEMP[16].x, TEMP[16].xxxx
>248: MUL TEMP[10].w, TEMP[10].xxxx, TEMP[16].xxxx
>249: MOV TEMP[7].w, TEMP[10].wwww
>250: MUL TEMP[16].xyz, CONST[10].xyww, IN[4].yyyy
>251: MOV TEMP[9].xyz, TEMP[16].xyzx
>252: MAD TEMP[16].xyz, IN[4].xxxx, CONST[9].xyww, TEMP[9]
>253: MOV TEMP[9].xyz, TEMP[16].xyzx
>254: MAD TEMP[16].xyz, IN[4].zzzz, CONST[11].xyww, TEMP[9]
>255: MOV TEMP[9].xyz, TEMP[16].xyzx
>256: ADD TEMP[16].xyz, TEMP[9], CONST[12].xyww
>257: MOV TEMP[9].xy, TEMP[16].xyzx
>258: RCP TEMP[16].x, TEMP[16].zzzz
>259: MOV TEMP[9].z, TEMP[16].xxxx
>260: MUL TEMP[16].xy, TEMP[16].xxxx, TEMP[9]
>261: MOV TEMP[9].xy, TEMP[16].xyxx
>262: MAD TEMP[16].xy, TEMP[9], IMM[3].zwzw, IMM[3].wwww
>263: MOV TEMP[9].xy, TEMP[16].xyxx
>264: MAD TEMP[16].xy, TEMP[7], IMM[4].xxxx, TEMP[9]
>265: MOV TEMP[7].xy, TEMP[16].xyxx
>266: MAD TEMP[16].xy, TEMP[18].wwww, IMM[1].yyyy, TEMP[7]
>267: MOV TEMP[16].xy, TEMP[16].xyyy
>268: TEX TEMP[16], TEMP[16], SAMP[4], 2D
>269: MAD TEMP[20].xy, TEMP[18].wwww, IMM[4].yzzw, TEMP[7]
>270: MOV TEMP[20].xy, TEMP[20].xyyy
>271: TEX TEMP[20], TEMP[20], SAMP[4], 2D
>272: MUL TEMP[12], TEMP[20], IMM[4].wwww
>273: MAD TEMP[13], TEMP[16], IMM[4].wwww, TEMP[12]
>274: MAD TEMP[12].xy, TEMP[18].wwww, IMM[4].zxzw, TEMP[7]
>275: MOV TEMP[12].xy, TEMP[12].xyyy
>276: TEX TEMP[12], TEMP[12], SAMP[4], 2D
>277: MAD TEMP[13], TEMP[12], IMM[4].wwww, TEMP[13]
>278: MAD TEMP[12].xy, TEMP[18].wwww, IMM[1].zzzz, TEMP[7]
>279: MOV TEMP[12].xy, TEMP[12].xyyy
>280: TEX TEMP[12], TEMP[12], SAMP[4], 2D
>281: MAD TEMP[13], TEMP[12], IMM[4].wwww, TEMP[13]
>282: ABS TEMP[12].x, TEMP[13].xxxx
>283: LG2 TEMP[9].x, TEMP[12].xxxx
>284: ABS TEMP[12].x, TEMP[13].yyyy
>285: LG2 TEMP[12].x, TEMP[12].xxxx
>286: MOV TEMP[9].y, TEMP[12].xxxx
>287: ABS TEMP[12].x, TEMP[13].zzzz
>288: LG2 TEMP[12].x, TEMP[12].xxxx
>289: MOV TEMP[9].z, TEMP[12].xxxx
>290: MUL TEMP[12].xyz, TEMP[9], IMM[0].wwww
>291: EX2 TEMP[13].x, TEMP[12].xxxx
>292: EX2 TEMP[16].x, TEMP[12].yyyy
>293: MOV TEMP[13].y, TEMP[16].xxxx
>294: EX2 TEMP[12].x, TEMP[12].zzzz
>295: MOV TEMP[13].z, TEMP[12].xxxx
>296: DP3 TEMP[3].x, TEMP[3].xyzz, TEMP[14].xyzz
>297: ADD TEMP[3].w, -TEMP[3].xxxx, IMM[0].xxxx
>298: MUL TEMP[12].x, TEMP[3].wwww, TEMP[3].wwww
>299: MUL TEMP[12].x, TEMP[12].xxxx, TEMP[12].xxxx
>300: MUL TEMP[14].y, TEMP[3].wwww, TEMP[12].xxxx
>301: MUL TEMP[16].xyz, TEMP[19].wwww, CONST[3]
>302: MOV TEMP[17], -TEMP[17].zzzz
>303: FSGE TEMP[18].x, TEMP[17].xxxx, IMM[0].zzzz
>304: UIF TEMP[18].xxxx :0
>305:   MOV TEMP[18].x, TEMP[16].xxxx
>306: ELSE :0
>307:   MOV TEMP[18].x, IMM[2].yyyy
>308: ENDIF
>309: MOV TEMP[18].x, TEMP[18].xxxx
>310: FSGE TEMP[19].x, TEMP[17].yyyy, IMM[0].zzzz
>311: UIF TEMP[19].xxxx :0
>312:   MOV TEMP[19].x, TEMP[16].yyyy
>313: ELSE :0
>314:   MOV TEMP[19].x, IMM[2].yyyy
>315: ENDIF
>316: MOV TEMP[18].y, TEMP[19].xxxx
>317: FSGE TEMP[19].x, TEMP[17].zzzz, IMM[0].zzzz
>318: UIF TEMP[19].xxxx :0
>319:   MOV TEMP[16].x, TEMP[16].zzzz
>320: ELSE :0
>321:   MOV TEMP[16].x, IMM[2].yyyy
>322: ENDIF
>323: MOV TEMP[18].z, TEMP[16].xxxx
>324: FSGE TEMP[16].x, TEMP[17].wwww, IMM[0].zzzz
>325: UIF TEMP[16].xxxx :0
>326: ELSE :0
>327: ENDIF
>328: MOV TEMP[9].xyz, TEMP[18].xyzx
>329: MUL TEMP[10].xyz, TEMP[10].wwww, CONST[4]
>330: MOV TEMP[8].xyz, TEMP[10].xyzx
>331: MUL TEMP[10].xyz, TEMP[8], IMM[5].xxxx
>332: MOV TEMP[5], -TEMP[5].zzzz
>333: FSGE TEMP[16].x, TEMP[5].xxxx, IMM[0].zzzz
>334: UIF TEMP[16].xxxx :0
>335:   MOV TEMP[16].x, TEMP[10].xxxx
>336: ELSE :0
>337:   MOV TEMP[16].x, IMM[2].yyyy
>338: ENDIF
>339: MOV TEMP[16].x, TEMP[16].xxxx
>340: FSGE TEMP[17].x, TEMP[5].yyyy, IMM[0].zzzz
>341: UIF TEMP[17].xxxx :0
>342:   MOV TEMP[17].x, TEMP[10].yyyy
>343: ELSE :0
>344:   MOV TEMP[17].x, IMM[2].yyyy
>345: ENDIF
>346: MOV TEMP[16].y, TEMP[17].xxxx
>347: FSGE TEMP[17].x, TEMP[5].zzzz, IMM[0].zzzz
>348: UIF TEMP[17].xxxx :0
>349:   MOV TEMP[10].x, TEMP[10].zzzz
>350: ELSE :0
>351:   MOV TEMP[10].x, IMM[2].yyyy
>352: ENDIF
>353: MOV TEMP[16].z, TEMP[10].xxxx
>354: FSGE TEMP[5].x, TEMP[5].wwww, IMM[0].zzzz
>355: UIF TEMP[5].xxxx :0
>356: ELSE :0
>357: ENDIF
>358: MOV TEMP[8].xyz, TEMP[16].xyzx
>359: ADD TEMP[5].xyz, TEMP[9], TEMP[8]
>360: MAD TEMP[4].yzw, TEMP[5].xxyz, TEMP[14].yyyy, TEMP[4].xxyz
>361: MOV TEMP[7].w, TEMP[4].zyzw
>362: MUL TEMP[5].xyz, TEMP[2], CONST[3]
>363: MOV TEMP[2].xyz, TEMP[5].xyzx
>364: MUL TEMP[5].xyz, TEMP[15].wwww, TEMP[2]
>365: MOV TEMP[8], -TEMP[11].wwww
>366: FSGE TEMP[10].x, TEMP[8].xxxx, IMM[0].zzzz
>367: UIF TEMP[10].xxxx :0
>368:   MOV TEMP[10].x, TEMP[5].xxxx
>369: ELSE :0
>370:   MOV TEMP[10].x, IMM[2].yyyy
>371: ENDIF
>372: MOV TEMP[10].x, TEMP[10].xxxx
>373: FSGE TEMP[11].x, TEMP[8].yyyy, IMM[0].zzzz
>374: UIF TEMP[11].xxxx :0
>375:   MOV TEMP[11].x, TEMP[5].yyyy
>376: ELSE :0
>377:   MOV TEMP[11].x, IMM[2].yyyy
>378: ENDIF
>379: MOV TEMP[10].y, TEMP[11].xxxx
>380: FSGE TEMP[11].x, TEMP[8].zzzz, IMM[0].zzzz
>381: UIF TEMP[11].xxxx :0
>382:   MOV TEMP[5].x, TEMP[5].zzzz
>383: ELSE :0
>384:   MOV TEMP[5].x, IMM[2].yyyy
>385: ENDIF
>386: MOV TEMP[10].z, TEMP[5].xxxx
>387: FSGE TEMP[5].x, TEMP[8].wwww, IMM[0].zzzz
>388: UIF TEMP[5].xxxx :0
>389: ELSE :0
>390: ENDIF
>391: MOV TEMP[2].xyz, TEMP[10].xyzx
>392: MAD TEMP[4].xyz, TEMP[4].yzww, TEMP[6].xxxx, TEMP[2]
>393: MOV TEMP[2].xyz, TEMP[4].xyzx
>394: MAD TEMP[4].xyz, TEMP[13], CONST[14].xxxx, TEMP[2]
>395: MOV TEMP[2].xyz, TEMP[4].xyzx
>396: MOV TEMP[4].xy, IN[3].xyyy
>397: TEX TEMP[4].w, TEMP[4], SAMP[5], 2D
>398: MOV TEMP[9].w, TEMP[4].wwww
>399: MAD TEMP[3].y, TEMP[3].wwww, TEMP[12].xxxx, IMM[2].xxxx
>400: MUL TEMP[3].xyz, TEMP[3].yyyy, TEMP[2]
>401: ADD TEMP[5], TEMP[13].wwww, IMM[0].xxxx
>402: MOV_SAT TEMP[5], TEMP[5]
>403: MUL TEMP[4].w, TEMP[4].wwww, TEMP[5].yyyy
>404: MOV TEMP[4].w, TEMP[4].wwww
>405: ABS TEMP[5].x, TEMP[3].xxxx
>406: LG2 TEMP[7].x, TEMP[5].xxxx
>407: ABS TEMP[5].x, TEMP[3].yyyy
>408: LG2 TEMP[5].x, TEMP[5].xxxx
>409: MOV TEMP[7].y, TEMP[5].xxxx
>410: ABS TEMP[3].x, TEMP[3].zzzz
>411: LG2 TEMP[3].x, TEMP[3].xxxx
>412: MOV TEMP[7].z, TEMP[3].xxxx
>413: MUL TEMP[3].xyz, TEMP[7], IMM[5].yyyy
>414: EX2 TEMP[7].x, TEMP[3].xxxx
>415: EX2 TEMP[5].x, TEMP[3].yyyy
>416: MOV TEMP[7].y, TEMP[5].xxxx
>417: EX2 TEMP[3].x, TEMP[3].zzzz
>418: MOV TEMP[7].z, TEMP[3].xxxx
>419: MOV TEMP[3].xyz, TEMP[7].xyzz
>420: TEX TEMP[3], TEMP[3], SAMP[11], 3D
>421: MOV TEMP[2].w, TEMP[3].wwww
>422: MAD TEMP[5].xy, IN[6], IMM[0].xyxx, IMM[0].zxzz
>423: MOV TEMP[5].xy, TEMP[5].xyyy
>424: TEX TEMP[5], TEMP[5], SAMP[8], 2D
>425: MOV TEMP[7].w, TEMP[5].wwww
>426: ADD TEMP[6].y, -TEMP[5].wwww, IMM[0].xxxx
>427: MAD TEMP[3].xyz, TEMP[3], TEMP[6].yyyy, TEMP[5]
>428: MOV TEMP[2].xyz, TEMP[3].xyzx
>429: LRP TEMP[2].xyz, TEMP[0].xxxx, TEMP[2], IMM[5].zzzz
>430: MOV TEMP[7].xyz, TEMP[2].xyzx
>431: MOV TEMP[2].xyz, TEMP[2].xyzz
>432: TEX TEMP[2], TEMP[2], SAMP[10], 3D
>433: MAD TEMP[0].x, TEMP[0].zzzz, -TEMP[0].xxxx, TEMP[0].xxxx
>434: LRP TEMP[0].xyz, TEMP[0].xxxx, TEMP[2], TEMP[7]
>435: MOV TEMP[9].xyz, TEMP[0].xyzx
>436: RCP TEMP[1].x, IN[5].zzzz
>437: MUL TEMP[2].x, TEMP[1].xxxx, IN[5].yyyy
>438: MAD TEMP[2].x, TEMP[2].xxxx, IMM[2].xxxx, IMM[2].xxxx
>439: MOV TEMP[1].x, TEMP[2].xxxx
>440: MOV TEMP[1].y, CONST[7].wwww
>441: MOV TEMP[2].xy, TEMP[1].xyyy
>442: TEX TEMP[2].x, TEMP[2], SAMP[7], 2D
>443: MOV TEMP[1].x, TEMP[2].xxxx
>444: ADD TEMP[3].yzw, -CONST[8].xxyz, IN[4].xxyz
>445: MOV TEMP[1].w, TEMP[3].zyzw
>446: DP3 TEMP[5].x, TEMP[3].yzww, TEMP[3].yzww
>447: MOV TEMP[1].y, TEMP[5].xxxx
>448: MUL TEMP[1].yz, TEMP[1].xyww, CONST[7].xyxw
>449: MUL TEMP[5].z, TEMP[1].zzzz, IMM[5].wwww
>450: EX2 TEMP[5].x, TEMP[5].zzzz
>451: ADD TEMP[5].z, -TEMP[5].xxxx, IMM[0].xxxx
>452: MUL TEMP[1].y, TEMP[5].zzzz, TEMP[1].yyyy
>453: RCP TEMP[3].x, TEMP[3].wwww
>454: MUL TEMP[1].y, TEMP[3].xxxx, TEMP[1].yyyy
>455: MUL TEMP[1].y, TEMP[1].yyyy, IMM[5].wwww
>456: EX2 TEMP[1].x, TEMP[1].yyyy
>457: MOV_SAT TEMP[1].x, TEMP[1].xxxx
>458: ADD TEMP[1].y, -TEMP[1].xxxx, IMM[0].xxxx
>459: MUL TEMP[1].x, TEMP[1].yyyy, TEMP[2].xxxx
>460: ADD TEMP[0].yzw, -TEMP[0].xxyz, CONST[13].xxyz
>461: MAD TEMP[0].xyz, TEMP[1].xxxx, TEMP[0].yzww, TEMP[9]
>462: MOV TEMP[4].xyz, TEMP[0].xyzx
>463: MOV OUT[0], TEMP[4]
>464: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %23 = load <16 x i8> addrspace(2)* %22, !tbaa !0
>  %24 = call float @llvm.SI.load.const(<16 x i8> %23, i32 0)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %23, i32 16)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %23, i32 20)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %23, i32 24)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %23, i32 32)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %23, i32 36)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %23, i32 40)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %23, i32 48)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %23, i32 52)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %23, i32 56)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %23, i32 60)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %23, i32 64)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %23, i32 68)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %23, i32 72)
>  %38 = call float @llvm.SI.load.const(<16 x i8> %23, i32 76)
>  %39 = call float @llvm.SI.load.const(<16 x i8> %23, i32 80)
>  %40 = call float @llvm.SI.load.const(<16 x i8> %23, i32 84)
>  %41 = call float @llvm.SI.load.const(<16 x i8> %23, i32 88)
>  %42 = call float @llvm.SI.load.const(<16 x i8> %23, i32 92)
>  %43 = call float @llvm.SI.load.const(<16 x i8> %23, i32 96)
>  %44 = call float @llvm.SI.load.const(<16 x i8> %23, i32 100)
>  %45 = call float @llvm.SI.load.const(<16 x i8> %23, i32 104)
>  %46 = call float @llvm.SI.load.const(<16 x i8> %23, i32 108)
>  %47 = call float @llvm.SI.load.const(<16 x i8> %23, i32 112)
>  %48 = call float @llvm.SI.load.const(<16 x i8> %23, i32 116)
>  %49 = call float @llvm.SI.load.const(<16 x i8> %23, i32 124)
>  %50 = call float @llvm.SI.load.const(<16 x i8> %23, i32 128)
>  %51 = call float @llvm.SI.load.const(<16 x i8> %23, i32 132)
>  %52 = call float @llvm.SI.load.const(<16 x i8> %23, i32 136)
>  %53 = call float @llvm.SI.load.const(<16 x i8> %23, i32 144)
>  %54 = call float @llvm.SI.load.const(<16 x i8> %23, i32 148)
>  %55 = call float @llvm.SI.load.const(<16 x i8> %23, i32 156)
>  %56 = call float @llvm.SI.load.const(<16 x i8> %23, i32 160)
>  %57 = call float @llvm.SI.load.const(<16 x i8> %23, i32 164)
>  %58 = call float @llvm.SI.load.const(<16 x i8> %23, i32 172)
>  %59 = call float @llvm.SI.load.const(<16 x i8> %23, i32 176)
>  %60 = call float @llvm.SI.load.const(<16 x i8> %23, i32 180)
>  %61 = call float @llvm.SI.load.const(<16 x i8> %23, i32 188)
>  %62 = call float @llvm.SI.load.const(<16 x i8> %23, i32 192)
>  %63 = call float @llvm.SI.load.const(<16 x i8> %23, i32 196)
>  %64 = call float @llvm.SI.load.const(<16 x i8> %23, i32 204)
>  %65 = call float @llvm.SI.load.const(<16 x i8> %23, i32 208)
>  %66 = call float @llvm.SI.load.const(<16 x i8> %23, i32 212)
>  %67 = call float @llvm.SI.load.const(<16 x i8> %23, i32 216)
>  %68 = call float @llvm.SI.load.const(<16 x i8> %23, i32 224)
>  %69 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %70 = load <8 x i32> addrspace(2)* %69, !tbaa !0
>  %71 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %72 = load <4 x i32> addrspace(2)* %71, !tbaa !0
>  %73 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 1
>  %74 = load <8 x i32> addrspace(2)* %73, !tbaa !0
>  %75 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 1
>  %76 = load <4 x i32> addrspace(2)* %75, !tbaa !0
>  %77 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 2
>  %78 = load <8 x i32> addrspace(2)* %77, !tbaa !0
>  %79 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 2
>  %80 = load <4 x i32> addrspace(2)* %79, !tbaa !0
>  %81 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 3
>  %82 = load <8 x i32> addrspace(2)* %81, !tbaa !0
>  %83 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 3
>  %84 = load <4 x i32> addrspace(2)* %83, !tbaa !0
>  %85 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 4
>  %86 = load <8 x i32> addrspace(2)* %85, !tbaa !0
>  %87 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 4
>  %88 = load <4 x i32> addrspace(2)* %87, !tbaa !0
>  %89 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 5
>  %90 = load <8 x i32> addrspace(2)* %89, !tbaa !0
>  %91 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 5
>  %92 = load <4 x i32> addrspace(2)* %91, !tbaa !0
>  %93 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 6
>  %94 = load <8 x i32> addrspace(2)* %93, !tbaa !0
>  %95 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 6
>  %96 = load <4 x i32> addrspace(2)* %95, !tbaa !0
>  %97 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 7
>  %98 = load <8 x i32> addrspace(2)* %97, !tbaa !0
>  %99 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 7
>  %100 = load <4 x i32> addrspace(2)* %99, !tbaa !0
>  %101 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 8
>  %102 = load <8 x i32> addrspace(2)* %101, !tbaa !0
>  %103 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 8
>  %104 = load <4 x i32> addrspace(2)* %103, !tbaa !0
>  %105 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 9
>  %106 = load <8 x i32> addrspace(2)* %105, !tbaa !0
>  %107 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 9
>  %108 = load <4 x i32> addrspace(2)* %107, !tbaa !0
>  %109 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 10
>  %110 = load <8 x i32> addrspace(2)* %109, !tbaa !0
>  %111 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 10
>  %112 = load <4 x i32> addrspace(2)* %111, !tbaa !0
>  %113 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 11
>  %114 = load <8 x i32> addrspace(2)* %113, !tbaa !0
>  %115 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 11
>  %116 = load <4 x i32> addrspace(2)* %115, !tbaa !0
>  %117 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %118 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %119 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %5, <2 x i32> %7)
>  %120 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %5, <2 x i32> %7)
>  %121 = call float @llvm.SI.fs.interp(i32 0, i32 2, i32 %5, <2 x i32> %7)
>  %122 = call float @llvm.SI.fs.interp(i32 1, i32 2, i32 %5, <2 x i32> %7)
>  %123 = call float @llvm.SI.fs.interp(i32 0, i32 3, i32 %5, <2 x i32> %7)
>  %124 = call float @llvm.SI.fs.interp(i32 1, i32 3, i32 %5, <2 x i32> %7)
>  %125 = call float @llvm.SI.fs.interp(i32 0, i32 4, i32 %5, <2 x i32> %7)
>  %126 = call float @llvm.SI.fs.interp(i32 1, i32 4, i32 %5, <2 x i32> %7)
>  %127 = call float @llvm.SI.fs.interp(i32 2, i32 4, i32 %5, <2 x i32> %7)
>  %128 = call float @llvm.SI.fs.interp(i32 1, i32 5, i32 %5, <2 x i32> %7)
>  %129 = call float @llvm.SI.fs.interp(i32 2, i32 5, i32 %5, <2 x i32> %7)
>  %130 = call float @llvm.SI.fs.interp(i32 0, i32 6, i32 %5, <2 x i32> %7)
>  %131 = call float @llvm.SI.fs.interp(i32 1, i32 6, i32 %5, <2 x i32> %7)
>  %132 = fmul float %130, 1.000000e+00
>  %133 = fadd float %132, 0.000000e+00
>  %134 = fmul float %131, -1.000000e+00
>  %135 = fadd float %134, 1.000000e+00
>  %136 = bitcast float %133 to i32
>  %137 = bitcast float %135 to i32
>  %138 = insertelement <2 x i32> undef, i32 %136, i32 0
>  %139 = insertelement <2 x i32> %138, i32 %137, i32 1
>  %140 = bitcast <8 x i32> %106 to <32 x i8>
>  %141 = bitcast <4 x i32> %108 to <16 x i8>
>  %142 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %139, <32 x i8> %140, <16 x i8> %141, i32 2)
>  %143 = extractelement <4 x float> %142, i32 0
>  %144 = extractelement <4 x float> %142, i32 2
>  %145 = call float @fabs(float %143)
>  %146 = fsub float -0.000000e+00, %145
>  %147 = fsub float -0.000000e+00, %145
>  %148 = fsub float -0.000000e+00, %145
>  %149 = fsub float -0.000000e+00, %145
>  %150 = fcmp oge float %146, 0.000000e+00
>  %151 = sext i1 %150 to i32
>  %152 = bitcast i32 %151 to float
>  %153 = bitcast float %152 to i32
>  %154 = icmp ne i32 %153, 0
>  %. = select i1 %154, float -1.000000e+00, float 0.000000e+00
>  %155 = fcmp oge float %147, 0.000000e+00
>  %156 = sext i1 %155 to i32
>  %157 = bitcast i32 %156 to float
>  %158 = bitcast float %157 to i32
>  %159 = icmp ne i32 %158, 0
>  %temp16.0 = select i1 %159, float -1.000000e+00, float 0.000000e+00
>  %160 = fcmp oge float %148, 0.000000e+00
>  %161 = sext i1 %160 to i32
>  %162 = bitcast i32 %161 to float
>  %163 = bitcast float %162 to i32
>  %164 = icmp ne i32 %163, 0
>  %.163 = select i1 %164, float -1.000000e+00, float 0.000000e+00
>  %165 = fcmp oge float %149, 0.000000e+00
>  %166 = sext i1 %165 to i32
>  %167 = bitcast i32 %166 to float
>  %168 = bitcast float %167 to i32
>  %169 = icmp ne i32 %168, 0
>  %170 = fcmp olt float %., 0.000000e+00
>  %171 = sext i1 %170 to i32
>  %172 = fcmp olt float %temp16.0, 0.000000e+00
>  %173 = sext i1 %172 to i32
>  %174 = fcmp olt float %.163, 0.000000e+00
>  %175 = sext i1 %174 to i32
>  %176 = bitcast i32 %171 to float
>  %177 = bitcast i32 %173 to float
>  %178 = bitcast i32 %175 to float
>  %179 = bitcast float %176 to i32
>  %180 = bitcast float %178 to i32
>  %181 = or i32 %179, %180
>  %182 = bitcast i32 %181 to float
>  %183 = bitcast float %182 to i32
>  %184 = bitcast float %177 to i32
>  %185 = or i32 %183, %184
>  %186 = bitcast i32 %185 to float
>  %187 = bitcast float %186 to i32
>  %188 = and i32 %187, 1065353216
>  %189 = bitcast i32 %188 to float
>  %190 = fsub float -0.000000e+00, %189
>  %191 = fsub float -0.000000e+00, %189
>  %192 = fsub float -0.000000e+00, %189
>  %193 = fsub float -0.000000e+00, %189
>  call void @llvm.AMDGPU.kill(float %190)
>  call void @llvm.AMDGPU.kill(float %191)
>  call void @llvm.AMDGPU.kill(float %192)
>  call void @llvm.AMDGPU.kill(float %193)
>  %194 = bitcast float %123 to i32
>  %195 = bitcast float %124 to i32
>  %196 = insertelement <2 x i32> undef, i32 %194, i32 0
>  %197 = insertelement <2 x i32> %196, i32 %195, i32 1
>  %198 = bitcast <8 x i32> %94 to <32 x i8>
>  %199 = bitcast <4 x i32> %96 to <16 x i8>
>  %200 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %197, <32 x i8> %198, <16 x i8> %199, i32 2)
>  %201 = extractelement <4 x float> %200, i32 0
>  %202 = extractelement <4 x float> %200, i32 1
>  %203 = extractelement <4 x float> %200, i32 2
>  %204 = call float @llvm.pow.f32(float %201, float 0x40019999A0000000)
>  %205 = call float @llvm.pow.f32(float %202, float 0x40019999A0000000)
>  %206 = call float @llvm.pow.f32(float %203, float 0x40019999A0000000)
>  %207 = fdiv float 1.000000e+00, %24
>  %208 = fmul float %207, 0x3FD54FDF40000000
>  %209 = fmul float %208, %208
>  %210 = bitcast float %117 to i32
>  %211 = bitcast float %118 to i32
>  %212 = insertelement <2 x i32> undef, i32 %210, i32 0
>  %213 = insertelement <2 x i32> %212, i32 %211, i32 1
>  %214 = bitcast <8 x i32> %70 to <32 x i8>
>  %215 = bitcast <4 x i32> %72 to <16 x i8>
>  %216 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %213, <32 x i8> %214, <16 x i8> %215, i32 2)
>  %217 = extractelement <4 x float> %216, i32 0
>  %218 = extractelement <4 x float> %216, i32 1
>  %219 = extractelement <4 x float> %216, i32 2
>  %220 = extractelement <4 x float> %216, i32 3
>  %221 = bitcast float %117 to i32
>  %222 = bitcast float %118 to i32
>  %223 = insertelement <2 x i32> undef, i32 %221, i32 0
>  %224 = insertelement <2 x i32> %223, i32 %222, i32 1
>  %225 = bitcast <8 x i32> %74 to <32 x i8>
>  %226 = bitcast <4 x i32> %76 to <16 x i8>
>  %227 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %224, <32 x i8> %225, <16 x i8> %226, i32 2)
>  %228 = extractelement <4 x float> %227, i32 0
>  %229 = fmul float %217, 2.000000e+00
>  %230 = fadd float %229, -1.000000e+00
>  %231 = fmul float %218, 2.000000e+00
>  %232 = fadd float %231, -1.000000e+00
>  %233 = fmul float %208, %230
>  %234 = fmul float %208, %232
>  %235 = fmul float %228, 2.000000e+00
>  %236 = fadd float %235, -1.000000e+00
>  %237 = bitcast float %119 to i32
>  %238 = bitcast float %120 to i32
>  %239 = insertelement <2 x i32> undef, i32 %237, i32 0
>  %240 = insertelement <2 x i32> %239, i32 %238, i32 1
>  %241 = bitcast <8 x i32> %78 to <32 x i8>
>  %242 = bitcast <4 x i32> %80 to <16 x i8>
>  %243 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %240, <32 x i8> %241, <16 x i8> %242, i32 2)
>  %244 = extractelement <4 x float> %243, i32 0
>  %245 = extractelement <4 x float> %243, i32 1
>  %246 = extractelement <4 x float> %243, i32 2
>  %247 = extractelement <4 x float> %243, i32 3
>  %248 = bitcast float %119 to i32
>  %249 = bitcast float %120 to i32
>  %250 = insertelement <2 x i32> undef, i32 %248, i32 0
>  %251 = insertelement <2 x i32> %250, i32 %249, i32 1
>  %252 = bitcast <8 x i32> %82 to <32 x i8>
>  %253 = bitcast <4 x i32> %84 to <16 x i8>
>  %254 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %251, <32 x i8> %252, <16 x i8> %253, i32 2)
>  %255 = extractelement <4 x float> %254, i32 0
>  %256 = fadd float %244, %244
>  %257 = fadd float %245, %245
>  %258 = fmul float %256, -1.000000e+00
>  %259 = fadd float %258, 1.000000e+00
>  %260 = fmul float %257, 1.000000e+00
>  %261 = fadd float %260, -1.000000e+00
>  %262 = fmul float %208, %259
>  %263 = fmul float %208, %261
>  %264 = fmul float %230, %208
>  %265 = fadd float %264, %262
>  %266 = fmul float %232, %208
>  %267 = fadd float %266, %263
>  %268 = fmul float %255, -2.000000e+00
>  %269 = fadd float %268, 1.000000e+00
>  %270 = fmul float %209, %246
>  %271 = fmul float %209, %247
>  %272 = fmul float %209, %269
>  %273 = fmul float %219, %209
>  %274 = fadd float %273, %270
>  %275 = fmul float %220, %209
>  %276 = fadd float %275, %271
>  %277 = fmul float %236, %209
>  %278 = fadd float %277, %272
>  %279 = fadd float %233, %233
>  %280 = fadd float %234, %234
>  %281 = fmul float %262, %279
>  %282 = fmul float %263, %280
>  %283 = fmul float %233, %263
>  %284 = fmul float %234, %262
>  %285 = fadd float %283, %284
>  %286 = fadd float %274, %281
>  %287 = fadd float %276, %282
>  %288 = fadd float %278, %285
>  %289 = bitcast float %121 to i32
>  %290 = bitcast float %122 to i32
>  %291 = insertelement <2 x i32> undef, i32 %289, i32 0
>  %292 = insertelement <2 x i32> %291, i32 %290, i32 1
>  %293 = bitcast <8 x i32> %78 to <32 x i8>
>  %294 = bitcast <4 x i32> %80 to <16 x i8>
>  %295 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %292, <32 x i8> %293, <16 x i8> %294, i32 2)
>  %296 = extractelement <4 x float> %295, i32 0
>  %297 = extractelement <4 x float> %295, i32 1
>  %298 = extractelement <4 x float> %295, i32 2
>  %299 = extractelement <4 x float> %295, i32 3
>  %300 = bitcast float %121 to i32
>  %301 = bitcast float %122 to i32
>  %302 = insertelement <2 x i32> undef, i32 %300, i32 0
>  %303 = insertelement <2 x i32> %302, i32 %301, i32 1
>  %304 = bitcast <8 x i32> %82 to <32 x i8>
>  %305 = bitcast <4 x i32> %84 to <16 x i8>
>  %306 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %303, <32 x i8> %304, <16 x i8> %305, i32 2)
>  %307 = extractelement <4 x float> %306, i32 0
>  %308 = fmul float %297, 2.000000e+00
>  %309 = fadd float %308, -1.000000e+00
>  %310 = fmul float %296, 2.000000e+00
>  %311 = fadd float %310, -1.000000e+00
>  %312 = fmul float %208, %309
>  %313 = fmul float %208, %311
>  %314 = fmul float %311, %208
>  %315 = fadd float %314, %265
>  %316 = fmul float %309, %208
>  %317 = fadd float %316, %267
>  %318 = fmul float %307, 2.000000e+00
>  %319 = fadd float %318, -1.000000e+00
>  %320 = fmul float %298, %209
>  %321 = fadd float %320, %286
>  %322 = fmul float %299, %209
>  %323 = fadd float %322, %287
>  %324 = fmul float %319, %209
>  %325 = fadd float %324, %288
>  %326 = fmul float %279, %313
>  %327 = fmul float %280, %312
>  %328 = fmul float %233, %312
>  %329 = fmul float %234, %313
>  %330 = fadd float %328, %329
>  %331 = fadd float %321, %326
>  %332 = fadd float %323, %327
>  %333 = fadd float %325, %330
>  %334 = fmul float %262, %313
>  %335 = fmul float %263, %312
>  %336 = fadd float %334, %334
>  %337 = fadd float %335, %335
>  %338 = fmul float %262, %312
>  %339 = fmul float %263, %313
>  %340 = fadd float %338, %339
>  %341 = fadd float %331, %336
>  %342 = fadd float %332, %337
>  %343 = fadd float %333, %340
>  %344 = fsub float -0.000000e+00, %315
>  %345 = fmul float %315, %344
>  %346 = fadd float %345, %341
>  %347 = fsub float -0.000000e+00, %317
>  %348 = fmul float %317, %347
>  %349 = fadd float %348, %342
>  %350 = fsub float -0.000000e+00, %317
>  %351 = fmul float %315, %350
>  %352 = fadd float %351, %343
>  %353 = fmul float %25, %315
>  %354 = fmul float %26, %317
>  %355 = fadd float %354, %353
>  %356 = fmul float %27, 1.000000e+00
>  %357 = fadd float %355, %356
>  %358 = call float @llvm.AMDIL.clamp.(float %357, float 0.000000e+00, float 1.000000e+00)
>  %359 = fsub float -0.000000e+00, %125
>  %360 = fadd float %50, %359
>  %361 = fsub float -0.000000e+00, %126
>  %362 = fadd float %51, %361
>  %363 = fsub float -0.000000e+00, %127
>  %364 = fadd float %52, %363
>  %365 = fmul float %360, %360
>  %366 = fmul float %362, %362
>  %367 = fadd float %366, %365
>  %368 = fmul float %364, %364
>  %369 = fadd float %367, %368
>  %370 = call float @llvm.maxnum.f32(float %369, float 0x3E7AD7F2A0000000)
>  %371 = call float @llvm.AMDGPU.rsq.clamped.f32(float %370)
>  %372 = fmul float %371, %360
>  %373 = fmul float %371, %362
>  %374 = fmul float %371, %364
>  %375 = fsub float -0.000000e+00, %39
>  %376 = fmul float %360, %371
>  %377 = fadd float %376, %375
>  %378 = fsub float -0.000000e+00, %40
>  %379 = fmul float %362, %371
>  %380 = fadd float %379, %378
>  %381 = fsub float -0.000000e+00, %41
>  %382 = fmul float %364, %371
>  %383 = fadd float %382, %381
>  %384 = fmul float %42, %377
>  %385 = fadd float %384, %39
>  %386 = fmul float %42, %380
>  %387 = fadd float %386, %40
>  %388 = fmul float %42, %383
>  %389 = fadd float %388, %41
>  %390 = fmul float %385, %385
>  %391 = fmul float %387, %387
>  %392 = fadd float %391, %390
>  %393 = fmul float %389, %389
>  %394 = fadd float %392, %393
>  %395 = call float @llvm.maxnum.f32(float %394, float 0x3E7AD7F2A0000000)
>  %396 = call float @llvm.AMDGPU.rsq.clamped.f32(float %395)
>  %397 = fmul float %385, %396
>  %398 = fadd float %397, %25
>  %399 = fmul float %387, %396
>  %400 = fadd float %399, %26
>  %401 = fmul float %389, %396
>  %402 = fadd float %401, %27
>  %403 = fdiv float 1.000000e+00, %402
>  %404 = fsub float -0.000000e+00, %315
>  %405 = fmul float %398, %403
>  %406 = fadd float %405, %404
>  %407 = fsub float -0.000000e+00, %317
>  %408 = fmul float %400, %403
>  %409 = fadd float %408, %407
>  %410 = fdiv float 1.000000e+00, %34
>  %411 = fadd float %410, %346
>  %412 = fadd float %410, %349
>  %413 = fmul float %352, %352
>  %414 = fsub float -0.000000e+00, %413
>  %415 = fmul float %411, %412
>  %416 = fadd float %415, %414
>  %417 = fmul float %406, %406
>  %418 = fadd float %352, %352
>  %419 = fmul float %406, %418
>  %420 = fsub float -0.000000e+00, %419
>  %421 = fmul float %411, %409
>  %422 = fadd float %421, %420
>  %423 = fmul float %409, %422
>  %424 = fmul float %417, %412
>  %425 = fadd float %424, %423
>  %426 = fmul float %425, 5.000000e-01
>  %427 = fdiv float 1.000000e+00, %416
>  %428 = fmul float %427, %426
>  %429 = fsub float -0.000000e+00, %416
>  %430 = fcmp oge float %429, 0.000000e+00
>  %431 = sext i1 %430 to i32
>  %432 = bitcast i32 %431 to float
>  %433 = bitcast float %432 to i32
>  %434 = icmp ne i32 %433, 0
>  %temp96.0 = select i1 %434, float 1.000000e+00, float -0.000000e+00
>  %435 = fmul float %426, %427
>  %436 = fadd float %435, -1.600000e+01
>  %437 = fcmp oge float %436, 0.000000e+00
>  %438 = sext i1 %437 to i32
>  %439 = bitcast i32 %438 to float
>  %440 = bitcast float %439 to i32
>  %441 = icmp ne i32 %440, 0
>  %.164 = select i1 %441, float 1.000000e+00, float -0.000000e+00
>  %442 = fadd float %.164, %temp96.0
>  %443 = fmul float %428, 0xBFF7154CA0000000
>  %444 = call float @llvm.AMDIL.exp.(float %443)
>  %445 = call float @llvm.maxnum.f32(float %416, float 0x3E7AD7F2A0000000)
>  %446 = call float @llvm.AMDGPU.rsq.clamped.f32(float %445)
>  %447 = fmul float %446, %444
>  %448 = fmul float %410, 1.600000e+01
>  %449 = fadd float %448, %346
>  %450 = fmul float %410, 1.600000e+01
>  %451 = fadd float %450, %349
>  %452 = fsub float -0.000000e+00, %413
>  %453 = fmul float %449, %451
>  %454 = fadd float %453, %452
>  %455 = fsub float -0.000000e+00, %419
>  %456 = fmul float %449, %409
>  %457 = fadd float %456, %455
>  %458 = fmul float %409, %457
>  %459 = fmul float %417, %451
>  %460 = fadd float %459, %458
>  %461 = fmul float %460, 5.000000e-01
>  %462 = fdiv float 1.000000e+00, %454
>  %463 = fmul float %461, %462
>  %464 = fsub float -0.000000e+00, %454
>  %465 = fcmp oge float %464, 0.000000e+00
>  %466 = sext i1 %465 to i32
>  %467 = bitcast i32 %466 to float
>  %468 = bitcast float %467 to i32
>  %469 = icmp ne i32 %468, 0
>  %temp80.0 = select i1 %469, float 1.000000e+00, float -0.000000e+00
>  %470 = fmul float %461, %462
>  %471 = fadd float %470, -1.600000e+01
>  %472 = fcmp oge float %471, 0.000000e+00
>  %473 = sext i1 %472 to i32
>  %474 = bitcast i32 %473 to float
>  %475 = bitcast float %474 to i32
>  %476 = icmp ne i32 %475, 0
>  %.165 = select i1 %476, float 1.000000e+00, float -0.000000e+00
>  %477 = fadd float %.165, %temp80.0
>  %478 = fmul float %463, 0xBFF7154CA0000000
>  %479 = call float @llvm.maxnum.f32(float %454, float 0x3E7AD7F2A0000000)
>  %480 = call float @llvm.AMDGPU.rsq.clamped.f32(float %479)
>  %481 = call float @llvm.AMDIL.exp.(float %478)
>  %482 = fmul float %480, %481
>  %483 = fmul float %482, 0x3F747AE140000000
>  %484 = fsub float -0.000000e+00, %43
>  %485 = fmul float %360, %371
>  %486 = fadd float %485, %484
>  %487 = fsub float -0.000000e+00, %44
>  %488 = fmul float %362, %371
>  %489 = fadd float %488, %487
>  %490 = fsub float -0.000000e+00, %45
>  %491 = fmul float %364, %371
>  %492 = fadd float %491, %490
>  %493 = fmul float %46, %486
>  %494 = fadd float %493, %43
>  %495 = fmul float %46, %489
>  %496 = fadd float %495, %44
>  %497 = fmul float %46, %492
>  %498 = fadd float %497, %45
>  %499 = fmul float %494, %494
>  %500 = fmul float %496, %496
>  %501 = fadd float %500, %499
>  %502 = fmul float %498, %498
>  %503 = fadd float %501, %502
>  %504 = call float @llvm.maxnum.f32(float %503, float 0x3E7AD7F2A0000000)
>  %505 = call float @llvm.AMDGPU.rsq.clamped.f32(float %504)
>  %506 = fmul float %494, %505
>  %507 = fadd float %506, %28
>  %508 = fmul float %496, %505
>  %509 = fadd float %508, %29
>  %510 = fmul float %498, %505
>  %511 = fadd float %510, %30
>  %512 = fdiv float 1.000000e+00, %511
>  %513 = fsub float -0.000000e+00, %315
>  %514 = fmul float %507, %512
>  %515 = fadd float %514, %513
>  %516 = fsub float -0.000000e+00, %317
>  %517 = fmul float %509, %512
>  %518 = fadd float %517, %516
>  %519 = fdiv float 1.000000e+00, %38
>  %520 = fadd float %519, %346
>  %521 = fadd float %519, %349
>  %522 = fsub float -0.000000e+00, %413
>  %523 = fmul float %520, %521
>  %524 = fadd float %523, %522
>  %525 = fmul float %515, %515
>  %526 = fmul float %418, %515
>  %527 = fsub float -0.000000e+00, %526
>  %528 = fmul float %520, %518
>  %529 = fadd float %528, %527
>  %530 = fmul float %518, %529
>  %531 = fmul float %525, %521
>  %532 = fadd float %531, %530
>  %533 = fmul float %532, 5.000000e-01
>  %534 = fdiv float 1.000000e+00, %524
>  %535 = fmul float %534, %533
>  %536 = fsub float -0.000000e+00, %524
>  %537 = fcmp oge float %536, 0.000000e+00
>  %538 = sext i1 %537 to i32
>  %539 = bitcast i32 %538 to float
>  %540 = bitcast float %539 to i32
>  %541 = icmp ne i32 %540, 0
>  %temp80.1 = select i1 %541, float 1.000000e+00, float -0.000000e+00
>  %542 = fmul float %533, %534
>  %543 = fadd float %542, -1.600000e+01
>  %544 = fcmp oge float %543, 0.000000e+00
>  %545 = sext i1 %544 to i32
>  %546 = bitcast i32 %545 to float
>  %547 = bitcast float %546 to i32
>  %548 = icmp ne i32 %547, 0
>  %.166 = select i1 %548, float 1.000000e+00, float -0.000000e+00
>  %549 = fadd float %.166, %temp80.1
>  %550 = fmul float %535, 0xBFF7154CA0000000
>  %551 = call float @llvm.AMDIL.exp.(float %550)
>  %552 = call float @llvm.maxnum.f32(float %524, float 0x3E7AD7F2A0000000)
>  %553 = call float @llvm.AMDGPU.rsq.clamped.f32(float %552)
>  %554 = fmul float %551, %553
>  %555 = fmul float %56, %126
>  %556 = fmul float %57, %126
>  %557 = fmul float %58, %126
>  %558 = fmul float %125, %53
>  %559 = fadd float %558, %555
>  %560 = fmul float %125, %54
>  %561 = fadd float %560, %556
>  %562 = fmul float %125, %55
>  %563 = fadd float %562, %557
>  %564 = fmul float %127, %59
>  %565 = fadd float %564, %559
>  %566 = fmul float %127, %60
>  %567 = fadd float %566, %561
>  %568 = fmul float %127, %61
>  %569 = fadd float %568, %563
>  %570 = fadd float %565, %62
>  %571 = fadd float %567, %63
>  %572 = fadd float %569, %64
>  %573 = fdiv float 1.000000e+00, %572
>  %574 = fmul float %573, %570
>  %575 = fmul float %573, %571
>  %576 = fmul float %574, 5.000000e-01
>  %577 = fadd float %576, -5.000000e-01
>  %578 = fmul float %575, -5.000000e-01
>  %579 = fadd float %578, -5.000000e-01
>  %580 = fmul float %346, 6.000000e+00
>  %581 = fadd float %580, %577
>  %582 = fmul float %349, 6.000000e+00
>  %583 = fadd float %582, %579
>  %584 = fmul float %524, 2.000000e+00
>  %585 = fadd float %584, %581
>  %586 = fmul float %524, 2.000000e+00
>  %587 = fadd float %586, %583
>  %588 = bitcast float %585 to i32
>  %589 = bitcast float %587 to i32
>  %590 = insertelement <2 x i32> undef, i32 %588, i32 0
>  %591 = insertelement <2 x i32> %590, i32 %589, i32 1
>  %592 = bitcast <8 x i32> %86 to <32 x i8>
>  %593 = bitcast <4 x i32> %88 to <16 x i8>
>  %594 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %591, <32 x i8> %592, <16 x i8> %593, i32 2)
>  %595 = extractelement <4 x float> %594, i32 0
>  %596 = extractelement <4 x float> %594, i32 1
>  %597 = extractelement <4 x float> %594, i32 2
>  %598 = extractelement <4 x float> %594, i32 3
>  %599 = fmul float %524, 2.000000e+00
>  %600 = fadd float %599, %581
>  %601 = fmul float %524, -2.000000e+00
>  %602 = fadd float %601, %583
>  %603 = bitcast float %600 to i32
>  %604 = bitcast float %602 to i32
>  %605 = insertelement <2 x i32> undef, i32 %603, i32 0
>  %606 = insertelement <2 x i32> %605, i32 %604, i32 1
>  %607 = bitcast <8 x i32> %86 to <32 x i8>
>  %608 = bitcast <4 x i32> %88 to <16 x i8>
>  %609 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %606, <32 x i8> %607, <16 x i8> %608, i32 2)
>  %610 = extractelement <4 x float> %609, i32 0
>  %611 = extractelement <4 x float> %609, i32 1
>  %612 = extractelement <4 x float> %609, i32 2
>  %613 = extractelement <4 x float> %609, i32 3
>  %614 = fmul float %610, 2.500000e-01
>  %615 = fmul float %611, 2.500000e-01
>  %616 = fmul float %612, 2.500000e-01
>  %617 = fmul float %613, 2.500000e-01
>  %618 = fmul float %595, 2.500000e-01
>  %619 = fadd float %618, %614
>  %620 = fmul float %596, 2.500000e-01
>  %621 = fadd float %620, %615
>  %622 = fmul float %597, 2.500000e-01
>  %623 = fadd float %622, %616
>  %624 = fmul float %598, 2.500000e-01
>  %625 = fadd float %624, %617
>  %626 = fmul float %524, -2.000000e+00
>  %627 = fadd float %626, %581
>  %628 = fmul float %524, 6.000000e+00
>  %629 = fadd float %628, %583
>  %630 = bitcast float %627 to i32
>  %631 = bitcast float %629 to i32
>  %632 = insertelement <2 x i32> undef, i32 %630, i32 0
>  %633 = insertelement <2 x i32> %632, i32 %631, i32 1
>  %634 = bitcast <8 x i32> %86 to <32 x i8>
>  %635 = bitcast <4 x i32> %88 to <16 x i8>
>  %636 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %633, <32 x i8> %634, <16 x i8> %635, i32 2)
>  %637 = extractelement <4 x float> %636, i32 0
>  %638 = extractelement <4 x float> %636, i32 1
>  %639 = extractelement <4 x float> %636, i32 2
>  %640 = extractelement <4 x float> %636, i32 3
>  %641 = fmul float %637, 2.500000e-01
>  %642 = fadd float %641, %619
>  %643 = fmul float %638, 2.500000e-01
>  %644 = fadd float %643, %621
>  %645 = fmul float %639, 2.500000e-01
>  %646 = fadd float %645, %623
>  %647 = fmul float %640, 2.500000e-01
>  %648 = fadd float %647, %625
>  %649 = fmul float %524, -2.000000e+00
>  %650 = fadd float %649, %581
>  %651 = fmul float %524, -2.000000e+00
>  %652 = fadd float %651, %583
>  %653 = bitcast float %650 to i32
>  %654 = bitcast float %652 to i32
>  %655 = insertelement <2 x i32> undef, i32 %653, i32 0
>  %656 = insertelement <2 x i32> %655, i32 %654, i32 1
>  %657 = bitcast <8 x i32> %86 to <32 x i8>
>  %658 = bitcast <4 x i32> %88 to <16 x i8>
>  %659 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %656, <32 x i8> %657, <16 x i8> %658, i32 2)
>  %660 = extractelement <4 x float> %659, i32 0
>  %661 = extractelement <4 x float> %659, i32 1
>  %662 = extractelement <4 x float> %659, i32 2
>  %663 = extractelement <4 x float> %659, i32 3
>  %664 = fmul float %660, 2.500000e-01
>  %665 = fadd float %664, %642
>  %666 = fmul float %661, 2.500000e-01
>  %667 = fadd float %666, %644
>  %668 = fmul float %662, 2.500000e-01
>  %669 = fadd float %668, %646
>  %670 = fmul float %663, 2.500000e-01
>  %671 = fadd float %670, %648
>  %672 = call float @fabs(float %665)
>  %673 = call float @llvm.log2.f32(float %672)
>  %674 = call float @fabs(float %667)
>  %675 = call float @llvm.log2.f32(float %674)
>  %676 = call float @fabs(float %669)
>  %677 = call float @llvm.log2.f32(float %676)
>  %678 = fmul float %673, 0x40019999A0000000
>  %679 = fmul float %675, 0x40019999A0000000
>  %680 = fmul float %677, 0x40019999A0000000
>  %681 = call float @llvm.AMDIL.exp.(float %678)
>  %682 = call float @llvm.AMDIL.exp.(float %679)
>  %683 = call float @llvm.AMDIL.exp.(float %680)
>  %684 = fmul float %315, %372
>  %685 = fmul float %317, %373
>  %686 = fadd float %685, %684
>  %687 = fmul float 1.000000e+00, %374
>  %688 = fadd float %686, %687
>  %689 = fsub float -0.000000e+00, %688
>  %690 = fadd float %689, 1.000000e+00
>  %691 = fmul float %690, %690
>  %692 = fmul float %691, %691
>  %693 = fmul float %690, %692
>  %694 = fmul float %447, %31
>  %695 = fmul float %447, %32
>  %696 = fmul float %447, %33
>  %697 = fsub float -0.000000e+00, %442
>  %698 = fsub float -0.000000e+00, %442
>  %699 = fsub float -0.000000e+00, %442
>  %700 = fcmp oge float %697, 0.000000e+00
>  %701 = sext i1 %700 to i32
>  %702 = bitcast i32 %701 to float
>  %703 = bitcast float %702 to i32
>  %704 = icmp ne i32 %703, 0
>  %temp72.0 = select i1 %704, float %694, float -0.000000e+00
>  %705 = fcmp oge float %698, 0.000000e+00
>  %706 = sext i1 %705 to i32
>  %707 = bitcast i32 %706 to float
>  %708 = bitcast float %707 to i32
>  %709 = icmp ne i32 %708, 0
>  %.167 = select i1 %709, float %695, float -0.000000e+00
>  %710 = fcmp oge float %699, 0.000000e+00
>  %711 = sext i1 %710 to i32
>  %712 = bitcast i32 %711 to float
>  %713 = bitcast float %712 to i32
>  %714 = icmp ne i32 %713, 0
>  %temp64.0 = select i1 %714, float %696, float -0.000000e+00
>  %715 = fmul float %554, %35
>  %716 = fmul float %554, %36
>  %717 = fmul float %554, %37
>  %718 = fmul float %715, 0x3FE99999A0000000
>  %719 = fmul float %716, 0x3FE99999A0000000
>  %720 = fmul float %717, 0x3FE99999A0000000
>  %721 = fsub float -0.000000e+00, %549
>  %722 = fsub float -0.000000e+00, %549
>  %723 = fsub float -0.000000e+00, %549
>  %724 = fsub float -0.000000e+00, %549
>  %725 = fcmp oge float %721, 0.000000e+00
>  %726 = sext i1 %725 to i32
>  %727 = bitcast i32 %726 to float
>  %728 = bitcast float %727 to i32
>  %729 = icmp ne i32 %728, 0
>  %.168 = select i1 %729, float %718, float -0.000000e+00
>  %730 = fcmp oge float %722, 0.000000e+00
>  %731 = sext i1 %730 to i32
>  %732 = bitcast i32 %731 to float
>  %733 = bitcast float %732 to i32
>  %734 = icmp ne i32 %733, 0
>  %temp68.1 = select i1 %734, float %719, float -0.000000e+00
>  %735 = fcmp oge float %723, 0.000000e+00
>  %736 = sext i1 %735 to i32
>  %737 = bitcast i32 %736 to float
>  %738 = bitcast float %737 to i32
>  %739 = icmp ne i32 %738, 0
>  %.169 = select i1 %739, float %720, float -0.000000e+00
>  %740 = fcmp oge float %724, 0.000000e+00
>  %741 = sext i1 %740 to i32
>  %742 = bitcast i32 %741 to float
>  %743 = bitcast float %742 to i32
>  %744 = icmp ne i32 %743, 0
>  %745 = fadd float %temp72.0, %.168
>  %746 = fadd float %.167, %temp68.1
>  %747 = fadd float %temp64.0, %.169
>  %748 = fmul float %745, %693
>  %749 = fadd float %748, %204
>  %750 = fmul float %746, %693
>  %751 = fadd float %750, %205
>  %752 = fmul float %747, %693
>  %753 = fadd float %752, %206
>  %754 = fmul float %204, %31
>  %755 = fmul float %205, %32
>  %756 = fmul float %206, %33
>  %757 = fmul float %483, %754
>  %758 = fmul float %483, %755
>  %759 = fmul float %483, %756
>  %760 = fsub float -0.000000e+00, %477
>  %761 = fsub float -0.000000e+00, %477
>  %762 = fsub float -0.000000e+00, %477
>  %763 = fcmp oge float %760, 0.000000e+00
>  %764 = sext i1 %763 to i32
>  %765 = bitcast i32 %764 to float
>  %766 = bitcast float %765 to i32
>  %767 = icmp ne i32 %766, 0
>  %temp40.1 = select i1 %767, float %757, float -0.000000e+00
>  %768 = fcmp oge float %761, 0.000000e+00
>  %769 = sext i1 %768 to i32
>  %770 = bitcast i32 %769 to float
>  %771 = bitcast float %770 to i32
>  %772 = icmp ne i32 %771, 0
>  %.170 = select i1 %772, float %758, float -0.000000e+00
>  %773 = fcmp oge float %762, 0.000000e+00
>  %774 = sext i1 %773 to i32
>  %775 = bitcast i32 %774 to float
>  %776 = bitcast float %775 to i32
>  %777 = icmp ne i32 %776, 0
>  %temp20.1 = select i1 %777, float %759, float -0.000000e+00
>  %778 = fmul float %749, %358
>  %779 = fadd float %778, %temp40.1
>  %780 = fmul float %751, %358
>  %781 = fadd float %780, %.170
>  %782 = fmul float %753, %358
>  %783 = fadd float %782, %temp20.1
>  %784 = fmul float %681, %68
>  %785 = fadd float %784, %779
>  %786 = fmul float %682, %68
>  %787 = fadd float %786, %781
>  %788 = fmul float %683, %68
>  %789 = fadd float %788, %783
>  %790 = bitcast float %123 to i32
>  %791 = bitcast float %124 to i32
>  %792 = insertelement <2 x i32> undef, i32 %790, i32 0
>  %793 = insertelement <2 x i32> %792, i32 %791, i32 1
>  %794 = bitcast <8 x i32> %90 to <32 x i8>
>  %795 = bitcast <4 x i32> %92 to <16 x i8>
>  %796 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %793, <32 x i8> %794, <16 x i8> %795, i32 2)
>  %797 = extractelement <4 x float> %796, i32 3
>  %798 = fmul float %690, %692
>  %799 = fadd float %798, 5.000000e-01
>  %800 = fmul float %799, %785
>  %801 = fmul float %799, %787
>  %802 = fmul float %799, %789
>  %803 = fadd float %671, 1.000000e+00
>  %804 = fadd float %671, 1.000000e+00
>  %805 = fadd float %671, 1.000000e+00
>  %806 = fadd float %671, 1.000000e+00
>  %807 = call float @llvm.AMDIL.clamp.(float %803, float 0.000000e+00, float 1.000000e+00)
>  %808 = call float @llvm.AMDIL.clamp.(float %804, float 0.000000e+00, float 1.000000e+00)
>  %809 = call float @llvm.AMDIL.clamp.(float %805, float 0.000000e+00, float 1.000000e+00)
>  %810 = call float @llvm.AMDIL.clamp.(float %806, float 0.000000e+00, float 1.000000e+00)
>  %811 = fmul float %797, %808
>  %812 = call float @fabs(float %800)
>  %813 = call float @llvm.log2.f32(float %812)
>  %814 = call float @fabs(float %801)
>  %815 = call float @llvm.log2.f32(float %814)
>  %816 = call float @fabs(float %802)
>  %817 = call float @llvm.log2.f32(float %816)
>  %818 = fmul float %813, 0x3FDD1743E0000000
>  %819 = fmul float %815, 0x3FDD1743E0000000
>  %820 = fmul float %817, 0x3FDD1743E0000000
>  %821 = call float @llvm.AMDIL.exp.(float %818)
>  %822 = call float @llvm.AMDIL.exp.(float %819)
>  %823 = call float @llvm.AMDIL.exp.(float %820)
>  %824 = bitcast float %821 to i32
>  %825 = bitcast float %822 to i32
>  %826 = bitcast float %823 to i32
>  %827 = insertelement <4 x i32> undef, i32 %824, i32 0
>  %828 = insertelement <4 x i32> %827, i32 %825, i32 1
>  %829 = insertelement <4 x i32> %828, i32 %826, i32 2
>  %830 = insertelement <4 x i32> %829, i32 undef, i32 3
>  %831 = bitcast <8 x i32> %114 to <32 x i8>
>  %832 = bitcast <4 x i32> %116 to <16 x i8>
>  %833 = call <4 x float> @llvm.SI.sample.v4i32(<4 x i32> %830, <32 x i8> %831, <16 x i8> %832, i32 3)
>  %834 = extractelement <4 x float> %833, i32 0
>  %835 = extractelement <4 x float> %833, i32 1
>  %836 = extractelement <4 x float> %833, i32 2
>  %837 = fmul float %130, 1.000000e+00
>  %838 = fadd float %837, 0.000000e+00
>  %839 = fmul float %131, -1.000000e+00
>  %840 = fadd float %839, 1.000000e+00
>  %841 = bitcast float %838 to i32
>  %842 = bitcast float %840 to i32
>  %843 = insertelement <2 x i32> undef, i32 %841, i32 0
>  %844 = insertelement <2 x i32> %843, i32 %842, i32 1
>  %845 = bitcast <8 x i32> %102 to <32 x i8>
>  %846 = bitcast <4 x i32> %104 to <16 x i8>
>  %847 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %844, <32 x i8> %845, <16 x i8> %846, i32 2)
>  %848 = extractelement <4 x float> %847, i32 0
>  %849 = extractelement <4 x float> %847, i32 1
>  %850 = extractelement <4 x float> %847, i32 2
>  %851 = extractelement <4 x float> %847, i32 3
>  %852 = fsub float -0.000000e+00, %851
>  %853 = fadd float %852, 1.000000e+00
>  %854 = fmul float %834, %853
>  %855 = fadd float %854, %848
>  %856 = fmul float %835, %853
>  %857 = fadd float %856, %849
>  %858 = fmul float %836, %853
>  %859 = fadd float %858, %850
>  %860 = call float @llvm.AMDGPU.lrp(float %143, float %855, float 0x3FD99999A0000000)
>  %861 = call float @llvm.AMDGPU.lrp(float %143, float %857, float 0x3FD99999A0000000)
>  %862 = call float @llvm.AMDGPU.lrp(float %143, float %859, float 0x3FD99999A0000000)
>  %863 = bitcast float %860 to i32
>  %864 = bitcast float %861 to i32
>  %865 = bitcast float %862 to i32
>  %866 = insertelement <4 x i32> undef, i32 %863, i32 0
>  %867 = insertelement <4 x i32> %866, i32 %864, i32 1
>  %868 = insertelement <4 x i32> %867, i32 %865, i32 2
>  %869 = insertelement <4 x i32> %868, i32 undef, i32 3
>  %870 = bitcast <8 x i32> %110 to <32 x i8>
>  %871 = bitcast <4 x i32> %112 to <16 x i8>
>  %872 = call <4 x float> @llvm.SI.sample.v4i32(<4 x i32> %869, <32 x i8> %870, <16 x i8> %871, i32 3)
>  %873 = extractelement <4 x float> %872, i32 0
>  %874 = extractelement <4 x float> %872, i32 1
>  %875 = extractelement <4 x float> %872, i32 2
>  %876 = fsub float -0.000000e+00, %143
>  %877 = fmul float %144, %876
>  %878 = fadd float %877, %143
>  %879 = call float @llvm.AMDGPU.lrp(float %878, float %873, float %860)
>  %880 = call float @llvm.AMDGPU.lrp(float %878, float %874, float %861)
>  %881 = call float @llvm.AMDGPU.lrp(float %878, float %875, float %862)
>  %882 = fdiv float 1.000000e+00, %129
>  %883 = fmul float %882, %128
>  %884 = fmul float %883, 5.000000e-01
>  %885 = fadd float %884, 5.000000e-01
>  %886 = bitcast float %885 to i32
>  %887 = bitcast float %49 to i32
>  %888 = insertelement <2 x i32> undef, i32 %886, i32 0
>  %889 = insertelement <2 x i32> %888, i32 %887, i32 1
>  %890 = bitcast <8 x i32> %98 to <32 x i8>
>  %891 = bitcast <4 x i32> %100 to <16 x i8>
>  %892 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %889, <32 x i8> %890, <16 x i8> %891, i32 2)
>  %893 = extractelement <4 x float> %892, i32 0
>  %894 = fsub float -0.000000e+00, %50
>  %895 = fadd float %894, %125
>  %896 = fsub float -0.000000e+00, %51
>  %897 = fadd float %896, %126
>  %898 = fsub float -0.000000e+00, %52
>  %899 = fadd float %898, %127
>  %900 = fmul float %895, %895
>  %901 = fmul float %897, %897
>  %902 = fadd float %901, %900
>  %903 = fmul float %899, %899
>  %904 = fadd float %902, %903
>  %905 = fmul float %904, %48
>  %906 = fmul float %899, %47
>  %907 = fmul float %906, 0x3FF7154CA0000000
>  %908 = call float @llvm.AMDIL.exp.(float %907)
>  %909 = fsub float -0.000000e+00, %908
>  %910 = fadd float %909, 1.000000e+00
>  %911 = fmul float %910, %905
>  %912 = fdiv float 1.000000e+00, %899
>  %913 = fmul float %912, %911
>  %914 = fmul float %913, 0x3FF7154CA0000000
>  %915 = call float @llvm.AMDIL.exp.(float %914)
>  %916 = call float @llvm.AMDIL.clamp.(float %915, float 0.000000e+00, float 1.000000e+00)
>  %917 = fsub float -0.000000e+00, %916
>  %918 = fadd float %917, 1.000000e+00
>  %919 = fmul float %918, %893
>  %920 = fsub float -0.000000e+00, %879
>  %921 = fadd float %920, %65
>  %922 = fsub float -0.000000e+00, %880
>  %923 = fadd float %922, %66
>  %924 = fsub float -0.000000e+00, %881
>  %925 = fadd float %924, %67
>  %926 = fmul float %919, %921
>  %927 = fadd float %926, %879
>  %928 = fmul float %919, %923
>  %929 = fadd float %928, %880
>  %930 = fmul float %919, %925
>  %931 = fadd float %930, %881
>  %932 = call i32 @llvm.SI.packf16(float %927, float %929)
>  %933 = bitcast i32 %932 to float
>  %934 = call i32 @llvm.SI.packf16(float %931, float %811)
>  %935 = bitcast i32 %934 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %933, float %935, float %933, float %935)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: readnone
>declare float @fabs(float) #2
>
>declare void @llvm.AMDGPU.kill(float)
>
>; Function Attrs: nounwind readnone
>declare float @llvm.pow.f32(float, float) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.clamp.(float, float, float) #2
>
>; Function Attrs: nounwind readnone
>declare float @llvm.maxnum.f32(float, float) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.AMDGPU.rsq.clamped.f32(float) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.exp.(float) #2
>
>; Function Attrs: nounwind readnone
>declare float @llvm.log2.f32(float) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v4i32(<4 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDGPU.lrp(float, float, float) #2
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                         ; BEFE0A7E
>	s_mov_b64 s[100:101], s[6:7]                                                 ; BEE40406
>	s_mov_b64 vcc, s[2:3]                                                        ; BEEA0402
>	s_mov_b32 m0, s9                                                             ; BEFC0309
>	v_interp_p1_f32 v2, v0, 1, 6, [m0]                                           ; C8081900
>	v_interp_p2_f32 v2, [v2], v1, 1, 6, [m0]                                     ; C8091901
>	v_sub_f32_e32 v5, 1.0, v2                                                    ; 080A04F2
>	v_interp_p1_f32 v2, v0, 0, 6, [m0]                                           ; C8081800
>	v_interp_p2_f32 v2, [v2], v1, 0, 6, [m0]                                     ; C8091801
>	v_add_f32_e32 v4, 0, v2                                                      ; 06080480
>	s_load_dwordx4 s[96:99], s[4:5], 0x0                                         ; C0B00500
>	s_load_dwordx4 s[76:79], s[4:5], 0x4                                         ; C0A60504
>	s_load_dwordx4 s[44:47], s[4:5], 0x8                                         ; C0960508
>	s_load_dwordx4 s[72:75], s[4:5], 0xc                                         ; C0A4050C
>	s_load_dwordx4 s[48:51], s[4:5], 0x10                                        ; C0980510
>	s_load_dwordx4 s[0:3], s[4:5], 0x14                                          ; C0800514
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_writelane_b32 v53, s0, 0                                                   ; 046B0000
>	v_writelane_b32 v53, s1, 1                                                   ; 046B0201
>	v_writelane_b32 v53, s2, 2                                                   ; 046B0402
>	v_writelane_b32 v53, s3, 3                                                   ; 046B0603
>	s_load_dwordx4 s[60:63], s[4:5], 0x18                                        ; C09E0518
>	s_load_dwordx4 s[0:3], s[4:5], 0x1c                                          ; C080051C
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_writelane_b32 v53, s0, 4                                                   ; 046B0800
>	v_writelane_b32 v53, s1, 5                                                   ; 046B0A01
>	v_writelane_b32 v53, s2, 6                                                   ; 046B0C02
>	v_writelane_b32 v53, s3, 7                                                   ; 046B0E03
>	s_load_dwordx4 s[36:39], s[4:5], 0x20                                        ; C0920520
>	s_load_dwordx4 s[16:19], s[4:5], 0x24                                        ; C0880524
>	s_load_dwordx4 s[32:35], s[4:5], 0x28                                        ; C0900528
>	s_load_dwordx4 s[40:43], s[4:5], 0x2c                                        ; C094052C
>	s_load_dwordx8 s[20:27], s[100:101], 0x48                                    ; C0CA6548
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample v[2:3], 5, 0, 0, 0, 0, 0, 0, 0, v[4:5], s[20:27], s[16:19]      ; F0800500 00850204
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_cmp_ge_f32_e64 s[0:1], -|v2|, 0                                            ; D00C0100 20010102
>	v_cndmask_b32_e64 v6, 0, -1, s[0:1]                                          ; D2000006 00018280
>	v_cmp_ne_i32_e64 s[0:1], v6, 0                                               ; D10A0000 00010106
>	v_cndmask_b32_e64 v6, 0, -1.0, s[0:1]                                        ; D2000006 0001E680
>	v_cmp_lt_f32_e64 s[0:1], v6, 0                                               ; D0020000 00010106
>	v_cndmask_b32_e64 v6, 0, -1, s[0:1]                                          ; D2000006 00018280
>	v_and_b32_e32 v6, 1.0, v6                                                    ; 360C0CF2
>	v_xor_b32_e32 v6, 0x80000000, v6                                             ; 3A0C0CFF 80000000
>	v_cmpx_le_f32_e32 vcc, 0, v6                                                 ; 7C260C80
>	v_cmpx_le_f32_e32 vcc, 0, v6                                                 ; 7C260C80
>	v_cmpx_le_f32_e32 vcc, 0, v6                                                 ; 7C260C80
>	v_cmpx_le_f32_e32 vcc, 0, v6                                                 ; 7C260C80
>	v_interp_p1_f32 v7, v0, 1, 0, [m0]                                           ; C81C0100
>	v_interp_p2_f32 v7, [v7], v1, 1, 0, [m0]                                     ; C81D0101
>	v_interp_p1_f32 v6, v0, 0, 0, [m0]                                           ; C8180000
>	v_interp_p2_f32 v6, [v6], v1, 0, 0, [m0]                                     ; C8190001
>	s_load_dwordx8 s[0:7], s[100:101], 0x0                                       ; C0C06500
>	s_load_dwordx8 s[88:95], s[100:101], 0x8                                     ; C0EC6508
>	s_load_dwordx8 s[8:15], s[100:101], 0x10                                     ; C0C46510
>	s_load_dwordx8 s[80:87], s[100:101], 0x18                                    ; C0E86518
>	s_load_dwordx8 s[52:59], s[100:101], 0x20                                    ; C0DA6520
>	s_load_dwordx8 s[16:23], s[100:101], 0x28                                    ; C0C86528
>	s_load_dwordx8 s[64:71], s[100:101], 0x30                                    ; C0E06530
>	s_load_dwordx8 s[24:31], s[100:101], 0x38                                    ; C0CC6538
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample v[8:11], 15, 0, 0, 0, 0, 0, 0, 0, v[6:7], s[0:7], s[96:99]      ; F0800F00 03000806
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v12, 2.0, v9, -1.0                                                 ; D282000C 03CE12F4
>	v_interp_p1_f32 v14, v0, 1, 1, [m0]                                          ; C8380500
>	v_interp_p2_f32 v14, [v14], v1, 1, 1, [m0]                                   ; C8390501
>	v_interp_p1_f32 v13, v0, 0, 1, [m0]                                          ; C8340400
>	v_interp_p2_f32 v13, [v13], v1, 0, 1, [m0]                                   ; C8350401
>	image_sample v[15:18], 15, 0, 0, 0, 0, 0, 0, 0, v[13:14], s[8:15], s[44:47]  ; F0800F00 01620F0D
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v19, 2.0, v16, -1.0                                                ; D2820013 03CE20F4
>	s_load_dwordx4 s[0:3], vcc, 0x0                                              ; C0806B00
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	s_buffer_load_dword s4, s[0:3], 0x0                                          ; C2020100
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_rcp_f32_e32 v20, s4                                                        ; 7E285404
>	v_mul_f32_e32 v20, 0x3eaa7efa, v20                                           ; 102828FF 3EAA7EFA
>	v_mul_f32_e32 v19, v19, v20                                                  ; 10262913
>	v_mad_f32 v21, v20, v12, v19                                                 ; D2820015 044E1914
>	v_interp_p1_f32 v23, v0, 1, 2, [m0]                                          ; C85C0900
>	v_interp_p2_f32 v23, [v23], v1, 1, 2, [m0]                                   ; C85D0901
>	v_interp_p1_f32 v22, v0, 0, 2, [m0]                                          ; C8580800
>	v_interp_p2_f32 v22, [v22], v1, 0, 2, [m0]                                   ; C8590801
>	image_sample v[24:27], 15, 0, 0, 0, 0, 0, 0, 0, v[22:23], s[8:15], s[44:47]  ; F0800F00 01621816
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v28, 2.0, v25, -1.0                                                ; D282001C 03CE32F4
>	v_mad_f32 v21, v20, v28, v21                                                 ; D2820015 04563914
>	v_interp_p1_f32 v29, v0, 1, 4, [m0]                                          ; C8741100
>	v_interp_p2_f32 v29, [v29], v1, 1, 4, [m0]                                   ; C8751101
>	s_buffer_load_dword s4, s[0:3], 0x21                                         ; C2020121
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v30, s4, v29                                                   ; 083C3A04
>	v_interp_p1_f32 v31, v0, 0, 4, [m0]                                          ; C87C1000
>	v_interp_p2_f32 v31, [v31], v1, 0, 4, [m0]                                   ; C87D1001
>	s_buffer_load_dword s5, s[0:3], 0x20                                         ; C2028120
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v32, s5, v31                                                   ; 08403E05
>	v_mul_f32_e32 v33, v32, v32                                                  ; 10424120
>	v_mad_f32 v33, v30, v30, v33                                                 ; D2820021 04863D1E
>	v_interp_p1_f32 v34, v0, 2, 4, [m0]                                          ; C8881200
>	v_interp_p2_f32 v34, [v34], v1, 2, 4, [m0]                                   ; C8891201
>	s_buffer_load_dword s44, s[0:3], 0x22                                        ; C2160122
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v35, s44, v34                                                  ; 0846442C
>	v_mad_f32 v33, v35, v35, v33                                                 ; D2820021 04864723
>	v_max_f32_e32 v33, 0x33d6bf95, v33                                           ; 204242FF 33D6BF95
>	v_rsq_clamp_f32_e32 v33, v33                                                 ; 7E425921
>	s_buffer_load_dword s6, s[0:3], 0x19                                         ; C2030119
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v36, v33, v30, -s6                                                 ; D2820024 801A3D21
>	s_buffer_load_dword s7, s[0:3], 0x1b                                         ; C203811B
>	v_mov_b32_e32 v37, s6                                                        ; 7E4A0206
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v36, v36, s7, v37                                                  ; D2820024 04940F24
>	s_buffer_load_dword s6, s[0:3], 0x18                                         ; C2030118
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v37, v33, v32, -s6                                                 ; D2820025 801A4121
>	v_mov_b32_e32 v38, s6                                                        ; 7E4C0206
>	v_mad_f32 v37, v37, s7, v38                                                  ; D2820025 04980F25
>	v_mul_f32_e32 v38, v37, v37                                                  ; 104C4B25
>	v_mad_f32 v38, v36, v36, v38                                                 ; D2820026 049A4924
>	s_buffer_load_dword s6, s[0:3], 0x1a                                         ; C203011A
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v39, v33, v35, -s6                                                 ; D2820027 801A4721
>	v_mov_b32_e32 v40, s6                                                        ; 7E500206
>	v_mad_f32 v39, v39, s7, v40                                                  ; D2820027 04A00F27
>	v_mad_f32 v38, v39, v39, v38                                                 ; D2820026 049A4F27
>	v_max_f32_e32 v38, 0x33d6bf95, v38                                           ; 204C4CFF 33D6BF95
>	v_rsq_clamp_f32_e32 v38, v38                                                 ; 7E4C5926
>	s_buffer_load_dword s6, s[0:3], 0x9                                          ; C2030109
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v36, v36, v38, s6                                                  ; D2820024 001A4D24
>	s_buffer_load_dword s6, s[0:3], 0xa                                          ; C203010A
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v39, v39, v38, s6                                                  ; D2820027 001A4D27
>	v_rcp_f32_e32 v39, v39                                                       ; 7E4E5527
>	v_mad_f32 v36, v36, v39, -v21                                                ; D2820024 84564F24
>	v_mad_f32 v40, 2.0, v8, -1.0                                                 ; D2820028 03CE10F4
>	v_mad_f32 v41, -2.0, v15, 1.0                                                ; D2820029 03CA1EF5
>	v_mul_f32_e32 v41, v41, v20                                                  ; 10522929
>	v_mad_f32 v42, v20, v40, v41                                                 ; D282002A 04A65114
>	v_mad_f32 v43, 2.0, v24, -1.0                                                ; D282002B 03CE30F4
>	v_mad_f32 v42, v20, v43, v42                                                 ; D282002A 04AA5714
>	s_buffer_load_dword s6, s[0:3], 0x8                                          ; C2030108
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v37, v37, v38, s6                                                  ; D2820025 001A4D25
>	v_mad_f32 v37, v37, v39, -v42                                                ; D2820025 84AA4F25
>	image_sample v6, 1, 0, 0, 0, 0, 0, 0, 0, v[6:7], s[88:95], s[76:79]          ; F0800100 02760606
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v6, 2.0, v6, -1.0                                                  ; D2820006 03CE0CF4
>	image_sample v7, 1, 0, 0, 0, 0, 0, 0, 0, v[13:14], s[80:87], s[72:75]        ; F0800100 0254070D
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v7, -2.0, v7, 1.0                                                  ; D2820007 03CA0EF5
>	v_mul_f32_e32 v13, v20, v20                                                  ; 101A2914
>	v_mul_f32_e32 v7, v7, v13                                                    ; 100E1B07
>	v_mad_f32 v6, v6, v13, v7                                                    ; D2820006 041E1B06
>	v_mul_f32_e32 v7, v12, v20                                                   ; 100E290C
>	v_mul_f32_e32 v14, v41, v7                                                   ; 101C0F29
>	v_mul_f32_e32 v38, v40, v20                                                  ; 104C2928
>	v_mad_f32 v14, v38, v19, v14                                                 ; D282000E 043A2726
>	v_add_f32_e32 v6, v14, v6                                                    ; 060C0D0E
>	image_sample v14, 1, 0, 0, 0, 0, 0, 0, 0, v[22:23], s[80:87], s[72:75]       ; F0800100 02540E16
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v14, 2.0, v14, -1.0                                                ; D282000E 03CE1CF4
>	v_mad_f32 v6, v14, v13, v6                                                   ; D2820006 041A1B0E
>	v_mul_f32_e32 v14, v43, v20                                                  ; 101C292B
>	v_mul_f32_e32 v22, v14, v7                                                   ; 102C0F0E
>	v_mul_f32_e32 v23, v28, v20                                                  ; 102E291C
>	v_mad_f32 v22, v38, v23, v22                                                 ; D2820016 045A2F26
>	v_add_f32_e32 v6, v22, v6                                                    ; 060C0D16
>	v_mul_f32_e32 v22, v14, v19                                                  ; 102C270E
>	v_mad_f32 v22, v41, v23, v22                                                 ; D2820016 045A2F29
>	v_add_f32_e32 v6, v22, v6                                                    ; 060C0D16
>	v_mad_f32 v6, -v42, v21, v6                                                  ; D2820006 241A2B2A
>	v_add_f32_e32 v22, v6, v6                                                    ; 062C0D06
>	v_mul_f32_e32 v28, v37, v22                                                  ; 10382D25
>	v_mul_f32_e32 v39, v17, v13                                                  ; 104E1B11
>	v_mad_f32 v39, v10, v13, v39                                                 ; D2820027 049E1B0A
>	v_mad_f32 v38, v20, v40, v38                                                 ; D2820026 049A5114
>	v_mad_f32 v39, v41, v38, v39                                                 ; D2820027 049E4D29
>	v_mad_f32 v39, v26, v13, v39                                                 ; D2820027 049E1B1A
>	v_mad_f32 v38, v38, v14, v39                                                 ; D2820026 049E1D26
>	v_mul_f32_e32 v14, v14, v41                                                  ; 101C530E
>	v_mad_f32 v14, 2.0, v14, v38                                                 ; D282000E 049A1CF4
>	v_mad_f32 v14, -v42, v42, v14                                                ; D282000E 243A552A
>	s_buffer_load_dword s6, s[0:3], 0x13                                         ; C2030113
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_rcp_f32_e32 v38, s6                                                        ; 7E4C5406
>	v_add_f32_e32 v39, v14, v38                                                  ; 064E4D0E
>	v_mad_f32 v28, v39, v36, -v28                                                ; D282001C 84724927
>	v_mul_f32_e32 v28, v28, v36                                                  ; 1038491C
>	v_mul_f32_e32 v36, v37, v37                                                  ; 10484B25
>	v_mul_f32_e32 v15, v18, v13                                                  ; 101E1B12
>	v_mad_f32 v8, v11, v13, v15                                                  ; D2820008 043E1B0B
>	v_mad_f32 v7, v20, v12, v7                                                   ; D2820007 041E1914
>	v_mad_f32 v8, v19, v7, v8                                                    ; D2820008 04220F13
>	v_mad_f32 v8, v27, v13, v8                                                   ; D2820008 04221B1B
>	v_mad_f32 v7, v7, v23, v8                                                    ; D2820007 04222F07
>	v_mul_f32_e32 v8, v23, v19                                                   ; 10102717
>	v_mad_f32 v7, 2.0, v8, v7                                                    ; D2820007 041E10F4
>	v_mad_f32 v7, -v21, v21, v7                                                  ; D2820007 241E2B15
>	v_add_f32_e32 v8, v7, v38                                                    ; 06104D07
>	v_mad_f32 v9, v36, v8, v28                                                   ; D2820009 04721124
>	v_mul_f32_e32 v9, 0.5, v9                                                    ; 101212F0
>	v_mul_f32_e32 v6, v6, v6                                                     ; 100C0D06
>	v_mad_f32 v8, v39, v8, -v6                                                   ; D2820008 841A1127
>	v_rcp_f32_e32 v10, v8                                                        ; 7E145508
>	v_mul_f32_e32 v11, v9, v10                                                   ; 10161509
>	v_mov_b32_e32 v12, 0xbfb8aa65                                                ; 7E1802FF BFB8AA65
>	v_mul_f32_e32 v11, v12, v11                                                  ; 1016170C
>	v_exp_f32_e32 v11, v11                                                       ; 7E164B0B
>	v_max_f32_e32 v13, 0x33d6bf95, v8                                            ; 201A10FF 33D6BF95
>	v_rsq_clamp_f32_e32 v13, v13                                                 ; 7E1A590D
>	v_mul_f32_e32 v11, v13, v11                                                  ; 1016170D
>	s_buffer_load_dword s6, s[0:3], 0x12                                         ; C2030112
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v13, s6, v11                                                   ; 101A1606
>	v_mul_f32_e32 v13, 0x3f4ccccd, v13                                           ; 101A1AFF 3F4CCCCD
>	v_mov_b32_e32 v15, 0xc1800000                                                ; 7E1E02FF C1800000
>	v_mad_f32 v9, v10, v9, v15                                                   ; D2820009 043E130A
>	v_cmp_ge_f32_e64 s[6:7], v9, 0                                               ; D00C0006 00010109
>	v_cndmask_b32_e64 v9, 0, -1, s[6:7]                                          ; D2000809 00198280
>	v_cmp_ne_i32_e64 s[6:7], v9, 0                                               ; D10A0006 00010109
>	v_mov_b32_e32 v9, 0x80000000                                                 ; 7E1202FF 80000000
>	v_cndmask_b32_e64 v10, v9, 1.0, s[6:7]                                       ; D200000A 0819E509
>	v_cmp_ge_f32_e64 s[6:7], -v8, 0                                              ; D00C0006 20010108
>	v_cndmask_b32_e64 v16, 0, -1, s[6:7]                                         ; D2000010 00198280
>	v_cmp_ne_i32_e64 s[6:7], v16, 0                                              ; D10A0006 00010110
>	v_cndmask_b32_e64 v16, v9, 1.0, s[6:7]                                       ; D2000010 0819E509
>	v_add_f32_e32 v10, v16, v10                                                  ; 06141510
>	v_cmp_ge_f32_e64 s[6:7], -v10, 0                                             ; D00C0006 2001010A
>	v_cndmask_b32_e64 v10, 0, -1, s[6:7]                                         ; D200000A 00198280
>	v_cmp_ne_i32_e64 s[46:47], v10, 0                                            ; D10A002E 0001010A
>	v_cndmask_b32_e64 v10, v9, v13, s[46:47]                                     ; D200000A 08BA1B09
>	s_buffer_load_dword s6, s[0:3], 0x15                                         ; C2030115
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v13, v33, v30, -s6                                                 ; D282000D 801A3D21
>	s_buffer_load_dword s7, s[0:3], 0x17                                         ; C2038117
>	v_mov_b32_e32 v16, s6                                                        ; 7E200206
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v13, v13, s7, v16                                                  ; D282000D 04400F0D
>	s_buffer_load_dword s6, s[0:3], 0x14                                         ; C2030114
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v16, v33, v32, -s6                                                 ; D2820010 801A4121
>	v_mov_b32_e32 v17, s6                                                        ; 7E220206
>	v_mad_f32 v16, v16, s7, v17                                                  ; D2820010 04440F10
>	v_mul_f32_e32 v17, v16, v16                                                  ; 10222110
>	v_mad_f32 v17, v13, v13, v17                                                 ; D2820011 04461B0D
>	s_buffer_load_dword s6, s[0:3], 0x16                                         ; C2030116
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v18, v33, v35, -s6                                                 ; D2820012 801A4721
>	v_mov_b32_e32 v19, s6                                                        ; 7E260206
>	v_mad_f32 v18, v18, s7, v19                                                  ; D2820012 044C0F12
>	v_mad_f32 v17, v18, v18, v17                                                 ; D2820011 04462512
>	v_max_f32_e32 v17, 0x33d6bf95, v17                                           ; 202222FF 33D6BF95
>	v_rsq_clamp_f32_e32 v17, v17                                                 ; 7E225911
>	s_buffer_load_dword s6, s[0:3], 0x5                                          ; C2030105
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v13, v13, v17, s6                                                  ; D282000D 001A230D
>	s_buffer_load_dword s7, s[0:3], 0x6                                          ; C2038106
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v18, v18, v17, s7                                                  ; D2820012 001E2312
>	v_rcp_f32_e32 v18, v18                                                       ; 7E245512
>	v_mad_f32 v13, v13, v18, -v21                                                ; D282000D 8456250D
>	s_buffer_load_dword s8, s[0:3], 0x4                                          ; C2040104
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v16, v16, v17, s8                                                  ; D2820010 00222310
>	v_mad_f32 v16, v16, v18, -v42                                                ; D2820010 84AA2510
>	v_mul_f32_e32 v17, v22, v16                                                  ; 10222116
>	s_buffer_load_dword s9, s[0:3], 0xf                                          ; C204810F
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_rcp_f32_e32 v18, s9                                                        ; 7E245409
>	v_add_f32_e32 v19, v14, v18                                                  ; 0626250E
>	v_mad_f32 v20, v19, v13, -v17                                                ; D2820014 84461B13
>	v_mul_f32_e32 v20, v20, v13                                                  ; 10281B14
>	v_mul_f32_e32 v16, v16, v16                                                  ; 10202110
>	v_add_f32_e32 v22, v7, v18                                                   ; 062C2507
>	v_mad_f32 v20, v16, v22, v20                                                 ; D2820014 04522D10
>	v_mul_f32_e32 v20, 0.5, v20                                                  ; 102828F0
>	v_mad_f32 v19, v19, v22, -v6                                                 ; D2820013 841A2D13
>	v_rcp_f32_e32 v22, v19                                                       ; 7E2C5513
>	v_mul_f32_e32 v23, v20, v22                                                  ; 102E2D14
>	v_mul_f32_e32 v23, v12, v23                                                  ; 102E2F0C
>	v_exp_f32_e32 v23, v23                                                       ; 7E2E4B17
>	v_max_f32_e32 v24, 0x33d6bf95, v19                                           ; 203026FF 33D6BF95
>	v_rsq_clamp_f32_e32 v24, v24                                                 ; 7E305918
>	v_mul_f32_e32 v23, v23, v24                                                  ; 102E3117
>	s_buffer_load_dword s9, s[0:3], 0xe                                          ; C204810E
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v24, s9, v23                                                   ; 10302E09
>	v_mad_f32 v20, v22, v20, v15                                                 ; D2820014 043E2916
>	v_cmp_ge_f32_e64 s[10:11], v20, 0                                            ; D00C000A 00010114
>	v_cndmask_b32_e64 v20, 0, -1, s[10:11]                                       ; D2000014 00298280
>	v_cmp_ne_i32_e64 s[10:11], v20, 0                                            ; D10A000A 00010114
>	v_cndmask_b32_e64 v20, v9, 1.0, s[10:11]                                     ; D2000014 0829E509
>	v_cmp_ge_f32_e64 s[10:11], -v19, 0                                           ; D00C000A 20010113
>	v_cndmask_b32_e64 v19, 0, -1, s[10:11]                                       ; D2000813 00298280
>	v_cmp_ne_i32_e64 s[10:11], v19, 0                                            ; D10A000A 00010113
>	v_cndmask_b32_e64 v19, v9, 1.0, s[10:11]                                     ; D2000813 0829E509
>	v_add_f32_e32 v19, v19, v20                                                  ; 06262913
>	v_cmp_ge_f32_e64 s[10:11], -v19, 0                                           ; D00C000A 20010113
>	v_cndmask_b32_e64 v19, 0, -1, s[10:11]                                       ; D2000813 00298280
>	v_cmp_ne_i32_e64 s[72:73], v19, 0                                            ; D10A0048 00010113
>	v_cndmask_b32_e64 v19, v9, v24, s[72:73]                                     ; D2000813 09223109
>	v_add_f32_e32 v10, v10, v19                                                  ; 0614270A
>	v_mul_f32_e32 v19, v32, v33                                                  ; 10264320
>	v_mul_f32_e32 v19, v19, v42                                                  ; 10265513
>	v_mul_f32_e32 v20, v30, v33                                                  ; 1028431E
>	v_mad_f32 v19, v21, v20, v19                                                 ; D2820013 044E2915
>	v_mad_f32 v19, v33, v35, v19                                                 ; D2820013 044E4721
>	v_sub_f32_e32 v19, 1.0, v19                                                  ; 082626F2
>	v_mul_f32_e32 v20, v19, v19                                                  ; 10282713
>	v_mul_f32_e32 v20, v20, v20                                                  ; 10282914
>	v_mul_f32_e32 v22, v20, v19                                                  ; 102C2714
>	v_interp_p1_f32 v25, v0, 1, 3, [m0]                                          ; C8640D00
>	v_interp_p2_f32 v25, [v25], v1, 1, 3, [m0]                                   ; C8650D01
>	v_interp_p1_f32 v24, v0, 0, 3, [m0]                                          ; C8600C00
>	v_interp_p2_f32 v24, [v24], v1, 0, 3, [m0]                                   ; C8610C01
>	image_sample v[26:28], 7, 0, 0, 0, 0, 0, 0, 0, v[24:25], s[64:71], s[60:63]  ; F0800700 01F01A18
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_log_f32_e32 v30, v28                                                       ; 7E3C4F1C
>	v_mov_b32_e32 v32, 0x400ccccd                                                ; 7E4002FF 400CCCCD
>	v_mul_legacy_f32_e32 v30, v32, v30                                           ; 0E3C3D20
>	v_exp_f32_e32 v30, v30                                                       ; 7E3C4B1E
>	v_mad_f32 v10, v10, v22, v30                                                 ; D282000A 047A2D0A
>	v_mov_b32_e32 v33, 0x41800000                                                ; 7E4202FF 41800000
>	v_mad_f32 v35, v18, v33, v14                                                 ; D2820023 043A4312
>	v_mad_f32 v17, v35, v13, -v17                                                ; D2820011 84461B23
>	v_mul_f32_e32 v13, v17, v13                                                  ; 101A1B11
>	v_mad_f32 v17, v18, v33, v7                                                  ; D2820011 041E4312
>	v_mad_f32 v13, v16, v17, v13                                                 ; D282000D 04362310
>	v_mul_f32_e32 v13, 0.5, v13                                                  ; 101A1AF0
>	v_mad_f32 v6, v35, v17, -v6                                                  ; D2820006 841A2323
>	v_rcp_f32_e32 v16, v6                                                        ; 7E205506
>	v_mul_f32_e32 v17, v16, v13                                                  ; 10221B10
>	v_mul_f32_e32 v12, v12, v17                                                  ; 1018230C
>	v_exp_f32_e32 v12, v12                                                       ; 7E184B0C
>	v_max_f32_e32 v17, 0x33d6bf95, v6                                            ; 20220CFF 33D6BF95
>	v_rsq_clamp_f32_e32 v17, v17                                                 ; 7E225911
>	v_mul_f32_e32 v12, v12, v17                                                  ; 1018230C
>	v_mul_f32_e32 v12, 0x3ba3d70a, v12                                           ; 101818FF 3BA3D70A
>	v_mul_f32_e32 v17, s9, v30                                                   ; 10223C09
>	v_mul_f32_e32 v17, v17, v12                                                  ; 10221911
>	v_mad_f32 v13, v13, v16, v15                                                 ; D282000D 043E210D
>	v_cmp_ge_f32_e64 s[10:11], v13, 0                                            ; D00C000A 0001010D
>	v_cndmask_b32_e64 v13, 0, -1, s[10:11]                                       ; D200080D 00298280
>	v_cmp_ne_i32_e64 s[10:11], v13, 0                                            ; D10A000A 0001010D
>	v_cndmask_b32_e64 v13, v9, 1.0, s[10:11]                                     ; D200080D 0829E509
>	v_cmp_ge_f32_e64 s[10:11], -v6, 0                                            ; D00C000A 20010106
>	v_cndmask_b32_e64 v6, 0, -1, s[10:11]                                        ; D2000006 00298280
>	v_cmp_ne_i32_e64 s[10:11], v6, 0                                             ; D10A000A 00010106
>	v_cndmask_b32_e64 v6, v9, 1.0, s[10:11]                                      ; D2000006 0829E509
>	v_add_f32_e32 v6, v6, v13                                                    ; 060C1B06
>	v_cmp_ge_f32_e64 s[10:11], -v6, 0                                            ; D00C000A 20010106
>	v_cndmask_b32_e64 v6, 0, -1, s[10:11]                                        ; D2000006 00298280
>	v_cmp_ne_i32_e64 s[60:61], v6, 0                                             ; D10A003C 00010106
>	v_cndmask_b32_e64 v6, v9, v17, s[60:61]                                      ; D2000006 08F22309
>	v_mul_f32_e32 v13, s8, v42                                                   ; 101A5408
>	v_mad_f32 v13, s6, v21, v13                                                  ; D282000D 04362A06
>	v_add_f32_e32 v13, s7, v13                                                   ; 061A1A07
>	v_add_f32_e64 v13, 0, v13 clamp                                              ; D206080D 00021A80
>	v_mad_f32 v6, v10, v13, v6                                                   ; D2820006 041A1B0A
>	s_buffer_load_dword s6, s[0:3], 0x29                                         ; C2030129
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v10, s6, v29                                                   ; 10143A06
>	s_buffer_load_dword s6, s[0:3], 0x25                                         ; C2030125
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v10, v31, s6, v10                                                  ; D282000A 04280D1F
>	s_buffer_load_dword s6, s[0:3], 0x2d                                         ; C203012D
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v10, v34, s6, v10                                                  ; D282000A 04280D22
>	s_buffer_load_dword s6, s[0:3], 0x31                                         ; C2030131
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_add_f32_e32 v10, s6, v10                                                   ; 06141406
>	s_buffer_load_dword s6, s[0:3], 0x2b                                         ; C203012B
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v15, s6, v29                                                   ; 101E3A06
>	s_buffer_load_dword s6, s[0:3], 0x27                                         ; C2030127
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v15, v31, s6, v15                                                  ; D282000F 043C0D1F
>	s_buffer_load_dword s6, s[0:3], 0x2f                                         ; C203012F
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v15, v34, s6, v15                                                  ; D282000F 043C0D22
>	s_buffer_load_dword s6, s[0:3], 0x33                                         ; C2030133
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_add_f32_e32 v15, s6, v15                                                   ; 061E1E06
>	v_rcp_f32_e32 v15, v15                                                       ; 7E1E550F
>	v_mul_f32_e32 v10, v10, v15                                                  ; 10141F0A
>	v_mad_f32 v10, -0.5, v10, -0.5                                               ; D282000A 03C614F1
>	v_mov_b32_e32 v16, 0x40c00000                                                ; 7E2002FF 40C00000
>	v_mad_f32 v7, v7, v16, v10                                                   ; D2820007 042A2107
>	v_mad_f32 v18, -2.0, v8, v7                                                  ; D2820012 041E10F5
>	s_buffer_load_dword s6, s[0:3], 0x28                                         ; C2030128
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v10, s6, v29                                                   ; 10143A06
>	s_buffer_load_dword s6, s[0:3], 0x24                                         ; C2030124
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v10, v31, s6, v10                                                  ; D282000A 04280D1F
>	s_buffer_load_dword s6, s[0:3], 0x2c                                         ; C203012C
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v10, v34, s6, v10                                                  ; D282000A 04280D22
>	s_buffer_load_dword s6, s[0:3], 0x30                                         ; C2030130
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_add_f32_e32 v10, s6, v10                                                   ; 06141406
>	v_mul_f32_e32 v10, v10, v15                                                  ; 10141F0A
>	v_mad_f32 v10, 0.5, v10, -0.5                                                ; D282000A 03C614F0
>	v_mad_f32 v10, v14, v16, v10                                                 ; D282000A 042A210E
>	v_mad_f32 v17, 2.0, v8, v10                                                  ; D2820011 042A10F4
>	image_sample v[35:38], 15, 0, 0, 0, 0, 0, 0, 0, v[17:18], s[52:59], s[48:51] ; F0800F00 018D2311
>	v_mov_b32_e32 v14, 0x3e800000                                                ; 7E1C02FF 3E800000
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mul_f32_e32 v15, 0x3e800000, v37                                           ; 101E4AFF 3E800000
>	v_mad_f32 v21, 2.0, v8, v7                                                   ; D2820015 041E10F4
>	v_mov_b32_e32 v39, v17                                                       ; 7E4E0311
>	v_mov_b32_e32 v40, v18                                                       ; 7E500312
>	v_mov_b32_e32 v40, v21                                                       ; 7E500315
>	image_sample v[39:42], 15, 0, 0, 0, 0, 0, 0, 0, v[39:40], s[52:59], s[48:51] ; F0800F00 018D2727
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v15, v41, v14, v15                                                 ; D282000F 043E1D29
>	v_mad_f32 v44, v8, v16, v7                                                   ; D282002C 041E2108
>	v_mad_f32 v43, -2.0, v8, v10                                                 ; D282002B 042A10F5
>	image_sample v[45:48], 15, 0, 0, 0, 0, 0, 0, 0, v[43:44], s[52:59], s[48:51] ; F0800F00 018D2D2B
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v7, v47, v14, v15                                                  ; D2820007 043E1D2F
>	v_mov_b32_e32 v44, v18                                                       ; 7E580312
>	image_sample v[15:18], 15, 0, 0, 0, 0, 0, 0, 0, v[43:44], s[52:59], s[48:51] ; F0800F00 018D0F2B
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v7, v17, v14, v7                                                   ; D2820007 041E1D11
>	v_log_f32_e64 v7, |v7|                                                       ; D34E0107 00000107
>	v_mul_f32_e32 v7, 0x400ccccd, v7                                             ; 100E0EFF 400CCCCD
>	v_exp_f32_e32 v7, v7                                                         ; 7E0E4B07
>	s_buffer_load_dword s6, s[0:3], 0x38                                         ; C2030138
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v6, v7, s6, v6                                                     ; D2820006 04180D07
>	v_mad_f32 v7, v19, v20, 0.5                                                  ; D2820007 03C22913
>	v_mul_f32_e32 v6, v6, v7                                                     ; 100C0F06
>	v_log_f32_e64 v6, |v6|                                                       ; D34E0106 00000106
>	v_mul_f32_e32 v6, 0x3ee8ba1f, v6                                             ; 100C0CFF 3EE8BA1F
>	v_exp_f32_e32 v51, v6                                                        ; 7E664B06
>	s_buffer_load_dword s7, s[0:3], 0x11                                         ; C2038111
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v6, s7, v11                                                    ; 100C1607
>	v_mul_f32_e32 v6, 0x3f4ccccd, v6                                             ; 100C0CFF 3F4CCCCD
>	v_cndmask_b32_e64 v6, v9, v6, s[46:47]                                       ; D2000006 08BA0D09
>	s_buffer_load_dword s7, s[0:3], 0xd                                          ; C203810D
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v8, s7, v23                                                    ; 10102E07
>	v_cndmask_b32_e64 v8, v9, v8, s[72:73]                                       ; D2000008 09221109
>	v_add_f32_e32 v6, v6, v8                                                     ; 060C1106
>	v_log_f32_e32 v8, v27                                                        ; 7E104F1B
>	v_mul_legacy_f32_e32 v8, v32, v8                                             ; 0E101120
>	v_exp_f32_e32 v8, v8                                                         ; 7E104B08
>	v_mad_f32 v6, v6, v22, v8                                                    ; D2820006 04222D06
>	v_mul_f32_e32 v8, s7, v8                                                     ; 10101007
>	v_mul_f32_e32 v8, v8, v12                                                    ; 10101908
>	v_cndmask_b32_e64 v8, v9, v8, s[60:61]                                       ; D2000008 08F21109
>	v_mad_f32 v6, v6, v13, v8                                                    ; D2820006 04221B06
>	v_mul_f32_e32 v8, 0x3e800000, v36                                            ; 101048FF 3E800000
>	v_mad_f32 v8, v40, v14, v8                                                   ; D2820008 04221D28
>	v_mad_f32 v8, v46, v14, v8                                                   ; D2820008 04221D2E
>	v_mad_f32 v8, v16, v14, v8                                                   ; D2820008 04221D10
>	v_log_f32_e64 v8, |v8|                                                       ; D34E0108 00000108
>	v_mul_f32_e32 v8, 0x400ccccd, v8                                             ; 101010FF 400CCCCD
>	v_exp_f32_e32 v8, v8                                                         ; 7E104B08
>	v_mad_f32 v6, v8, s6, v6                                                     ; D2820006 04180D08
>	v_mul_f32_e32 v6, v6, v7                                                     ; 100C0F06
>	v_log_f32_e64 v6, |v6|                                                       ; D34E0106 00000106
>	v_mul_f32_e32 v6, 0x3ee8ba1f, v6                                             ; 100C0CFF 3EE8BA1F
>	v_exp_f32_e32 v50, v6                                                        ; 7E644B06
>	s_buffer_load_dword s7, s[0:3], 0x10                                         ; C2038110
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v6, s7, v11                                                    ; 100C1607
>	v_mul_f32_e32 v6, 0x3f4ccccd, v6                                             ; 100C0CFF 3F4CCCCD
>	v_cndmask_b32_e64 v6, v9, v6, s[46:47]                                       ; D2000006 08BA0D09
>	s_buffer_load_dword s7, s[0:3], 0xc                                          ; C203810C
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v8, s7, v23                                                    ; 10102E07
>	v_cndmask_b32_e64 v8, v9, v8, s[72:73]                                       ; D2000008 09221109
>	v_add_f32_e32 v6, v6, v8                                                     ; 060C1106
>	v_log_f32_e32 v8, v26                                                        ; 7E104F1A
>	v_mul_legacy_f32_e32 v8, v32, v8                                             ; 0E101120
>	v_exp_f32_e32 v8, v8                                                         ; 7E104B08
>	v_mad_f32 v6, v6, v22, v8                                                    ; D2820006 04222D06
>	v_mul_f32_e32 v8, s7, v8                                                     ; 10101007
>	v_mul_f32_e32 v8, v8, v12                                                    ; 10101908
>	v_cndmask_b32_e64 v8, v9, v8, s[60:61]                                       ; D2000008 08F21109
>	v_mad_f32 v6, v6, v13, v8                                                    ; D2820006 04221B06
>	v_mul_f32_e32 v8, 0x3e800000, v35                                            ; 101046FF 3E800000
>	v_mad_f32 v8, v39, v14, v8                                                   ; D2820008 04221D27
>	v_mad_f32 v8, v45, v14, v8                                                   ; D2820008 04221D2D
>	v_mad_f32 v8, v15, v14, v8                                                   ; D2820008 04221D0F
>	v_log_f32_e64 v8, |v8|                                                       ; D34E0108 00000108
>	v_mul_f32_e32 v8, 0x400ccccd, v8                                             ; 101010FF 400CCCCD
>	v_exp_f32_e32 v8, v8                                                         ; 7E104B08
>	v_mad_f32 v6, v8, s6, v6                                                     ; D2820006 04180D08
>	v_mul_f32_e32 v6, v6, v7                                                     ; 100C0F06
>	v_log_f32_e64 v6, |v6|                                                       ; D34E0106 00000106
>	v_mul_f32_e32 v6, 0x3ee8ba1f, v6                                             ; 100C0CFF 3EE8BA1F
>	v_exp_f32_e32 v49, v6                                                        ; 7E624B06
>	s_load_dwordx8 s[8:15], s[100:101], 0x58                                     ; C0C46558
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample v[6:8], 7, 0, 0, 0, 0, 0, 0, 0, v[49:52], s[8:15], s[40:43]     ; F0800700 01420631
>	s_load_dwordx8 s[8:15], s[100:101], 0x40                                     ; C0C46540
>	s_waitcnt vmcnt(0) lgkmcnt(0)                                                ; BF8C0070
>	image_sample v[9:12], 15, 0, 0, 0, 0, 0, 0, 0, v[4:5], s[8:15], s[36:39]     ; F0800F00 01220904
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_sub_f32_e32 v4, 1.0, v12                                                   ; 080818F2
>	v_mad_f32 v5, v8, v4, v11                                                    ; D2820005 042E0908
>	v_sub_f32_e32 v13, 1.0, v2                                                   ; 081A04F2
>	v_mul_f32_e32 v13, 0x3ecccccd, v13                                           ; 101A1AFF 3ECCCCCD
>	v_mad_f32 v21, v2, v5, v13                                                   ; D2820015 04360B02
>	v_mad_f32 v5, v7, v4, v10                                                    ; D2820005 042A0907
>	v_mad_f32 v20, v2, v5, v13                                                   ; D2820014 04360B02
>	v_mad_f32 v4, v6, v4, v9                                                     ; D2820004 04260906
>	v_mad_f32 v19, v2, v4, v13                                                   ; D2820013 04360902
>	s_load_dwordx8 s[8:15], s[100:101], 0x50                                     ; C0C46550
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample v[4:6], 7, 0, 0, 0, 0, 0, 0, 0, v[19:22], s[8:15], s[32:35]     ; F0800700 01020413
>	v_mad_f32 v2, -v3, v2, v2                                                    ; D2820002 240A0503
>	v_sub_f32_e32 v3, 1.0, v2                                                    ; 080604F2
>	v_mul_f32_e32 v7, v20, v3                                                    ; 100E0714
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v7, v2, v5, v7                                                     ; D2820007 041E0B02
>	s_buffer_load_dword s6, s[0:3], 0x35                                         ; C2030135
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v8, s6, v7                                                     ; 08100E06
>	v_subrev_f32_e32 v9, s4, v29                                                 ; 0A123A04
>	v_subrev_f32_e32 v10, s5, v31                                                ; 0A143E05
>	v_mul_f32_e32 v10, v10, v10                                                  ; 1014150A
>	v_mad_f32 v9, v9, v9, v10                                                    ; D2820009 042A1309
>	v_subrev_f32_e32 v10, s44, v34                                               ; 0A14442C
>	v_mad_f32 v9, v10, v10, v9                                                   ; D2820009 0426150A
>	s_buffer_load_dword s4, s[0:3], 0x1d                                         ; C202011D
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v9, s4, v9                                                     ; 10121204
>	s_buffer_load_dword s4, s[0:3], 0x1c                                         ; C202011C
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v11, s4, v10                                                   ; 10161404
>	v_mul_f32_e32 v11, 0x3fb8aa65, v11                                           ; 101616FF 3FB8AA65
>	v_exp_f32_e32 v11, v11                                                       ; 7E164B0B
>	v_sub_f32_e32 v11, 1.0, v11                                                  ; 081616F2
>	v_mul_f32_e32 v9, v9, v11                                                    ; 10121709
>	v_rcp_f32_e32 v10, v10                                                       ; 7E14550A
>	v_mul_f32_e32 v9, v9, v10                                                    ; 10121509
>	v_mul_f32_e32 v9, 0x3fb8aa65, v9                                             ; 101212FF 3FB8AA65
>	v_exp_f32_e32 v9, v9                                                         ; 7E124B09
>	v_add_f32_e64 v9, 0, v9 clamp                                                ; D2060809 00021280
>	v_sub_f32_e32 v9, 1.0, v9                                                    ; 081212F2
>	v_interp_p1_f32 v10, v0, 1, 5, [m0]                                          ; C8281500
>	v_interp_p2_f32 v10, [v10], v1, 1, 5, [m0]                                   ; C8291501
>	v_interp_p1_f32 v11, v0, 2, 5, [m0]                                          ; C82C1600
>	v_interp_p2_f32 v11, [v11], v1, 2, 5, [m0]                                   ; C82D1601
>	v_rcp_f32_e32 v0, v11                                                        ; 7E00550B
>	v_mul_f32_e32 v0, v10, v0                                                    ; 1000010A
>	v_mad_f32 v0, 0.5, v0, 0.5                                                   ; D2820000 03C200F0
>	s_buffer_load_dword s4, s[0:3], 0x1f                                         ; C202011F
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mov_b32_e32 v1, s4                                                         ; 7E020204
>	v_readlane_b32 s4, v53, 4                                                    ; 02090935
>	v_readlane_b32 s5, v53, 5                                                    ; 020B0B35
>	v_readlane_b32 s6, v53, 6                                                    ; 020D0D35
>	v_readlane_b32 s7, v53, 7                                                    ; 020F0F35
>	s_nop 2                                                                      ; BF800002
>	image_sample v0, 1, 0, 0, 0, 0, 0, 0, 0, v[0:1], s[24:31], s[4:7]            ; F0800100 00260000
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mul_f32_e32 v0, v0, v9                                                     ; 10001300
>	v_mad_f32 v1, v0, v8, v7                                                     ; D2820001 041E1100
>	v_mul_f32_e32 v7, v19, v3                                                    ; 100E0713
>	v_mad_f32 v7, v2, v4, v7                                                     ; D2820007 041E0902
>	s_buffer_load_dword s4, s[0:3], 0x34                                         ; C2020134
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v8, s4, v7                                                     ; 08100E04
>	v_mad_f32 v7, v0, v8, v7                                                     ; D2820007 041E1100
>	v_cvt_pkrtz_f16_f32_e32 v1, v7, v1                                           ; 5E020307
>	v_mul_f32_e32 v3, v21, v3                                                    ; 10060715
>	v_mad_f32 v2, v2, v6, v3                                                     ; D2820002 040E0D02
>	s_buffer_load_dword s0, s[0:3], 0x36                                         ; C2000136
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v3, s0, v2                                                     ; 08060400
>	v_mad_f32 v0, v0, v3, v2                                                     ; D2820000 040A0700
>	v_mul_f32_e32 v2, 0x3e800000, v38                                            ; 10044CFF 3E800000
>	v_mad_f32 v2, v42, v14, v2                                                   ; D2820002 040A1D2A
>	v_mad_f32 v2, v48, v14, v2                                                   ; D2820002 040A1D30
>	v_mad_f32 v2, v18, v14, v2                                                   ; D2820002 040A1D12
>	v_add_f32_e32 v2, 1.0, v2                                                    ; 060404F2
>	v_add_f32_e64 v2, 0, v2 clamp                                                ; D2060802 00020480
>	v_readlane_b32 s0, v53, 0                                                    ; 02010135
>	v_readlane_b32 s1, v53, 1                                                    ; 02030335
>	v_readlane_b32 s2, v53, 2                                                    ; 02050535
>	v_readlane_b32 s3, v53, 3                                                    ; 02070735
>	s_nop 2                                                                      ; BF800002
>	image_sample v3, 8, 0, 0, 0, 0, 0, 0, 0, v[24:25], s[16:23], s[0:3]          ; F0800800 00040318
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mul_f32_e32 v2, v2, v3                                                     ; 10040702
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v2                                           ; 5E000500
>	exp 15, 0, 1, 1, 1, v1, v0, v1, v0                                           ; F8001C0F 00010001
>	s_endpgm                                                                     ; BF810000
>
>SGPRS: 104
>VGPRS: 56
>Code Size: 3340 bytes
>LDS: 0 blocks
>Scratch: 2048 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL OUT[2], GENERIC[10]
>DCL OUT[3], GENERIC[11]
>DCL OUT[4], GENERIC[12]
>DCL OUT[5], GENERIC[13]
>DCL OUT[6], GENERIC[14]
>DCL CONST[0..13]
>DCL TEMP[0..9], LOCAL
>IMM[0] FLT32 {    2.0000,    -1.0000,     0.5000,    -0.5000}
>IMM[1] FLT32 {    1.0000,     0.0000,     0.0000,     0.0000}
>  0: MAD TEMP[0].xyz, IN[1], IMM[0].xxxx, IMM[0].yyyy
>  1: MOV TEMP[0].xyz, TEMP[0].xyzx
>  2: MAD TEMP[1].x, IN[0].xxxx, IMM[0].zzzz, IMM[0].zzzz
>  3: MOV TEMP[2].xy, CONST[4].xyxx
>  4: ADD TEMP[3].yz, -TEMP[2].xxyw, CONST[5].xxyw
>  5: MAD TEMP[4].x, TEMP[1].xxxx, TEMP[3].yyyy, CONST[4].xxxx
>  6: MOV TEMP[2].x, TEMP[4].xxxx
>  7: MAD TEMP[5].x, IN[0].yyyy, IMM[0].zzzz, IMM[0].zzzz
>  8: MAD TEMP[3].y, TEMP[5].xxxx, TEMP[3].zzzz, CONST[4].yyyy
>  9: MOV TEMP[2].y, TEMP[3].yyyy
> 10: MAD TEMP[5].z, IN[1].wwww, CONST[6].xxxx, CONST[7].xxxx
> 11: MOV TEMP[2].z, TEMP[5].zzzz
> 12: ADD TEMP[6].xyz, -TEMP[2], CONST[12]
> 13: MOV TEMP[6].xyz, TEMP[6].xyzx
> 14: MUL TEMP[1], TEMP[3].yyyy, CONST[1]
> 15: MAD TEMP[1], TEMP[4].xxxx, CONST[0], TEMP[1]
> 16: MAD TEMP[1], TEMP[5].zzzz, CONST[2], TEMP[1]
> 17: ADD TEMP[1], TEMP[1], CONST[3]
> 18: RCP TEMP[7].x, TEMP[1].wwww
> 19: MOV TEMP[2].w, TEMP[7].xxxx
> 20: MUL TEMP[7].xy, TEMP[1], TEMP[7].xxxx
> 21: MOV TEMP[7].xy, TEMP[7].xyxx
> 22: MOV TEMP[8], TEMP[1]
> 23: ADD TEMP[9].xy, TEMP[2], CONST[13].zwzw
> 24: MOV TEMP[1].xy, TEMP[9].xyxx
> 25: MUL TEMP[9].xy, TEMP[1], CONST[13]
> 26: MOV TEMP[9].xy, TEMP[9].xyxx
> 27: MUL TEMP[3].xy, TEMP[3].yyyy, CONST[9]
> 28: MOV TEMP[1].xy, TEMP[3].xyxx
> 29: MOV TEMP[2].xyz, TEMP[2].xyzx
> 30: MAD TEMP[3].xy, TEMP[4].xxxx, CONST[8], TEMP[1]
> 31: MOV TEMP[1].xy, TEMP[3].xyxx
> 32: MAD TEMP[3].xy, TEMP[5].zzzz, CONST[10], TEMP[1]
> 33: MOV TEMP[1].xy, TEMP[3].xyxx
> 34: ADD TEMP[3].xy, TEMP[1], CONST[11]
> 35: MOV TEMP[1].xy, TEMP[3].xyxx
> 36: MAD TEMP[1].xy, TEMP[1], IMM[0].zwyw, IMM[0].wwww
> 37: MOV TEMP[1].xy, TEMP[1].xyxx
> 38: MOV TEMP[0].w, IMM[1].xxxx
> 39: MOV TEMP[2].w, IMM[1].xxxx
> 40: MOV TEMP[6].w, IMM[1].xxxx
> 41: MOV TEMP[7].zw, IMM[1].xxyx
> 42: MOV TEMP[1].zw, IMM[1].xxyx
> 43: MOV TEMP[9].zw, IMM[1].xxyx
> 44: MOV OUT[6], TEMP[9]
> 45: MOV OUT[1], TEMP[0]
> 46: MOV OUT[2], TEMP[2]
> 47: MOV OUT[0], TEMP[8]
> 48: MOV OUT[3], TEMP[6]
> 49: MOV OUT[4], TEMP[7]
> 50: MOV OUT[5], TEMP[1]
> 51: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 0)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 4)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %12, i32 8)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %12, i32 12)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %12, i32 16)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %12, i32 20)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %12, i32 24)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %12, i32 28)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %12, i32 32)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %12, i32 36)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %12, i32 40)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %12, i32 44)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %12, i32 48)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %12, i32 52)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %12, i32 56)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %12, i32 60)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %12, i32 64)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %12, i32 68)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %12, i32 80)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %12, i32 84)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %12, i32 96)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %12, i32 112)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %12, i32 128)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %12, i32 132)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %12, i32 144)
>  %38 = call float @llvm.SI.load.const(<16 x i8> %12, i32 148)
>  %39 = call float @llvm.SI.load.const(<16 x i8> %12, i32 160)
>  %40 = call float @llvm.SI.load.const(<16 x i8> %12, i32 164)
>  %41 = call float @llvm.SI.load.const(<16 x i8> %12, i32 176)
>  %42 = call float @llvm.SI.load.const(<16 x i8> %12, i32 180)
>  %43 = call float @llvm.SI.load.const(<16 x i8> %12, i32 192)
>  %44 = call float @llvm.SI.load.const(<16 x i8> %12, i32 196)
>  %45 = call float @llvm.SI.load.const(<16 x i8> %12, i32 200)
>  %46 = call float @llvm.SI.load.const(<16 x i8> %12, i32 208)
>  %47 = call float @llvm.SI.load.const(<16 x i8> %12, i32 212)
>  %48 = call float @llvm.SI.load.const(<16 x i8> %12, i32 216)
>  %49 = call float @llvm.SI.load.const(<16 x i8> %12, i32 220)
>  %50 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %51 = load <16 x i8> addrspace(2)* %50, !tbaa !0
>  %52 = add i32 %5, %7
>  %53 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %51, i32 0, i32 %52)
>  %54 = extractelement <4 x float> %53, i32 0
>  %55 = extractelement <4 x float> %53, i32 1
>  %56 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %57 = load <16 x i8> addrspace(2)* %56, !tbaa !0
>  %58 = add i32 %5, %7
>  %59 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %57, i32 0, i32 %58)
>  %60 = extractelement <4 x float> %59, i32 0
>  %61 = extractelement <4 x float> %59, i32 1
>  %62 = extractelement <4 x float> %59, i32 2
>  %63 = extractelement <4 x float> %59, i32 3
>  %64 = fmul float %60, 2.000000e+00
>  %65 = fadd float %64, -1.000000e+00
>  %66 = fmul float %61, 2.000000e+00
>  %67 = fadd float %66, -1.000000e+00
>  %68 = fmul float %62, 2.000000e+00
>  %69 = fadd float %68, -1.000000e+00
>  %70 = fmul float %54, 5.000000e-01
>  %71 = fadd float %70, 5.000000e-01
>  %72 = fsub float -0.000000e+00, %29
>  %73 = fadd float %72, %31
>  %74 = fsub float -0.000000e+00, %30
>  %75 = fadd float %74, %32
>  %76 = fmul float %71, %73
>  %77 = fadd float %76, %29
>  %78 = fmul float %55, 5.000000e-01
>  %79 = fadd float %78, 5.000000e-01
>  %80 = fmul float %79, %75
>  %81 = fadd float %80, %30
>  %82 = fmul float %63, %33
>  %83 = fadd float %82, %34
>  %84 = fsub float -0.000000e+00, %77
>  %85 = fadd float %84, %43
>  %86 = fsub float -0.000000e+00, %81
>  %87 = fadd float %86, %44
>  %88 = fsub float -0.000000e+00, %83
>  %89 = fadd float %88, %45
>  %90 = fmul float %81, %17
>  %91 = fmul float %81, %18
>  %92 = fmul float %81, %19
>  %93 = fmul float %81, %20
>  %94 = fmul float %77, %13
>  %95 = fadd float %94, %90
>  %96 = fmul float %77, %14
>  %97 = fadd float %96, %91
>  %98 = fmul float %77, %15
>  %99 = fadd float %98, %92
>  %100 = fmul float %77, %16
>  %101 = fadd float %100, %93
>  %102 = fmul float %83, %21
>  %103 = fadd float %102, %95
>  %104 = fmul float %83, %22
>  %105 = fadd float %104, %97
>  %106 = fmul float %83, %23
>  %107 = fadd float %106, %99
>  %108 = fmul float %83, %24
>  %109 = fadd float %108, %101
>  %110 = fadd float %103, %25
>  %111 = fadd float %105, %26
>  %112 = fadd float %107, %27
>  %113 = fadd float %109, %28
>  %114 = fdiv float 1.000000e+00, %113
>  %115 = fmul float %110, %114
>  %116 = fmul float %111, %114
>  %117 = fadd float %77, %48
>  %118 = fadd float %81, %49
>  %119 = fmul float %117, %46
>  %120 = fmul float %118, %47
>  %121 = fmul float %81, %37
>  %122 = fmul float %81, %38
>  %123 = fmul float %77, %35
>  %124 = fadd float %123, %121
>  %125 = fmul float %77, %36
>  %126 = fadd float %125, %122
>  %127 = fmul float %83, %39
>  %128 = fadd float %127, %124
>  %129 = fmul float %83, %40
>  %130 = fadd float %129, %126
>  %131 = fadd float %128, %41
>  %132 = fadd float %130, %42
>  %133 = fmul float %131, 5.000000e-01
>  %134 = fadd float %133, -5.000000e-01
>  %135 = fmul float %132, -5.000000e-01
>  %136 = fadd float %135, -5.000000e-01
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %65, float %67, float %69, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %77, float %81, float %83, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 34, i32 0, float %85, float %87, float %89, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 35, i32 0, float %115, float %116, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 36, i32 0, float %134, float %136, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 37, i32 0, float %119, float %120, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %110, float %111, float %112, float %113)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                            ; 4A00000A
>	s_load_dwordx4 s[4:7], s[8:9], 0x0                   ; C0820900
>	s_load_dwordx4 s[8:11], s[8:9], 0x4                  ; C0840904
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[8:11], 0 idxen ; E00C2000 80020100
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_mad_f32 v5, 2.0, v3, -1.0                          ; D2820005 03CE06F4
>	v_mad_f32 v6, 2.0, v2, -1.0                          ; D2820006 03CE04F4
>	v_mad_f32 v7, 2.0, v1, -1.0                          ; D2820007 03CE02F4
>	v_mov_b32_e32 v8, 1.0                                ; 7E1002F2
>	exp 15, 32, 0, 0, 0, v7, v6, v5, v8                  ; F800020F 08050607
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                   ; C0800300
>	s_waitcnt expcnt(0) lgkmcnt(0)                       ; BF8C000F
>	s_buffer_load_dword s8, s[0:3], 0x11                 ; C2040111
>	s_buffer_load_dword s9, s[0:3], 0x15                 ; C2048115
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mov_b32_e32 v5, s8                                 ; 7E0A0208
>	v_sub_f32_e32 v5, s9, v5                             ; 080A0A09
>	buffer_load_format_xyzw v[9:12], v0, s[4:7], 0 idxen ; E00C2000 80010900
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_mad_f32 v0, 0.5, v10, 0.5                          ; D2820000 03C214F0
>	v_mad_f32 v0, v0, v5, s8                             ; D2820000 00220B00
>	s_buffer_load_dword s4, s[0:3], 0x10                 ; C2020110
>	s_buffer_load_dword s5, s[0:3], 0x14                 ; C2028114
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mov_b32_e32 v5, s4                                 ; 7E0A0204
>	v_sub_f32_e32 v5, s5, v5                             ; 080A0A05
>	v_mad_f32 v6, 0.5, v9, 0.5                           ; D2820006 03C212F0
>	v_mad_f32 v5, v6, v5, s4                             ; D2820005 00120B06
>	s_buffer_load_dword s4, s[0:3], 0x1c                 ; C202011C
>	s_buffer_load_dword s5, s[0:3], 0x18                 ; C2028118
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mov_b32_e32 v6, s4                                 ; 7E0C0204
>	v_mad_f32 v1, s5, v4, v6                             ; D2820001 041A0805
>	exp 15, 33, 0, 0, 0, v5, v0, v1, v8                  ; F800021F 08010005
>	s_buffer_load_dword s4, s[0:3], 0x31                 ; C2020131
>	s_waitcnt expcnt(0) lgkmcnt(0)                       ; BF8C000F
>	v_sub_f32_e32 v2, s4, v0                             ; 08040004
>	s_buffer_load_dword s4, s[0:3], 0x30                 ; C2020130
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_sub_f32_e32 v3, s4, v5                             ; 08060A04
>	s_buffer_load_dword s4, s[0:3], 0x32                 ; C2020132
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_sub_f32_e32 v4, s4, v1                             ; 08080204
>	exp 15, 34, 0, 0, 0, v3, v2, v4, v8                  ; F800022F 08040203
>	s_buffer_load_dword s4, s[0:3], 0x5                  ; C2020105
>	s_waitcnt expcnt(0) lgkmcnt(0)                       ; BF8C000F
>	v_mul_f32_e32 v2, s4, v0                             ; 10040004
>	s_buffer_load_dword s4, s[0:3], 0x1                  ; C2020101
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v2, v5, s4, v2                             ; D2820002 04080905
>	s_buffer_load_dword s4, s[0:3], 0x9                  ; C2020109
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v2, v1, s4, v2                             ; D2820002 04080901
>	s_buffer_load_dword s4, s[0:3], 0xd                  ; C202010D
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_add_f32_e32 v2, s4, v2                             ; 06040404
>	s_buffer_load_dword s4, s[0:3], 0x7                  ; C2020107
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v3, s4, v0                             ; 10060004
>	s_buffer_load_dword s4, s[0:3], 0x3                  ; C2020103
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v3, v5, s4, v3                             ; D2820003 040C0905
>	s_buffer_load_dword s4, s[0:3], 0xb                  ; C202010B
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v3, v1, s4, v3                             ; D2820003 040C0901
>	s_buffer_load_dword s4, s[0:3], 0xf                  ; C202010F
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_add_f32_e32 v3, s4, v3                             ; 06060604
>	v_rcp_f32_e32 v4, v3                                 ; 7E085503
>	v_mul_f32_e32 v6, v4, v2                             ; 100C0504
>	s_buffer_load_dword s4, s[0:3], 0x4                  ; C2020104
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v7, s4, v0                             ; 100E0004
>	s_buffer_load_dword s4, s[0:3], 0x0                  ; C2020100
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v7, v5, s4, v7                             ; D2820007 041C0905
>	s_buffer_load_dword s4, s[0:3], 0x8                  ; C2020108
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v7, v1, s4, v7                             ; D2820007 041C0901
>	s_buffer_load_dword s4, s[0:3], 0xc                  ; C202010C
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_add_f32_e32 v7, s4, v7                             ; 060E0E04
>	v_mul_f32_e32 v4, v4, v7                             ; 10080F04
>	v_mov_b32_e32 v9, 0                                  ; 7E120280
>	exp 15, 35, 0, 0, 0, v4, v6, v9, v8                  ; F800023F 08090604
>	s_buffer_load_dword s4, s[0:3], 0x24                 ; C2020124
>	s_waitcnt expcnt(0) lgkmcnt(0)                       ; BF8C000F
>	v_mul_f32_e32 v4, s4, v0                             ; 10080004
>	s_buffer_load_dword s4, s[0:3], 0x20                 ; C2020120
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v4, v5, s4, v4                             ; D2820004 04100905
>	s_buffer_load_dword s4, s[0:3], 0x28                 ; C2020128
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v4, v1, s4, v4                             ; D2820004 04100901
>	s_buffer_load_dword s4, s[0:3], 0x2c                 ; C202012C
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_add_f32_e32 v4, s4, v4                             ; 06080804
>	v_mad_f32 v4, 0.5, v4, -0.5                          ; D2820004 03C608F0
>	s_buffer_load_dword s4, s[0:3], 0x25                 ; C2020125
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v6, s4, v0                             ; 100C0004
>	s_buffer_load_dword s4, s[0:3], 0x21                 ; C2020121
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v6, v5, s4, v6                             ; D2820006 04180905
>	s_buffer_load_dword s4, s[0:3], 0x29                 ; C2020129
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v6, v1, s4, v6                             ; D2820006 04180901
>	s_buffer_load_dword s4, s[0:3], 0x2d                 ; C202012D
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_add_f32_e32 v6, s4, v6                             ; 060C0C04
>	v_mad_f32 v6, -0.5, v6, -0.5                         ; D2820006 03C60CF1
>	exp 15, 36, 0, 0, 0, v4, v6, v9, v8                  ; F800024F 08090604
>	s_buffer_load_dword s4, s[0:3], 0x37                 ; C2020137
>	s_waitcnt expcnt(0) lgkmcnt(0)                       ; BF8C000F
>	v_add_f32_e32 v4, s4, v0                             ; 06080004
>	s_buffer_load_dword s4, s[0:3], 0x35                 ; C2020135
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v4, s4, v4                             ; 10080804
>	s_buffer_load_dword s4, s[0:3], 0x36                 ; C2020136
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_add_f32_e32 v6, s4, v5                             ; 060C0A04
>	s_buffer_load_dword s4, s[0:3], 0x34                 ; C2020134
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v6, s4, v6                             ; 100C0C04
>	exp 15, 37, 0, 0, 0, v6, v4, v9, v8                  ; F800025F 08090406
>	s_buffer_load_dword s4, s[0:3], 0x6                  ; C2020106
>	s_waitcnt expcnt(0) lgkmcnt(0)                       ; BF8C000F
>	v_mul_f32_e32 v0, s4, v0                             ; 10000004
>	s_buffer_load_dword s4, s[0:3], 0x2                  ; C2020102
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v0, v5, s4, v0                             ; D2820000 04000905
>	s_buffer_load_dword s4, s[0:3], 0xa                  ; C202010A
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v0, v1, s4, v0                             ; D2820000 04000901
>	s_buffer_load_dword s0, s[0:3], 0xe                  ; C200010E
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_add_f32_e32 v0, s0, v0                             ; 06000000
>	exp 15, 12, 0, 1, 0, v7, v2, v0, v3                  ; F80008CF 03000207
>	s_endpgm                                             ; BF810000
>
>SGPRS: 16
>VGPRS: 16
>Code Size: 684 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL IN[1], GENERIC[10], PERSPECTIVE
>DCL IN[2], GENERIC[11], PERSPECTIVE
>DCL IN[3], GENERIC[12], PERSPECTIVE
>DCL IN[4], GENERIC[13], PERSPECTIVE
>DCL IN[5], GENERIC[14], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SAMP[1]
>DCL SAMP[2]
>DCL SAMP[3]
>DCL SAMP[4]
>DCL SAMP[5]
>DCL SAMP[6]
>DCL SAMP[7]
>DCL SAMP[8]
>DCL SAMP[9]
>DCL CONST[0..16]
>DCL TEMP[0..28], LOCAL
>IMM[0] FLT32 {    1.0000,    -1.0000,     0.0000,    -0.0000}
>IMM[1] FLT32 {    2.2000,     1.0000,    -1.0000,     2.0000}
>IMM[2] FLT32 {    0.5000,   -16.0000,    -1.4427,     0.0000}
>IMM[3] FLT32 {    0.3000,     0.5900,     0.1100,     1.4427}
>IMM[4] FLT32 {    0.4545,     0.0000,     0.0000,     0.0000}
>  0: MAD TEMP[0].xy, IN[4], IMM[0].xyxx, IMM[0].zxzz
>  1: MOV TEMP[0].xy, TEMP[0].xyyy
>  2: TEX TEMP[0], TEMP[0], SAMP[7], 2D
>  3: ABS TEMP[1].x, TEMP[0]
>  4: MOV TEMP[2], -TEMP[1].xxxx
>  5: FSGE TEMP[3].x, TEMP[2].xxxx, IMM[0].zzzz
>  6: UIF TEMP[3].xxxx :0
>  7:   MOV TEMP[3].x, IMM[0].yyyy
>  8: ELSE :0
>  9:   MOV TEMP[3].x, IMM[0].wwww
> 10: ENDIF
> 11: MOV TEMP[3].x, TEMP[3].xxxx
> 12: FSGE TEMP[4].x, TEMP[2].yyyy, IMM[0].zzzz
> 13: UIF TEMP[4].xxxx :0
> 14:   MOV TEMP[4].x, IMM[0].yyyy
> 15: ELSE :0
> 16:   MOV TEMP[4].x, IMM[0].wwww
> 17: ENDIF
> 18: MOV TEMP[3].y, TEMP[4].xxxx
> 19: FSGE TEMP[4].x, TEMP[2].zzzz, IMM[0].zzzz
> 20: UIF TEMP[4].xxxx :0
> 21:   MOV TEMP[4].x, IMM[0].yyyy
> 22: ELSE :0
> 23:   MOV TEMP[4].x, IMM[0].wwww
> 24: ENDIF
> 25: MOV TEMP[3].z, TEMP[4].xxxx
> 26: FSGE TEMP[2].x, TEMP[2].wwww, IMM[0].zzzz
> 27: UIF TEMP[2].xxxx :0
> 28:   MOV TEMP[2].x, IMM[0].yyyy
> 29: ELSE :0
> 30:   MOV TEMP[2].x, IMM[0].wwww
> 31: ENDIF
> 32: MOV TEMP[3].w, TEMP[2].xxxx
> 33: MOV TEMP[2].w, TEMP[3]
> 34: FSLT TEMP[3].xyz, TEMP[3].xyzz, IMM[0].zzzz
> 35: OR TEMP[4].x, TEMP[3].xxxx, TEMP[3].zzzz
> 36: OR TEMP[3].x, TEMP[4].xxxx, TEMP[3].yyyy
> 37: AND TEMP[3].x, TEMP[3].xxxx, IMM[0].xxxx
> 38: KILL_IF -TEMP[3].xxxx
> 39: MOV TEMP[3].w, IMM[0].zzzz
> 40: MOV TEMP[3].x, IN[0].xxxx
> 41: MOV TEMP[3].y, IN[0].yyyy
> 42: MOV TEMP[3].z, IN[0].zzzz
> 43: DP4 TEMP[4].x, TEMP[3], TEMP[3]
> 44: RSQ TEMP[4].x, TEMP[4].xxxx
> 45: MUL TEMP[3].xyz, TEMP[3], TEMP[4].xxxx
> 46: MOV TEMP[2].xyz, TEMP[3].xyzx
> 47: MOV TEMP[4].w, IMM[0].zzzz
> 48: MOV TEMP[4].x, IN[2].xxxx
> 49: MOV TEMP[4].y, IN[2].yyyy
> 50: MOV TEMP[4].z, IN[2].zzzz
> 51: DP4 TEMP[5].x, TEMP[4], TEMP[4]
> 52: RSQ TEMP[5].x, TEMP[5].xxxx
> 53: MUL TEMP[4].xyz, TEMP[4], TEMP[5].xxxx
> 54: ABS TEMP[5], TEMP[2]
> 55: ABS TEMP[6], TEMP[2]
> 56: MUL TEMP[5].xyz, TEMP[5], TEMP[6]
> 57: MUL TEMP[6].xyw, TEMP[5].xyzz, TEMP[5].xyzz
> 58: ADD TEMP[7].y, TEMP[6].yyyy, TEMP[6].xxxx
> 59: MAD TEMP[7].y, TEMP[5].zzzz, TEMP[5].zzzz, TEMP[7].yyyy
> 60: RCP TEMP[7].x, TEMP[7].yyyy
> 61: MUL TEMP[6].xyz, TEMP[7].xxxx, TEMP[6].xyww
> 62: MUL TEMP[7].yw, CONST[9].xxzy, IN[1].xyzz
> 63: MOV TEMP[7].xy, TEMP[7].ywww
> 64: TEX TEMP[7], TEMP[7], SAMP[3], 2D
> 65: POW TEMP[8].x, TEMP[7].xxxx, IMM[1].xxxx
> 66: POW TEMP[8].y, TEMP[7].yyyy, IMM[1].xxxx
> 67: POW TEMP[8].z, TEMP[7].zzzz, IMM[1].xxxx
> 68: POW TEMP[8].w, TEMP[7].wwww, IMM[1].yyyy
> 69: MOV TEMP[7].w, TEMP[8].wwww
> 70: MUL TEMP[9].yw, CONST[9].xxzy, IN[1].xxzz
> 71: MOV TEMP[9].xy, TEMP[9].ywww
> 72: TEX TEMP[9], TEMP[9], SAMP[3], 2D
> 73: POW TEMP[10].x, TEMP[9].xxxx, IMM[1].xxxx
> 74: POW TEMP[10].y, TEMP[9].yyyy, IMM[1].xxxx
> 75: POW TEMP[10].z, TEMP[9].zzzz, IMM[1].xxxx
> 76: POW TEMP[10].w, TEMP[9].wwww, IMM[1].yyyy
> 77: MOV TEMP[9].w, TEMP[10].wwww
> 78: MUL TEMP[10].xyz, TEMP[6].yyyy, TEMP[10]
> 79: MOV TEMP[9].xyz, TEMP[10].xyzx
> 80: MAD TEMP[8].xyz, TEMP[8], TEMP[6].xxxx, TEMP[9]
> 81: MOV TEMP[7].xyz, TEMP[8].xyzx
> 82: MUL TEMP[8].yw, CONST[4].xxxx, IN[1].xxzy
> 83: MOV TEMP[8].xy, TEMP[8].ywww
> 84: TEX TEMP[8], TEMP[8], SAMP[1], 2D
> 85: POW TEMP[10].x, TEMP[8].xxxx, IMM[1].xxxx
> 86: POW TEMP[10].y, TEMP[8].yyyy, IMM[1].xxxx
> 87: POW TEMP[10].z, TEMP[8].zzzz, IMM[1].xxxx
> 88: POW TEMP[10].w, TEMP[8].wwww, IMM[1].yyyy
> 89: MAD TEMP[8].xyz, TEMP[10], TEMP[6].zzzz, TEMP[7]
> 90: MOV TEMP[7].xyz, TEMP[8].xyzx
> 91: MUL TEMP[10].yw, CONST[10].xxzy, IN[1].xyzz
> 92: MOV TEMP[2].w, -TEMP[3].zzzz
> 93: DP3 TEMP[11].x, TEMP[2].wyxx, TEMP[4].xyzz
> 94: MOV TEMP[9].y, TEMP[11].xxxx
> 95: MUL TEMP[11].xyz, TEMP[2].xzyw, IMM[1].yyzw
> 96: DP3 TEMP[12].x, TEMP[11].zxyy, TEMP[4].xyzz
> 97: MOV TEMP[9].w, TEMP[12].xxxx
> 98: DP3 TEMP[12].x, TEMP[3].xyzz, TEMP[4].xyzz
> 99: MOV TEMP[9].z, TEMP[12].xxxx
>100: MUL TEMP[12].xyz, TEMP[2].zyxw, IMM[1].zyyw
>101: DP3 TEMP[13].x, TEMP[12].xyzz, CONST[1].xyzz
>102: MOV TEMP[13].z, TEMP[13].xxxx
>103: DP3 TEMP[14].x, TEMP[11].zxyy, CONST[1].xyzz
>104: MOV TEMP[13].w, TEMP[14].xxxx
>105: DP3 TEMP[13].x, TEMP[3].xyzz, CONST[1].xyzz
>106: DP3 TEMP[14].x, TEMP[12].xyzz, CONST[2].xyzz
>107: MOV TEMP[12].y, TEMP[14].xxxx
>108: DP3 TEMP[14].x, TEMP[11].zxyy, CONST[2].xyzz
>109: MOV TEMP[12].w, TEMP[14].xxxx
>110: DP3 TEMP[14].x, TEMP[3].xyzz, CONST[2].xyzz
>111: MOV TEMP[12].z, TEMP[14].xxxx
>112: MOV TEMP[14].xy, TEMP[10].ywww
>113: TEX TEMP[14].zw, TEMP[14], SAMP[5], 2D
>114: MOV TEMP[15].zw, TEMP[14].wwzw
>115: MOV TEMP[16].xy, TEMP[10].ywww
>116: TEX TEMP[16], TEMP[16], SAMP[4], 2D
>117: MOV TEMP[17].w, TEMP[16].wwww
>118: MUL TEMP[18].y, CONST[3].xxxx, CONST[3].xxxx
>119: MAD TEMP[19].xy, TEMP[16], IMM[1].wwww, IMM[0].yyyy
>120: MOV TEMP[15].xy, TEMP[19].xyxx
>121: MUL TEMP[19].xy, TEMP[15], CONST[3].xxxx
>122: MOV TEMP[15].xy, TEMP[19].xyxx
>123: MOV TEMP[15].z, IMM[0].xxxx
>124: MOV TEMP[20].w, IMM[0].zzzz
>125: MOV TEMP[20].x, TEMP[19].xxxx
>126: MOV TEMP[20].y, TEMP[19].yyyy
>127: MOV TEMP[20].z, IMM[0].xxxx
>128: DP4 TEMP[19].x, TEMP[20], TEMP[20]
>129: RSQ TEMP[19].x, TEMP[19].xxxx
>130: MUL TEMP[19].xyz, TEMP[20], TEMP[19].xxxx
>131: MAD TEMP[14].z, TEMP[14].wwww, IMM[1].wwww, IMM[0].yyyy
>132: MOV TEMP[14].z, TEMP[14].zzzz
>133: MOV TEMP[14].xy, TEMP[16].zwzz
>134: MUL TEMP[16].xyz, TEMP[15].xyyw, TEMP[15].xyxw
>135: MOV TEMP[17].xyz, TEMP[16].xyzx
>136: MAD TEMP[16].xyz, TEMP[14], TEMP[18].yyyy, -TEMP[17]
>137: DP3 TEMP[20].x, TEMP[13].zwxx, TEMP[19].xyzz
>138: MOV_SAT TEMP[20].x, TEMP[20].xxxx
>139: DP3 TEMP[21].x, TEMP[12].ywzz, TEMP[19].xyzz
>140: MOV_SAT TEMP[21].x, TEMP[21].xxxx
>141: MOV TEMP[2].w, TEMP[21].xxxx
>142: ADD TEMP[22].xyz, TEMP[9].ywzw, TEMP[12].ywzw
>143: MOV TEMP[14].xyz, TEMP[22].xyzx
>144: RCP TEMP[22].x, TEMP[22].zzzz
>145: MAD TEMP[22].xy, TEMP[14], TEMP[22].xxxx, -TEMP[15]
>146: RCP TEMP[23].x, CONST[6].xxxx
>147: ADD TEMP[24].zw, TEMP[23].xxxx, TEMP[16].xyxy
>148: MUL TEMP[25].w, TEMP[16].zzzz, TEMP[16].zzzz
>149: MAD TEMP[25].w, TEMP[24].zzzz, TEMP[24].wwww, -TEMP[25].wwww
>150: MUL TEMP[26].w, TEMP[22].xxxx, TEMP[22].xxxx
>151: MUL TEMP[27].w, TEMP[22].yyyy, TEMP[24].wwww
>152: DP2 TEMP[16].x, TEMP[22].xxxx, -TEMP[16].zzzz
>153: ADD TEMP[16].x, TEMP[16].xxxx, TEMP[27].wwww
>154: MUL TEMP[16].w, TEMP[22].yyyy, TEMP[16].xxxx
>155: MAD TEMP[16].w, TEMP[26].wwww, TEMP[24].wwww, TEMP[16].wwww
>156: MUL TEMP[16].w, TEMP[16].wwww, IMM[2].xxxx
>157: RCP TEMP[22].x, TEMP[25].wwww
>158: MUL TEMP[24].z, TEMP[16].wwww, TEMP[22].xxxx
>159: MOV TEMP[26].x, -TEMP[25].wwww
>160: FSGE TEMP[26].x, TEMP[26].xxxx, IMM[0].zzzz
>161: UIF TEMP[26].xxxx :0
>162:   MOV TEMP[26].x, IMM[0].xxxx
>163: ELSE :0
>164:   MOV TEMP[26].x, IMM[0].zzzz
>165: ENDIF
>166: MAD TEMP[16].w, TEMP[16].wwww, TEMP[22].xxxx, IMM[2].yyyy
>167: FSGE TEMP[16].x, TEMP[16].wwww, IMM[0].zzzz
>168: UIF TEMP[16].xxxx :0
>169:   MOV TEMP[16].x, IMM[0].xxxx
>170: ELSE :0
>171:   MOV TEMP[16].x, IMM[0].zzzz
>172: ENDIF
>173: ADD TEMP[16].w, TEMP[16].xxxx, TEMP[26].xxxx
>174: MUL TEMP[22].w, TEMP[24].zzzz, IMM[2].zzzz
>175: EX2 TEMP[22].x, TEMP[22].wwww
>176: MAX TEMP[24].x, TEMP[25].wwww, IMM[2].wwww
>177: RSQ TEMP[24].x, TEMP[24].xxxx
>178: MUL TEMP[22].w, TEMP[24].xxxx, TEMP[22].xxxx
>179: MOV TEMP[16].x, -TEMP[16].wwww
>180: FSGE TEMP[16].x, TEMP[16].xxxx, IMM[0].zzzz
>181: UIF TEMP[16].xxxx :0
>182:   MOV TEMP[16].x, TEMP[22].wwww
>183: ELSE :0
>184:   MOV TEMP[16].x, IMM[0].zzzz
>185: ENDIF
>186: DP3 TEMP[22].x, TEMP[19].xzyy, TEMP[9].yzww
>187: ADD TEMP[22].w, -TEMP[22].xxxx, IMM[0].xxxx
>188: MUL TEMP[24].w, TEMP[22].wwww, TEMP[22].wwww
>189: MUL TEMP[24].w, TEMP[24].wwww, TEMP[24].wwww
>190: MUL TEMP[22].w, TEMP[22].wwww, TEMP[24].wwww
>191: MOV TEMP[7].w, TEMP[22].wwww
>192: MUL TEMP[24].xyz, TEMP[2].zyxw, IMM[1].zzyw
>193: DP3 TEMP[15].x, TEMP[19].xyzz, TEMP[24].xyzz
>194: DP3 TEMP[24].x, TEMP[19].yxzz, TEMP[3].xyyy
>195: MOV TEMP[15].y, TEMP[24].xxxx
>196: DP3 TEMP[19].x, TEMP[19].xyzz, TEMP[3].xzzz
>197: MOV TEMP[15].z, TEMP[19].xxxx
>198: MOV TEMP[19].xyz, TEMP[15].xyzz
>199: TEX TEMP[19], TEMP[19], SAMP[0], CUBE
>200: POW TEMP[24].x, TEMP[19].xxxx, IMM[1].xxxx
>201: POW TEMP[24].y, TEMP[19].yyyy, IMM[1].xxxx
>202: POW TEMP[24].z, TEMP[19].zzzz, IMM[1].xxxx
>203: POW TEMP[24].w, TEMP[19].wwww, IMM[1].yyyy
>204: MOV TEMP[15].w, TEMP[24].wwww
>205: ADD TEMP[19].xyz, TEMP[20].xxxx, TEMP[24]
>206: MOV TEMP[15].xyz, TEMP[19].xyzx
>207: MUL TEMP[19].xyz, TEMP[7], TEMP[15]
>208: MOV TEMP[15].xyz, TEMP[19].xyzx
>209: MUL TEMP[16], TEMP[16].xxxx, TEMP[22].wwww
>210: MOV_SAT TEMP[16], TEMP[16]
>211: MAD TEMP[15].xyz, TEMP[16].wwww, TEMP[21].xxxx, TEMP[15]
>212: MUL TEMP[16].xy, CONST[10], IN[1].xzzw
>213: MUL TEMP[19].xyz, TEMP[2].zyxw, IMM[1].yyzw
>214: DP3 TEMP[9].x, TEMP[19].zxyy, TEMP[4].xyzz
>215: DP3 TEMP[20].x, TEMP[19].zxyy, CONST[1].xyzz
>216: MOV TEMP[13].y, TEMP[20].xxxx
>217: DP3 TEMP[12].x, TEMP[19].zxyy, CONST[2].xyzz
>218: MOV TEMP[20].xy, TEMP[16].xyyy
>219: TEX TEMP[20].w, TEMP[20], SAMP[5], 2D
>220: MOV TEMP[14].w, TEMP[20].wwww
>221: MOV TEMP[16].xy, TEMP[16].xyyy
>222: TEX TEMP[16], TEMP[16], SAMP[4], 2D
>223: MOV TEMP[17].zw, TEMP[16].wwzw
>224: MAD TEMP[21].xy, TEMP[16], IMM[1].wwww, IMM[0].yyyy
>225: MOV TEMP[17].xy, TEMP[21].xyxx
>226: MUL TEMP[21].xy, TEMP[17], CONST[3].xxxx
>227: MOV TEMP[14].xy, TEMP[21].xyxx
>228: MOV TEMP[22].w, IMM[0].zzzz
>229: MOV TEMP[22].x, TEMP[21].xxxx
>230: MOV TEMP[22].y, TEMP[21].yyyy
>231: MOV TEMP[22].z, IMM[0].xxxx
>232: DP4 TEMP[21].x, TEMP[22], TEMP[22]
>233: RSQ TEMP[21].x, TEMP[21].xxxx
>234: MUL TEMP[21].xyz, TEMP[22], TEMP[21].xxxx
>235: MAD TEMP[20].z, TEMP[20].wwww, IMM[1].wwww, IMM[0].yyyy
>236: MOV TEMP[20].z, TEMP[20].zzzz
>237: MOV TEMP[20].xy, TEMP[16].zwzz
>238: MUL TEMP[16].xyz, TEMP[14].xyyw, TEMP[14].xyxw
>239: MOV TEMP[17].xyz, TEMP[16].xyzx
>240: MAD TEMP[16].xyz, TEMP[20], TEMP[18].yyyy, -TEMP[17]
>241: MOV TEMP[17].xyz, TEMP[16].xyzx
>242: DP3 TEMP[20].x, TEMP[13].ywxx, TEMP[21].xyzz
>243: MOV_SAT TEMP[20].x, TEMP[20].xxxx
>244: DP3 TEMP[22].x, TEMP[12].xwzz, TEMP[21].xyzz
>245: MOV_SAT TEMP[22].x, TEMP[22].xxxx
>246: ADD TEMP[24].yzw, TEMP[9].xxwz, TEMP[12].xxwz
>247: MOV TEMP[13].yzw, TEMP[24].zyzw
>248: RCP TEMP[24].x, TEMP[24].wwww
>249: MAD TEMP[24].yz, TEMP[13], TEMP[24].xxxx, -TEMP[14].xxyw
>250: ADD TEMP[25].xy, TEMP[23].xxxx, TEMP[17]
>251: MUL TEMP[26].w, TEMP[16].zzzz, TEMP[16].zzzz
>252: MAD TEMP[26].w, TEMP[25].xxxx, TEMP[25].yyyy, -TEMP[26].wwww
>253: MUL TEMP[27].w, TEMP[24].yyyy, TEMP[24].yyyy
>254: MUL TEMP[28].w, TEMP[24].zzzz, TEMP[25].yyyy
>255: DP2 TEMP[16].x, TEMP[24].yyyy, -TEMP[16].zzzz
>256: ADD TEMP[16].x, TEMP[16].xxxx, TEMP[28].wwww
>257: MUL TEMP[16].w, TEMP[24].zzzz, TEMP[16].xxxx
>258: MAD TEMP[16].w, TEMP[27].wwww, TEMP[25].yyyy, TEMP[16].wwww
>259: MUL TEMP[16].w, TEMP[16].wwww, IMM[2].xxxx
>260: RCP TEMP[24].x, TEMP[26].wwww
>261: MUL TEMP[25].w, TEMP[16].wwww, TEMP[24].xxxx
>262: MOV TEMP[12].w, TEMP[25].wwww
>263: MOV TEMP[27].x, -TEMP[26].wwww
>264: FSGE TEMP[27].x, TEMP[27].xxxx, IMM[0].zzzz
>265: UIF TEMP[27].xxxx :0
>266:   MOV TEMP[27].x, IMM[0].xxxx
>267: ELSE :0
>268:   MOV TEMP[27].x, IMM[0].zzzz
>269: ENDIF
>270: MAD TEMP[16].w, TEMP[16].wwww, TEMP[24].xxxx, IMM[2].yyyy
>271: FSGE TEMP[16].x, TEMP[16].wwww, IMM[0].zzzz
>272: UIF TEMP[16].xxxx :0
>273:   MOV TEMP[16].x, IMM[0].xxxx
>274: ELSE :0
>275:   MOV TEMP[16].x, IMM[0].zzzz
>276: ENDIF
>277: ADD TEMP[16].w, TEMP[16].xxxx, TEMP[27].xxxx
>278: MUL TEMP[24].w, TEMP[25].wwww, IMM[2].zzzz
>279: EX2 TEMP[24].x, TEMP[24].wwww
>280: MAX TEMP[25].x, TEMP[26].wwww, IMM[2].wwww
>281: RSQ TEMP[25].x, TEMP[25].xxxx
>282: MUL TEMP[24].w, TEMP[25].xxxx, TEMP[24].xxxx
>283: MOV TEMP[16].x, -TEMP[16].wwww
>284: FSGE TEMP[16].x, TEMP[16].xxxx, IMM[0].zzzz
>285: UIF TEMP[16].xxxx :0
>286:   MOV TEMP[16].x, TEMP[24].wwww
>287: ELSE :0
>288:   MOV TEMP[16].x, IMM[0].zzzz
>289: ENDIF
>290: DP3 TEMP[24].x, TEMP[21].xzyy, TEMP[9].xzww
>291: ADD TEMP[24].w, -TEMP[24].xxxx, IMM[0].xxxx
>292: MUL TEMP[25].w, TEMP[24].wwww, TEMP[24].wwww
>293: MUL TEMP[25].w, TEMP[25].wwww, TEMP[25].wwww
>294: MOV TEMP[9].w, TEMP[25].wwww
>295: MUL TEMP[24].w, TEMP[24].wwww, TEMP[25].wwww
>296: MUL TEMP[25].yzw, TEMP[3].xxyx, IMM[0].yyyx
>297: DP3 TEMP[17].x, TEMP[21].xyzz, TEMP[25].yzww
>298: DP3 TEMP[25].x, TEMP[21].yzxx, TEMP[3].xyzz
>299: MOV TEMP[17].y, TEMP[25].xxxx
>300: DP3 TEMP[21].x, TEMP[21].xyzz, TEMP[3].yzzz
>301: MOV TEMP[17].z, TEMP[21].xxxx
>302: MOV TEMP[21].xyz, TEMP[17].xyzz
>303: TEX TEMP[21], TEMP[21], SAMP[0], CUBE
>304: POW TEMP[25].x, TEMP[21].xxxx, IMM[1].xxxx
>305: POW TEMP[25].y, TEMP[21].yyyy, IMM[1].xxxx
>306: POW TEMP[25].z, TEMP[21].zzzz, IMM[1].xxxx
>307: POW TEMP[25].w, TEMP[21].wwww, IMM[1].yyyy
>308: MOV TEMP[17].w, TEMP[25].wwzw
>309: ADD TEMP[20].yzw, TEMP[20].xxxx, TEMP[25].xxyz
>310: MOV TEMP[13].yzw, TEMP[20].zyzw
>311: MUL TEMP[20].yzw, TEMP[8].xxyz, TEMP[13]
>312: MOV TEMP[13].yzw, TEMP[20].zyzw
>313: MUL TEMP[16], TEMP[16].xxxx, TEMP[24].wwww
>314: MOV_SAT TEMP[16], TEMP[16]
>315: MAD TEMP[16].yzw, TEMP[16].wwww, TEMP[22].xxxx, TEMP[13]
>316: MUL TEMP[20].xy, CONST[5].xxxx, IN[1]
>317: DP3 TEMP[9].x, TEMP[19].xyzz, TEMP[4].xyzz
>318: DP3 TEMP[21].x, TEMP[11].xyzz, TEMP[4].xyzz
>319: MOV TEMP[9].y, TEMP[21].xxxx
>320: DP3 TEMP[14].x, TEMP[19].xyzz, CONST[1].xyzz
>321: DP3 TEMP[21].x, TEMP[11].xyzz, CONST[1].xyzz
>322: MOV TEMP[14].y, TEMP[21].xxxx
>323: DP3 TEMP[12].x, TEMP[19].xyzz, CONST[2].xyzz
>324: DP3 TEMP[19].x, TEMP[11].xyzz, CONST[2].xyzz
>325: MOV TEMP[12].y, TEMP[19].xxxx
>326: MOV TEMP[19].xy, TEMP[20].xyyy
>327: TEX TEMP[19], TEMP[19], SAMP[2], 2D
>328: MOV TEMP[10].zw, TEMP[19].wwzw
>329: MAD TEMP[20].xy, TEMP[19], IMM[1].wwww, IMM[0].yyyy
>330: MOV TEMP[10].xy, TEMP[20].xyxx
>331: MUL TEMP[20].xy, TEMP[10], CONST[3].xxxx
>332: MOV TEMP[17].xy, TEMP[20].xyxx
>333: MOV TEMP[17].z, IMM[0].xxxx
>334: MOV TEMP[21].w, IMM[0].zzzz
>335: MOV TEMP[21].x, TEMP[20].xxxx
>336: MOV TEMP[21].y, TEMP[20].yyyy
>337: MOV TEMP[21].z, IMM[0].xxxx
>338: DP4 TEMP[20].x, TEMP[21], TEMP[21]
>339: RSQ TEMP[20].x, TEMP[20].xxxx
>340: MUL TEMP[20].xyz, TEMP[21], TEMP[20].xxxx
>341: MAD TEMP[19].xyz, TEMP[19].zwww, IMM[1].yywz, IMM[0].zzyy
>342: MOV TEMP[10].xyz, TEMP[19].xyzx
>343: MUL TEMP[19].xyz, TEMP[17].xyyw, TEMP[17].xyxw
>344: MOV TEMP[11].xyz, TEMP[19].xyzx
>345: MAD TEMP[11].xyz, TEMP[10], TEMP[18].yyyy, -TEMP[11]
>346: MOV TEMP[10].xyz, TEMP[11].xyzx
>347: MOV TEMP[14].z, TEMP[13].xxxx
>348: DP3 TEMP[14].x, TEMP[14].xyzz, TEMP[20].xyzz
>349: MOV_SAT TEMP[14].x, TEMP[14].xxxx
>350: DP3 TEMP[18].x, TEMP[12].xyzz, TEMP[20].xyzz
>351: MOV_SAT TEMP[18].x, TEMP[18].xxxx
>352: ADD TEMP[19].xyz, TEMP[9], TEMP[12]
>353: MOV TEMP[12].xyz, TEMP[19].xyzx
>354: RCP TEMP[19].x, TEMP[19].zzzz
>355: MAD TEMP[12].xy, TEMP[12], TEMP[19].xxxx, -TEMP[17]
>356: ADD TEMP[10].xy, TEMP[23].xxxx, TEMP[10]
>357: MUL TEMP[17].w, TEMP[11].zzzz, TEMP[11].zzzz
>358: MAD TEMP[17].w, TEMP[10].xxxx, TEMP[10].yyyy, -TEMP[17].wwww
>359: MUL TEMP[19].w, TEMP[12].xxxx, TEMP[12].xxxx
>360: MUL TEMP[21].w, TEMP[12].yyyy, TEMP[10].yyyy
>361: DP2 TEMP[11].x, TEMP[12].xxxx, -TEMP[11].zzzz
>362: ADD TEMP[11].x, TEMP[11].xxxx, TEMP[21].wwww
>363: MUL TEMP[11].w, TEMP[12].yyyy, TEMP[11].xxxx
>364: MAD TEMP[10].w, TEMP[19].wwww, TEMP[10].yyyy, TEMP[11].wwww
>365: MUL TEMP[10].w, TEMP[10].wwww, IMM[2].xxxx
>366: RCP TEMP[11].x, TEMP[17].wwww
>367: MUL TEMP[12].w, TEMP[10].wwww, TEMP[11].xxxx
>368: MOV TEMP[7].w, TEMP[12].wwww
>369: MOV TEMP[19].x, -TEMP[17].wwww
>370: FSGE TEMP[19].x, TEMP[19].xxxx, IMM[0].zzzz
>371: UIF TEMP[19].xxxx :0
>372:   MOV TEMP[19].x, IMM[0].xxxx
>373: ELSE :0
>374:   MOV TEMP[19].x, IMM[0].zzzz
>375: ENDIF
>376: MAD TEMP[10].w, TEMP[10].wwww, TEMP[11].xxxx, IMM[2].yyyy
>377: FSGE TEMP[10].x, TEMP[10].wwww, IMM[0].zzzz
>378: UIF TEMP[10].xxxx :0
>379:   MOV TEMP[10].x, IMM[0].xxxx
>380: ELSE :0
>381:   MOV TEMP[10].x, IMM[0].zzzz
>382: ENDIF
>383: ADD TEMP[10].w, TEMP[10].xxxx, TEMP[19].xxxx
>384: MUL TEMP[11].w, TEMP[12].wwww, IMM[2].zzzz
>385: EX2 TEMP[11].x, TEMP[11].wwww
>386: MAX TEMP[12].x, TEMP[17].wwww, IMM[2].wwww
>387: RSQ TEMP[12].x, TEMP[12].xxxx
>388: MUL TEMP[11].w, TEMP[12].xxxx, TEMP[11].xxxx
>389: MOV TEMP[10].x, -TEMP[10].wwww
>390: FSGE TEMP[10].x, TEMP[10].xxxx, IMM[0].zzzz
>391: UIF TEMP[10].xxxx :0
>392:   MOV TEMP[10].x, TEMP[11].wwww
>393: ELSE :0
>394:   MOV TEMP[10].x, IMM[0].zzzz
>395: ENDIF
>396: MOV TEMP[2].w, TEMP[10].xxxx
>397: DP3 TEMP[11].x, TEMP[20].xyzz, TEMP[9].xyzz
>398: ADD TEMP[11].w, -TEMP[11].xxxx, IMM[0].xxxx
>399: MUL TEMP[12].w, TEMP[11].wwww, TEMP[11].wwww
>400: MUL TEMP[12].w, TEMP[12].wwww, TEMP[12].wwww
>401: MUL TEMP[11].w, TEMP[11].wwww, TEMP[12].wwww
>402: DP3 TEMP[9].x, TEMP[20].yzxx, TEMP[3].xxzz
>403: DP3 TEMP[12].x, TEMP[20].xzyy, TEMP[3].yyzz
>404: MOV TEMP[9].y, TEMP[12].xxxx
>405: MUL TEMP[12].xyz, TEMP[2], IMM[1].zzyw
>406: DP3 TEMP[12].x, TEMP[20].xyzz, TEMP[12].xyzz
>407: MOV TEMP[9].z, TEMP[12].xxxx
>408: MOV TEMP[12].xyz, TEMP[9].xyzz
>409: TEX TEMP[12], TEMP[12], SAMP[0], CUBE
>410: POW TEMP[17].x, TEMP[12].xxxx, IMM[1].xxxx
>411: POW TEMP[17].y, TEMP[12].yyyy, IMM[1].xxxx
>412: POW TEMP[17].z, TEMP[12].zzzz, IMM[1].xxxx
>413: POW TEMP[17].w, TEMP[12].wwww, IMM[1].yyyy
>414: MOV TEMP[9].w, TEMP[17].wwww
>415: ADD TEMP[12].xyz, TEMP[14].xxxx, TEMP[17]
>416: MOV TEMP[9].xyz, TEMP[12].xyzx
>417: MUL TEMP[12].xyz, TEMP[7], TEMP[9]
>418: MOV TEMP[9].xyz, TEMP[12].xyzx
>419: MUL TEMP[10], TEMP[10].xxxx, TEMP[11].wwww
>420: MOV_SAT TEMP[10], TEMP[10]
>421: MAD TEMP[10].xyz, TEMP[10].yyyy, TEMP[18].xxxx, TEMP[9]
>422: MOV TEMP[9].xyz, TEMP[10].xyzx
>423: MUL TEMP[10].xyz, TEMP[6].yyyy, TEMP[16].yzww
>424: MAD TEMP[10].xyw, TEMP[15].xyzz, TEMP[6].xxxx, TEMP[10].xyzz
>425: MOV TEMP[5].w, TEMP[10].xyxw
>426: MAD TEMP[6].xyz, TEMP[9], TEMP[6].zzzz, TEMP[10].xyww
>427: MOV TEMP[5].xyz, TEMP[6].xyzx
>428: DP3 TEMP[6].x, TEMP[8].xyzz, IMM[3].xyzz
>429: ADD TEMP[6].y, -TEMP[6].xxxx, IMM[0].xxxx
>430: MUL TEMP[6].xyz, TEMP[6].yyyy, CONST[7]
>431: MOV TEMP[7].xyz, TEMP[6].xyzx
>432: MOV_SAT TEMP[6].x, TEMP[13].xxxx
>433: MOV TEMP[8].xyz, TEMP[3].xyzz
>434: TEX TEMP[8], TEMP[8], SAMP[0], CUBE
>435: POW TEMP[9].x, TEMP[8].xxxx, IMM[1].xxxx
>436: POW TEMP[9].y, TEMP[8].yyyy, IMM[1].xxxx
>437: POW TEMP[9].z, TEMP[8].zzzz, IMM[1].xxxx
>438: POW TEMP[9].w, TEMP[8].wwww, IMM[1].yyyy
>439: MOV TEMP[2].w, TEMP[9].wwww
>440: ADD TEMP[6].xyz, TEMP[9], TEMP[6].xxxx
>441: MOV TEMP[2].xyz, TEMP[6].xyzx
>442: MAD TEMP[6].xyz, TEMP[7], TEMP[2], TEMP[5]
>443: MAD TEMP[8].x, IN[3].yyyy, IMM[2].xxxx, IMM[2].xxxx
>444: MOV TEMP[5].x, TEMP[8].xxxx
>445: MOV TEMP[5].y, CONST[16].wwww
>446: MOV TEMP[8].xy, TEMP[5].xyyy
>447: TEX TEMP[8].x, TEMP[8], SAMP[6], 2D
>448: MOV TEMP[5].x, TEMP[8].xxxx
>449: ADD TEMP[9].yzw, -CONST[0].xxyz, IN[1].xxyz
>450: DP3 TEMP[10].x, TEMP[9].yzww, TEMP[9].yzww
>451: MUL TEMP[10].y, TEMP[10].xxxx, CONST[16].yyyy
>452: MUL TEMP[11].w, TEMP[9].wwww, CONST[16].xxxx
>453: MUL TEMP[11].w, TEMP[11].wwww, IMM[3].wwww
>454: EX2 TEMP[11].x, TEMP[11].wwww
>455: ADD TEMP[11].w, -TEMP[11].xxxx, IMM[0].xxxx
>456: MUL TEMP[10].y, TEMP[11].wwww, TEMP[10].yyyy
>457: RCP TEMP[9].x, TEMP[9].wwww
>458: MUL TEMP[9].y, TEMP[9].xxxx, TEMP[10].yyyy
>459: MUL TEMP[9].y, TEMP[9].yyyy, IMM[3].wwww
>460: EX2 TEMP[9].x, TEMP[9].yyyy
>461: MOV_SAT TEMP[9].x, TEMP[9].xxxx
>462: ADD TEMP[9].y, -TEMP[9].xxxx, IMM[0].xxxx
>463: MUL TEMP[10].w, TEMP[9].yyyy, TEMP[8].xxxx
>464: ADD TEMP[11].w, -CONST[8].xxxx, IN[1].zzzz
>465: MOV TEMP[1].w, TEMP[11].wwww
>466: FSGE TEMP[12].x, TEMP[11].wwww, IMM[0].zzzz
>467: UIF TEMP[12].xxxx :0
>468:   MOV TEMP[12].x, IMM[0].xxxx
>469: ELSE :0
>470:   MOV TEMP[12].x, IMM[0].zzzz
>471: ENDIF
>472: MOV TEMP[12].w, TEMP[12].xxxx
>473: ADD TEMP[13].w, CONST[8].xxxx, -IN[1].zzzz
>474: MOV TEMP[7].xyz, CONST[11].xyzx
>475: ADD TEMP[14].xyz, -TEMP[7], CONST[12]
>476: MUL TEMP[15], TEMP[13].wwww, CONST[13].xxxx
>477: MOV_SAT TEMP[15], TEMP[15]
>478: MAD TEMP[14].yzw, TEMP[15].yyyy, TEMP[14].xxyz, CONST[11].xxyz
>479: MOV TEMP[5].yzw, TEMP[14].zyzw
>480: MUL TEMP[14].yzw, TEMP[6].xxyz, TEMP[5]
>481: RCP TEMP[7].x, -TEMP[4].zzzz
>482: MUL TEMP[7].w, TEMP[13].wwww, TEMP[7].xxxx
>483: MUL TEMP[7].xyw, TEMP[7].wwww, TEMP[4].xyzz
>484: MOV TEMP[3].w, TEMP[7].xyxw
>485: DP3 TEMP[3].x, TEMP[7].xyww, TEMP[7].xyww
>486: MAX TEMP[7].x, TEMP[3].xxxx, IMM[2].wwww
>487: SQRT TEMP[3].x, TEMP[7].xxxx
>488: MUL TEMP[7].x, -TEMP[3].xxxx, CONST[14].xxxx
>489: MUL TEMP[7].x, TEMP[7].xxxx, IMM[3].wwww
>490: EX2 TEMP[7].x, TEMP[7].xxxx
>491: MUL TEMP[13].y, TEMP[4].zzzz, TEMP[4].zzzz
>492: MUL TEMP[13].y, TEMP[13].yyyy, TEMP[13].yyyy
>493: MAD TEMP[4].y, TEMP[4].zzzz, -TEMP[13].yyyy, IMM[0].xxxx
>494: ADD TEMP[4].y, -TEMP[4].yyyy, IMM[0].xxxx
>495: MUL TEMP[4].x, TEMP[7].xxxx, TEMP[4].yyyy
>496: MUL TEMP[4].xyz, TEMP[4].xxxx, TEMP[14].yzww
>497: MOV TEMP[3].xyz, TEMP[4].xyzx
>498: MAD TEMP[4].y, TEMP[8].xxxx, -TEMP[9].yyyy, IMM[0].xxxx
>499: MUL TEMP[4].xyz, TEMP[4].yyyy, TEMP[3]
>500: MOV TEMP[3].w, IMM[0].zzzz
>501: FSGE TEMP[7].x, TEMP[11].wwww, IMM[0].zzzz
>502: UIF TEMP[7].xxxx :0
>503:   MOV TEMP[7].x, TEMP[6].xxxx
>504: ELSE :0
>505:   MOV TEMP[7].x, TEMP[4].xxxx
>506: ENDIF
>507: FSGE TEMP[8].x, TEMP[11].wwww, IMM[0].zzzz
>508: UIF TEMP[8].xxxx :0
>509:   MOV TEMP[8].x, TEMP[6].yyyy
>510: ELSE :0
>511:   MOV TEMP[8].x, TEMP[4].yyyy
>512: ENDIF
>513: FSGE TEMP[9].x, TEMP[11].wwww, IMM[0].zzzz
>514: UIF TEMP[9].xxxx :0
>515:   MOV TEMP[6].x, TEMP[6].zzzz
>516: ELSE :0
>517:   MOV TEMP[6].x, TEMP[4].zzzz
>518: ENDIF
>519: FSGE TEMP[4].x, TEMP[11].wwww, IMM[0].zzzz
>520: UIF TEMP[4].xxxx :0
>521:   MOV TEMP[4].x, TEMP[10].wwww
>522: ELSE :0
>523:   MOV TEMP[4].x, IMM[0].zzzz
>524: ENDIF
>525: MOV TEMP[2].w, TEMP[4].xxxx
>526: ABS TEMP[7].x, TEMP[7].xxxx
>527: LG2 TEMP[3].x, TEMP[7].xxxx
>528: ABS TEMP[7].x, TEMP[8].xxxx
>529: LG2 TEMP[7].x, TEMP[7].xxxx
>530: MOV TEMP[3].y, TEMP[7].xxxx
>531: ABS TEMP[6].x, TEMP[6].xxxx
>532: LG2 TEMP[6].x, TEMP[6].xxxx
>533: MOV TEMP[3].z, TEMP[6].xxxx
>534: MUL TEMP[6].xyz, TEMP[3], IMM[4].xxxx
>535: EX2 TEMP[3].x, TEMP[6].xxxx
>536: EX2 TEMP[7].x, TEMP[6].yyyy
>537: MOV TEMP[3].y, TEMP[7].xxxx
>538: EX2 TEMP[6].x, TEMP[6].zzzz
>539: MOV TEMP[3].z, TEMP[6].xxxx
>540: MOV TEMP[6].xy, IN[5].xyyy
>541: TEX TEMP[6], TEMP[6], SAMP[8], 2D
>542: MOV TEMP[5].w, TEMP[6].wwww
>543: LRP TEMP[3].xyz, TEMP[0].xxxx, TEMP[3], TEMP[6]
>544: MOV TEMP[2].xyz, TEMP[3].xyzx
>545: MOV TEMP[3].xyz, TEMP[3].xyzz
>546: TEX TEMP[3], TEMP[3], SAMP[9], 3D
>547: MAD TEMP[0].x, TEMP[0].zzzz, -TEMP[0].xxxx, TEMP[0].xxxx
>548: LRP TEMP[0].xyz, TEMP[0].xxxx, TEMP[3], TEMP[2]
>549: MOV TEMP[5].xyz, TEMP[0].xyzx
>550: ADD TEMP[0].xyz, -TEMP[5], CONST[15]
>551: MOV TEMP[1].xyz, TEMP[0].xyzx
>552: MAD TEMP[0].xyz, TEMP[4].xxxx, TEMP[1], TEMP[5]
>553: MOV TEMP[12].xyz, TEMP[0].xyzx
>554: MOV OUT[0], TEMP[12]
>555: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %23 = load <16 x i8> addrspace(2)* %22, !tbaa !0
>  %24 = call float @llvm.SI.load.const(<16 x i8> %23, i32 0)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %23, i32 4)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %23, i32 8)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %23, i32 16)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %23, i32 20)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %23, i32 24)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %23, i32 32)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %23, i32 36)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %23, i32 40)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %23, i32 48)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %23, i32 64)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %23, i32 80)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %23, i32 96)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %23, i32 112)
>  %38 = call float @llvm.SI.load.const(<16 x i8> %23, i32 116)
>  %39 = call float @llvm.SI.load.const(<16 x i8> %23, i32 120)
>  %40 = call float @llvm.SI.load.const(<16 x i8> %23, i32 128)
>  %41 = call float @llvm.SI.load.const(<16 x i8> %23, i32 144)
>  %42 = call float @llvm.SI.load.const(<16 x i8> %23, i32 148)
>  %43 = call float @llvm.SI.load.const(<16 x i8> %23, i32 160)
>  %44 = call float @llvm.SI.load.const(<16 x i8> %23, i32 164)
>  %45 = call float @llvm.SI.load.const(<16 x i8> %23, i32 176)
>  %46 = call float @llvm.SI.load.const(<16 x i8> %23, i32 180)
>  %47 = call float @llvm.SI.load.const(<16 x i8> %23, i32 184)
>  %48 = call float @llvm.SI.load.const(<16 x i8> %23, i32 192)
>  %49 = call float @llvm.SI.load.const(<16 x i8> %23, i32 196)
>  %50 = call float @llvm.SI.load.const(<16 x i8> %23, i32 200)
>  %51 = call float @llvm.SI.load.const(<16 x i8> %23, i32 208)
>  %52 = call float @llvm.SI.load.const(<16 x i8> %23, i32 224)
>  %53 = call float @llvm.SI.load.const(<16 x i8> %23, i32 240)
>  %54 = call float @llvm.SI.load.const(<16 x i8> %23, i32 244)
>  %55 = call float @llvm.SI.load.const(<16 x i8> %23, i32 248)
>  %56 = call float @llvm.SI.load.const(<16 x i8> %23, i32 256)
>  %57 = call float @llvm.SI.load.const(<16 x i8> %23, i32 260)
>  %58 = call float @llvm.SI.load.const(<16 x i8> %23, i32 268)
>  %59 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %60 = load <8 x i32> addrspace(2)* %59, !tbaa !0
>  %61 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %62 = load <4 x i32> addrspace(2)* %61, !tbaa !0
>  %63 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 1
>  %64 = load <8 x i32> addrspace(2)* %63, !tbaa !0
>  %65 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 1
>  %66 = load <4 x i32> addrspace(2)* %65, !tbaa !0
>  %67 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 2
>  %68 = load <8 x i32> addrspace(2)* %67, !tbaa !0
>  %69 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 2
>  %70 = load <4 x i32> addrspace(2)* %69, !tbaa !0
>  %71 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 3
>  %72 = load <8 x i32> addrspace(2)* %71, !tbaa !0
>  %73 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 3
>  %74 = load <4 x i32> addrspace(2)* %73, !tbaa !0
>  %75 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 4
>  %76 = load <8 x i32> addrspace(2)* %75, !tbaa !0
>  %77 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 4
>  %78 = load <4 x i32> addrspace(2)* %77, !tbaa !0
>  %79 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 5
>  %80 = load <8 x i32> addrspace(2)* %79, !tbaa !0
>  %81 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 5
>  %82 = load <4 x i32> addrspace(2)* %81, !tbaa !0
>  %83 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 6
>  %84 = load <8 x i32> addrspace(2)* %83, !tbaa !0
>  %85 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 6
>  %86 = load <4 x i32> addrspace(2)* %85, !tbaa !0
>  %87 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 7
>  %88 = load <8 x i32> addrspace(2)* %87, !tbaa !0
>  %89 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 7
>  %90 = load <4 x i32> addrspace(2)* %89, !tbaa !0
>  %91 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 8
>  %92 = load <8 x i32> addrspace(2)* %91, !tbaa !0
>  %93 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 8
>  %94 = load <4 x i32> addrspace(2)* %93, !tbaa !0
>  %95 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 9
>  %96 = load <8 x i32> addrspace(2)* %95, !tbaa !0
>  %97 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 9
>  %98 = load <4 x i32> addrspace(2)* %97, !tbaa !0
>  %99 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %100 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %101 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %5, <2 x i32> %7)
>  %102 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %5, <2 x i32> %7)
>  %103 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %5, <2 x i32> %7)
>  %104 = call float @llvm.SI.fs.interp(i32 2, i32 1, i32 %5, <2 x i32> %7)
>  %105 = call float @llvm.SI.fs.interp(i32 0, i32 2, i32 %5, <2 x i32> %7)
>  %106 = call float @llvm.SI.fs.interp(i32 1, i32 2, i32 %5, <2 x i32> %7)
>  %107 = call float @llvm.SI.fs.interp(i32 2, i32 2, i32 %5, <2 x i32> %7)
>  %108 = call float @llvm.SI.fs.interp(i32 1, i32 3, i32 %5, <2 x i32> %7)
>  %109 = call float @llvm.SI.fs.interp(i32 0, i32 4, i32 %5, <2 x i32> %7)
>  %110 = call float @llvm.SI.fs.interp(i32 1, i32 4, i32 %5, <2 x i32> %7)
>  %111 = call float @llvm.SI.fs.interp(i32 0, i32 5, i32 %5, <2 x i32> %7)
>  %112 = call float @llvm.SI.fs.interp(i32 1, i32 5, i32 %5, <2 x i32> %7)
>  %113 = fmul float %109, 1.000000e+00
>  %114 = fadd float %113, 0.000000e+00
>  %115 = fmul float %110, -1.000000e+00
>  %116 = fadd float %115, 1.000000e+00
>  %117 = bitcast float %114 to i32
>  %118 = bitcast float %116 to i32
>  %119 = insertelement <2 x i32> undef, i32 %117, i32 0
>  %120 = insertelement <2 x i32> %119, i32 %118, i32 1
>  %121 = bitcast <8 x i32> %88 to <32 x i8>
>  %122 = bitcast <4 x i32> %90 to <16 x i8>
>  %123 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %120, <32 x i8> %121, <16 x i8> %122, i32 2)
>  %124 = extractelement <4 x float> %123, i32 0
>  %125 = extractelement <4 x float> %123, i32 2
>  %126 = call float @fabs(float %124)
>  %127 = fsub float -0.000000e+00, %126
>  %128 = fsub float -0.000000e+00, %126
>  %129 = fsub float -0.000000e+00, %126
>  %130 = fsub float -0.000000e+00, %126
>  %131 = fcmp oge float %127, 0.000000e+00
>  %132 = sext i1 %131 to i32
>  %133 = bitcast i32 %132 to float
>  %134 = bitcast float %133 to i32
>  %135 = icmp ne i32 %134, 0
>  %. = select i1 %135, float -1.000000e+00, float -0.000000e+00
>  %136 = fcmp oge float %128, 0.000000e+00
>  %137 = sext i1 %136 to i32
>  %138 = bitcast i32 %137 to float
>  %139 = bitcast float %138 to i32
>  %140 = icmp ne i32 %139, 0
>  %temp16.0 = select i1 %140, float -1.000000e+00, float -0.000000e+00
>  %141 = fcmp oge float %129, 0.000000e+00
>  %142 = sext i1 %141 to i32
>  %143 = bitcast i32 %142 to float
>  %144 = bitcast float %143 to i32
>  %145 = icmp ne i32 %144, 0
>  %.167 = select i1 %145, float -1.000000e+00, float -0.000000e+00
>  %146 = fcmp oge float %130, 0.000000e+00
>  %147 = sext i1 %146 to i32
>  %148 = bitcast i32 %147 to float
>  %149 = bitcast float %148 to i32
>  %150 = icmp ne i32 %149, 0
>  %temp8.0 = select i1 %150, float -1.000000e+00, float -0.000000e+00
>  %151 = fcmp olt float %., 0.000000e+00
>  %152 = sext i1 %151 to i32
>  %153 = fcmp olt float %temp16.0, 0.000000e+00
>  %154 = sext i1 %153 to i32
>  %155 = fcmp olt float %.167, 0.000000e+00
>  %156 = sext i1 %155 to i32
>  %157 = bitcast i32 %152 to float
>  %158 = bitcast i32 %154 to float
>  %159 = bitcast i32 %156 to float
>  %160 = bitcast float %157 to i32
>  %161 = bitcast float %159 to i32
>  %162 = or i32 %160, %161
>  %163 = bitcast i32 %162 to float
>  %164 = bitcast float %163 to i32
>  %165 = bitcast float %158 to i32
>  %166 = or i32 %164, %165
>  %167 = bitcast i32 %166 to float
>  %168 = bitcast float %167 to i32
>  %169 = and i32 %168, 1065353216
>  %170 = bitcast i32 %169 to float
>  %171 = fsub float -0.000000e+00, %170
>  %172 = fsub float -0.000000e+00, %170
>  %173 = fsub float -0.000000e+00, %170
>  %174 = fsub float -0.000000e+00, %170
>  call void @llvm.AMDGPU.kill(float %171)
>  call void @llvm.AMDGPU.kill(float %172)
>  call void @llvm.AMDGPU.kill(float %173)
>  call void @llvm.AMDGPU.kill(float %174)
>  %175 = fmul float %99, %99
>  %176 = fmul float %100, %100
>  %177 = fadd float %175, %176
>  %178 = fmul float %101, %101
>  %179 = fadd float %177, %178
>  %180 = fmul float 0.000000e+00, 0.000000e+00
>  %181 = fadd float %179, %180
>  %182 = call float @llvm.AMDGPU.rsq.clamped.f32(float %181)
>  %183 = fmul float %99, %182
>  %184 = fmul float %100, %182
>  %185 = fmul float %101, %182
>  %186 = fmul float %105, %105
>  %187 = fmul float %106, %106
>  %188 = fadd float %186, %187
>  %189 = fmul float %107, %107
>  %190 = fadd float %188, %189
>  %191 = fmul float 0.000000e+00, 0.000000e+00
>  %192 = fadd float %190, %191
>  %193 = call float @llvm.AMDGPU.rsq.clamped.f32(float %192)
>  %194 = fmul float %105, %193
>  %195 = fmul float %106, %193
>  %196 = fmul float %107, %193
>  %197 = call float @fabs(float %183)
>  %198 = call float @fabs(float %184)
>  %199 = call float @fabs(float %185)
>  %200 = call float @fabs(float %temp8.0)
>  %201 = call float @fabs(float %183)
>  %202 = call float @fabs(float %184)
>  %203 = call float @fabs(float %185)
>  %204 = call float @fabs(float %temp8.0)
>  %205 = fmul float %197, %201
>  %206 = fmul float %198, %202
>  %207 = fmul float %199, %203
>  %208 = fmul float %205, %205
>  %209 = fmul float %206, %206
>  %210 = fmul float %207, %207
>  %211 = fadd float %209, %208
>  %212 = fmul float %207, %207
>  %213 = fadd float %212, %211
>  %214 = fdiv float 1.000000e+00, %213
>  %215 = fmul float %214, %208
>  %216 = fmul float %214, %209
>  %217 = fmul float %214, %210
>  %218 = fmul float %41, %103
>  %219 = fmul float %42, %104
>  %220 = bitcast float %218 to i32
>  %221 = bitcast float %219 to i32
>  %222 = insertelement <2 x i32> undef, i32 %220, i32 0
>  %223 = insertelement <2 x i32> %222, i32 %221, i32 1
>  %224 = bitcast <8 x i32> %72 to <32 x i8>
>  %225 = bitcast <4 x i32> %74 to <16 x i8>
>  %226 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %223, <32 x i8> %224, <16 x i8> %225, i32 2)
>  %227 = extractelement <4 x float> %226, i32 0
>  %228 = extractelement <4 x float> %226, i32 1
>  %229 = extractelement <4 x float> %226, i32 2
>  %230 = call float @llvm.pow.f32(float %227, float 0x40019999A0000000)
>  %231 = call float @llvm.pow.f32(float %228, float 0x40019999A0000000)
>  %232 = call float @llvm.pow.f32(float %229, float 0x40019999A0000000)
>  %233 = fmul float %41, %102
>  %234 = fmul float %42, %104
>  %235 = bitcast float %233 to i32
>  %236 = bitcast float %234 to i32
>  %237 = insertelement <2 x i32> undef, i32 %235, i32 0
>  %238 = insertelement <2 x i32> %237, i32 %236, i32 1
>  %239 = bitcast <8 x i32> %72 to <32 x i8>
>  %240 = bitcast <4 x i32> %74 to <16 x i8>
>  %241 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %238, <32 x i8> %239, <16 x i8> %240, i32 2)
>  %242 = extractelement <4 x float> %241, i32 0
>  %243 = extractelement <4 x float> %241, i32 1
>  %244 = extractelement <4 x float> %241, i32 2
>  %245 = call float @llvm.pow.f32(float %242, float 0x40019999A0000000)
>  %246 = call float @llvm.pow.f32(float %243, float 0x40019999A0000000)
>  %247 = call float @llvm.pow.f32(float %244, float 0x40019999A0000000)
>  %248 = fmul float %216, %245
>  %249 = fmul float %216, %246
>  %250 = fmul float %216, %247
>  %251 = fmul float %230, %215
>  %252 = fadd float %251, %248
>  %253 = fmul float %231, %215
>  %254 = fadd float %253, %249
>  %255 = fmul float %232, %215
>  %256 = fadd float %255, %250
>  %257 = fmul float %34, %102
>  %258 = fmul float %34, %103
>  %259 = bitcast float %257 to i32
>  %260 = bitcast float %258 to i32
>  %261 = insertelement <2 x i32> undef, i32 %259, i32 0
>  %262 = insertelement <2 x i32> %261, i32 %260, i32 1
>  %263 = bitcast <8 x i32> %64 to <32 x i8>
>  %264 = bitcast <4 x i32> %66 to <16 x i8>
>  %265 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %262, <32 x i8> %263, <16 x i8> %264, i32 2)
>  %266 = extractelement <4 x float> %265, i32 0
>  %267 = extractelement <4 x float> %265, i32 1
>  %268 = extractelement <4 x float> %265, i32 2
>  %269 = extractelement <4 x float> %265, i32 3
>  %270 = call float @llvm.pow.f32(float %266, float 0x40019999A0000000)
>  %271 = call float @llvm.pow.f32(float %267, float 0x40019999A0000000)
>  %272 = call float @llvm.pow.f32(float %268, float 0x40019999A0000000)
>  %273 = fmul float %270, %217
>  %274 = fadd float %273, %252
>  %275 = fmul float %271, %217
>  %276 = fadd float %275, %254
>  %277 = fmul float %272, %217
>  %278 = fadd float %277, %256
>  %279 = fmul float %43, %103
>  %280 = fmul float %44, %104
>  %281 = fsub float -0.000000e+00, %185
>  %282 = fmul float %281, %194
>  %283 = fmul float %184, %195
>  %284 = fadd float %283, %282
>  %285 = fmul float %183, %196
>  %286 = fadd float %284, %285
>  %287 = fmul float %183, 1.000000e+00
>  %288 = fmul float %185, 1.000000e+00
>  %289 = fmul float %184, -1.000000e+00
>  %290 = fmul float %289, %194
>  %291 = fmul float %287, %195
>  %292 = fadd float %291, %290
>  %293 = fmul float %288, %196
>  %294 = fadd float %292, %293
>  %295 = fmul float %183, %194
>  %296 = fmul float %184, %195
>  %297 = fadd float %296, %295
>  %298 = fmul float %185, %196
>  %299 = fadd float %297, %298
>  %300 = fmul float %185, -1.000000e+00
>  %301 = fmul float %184, 1.000000e+00
>  %302 = fmul float %183, 1.000000e+00
>  %303 = fmul float %300, %27
>  %304 = fmul float %301, %28
>  %305 = fadd float %304, %303
>  %306 = fmul float %302, %29
>  %307 = fadd float %305, %306
>  %308 = fmul float %289, %27
>  %309 = fmul float %287, %28
>  %310 = fadd float %309, %308
>  %311 = fmul float %288, %29
>  %312 = fadd float %310, %311
>  %313 = fmul float %183, %27
>  %314 = fmul float %184, %28
>  %315 = fadd float %314, %313
>  %316 = fmul float %185, %29
>  %317 = fadd float %315, %316
>  %318 = fmul float %300, %30
>  %319 = fmul float %301, %31
>  %320 = fadd float %319, %318
>  %321 = fmul float %302, %32
>  %322 = fadd float %320, %321
>  %323 = fmul float %289, %30
>  %324 = fmul float %287, %31
>  %325 = fadd float %324, %323
>  %326 = fmul float %288, %32
>  %327 = fadd float %325, %326
>  %328 = fmul float %183, %30
>  %329 = fmul float %184, %31
>  %330 = fadd float %329, %328
>  %331 = fmul float %185, %32
>  %332 = fadd float %330, %331
>  %333 = bitcast float %279 to i32
>  %334 = bitcast float %280 to i32
>  %335 = insertelement <2 x i32> undef, i32 %333, i32 0
>  %336 = insertelement <2 x i32> %335, i32 %334, i32 1
>  %337 = bitcast <8 x i32> %80 to <32 x i8>
>  %338 = bitcast <4 x i32> %82 to <16 x i8>
>  %339 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %336, <32 x i8> %337, <16 x i8> %338, i32 2)
>  %340 = extractelement <4 x float> %339, i32 3
>  %341 = bitcast float %279 to i32
>  %342 = bitcast float %280 to i32
>  %343 = insertelement <2 x i32> undef, i32 %341, i32 0
>  %344 = insertelement <2 x i32> %343, i32 %342, i32 1
>  %345 = bitcast <8 x i32> %76 to <32 x i8>
>  %346 = bitcast <4 x i32> %78 to <16 x i8>
>  %347 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %344, <32 x i8> %345, <16 x i8> %346, i32 2)
>  %348 = extractelement <4 x float> %347, i32 0
>  %349 = extractelement <4 x float> %347, i32 1
>  %350 = extractelement <4 x float> %347, i32 2
>  %351 = extractelement <4 x float> %347, i32 3
>  %352 = fmul float %33, %33
>  %353 = fmul float %348, 2.000000e+00
>  %354 = fadd float %353, -1.000000e+00
>  %355 = fmul float %349, 2.000000e+00
>  %356 = fadd float %355, -1.000000e+00
>  %357 = fmul float %354, %33
>  %358 = fmul float %356, %33
>  %359 = fmul float %357, %357
>  %360 = fmul float %358, %358
>  %361 = fadd float %359, %360
>  %362 = fmul float 1.000000e+00, 1.000000e+00
>  %363 = fadd float %361, %362
>  %364 = fmul float 0.000000e+00, 0.000000e+00
>  %365 = fadd float %363, %364
>  %366 = call float @llvm.AMDGPU.rsq.clamped.f32(float %365)
>  %367 = fmul float %357, %366
>  %368 = fmul float %358, %366
>  %369 = fmul float 1.000000e+00, %366
>  %370 = fmul float %340, 2.000000e+00
>  %371 = fadd float %370, -1.000000e+00
>  %372 = fmul float %357, %357
>  %373 = fmul float %358, %358
>  %374 = fmul float %358, %357
>  %375 = fsub float -0.000000e+00, %372
>  %376 = fmul float %350, %352
>  %377 = fadd float %376, %375
>  %378 = fsub float -0.000000e+00, %373
>  %379 = fmul float %351, %352
>  %380 = fadd float %379, %378
>  %381 = fsub float -0.000000e+00, %374
>  %382 = fmul float %371, %352
>  %383 = fadd float %382, %381
>  %384 = fmul float %307, %367
>  %385 = fmul float %312, %368
>  %386 = fadd float %385, %384
>  %387 = fmul float %317, %369
>  %388 = fadd float %386, %387
>  %389 = call float @llvm.AMDIL.clamp.(float %388, float 0.000000e+00, float 1.000000e+00)
>  %390 = fmul float %322, %367
>  %391 = fmul float %327, %368
>  %392 = fadd float %391, %390
>  %393 = fmul float %332, %369
>  %394 = fadd float %392, %393
>  %395 = call float @llvm.AMDIL.clamp.(float %394, float 0.000000e+00, float 1.000000e+00)
>  %396 = fadd float %286, %322
>  %397 = fadd float %294, %327
>  %398 = fadd float %299, %332
>  %399 = fdiv float 1.000000e+00, %398
>  %400 = fsub float -0.000000e+00, %357
>  %401 = fmul float %396, %399
>  %402 = fadd float %401, %400
>  %403 = fsub float -0.000000e+00, %358
>  %404 = fmul float %397, %399
>  %405 = fadd float %404, %403
>  %406 = fdiv float 1.000000e+00, %36
>  %407 = fadd float %406, %377
>  %408 = fadd float %406, %380
>  %409 = fmul float %383, %383
>  %410 = fsub float -0.000000e+00, %409
>  %411 = fmul float %407, %408
>  %412 = fadd float %411, %410
>  %413 = fmul float %402, %402
>  %414 = fmul float %405, %408
>  %415 = fsub float -0.000000e+00, %383
>  %416 = fsub float -0.000000e+00, %383
>  %417 = fmul float %402, %415
>  %418 = fmul float %402, %416
>  %419 = fadd float %417, %418
>  %420 = fadd float %419, %414
>  %421 = fmul float %405, %420
>  %422 = fmul float %413, %408
>  %423 = fadd float %422, %421
>  %424 = fmul float %423, 5.000000e-01
>  %425 = fdiv float 1.000000e+00, %412
>  %426 = fmul float %424, %425
>  %427 = fsub float -0.000000e+00, %412
>  %428 = fcmp oge float %427, 0.000000e+00
>  %429 = sext i1 %428 to i32
>  %430 = bitcast i32 %429 to float
>  %431 = bitcast float %430 to i32
>  %432 = icmp ne i32 %431, 0
>  %.168 = select i1 %432, float 1.000000e+00, float 0.000000e+00
>  %433 = fmul float %424, %425
>  %434 = fadd float %433, -1.600000e+01
>  %435 = fcmp oge float %434, 0.000000e+00
>  %436 = sext i1 %435 to i32
>  %437 = bitcast i32 %436 to float
>  %438 = bitcast float %437 to i32
>  %439 = icmp ne i32 %438, 0
>  %temp64.0 = select i1 %439, float 1.000000e+00, float 0.000000e+00
>  %440 = fadd float %temp64.0, %.168
>  %441 = fmul float %426, 0xBFF7154CA0000000
>  %442 = call float @llvm.AMDIL.exp.(float %441)
>  %443 = call float @llvm.maxnum.f32(float %412, float 0x3E7AD7F2A0000000)
>  %444 = call float @llvm.AMDGPU.rsq.clamped.f32(float %443)
>  %445 = fmul float %444, %442
>  %446 = fsub float -0.000000e+00, %440
>  %447 = fcmp oge float %446, 0.000000e+00
>  %448 = sext i1 %447 to i32
>  %449 = bitcast i32 %448 to float
>  %450 = bitcast float %449 to i32
>  %451 = icmp ne i32 %450, 0
>  %.169 = select i1 %451, float %445, float 0.000000e+00
>  %452 = fmul float %367, %286
>  %453 = fmul float %369, %299
>  %454 = fadd float %453, %452
>  %455 = fmul float %368, %294
>  %456 = fadd float %454, %455
>  %457 = fsub float -0.000000e+00, %456
>  %458 = fadd float %457, 1.000000e+00
>  %459 = fmul float %458, %458
>  %460 = fmul float %459, %459
>  %461 = fmul float %458, %460
>  %462 = fmul float %185, -1.000000e+00
>  %463 = fmul float %184, -1.000000e+00
>  %464 = fmul float %183, 1.000000e+00
>  %465 = fmul float %367, %462
>  %466 = fmul float %368, %463
>  %467 = fadd float %466, %465
>  %468 = fmul float %369, %464
>  %469 = fadd float %467, %468
>  %470 = fmul float %368, %183
>  %471 = fmul float %367, %184
>  %472 = fadd float %471, %470
>  %473 = fmul float %369, %184
>  %474 = fadd float %472, %473
>  %475 = fmul float %367, %183
>  %476 = fmul float %368, %185
>  %477 = fadd float %476, %475
>  %478 = fmul float %369, %185
>  %479 = fadd float %477, %478
>  %480 = insertelement <4 x float> undef, float %469, i32 0
>  %481 = insertelement <4 x float> %480, float %474, i32 1
>  %482 = insertelement <4 x float> %481, float %479, i32 2
>  %483 = insertelement <4 x float> %482, float 0.000000e+00, i32 3
>  %484 = call <4 x float> @llvm.AMDGPU.cube(<4 x float> %483)
>  %485 = extractelement <4 x float> %484, i32 0
>  %486 = extractelement <4 x float> %484, i32 1
>  %487 = extractelement <4 x float> %484, i32 2
>  %488 = extractelement <4 x float> %484, i32 3
>  %489 = call float @fabs(float %487)
>  %490 = fdiv float 1.000000e+00, %489
>  %491 = fmul float %485, %490
>  %492 = fadd float %491, 1.500000e+00
>  %493 = fmul float %486, %490
>  %494 = fadd float %493, 1.500000e+00
>  %495 = bitcast float %494 to i32
>  %496 = bitcast float %492 to i32
>  %497 = bitcast float %488 to i32
>  %498 = insertelement <4 x i32> undef, i32 %495, i32 0
>  %499 = insertelement <4 x i32> %498, i32 %496, i32 1
>  %500 = insertelement <4 x i32> %499, i32 %497, i32 2
>  %501 = insertelement <4 x i32> %500, i32 undef, i32 3
>  %502 = bitcast <8 x i32> %60 to <32 x i8>
>  %503 = bitcast <4 x i32> %62 to <16 x i8>
>  %504 = call <4 x float> @llvm.SI.sample.v4i32(<4 x i32> %501, <32 x i8> %502, <16 x i8> %503, i32 4)
>  %505 = extractelement <4 x float> %504, i32 0
>  %506 = extractelement <4 x float> %504, i32 1
>  %507 = extractelement <4 x float> %504, i32 2
>  %508 = call float @llvm.pow.f32(float %505, float 0x40019999A0000000)
>  %509 = call float @llvm.pow.f32(float %506, float 0x40019999A0000000)
>  %510 = call float @llvm.pow.f32(float %507, float 0x40019999A0000000)
>  %511 = fadd float %389, %508
>  %512 = fadd float %389, %509
>  %513 = fadd float %389, %510
>  %514 = fmul float %274, %511
>  %515 = fmul float %276, %512
>  %516 = fmul float %278, %513
>  %517 = fmul float %.169, %461
>  %518 = fmul float %.169, %461
>  %519 = fmul float %.169, %461
>  %520 = fmul float %.169, %461
>  %521 = call float @llvm.AMDIL.clamp.(float %517, float 0.000000e+00, float 1.000000e+00)
>  %522 = call float @llvm.AMDIL.clamp.(float %518, float 0.000000e+00, float 1.000000e+00)
>  %523 = call float @llvm.AMDIL.clamp.(float %519, float 0.000000e+00, float 1.000000e+00)
>  %524 = call float @llvm.AMDIL.clamp.(float %520, float 0.000000e+00, float 1.000000e+00)
>  %525 = fmul float %524, %395
>  %526 = fadd float %525, %514
>  %527 = fmul float %524, %395
>  %528 = fadd float %527, %515
>  %529 = fmul float %524, %395
>  %530 = fadd float %529, %516
>  %531 = fmul float %43, %102
>  %532 = fmul float %44, %104
>  %533 = fmul float %185, 1.000000e+00
>  %534 = fmul float %184, 1.000000e+00
>  %535 = fmul float %183, -1.000000e+00
>  %536 = fmul float %535, %194
>  %537 = fmul float %533, %195
>  %538 = fadd float %537, %536
>  %539 = fmul float %534, %196
>  %540 = fadd float %538, %539
>  %541 = fmul float %535, %27
>  %542 = fmul float %533, %28
>  %543 = fadd float %542, %541
>  %544 = fmul float %534, %29
>  %545 = fadd float %543, %544
>  %546 = fmul float %535, %30
>  %547 = fmul float %533, %31
>  %548 = fadd float %547, %546
>  %549 = fmul float %534, %32
>  %550 = fadd float %548, %549
>  %551 = bitcast float %531 to i32
>  %552 = bitcast float %532 to i32
>  %553 = insertelement <2 x i32> undef, i32 %551, i32 0
>  %554 = insertelement <2 x i32> %553, i32 %552, i32 1
>  %555 = bitcast <8 x i32> %80 to <32 x i8>
>  %556 = bitcast <4 x i32> %82 to <16 x i8>
>  %557 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %554, <32 x i8> %555, <16 x i8> %556, i32 2)
>  %558 = extractelement <4 x float> %557, i32 3
>  %559 = bitcast float %531 to i32
>  %560 = bitcast float %532 to i32
>  %561 = insertelement <2 x i32> undef, i32 %559, i32 0
>  %562 = insertelement <2 x i32> %561, i32 %560, i32 1
>  %563 = bitcast <8 x i32> %76 to <32 x i8>
>  %564 = bitcast <4 x i32> %78 to <16 x i8>
>  %565 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %562, <32 x i8> %563, <16 x i8> %564, i32 2)
>  %566 = extractelement <4 x float> %565, i32 0
>  %567 = extractelement <4 x float> %565, i32 1
>  %568 = extractelement <4 x float> %565, i32 2
>  %569 = extractelement <4 x float> %565, i32 3
>  %570 = fmul float %566, 2.000000e+00
>  %571 = fadd float %570, -1.000000e+00
>  %572 = fmul float %567, 2.000000e+00
>  %573 = fadd float %572, -1.000000e+00
>  %574 = fmul float %571, %33
>  %575 = fmul float %573, %33
>  %576 = fmul float %574, %574
>  %577 = fmul float %575, %575
>  %578 = fadd float %576, %577
>  %579 = fmul float 1.000000e+00, 1.000000e+00
>  %580 = fadd float %578, %579
>  %581 = fmul float 0.000000e+00, 0.000000e+00
>  %582 = fadd float %580, %581
>  %583 = call float @llvm.AMDGPU.rsq.clamped.f32(float %582)
>  %584 = fmul float %574, %583
>  %585 = fmul float %575, %583
>  %586 = fmul float 1.000000e+00, %583
>  %587 = fmul float %558, 2.000000e+00
>  %588 = fadd float %587, -1.000000e+00
>  %589 = fmul float %574, %574
>  %590 = fmul float %575, %575
>  %591 = fmul float %575, %574
>  %592 = fsub float -0.000000e+00, %589
>  %593 = fmul float %568, %352
>  %594 = fadd float %593, %592
>  %595 = fsub float -0.000000e+00, %590
>  %596 = fmul float %569, %352
>  %597 = fadd float %596, %595
>  %598 = fsub float -0.000000e+00, %591
>  %599 = fmul float %588, %352
>  %600 = fadd float %599, %598
>  %601 = fmul float %545, %584
>  %602 = fmul float %312, %585
>  %603 = fadd float %602, %601
>  %604 = fmul float %317, %586
>  %605 = fadd float %603, %604
>  %606 = call float @llvm.AMDIL.clamp.(float %605, float 0.000000e+00, float 1.000000e+00)
>  %607 = fmul float %550, %584
>  %608 = fmul float %327, %585
>  %609 = fadd float %608, %607
>  %610 = fmul float %332, %586
>  %611 = fadd float %609, %610
>  %612 = call float @llvm.AMDIL.clamp.(float %611, float 0.000000e+00, float 1.000000e+00)
>  %613 = fadd float %540, %550
>  %614 = fadd float %294, %327
>  %615 = fadd float %299, %332
>  %616 = fdiv float 1.000000e+00, %615
>  %617 = fsub float -0.000000e+00, %574
>  %618 = fmul float %613, %616
>  %619 = fadd float %618, %617
>  %620 = fsub float -0.000000e+00, %575
>  %621 = fmul float %614, %616
>  %622 = fadd float %621, %620
>  %623 = fadd float %406, %594
>  %624 = fadd float %406, %597
>  %625 = fmul float %600, %600
>  %626 = fsub float -0.000000e+00, %625
>  %627 = fmul float %623, %624
>  %628 = fadd float %627, %626
>  %629 = fmul float %619, %619
>  %630 = fmul float %622, %624
>  %631 = fsub float -0.000000e+00, %600
>  %632 = fsub float -0.000000e+00, %600
>  %633 = fmul float %619, %631
>  %634 = fmul float %619, %632
>  %635 = fadd float %633, %634
>  %636 = fadd float %635, %630
>  %637 = fmul float %622, %636
>  %638 = fmul float %629, %624
>  %639 = fadd float %638, %637
>  %640 = fmul float %639, 5.000000e-01
>  %641 = fdiv float 1.000000e+00, %628
>  %642 = fmul float %640, %641
>  %643 = fsub float -0.000000e+00, %628
>  %644 = fcmp oge float %643, 0.000000e+00
>  %645 = sext i1 %644 to i32
>  %646 = bitcast i32 %645 to float
>  %647 = bitcast float %646 to i32
>  %648 = icmp ne i32 %647, 0
>  %temp108.0 = select i1 %648, float 1.000000e+00, float 0.000000e+00
>  %649 = fmul float %640, %641
>  %650 = fadd float %649, -1.600000e+01
>  %651 = fcmp oge float %650, 0.000000e+00
>  %652 = sext i1 %651 to i32
>  %653 = bitcast i32 %652 to float
>  %654 = bitcast float %653 to i32
>  %655 = icmp ne i32 %654, 0
>  %.170 = select i1 %655, float 1.000000e+00, float 0.000000e+00
>  %656 = fadd float %.170, %temp108.0
>  %657 = fmul float %642, 0xBFF7154CA0000000
>  %658 = call float @llvm.AMDIL.exp.(float %657)
>  %659 = call float @llvm.maxnum.f32(float %628, float 0x3E7AD7F2A0000000)
>  %660 = call float @llvm.AMDGPU.rsq.clamped.f32(float %659)
>  %661 = fmul float %660, %658
>  %662 = fsub float -0.000000e+00, %656
>  %663 = fcmp oge float %662, 0.000000e+00
>  %664 = sext i1 %663 to i32
>  %665 = bitcast i32 %664 to float
>  %666 = bitcast float %665 to i32
>  %667 = icmp ne i32 %666, 0
>  %temp64.3 = select i1 %667, float %661, float 0.000000e+00
>  %668 = fmul float %584, %540
>  %669 = fmul float %586, %299
>  %670 = fadd float %669, %668
>  %671 = fmul float %585, %294
>  %672 = fadd float %670, %671
>  %673 = fsub float -0.000000e+00, %672
>  %674 = fadd float %673, 1.000000e+00
>  %675 = fmul float %674, %674
>  %676 = fmul float %675, %675
>  %677 = fmul float %674, %676
>  %678 = fmul float %183, -1.000000e+00
>  %679 = fmul float %184, -1.000000e+00
>  %680 = fmul float %183, 1.000000e+00
>  %681 = fmul float %584, %678
>  %682 = fmul float %585, %679
>  %683 = fadd float %682, %681
>  %684 = fmul float %586, %680
>  %685 = fadd float %683, %684
>  %686 = fmul float %585, %183
>  %687 = fmul float %586, %184
>  %688 = fadd float %687, %686
>  %689 = fmul float %584, %185
>  %690 = fadd float %688, %689
>  %691 = fmul float %584, %184
>  %692 = fmul float %585, %185
>  %693 = fadd float %692, %691
>  %694 = fmul float %586, %185
>  %695 = fadd float %693, %694
>  %696 = insertelement <4 x float> undef, float %685, i32 0
>  %697 = insertelement <4 x float> %696, float %690, i32 1
>  %698 = insertelement <4 x float> %697, float %695, i32 2
>  %699 = insertelement <4 x float> %698, float 0.000000e+00, i32 3
>  %700 = call <4 x float> @llvm.AMDGPU.cube(<4 x float> %699)
>  %701 = extractelement <4 x float> %700, i32 0
>  %702 = extractelement <4 x float> %700, i32 1
>  %703 = extractelement <4 x float> %700, i32 2
>  %704 = extractelement <4 x float> %700, i32 3
>  %705 = call float @fabs(float %703)
>  %706 = fdiv float 1.000000e+00, %705
>  %707 = fmul float %701, %706
>  %708 = fadd float %707, 1.500000e+00
>  %709 = fmul float %702, %706
>  %710 = fadd float %709, 1.500000e+00
>  %711 = bitcast float %710 to i32
>  %712 = bitcast float %708 to i32
>  %713 = bitcast float %704 to i32
>  %714 = insertelement <4 x i32> undef, i32 %711, i32 0
>  %715 = insertelement <4 x i32> %714, i32 %712, i32 1
>  %716 = insertelement <4 x i32> %715, i32 %713, i32 2
>  %717 = insertelement <4 x i32> %716, i32 undef, i32 3
>  %718 = bitcast <8 x i32> %60 to <32 x i8>
>  %719 = bitcast <4 x i32> %62 to <16 x i8>
>  %720 = call <4 x float> @llvm.SI.sample.v4i32(<4 x i32> %717, <32 x i8> %718, <16 x i8> %719, i32 4)
>  %721 = extractelement <4 x float> %720, i32 0
>  %722 = extractelement <4 x float> %720, i32 1
>  %723 = extractelement <4 x float> %720, i32 2
>  %724 = call float @llvm.pow.f32(float %721, float 0x40019999A0000000)
>  %725 = call float @llvm.pow.f32(float %722, float 0x40019999A0000000)
>  %726 = call float @llvm.pow.f32(float %723, float 0x40019999A0000000)
>  %727 = fadd float %606, %724
>  %728 = fadd float %606, %725
>  %729 = fadd float %606, %726
>  %730 = fmul float %274, %727
>  %731 = fmul float %276, %728
>  %732 = fmul float %278, %729
>  %733 = fmul float %temp64.3, %677
>  %734 = fmul float %temp64.3, %677
>  %735 = fmul float %temp64.3, %677
>  %736 = fmul float %temp64.3, %677
>  %737 = call float @llvm.AMDIL.clamp.(float %733, float 0.000000e+00, float 1.000000e+00)
>  %738 = call float @llvm.AMDIL.clamp.(float %734, float 0.000000e+00, float 1.000000e+00)
>  %739 = call float @llvm.AMDIL.clamp.(float %735, float 0.000000e+00, float 1.000000e+00)
>  %740 = call float @llvm.AMDIL.clamp.(float %736, float 0.000000e+00, float 1.000000e+00)
>  %741 = fmul float %740, %612
>  %742 = fadd float %741, %730
>  %743 = fmul float %740, %612
>  %744 = fadd float %743, %731
>  %745 = fmul float %740, %612
>  %746 = fadd float %745, %732
>  %747 = fmul float %35, %102
>  %748 = fmul float %35, %103
>  %749 = fmul float %533, %194
>  %750 = fmul float %534, %195
>  %751 = fadd float %750, %749
>  %752 = fmul float %535, %196
>  %753 = fadd float %751, %752
>  %754 = fmul float %287, %194
>  %755 = fmul float %288, %195
>  %756 = fadd float %755, %754
>  %757 = fmul float %289, %196
>  %758 = fadd float %756, %757
>  %759 = fmul float %533, %27
>  %760 = fmul float %534, %28
>  %761 = fadd float %760, %759
>  %762 = fmul float %535, %29
>  %763 = fadd float %761, %762
>  %764 = fmul float %287, %27
>  %765 = fmul float %288, %28
>  %766 = fadd float %765, %764
>  %767 = fmul float %289, %29
>  %768 = fadd float %766, %767
>  %769 = fmul float %533, %30
>  %770 = fmul float %534, %31
>  %771 = fadd float %770, %769
>  %772 = fmul float %535, %32
>  %773 = fadd float %771, %772
>  %774 = fmul float %287, %30
>  %775 = fmul float %288, %31
>  %776 = fadd float %775, %774
>  %777 = fmul float %289, %32
>  %778 = fadd float %776, %777
>  %779 = bitcast float %747 to i32
>  %780 = bitcast float %748 to i32
>  %781 = insertelement <2 x i32> undef, i32 %779, i32 0
>  %782 = insertelement <2 x i32> %781, i32 %780, i32 1
>  %783 = bitcast <8 x i32> %68 to <32 x i8>
>  %784 = bitcast <4 x i32> %70 to <16 x i8>
>  %785 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %782, <32 x i8> %783, <16 x i8> %784, i32 2)
>  %786 = extractelement <4 x float> %785, i32 0
>  %787 = extractelement <4 x float> %785, i32 1
>  %788 = extractelement <4 x float> %785, i32 2
>  %789 = extractelement <4 x float> %785, i32 3
>  %790 = fmul float %786, 2.000000e+00
>  %791 = fadd float %790, -1.000000e+00
>  %792 = fmul float %787, 2.000000e+00
>  %793 = fadd float %792, -1.000000e+00
>  %794 = fmul float %791, %33
>  %795 = fmul float %793, %33
>  %796 = fmul float %794, %794
>  %797 = fmul float %795, %795
>  %798 = fadd float %796, %797
>  %799 = fmul float 1.000000e+00, 1.000000e+00
>  %800 = fadd float %798, %799
>  %801 = fmul float 0.000000e+00, 0.000000e+00
>  %802 = fadd float %800, %801
>  %803 = call float @llvm.AMDGPU.rsq.clamped.f32(float %802)
>  %804 = fmul float %794, %803
>  %805 = fmul float %795, %803
>  %806 = fmul float 1.000000e+00, %803
>  %807 = fmul float %788, 1.000000e+00
>  %808 = fadd float %807, 0.000000e+00
>  %809 = fmul float %789, 1.000000e+00
>  %810 = fadd float %809, 0.000000e+00
>  %811 = fmul float %789, 2.000000e+00
>  %812 = fadd float %811, -1.000000e+00
>  %813 = fmul float %794, %794
>  %814 = fmul float %795, %795
>  %815 = fmul float %795, %794
>  %816 = fsub float -0.000000e+00, %813
>  %817 = fmul float %808, %352
>  %818 = fadd float %817, %816
>  %819 = fsub float -0.000000e+00, %814
>  %820 = fmul float %810, %352
>  %821 = fadd float %820, %819
>  %822 = fsub float -0.000000e+00, %815
>  %823 = fmul float %812, %352
>  %824 = fadd float %823, %822
>  %825 = fmul float %763, %804
>  %826 = fmul float %768, %805
>  %827 = fadd float %826, %825
>  %828 = fmul float %317, %806
>  %829 = fadd float %827, %828
>  %830 = call float @llvm.AMDIL.clamp.(float %829, float 0.000000e+00, float 1.000000e+00)
>  %831 = fmul float %773, %804
>  %832 = fmul float %778, %805
>  %833 = fadd float %832, %831
>  %834 = fmul float %332, %806
>  %835 = fadd float %833, %834
>  %836 = call float @llvm.AMDIL.clamp.(float %835, float 0.000000e+00, float 1.000000e+00)
>  %837 = fadd float %753, %773
>  %838 = fadd float %758, %778
>  %839 = fadd float %299, %332
>  %840 = fdiv float 1.000000e+00, %839
>  %841 = fsub float -0.000000e+00, %794
>  %842 = fmul float %837, %840
>  %843 = fadd float %842, %841
>  %844 = fsub float -0.000000e+00, %795
>  %845 = fmul float %838, %840
>  %846 = fadd float %845, %844
>  %847 = fadd float %406, %818
>  %848 = fadd float %406, %821
>  %849 = fmul float %824, %824
>  %850 = fsub float -0.000000e+00, %849
>  %851 = fmul float %847, %848
>  %852 = fadd float %851, %850
>  %853 = fmul float %843, %843
>  %854 = fmul float %846, %848
>  %855 = fsub float -0.000000e+00, %824
>  %856 = fsub float -0.000000e+00, %824
>  %857 = fmul float %843, %855
>  %858 = fmul float %843, %856
>  %859 = fadd float %857, %858
>  %860 = fadd float %859, %854
>  %861 = fmul float %846, %860
>  %862 = fmul float %853, %848
>  %863 = fadd float %862, %861
>  %864 = fmul float %863, 5.000000e-01
>  %865 = fdiv float 1.000000e+00, %852
>  %866 = fmul float %864, %865
>  %867 = fsub float -0.000000e+00, %852
>  %868 = fcmp oge float %867, 0.000000e+00
>  %869 = sext i1 %868 to i32
>  %870 = bitcast i32 %869 to float
>  %871 = bitcast float %870 to i32
>  %872 = icmp ne i32 %871, 0
>  %.171 = select i1 %872, float 1.000000e+00, float 0.000000e+00
>  %873 = fmul float %864, %865
>  %874 = fadd float %873, -1.600000e+01
>  %875 = fcmp oge float %874, 0.000000e+00
>  %876 = sext i1 %875 to i32
>  %877 = bitcast i32 %876 to float
>  %878 = bitcast float %877 to i32
>  %879 = icmp ne i32 %878, 0
>  %temp40.0 = select i1 %879, float 1.000000e+00, float 0.000000e+00
>  %880 = fadd float %temp40.0, %.171
>  %881 = fmul float %866, 0xBFF7154CA0000000
>  %882 = call float @llvm.AMDIL.exp.(float %881)
>  %883 = call float @llvm.maxnum.f32(float %852, float 0x3E7AD7F2A0000000)
>  %884 = call float @llvm.AMDGPU.rsq.clamped.f32(float %883)
>  %885 = fmul float %884, %882
>  %886 = fsub float -0.000000e+00, %880
>  %887 = fcmp oge float %886, 0.000000e+00
>  %888 = sext i1 %887 to i32
>  %889 = bitcast i32 %888 to float
>  %890 = bitcast float %889 to i32
>  %891 = icmp ne i32 %890, 0
>  %.172 = select i1 %891, float %885, float 0.000000e+00
>  %892 = fmul float %804, %753
>  %893 = fmul float %805, %758
>  %894 = fadd float %893, %892
>  %895 = fmul float %806, %299
>  %896 = fadd float %894, %895
>  %897 = fsub float -0.000000e+00, %896
>  %898 = fadd float %897, 1.000000e+00
>  %899 = fmul float %898, %898
>  %900 = fmul float %899, %899
>  %901 = fmul float %898, %900
>  %902 = fmul float %805, %183
>  %903 = fmul float %806, %183
>  %904 = fadd float %903, %902
>  %905 = fmul float %804, %185
>  %906 = fadd float %904, %905
>  %907 = fmul float %804, %184
>  %908 = fmul float %806, %184
>  %909 = fadd float %908, %907
>  %910 = fmul float %805, %185
>  %911 = fadd float %909, %910
>  %912 = fmul float %183, -1.000000e+00
>  %913 = fmul float %184, -1.000000e+00
>  %914 = fmul float %185, 1.000000e+00
>  %915 = fmul float %804, %912
>  %916 = fmul float %805, %913
>  %917 = fadd float %916, %915
>  %918 = fmul float %806, %914
>  %919 = fadd float %917, %918
>  %920 = insertelement <4 x float> undef, float %906, i32 0
>  %921 = insertelement <4 x float> %920, float %911, i32 1
>  %922 = insertelement <4 x float> %921, float %919, i32 2
>  %923 = insertelement <4 x float> %922, float %900, i32 3
>  %924 = call <4 x float> @llvm.AMDGPU.cube(<4 x float> %923)
>  %925 = extractelement <4 x float> %924, i32 0
>  %926 = extractelement <4 x float> %924, i32 1
>  %927 = extractelement <4 x float> %924, i32 2
>  %928 = extractelement <4 x float> %924, i32 3
>  %929 = call float @fabs(float %927)
>  %930 = fdiv float 1.000000e+00, %929
>  %931 = fmul float %925, %930
>  %932 = fadd float %931, 1.500000e+00
>  %933 = fmul float %926, %930
>  %934 = fadd float %933, 1.500000e+00
>  %935 = bitcast float %934 to i32
>  %936 = bitcast float %932 to i32
>  %937 = bitcast float %928 to i32
>  %938 = insertelement <4 x i32> undef, i32 %935, i32 0
>  %939 = insertelement <4 x i32> %938, i32 %936, i32 1
>  %940 = insertelement <4 x i32> %939, i32 %937, i32 2
>  %941 = insertelement <4 x i32> %940, i32 undef, i32 3
>  %942 = bitcast <8 x i32> %60 to <32 x i8>
>  %943 = bitcast <4 x i32> %62 to <16 x i8>
>  %944 = call <4 x float> @llvm.SI.sample.v4i32(<4 x i32> %941, <32 x i8> %942, <16 x i8> %943, i32 4)
>  %945 = extractelement <4 x float> %944, i32 0
>  %946 = extractelement <4 x float> %944, i32 1
>  %947 = extractelement <4 x float> %944, i32 2
>  %948 = call float @llvm.pow.f32(float %945, float 0x40019999A0000000)
>  %949 = call float @llvm.pow.f32(float %946, float 0x40019999A0000000)
>  %950 = call float @llvm.pow.f32(float %947, float 0x40019999A0000000)
>  %951 = fadd float %830, %948
>  %952 = fadd float %830, %949
>  %953 = fadd float %830, %950
>  %954 = fmul float %274, %951
>  %955 = fmul float %276, %952
>  %956 = fmul float %278, %953
>  %957 = fmul float %.172, %901
>  %958 = fmul float %.172, %901
>  %959 = fmul float %.172, %901
>  %960 = fmul float %.172, %901
>  %961 = call float @llvm.AMDIL.clamp.(float %957, float 0.000000e+00, float 1.000000e+00)
>  %962 = call float @llvm.AMDIL.clamp.(float %958, float 0.000000e+00, float 1.000000e+00)
>  %963 = call float @llvm.AMDIL.clamp.(float %959, float 0.000000e+00, float 1.000000e+00)
>  %964 = call float @llvm.AMDIL.clamp.(float %960, float 0.000000e+00, float 1.000000e+00)
>  %965 = fmul float %962, %836
>  %966 = fadd float %965, %954
>  %967 = fmul float %962, %836
>  %968 = fadd float %967, %955
>  %969 = fmul float %962, %836
>  %970 = fadd float %969, %956
>  %971 = fmul float %216, %742
>  %972 = fmul float %216, %744
>  %973 = fmul float %216, %746
>  %974 = fmul float %526, %215
>  %975 = fadd float %974, %971
>  %976 = fmul float %528, %215
>  %977 = fadd float %976, %972
>  %978 = fmul float %530, %215
>  %979 = fadd float %978, %973
>  %980 = fmul float %966, %217
>  %981 = fadd float %980, %975
>  %982 = fmul float %968, %217
>  %983 = fadd float %982, %977
>  %984 = fmul float %970, %217
>  %985 = fadd float %984, %979
>  %986 = fmul float %274, 0x3FD3333340000000
>  %987 = fmul float %276, 0x3FE2E147A0000000
>  %988 = fadd float %987, %986
>  %989 = fmul float %278, 0x3FBC28F5C0000000
>  %990 = fadd float %988, %989
>  %991 = fsub float -0.000000e+00, %990
>  %992 = fadd float %991, 1.000000e+00
>  %993 = fmul float %992, %37
>  %994 = fmul float %992, %38
>  %995 = fmul float %992, %39
>  %996 = call float @llvm.AMDIL.clamp.(float %317, float 0.000000e+00, float 1.000000e+00)
>  %997 = insertelement <4 x float> undef, float %183, i32 0
>  %998 = insertelement <4 x float> %997, float %184, i32 1
>  %999 = insertelement <4 x float> %998, float %185, i32 2
>  %1000 = insertelement <4 x float> %999, float %269, i32 3
>  %1001 = call <4 x float> @llvm.AMDGPU.cube(<4 x float> %1000)
>  %1002 = extractelement <4 x float> %1001, i32 0
>  %1003 = extractelement <4 x float> %1001, i32 1
>  %1004 = extractelement <4 x float> %1001, i32 2
>  %1005 = extractelement <4 x float> %1001, i32 3
>  %1006 = call float @fabs(float %1004)
>  %1007 = fdiv float 1.000000e+00, %1006
>  %1008 = fmul float %1002, %1007
>  %1009 = fadd float %1008, 1.500000e+00
>  %1010 = fmul float %1003, %1007
>  %1011 = fadd float %1010, 1.500000e+00
>  %1012 = bitcast float %1011 to i32
>  %1013 = bitcast float %1009 to i32
>  %1014 = bitcast float %1005 to i32
>  %1015 = insertelement <4 x i32> undef, i32 %1012, i32 0
>  %1016 = insertelement <4 x i32> %1015, i32 %1013, i32 1
>  %1017 = insertelement <4 x i32> %1016, i32 %1014, i32 2
>  %1018 = insertelement <4 x i32> %1017, i32 undef, i32 3
>  %1019 = bitcast <8 x i32> %60 to <32 x i8>
>  %1020 = bitcast <4 x i32> %62 to <16 x i8>
>  %1021 = call <4 x float> @llvm.SI.sample.v4i32(<4 x i32> %1018, <32 x i8> %1019, <16 x i8> %1020, i32 4)
>  %1022 = extractelement <4 x float> %1021, i32 0
>  %1023 = extractelement <4 x float> %1021, i32 1
>  %1024 = extractelement <4 x float> %1021, i32 2
>  %1025 = call float @llvm.pow.f32(float %1022, float 0x40019999A0000000)
>  %1026 = call float @llvm.pow.f32(float %1023, float 0x40019999A0000000)
>  %1027 = call float @llvm.pow.f32(float %1024, float 0x40019999A0000000)
>  %1028 = fadd float %1025, %996
>  %1029 = fadd float %1026, %996
>  %1030 = fadd float %1027, %996
>  %1031 = fmul float %993, %1028
>  %1032 = fadd float %1031, %981
>  %1033 = fmul float %994, %1029
>  %1034 = fadd float %1033, %983
>  %1035 = fmul float %995, %1030
>  %1036 = fadd float %1035, %985
>  %1037 = fmul float %108, 5.000000e-01
>  %1038 = fadd float %1037, 5.000000e-01
>  %1039 = bitcast float %1038 to i32
>  %1040 = bitcast float %58 to i32
>  %1041 = insertelement <2 x i32> undef, i32 %1039, i32 0
>  %1042 = insertelement <2 x i32> %1041, i32 %1040, i32 1
>  %1043 = bitcast <8 x i32> %84 to <32 x i8>
>  %1044 = bitcast <4 x i32> %86 to <16 x i8>
>  %1045 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %1042, <32 x i8> %1043, <16 x i8> %1044, i32 2)
>  %1046 = extractelement <4 x float> %1045, i32 0
>  %1047 = fsub float -0.000000e+00, %24
>  %1048 = fadd float %1047, %102
>  %1049 = fsub float -0.000000e+00, %25
>  %1050 = fadd float %1049, %103
>  %1051 = fsub float -0.000000e+00, %26
>  %1052 = fadd float %1051, %104
>  %1053 = fmul float %1048, %1048
>  %1054 = fmul float %1050, %1050
>  %1055 = fadd float %1054, %1053
>  %1056 = fmul float %1052, %1052
>  %1057 = fadd float %1055, %1056
>  %1058 = fmul float %1057, %57
>  %1059 = fmul float %1052, %56
>  %1060 = fmul float %1059, 0x3FF7154CA0000000
>  %1061 = call float @llvm.AMDIL.exp.(float %1060)
>  %1062 = fsub float -0.000000e+00, %1061
>  %1063 = fadd float %1062, 1.000000e+00
>  %1064 = fmul float %1063, %1058
>  %1065 = fdiv float 1.000000e+00, %1052
>  %1066 = fmul float %1065, %1064
>  %1067 = fmul float %1066, 0x3FF7154CA0000000
>  %1068 = call float @llvm.AMDIL.exp.(float %1067)
>  %1069 = call float @llvm.AMDIL.clamp.(float %1068, float 0.000000e+00, float 1.000000e+00)
>  %1070 = fsub float -0.000000e+00, %1069
>  %1071 = fadd float %1070, 1.000000e+00
>  %1072 = fmul float %1071, %1046
>  %1073 = fsub float -0.000000e+00, %40
>  %1074 = fadd float %1073, %104
>  %1075 = fcmp oge float %1074, 0.000000e+00
>  %1076 = sext i1 %1075 to i32
>  %1077 = bitcast i32 %1076 to float
>  %1078 = bitcast float %1077 to i32
>  %1079 = icmp ne i32 %1078, 0
>  %temp48.0 = select i1 %1079, float 1.000000e+00, float 0.000000e+00
>  %1080 = fsub float -0.000000e+00, %104
>  %1081 = fadd float %40, %1080
>  %1082 = fsub float -0.000000e+00, %45
>  %1083 = fadd float %1082, %48
>  %1084 = fsub float -0.000000e+00, %46
>  %1085 = fadd float %1084, %49
>  %1086 = fsub float -0.000000e+00, %47
>  %1087 = fadd float %1086, %50
>  %1088 = fmul float %1081, %51
>  %1089 = fmul float %1081, %51
>  %1090 = fmul float %1081, %51
>  %1091 = fmul float %1081, %51
>  %1092 = call float @llvm.AMDIL.clamp.(float %1088, float 0.000000e+00, float 1.000000e+00)
>  %1093 = call float @llvm.AMDIL.clamp.(float %1089, float 0.000000e+00, float 1.000000e+00)
>  %1094 = call float @llvm.AMDIL.clamp.(float %1090, float 0.000000e+00, float 1.000000e+00)
>  %1095 = call float @llvm.AMDIL.clamp.(float %1091, float 0.000000e+00, float 1.000000e+00)
>  %1096 = fmul float %1093, %1083
>  %1097 = fadd float %1096, %45
>  %1098 = fmul float %1093, %1085
>  %1099 = fadd float %1098, %46
>  %1100 = fmul float %1093, %1087
>  %1101 = fadd float %1100, %47
>  %1102 = fmul float %1032, %1097
>  %1103 = fmul float %1034, %1099
>  %1104 = fmul float %1036, %1101
>  %1105 = fsub float -0.000000e+00, %196
>  %1106 = fdiv float 1.000000e+00, %1105
>  %1107 = fmul float %1081, %1106
>  %1108 = fmul float %1107, %194
>  %1109 = fmul float %1107, %195
>  %1110 = fmul float %1107, %196
>  %1111 = fmul float %1108, %1108
>  %1112 = fmul float %1109, %1109
>  %1113 = fadd float %1112, %1111
>  %1114 = fmul float %1110, %1110
>  %1115 = fadd float %1113, %1114
>  %1116 = call float @llvm.maxnum.f32(float %1115, float 0x3E7AD7F2A0000000)
>  %1117 = call float @llvm.sqrt.f32(float %1116)
>  %1118 = fsub float -0.000000e+00, %1117
>  %1119 = fmul float %1118, %52
>  %1120 = fmul float %1119, 0x3FF7154CA0000000
>  %1121 = call float @llvm.AMDIL.exp.(float %1120)
>  %1122 = fmul float %196, %196
>  %1123 = fmul float %1122, %1122
>  %1124 = fsub float -0.000000e+00, %1123
>  %1125 = fmul float %196, %1124
>  %1126 = fadd float %1125, 1.000000e+00
>  %1127 = fsub float -0.000000e+00, %1126
>  %1128 = fadd float %1127, 1.000000e+00
>  %1129 = fmul float %1121, %1128
>  %1130 = fmul float %1129, %1102
>  %1131 = fmul float %1129, %1103
>  %1132 = fmul float %1129, %1104
>  %1133 = fsub float -0.000000e+00, %1071
>  %1134 = fmul float %1046, %1133
>  %1135 = fadd float %1134, 1.000000e+00
>  %1136 = fmul float %1135, %1130
>  %1137 = fmul float %1135, %1131
>  %1138 = fmul float %1135, %1132
>  %1139 = fcmp oge float %1074, 0.000000e+00
>  %1140 = sext i1 %1139 to i32
>  %1141 = bitcast i32 %1140 to float
>  %1142 = bitcast float %1141 to i32
>  %1143 = icmp ne i32 %1142, 0
>  %.173 = select i1 %1143, float %1032, float %1136
>  %1144 = fcmp oge float %1074, 0.000000e+00
>  %1145 = sext i1 %1144 to i32
>  %1146 = bitcast i32 %1145 to float
>  %1147 = bitcast float %1146 to i32
>  %1148 = icmp ne i32 %1147, 0
>  %temp32.0 = select i1 %1148, float %1034, float %1137
>  %1149 = fcmp oge float %1074, 0.000000e+00
>  %1150 = sext i1 %1149 to i32
>  %1151 = bitcast i32 %1150 to float
>  %1152 = bitcast float %1151 to i32
>  %1153 = icmp ne i32 %1152, 0
>  %.174 = select i1 %1153, float %1036, float %1138
>  %1154 = fcmp oge float %1074, 0.000000e+00
>  %1155 = sext i1 %1154 to i32
>  %1156 = bitcast i32 %1155 to float
>  %1157 = bitcast float %1156 to i32
>  %1158 = icmp ne i32 %1157, 0
>  %temp16.2 = select i1 %1158, float %1072, float 0.000000e+00
>  %1159 = call float @fabs(float %.173)
>  %1160 = call float @llvm.log2.f32(float %1159)
>  %1161 = call float @fabs(float %temp32.0)
>  %1162 = call float @llvm.log2.f32(float %1161)
>  %1163 = call float @fabs(float %.174)
>  %1164 = call float @llvm.log2.f32(float %1163)
>  %1165 = fmul float %1160, 0x3FDD1743E0000000
>  %1166 = fmul float %1162, 0x3FDD1743E0000000
>  %1167 = fmul float %1164, 0x3FDD1743E0000000
>  %1168 = call float @llvm.AMDIL.exp.(float %1165)
>  %1169 = call float @llvm.AMDIL.exp.(float %1166)
>  %1170 = call float @llvm.AMDIL.exp.(float %1167)
>  %1171 = bitcast float %111 to i32
>  %1172 = bitcast float %112 to i32
>  %1173 = insertelement <2 x i32> undef, i32 %1171, i32 0
>  %1174 = insertelement <2 x i32> %1173, i32 %1172, i32 1
>  %1175 = bitcast <8 x i32> %92 to <32 x i8>
>  %1176 = bitcast <4 x i32> %94 to <16 x i8>
>  %1177 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %1174, <32 x i8> %1175, <16 x i8> %1176, i32 2)
>  %1178 = extractelement <4 x float> %1177, i32 0
>  %1179 = extractelement <4 x float> %1177, i32 1
>  %1180 = extractelement <4 x float> %1177, i32 2
>  %1181 = call float @llvm.AMDGPU.lrp(float %124, float %1168, float %1178)
>  %1182 = call float @llvm.AMDGPU.lrp(float %124, float %1169, float %1179)
>  %1183 = call float @llvm.AMDGPU.lrp(float %124, float %1170, float %1180)
>  %1184 = bitcast float %1181 to i32
>  %1185 = bitcast float %1182 to i32
>  %1186 = bitcast float %1183 to i32
>  %1187 = insertelement <4 x i32> undef, i32 %1184, i32 0
>  %1188 = insertelement <4 x i32> %1187, i32 %1185, i32 1
>  %1189 = insertelement <4 x i32> %1188, i32 %1186, i32 2
>  %1190 = insertelement <4 x i32> %1189, i32 undef, i32 3
>  %1191 = bitcast <8 x i32> %96 to <32 x i8>
>  %1192 = bitcast <4 x i32> %98 to <16 x i8>
>  %1193 = call <4 x float> @llvm.SI.sample.v4i32(<4 x i32> %1190, <32 x i8> %1191, <16 x i8> %1192, i32 3)
>  %1194 = extractelement <4 x float> %1193, i32 0
>  %1195 = extractelement <4 x float> %1193, i32 1
>  %1196 = extractelement <4 x float> %1193, i32 2
>  %1197 = fsub float -0.000000e+00, %124
>  %1198 = fmul float %125, %1197
>  %1199 = fadd float %1198, %124
>  %1200 = call float @llvm.AMDGPU.lrp(float %1199, float %1194, float %1181)
>  %1201 = call float @llvm.AMDGPU.lrp(float %1199, float %1195, float %1182)
>  %1202 = call float @llvm.AMDGPU.lrp(float %1199, float %1196, float %1183)
>  %1203 = fsub float -0.000000e+00, %1200
>  %1204 = fadd float %1203, %53
>  %1205 = fsub float -0.000000e+00, %1201
>  %1206 = fadd float %1205, %54
>  %1207 = fsub float -0.000000e+00, %1202
>  %1208 = fadd float %1207, %55
>  %1209 = fmul float %temp16.2, %1204
>  %1210 = fadd float %1209, %1200
>  %1211 = fmul float %temp16.2, %1206
>  %1212 = fadd float %1211, %1201
>  %1213 = fmul float %temp16.2, %1208
>  %1214 = fadd float %1213, %1202
>  %1215 = call i32 @llvm.SI.packf16(float %1210, float %1212)
>  %1216 = bitcast i32 %1215 to float
>  %1217 = call i32 @llvm.SI.packf16(float %1214, float %temp48.0)
>  %1218 = bitcast i32 %1217 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %1216, float %1218, float %1216, float %1218)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: readnone
>declare float @fabs(float) #2
>
>declare void @llvm.AMDGPU.kill(float)
>
>; Function Attrs: nounwind readnone
>declare float @llvm.AMDGPU.rsq.clamped.f32(float) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.pow.f32(float, float) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.clamp.(float, float, float) #2
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.exp.(float) #2
>
>; Function Attrs: nounwind readnone
>declare float @llvm.maxnum.f32(float, float) #1
>
>; Function Attrs: readnone
>declare <4 x float> @llvm.AMDGPU.cube(<4 x float>) #2
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v4i32(<4 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.sqrt.f32(float) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.log2.f32(float) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDGPU.lrp(float, float, float) #2
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                         ; BEFE0A7E
>	s_mov_b32 m0, s9                                                             ; BEFC0309
>	v_interp_p1_f32 v2, v0, 1, 4, [m0]                                           ; C8081100
>	v_interp_p2_f32 v2, [v2], v1, 1, 4, [m0]                                     ; C8091101
>	v_sub_f32_e32 v3, 1.0, v2                                                    ; 080604F2
>	v_interp_p1_f32 v4, v0, 0, 4, [m0]                                           ; C8101000
>	v_interp_p2_f32 v4, [v4], v1, 0, 4, [m0]                                     ; C8111001
>	v_mov_b32_e32 v18, 0                                                         ; 7E240280
>	v_add_f32_e32 v2, 0, v4                                                      ; 06040880
>	s_load_dwordx4 s[28:31], s[4:5], 0x0                                         ; C08E0500
>	s_load_dwordx4 s[80:83], s[4:5], 0x4                                         ; C0A80504
>	s_load_dwordx4 s[40:43], s[4:5], 0x8                                         ; C0940508
>	s_load_dwordx4 s[92:95], s[4:5], 0xc                                         ; C0AE050C
>	s_load_dwordx4 s[68:71], s[4:5], 0x10                                        ; C0A20510
>	s_load_dwordx4 s[52:55], s[4:5], 0x14                                        ; C09A0514
>	s_load_dwordx4 s[24:27], s[4:5], 0x18                                        ; C08C0518
>	s_load_dwordx4 s[56:59], s[4:5], 0x1c                                        ; C09C051C
>	s_load_dwordx4 s[8:11], s[4:5], 0x20                                         ; C0840520
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_writelane_b32 v82, s8, 0                                                   ; 04A50008
>	v_writelane_b32 v82, s9, 1                                                   ; 04A50209
>	v_writelane_b32 v82, s10, 2                                                  ; 04A5040A
>	v_writelane_b32 v82, s11, 3                                                  ; 04A5060B
>	s_load_dwordx4 s[96:99], s[4:5], 0x24                                        ; C0B00524
>	s_load_dwordx8 s[32:39], s[6:7], 0x0                                         ; C0D00700
>	s_load_dwordx8 s[84:91], s[6:7], 0x8                                         ; C0EA0708
>	s_load_dwordx8 s[44:51], s[6:7], 0x10                                        ; C0D60710
>	s_load_dwordx8 s[8:15], s[6:7], 0x18                                         ; C0C40718
>	s_load_dwordx8 s[72:79], s[6:7], 0x20                                        ; C0E40720
>	s_load_dwordx8 s[60:67], s[6:7], 0x28                                        ; C0DE0728
>	s_load_dwordx8 s[16:23], s[6:7], 0x30                                        ; C0C80730
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_writelane_b32 v82, s16, 4                                                  ; 04A50810
>	v_writelane_b32 v82, s17, 5                                                  ; 04A50A11
>	v_writelane_b32 v82, s18, 6                                                  ; 04A50C12
>	v_writelane_b32 v82, s19, 7                                                  ; 04A50E13
>	v_writelane_b32 v82, s20, 8                                                  ; 04A51014
>	v_writelane_b32 v82, s21, 9                                                  ; 04A51215
>	v_writelane_b32 v82, s22, 10                                                 ; 04A51416
>	v_writelane_b32 v82, s23, 11                                                 ; 04A51617
>	s_load_dwordx8 s[16:23], s[6:7], 0x38                                        ; C0C80738
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample v[2:3], 5, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[16:23], s[56:59]      ; F0800500 01C40202
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_cmp_ge_f32_e64 s[0:1], -|v2|, 0                                            ; D00C0100 20010102
>	v_cndmask_b32_e64 v4, 0, -1, s[0:1]                                          ; D2000004 00018280
>	v_cmp_ne_i32_e64 s[0:1], v4, 0                                               ; D10A0000 00010104
>	v_mov_b32_e32 v4, 0x80000000                                                 ; 7E0802FF 80000000
>	v_cndmask_b32_e64 v4, v4, -1.0, s[0:1]                                       ; D2000004 0001E704
>	v_cmp_lt_f32_e64 s[0:1], v4, 0                                               ; D0020000 00010104
>	v_cndmask_b32_e64 v4, 0, -1, s[0:1]                                          ; D2000004 00018280
>	v_and_b32_e32 v4, 1.0, v4                                                    ; 360808F2
>	v_xor_b32_e32 v4, 0x80000000, v4                                             ; 3A0808FF 80000000
>	v_cmpx_le_f32_e32 vcc, 0, v4                                                 ; 7C260880
>	v_cmpx_le_f32_e32 vcc, 0, v4                                                 ; 7C260880
>	v_cmpx_le_f32_e32 vcc, 0, v4                                                 ; 7C260880
>	v_cmpx_le_f32_e32 vcc, 0, v4                                                 ; 7C260880
>	v_interp_p1_f32 v5, v0, 2, 1, [m0]                                           ; C8140600
>	v_interp_p2_f32 v5, [v5], v1, 2, 1, [m0]                                     ; C8150601
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                                           ; C0800300
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	s_buffer_load_dword s4, s[0:3], 0x29                                         ; C2020129
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v20, s4, v5                                                    ; 10280A04
>	v_interp_p1_f32 v12, v0, 1, 1, [m0]                                          ; C8300500
>	v_interp_p2_f32 v12, [v12], v1, 1, 1, [m0]                                   ; C8310501
>	s_buffer_load_dword s59, s[0:3], 0x28                                        ; C21D8128
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v19, s59, v12                                                  ; 1026183B
>	image_sample v[21:24], 15, 0, 0, 0, 0, 0, 0, 0, v[19:20], s[72:79], s[68:71] ; F0800F00 02321513
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v4, 2.0, v21, -1.0                                                 ; D2820004 03CE2AF4
>	s_buffer_load_dword s4, s[0:3], 0xc                                          ; C202010C
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v25, s4, v4                                                    ; 10320804
>	v_mad_f32 v4, 2.0, v22, -1.0                                                 ; D2820004 03CE2CF4
>	v_mul_f32_e32 v26, s4, v4                                                    ; 10340804
>	v_mul_f32_e32 v27, v26, v26                                                  ; 1036351A
>	v_mad_f32 v4, v25, v25, v27                                                  ; D2820004 046E3319
>	v_add_f32_e32 v4, 1.0, v4                                                    ; 060808F2
>	v_add_f32_e32 v4, 0, v4                                                      ; 06080880
>	v_rsq_clamp_f32_e32 v28, v4                                                  ; 7E385904
>	v_mul_f32_e32 v29, v28, v26                                                  ; 103A351C
>	v_mul_f32_e32 v30, v28, v25                                                  ; 103C331C
>	v_interp_p1_f32 v4, v0, 0, 0, [m0]                                           ; C8100000
>	v_interp_p2_f32 v4, [v4], v1, 0, 0, [m0]                                     ; C8110001
>	v_interp_p1_f32 v6, v0, 1, 0, [m0]                                           ; C8180100
>	v_interp_p2_f32 v6, [v6], v1, 1, 0, [m0]                                     ; C8190101
>	v_mul_f32_e32 v7, v6, v6                                                     ; 100E0D06
>	v_mad_f32 v7, v4, v4, v7                                                     ; D2820007 041E0904
>	v_interp_p1_f32 v8, v0, 2, 0, [m0]                                           ; C8200200
>	v_interp_p2_f32 v8, [v8], v1, 2, 0, [m0]                                     ; C8210201
>	v_mad_f32 v7, v8, v8, v7                                                     ; D2820007 041E1108
>	v_add_f32_e32 v7, 0, v7                                                      ; 060E0E80
>	v_rsq_clamp_f32_e32 v7, v7                                                   ; 7E0E5907
>	v_mul_f32_e32 v31, v7, v4                                                    ; 103E0907
>	v_mul_f32_e32 v4, v31, v30                                                   ; 10083D1F
>	v_mul_f32_e32 v33, v7, v8                                                    ; 10421107
>	v_mad_f32 v4, v29, v33, v4                                                   ; D2820004 0412431D
>	v_mad_f32 v17, v28, v33, v4                                                  ; D2820011 0412431C
>	v_mul_f32_e32 v4, v31, v29                                                   ; 10083B1F
>	v_mul_f32_e32 v32, v7, v6                                                    ; 10400D07
>	v_mad_f32 v4, v30, v32, v4                                                   ; D2820004 0412411E
>	v_mad_f32 v16, v28, v32, v4                                                  ; D2820010 0412411C
>	v_mul_f32_e32 v4, v33, v30                                                   ; 10083D21
>	v_mad_f32 v4, v29, -v32, -v4                                                 ; D2820004 C412411D
>	v_mad_f32 v15, v28, v31, v4                                                  ; D282000F 04123F1C
>	v_cubeid_f32 v38, v15, v16, v17                                              ; D2880026 0446210F
>	v_cubema_f32 v37, v15, v16, v17                                              ; D28E0025 0446210F
>	v_cubesc_f32 v36, v15, v16, v17                                              ; D28A0024 0446210F
>	v_cubetc_f32 v35, v15, v16, v17                                              ; D28C0023 0446210F
>	v_rcp_f32_e64 v4, |v37|                                                      ; D3540104 00000125
>	v_mov_b32_e32 v43, 0x3fc00000                                                ; 7E5602FF 3FC00000
>	v_mad_f32 v37, v35, v4, v43                                                  ; D2820025 04AE0923
>	v_mad_f32 v36, v36, v4, v43                                                  ; D2820024 04AE0924
>	image_sample v[6:8], 7, 0, 0, 0, 0, 0, 0, 0, v[36:39], s[32:39], s[28:31]    ; F0800700 00E80624
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_log_f32_e32 v9, v8                                                         ; 7E124F08
>	v_mov_b32_e32 v4, 0x400ccccd                                                 ; 7E0802FF 400CCCCD
>	v_mul_legacy_f32_e32 v9, v4, v9                                              ; 0E121304
>	v_exp_f32_e32 v9, v9                                                         ; 7E124B09
>	s_buffer_load_dword s16, s[0:3], 0x4                                         ; C2080104
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v35, s16, v33                                                  ; 10464210
>	s_buffer_load_dword s5, s[0:3], 0x5                                          ; C2028105
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v10, v32, s5, -v35                                                 ; D282000A 848C0B20
>	s_buffer_load_dword s56, s[0:3], 0x6                                         ; C21C0106
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v10, v31, s56, v10                                                 ; D282000A 0428711F
>	v_mul_f32_e32 v10, v30, v10                                                  ; 1014151E
>	v_mul_f32_e32 v11, s16, v32                                                  ; 10164010
>	v_mad_f32 v11, v31, s5, -v11                                                 ; D282000B 842C0B1F
>	v_mad_f32 v36, v33, s56, v11                                                 ; D2820024 042C7121
>	v_mad_f32 v10, v36, v29, v10                                                 ; D282000A 042A3B24
>	v_mul_f32_e32 v37, s16, v31                                                  ; 104A3E10
>	v_mad_f32 v11, v32, s5, v37                                                  ; D282000B 04940B20
>	v_mad_f32 v38, v33, s56, v11                                                 ; D2820026 042C7121
>	v_mad_f32 v10, v38, v28, v10                                                 ; D282000A 042A3926
>	v_add_f32_e64 v10, 0, v10 clamp                                              ; D206080A 00021480
>	v_add_f32_e32 v39, v9, v10                                                   ; 064E1509
>	v_mul_f32_e64 v11, |v32|, |v32|                                              ; D210030B 00024120
>	v_mul_f32_e64 v9, |v31|, |v31|                                               ; D2100309 00023F1F
>	v_mul_f32_e32 v9, v9, v9                                                     ; 10121309
>	v_mad_f32 v13, v11, v11, v9                                                  ; D282000D 0426170B
>	v_mul_f32_e64 v40, |v33|, |v33|                                              ; D2100328 00024321
>	v_mad_f32 v13, v40, v40, v13                                                 ; D282000D 04365128
>	v_rcp_f32_e32 v13, v13                                                       ; 7E1A550D
>	v_mul_f32_e32 v9, v9, v13                                                    ; 10121B09
>	v_mul_f32_e32 v11, v11, v11                                                  ; 1016170B
>	v_mul_f32_e32 v11, v11, v13                                                  ; 10161B0B
>	s_buffer_load_dword s16, s[0:3], 0x25                                        ; C2080125
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v42, s16, v5                                                   ; 10540A10
>	v_interp_p1_f32 v14, v0, 0, 1, [m0]                                          ; C8380400
>	v_interp_p2_f32 v14, [v14], v1, 0, 1, [m0]                                   ; C8390401
>	s_buffer_load_dword s16, s[0:3], 0x24                                        ; C2080124
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v41, s16, v14                                                  ; 10521C10
>	image_sample v[44:46], 7, 0, 0, 0, 0, 0, 0, 0, v[41:42], s[8:15], s[92:95]   ; F0800700 02E22C29
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_log_f32_e32 v47, v46                                                       ; 7E5E4F2E
>	v_mul_legacy_f32_e32 v47, v4, v47                                            ; 0E5E5F04
>	v_exp_f32_e32 v47, v47                                                       ; 7E5E4B2F
>	v_mul_f32_e32 v47, v47, v11                                                  ; 105E172F
>	v_mul_f32_e32 v41, s16, v12                                                  ; 10521810
>	image_sample v[48:50], 7, 0, 0, 0, 0, 0, 0, 0, v[41:42], s[8:15], s[92:95]   ; F0800700 02E23029
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_log_f32_e32 v41, v50                                                       ; 7E524F32
>	v_mul_legacy_f32_e32 v41, v4, v41                                            ; 0E525304
>	v_exp_f32_e32 v41, v41                                                       ; 7E524B29
>	v_mad_f32 v41, v41, v9, v47                                                  ; D2820029 04BE1329
>	v_mul_f32_e32 v40, v40, v40                                                  ; 10505128
>	v_mul_f32_e32 v13, v40, v13                                                  ; 101A1B28
>	s_buffer_load_dword s8, s[0:3], 0x10                                         ; C2040110
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v52, s8, v12                                                   ; 10681808
>	v_mul_f32_e32 v51, s8, v14                                                   ; 10661C08
>	image_sample v[51:54], 15, 0, 0, 0, 0, 0, 0, 0, v[51:52], s[84:91], s[80:83] ; F0800F00 02953333
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_log_f32_e32 v40, v53                                                       ; 7E504F35
>	v_mul_legacy_f32_e32 v40, v4, v40                                            ; 0E505104
>	v_exp_f32_e32 v40, v40                                                       ; 7E504B28
>	v_mad_f32 v40, v40, v13, v41                                                 ; D2820028 04A61B28
>	v_mul_f32_e32 v39, v39, v40                                                  ; 104E5127
>	s_buffer_load_dword s8, s[0:3], 0x8                                          ; C2040108
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v41, s8, v32                                                   ; 10524008
>	s_buffer_load_dword s57, s[0:3], 0x9                                         ; C21C8109
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v41, v31, s57, -v41                                                ; D2820029 84A4731F
>	s_buffer_load_dword s58, s[0:3], 0xa                                         ; C21D010A
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v41, v33, s58, v41                                                 ; D2820029 04A47521
>	v_interp_p1_f32 v42, v0, 0, 2, [m0]                                          ; C8A80800
>	v_interp_p2_f32 v42, [v42], v1, 0, 2, [m0]                                   ; C8A90801
>	v_interp_p1_f32 v47, v0, 1, 2, [m0]                                          ; C8BC0900
>	v_interp_p2_f32 v47, [v47], v1, 1, 2, [m0]                                   ; C8BD0901
>	v_mul_f32_e32 v55, v47, v47                                                  ; 106E5F2F
>	v_mad_f32 v55, v42, v42, v55                                                 ; D2820037 04DE552A
>	v_interp_p1_f32 v56, v0, 2, 2, [m0]                                          ; C8E00A00
>	v_interp_p2_f32 v56, [v56], v1, 2, 2, [m0]                                   ; C8E10A01
>	v_mad_f32 v55, v56, v56, v55                                                 ; D2820037 04DE7138
>	v_add_f32_e32 v55, 0, v55                                                    ; 066E6E80
>	v_rsq_clamp_f32_e32 v55, v55                                                 ; 7E6E5937
>	v_mul_f32_e32 v42, v55, v42                                                  ; 10545537
>	v_mul_f32_e32 v57, v42, v32                                                  ; 1072412A
>	v_mul_f32_e32 v47, v55, v47                                                  ; 105E5F37
>	v_mad_f32 v57, v31, v47, -v57                                                ; D2820039 84E65F1F
>	v_mul_f32_e32 v55, v55, v56                                                  ; 106E7137
>	v_mad_f32 v56, v33, v55, v57                                                 ; D2820038 04E66F21
>	v_add_f32_e32 v57, v41, v56                                                  ; 06727129
>	v_mul_f32_e32 v58, s8, v31                                                   ; 10743E08
>	v_mad_f32 v59, v32, s57, v58                                                 ; D282003B 04E87320
>	v_mad_f32 v59, v33, s58, v59                                                 ; D282003B 04EC7521
>	v_mul_f32_e32 v60, v42, v31                                                  ; 10783F2A
>	v_mad_f32 v61, v32, v47, v60                                                 ; D282003D 04F25F20
>	v_mad_f32 v61, v33, v55, v61                                                 ; D282003D 04F66F21
>	v_add_f32_e32 v62, v59, v61                                                  ; 067C7B3B
>	v_rcp_f32_e32 v62, v62                                                       ; 7E7C553E
>	v_mad_f32 v63, v57, v62, -v26                                                ; D282003F 846A7D39
>	v_mul_f32_e64 v64, s4, s4                                                    ; D2100040 00000804
>	v_mad_f32 v27, v24, v64, -v27                                                ; D282001B 846E8118
>	s_buffer_load_dword s9, s[0:3], 0x18                                         ; C2048118
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_rcp_f32_e32 v65, s9                                                        ; 7E825409
>	v_add_f32_e32 v27, v27, v65                                                  ; 0636831B
>	v_mul_f32_e32 v66, v27, v63                                                  ; 10847F1B
>	v_mul_f32_e32 v67, s8, v33                                                   ; 10864208
>	v_mad_f32 v68, v32, s57, -v67                                                ; D2820044 850C7320
>	v_mad_f32 v68, v31, s58, v68                                                 ; D2820044 0510751F
>	v_mul_f32_e32 v69, v42, v33                                                  ; 108A432A
>	v_mad_f32 v70, v32, v47, -v69                                                ; D2820046 85165F20
>	v_mad_f32 v70, v31, v55, v70                                                 ; D2820046 051A6F1F
>	v_add_f32_e32 v71, v68, v70                                                  ; 068E8D44
>	v_mad_f32 v71, v71, v62, -v25                                                ; D2820047 84667D47
>	v_mul_f32_e32 v26, v25, v26                                                  ; 10343519
>	image_sample v72, 8, 0, 0, 0, 0, 0, 0, 0, v[19:20], s[60:67], s[52:55]       ; F0800800 01AF4813
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v72, 2.0, v72, -1.0                                                ; D2820048 03CE90F4
>	v_mad_f32 v26, v72, v64, -v26                                                ; D282001A 846A8148
>	v_mul_f32_e64 v72, v71, -v26                                                 ; D2100048 40023547
>	v_mad_f32 v66, 2.0, v72, v66                                                 ; D2820042 050A90F4
>	v_mul_f32_e32 v63, v66, v63                                                  ; 107E7F42
>	v_mul_f32_e32 v66, v71, v71                                                  ; 10848F47
>	v_mad_f32 v63, v66, v27, v63                                                 ; D282003F 04FE3742
>	v_mul_f32_e32 v63, 0.5, v63                                                  ; 107E7EF0
>	v_mul_f32_e32 v25, v25, v25                                                  ; 10323319
>	v_mad_f32 v21, v23, v64, -v25                                                ; D2820015 84668117
>	v_add_f32_e32 v21, v21, v65                                                  ; 062A8315
>	v_mul_f32_e32 v22, v26, v26                                                  ; 102C351A
>	v_mad_f32 v21, v21, v27, -v22                                                ; D2820015 845A3715
>	v_rcp_f32_e32 v22, v21                                                       ; 7E2C5515
>	v_mul_f32_e32 v23, v22, v63                                                  ; 102E7F16
>	v_mov_b32_e32 v24, 0xbfb8aa65                                                ; 7E3002FF BFB8AA65
>	v_mul_f32_e32 v23, v24, v23                                                  ; 102E2F18
>	v_exp_f32_e32 v23, v23                                                       ; 7E2E4B17
>	v_max_f32_e32 v25, 0x33d6bf95, v21                                           ; 20322AFF 33D6BF95
>	v_rsq_clamp_f32_e32 v25, v25                                                 ; 7E325919
>	v_mul_f32_e32 v23, v23, v25                                                  ; 102E3317
>	v_mov_b32_e32 v25, 0xc1800000                                                ; 7E3202FF C1800000
>	v_mad_f32 v22, v63, v22, v25                                                 ; D2820016 04662D3F
>	v_cmp_ge_f32_e64 s[8:9], v22, 0                                              ; D00C0008 00010116
>	v_cndmask_b32_e64 v22, 0, -1, s[8:9]                                         ; D2000016 00218280
>	v_cmp_ne_i32_e64 s[8:9], v22, 0                                              ; D10A0008 00010116
>	v_cndmask_b32_e64 v22, 0, 1.0, s[8:9]                                        ; D2000016 0021E480
>	v_cmp_ge_f32_e64 s[8:9], -v21, 0                                             ; D00C0008 20010115
>	v_cndmask_b32_e64 v21, 0, -1, s[8:9]                                         ; D2000815 00218280
>	v_cmp_ne_i32_e64 s[8:9], v21, 0                                              ; D10A0008 00010115
>	v_cndmask_b32_e64 v21, 0, 1.0, s[8:9]                                        ; D2000815 0021E480
>	v_add_f32_e32 v21, v21, v22                                                  ; 062A2D15
>	v_cmp_ge_f32_e64 s[8:9], -v21, 0                                             ; D00C0008 20010115
>	v_cndmask_b32_e64 v21, 0, -1, s[8:9]                                         ; D2000815 00218280
>	v_cmp_ne_i32_e64 s[8:9], v21, 0                                              ; D10A0008 00010115
>	v_cndmask_b32_e64 v21, 0, v23, s[8:9]                                        ; D2000815 00222E80
>	v_mul_f32_e32 v22, v70, v30                                                  ; 102C3D46
>	v_mad_f32 v22, v28, v61, v22                                                 ; D2820016 045A7B1C
>	v_mad_f32 v22, v29, v56, v22                                                 ; D2820016 045A711D
>	v_sub_f32_e32 v22, 1.0, v22                                                  ; 082C2CF2
>	v_mul_f32_e32 v23, v22, v22                                                  ; 102E2D16
>	v_mul_f32_e32 v23, v23, v23                                                  ; 102E2F17
>	v_mul_f32_e32 v22, v23, v22                                                  ; 102C2D17
>	v_mul_f32_e32 v21, v22, v21                                                  ; 102A2B16
>	v_add_f32_e64 v21, 0, v21 clamp                                              ; D2060815 00022A80
>	v_mul_f32_e32 v22, v30, v68                                                  ; 102C891E
>	v_mad_f32 v22, v41, v29, v22                                                 ; D2820016 045A3B29
>	v_mad_f32 v22, v59, v28, v22                                                 ; D2820016 045A393B
>	v_add_f32_e64 v22, 0, v22 clamp                                              ; D2060816 00022C80
>	v_mad_f32 v23, v21, v22, v39                                                 ; D2820017 049E2D15
>	v_mul_f32_e32 v19, s59, v14                                                  ; 10261C3B
>	image_sample v[26:29], 15, 0, 0, 0, 0, 0, 0, 0, v[19:20], s[72:79], s[68:71] ; F0800F00 02321A13
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v30, 2.0, v26, -1.0                                                ; D282001E 03CE34F4
>	v_mul_f32_e32 v30, s4, v30                                                   ; 103C3C04
>	v_mad_f32 v39, 2.0, v27, -1.0                                                ; D2820027 03CE36F4
>	v_mul_f32_e32 v39, s4, v39                                                   ; 104E4E04
>	v_mul_f32_e32 v63, v39, v39                                                  ; 107E4F27
>	v_mad_f32 v66, v30, v30, v63                                                 ; D2820042 04FE3D1E
>	v_add_f32_e32 v66, 1.0, v66                                                  ; 068484F2
>	v_add_f32_e32 v66, 0, v66                                                    ; 06848480
>	v_rsq_clamp_f32_e32 v66, v66                                                 ; 7E845942
>	v_mul_f32_e32 v68, v66, v39                                                  ; 10884F42
>	v_mul_f32_e32 v70, v66, v30                                                  ; 108C3D42
>	v_mul_f32_e32 v71, v32, v70                                                  ; 108E8D20
>	v_mad_f32 v71, v68, v33, v71                                                 ; D2820047 051E4344
>	v_mad_f32 v17, v66, v33, v71                                                 ; D2820011 051E4342
>	v_mul_f32_e32 v71, v31, v68                                                  ; 108E891F
>	v_mad_f32 v71, v66, v32, v71                                                 ; D2820047 051E4142
>	v_mad_f32 v16, v70, v33, v71                                                 ; D2820010 051E4346
>	v_mul_f32_e32 v71, v31, v70                                                  ; 108E8D1F
>	v_mad_f32 v71, v68, -v32, -v71                                               ; D2820047 C51E4144
>	v_mad_f32 v15, v66, v31, v71                                                 ; D282000F 051E3F42
>	v_cubeid_f32 v74, v15, v16, v17                                              ; D288004A 0446210F
>	v_cubema_f32 v73, v15, v16, v17                                              ; D28E0049 0446210F
>	v_cubesc_f32 v72, v15, v16, v17                                              ; D28A0048 0446210F
>	v_cubetc_f32 v71, v15, v16, v17                                              ; D28C0047 0446210F
>	v_rcp_f32_e64 v15, |v73|                                                     ; D354010F 00000149
>	v_mad_f32 v73, v71, v15, v43                                                 ; D2820049 04AE1F47
>	v_mad_f32 v72, v72, v15, v43                                                 ; D2820048 04AE1F48
>	image_sample v[15:17], 7, 0, 0, 0, 0, 0, 0, 0, v[72:75], s[32:39], s[28:31]  ; F0800700 00E80F48
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_log_f32_e32 v18, v17                                                       ; 7E244F11
>	v_mul_legacy_f32_e32 v18, v4, v18                                            ; 0E242504
>	v_exp_f32_e32 v18, v18                                                       ; 7E244B12
>	v_mad_f32 v71, v33, s5, -v37                                                 ; D2820047 84940B21
>	v_mad_f32 v71, v32, s56, v71                                                 ; D2820047 051C7120
>	v_mul_f32_e32 v71, v70, v71                                                  ; 108E8F46
>	v_mad_f32 v36, v36, v68, v71                                                 ; D2820024 051E8924
>	v_mad_f32 v36, v38, v66, v36                                                 ; D2820024 04928526
>	v_add_f32_e64 v36, 0, v36 clamp                                              ; D2060824 00024880
>	v_add_f32_e32 v18, v18, v36                                                  ; 06244912
>	v_mul_f32_e32 v18, v18, v40                                                  ; 10245112
>	v_mad_f32 v57, v57, v62, -v39                                                ; D2820039 849E7D39
>	v_mad_f32 v63, v29, v64, -v63                                                ; D282003F 84FE811D
>	v_add_f32_e32 v63, v63, v65                                                  ; 067E833F
>	v_mul_f32_e32 v71, v63, v57                                                  ; 108E733F
>	v_mad_f32 v72, v33, s57, -v58                                                ; D2820048 84E87321
>	v_mad_f32 v72, v32, s58, v72                                                 ; D2820048 05207520
>	v_mad_f32 v73, v33, v47, -v60                                                ; D2820049 84F25F21
>	v_mad_f32 v73, v32, v55, v73                                                 ; D2820049 05266F20
>	v_add_f32_e32 v74, v72, v73                                                  ; 06949348
>	v_mad_f32 v74, v74, v62, -v30                                                ; D282004A 847A7D4A
>	v_mul_f32_e32 v39, v30, v39                                                  ; 104E4F1E
>	image_sample v19, 8, 0, 0, 0, 0, 0, 0, 0, v[19:20], s[60:67], s[52:55]       ; F0800800 01AF1313
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v19, 2.0, v19, -1.0                                                ; D2820013 03CE26F4
>	v_mad_f32 v19, v19, v64, -v39                                                ; D2820013 849E8113
>	v_mul_f32_e64 v20, v74, -v19                                                 ; D2100014 4002274A
>	v_mad_f32 v20, 2.0, v20, v71                                                 ; D2820014 051E28F4
>	v_mul_f32_e32 v20, v20, v57                                                  ; 10287314
>	v_mul_f32_e32 v39, v74, v74                                                  ; 104E954A
>	v_mad_f32 v20, v39, v63, v20                                                 ; D2820014 04527F27
>	v_mul_f32_e32 v20, 0.5, v20                                                  ; 102828F0
>	v_mul_f32_e32 v30, v30, v30                                                  ; 103C3D1E
>	v_mad_f32 v26, v28, v64, -v30                                                ; D282001A 847A811C
>	v_add_f32_e32 v26, v26, v65                                                  ; 0634831A
>	v_mul_f32_e32 v19, v19, v19                                                  ; 10262713
>	v_mad_f32 v19, v26, v63, -v19                                                ; D2820013 844E7F1A
>	v_rcp_f32_e32 v26, v19                                                       ; 7E345513
>	v_mul_f32_e32 v27, v26, v20                                                  ; 1036291A
>	v_mul_f32_e32 v27, v24, v27                                                  ; 10363718
>	v_exp_f32_e32 v27, v27                                                       ; 7E364B1B
>	v_max_f32_e32 v28, 0x33d6bf95, v19                                           ; 203826FF 33D6BF95
>	v_rsq_clamp_f32_e32 v28, v28                                                 ; 7E38591C
>	v_mul_f32_e32 v27, v27, v28                                                  ; 1036391B
>	v_mad_f32 v20, v20, v26, v25                                                 ; D2820014 04663514
>	v_cmp_ge_f32_e64 s[8:9], v20, 0                                              ; D00C0008 00010114
>	v_cndmask_b32_e64 v20, 0, -1, s[8:9]                                         ; D2000014 00218280
>	v_cmp_ne_i32_e64 s[8:9], v20, 0                                              ; D10A0008 00010114
>	v_cndmask_b32_e64 v20, 0, 1.0, s[8:9]                                        ; D2000014 0021E480
>	v_cmp_ge_f32_e64 s[8:9], -v19, 0                                             ; D00C0008 20010113
>	v_cndmask_b32_e64 v19, 0, -1, s[8:9]                                         ; D2000813 00218280
>	v_cmp_ne_i32_e64 s[8:9], v19, 0                                              ; D10A0008 00010113
>	v_cndmask_b32_e64 v19, 0, 1.0, s[8:9]                                        ; D2000813 0021E480
>	v_add_f32_e32 v19, v19, v20                                                  ; 06262913
>	v_cmp_ge_f32_e64 s[8:9], -v19, 0                                             ; D00C0008 20010113
>	v_cndmask_b32_e64 v19, 0, -1, s[8:9]                                         ; D2000813 00218280
>	v_cmp_ne_i32_e64 s[8:9], v19, 0                                              ; D10A0008 00010113
>	v_cndmask_b32_e64 v19, 0, v27, s[8:9]                                        ; D2000813 00223680
>	v_mul_f32_e32 v20, v73, v70                                                  ; 10288D49
>	v_mad_f32 v20, v66, v61, v20                                                 ; D2820014 04527B42
>	v_mad_f32 v20, v68, v56, v20                                                 ; D2820014 04527144
>	v_sub_f32_e32 v20, 1.0, v20                                                  ; 082828F2
>	v_mul_f32_e32 v26, v20, v20                                                  ; 10342914
>	v_mul_f32_e32 v26, v26, v26                                                  ; 1034351A
>	v_mul_f32_e32 v20, v26, v20                                                  ; 1028291A
>	v_mul_f32_e32 v19, v20, v19                                                  ; 10262714
>	v_add_f32_e64 v19, 0, v19 clamp                                              ; D2060813 00022680
>	v_mul_f32_e32 v20, v70, v72                                                  ; 10289146
>	v_mad_f32 v20, v41, v68, v20                                                 ; D2820014 04528929
>	v_mad_f32 v20, v59, v66, v20                                                 ; D2820014 0452853B
>	v_add_f32_e64 v20, 0, v20 clamp                                              ; D2060814 00022880
>	v_mad_f32 v18, v19, v20, v18                                                 ; D2820012 044A2913
>	v_mul_f32_e32 v18, v18, v11                                                  ; 10241712
>	v_mad_f32 v18, v23, v9, v18                                                  ; D2820012 044A1317
>	s_buffer_load_dword s8, s[0:3], 0x14                                         ; C2040114
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v27, s8, v12                                                   ; 10361808
>	v_mul_f32_e32 v26, s8, v14                                                   ; 10341C08
>	image_sample v[26:29], 15, 0, 0, 0, 0, 0, 0, 0, v[26:27], s[44:51], s[40:43] ; F0800F00 014B1A1A
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v23, 2.0, v26, -1.0                                                ; D2820017 03CE34F4
>	v_mul_f32_e32 v23, s4, v23                                                   ; 102E2E04
>	v_mad_f32 v30, 2.0, v27, -1.0                                                ; D282001E 03CE36F4
>	v_mul_f32_e32 v30, s4, v30                                                   ; 103C3C04
>	v_mul_f32_e32 v39, v30, v30                                                  ; 104E3D1E
>	v_mad_f32 v41, v23, v23, v39                                                 ; D2820029 049E2F17
>	v_add_f32_e32 v41, 1.0, v41                                                  ; 065252F2
>	v_add_f32_e32 v41, 0, v41                                                    ; 06525280
>	v_rsq_clamp_f32_e32 v41, v41                                                 ; 7E525929
>	v_mul_f32_e32 v56, v41, v30                                                  ; 10703D29
>	v_mul_f32_e32 v57, v41, v23                                                  ; 10722F29
>	v_mul_f32_e32 v63, v31, v57                                                  ; 107E731F
>	v_mad_f32 v63, v56, -v32, -v63                                               ; D282003F C4FE4138
>	v_mad_f32 v72, v41, v33, v63                                                 ; D2820048 04FE4329
>	v_mul_f32_e32 v63, v32, v57                                                  ; 107E7320
>	v_mad_f32 v63, v41, v32, v63                                                 ; D282003F 04FE4129
>	v_mad_f32 v71, v56, v33, v63                                                 ; D2820047 04FE4338
>	v_mul_f32_e32 v63, v31, v56                                                  ; 107E711F
>	v_mad_f32 v63, v41, v31, v63                                                 ; D282003F 04FE3F29
>	v_mad_f32 v70, v57, v33, v63                                                 ; D2820046 04FE4339
>	v_mad_f32 v63, v32, v47, v69                                                 ; D282003F 05165F20
>	v_mad_f32 v63, -v31, v55, v63                                                ; D282003F 24FE6F1F
>	v_mul_f32_e32 v66, v63, v57                                                  ; 1084733F
>	v_mad_f32 v60, v33, v47, v60                                                 ; D282003C 04F25F21
>	v_mad_f32 v60, -v32, v55, v60                                                ; D282003C 24F26F20
>	v_mad_f32 v66, v56, v60, v66                                                 ; D2820042 050A7938
>	v_mad_f32 v61, v41, v61, v66                                                 ; D282003D 050A7B29
>	v_sub_f32_e32 v61, 1.0, v61                                                  ; 087A7AF2
>	v_mul_f32_e32 v66, v61, v61                                                  ; 10847B3D
>	v_mul_f32_e32 v73, v66, v66                                                  ; 10928542
>	v_cubeid_f32 v77, v70, v71, v72                                              ; D288004D 05228F46
>	v_cubema_f32 v76, v70, v71, v72                                              ; D28E004C 05228F46
>	v_cubesc_f32 v75, v70, v71, v72                                              ; D28A004B 05228F46
>	v_cubetc_f32 v74, v70, v71, v72                                              ; D28C004A 05228F46
>	v_rcp_f32_e64 v66, |v76|                                                     ; D3540142 0000014C
>	v_mad_f32 v76, v74, v66, v43                                                 ; D282004C 04AE854A
>	v_mad_f32 v75, v75, v66, v43                                                 ; D282004B 04AE854B
>	image_sample v[74:76], 7, 0, 0, 0, 0, 0, 0, 0, v[75:78], s[32:39], s[28:31]  ; F0800700 00E84A4B
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_log_f32_e32 v66, v76                                                       ; 7E844F4C
>	v_mul_legacy_f32_e32 v66, v4, v66                                            ; 0E848504
>	v_exp_f32_e32 v66, v66                                                       ; 7E844B42
>	v_mad_f32 v35, v32, s5, v35                                                  ; D2820023 048C0B20
>	v_mad_f32 v35, -v31, s56, v35                                                ; D2820023 248C711F
>	v_mul_f32_e32 v35, v57, v35                                                  ; 10464739
>	v_mad_f32 v37, v33, s5, v37                                                  ; D2820025 04940B21
>	v_mad_f32 v37, -v32, s56, v37                                                ; D2820025 24947120
>	v_mad_f32 v35, v37, v56, v35                                                 ; D2820023 048E7125
>	v_mad_f32 v35, v38, v41, v35                                                 ; D2820023 048E5326
>	v_add_f32_e64 v35, 0, v35 clamp                                              ; D2060823 00024680
>	v_add_f32_e32 v37, v66, v35                                                  ; 064A4742
>	v_mul_f32_e32 v37, v37, v40                                                  ; 104A5125
>	v_mad_f32 v58, v33, s57, v58                                                 ; D282003A 04E87321
>	v_mad_f32 v58, -v32, s58, v58                                                ; D282003A 24E87520
>	v_add_f32_e32 v60, v58, v60                                                  ; 0678793A
>	v_mad_f32 v60, v60, v62, -v30                                                ; D282003C 847A7D3C
>	v_add_f32_e32 v66, 0, v29                                                    ; 06843A80
>	v_mad_f32 v39, v66, v64, -v39                                                ; D2820027 849E8142
>	v_add_f32_e32 v39, v39, v65                                                  ; 064E8327
>	v_mul_f32_e32 v66, v39, v60                                                  ; 10847927
>	v_mad_f32 v67, v32, s57, v67                                                 ; D2820043 050C7320
>	v_mad_f32 v67, -v31, s58, v67                                                ; D2820043 250C751F
>	v_add_f32_e32 v63, v67, v63                                                  ; 067E7F43
>	v_mad_f32 v62, v63, v62, -v23                                                ; D282003E 845E7D3F
>	v_mul_f32_e32 v30, v23, v30                                                  ; 103C3D17
>	v_mad_f32 v63, 2.0, v29, -1.0                                                ; D282003F 03CE3AF4
>	v_mad_f32 v30, v63, v64, -v30                                                ; D282001E 847A813F
>	v_mul_f32_e64 v63, v62, -v30                                                 ; D210003F 40023D3E
>	v_mad_f32 v63, 2.0, v63, v66                                                 ; D282003F 050A7EF4
>	v_mul_f32_e32 v60, v63, v60                                                  ; 1078793F
>	v_mul_f32_e32 v62, v62, v62                                                  ; 107C7D3E
>	v_mad_f32 v60, v62, v39, v60                                                 ; D282003C 04F24F3E
>	v_mul_f32_e32 v60, 0.5, v60                                                  ; 107878F0
>	v_add_f32_e32 v26, 0, v28                                                    ; 06343880
>	v_mul_f32_e32 v23, v23, v23                                                  ; 102E2F17
>	v_mad_f32 v23, v26, v64, -v23                                                ; D2820017 845E811A
>	v_add_f32_e32 v23, v23, v65                                                  ; 062E8317
>	v_mul_f32_e32 v26, v30, v30                                                  ; 10343D1E
>	v_mad_f32 v23, v23, v39, -v26                                                ; D2820017 846A4F17
>	v_rcp_f32_e32 v26, v23                                                       ; 7E345517
>	v_mul_f32_e32 v27, v26, v60                                                  ; 1036791A
>	v_mul_f32_e32 v24, v24, v27                                                  ; 10303718
>	v_exp_f32_e32 v24, v24                                                       ; 7E304B18
>	v_max_f32_e32 v27, 0x33d6bf95, v23                                           ; 20362EFF 33D6BF95
>	v_rsq_clamp_f32_e32 v27, v27                                                 ; 7E36591B
>	v_mul_f32_e32 v24, v24, v27                                                  ; 10303718
>	v_mad_f32 v25, v60, v26, v25                                                 ; D2820019 0466353C
>	v_cmp_ge_f32_e64 s[4:5], v25, 0                                              ; D00C0004 00010119
>	v_cndmask_b32_e64 v25, 0, -1, s[4:5]                                         ; D2000819 00118280
>	v_cmp_ne_i32_e64 s[4:5], v25, 0                                              ; D10A0004 00010119
>	v_cndmask_b32_e64 v25, 0, 1.0, s[4:5]                                        ; D2000819 0011E480
>	v_cmp_ge_f32_e64 s[4:5], -v23, 0                                             ; D00C0004 20010117
>	v_cndmask_b32_e64 v23, 0, -1, s[4:5]                                         ; D2000817 00118280
>	v_cmp_ne_i32_e64 s[4:5], v23, 0                                              ; D10A0004 00010117
>	v_cndmask_b32_e64 v23, 0, 1.0, s[4:5]                                        ; D2000817 0011E480
>	v_add_f32_e32 v23, v23, v25                                                  ; 062E3317
>	v_cmp_ge_f32_e64 s[4:5], -v23, 0                                             ; D00C0004 20010117
>	v_cndmask_b32_e64 v23, 0, -1, s[4:5]                                         ; D2000817 00118280
>	v_cmp_ne_i32_e64 s[4:5], v23, 0                                              ; D10A0004 00010117
>	v_cndmask_b32_e64 v23, 0, v24, s[4:5]                                        ; D2000817 00123080
>	v_mul_f32_e32 v24, v73, v61                                                  ; 10307B49
>	v_mul_f32_e32 v23, v24, v23                                                  ; 102E2F18
>	v_add_f32_e64 v23, 0, v23 clamp                                              ; D2060817 00022E80
>	v_mul_f32_e32 v24, v57, v67                                                  ; 10308739
>	v_mad_f32 v24, v58, v56, v24                                                 ; D2820018 0462713A
>	v_mad_f32 v24, v59, v41, v24                                                 ; D2820018 0462533B
>	v_add_f32_e64 v24, 0, v24 clamp                                              ; D2060818 00023080
>	v_mad_f32 v25, v23, v24, v37                                                 ; D2820019 04963117
>	v_mad_f32 v18, v25, v13, v18                                                 ; D2820012 044A1B19
>	v_mov_b32_e32 v34, v54                                                       ; 7E440336
>	v_cubeid_f32 v59, v31, v32, v33                                              ; D288003B 0486411F
>	v_cubema_f32 v58, v31, v32, v33                                              ; D28E003A 0486411F
>	v_cubesc_f32 v57, v31, v32, v33                                              ; D28A0039 0486411F
>	v_cubetc_f32 v56, v31, v32, v33                                              ; D28C0038 0486411F
>	v_rcp_f32_e64 v25, |v58|                                                     ; D3540119 0000013A
>	v_mad_f32 v58, v56, v25, v43                                                 ; D282003A 04AE3338
>	v_mad_f32 v57, v57, v25, v43                                                 ; D2820039 04AE3339
>	image_sample v[25:27], 7, 0, 0, 0, 0, 0, 0, 0, v[57:60], s[32:39], s[28:31]  ; F0800700 00E81939
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_log_f32_e32 v28, v27                                                       ; 7E384F1B
>	v_mul_legacy_f32_e32 v28, v4, v28                                            ; 0E383904
>	v_exp_f32_e32 v28, v28                                                       ; 7E384B1C
>	v_add_f32_e64 v29, 0, v38 clamp                                              ; D206081D 00024C80
>	v_add_f32_e32 v28, v29, v28                                                  ; 0638391D
>	v_log_f32_e32 v30, v45                                                       ; 7E3C4F2D
>	v_mul_legacy_f32_e32 v30, v4, v30                                            ; 0E3C3D04
>	v_exp_f32_e32 v30, v30                                                       ; 7E3C4B1E
>	v_mul_f32_e32 v30, v30, v11                                                  ; 103C171E
>	v_log_f32_e32 v31, v49                                                       ; 7E3E4F31
>	v_mul_legacy_f32_e32 v31, v4, v31                                            ; 0E3E3F04
>	v_exp_f32_e32 v31, v31                                                       ; 7E3E4B1F
>	v_mad_f32 v30, v31, v9, v30                                                  ; D282001E 047A131F
>	v_log_f32_e32 v31, v52                                                       ; 7E3E4F34
>	v_mul_legacy_f32_e32 v31, v4, v31                                            ; 0E3E3F04
>	v_exp_f32_e32 v31, v31                                                       ; 7E3E4B1F
>	v_mad_f32 v30, v31, v13, v30                                                 ; D282001E 047A1B1F
>	v_log_f32_e32 v31, v44                                                       ; 7E3E4F2C
>	v_mul_legacy_f32_e32 v31, v4, v31                                            ; 0E3E3F04
>	v_exp_f32_e32 v31, v31                                                       ; 7E3E4B1F
>	v_mul_f32_e32 v31, v31, v11                                                  ; 103E171F
>	v_log_f32_e32 v32, v48                                                       ; 7E404F30
>	v_mul_legacy_f32_e32 v32, v4, v32                                            ; 0E404104
>	v_exp_f32_e32 v32, v32                                                       ; 7E404B20
>	v_mad_f32 v31, v32, v9, v31                                                  ; D282001F 047E1320
>	v_log_f32_e32 v32, v51                                                       ; 7E404F33
>	v_mul_legacy_f32_e32 v32, v4, v32                                            ; 0E404104
>	v_exp_f32_e32 v32, v32                                                       ; 7E404B20
>	v_mad_f32 v31, v32, v13, v31                                                 ; D282001F 047E1B20
>	v_mul_f32_e32 v32, 0x3e99999a, v31                                           ; 10403EFF 3E99999A
>	v_mov_b32_e32 v33, 0x3f170a3d                                                ; 7E4202FF 3F170A3D
>	v_mad_f32 v32, v33, v30, v32                                                 ; D2820020 04823D21
>	v_mov_b32_e32 v33, 0x3de147ae                                                ; 7E4202FF 3DE147AE
>	v_mad_f32 v32, v33, v40, v32                                                 ; D2820020 04825121
>	v_sub_f32_e32 v32, 1.0, v32                                                  ; 084040F2
>	s_buffer_load_dword s4, s[0:3], 0x1e                                         ; C202011E
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v33, s4, v32                                                   ; 10424004
>	v_mad_f32 v18, v33, v28, v18                                                 ; D2820012 044A3921
>	s_buffer_load_dword s4, s[0:3], 0x20                                         ; C2020120
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v28, s4, v5                                                    ; 08380A04
>	s_buffer_load_dword s5, s[0:3], 0x34                                         ; C2028134
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v33, s5, v28                                                   ; 10423805
>	v_add_f32_e64 v33, 0, v33 clamp                                              ; D2060821 00024280
>	s_buffer_load_dword s5, s[0:3], 0x2e                                         ; C202812E
>	s_buffer_load_dword s8, s[0:3], 0x32                                         ; C2040132
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mov_b32_e32 v34, s5                                                        ; 7E440205
>	v_sub_f32_e32 v34, s8, v34                                                   ; 08444408
>	v_mad_f32 v34, v33, v34, s5                                                  ; D2820022 00164521
>	v_mul_f32_e32 v34, v34, v18                                                  ; 10442522
>	v_rcp_f32_e64 v37, -v55                                                      ; D3540025 20000137
>	v_mul_f32_e32 v28, v37, v28                                                  ; 10383925
>	v_mul_f32_e32 v37, v47, v28                                                  ; 104A392F
>	v_mul_f32_e32 v38, v42, v28                                                  ; 104C392A
>	v_mul_f32_e32 v38, v38, v38                                                  ; 104C4D26
>	v_mad_f32 v37, v37, v37, v38                                                 ; D2820025 049A4B25
>	v_mul_f32_e32 v28, v55, v28                                                  ; 10383937
>	v_mad_f32 v28, v28, v28, v37                                                 ; D282001C 0496391C
>	v_max_f32_e32 v28, 0x33d6bf95, v28                                           ; 203838FF 33D6BF95
>	v_sqrt_f32_e32 v28, v28                                                      ; 7E38671C
>	s_buffer_load_dword s5, s[0:3], 0x38                                         ; C2028138
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e64 v28, -v28, s5                                                  ; D210001C 20000B1C
>	v_mul_f32_e32 v28, 0x3fb8aa65, v28                                           ; 103838FF 3FB8AA65
>	v_exp_f32_e32 v28, v28                                                       ; 7E384B1C
>	v_mul_f32_e32 v37, v55, v55                                                  ; 104A6F37
>	v_mul_f32_e32 v37, v37, v37                                                  ; 104A4B25
>	v_mad_f32 v37, -v55, v37, 1.0                                                ; D2820025 23CA4B37
>	v_sub_f32_e32 v37, 1.0, v37                                                  ; 084A4AF2
>	v_mul_f32_e32 v28, v37, v28                                                  ; 10383925
>	v_mul_f32_e32 v34, v34, v28                                                  ; 10443922
>	s_buffer_load_dword s5, s[0:3], 0x1                                          ; C2028101
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_subrev_f32_e32 v12, s5, v12                                                ; 0A181805
>	s_buffer_load_dword s5, s[0:3], 0x0                                          ; C2028100
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_subrev_f32_e32 v14, s5, v14                                                ; 0A1C1C05
>	v_mul_f32_e32 v14, v14, v14                                                  ; 101C1D0E
>	v_mad_f32 v12, v12, v12, v14                                                 ; D282000C 043A190C
>	s_buffer_load_dword s5, s[0:3], 0x2                                          ; C2028102
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_subrev_f32_e32 v14, s5, v5                                                 ; 0A1C0A05
>	v_mad_f32 v12, v14, v14, v12                                                 ; D282000C 04321D0E
>	s_buffer_load_dword s5, s[0:3], 0x41                                         ; C2028141
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v12, s5, v12                                                   ; 10181805
>	s_buffer_load_dword s5, s[0:3], 0x40                                         ; C2028140
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v37, s5, v14                                                   ; 104A1C05
>	v_mul_f32_e32 v37, 0x3fb8aa65, v37                                           ; 104A4AFF 3FB8AA65
>	v_exp_f32_e32 v37, v37                                                       ; 7E4A4B25
>	v_sub_f32_e32 v37, 1.0, v37                                                  ; 084A4AF2
>	v_mul_f32_e32 v12, v12, v37                                                  ; 10184B0C
>	v_rcp_f32_e32 v14, v14                                                       ; 7E1C550E
>	v_mul_f32_e32 v12, v12, v14                                                  ; 10181D0C
>	v_mul_f32_e32 v12, 0x3fb8aa65, v12                                           ; 101818FF 3FB8AA65
>	v_exp_f32_e32 v12, v12                                                       ; 7E184B0C
>	v_add_f32_e64 v12, 0, v12 clamp                                              ; D206080C 00021880
>	v_sub_f32_e32 v12, 1.0, v12                                                  ; 081818F2
>	v_interp_p1_f32 v14, v0, 1, 3, [m0]                                          ; C8380D00
>	v_interp_p2_f32 v14, [v14], v1, 1, 3, [m0]                                   ; C8390D01
>	v_mad_f32 v37, 0.5, v14, 0.5                                                 ; D2820025 03C21CF0
>	s_buffer_load_dword s5, s[0:3], 0x43                                         ; C2028143
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mov_b32_e32 v38, s5                                                        ; 7E4C0205
>	v_readlane_b32 s8, v82, 4                                                    ; 02110952
>	v_readlane_b32 s9, v82, 5                                                    ; 02130B52
>	v_readlane_b32 s10, v82, 6                                                   ; 02150D52
>	v_readlane_b32 s11, v82, 7                                                   ; 02170F52
>	v_readlane_b32 s12, v82, 8                                                   ; 02191152
>	v_readlane_b32 s13, v82, 9                                                   ; 021B1352
>	v_readlane_b32 s14, v82, 10                                                  ; 021D1552
>	v_readlane_b32 s15, v82, 11                                                  ; 021F1752
>	s_nop 2                                                                      ; BF800002
>	image_sample v14, 1, 0, 0, 0, 0, 0, 0, 0, v[37:38], s[8:15], s[24:27]        ; F0800100 00C20E25
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v37, -v12, v14, 1.0                                                ; D2820025 23CA1D0C
>	v_mul_f32_e32 v34, v34, v37                                                  ; 10444B22
>	v_subrev_f32_e32 v5, s4, v5                                                  ; 0A0A0A04
>	v_cmp_ge_f32_e64 s[4:5], v5, 0                                               ; D00C0004 00010105
>	v_cndmask_b32_e64 v5, 0, -1, s[4:5]                                          ; D2000805 00118280
>	v_cmp_ne_i32_e64 s[4:5], v5, 0                                               ; D10A0004 00010105
>	v_cndmask_b32_e64 v5, v34, v18, s[4:5]                                       ; D2000805 10122522
>	v_log_f32_e64 v5, |v5|                                                       ; D34E0105 00000105
>	v_mul_f32_e32 v5, 0x3ee8ba1f, v5                                             ; 100A0AFF 3EE8BA1F
>	v_exp_f32_e32 v5, v5                                                         ; 7E0A4B05
>	v_sub_f32_e32 v18, 1.0, v2                                                   ; 082404F2
>	v_interp_p1_f32 v39, v0, 1, 5, [m0]                                          ; C89C1500
>	v_interp_p2_f32 v39, [v39], v1, 1, 5, [m0]                                   ; C89D1501
>	v_interp_p1_f32 v38, v0, 0, 5, [m0]                                          ; C8981400
>	v_interp_p2_f32 v38, [v38], v1, 0, 5, [m0]                                   ; C8991401
>	s_load_dwordx8 s[8:15], s[6:7], 0x40                                         ; C0C40740
>	v_readlane_b32 s16, v82, 0                                                   ; 02210152
>	v_readlane_b32 s17, v82, 1                                                   ; 02230352
>	v_readlane_b32 s18, v82, 2                                                   ; 02250552
>	v_readlane_b32 s19, v82, 3                                                   ; 02270752
>	s_nop 2                                                                      ; BF800002
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample v[38:40], 7, 0, 0, 0, 0, 0, 0, 0, v[38:39], s[8:15], s[16:19]   ; F0800700 00822626
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mul_f32_e32 v0, v40, v18                                                   ; 10002528
>	v_mad_f32 v43, v2, v5, v0                                                    ; D282002B 04020B02
>	v_log_f32_e32 v0, v7                                                         ; 7E004F07
>	v_mul_legacy_f32_e32 v0, v4, v0                                              ; 0E000104
>	v_exp_f32_e32 v0, v0                                                         ; 7E004B00
>	v_add_f32_e32 v0, v0, v10                                                    ; 06001500
>	v_mul_f32_e32 v0, v0, v30                                                    ; 10003D00
>	v_mad_f32 v0, v21, v22, v0                                                   ; D2820000 04022D15
>	v_log_f32_e32 v1, v16                                                        ; 7E024F10
>	v_mul_legacy_f32_e32 v1, v4, v1                                              ; 0E020304
>	v_exp_f32_e32 v1, v1                                                         ; 7E024B01
>	v_add_f32_e32 v1, v1, v36                                                    ; 06024901
>	v_mul_f32_e32 v1, v1, v30                                                    ; 10023D01
>	v_mad_f32 v1, v19, v20, v1                                                   ; D2820001 04062913
>	v_mul_f32_e32 v1, v1, v11                                                    ; 10021701
>	v_mad_f32 v0, v0, v9, v1                                                     ; D2820000 04061300
>	v_log_f32_e32 v1, v75                                                        ; 7E024F4B
>	v_mul_legacy_f32_e32 v1, v4, v1                                              ; 0E020304
>	v_exp_f32_e32 v1, v1                                                         ; 7E024B01
>	v_add_f32_e32 v1, v1, v35                                                    ; 06024701
>	v_mul_f32_e32 v1, v1, v30                                                    ; 10023D01
>	v_mad_f32 v1, v23, v24, v1                                                   ; D2820001 04063117
>	v_mad_f32 v0, v1, v13, v0                                                    ; D2820000 04021B01
>	v_log_f32_e32 v1, v26                                                        ; 7E024F1A
>	v_mul_legacy_f32_e32 v1, v4, v1                                              ; 0E020304
>	v_exp_f32_e32 v1, v1                                                         ; 7E024B01
>	v_add_f32_e32 v1, v29, v1                                                    ; 0602031D
>	s_buffer_load_dword s8, s[0:3], 0x1d                                         ; C204011D
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v5, s8, v32                                                    ; 100A4008
>	v_mad_f32 v0, v5, v1, v0                                                     ; D2820000 04020305
>	s_buffer_load_dword s8, s[0:3], 0x2d                                         ; C204012D
>	s_buffer_load_dword s9, s[0:3], 0x31                                         ; C2048131
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mov_b32_e32 v1, s8                                                         ; 7E020208
>	v_sub_f32_e32 v1, s9, v1                                                     ; 08020209
>	v_mad_f32 v1, v33, v1, s8                                                    ; D2820001 00220321
>	v_mul_f32_e32 v1, v1, v0                                                     ; 10020101
>	v_mul_f32_e32 v1, v1, v28                                                    ; 10023901
>	v_mul_f32_e32 v1, v1, v37                                                    ; 10024B01
>	v_cndmask_b32_e64 v0, v1, v0, s[4:5]                                         ; D2000000 08120101
>	v_log_f32_e64 v0, |v0|                                                       ; D34E0100 00000100
>	v_mul_f32_e32 v0, 0x3ee8ba1f, v0                                             ; 100000FF 3EE8BA1F
>	v_exp_f32_e32 v0, v0                                                         ; 7E004B00
>	v_mul_f32_e32 v1, v39, v18                                                   ; 10022527
>	v_mad_f32 v42, v2, v0, v1                                                    ; D282002A 04060102
>	v_log_f32_e32 v0, v6                                                         ; 7E004F06
>	v_mul_legacy_f32_e32 v0, v4, v0                                              ; 0E000104
>	v_exp_f32_e32 v0, v0                                                         ; 7E004B00
>	v_add_f32_e32 v0, v0, v10                                                    ; 06001500
>	v_mul_f32_e32 v0, v0, v31                                                    ; 10003F00
>	v_mad_f32 v0, v21, v22, v0                                                   ; D2820000 04022D15
>	v_log_f32_e32 v1, v15                                                        ; 7E024F0F
>	v_mul_legacy_f32_e32 v1, v4, v1                                              ; 0E020304
>	v_exp_f32_e32 v1, v1                                                         ; 7E024B01
>	v_add_f32_e32 v1, v1, v36                                                    ; 06024901
>	v_mul_f32_e32 v1, v1, v31                                                    ; 10023F01
>	v_mad_f32 v1, v19, v20, v1                                                   ; D2820001 04062913
>	v_mul_f32_e32 v1, v1, v11                                                    ; 10021701
>	v_mad_f32 v0, v0, v9, v1                                                     ; D2820000 04061300
>	v_log_f32_e32 v1, v74                                                        ; 7E024F4A
>	v_mul_legacy_f32_e32 v1, v4, v1                                              ; 0E020304
>	v_exp_f32_e32 v1, v1                                                         ; 7E024B01
>	v_add_f32_e32 v1, v1, v35                                                    ; 06024701
>	v_mul_f32_e32 v1, v1, v31                                                    ; 10023F01
>	v_mad_f32 v1, v23, v24, v1                                                   ; D2820001 04063117
>	v_mad_f32 v0, v1, v13, v0                                                    ; D2820000 04021B01
>	v_log_f32_e32 v1, v25                                                        ; 7E024F19
>	v_mul_legacy_f32_e32 v1, v4, v1                                              ; 0E020304
>	v_exp_f32_e32 v1, v1                                                         ; 7E024B01
>	v_add_f32_e32 v1, v29, v1                                                    ; 0602031D
>	s_buffer_load_dword s8, s[0:3], 0x1c                                         ; C204011C
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v4, s8, v32                                                    ; 10084008
>	v_mad_f32 v0, v4, v1, v0                                                     ; D2820000 04020304
>	s_buffer_load_dword s8, s[0:3], 0x2c                                         ; C204012C
>	s_buffer_load_dword s9, s[0:3], 0x30                                         ; C2048130
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mov_b32_e32 v1, s8                                                         ; 7E020208
>	v_sub_f32_e32 v1, s9, v1                                                     ; 08020209
>	v_mad_f32 v1, v33, v1, s8                                                    ; D2820001 00220321
>	v_mul_f32_e32 v1, v1, v0                                                     ; 10020101
>	v_mul_f32_e32 v1, v1, v28                                                    ; 10023901
>	v_mul_f32_e32 v1, v1, v37                                                    ; 10024B01
>	v_cndmask_b32_e64 v0, v1, v0, s[4:5]                                         ; D2000000 08120101
>	v_log_f32_e64 v0, |v0|                                                       ; D34E0100 00000100
>	v_mul_f32_e32 v0, 0x3ee8ba1f, v0                                             ; 100000FF 3EE8BA1F
>	v_exp_f32_e32 v0, v0                                                         ; 7E004B00
>	v_mul_f32_e32 v1, v38, v18                                                   ; 10022526
>	v_mad_f32 v41, v2, v0, v1                                                    ; D2820029 04060102
>	s_load_dwordx8 s[8:15], s[6:7], 0x48                                         ; C0C40748
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample v[4:6], 7, 0, 0, 0, 0, 0, 0, 0, v[41:44], s[8:15], s[96:99]     ; F0800700 03020429
>	v_mad_f32 v0, -v3, v2, v2                                                    ; D2820000 240A0503
>	v_sub_f32_e32 v1, 1.0, v0                                                    ; 080200F2
>	v_mul_f32_e32 v2, v42, v1                                                    ; 1004032A
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v2, v0, v5, v2                                                     ; D2820002 040A0B00
>	s_buffer_load_dword s6, s[0:3], 0x3d                                         ; C203013D
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v3, s6, v2                                                     ; 08060406
>	v_mul_f32_e32 v7, v14, v12                                                   ; 100E190E
>	v_cndmask_b32_e64 v7, 0, v7, s[4:5]                                          ; D2000807 00120E80
>	v_mad_f32 v2, v7, v3, v2                                                     ; D2820002 040A0707
>	v_mul_f32_e32 v3, v41, v1                                                    ; 10060329
>	v_mad_f32 v3, v0, v4, v3                                                     ; D2820003 040E0900
>	s_buffer_load_dword s6, s[0:3], 0x3c                                         ; C203013C
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v8, s6, v3                                                     ; 08100606
>	v_mad_f32 v3, v7, v8, v3                                                     ; D2820003 040E1107
>	v_cvt_pkrtz_f16_f32_e32 v2, v3, v2                                           ; 5E040503
>	v_mul_f32_e32 v1, v43, v1                                                    ; 1002032B
>	v_mad_f32 v0, v0, v6, v1                                                     ; D2820000 04060D00
>	s_buffer_load_dword s0, s[0:3], 0x3e                                         ; C200013E
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v1, s0, v0                                                     ; 08020000
>	v_mad_f32 v0, v7, v1, v0                                                     ; D2820000 04020307
>	v_cndmask_b32_e64 v1, 0, 1.0, s[4:5]                                         ; D2000801 0011E480
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v1                                           ; 5E000300
>	exp 15, 0, 1, 1, 1, v2, v0, v2, v0                                           ; F8001C0F 00020002
>	s_endpgm                                                                     ; BF810000
>
>SGPRS: 104
>VGPRS: 84
>Code Size: 4340 bytes
>LDS: 0 blocks
>Scratch: 4096 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL SAMP[0]
>DCL CONST[0..15]
>DCL TEMP[0..3], LOCAL
>IMM[0] FLT32 {    0.5000,     0.0000,   255.0000,    -0.5000}
>IMM[1] FLT32 {    0.5000,    -0.5000,     1.0000,     0.0000}
>  0: ADD TEMP[0].xy, CONST[15], IN[0]
>  1: MOV TEMP[1].xy, TEMP[0].xyxx
>  2: ADD TEMP[0].zw, TEMP[0].xyxy, CONST[15]
>  3: MOV TEMP[1].zw, TEMP[0].wwzw
>  4: MUL TEMP[0].xy, TEMP[1], CONST[9]
>  5: MOV TEMP[1].xy, TEMP[0].xyxx
>  6: ADD TEMP[2].zw, TEMP[1], IMM[0].xxxx
>  7: MUL TEMP[2].xy, TEMP[2].zwzw, CONST[10]
>  8: MOV TEMP[3].xy, TEMP[2].xyyy
>  9: MOV TEMP[3].w, IMM[0].yyyy
> 10: TXL TEMP[3].w, TEMP[3], SAMP[0], 2D
> 11: MUL TEMP[3].z, TEMP[3].wwww, CONST[9].zzzz
> 12: MUL TEMP[3].z, TEMP[3].zzzz, IMM[0].zzzz
> 13: MUL TEMP[2], TEMP[0].yyyy, CONST[12]
> 14: MAD TEMP[2], TEMP[0].xxxx, CONST[11], TEMP[2]
> 15: MAD TEMP[1], TEMP[3].zzzz, CONST[13], TEMP[2]
> 16: ADD TEMP[1], TEMP[1], CONST[14]
> 17: MUL TEMP[2], TEMP[1].yyyy, CONST[1]
> 18: MAD TEMP[2], TEMP[1].xxxx, CONST[0], TEMP[2]
> 19: MAD TEMP[2], TEMP[1].zzzz, CONST[2], TEMP[2]
> 20: MAD TEMP[2], TEMP[1].wwww, CONST[3], TEMP[2]
> 21: MOV TEMP[0], TEMP[2]
> 22: MUL TEMP[2], TEMP[1].yyyy, CONST[5]
> 23: MAD TEMP[2], TEMP[1].xxxx, CONST[4], TEMP[2]
> 24: MAD TEMP[2], TEMP[1].zzzz, CONST[6], TEMP[2]
> 25: MAD TEMP[2], TEMP[1].wwww, CONST[7], TEMP[2]
> 26: MAD TEMP[1], TEMP[2], IMM[1].xyzz, IMM[0].xxyy
> 27: MOV OUT[1], TEMP[1]
> 28: MOV OUT[0], TEMP[0]
> 29: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 0)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 4)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %12, i32 8)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %12, i32 12)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %12, i32 16)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %12, i32 20)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %12, i32 24)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %12, i32 28)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %12, i32 32)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %12, i32 36)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %12, i32 40)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %12, i32 44)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %12, i32 48)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %12, i32 52)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %12, i32 56)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %12, i32 60)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %12, i32 64)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %12, i32 68)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %12, i32 72)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %12, i32 76)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %12, i32 80)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %12, i32 84)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %12, i32 88)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %12, i32 92)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %12, i32 96)
>  %38 = call float @llvm.SI.load.const(<16 x i8> %12, i32 100)
>  %39 = call float @llvm.SI.load.const(<16 x i8> %12, i32 104)
>  %40 = call float @llvm.SI.load.const(<16 x i8> %12, i32 108)
>  %41 = call float @llvm.SI.load.const(<16 x i8> %12, i32 112)
>  %42 = call float @llvm.SI.load.const(<16 x i8> %12, i32 116)
>  %43 = call float @llvm.SI.load.const(<16 x i8> %12, i32 120)
>  %44 = call float @llvm.SI.load.const(<16 x i8> %12, i32 124)
>  %45 = call float @llvm.SI.load.const(<16 x i8> %12, i32 144)
>  %46 = call float @llvm.SI.load.const(<16 x i8> %12, i32 148)
>  %47 = call float @llvm.SI.load.const(<16 x i8> %12, i32 152)
>  %48 = call float @llvm.SI.load.const(<16 x i8> %12, i32 160)
>  %49 = call float @llvm.SI.load.const(<16 x i8> %12, i32 164)
>  %50 = call float @llvm.SI.load.const(<16 x i8> %12, i32 176)
>  %51 = call float @llvm.SI.load.const(<16 x i8> %12, i32 180)
>  %52 = call float @llvm.SI.load.const(<16 x i8> %12, i32 184)
>  %53 = call float @llvm.SI.load.const(<16 x i8> %12, i32 188)
>  %54 = call float @llvm.SI.load.const(<16 x i8> %12, i32 192)
>  %55 = call float @llvm.SI.load.const(<16 x i8> %12, i32 196)
>  %56 = call float @llvm.SI.load.const(<16 x i8> %12, i32 200)
>  %57 = call float @llvm.SI.load.const(<16 x i8> %12, i32 204)
>  %58 = call float @llvm.SI.load.const(<16 x i8> %12, i32 208)
>  %59 = call float @llvm.SI.load.const(<16 x i8> %12, i32 212)
>  %60 = call float @llvm.SI.load.const(<16 x i8> %12, i32 216)
>  %61 = call float @llvm.SI.load.const(<16 x i8> %12, i32 220)
>  %62 = call float @llvm.SI.load.const(<16 x i8> %12, i32 224)
>  %63 = call float @llvm.SI.load.const(<16 x i8> %12, i32 228)
>  %64 = call float @llvm.SI.load.const(<16 x i8> %12, i32 232)
>  %65 = call float @llvm.SI.load.const(<16 x i8> %12, i32 236)
>  %66 = call float @llvm.SI.load.const(<16 x i8> %12, i32 240)
>  %67 = call float @llvm.SI.load.const(<16 x i8> %12, i32 244)
>  %68 = call float @llvm.SI.load.const(<16 x i8> %12, i32 248)
>  %69 = call float @llvm.SI.load.const(<16 x i8> %12, i32 252)
>  %70 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %71 = load <8 x i32> addrspace(2)* %70, !tbaa !0
>  %72 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %73 = load <4 x i32> addrspace(2)* %72, !tbaa !0
>  %74 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %75 = load <16 x i8> addrspace(2)* %74, !tbaa !0
>  %76 = add i32 %5, %7
>  %77 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %75, i32 0, i32 %76)
>  %78 = extractelement <4 x float> %77, i32 0
>  %79 = extractelement <4 x float> %77, i32 1
>  %80 = fadd float %66, %78
>  %81 = fadd float %67, %79
>  %82 = fadd float %80, %68
>  %83 = fadd float %81, %69
>  %84 = fmul float %80, %45
>  %85 = fmul float %81, %46
>  %86 = fadd float %82, 5.000000e-01
>  %87 = fadd float %83, 5.000000e-01
>  %88 = fmul float %86, %48
>  %89 = fmul float %87, %49
>  %90 = bitcast float %88 to i32
>  %91 = bitcast float %89 to i32
>  %92 = bitcast float 0.000000e+00 to i32
>  %93 = insertelement <4 x i32> undef, i32 %90, i32 0
>  %94 = insertelement <4 x i32> %93, i32 %91, i32 1
>  %95 = insertelement <4 x i32> %94, i32 %92, i32 2
>  %96 = insertelement <4 x i32> %95, i32 undef, i32 3
>  %97 = bitcast <8 x i32> %71 to <32 x i8>
>  %98 = bitcast <4 x i32> %73 to <16 x i8>
>  %99 = call <4 x float> @llvm.SI.samplel.v4i32(<4 x i32> %96, <32 x i8> %97, <16 x i8> %98, i32 2)
>  %100 = extractelement <4 x float> %99, i32 3
>  %101 = fmul float %100, %47
>  %102 = fmul float %101, 2.550000e+02
>  %103 = fmul float %85, %54
>  %104 = fmul float %85, %55
>  %105 = fmul float %85, %56
>  %106 = fmul float %85, %57
>  %107 = fmul float %84, %50
>  %108 = fadd float %107, %103
>  %109 = fmul float %84, %51
>  %110 = fadd float %109, %104
>  %111 = fmul float %84, %52
>  %112 = fadd float %111, %105
>  %113 = fmul float %84, %53
>  %114 = fadd float %113, %106
>  %115 = fmul float %102, %58
>  %116 = fadd float %115, %108
>  %117 = fmul float %102, %59
>  %118 = fadd float %117, %110
>  %119 = fmul float %102, %60
>  %120 = fadd float %119, %112
>  %121 = fmul float %102, %61
>  %122 = fadd float %121, %114
>  %123 = fadd float %116, %62
>  %124 = fadd float %118, %63
>  %125 = fadd float %120, %64
>  %126 = fadd float %122, %65
>  %127 = fmul float %124, %17
>  %128 = fmul float %124, %18
>  %129 = fmul float %124, %19
>  %130 = fmul float %124, %20
>  %131 = fmul float %123, %13
>  %132 = fadd float %131, %127
>  %133 = fmul float %123, %14
>  %134 = fadd float %133, %128
>  %135 = fmul float %123, %15
>  %136 = fadd float %135, %129
>  %137 = fmul float %123, %16
>  %138 = fadd float %137, %130
>  %139 = fmul float %125, %21
>  %140 = fadd float %139, %132
>  %141 = fmul float %125, %22
>  %142 = fadd float %141, %134
>  %143 = fmul float %125, %23
>  %144 = fadd float %143, %136
>  %145 = fmul float %125, %24
>  %146 = fadd float %145, %138
>  %147 = fmul float %126, %25
>  %148 = fadd float %147, %140
>  %149 = fmul float %126, %26
>  %150 = fadd float %149, %142
>  %151 = fmul float %126, %27
>  %152 = fadd float %151, %144
>  %153 = fmul float %126, %28
>  %154 = fadd float %153, %146
>  %155 = fmul float %124, %33
>  %156 = fmul float %124, %34
>  %157 = fmul float %124, %35
>  %158 = fmul float %124, %36
>  %159 = fmul float %123, %29
>  %160 = fadd float %159, %155
>  %161 = fmul float %123, %30
>  %162 = fadd float %161, %156
>  %163 = fmul float %123, %31
>  %164 = fadd float %163, %157
>  %165 = fmul float %123, %32
>  %166 = fadd float %165, %158
>  %167 = fmul float %125, %37
>  %168 = fadd float %167, %160
>  %169 = fmul float %125, %38
>  %170 = fadd float %169, %162
>  %171 = fmul float %125, %39
>  %172 = fadd float %171, %164
>  %173 = fmul float %125, %40
>  %174 = fadd float %173, %166
>  %175 = fmul float %126, %41
>  %176 = fadd float %175, %168
>  %177 = fmul float %126, %42
>  %178 = fadd float %177, %170
>  %179 = fmul float %126, %43
>  %180 = fadd float %179, %172
>  %181 = fmul float %126, %44
>  %182 = fadd float %181, %174
>  %183 = fmul float %176, 5.000000e-01
>  %184 = fadd float %183, 5.000000e-01
>  %185 = fmul float %178, -5.000000e-01
>  %186 = fadd float %185, 5.000000e-01
>  %187 = fmul float %180, 1.000000e+00
>  %188 = fadd float %187, 0.000000e+00
>  %189 = fmul float %182, 1.000000e+00
>  %190 = fadd float %189, 0.000000e+00
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %184, float %186, float %188, float %190)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %148, float %150, float %152, float %154)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.samplel.v4i32(<4 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                                            ; 4A00000A
>	s_load_dwordx4 s[8:11], s[8:9], 0x0                                  ; C0840900
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	buffer_load_format_xyzw v[0:3], v0, s[8:11], 0 idxen                 ; E00C2000 80020000
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                                   ; C0800300
>	s_waitcnt vmcnt(0) lgkmcnt(0)                                        ; BF8C0070
>	s_buffer_load_dword s8, s[0:3], 0x3c                                 ; C204013C
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_add_f32_e32 v4, s8, v0                                             ; 06080008
>	s_buffer_load_dword s8, s[0:3], 0x24                                 ; C2040124
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mul_f32_e32 v5, s8, v4                                             ; 100A0808
>	s_buffer_load_dword s8, s[0:3], 0x3d                                 ; C204013D
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_add_f32_e32 v0, s8, v1                                             ; 06000208
>	s_buffer_load_dword s8, s[0:3], 0x25                                 ; C2040125
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mul_f32_e32 v1, s8, v0                                             ; 10020008
>	s_buffer_load_dword s8, s[0:3], 0x30                                 ; C2040130
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mul_f32_e32 v2, s8, v1                                             ; 10040208
>	s_buffer_load_dword s8, s[0:3], 0x2c                                 ; C204012C
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mad_f32 v2, v5, s8, v2                                             ; D2820002 04081105
>	s_buffer_load_dword s8, s[0:3], 0x3f                                 ; C204013F
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_add_f32_e32 v0, s8, v0                                             ; 06000008
>	v_add_f32_e32 v0, 0.5, v0                                            ; 060000F0
>	s_buffer_load_dword s8, s[0:3], 0x29                                 ; C2040129
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mul_f32_e32 v7, s8, v0                                             ; 100E0008
>	s_buffer_load_dword s8, s[0:3], 0x3e                                 ; C204013E
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_add_f32_e32 v0, s8, v4                                             ; 06000808
>	v_add_f32_e32 v0, 0.5, v0                                            ; 060000F0
>	s_buffer_load_dword s8, s[0:3], 0x28                                 ; C2040128
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mul_f32_e32 v6, s8, v0                                             ; 100C0008
>	v_mov_b32_e32 v8, 0                                                  ; 7E100280
>	s_load_dwordx4 s[8:11], s[4:5], 0x0                                  ; C0840500
>	s_load_dwordx8 s[12:19], s[6:7], 0x0                                 ; C0C60700
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	image_sample_l v0, 8, 0, 0, 0, 0, 0, 0, 0, v[6:9], s[12:19], s[8:11] ; F0900800 00430006
>	s_buffer_load_dword s4, s[0:3], 0x26                                 ; C2020126
>	s_waitcnt vmcnt(0) lgkmcnt(0)                                        ; BF8C0070
>	v_mul_f32_e32 v0, s4, v0                                             ; 10000004
>	v_mul_f32_e32 v0, 0x437f0000, v0                                     ; 100000FF 437F0000
>	s_buffer_load_dword s4, s[0:3], 0x34                                 ; C2020134
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mad_f32 v2, v0, s4, v2                                             ; D2820002 04080900
>	s_buffer_load_dword s4, s[0:3], 0x38                                 ; C2020138
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_add_f32_e32 v2, s4, v2                                             ; 06040404
>	s_buffer_load_dword s4, s[0:3], 0x31                                 ; C2020131
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mul_f32_e32 v3, s4, v1                                             ; 10060204
>	s_buffer_load_dword s4, s[0:3], 0x2d                                 ; C202012D
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mad_f32 v3, v5, s4, v3                                             ; D2820003 040C0905
>	s_buffer_load_dword s4, s[0:3], 0x35                                 ; C2020135
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mad_f32 v3, v0, s4, v3                                             ; D2820003 040C0900
>	s_buffer_load_dword s4, s[0:3], 0x39                                 ; C2020139
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_add_f32_e32 v3, s4, v3                                             ; 06060604
>	s_buffer_load_dword s4, s[0:3], 0x15                                 ; C2020115
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mul_f32_e32 v4, s4, v3                                             ; 10080604
>	s_buffer_load_dword s4, s[0:3], 0x11                                 ; C2020111
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mad_f32 v4, v2, s4, v4                                             ; D2820004 04100902
>	s_buffer_load_dword s4, s[0:3], 0x32                                 ; C2020132
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mul_f32_e32 v6, s4, v1                                             ; 100C0204
>	s_buffer_load_dword s4, s[0:3], 0x2e                                 ; C202012E
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mad_f32 v6, v5, s4, v6                                             ; D2820006 04180905
>	s_buffer_load_dword s4, s[0:3], 0x36                                 ; C2020136
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mad_f32 v6, v0, s4, v6                                             ; D2820006 04180900
>	s_buffer_load_dword s4, s[0:3], 0x3a                                 ; C202013A
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_add_f32_e32 v6, s4, v6                                             ; 060C0C04
>	s_buffer_load_dword s4, s[0:3], 0x19                                 ; C2020119
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mad_f32 v4, v6, s4, v4                                             ; D2820004 04100906
>	s_buffer_load_dword s4, s[0:3], 0x33                                 ; C2020133
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mul_f32_e32 v1, s4, v1                                             ; 10020204
>	s_buffer_load_dword s4, s[0:3], 0x2f                                 ; C202012F
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mad_f32 v1, v5, s4, v1                                             ; D2820001 04040905
>	s_buffer_load_dword s4, s[0:3], 0x37                                 ; C2020137
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mad_f32 v0, v0, s4, v1                                             ; D2820000 04040900
>	s_buffer_load_dword s4, s[0:3], 0x3b                                 ; C202013B
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_add_f32_e32 v0, s4, v0                                             ; 06000004
>	s_buffer_load_dword s4, s[0:3], 0x1d                                 ; C202011D
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mad_f32 v1, v0, s4, v4                                             ; D2820001 04100900
>	v_mad_f32 v1, -0.5, v1, 0.5                                          ; D2820001 03C202F1
>	s_buffer_load_dword s4, s[0:3], 0x17                                 ; C2020117
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mul_f32_e32 v4, s4, v3                                             ; 10080604
>	s_buffer_load_dword s4, s[0:3], 0x13                                 ; C2020113
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mad_f32 v4, v2, s4, v4                                             ; D2820004 04100902
>	s_buffer_load_dword s4, s[0:3], 0x1b                                 ; C202011B
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mad_f32 v4, v6, s4, v4                                             ; D2820004 04100906
>	s_buffer_load_dword s4, s[0:3], 0x1f                                 ; C202011F
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mad_f32 v4, v0, s4, v4                                             ; D2820004 04100900
>	v_add_f32_e32 v4, 0, v4                                              ; 06080880
>	s_buffer_load_dword s4, s[0:3], 0x16                                 ; C2020116
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mul_f32_e32 v5, s4, v3                                             ; 100A0604
>	s_buffer_load_dword s4, s[0:3], 0x12                                 ; C2020112
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mad_f32 v5, v2, s4, v5                                             ; D2820005 04140902
>	s_buffer_load_dword s4, s[0:3], 0x1a                                 ; C202011A
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mad_f32 v5, v6, s4, v5                                             ; D2820005 04140906
>	s_buffer_load_dword s4, s[0:3], 0x1e                                 ; C202011E
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mad_f32 v5, v0, s4, v5                                             ; D2820005 04140900
>	v_add_f32_e32 v5, 0, v5                                              ; 060A0A80
>	s_buffer_load_dword s4, s[0:3], 0x14                                 ; C2020114
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mul_f32_e32 v7, s4, v3                                             ; 100E0604
>	s_buffer_load_dword s4, s[0:3], 0x10                                 ; C2020110
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mad_f32 v7, v2, s4, v7                                             ; D2820007 041C0902
>	s_buffer_load_dword s4, s[0:3], 0x18                                 ; C2020118
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mad_f32 v7, v6, s4, v7                                             ; D2820007 041C0906
>	s_buffer_load_dword s4, s[0:3], 0x1c                                 ; C202011C
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mad_f32 v7, v0, s4, v7                                             ; D2820007 041C0900
>	v_mad_f32 v7, 0.5, v7, 0.5                                           ; D2820007 03C20EF0
>	exp 15, 32, 0, 0, 0, v7, v1, v5, v4                                  ; F800020F 04050107
>	s_buffer_load_dword s4, s[0:3], 0x7                                  ; C2020107
>	s_waitcnt expcnt(0) lgkmcnt(0)                                       ; BF8C000F
>	v_mul_f32_e32 v1, s4, v3                                             ; 10020604
>	s_buffer_load_dword s4, s[0:3], 0x3                                  ; C2020103
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mad_f32 v1, v2, s4, v1                                             ; D2820001 04040902
>	s_buffer_load_dword s4, s[0:3], 0xb                                  ; C202010B
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mad_f32 v1, v6, s4, v1                                             ; D2820001 04040906
>	s_buffer_load_dword s4, s[0:3], 0xf                                  ; C202010F
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mad_f32 v1, v0, s4, v1                                             ; D2820001 04040900
>	s_buffer_load_dword s4, s[0:3], 0x6                                  ; C2020106
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mul_f32_e32 v4, s4, v3                                             ; 10080604
>	s_buffer_load_dword s4, s[0:3], 0x2                                  ; C2020102
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mad_f32 v4, v2, s4, v4                                             ; D2820004 04100902
>	s_buffer_load_dword s4, s[0:3], 0xa                                  ; C202010A
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mad_f32 v4, v6, s4, v4                                             ; D2820004 04100906
>	s_buffer_load_dword s4, s[0:3], 0xe                                  ; C202010E
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mad_f32 v4, v0, s4, v4                                             ; D2820004 04100900
>	s_buffer_load_dword s4, s[0:3], 0x5                                  ; C2020105
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mul_f32_e32 v5, s4, v3                                             ; 100A0604
>	s_buffer_load_dword s4, s[0:3], 0x1                                  ; C2020101
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mad_f32 v5, v2, s4, v5                                             ; D2820005 04140902
>	s_buffer_load_dword s4, s[0:3], 0x9                                  ; C2020109
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mad_f32 v5, v6, s4, v5                                             ; D2820005 04140906
>	s_buffer_load_dword s4, s[0:3], 0xd                                  ; C202010D
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mad_f32 v5, v0, s4, v5                                             ; D2820005 04140900
>	s_buffer_load_dword s4, s[0:3], 0x4                                  ; C2020104
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mul_f32_e32 v3, s4, v3                                             ; 10060604
>	s_buffer_load_dword s4, s[0:3], 0x0                                  ; C2020100
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mad_f32 v2, v2, s4, v3                                             ; D2820002 040C0902
>	s_buffer_load_dword s4, s[0:3], 0x8                                  ; C2020108
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mad_f32 v2, v6, s4, v2                                             ; D2820002 04080906
>	s_buffer_load_dword s0, s[0:3], 0xc                                  ; C200010C
>	s_waitcnt lgkmcnt(0)                                                 ; BF8C007F
>	v_mad_f32 v0, v0, s0, v2                                             ; D2820000 04080100
>	exp 15, 12, 0, 1, 0, v0, v5, v4, v1                                  ; F80008CF 01040500
>	s_endpgm                                                             ; BF810000
>
>SGPRS: 24
>VGPRS: 12
>Code Size: 924 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL TEMP[0], LOCAL
>IMM[0] FLT32 {    1.0000,    -1.0000,     0.0000,     0.0000}
>  0: MAD TEMP[0].xy, IN[0], IMM[0].xyxx, IMM[0].zxzz
>  1: MOV TEMP[0].xy, TEMP[0].xyyy
>  2: TEX TEMP[0], TEMP[0], SAMP[0], 2D
>  3: MOV OUT[0], TEMP[0]
>  4: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %23 = load <8 x i32> addrspace(2)* %22, !tbaa !0
>  %24 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %25 = load <4 x i32> addrspace(2)* %24, !tbaa !0
>  %26 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %27 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %28 = fmul float %26, 1.000000e+00
>  %29 = fadd float %28, 0.000000e+00
>  %30 = fmul float %27, -1.000000e+00
>  %31 = fadd float %30, 1.000000e+00
>  %32 = bitcast float %29 to i32
>  %33 = bitcast float %31 to i32
>  %34 = insertelement <2 x i32> undef, i32 %32, i32 0
>  %35 = insertelement <2 x i32> %34, i32 %33, i32 1
>  %36 = bitcast <8 x i32> %23 to <32 x i8>
>  %37 = bitcast <4 x i32> %25 to <16 x i8>
>  %38 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %35, <32 x i8> %36, <16 x i8> %37, i32 2)
>  %39 = extractelement <4 x float> %38, i32 0
>  %40 = extractelement <4 x float> %38, i32 1
>  %41 = extractelement <4 x float> %38, i32 2
>  %42 = extractelement <4 x float> %38, i32 3
>  %43 = call i32 @llvm.SI.packf16(float %39, float %40)
>  %44 = bitcast i32 %43 to float
>  %45 = call i32 @llvm.SI.packf16(float %41, float %42)
>  %46 = bitcast i32 %45 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %44, float %46, float %44, float %46)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                  ; BEFE0A7E
>	s_mov_b32 m0, s9                                                      ; BEFC0309
>	v_interp_p1_f32 v2, v0, 1, 0, [m0]                                    ; C8080100
>	v_interp_p2_f32 v2, [v2], v1, 1, 0, [m0]                              ; C8090101
>	v_sub_f32_e32 v3, 1.0, v2                                             ; 080604F2
>	v_interp_p1_f32 v4, v0, 0, 0, [m0]                                    ; C8100000
>	v_interp_p2_f32 v4, [v4], v1, 0, 0, [m0]                              ; C8110001
>	v_add_f32_e32 v2, 0, v4                                               ; 06040880
>	s_load_dwordx4 s[0:3], s[4:5], 0x0                                    ; C0800500
>	s_load_dwordx8 s[4:11], s[6:7], 0x0                                   ; C0C20700
>	s_waitcnt lgkmcnt(0)                                                  ; BF8C007F
>	image_sample v[0:3], 15, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[4:11], s[0:3] ; F0800F00 00010002
>	s_waitcnt vmcnt(0)                                                    ; BF8C0770
>	v_cvt_pkrtz_f16_f32_e32 v4, v2, v3                                    ; 5E080702
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v1                                    ; 5E000300
>	exp 15, 0, 1, 1, 1, v0, v4, v0, v4                                    ; F8001C0F 04000400
>	s_endpgm                                                              ; BF810000
>
>SGPRS: 16
>VGPRS: 8
>Code Size: 76 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL IN[2]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL OUT[2], GENERIC[10]
>DCL CONST[0..254]
>DCL TEMP[0..4], LOCAL
>DCL ADDR[0]
>IMM[0] FLT32 {    1.0000,     2.0000,    -1.0000,     0.0000}
>  0: F2I TEMP[0].x, IN[2].xxxx
>  1: UARL ADDR[0].x, TEMP[0].xxxx
>  2: MOV TEMP[0].xyz, CONST[ADDR[0].x+7].xyzx
>  3: MAD TEMP[1].xy, CONST[6], IN[1].zwzw, TEMP[0]
>  4: MOV TEMP[1].xy, TEMP[1].xyxx
>  5: ADD TEMP[2].xy, TEMP[0].zzzz, -CONST[5].xzzw
>  6: MOV TEMP[0].xy, TEMP[2].xyxx
>  7: MUL TEMP[2].z, TEMP[0].zzzz, CONST[4].zzzz
>  8: MOV TEMP[0].z, TEMP[2].zzzz
>  9: LRP TEMP[2].x, TEMP[2].zzzz, CONST[4].yyyy, CONST[4].xxxx
> 10: MUL TEMP[3], TEMP[0], CONST[5].ywzw
> 11: MOV_SAT TEMP[3], TEMP[3]
> 12: ADD TEMP[4].y, -TEMP[3].yyyy, IMM[0].xxxx
> 13: MUL TEMP[3].xy, TEMP[4].yyyy, TEMP[3].xxxx
> 14: MOV TEMP[3].xy, TEMP[3].xyxx
> 15: MAD TEMP[4].xy, IN[1], IMM[0].yyyy, IMM[0].zzzz
> 16: MOV TEMP[0].xy, TEMP[4].xyxx
> 17: MAD TEMP[4].xy, TEMP[0], -TEMP[2].xxxx, IN[0]
> 18: MUL TEMP[2], TEMP[4].yyyy, CONST[1]
> 19: MAD TEMP[0], TEMP[4].xxxx, CONST[0], TEMP[2]
> 20: MAD TEMP[0], CONST[4].wwww, CONST[2], TEMP[0]
> 21: ADD TEMP[0], TEMP[0], CONST[3]
> 22: MOV TEMP[1].zw, IMM[0].xxwx
> 23: MOV TEMP[3].zw, IMM[0].xxwx
> 24: MOV OUT[1], TEMP[1]
> 25: MOV OUT[0], TEMP[0]
> 26: MOV OUT[2], TEMP[3]
> 27: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 0)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 4)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %12, i32 8)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %12, i32 12)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %12, i32 16)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %12, i32 20)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %12, i32 24)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %12, i32 28)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %12, i32 32)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %12, i32 36)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %12, i32 40)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %12, i32 44)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %12, i32 48)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %12, i32 52)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %12, i32 56)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %12, i32 60)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %12, i32 64)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %12, i32 68)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %12, i32 72)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %12, i32 76)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %12, i32 80)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %12, i32 84)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %12, i32 88)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %12, i32 92)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %12, i32 96)
>  %38 = call float @llvm.SI.load.const(<16 x i8> %12, i32 100)
>  %39 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %40 = load <16 x i8> addrspace(2)* %39, !tbaa !0
>  %41 = add i32 %5, %7
>  %42 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %40, i32 0, i32 %41)
>  %43 = extractelement <4 x float> %42, i32 0
>  %44 = extractelement <4 x float> %42, i32 1
>  %45 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %46 = load <16 x i8> addrspace(2)* %45, !tbaa !0
>  %47 = add i32 %5, %7
>  %48 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %46, i32 0, i32 %47)
>  %49 = extractelement <4 x float> %48, i32 0
>  %50 = extractelement <4 x float> %48, i32 1
>  %51 = extractelement <4 x float> %48, i32 2
>  %52 = extractelement <4 x float> %48, i32 3
>  %53 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 2
>  %54 = load <16 x i8> addrspace(2)* %53, !tbaa !0
>  %55 = add i32 %5, %7
>  %56 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %54, i32 0, i32 %55)
>  %57 = extractelement <4 x float> %56, i32 0
>  %58 = fptosi float %57 to i32
>  %59 = bitcast i32 %58 to float
>  %60 = bitcast float %59 to i32
>  %61 = shl i32 %60, 4
>  %62 = add i32 %61, 112
>  %63 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %62)
>  %64 = shl i32 %60, 4
>  %65 = add i32 %64, 116
>  %66 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %65)
>  %67 = shl i32 %60, 4
>  %68 = add i32 %67, 120
>  %69 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %68)
>  %70 = fmul float %37, %51
>  %71 = fadd float %70, %63
>  %72 = fmul float %38, %52
>  %73 = fadd float %72, %66
>  %74 = fsub float -0.000000e+00, %33
>  %75 = fadd float %69, %74
>  %76 = fsub float -0.000000e+00, %35
>  %77 = fadd float %69, %76
>  %78 = fmul float %69, %31
>  %79 = call float @llvm.AMDGPU.lrp(float %78, float %30, float %29)
>  %80 = fmul float %75, %34
>  %81 = fmul float %77, %36
>  %82 = fmul float %78, %35
>  %83 = fmul float 0.000000e+00, %36
>  %84 = call float @llvm.AMDIL.clamp.(float %80, float 0.000000e+00, float 1.000000e+00)
>  %85 = call float @llvm.AMDIL.clamp.(float %81, float 0.000000e+00, float 1.000000e+00)
>  %86 = call float @llvm.AMDIL.clamp.(float %82, float 0.000000e+00, float 1.000000e+00)
>  %87 = call float @llvm.AMDIL.clamp.(float %83, float 0.000000e+00, float 1.000000e+00)
>  %88 = fsub float -0.000000e+00, %85
>  %89 = fadd float %88, 1.000000e+00
>  %90 = fmul float %89, %84
>  %91 = fmul float %89, %84
>  %92 = fmul float %49, 2.000000e+00
>  %93 = fadd float %92, -1.000000e+00
>  %94 = fmul float %50, 2.000000e+00
>  %95 = fadd float %94, -1.000000e+00
>  %96 = fsub float -0.000000e+00, %79
>  %97 = fmul float %93, %96
>  %98 = fadd float %97, %43
>  %99 = fsub float -0.000000e+00, %79
>  %100 = fmul float %95, %99
>  %101 = fadd float %100, %44
>  %102 = fmul float %101, %17
>  %103 = fmul float %101, %18
>  %104 = fmul float %101, %19
>  %105 = fmul float %101, %20
>  %106 = fmul float %98, %13
>  %107 = fadd float %106, %102
>  %108 = fmul float %98, %14
>  %109 = fadd float %108, %103
>  %110 = fmul float %98, %15
>  %111 = fadd float %110, %104
>  %112 = fmul float %98, %16
>  %113 = fadd float %112, %105
>  %114 = fmul float %32, %21
>  %115 = fadd float %114, %107
>  %116 = fmul float %32, %22
>  %117 = fadd float %116, %109
>  %118 = fmul float %32, %23
>  %119 = fadd float %118, %111
>  %120 = fmul float %32, %24
>  %121 = fadd float %120, %113
>  %122 = fadd float %115, %25
>  %123 = fadd float %117, %26
>  %124 = fadd float %119, %27
>  %125 = fadd float %121, %28
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %71, float %73, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %90, float %91, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %122, float %123, float %124, float %125)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDGPU.lrp(float, float, float) #2
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.clamp.(float, float, float) #2
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                             ; 4A00000A
>	s_load_dwordx4 s[4:7], s[8:9], 0x0                    ; C0820900
>	s_load_dwordx4 s[12:15], s[8:9], 0x4                  ; C0860904
>	s_load_dwordx4 s[8:11], s[8:9], 0x8                   ; C0840908
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[8:11], 0 idxen  ; E00C2000 80020100
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_cvt_i32_f32_e32 v1, v1                              ; 7E021101
>	v_lshlrev_b32_e32 v1, 4, v1                           ; 34020284
>	v_add_i32_e32 v2, 0x74, v1                            ; 4A0402FF 00000074
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                    ; C0800300
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	buffer_load_dword v2, v2, s[0:3], 0 offen             ; E0301000 80000202
>	buffer_load_format_xyzw v[3:6], v0, s[12:15], 0 idxen ; E00C2000 80030300
>	s_buffer_load_dword s8, s[0:3], 0x19                  ; C2040119
>	s_waitcnt vmcnt(0) lgkmcnt(0)                         ; BF8C0070
>	v_mad_f32 v2, s8, v6, v2                              ; D2820002 040A0C08
>	v_add_i32_e32 v7, 0x70, v1                            ; 4A0E02FF 00000070
>	buffer_load_dword v7, v7, s[0:3], 0 offen             ; E0301000 80000707
>	s_buffer_load_dword s8, s[0:3], 0x18                  ; C2040118
>	s_waitcnt vmcnt(0) lgkmcnt(0)                         ; BF8C0070
>	v_mad_f32 v7, s8, v5, v7                              ; D2820007 041E0A08
>	v_mov_b32_e32 v8, 1.0                                 ; 7E1002F2
>	v_mov_b32_e32 v9, 0                                   ; 7E120280
>	exp 15, 32, 0, 0, 0, v7, v2, v9, v8                   ; F800020F 08090207
>	v_add_i32_e32 v1, 0x78, v1                            ; 4A0202FF 00000078
>	buffer_load_dword v1, v1, s[0:3], 0 offen             ; E0301000 80000101
>	s_buffer_load_dword s8, s[0:3], 0x16                  ; C2040116
>	s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)               ; BF8C0000
>	v_subrev_f32_e32 v2, s8, v1                           ; 0A040208
>	s_buffer_load_dword s8, s[0:3], 0x17                  ; C2040117
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v2, s8, v2                              ; 10040408
>	v_add_f32_e64 v2, 0, v2 clamp                         ; D2060802 00020480
>	v_sub_f32_e32 v2, 1.0, v2                             ; 080404F2
>	s_buffer_load_dword s8, s[0:3], 0x14                  ; C2040114
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_subrev_f32_e32 v7, s8, v1                           ; 0A0E0208
>	s_buffer_load_dword s8, s[0:3], 0x15                  ; C2040115
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v7, s8, v7                              ; 100E0E08
>	v_add_f32_e64 v7, 0, v7 clamp                         ; D2060807 00020E80
>	v_mul_f32_e32 v2, v7, v2                              ; 10040507
>	exp 15, 33, 0, 0, 0, v2, v2, v9, v8                   ; F800021F 08090202
>	s_buffer_load_dword s8, s[0:3], 0x12                  ; C2040112
>	s_waitcnt expcnt(0) lgkmcnt(0)                        ; BF8C000F
>	v_mul_f32_e32 v2, s8, v1                              ; 10040208
>	v_mad_f32 v1, -v1, s8, 1.0                            ; D2820001 23C81101
>	s_buffer_load_dword s8, s[0:3], 0x10                  ; C2040110
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v1, s8, v1                              ; 10020208
>	s_buffer_load_dword s8, s[0:3], 0x11                  ; C2040111
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v1, v2, s8, v1                              ; D2820001 04041102
>	v_mad_f32 v2, 2.0, v3, -1.0                           ; D2820002 03CE06F4
>	buffer_load_format_xyzw v[7:10], v0, s[4:7], 0 idxen  ; E00C2000 80010700
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mad_f32 v0, -v2, v1, v7                             ; D2820000 241E0302
>	v_mad_f32 v2, 2.0, v4, -1.0                           ; D2820002 03CE08F4
>	v_mad_f32 v1, -v2, v1, v8                             ; D2820001 24220302
>	s_buffer_load_dword s4, s[0:3], 0x7                   ; C2020107
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v2, s4, v1                              ; 10040204
>	s_buffer_load_dword s4, s[0:3], 0x3                   ; C2020103
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v2, v0, s4, v2                              ; D2820002 04080900
>	s_buffer_load_dword s4, s[0:3], 0xb                   ; C202010B
>	s_buffer_load_dword s5, s[0:3], 0x13                  ; C2028113
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mov_b32_e32 v3, s4                                  ; 7E060204
>	v_mad_f32 v2, v3, s5, v2                              ; D2820002 04080B03
>	s_buffer_load_dword s4, s[0:3], 0xf                   ; C202010F
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v2, s4, v2                              ; 06040404
>	s_buffer_load_dword s4, s[0:3], 0x6                   ; C2020106
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v3, s4, v1                              ; 10060204
>	s_buffer_load_dword s4, s[0:3], 0x2                   ; C2020102
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v3, v0, s4, v3                              ; D2820003 040C0900
>	s_buffer_load_dword s4, s[0:3], 0xa                   ; C202010A
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mov_b32_e32 v4, s4                                  ; 7E080204
>	v_mad_f32 v3, v4, s5, v3                              ; D2820003 040C0B04
>	s_buffer_load_dword s4, s[0:3], 0xe                   ; C202010E
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v3, s4, v3                              ; 06060604
>	s_buffer_load_dword s4, s[0:3], 0x5                   ; C2020105
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v4, s4, v1                              ; 10080204
>	s_buffer_load_dword s4, s[0:3], 0x1                   ; C2020101
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v4, v0, s4, v4                              ; D2820004 04100900
>	s_buffer_load_dword s4, s[0:3], 0x9                   ; C2020109
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mov_b32_e32 v5, s4                                  ; 7E0A0204
>	v_mad_f32 v4, v5, s5, v4                              ; D2820004 04100B05
>	s_buffer_load_dword s4, s[0:3], 0xd                   ; C202010D
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v4, s4, v4                              ; 06080804
>	s_buffer_load_dword s4, s[0:3], 0x4                   ; C2020104
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v1, s4, v1                              ; 10020204
>	s_buffer_load_dword s4, s[0:3], 0x0                   ; C2020100
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v0, v0, s4, v1                              ; D2820000 04040900
>	s_buffer_load_dword s4, s[0:3], 0x8                   ; C2020108
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mov_b32_e32 v1, s4                                  ; 7E020204
>	v_mad_f32 v0, v1, s5, v0                              ; D2820000 04000B01
>	s_buffer_load_dword s0, s[0:3], 0xc                   ; C200010C
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v0, s0, v0                              ; 06000000
>	exp 15, 12, 0, 1, 0, v0, v4, v3, v2                   ; F80008CF 02030400
>	s_endpgm                                              ; BF810000
>
>SGPRS: 24
>VGPRS: 12
>Code Size: 580 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL IN[1], GENERIC[10], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL CONST[0]
>DCL TEMP[0..2], LOCAL
>IMM[0] FLT32 {    2.2000,     1.0000,     0.4545,     0.0000}
>  0: MOV TEMP[0].xy, IN[0].xyyy
>  1: TEX TEMP[0], TEMP[0], SAMP[0], 2D
>  2: POW TEMP[1].x, TEMP[0].xxxx, IMM[0].xxxx
>  3: POW TEMP[1].y, TEMP[0].yyyy, IMM[0].xxxx
>  4: POW TEMP[1].z, TEMP[0].zzzz, IMM[0].xxxx
>  5: POW TEMP[1].w, TEMP[0].wwww, IMM[0].yyyy
>  6: MOV TEMP[0].w, TEMP[1].wwww
>  7: MUL TEMP[1].xyz, TEMP[1], IN[1].xxxx
>  8: MOV TEMP[0].xyz, TEMP[1].xyzx
>  9: MUL TEMP[0].xyz, TEMP[0], CONST[0]
> 10: ABS TEMP[1].x, TEMP[0].xxxx
> 11: LG2 TEMP[1].x, TEMP[1].xxxx
> 12: ABS TEMP[2].x, TEMP[0].yyyy
> 13: LG2 TEMP[2].x, TEMP[2].xxxx
> 14: MOV TEMP[1].y, TEMP[2].xxxx
> 15: ABS TEMP[0].x, TEMP[0].zzzz
> 16: LG2 TEMP[0].x, TEMP[0].xxxx
> 17: MOV TEMP[1].z, TEMP[0].xxxx
> 18: MUL TEMP[0].xyz, TEMP[1], IMM[0].zzzz
> 19: EX2 TEMP[1].x, TEMP[0].xxxx
> 20: EX2 TEMP[2].x, TEMP[0].yyyy
> 21: MOV TEMP[1].y, TEMP[2].xxxx
> 22: EX2 TEMP[0].x, TEMP[0].zzzz
> 23: MOV TEMP[1].z, TEMP[0].xxxx
> 24: MOV TEMP[1].w, IMM[0].yyyy
> 25: MOV OUT[0], TEMP[1]
> 26: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %23 = load <16 x i8> addrspace(2)* %22, !tbaa !0
>  %24 = call float @llvm.SI.load.const(<16 x i8> %23, i32 0)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %23, i32 4)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %23, i32 8)
>  %27 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %28 = load <8 x i32> addrspace(2)* %27, !tbaa !0
>  %29 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %30 = load <4 x i32> addrspace(2)* %29, !tbaa !0
>  %31 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %32 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %33 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %5, <2 x i32> %7)
>  %34 = bitcast float %31 to i32
>  %35 = bitcast float %32 to i32
>  %36 = insertelement <2 x i32> undef, i32 %34, i32 0
>  %37 = insertelement <2 x i32> %36, i32 %35, i32 1
>  %38 = bitcast <8 x i32> %28 to <32 x i8>
>  %39 = bitcast <4 x i32> %30 to <16 x i8>
>  %40 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %37, <32 x i8> %38, <16 x i8> %39, i32 2)
>  %41 = extractelement <4 x float> %40, i32 0
>  %42 = extractelement <4 x float> %40, i32 1
>  %43 = extractelement <4 x float> %40, i32 2
>  %44 = call float @llvm.pow.f32(float %41, float 0x40019999A0000000)
>  %45 = call float @llvm.pow.f32(float %42, float 0x40019999A0000000)
>  %46 = call float @llvm.pow.f32(float %43, float 0x40019999A0000000)
>  %47 = fmul float %44, %33
>  %48 = fmul float %45, %33
>  %49 = fmul float %46, %33
>  %50 = fmul float %47, %24
>  %51 = fmul float %48, %25
>  %52 = fmul float %49, %26
>  %53 = call float @fabs(float %50)
>  %54 = call float @llvm.log2.f32(float %53)
>  %55 = call float @fabs(float %51)
>  %56 = call float @llvm.log2.f32(float %55)
>  %57 = call float @fabs(float %52)
>  %58 = call float @llvm.log2.f32(float %57)
>  %59 = fmul float %54, 0x3FDD1743E0000000
>  %60 = fmul float %56, 0x3FDD1743E0000000
>  %61 = fmul float %58, 0x3FDD1743E0000000
>  %62 = call float @llvm.AMDIL.exp.(float %59)
>  %63 = call float @llvm.AMDIL.exp.(float %60)
>  %64 = call float @llvm.AMDIL.exp.(float %61)
>  %65 = call i32 @llvm.SI.packf16(float %62, float %63)
>  %66 = bitcast i32 %65 to float
>  %67 = call i32 @llvm.SI.packf16(float %64, float 1.000000e+00)
>  %68 = bitcast i32 %67 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %66, float %68, float %66, float %68)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.pow.f32(float, float) #1
>
>; Function Attrs: readnone
>declare float @fabs(float) #2
>
>; Function Attrs: nounwind readnone
>declare float @llvm.log2.f32(float) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.exp.(float) #2
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                   ; BEFE0A7E
>	s_mov_b32 m0, s9                                                       ; BEFC0309
>	v_interp_p1_f32 v3, v0, 1, 0, [m0]                                     ; C80C0100
>	v_interp_p2_f32 v3, [v3], v1, 1, 0, [m0]                               ; C80D0101
>	v_interp_p1_f32 v2, v0, 0, 0, [m0]                                     ; C8080000
>	v_interp_p2_f32 v2, [v2], v1, 0, 0, [m0]                               ; C8090001
>	s_load_dwordx4 s[8:11], s[4:5], 0x0                                    ; C0840500
>	s_load_dwordx8 s[12:19], s[6:7], 0x0                                   ; C0C60700
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	image_sample v[2:4], 7, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[12:19], s[8:11] ; F0800700 00430202
>	s_waitcnt vmcnt(0)                                                     ; BF8C0770
>	v_log_f32_e32 v5, v3                                                   ; 7E0A4F03
>	v_mov_b32_e32 v6, 0x400ccccd                                           ; 7E0C02FF 400CCCCD
>	v_mul_legacy_f32_e32 v5, v6, v5                                        ; 0E0A0B06
>	v_exp_f32_e32 v5, v5                                                   ; 7E0A4B05
>	v_interp_p1_f32 v7, v0, 0, 1, [m0]                                     ; C81C0400
>	v_interp_p2_f32 v7, [v7], v1, 0, 1, [m0]                               ; C81D0401
>	v_mul_f32_e32 v0, v7, v5                                               ; 10000B07
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                                     ; C0800300
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	s_buffer_load_dword s4, s[0:3], 0x1                                    ; C2020101
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mul_f32_e32 v0, s4, v0                                               ; 10000004
>	v_log_f32_e64 v0, |v0|                                                 ; D34E0100 00000100
>	v_mul_f32_e32 v0, 0x3ee8ba1f, v0                                       ; 100000FF 3EE8BA1F
>	v_exp_f32_e32 v0, v0                                                   ; 7E004B00
>	v_log_f32_e32 v1, v2                                                   ; 7E024F02
>	v_mul_legacy_f32_e32 v1, v6, v1                                        ; 0E020306
>	v_exp_f32_e32 v1, v1                                                   ; 7E024B01
>	v_mul_f32_e32 v1, v7, v1                                               ; 10020307
>	s_buffer_load_dword s4, s[0:3], 0x0                                    ; C2020100
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mul_f32_e32 v1, s4, v1                                               ; 10020204
>	v_log_f32_e64 v1, |v1|                                                 ; D34E0101 00000101
>	v_mul_f32_e32 v1, 0x3ee8ba1f, v1                                       ; 100202FF 3EE8BA1F
>	v_exp_f32_e32 v1, v1                                                   ; 7E024B01
>	v_cvt_pkrtz_f16_f32_e32 v0, v1, v0                                     ; 5E000101
>	v_log_f32_e32 v1, v4                                                   ; 7E024F04
>	v_mul_legacy_f32_e32 v1, v6, v1                                        ; 0E020306
>	v_exp_f32_e32 v1, v1                                                   ; 7E024B01
>	v_mul_f32_e32 v1, v7, v1                                               ; 10020307
>	s_buffer_load_dword s0, s[0:3], 0x2                                    ; C2000102
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	v_mul_f32_e32 v1, s0, v1                                               ; 10020200
>	v_log_f32_e64 v1, |v1|                                                 ; D34E0101 00000101
>	v_mul_f32_e32 v1, 0x3ee8ba1f, v1                                       ; 100202FF 3EE8BA1F
>	v_exp_f32_e32 v1, v1                                                   ; 7E024B01
>	v_cvt_pkrtz_f16_f32_e64 v1, v1, 1.0                                    ; D25E0001 0001E501
>	exp 15, 0, 1, 1, 1, v0, v1, v0, v1                                     ; F8001C0F 01000100
>	s_endpgm                                                               ; BF810000
>
>SGPRS: 24
>VGPRS: 8
>Code Size: 240 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL IN[2]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL OUT[2], GENERIC[10]
>DCL OUT[3], GENERIC[11]
>DCL OUT[4], GENERIC[12]
>DCL OUT[5], GENERIC[13]
>DCL CONST[0..12]
>DCL TEMP[0..6], LOCAL
>IMM[0] FLT32 {    0.0000,     0.5000,    -0.5000,     0.0000}
>IMM[1] FLT32 {    1.0000,     0.0000,     0.0000,     0.0000}
>  0: MUL TEMP[0].xyz, CONST[10], IN[2].yyyy
>  1: MOV TEMP[0].xyz, TEMP[0].xyzx
>  2: MAD TEMP[1].xyz, IN[2].xxxx, CONST[9], TEMP[0]
>  3: MOV TEMP[0].xyz, TEMP[1].xyzx
>  4: MAD TEMP[1].xyz, IN[2].zzzz, CONST[11], TEMP[0]
>  5: MOV TEMP[0].xyz, TEMP[1].xyzx
>  6: DP3 TEMP[1].x, TEMP[1].xyzz, TEMP[1].xyzz
>  7: MAX TEMP[1].x, TEMP[1].xxxx, IMM[0].xxxx
>  8: RSQ TEMP[1].x, TEMP[1].xxxx
>  9: MOV TEMP[0].w, TEMP[1].xxxx
> 10: MUL TEMP[1].xyz, TEMP[1].xxxx, TEMP[0]
> 11: MOV TEMP[1].xyz, TEMP[1].xyzx
> 12: MUL TEMP[0], CONST[10], IN[0].yyyy
> 13: MAD TEMP[0], IN[0].xxxx, CONST[9], TEMP[0]
> 14: MAD TEMP[0], IN[0].zzzz, CONST[11], TEMP[0]
> 15: ADD TEMP[0], TEMP[0], CONST[12]
> 16: MUL TEMP[2], TEMP[0].yyyy, CONST[1]
> 17: MAD TEMP[2], TEMP[0].xxxx, CONST[0], TEMP[2]
> 18: MAD TEMP[2], TEMP[0].zzzz, CONST[2], TEMP[2]
> 19: MAD TEMP[2], TEMP[0].wwww, CONST[3], TEMP[2]
> 20: RCP TEMP[3].x, TEMP[2].wwww
> 21: MOV TEMP[0].w, TEMP[3].xxxx
> 22: MUL TEMP[3].xy, TEMP[3].xxxx, TEMP[2]
> 23: MOV TEMP[3].xy, TEMP[3].xyxx
> 24: MOV TEMP[4], TEMP[2]
> 25: ADD TEMP[5].xy, TEMP[0], CONST[8].zwzw
> 26: MUL TEMP[5].zw, TEMP[5].xyxy, CONST[8].xyxy
> 27: MOV TEMP[5].zw, TEMP[5].wwzw
> 28: MUL TEMP[6].xy, TEMP[0].yyyy, CONST[5]
> 29: MOV TEMP[2].xy, TEMP[6].xyxx
> 30: MAD TEMP[6].xy, TEMP[0].xxxx, CONST[4], TEMP[2]
> 31: MOV TEMP[2].xy, TEMP[6].xyxx
> 32: MAD TEMP[6].xy, TEMP[0].zzzz, CONST[6], TEMP[2]
> 33: MOV TEMP[2].xy, TEMP[6].xyxx
> 34: MOV TEMP[6].xyz, TEMP[0].xyzx
> 35: ADD TEMP[2].xy, TEMP[2], CONST[7]
> 36: MOV TEMP[0].xy, TEMP[2].xyxx
> 37: MAD TEMP[0].xy, TEMP[0], IMM[0].yzww, IMM[0].zzzz
> 38: MOV TEMP[0].xy, TEMP[0].xyxx
> 39: MOV TEMP[5].xy, IN[1].xyxx
> 40: MOV TEMP[6].w, IMM[1].xxxx
> 41: MOV TEMP[3].zw, IMM[1].xxyx
> 42: MOV TEMP[0].zw, IMM[1].xxyx
> 43: MOV OUT[2], TEMP[5]
> 44: MOV OUT[3], TEMP[6]
> 45: MOV OUT[0], TEMP[4]
> 46: MOV OUT[1], TEMP[1]
> 47: MOV OUT[4], TEMP[3]
> 48: MOV OUT[5], TEMP[0]
> 49: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 0)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 4)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %12, i32 8)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %12, i32 12)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %12, i32 16)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %12, i32 20)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %12, i32 24)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %12, i32 28)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %12, i32 32)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %12, i32 36)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %12, i32 40)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %12, i32 44)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %12, i32 48)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %12, i32 52)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %12, i32 56)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %12, i32 60)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %12, i32 64)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %12, i32 68)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %12, i32 80)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %12, i32 84)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %12, i32 96)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %12, i32 100)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %12, i32 112)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %12, i32 116)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %12, i32 128)
>  %38 = call float @llvm.SI.load.const(<16 x i8> %12, i32 132)
>  %39 = call float @llvm.SI.load.const(<16 x i8> %12, i32 136)
>  %40 = call float @llvm.SI.load.const(<16 x i8> %12, i32 140)
>  %41 = call float @llvm.SI.load.const(<16 x i8> %12, i32 144)
>  %42 = call float @llvm.SI.load.const(<16 x i8> %12, i32 148)
>  %43 = call float @llvm.SI.load.const(<16 x i8> %12, i32 152)
>  %44 = call float @llvm.SI.load.const(<16 x i8> %12, i32 156)
>  %45 = call float @llvm.SI.load.const(<16 x i8> %12, i32 160)
>  %46 = call float @llvm.SI.load.const(<16 x i8> %12, i32 164)
>  %47 = call float @llvm.SI.load.const(<16 x i8> %12, i32 168)
>  %48 = call float @llvm.SI.load.const(<16 x i8> %12, i32 172)
>  %49 = call float @llvm.SI.load.const(<16 x i8> %12, i32 176)
>  %50 = call float @llvm.SI.load.const(<16 x i8> %12, i32 180)
>  %51 = call float @llvm.SI.load.const(<16 x i8> %12, i32 184)
>  %52 = call float @llvm.SI.load.const(<16 x i8> %12, i32 188)
>  %53 = call float @llvm.SI.load.const(<16 x i8> %12, i32 192)
>  %54 = call float @llvm.SI.load.const(<16 x i8> %12, i32 196)
>  %55 = call float @llvm.SI.load.const(<16 x i8> %12, i32 200)
>  %56 = call float @llvm.SI.load.const(<16 x i8> %12, i32 204)
>  %57 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %58 = load <16 x i8> addrspace(2)* %57, !tbaa !0
>  %59 = add i32 %5, %7
>  %60 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %58, i32 0, i32 %59)
>  %61 = extractelement <4 x float> %60, i32 0
>  %62 = extractelement <4 x float> %60, i32 1
>  %63 = extractelement <4 x float> %60, i32 2
>  %64 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %65 = load <16 x i8> addrspace(2)* %64, !tbaa !0
>  %66 = add i32 %5, %7
>  %67 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %65, i32 0, i32 %66)
>  %68 = extractelement <4 x float> %67, i32 0
>  %69 = extractelement <4 x float> %67, i32 1
>  %70 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 2
>  %71 = load <16 x i8> addrspace(2)* %70, !tbaa !0
>  %72 = add i32 %5, %7
>  %73 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %71, i32 0, i32 %72)
>  %74 = extractelement <4 x float> %73, i32 0
>  %75 = extractelement <4 x float> %73, i32 1
>  %76 = extractelement <4 x float> %73, i32 2
>  %77 = fmul float %45, %75
>  %78 = fmul float %46, %75
>  %79 = fmul float %47, %75
>  %80 = fmul float %74, %41
>  %81 = fadd float %80, %77
>  %82 = fmul float %74, %42
>  %83 = fadd float %82, %78
>  %84 = fmul float %74, %43
>  %85 = fadd float %84, %79
>  %86 = fmul float %76, %49
>  %87 = fadd float %86, %81
>  %88 = fmul float %76, %50
>  %89 = fadd float %88, %83
>  %90 = fmul float %76, %51
>  %91 = fadd float %90, %85
>  %92 = fmul float %87, %87
>  %93 = fmul float %89, %89
>  %94 = fadd float %93, %92
>  %95 = fmul float %91, %91
>  %96 = fadd float %94, %95
>  %97 = call float @llvm.maxnum.f32(float %96, float 0x3E7AD7F2A0000000)
>  %98 = call float @llvm.AMDGPU.rsq.clamped.f32(float %97)
>  %99 = fmul float %98, %87
>  %100 = fmul float %98, %89
>  %101 = fmul float %98, %91
>  %102 = fmul float %45, %62
>  %103 = fmul float %46, %62
>  %104 = fmul float %47, %62
>  %105 = fmul float %48, %62
>  %106 = fmul float %61, %41
>  %107 = fadd float %106, %102
>  %108 = fmul float %61, %42
>  %109 = fadd float %108, %103
>  %110 = fmul float %61, %43
>  %111 = fadd float %110, %104
>  %112 = fmul float %61, %44
>  %113 = fadd float %112, %105
>  %114 = fmul float %63, %49
>  %115 = fadd float %114, %107
>  %116 = fmul float %63, %50
>  %117 = fadd float %116, %109
>  %118 = fmul float %63, %51
>  %119 = fadd float %118, %111
>  %120 = fmul float %63, %52
>  %121 = fadd float %120, %113
>  %122 = fadd float %115, %53
>  %123 = fadd float %117, %54
>  %124 = fadd float %119, %55
>  %125 = fadd float %121, %56
>  %126 = fmul float %123, %17
>  %127 = fmul float %123, %18
>  %128 = fmul float %123, %19
>  %129 = fmul float %123, %20
>  %130 = fmul float %122, %13
>  %131 = fadd float %130, %126
>  %132 = fmul float %122, %14
>  %133 = fadd float %132, %127
>  %134 = fmul float %122, %15
>  %135 = fadd float %134, %128
>  %136 = fmul float %122, %16
>  %137 = fadd float %136, %129
>  %138 = fmul float %124, %21
>  %139 = fadd float %138, %131
>  %140 = fmul float %124, %22
>  %141 = fadd float %140, %133
>  %142 = fmul float %124, %23
>  %143 = fadd float %142, %135
>  %144 = fmul float %124, %24
>  %145 = fadd float %144, %137
>  %146 = fmul float %125, %25
>  %147 = fadd float %146, %139
>  %148 = fmul float %125, %26
>  %149 = fadd float %148, %141
>  %150 = fmul float %125, %27
>  %151 = fadd float %150, %143
>  %152 = fmul float %125, %28
>  %153 = fadd float %152, %145
>  %154 = fdiv float 1.000000e+00, %153
>  %155 = fmul float %154, %147
>  %156 = fmul float %154, %149
>  %157 = fadd float %122, %39
>  %158 = fadd float %123, %40
>  %159 = fmul float %157, %37
>  %160 = fmul float %158, %38
>  %161 = fmul float %123, %31
>  %162 = fmul float %123, %32
>  %163 = fmul float %122, %29
>  %164 = fadd float %163, %161
>  %165 = fmul float %122, %30
>  %166 = fadd float %165, %162
>  %167 = fmul float %124, %33
>  %168 = fadd float %167, %164
>  %169 = fmul float %124, %34
>  %170 = fadd float %169, %166
>  %171 = fadd float %168, %35
>  %172 = fadd float %170, %36
>  %173 = fmul float %171, 5.000000e-01
>  %174 = fadd float %173, -5.000000e-01
>  %175 = fmul float %172, -5.000000e-01
>  %176 = fadd float %175, -5.000000e-01
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %99, float %100, float %101, float 0.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %68, float %69, float %159, float %160)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 34, i32 0, float %122, float %123, float %124, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 35, i32 0, float %155, float %156, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 36, i32 0, float %174, float %176, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %147, float %149, float %151, float %153)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.maxnum.f32(float, float) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.AMDGPU.rsq.clamped.f32(float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                             ; 4A00000A
>	s_load_dwordx4 s[12:15], s[8:9], 0x0                  ; C0860900
>	s_load_dwordx4 s[4:7], s[8:9], 0x4                    ; C0820904
>	s_load_dwordx4 s[8:11], s[8:9], 0x8                   ; C0840908
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[8:11], 0 idxen  ; E00C2000 80020100
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                    ; C0800300
>	s_waitcnt vmcnt(0) lgkmcnt(0)                         ; BF8C0070
>	s_buffer_load_dword s8, s[0:3], 0x29                  ; C2040129
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v5, s8, v2                              ; 100A0408
>	s_buffer_load_dword s9, s[0:3], 0x25                  ; C2048125
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v5, v1, s9, v5                              ; D2820005 04141301
>	s_buffer_load_dword s10, s[0:3], 0x2d                 ; C205012D
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v5, v3, s10, v5                             ; D2820005 04141503
>	s_buffer_load_dword s11, s[0:3], 0x28                 ; C2058128
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v6, s11, v2                             ; 100C040B
>	s_buffer_load_dword s16, s[0:3], 0x24                 ; C2080124
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v6, v1, s16, v6                             ; D2820006 04182101
>	s_buffer_load_dword s17, s[0:3], 0x2c                 ; C208812C
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v6, v3, s17, v6                             ; D2820006 04182303
>	v_mul_f32_e32 v7, v6, v6                              ; 100E0D06
>	v_mad_f32 v7, v5, v5, v7                              ; D2820007 041E0B05
>	s_buffer_load_dword s18, s[0:3], 0x2a                 ; C209012A
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v8, s18, v2                             ; 10100412
>	s_buffer_load_dword s19, s[0:3], 0x26                 ; C2098126
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v8, v1, s19, v8                             ; D2820008 04202701
>	s_buffer_load_dword s20, s[0:3], 0x2e                 ; C20A012E
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v1, v3, s20, v8                             ; D2820001 04202903
>	v_mad_f32 v2, v1, v1, v7                              ; D2820002 041E0301
>	v_max_f32_e32 v2, 0x33d6bf95, v2                      ; 200404FF 33D6BF95
>	v_rsq_clamp_f32_e32 v2, v2                            ; 7E045902
>	v_mul_f32_e32 v1, v1, v2                              ; 10020501
>	v_mul_f32_e32 v3, v5, v2                              ; 10060505
>	v_mul_f32_e32 v2, v6, v2                              ; 10040506
>	v_mov_b32_e32 v4, 0                                   ; 7E080280
>	exp 15, 32, 0, 0, 0, v2, v3, v1, v4                   ; F800020F 04010302
>	buffer_load_format_xyzw v[5:8], v0, s[12:15], 0 idxen ; E00C2000 80030500
>	s_waitcnt vmcnt(0) expcnt(0)                          ; BF8C0700
>	v_mul_f32_e32 v1, s8, v6                              ; 10020C08
>	v_mad_f32 v1, v5, s9, v1                              ; D2820001 04041305
>	v_mad_f32 v1, v7, s10, v1                             ; D2820001 04041507
>	s_buffer_load_dword s8, s[0:3], 0x31                  ; C2040131
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v1, s8, v1                              ; 06020208
>	s_buffer_load_dword s8, s[0:3], 0x23                  ; C2040123
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v2, s8, v1                              ; 06040208
>	s_buffer_load_dword s8, s[0:3], 0x21                  ; C2040121
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v2, s8, v2                              ; 10040408
>	v_mul_f32_e32 v3, s11, v6                             ; 10060C0B
>	v_mad_f32 v3, v5, s16, v3                             ; D2820003 040C2105
>	v_mad_f32 v3, v7, s17, v3                             ; D2820003 040C2307
>	s_buffer_load_dword s8, s[0:3], 0x30                  ; C2040130
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v3, s8, v3                              ; 06060608
>	s_buffer_load_dword s8, s[0:3], 0x22                  ; C2040122
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v9, s8, v3                              ; 06120608
>	s_buffer_load_dword s8, s[0:3], 0x20                  ; C2040120
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v9, s8, v9                              ; 10121208
>	buffer_load_format_xyzw v[10:13], v0, s[4:7], 0 idxen ; E00C2000 80010A00
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	exp 15, 33, 0, 0, 0, v10, v11, v9, v2                 ; F800021F 02090B0A
>	v_mul_f32_e32 v0, s18, v6                             ; 10000C12
>	v_mad_f32 v0, v5, s19, v0                             ; D2820000 04002705
>	v_mad_f32 v0, v7, s20, v0                             ; D2820000 04002907
>	s_buffer_load_dword s4, s[0:3], 0x32                  ; C2020132
>	s_waitcnt expcnt(0) lgkmcnt(0)                        ; BF8C000F
>	v_add_f32_e32 v0, s4, v0                              ; 06000004
>	v_mov_b32_e32 v2, 1.0                                 ; 7E0402F2
>	exp 15, 34, 0, 0, 0, v3, v1, v0, v2                   ; F800022F 02000103
>	s_buffer_load_dword s4, s[0:3], 0x5                   ; C2020105
>	s_waitcnt expcnt(0) lgkmcnt(0)                        ; BF8C000F
>	v_mul_f32_e32 v9, s4, v1                              ; 10120204
>	s_buffer_load_dword s4, s[0:3], 0x1                   ; C2020101
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v9, v3, s4, v9                              ; D2820009 04240903
>	s_buffer_load_dword s4, s[0:3], 0x9                   ; C2020109
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v9, v0, s4, v9                              ; D2820009 04240900
>	s_buffer_load_dword s4, s[0:3], 0x2b                  ; C202012B
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v10, s4, v6                             ; 10140C04
>	s_buffer_load_dword s4, s[0:3], 0x27                  ; C2020127
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v10, v5, s4, v10                            ; D282000A 04280905
>	s_buffer_load_dword s4, s[0:3], 0x2f                  ; C202012F
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v5, v7, s4, v10                             ; D2820005 04280907
>	s_buffer_load_dword s4, s[0:3], 0x33                  ; C2020133
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v5, s4, v5                              ; 060A0A04
>	s_buffer_load_dword s4, s[0:3], 0xd                   ; C202010D
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v6, v5, s4, v9                              ; D2820006 04240905
>	s_buffer_load_dword s4, s[0:3], 0x7                   ; C2020107
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v7, s4, v1                              ; 100E0204
>	s_buffer_load_dword s4, s[0:3], 0x3                   ; C2020103
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v7, v3, s4, v7                              ; D2820007 041C0903
>	s_buffer_load_dword s4, s[0:3], 0xb                   ; C202010B
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v7, v0, s4, v7                              ; D2820007 041C0900
>	s_buffer_load_dword s4, s[0:3], 0xf                   ; C202010F
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v7, v5, s4, v7                              ; D2820007 041C0905
>	v_rcp_f32_e32 v8, v7                                  ; 7E105507
>	v_mul_f32_e32 v9, v6, v8                              ; 10121106
>	s_buffer_load_dword s4, s[0:3], 0x4                   ; C2020104
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v10, s4, v1                             ; 10140204
>	s_buffer_load_dword s4, s[0:3], 0x0                   ; C2020100
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v10, v3, s4, v10                            ; D282000A 04280903
>	s_buffer_load_dword s4, s[0:3], 0x8                   ; C2020108
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v10, v0, s4, v10                            ; D282000A 04280900
>	s_buffer_load_dword s4, s[0:3], 0xc                   ; C202010C
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v10, v5, s4, v10                            ; D282000A 04280905
>	v_mul_f32_e32 v8, v10, v8                             ; 1010110A
>	exp 15, 35, 0, 0, 0, v8, v9, v4, v2                   ; F800023F 02040908
>	s_buffer_load_dword s4, s[0:3], 0x14                  ; C2020114
>	s_waitcnt expcnt(0) lgkmcnt(0)                        ; BF8C000F
>	v_mul_f32_e32 v8, s4, v1                              ; 10100204
>	s_buffer_load_dword s4, s[0:3], 0x10                  ; C2020110
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v8, v3, s4, v8                              ; D2820008 04200903
>	s_buffer_load_dword s4, s[0:3], 0x18                  ; C2020118
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v8, v0, s4, v8                              ; D2820008 04200900
>	s_buffer_load_dword s4, s[0:3], 0x1c                  ; C202011C
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v8, s4, v8                              ; 06101004
>	v_mad_f32 v8, 0.5, v8, -0.5                           ; D2820008 03C610F0
>	s_buffer_load_dword s4, s[0:3], 0x15                  ; C2020115
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mul_f32_e32 v9, s4, v1                              ; 10120204
>	s_buffer_load_dword s4, s[0:3], 0x11                  ; C2020111
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v9, v3, s4, v9                              ; D2820009 04240903
>	s_buffer_load_dword s4, s[0:3], 0x19                  ; C2020119
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v9, v0, s4, v9                              ; D2820009 04240900
>	s_buffer_load_dword s4, s[0:3], 0x1d                  ; C202011D
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_add_f32_e32 v9, s4, v9                              ; 06121204
>	v_mad_f32 v9, -0.5, v9, -0.5                          ; D2820009 03C612F1
>	exp 15, 36, 0, 0, 0, v8, v9, v4, v2                   ; F800024F 02040908
>	s_buffer_load_dword s4, s[0:3], 0x6                   ; C2020106
>	s_waitcnt expcnt(0) lgkmcnt(0)                        ; BF8C000F
>	v_mul_f32_e32 v1, s4, v1                              ; 10020204
>	s_buffer_load_dword s4, s[0:3], 0x2                   ; C2020102
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v1, v3, s4, v1                              ; D2820001 04040903
>	s_buffer_load_dword s4, s[0:3], 0xa                   ; C202010A
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v0, v0, s4, v1                              ; D2820000 04040900
>	s_buffer_load_dword s0, s[0:3], 0xe                   ; C200010E
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mad_f32 v0, v5, s0, v0                              ; D2820000 04000105
>	exp 15, 12, 0, 1, 0, v10, v6, v0, v7                  ; F80008CF 0700060A
>	s_endpgm                                              ; BF810000
>
>SGPRS: 24
>VGPRS: 16
>Code Size: 876 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL IN[1], GENERIC[10], PERSPECTIVE
>DCL IN[2], GENERIC[11], PERSPECTIVE
>DCL IN[3], GENERIC[12], PERSPECTIVE
>DCL IN[4], GENERIC[13], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SAMP[1]
>DCL SAMP[2]
>DCL SAMP[3]
>DCL SAMP[4]
>DCL SAMP[5]
>DCL SAMP[6]
>DCL SAMP[7]
>DCL SAMP[8]
>DCL SAMP[9]
>DCL CONST[0..8]
>DCL TEMP[0..9], LOCAL
>IMM[0] FLT32 {    2.2000,     1.0000,    -0.1000,     0.0000}
>IMM[1] FLT32 {    0.0000,     0.0529,     0.8460,     0.5290}
>IMM[2] FLT32 {  199.0000,     0.1000,     0.4545,     1.4427}
>IMM[3] FLT32 {    1.0000,    -1.0000,     0.5000,     0.0000}
>  0: MOV TEMP[0].xy, IN[1].xyyy
>  1: TEX TEMP[0], TEMP[0], SAMP[0], 2D
>  2: POW TEMP[1].x, TEMP[0].xxxx, IMM[0].xxxx
>  3: POW TEMP[1].y, TEMP[0].yyyy, IMM[0].xxxx
>  4: POW TEMP[1].z, TEMP[0].zzzz, IMM[0].xxxx
>  5: POW TEMP[1].w, TEMP[0].wwww, IMM[0].yyyy
>  6: MOV TEMP[0].w, TEMP[1]
>  7: ABS TEMP[2].x, TEMP[1].wwww
>  8: POW TEMP[2].x, TEMP[2].xxxx, IMM[0].xxxx
>  9: MOV TEMP[3].x, TEMP[2].xxxx
> 10: ADD TEMP[4].xyz, TEMP[2].xxxx, IMM[0].zzzz
> 11: FSLT TEMP[5].xyz, TEMP[4].xyzz, IMM[0].wwww
> 12: OR TEMP[6].x, TEMP[5].xxxx, TEMP[5].zzzz
> 13: OR TEMP[5].x, TEMP[6].xxxx, TEMP[5].yyyy
> 14: AND TEMP[5].x, TEMP[5].xxxx, IMM[0].yyyy
> 15: KILL_IF -TEMP[5].xxxx
> 16: MOV TEMP[5].xyz, IN[0].xyzz
> 17: TEX TEMP[5], TEMP[5], SAMP[2], CUBE
> 18: POW TEMP[6].x, TEMP[5].xxxx, IMM[0].xxxx
> 19: POW TEMP[6].y, TEMP[5].yyyy, IMM[0].xxxx
> 20: POW TEMP[6].z, TEMP[5].zzzz, IMM[0].xxxx
> 21: POW TEMP[6].w, TEMP[5].wwww, IMM[0].yyyy
> 22: MUL TEMP[1].xyz, TEMP[1], TEMP[6]
> 23: MOV TEMP[0].xyz, TEMP[1].xyzx
> 24: ADD TEMP[1].xyz, TEMP[0], TEMP[0]
> 25: MOV TEMP[0].xyz, TEMP[1].xyzx
> 26: ADD TEMP[1].yzw, CONST[3].xxyz, -IN[2].xxyz
> 27: MOV TEMP[3].yzw, TEMP[1].zyzw
> 28: DP3 TEMP[5].x, TEMP[1].yzww, TEMP[1].yzww
> 29: MAX TEMP[5].x, TEMP[5].xxxx, IMM[1].xxxx
> 30: RSQ TEMP[5].x, TEMP[5].xxxx
> 31: MUL TEMP[1].xyz, TEMP[5].xxxx, TEMP[1].yzww
> 32: MAD TEMP[3].yzw, TEMP[3], TEMP[5].xxxx, IMM[1].yyzw
> 33: MOV TEMP[5].w, IMM[0].wwww
> 34: MOV TEMP[5].x, TEMP[3].yyyy
> 35: MOV TEMP[5].y, TEMP[3].zzzz
> 36: MOV TEMP[5].z, TEMP[3].wwww
> 37: DP4 TEMP[3].x, TEMP[5], TEMP[5]
> 38: RSQ TEMP[3].x, TEMP[3].xxxx
> 39: MUL TEMP[3].xyz, TEMP[5], TEMP[3].xxxx
> 40: DP3 TEMP[3].x, TEMP[3].xyzz, IN[0].xyzz
> 41: MOV_SAT TEMP[3].x, TEMP[3].xxxx
> 42: DP3 TEMP[5].x, IN[0].xyzz, TEMP[1].xyzz
> 43: ADD TEMP[5].y, TEMP[5].xxxx, TEMP[5].xxxx
> 44: MAD TEMP[5].yzw, TEMP[5].yyyy, IN[0].xxyz, -TEMP[1].xxyz
> 45: MOV TEMP[6].xyz, TEMP[5].yzww
> 46: TEX TEMP[6], TEMP[6], SAMP[4], CUBE
> 47: POW TEMP[7].x, TEMP[6].xxxx, IMM[0].xxxx
> 48: POW TEMP[7].y, TEMP[6].yyyy, IMM[0].xxxx
> 49: POW TEMP[7].z, TEMP[6].zzzz, IMM[0].xxxx
> 50: POW TEMP[7].w, TEMP[6].wwww, IMM[0].yyyy
> 51: MOV TEMP[4].w, TEMP[7].wwww
> 52: MOV TEMP[6].xy, IN[1].xyyy
> 53: TEX TEMP[6], TEMP[6], SAMP[1], 2D
> 54: POW TEMP[8].x, TEMP[6].xxxx, IMM[0].xxxx
> 55: POW TEMP[8].y, TEMP[6].yyyy, IMM[0].xxxx
> 56: POW TEMP[8].z, TEMP[6].zzzz, IMM[0].xxxx
> 57: POW TEMP[8].w, TEMP[6].wwww, IMM[0].yyyy
> 58: ABS TEMP[6].x, TEMP[8].wwww
> 59: POW TEMP[6].x, TEMP[6].xxxx, IMM[0].xxxx
> 60: MOV TEMP[5].xyz, TEMP[5].yzww
> 61: TEX TEMP[5], TEMP[5], SAMP[3], CUBE
> 62: POW TEMP[9].x, TEMP[5].xxxx, IMM[0].xxxx
> 63: POW TEMP[9].y, TEMP[5].yyyy, IMM[0].xxxx
> 64: POW TEMP[9].z, TEMP[5].zzzz, IMM[0].xxxx
> 65: POW TEMP[9].w, TEMP[5].wwww, IMM[0].yyyy
> 66: LRP TEMP[5].xyz, TEMP[6].xxxx, TEMP[9], TEMP[7]
> 67: MAD TEMP[6].y, TEMP[6].xxxx, IMM[2].xxxx, IMM[0].yyyy
> 68: ABS TEMP[3].x, TEMP[3].xxxx
> 69: POW TEMP[3].x, TEMP[3].xxxx, TEMP[6].yyyy
> 70: MUL TEMP[6].w, TEMP[6].yyyy, IMM[2].yyyy
> 71: MOV TEMP[0].w, TEMP[6].wwww
> 72: MAD TEMP[3].yzw, TEMP[6].wwww, TEMP[3].xxxx, TEMP[5].xxyz
> 73: MAD TEMP[3].xyz, TEMP[3].yzww, TEMP[8], TEMP[0]
> 74: ABS TEMP[5].x, TEMP[3].xxxx
> 75: LG2 TEMP[4].x, TEMP[5].xxxx
> 76: ABS TEMP[5].x, TEMP[3].yyyy
> 77: LG2 TEMP[5].x, TEMP[5].xxxx
> 78: MOV TEMP[4].y, TEMP[5].xxxx
> 79: ABS TEMP[3].x, TEMP[3].zzzz
> 80: LG2 TEMP[3].x, TEMP[3].xxxx
> 81: MOV TEMP[4].z, TEMP[3].xxxx
> 82: MUL TEMP[3].xyz, TEMP[4], IMM[2].zzzz
> 83: EX2 TEMP[4].x, TEMP[3].xxxx
> 84: EX2 TEMP[5].x, TEMP[3].yyyy
> 85: MOV TEMP[4].y, TEMP[5].xxxx
> 86: EX2 TEMP[3].x, TEMP[3].zzzz
> 87: MOV TEMP[4].z, TEMP[3].xxxx
> 88: MOV TEMP[3].xyz, TEMP[4].xyzz
> 89: TEX TEMP[3].xyz, TEMP[3], SAMP[6], 3D
> 90: MAD TEMP[5].xy, IN[4], IMM[3].xyxx, IMM[0].wyww
> 91: MOV TEMP[5].xy, TEMP[5].xyyy
> 92: TEX TEMP[5].xzw, TEMP[5], SAMP[7], 2D
> 93: MOV TEMP[1].w, TEMP[5].wwww
> 94: MOV TEMP[6].xy, IN[1].zwww
> 95: TEX TEMP[6].xyz, TEMP[6], SAMP[8], 2D
> 96: LRP TEMP[3].yzw, TEMP[5].xxxx, TEMP[3].xxyz, TEMP[6].xxyz
> 97: MAD TEMP[5].x, TEMP[5].zzzz, -TEMP[5].xxxx, TEMP[5].xxxx
> 98: MOV TEMP[6].xyz, TEMP[3].yzww
> 99: TEX TEMP[6], TEMP[6], SAMP[9], 3D
>100: LRP TEMP[3].xyz, TEMP[5].xxxx, TEMP[6], TEMP[3].yzww
>101: MOV TEMP[1].xyz, TEMP[3].xyzx
>102: ADD TEMP[3].xyz, -TEMP[1], CONST[5]
>103: MOV TEMP[0].xyz, TEMP[3].xyzx
>104: MUL TEMP[3].z, CONST[8].xxxx, IN[2].zzzz
>105: MOV TEMP[4].z, TEMP[3].zzzz
>106: MOV TEMP[4].xy, IN[2].xyxx
>107: ADD TEMP[3].yzw, TEMP[4].xxyz, -CONST[6].xxyz
>108: MUL TEMP[5].w, TEMP[3].wwww, CONST[4].xxxx
>109: MUL TEMP[5].w, TEMP[5].wwww, IMM[2].wwww
>110: EX2 TEMP[5].x, TEMP[5].wwww
>111: ADD TEMP[5].w, -TEMP[5].xxxx, IMM[0].yyyy
>112: DP3 TEMP[6].x, TEMP[3].yzww, TEMP[3].yzww
>113: RCP TEMP[3].x, TEMP[3].wwww
>114: MUL TEMP[6].y, TEMP[6].xxxx, CONST[4].yyyy
>115: MUL TEMP[5].w, TEMP[5].wwww, TEMP[6].yyyy
>116: MUL TEMP[3].w, TEMP[3].xxxx, TEMP[5].wwww
>117: MUL TEMP[3].w, TEMP[3].wwww, IMM[2].wwww
>118: EX2 TEMP[3].x, TEMP[3].wwww
>119: MOV_SAT TEMP[3].x, TEMP[3].xxxx
>120: ADD TEMP[3].w, -TEMP[3].xxxx, IMM[0].yyyy
>121: MAD TEMP[5].x, IN[3].yyyy, IMM[3].zzzz, IMM[3].zzzz
>122: MOV TEMP[4].x, TEMP[5].xxxx
>123: MOV TEMP[4].y, CONST[4].wwww
>124: MOV TEMP[4].xy, TEMP[4].xyyy
>125: TEX TEMP[4].x, TEMP[4], SAMP[5], 2D
>126: MUL TEMP[3].w, TEMP[3].wwww, TEMP[4].xxxx
>127: MOV TEMP[0].w, TEMP[3].wwww
>128: MAD TEMP[0].xyz, TEMP[3].wwww, TEMP[0], TEMP[1]
>129: MOV TEMP[0].xyz, TEMP[0].xyzx
>130: MUL TEMP[1].x, TEMP[2].xxxx, IMM[3].zzzz
>131: ADD TEMP[3].y, -CONST[7].xxxx, IN[2].zzzz
>132: FSGE TEMP[3].x, TEMP[3].yyyy, IMM[0].wwww
>133: UIF TEMP[3].xxxx :0
>134:   MOV TEMP[2].x, TEMP[2].xxxx
>135: ELSE :0
>136:   MOV TEMP[2].x, TEMP[1].xxxx
>137: ENDIF
>138: MOV TEMP[0].w, TEMP[2].xxxx
>139: MOV OUT[0], TEMP[0]
>140: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %23 = load <16 x i8> addrspace(2)* %22, !tbaa !0
>  %24 = call float @llvm.SI.load.const(<16 x i8> %23, i32 48)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %23, i32 52)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %23, i32 56)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %23, i32 64)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %23, i32 68)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %23, i32 76)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %23, i32 80)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %23, i32 84)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %23, i32 88)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %23, i32 96)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %23, i32 100)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %23, i32 104)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %23, i32 112)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %23, i32 128)
>  %38 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %39 = load <8 x i32> addrspace(2)* %38, !tbaa !0
>  %40 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %41 = load <4 x i32> addrspace(2)* %40, !tbaa !0
>  %42 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 1
>  %43 = load <8 x i32> addrspace(2)* %42, !tbaa !0
>  %44 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 1
>  %45 = load <4 x i32> addrspace(2)* %44, !tbaa !0
>  %46 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 2
>  %47 = load <8 x i32> addrspace(2)* %46, !tbaa !0
>  %48 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 2
>  %49 = load <4 x i32> addrspace(2)* %48, !tbaa !0
>  %50 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 3
>  %51 = load <8 x i32> addrspace(2)* %50, !tbaa !0
>  %52 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 3
>  %53 = load <4 x i32> addrspace(2)* %52, !tbaa !0
>  %54 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 4
>  %55 = load <8 x i32> addrspace(2)* %54, !tbaa !0
>  %56 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 4
>  %57 = load <4 x i32> addrspace(2)* %56, !tbaa !0
>  %58 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 5
>  %59 = load <8 x i32> addrspace(2)* %58, !tbaa !0
>  %60 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 5
>  %61 = load <4 x i32> addrspace(2)* %60, !tbaa !0
>  %62 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 6
>  %63 = load <8 x i32> addrspace(2)* %62, !tbaa !0
>  %64 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 6
>  %65 = load <4 x i32> addrspace(2)* %64, !tbaa !0
>  %66 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 7
>  %67 = load <8 x i32> addrspace(2)* %66, !tbaa !0
>  %68 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 7
>  %69 = load <4 x i32> addrspace(2)* %68, !tbaa !0
>  %70 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 8
>  %71 = load <8 x i32> addrspace(2)* %70, !tbaa !0
>  %72 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 8
>  %73 = load <4 x i32> addrspace(2)* %72, !tbaa !0
>  %74 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 9
>  %75 = load <8 x i32> addrspace(2)* %74, !tbaa !0
>  %76 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 9
>  %77 = load <4 x i32> addrspace(2)* %76, !tbaa !0
>  %78 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %79 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %80 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %5, <2 x i32> %7)
>  %81 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %5, <2 x i32> %7)
>  %82 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %5, <2 x i32> %7)
>  %83 = call float @llvm.SI.fs.interp(i32 2, i32 1, i32 %5, <2 x i32> %7)
>  %84 = call float @llvm.SI.fs.interp(i32 3, i32 1, i32 %5, <2 x i32> %7)
>  %85 = call float @llvm.SI.fs.interp(i32 0, i32 2, i32 %5, <2 x i32> %7)
>  %86 = call float @llvm.SI.fs.interp(i32 1, i32 2, i32 %5, <2 x i32> %7)
>  %87 = call float @llvm.SI.fs.interp(i32 2, i32 2, i32 %5, <2 x i32> %7)
>  %88 = call float @llvm.SI.fs.interp(i32 1, i32 3, i32 %5, <2 x i32> %7)
>  %89 = call float @llvm.SI.fs.interp(i32 0, i32 4, i32 %5, <2 x i32> %7)
>  %90 = call float @llvm.SI.fs.interp(i32 1, i32 4, i32 %5, <2 x i32> %7)
>  %91 = bitcast float %81 to i32
>  %92 = bitcast float %82 to i32
>  %93 = insertelement <2 x i32> undef, i32 %91, i32 0
>  %94 = insertelement <2 x i32> %93, i32 %92, i32 1
>  %95 = bitcast <8 x i32> %39 to <32 x i8>
>  %96 = bitcast <4 x i32> %41 to <16 x i8>
>  %97 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %94, <32 x i8> %95, <16 x i8> %96, i32 2)
>  %98 = extractelement <4 x float> %97, i32 0
>  %99 = extractelement <4 x float> %97, i32 1
>  %100 = extractelement <4 x float> %97, i32 2
>  %101 = extractelement <4 x float> %97, i32 3
>  %102 = call float @llvm.pow.f32(float %98, float 0x40019999A0000000)
>  %103 = call float @llvm.pow.f32(float %99, float 0x40019999A0000000)
>  %104 = call float @llvm.pow.f32(float %100, float 0x40019999A0000000)
>  %105 = call float @llvm.pow.f32(float %101, float 1.000000e+00)
>  %106 = call float @fabs(float %105)
>  %107 = call float @llvm.pow.f32(float %106, float 0x40019999A0000000)
>  %108 = fadd float %107, 0xBFB99999A0000000
>  %109 = fadd float %107, 0xBFB99999A0000000
>  %110 = fadd float %107, 0xBFB99999A0000000
>  %111 = fcmp olt float %108, 0.000000e+00
>  %112 = sext i1 %111 to i32
>  %113 = fcmp olt float %109, 0.000000e+00
>  %114 = sext i1 %113 to i32
>  %115 = fcmp olt float %110, 0.000000e+00
>  %116 = sext i1 %115 to i32
>  %117 = bitcast i32 %112 to float
>  %118 = bitcast i32 %114 to float
>  %119 = bitcast i32 %116 to float
>  %120 = bitcast float %117 to i32
>  %121 = bitcast float %119 to i32
>  %122 = or i32 %120, %121
>  %123 = bitcast i32 %122 to float
>  %124 = bitcast float %123 to i32
>  %125 = bitcast float %118 to i32
>  %126 = or i32 %124, %125
>  %127 = bitcast i32 %126 to float
>  %128 = bitcast float %127 to i32
>  %129 = and i32 %128, 1065353216
>  %130 = bitcast i32 %129 to float
>  %131 = fsub float -0.000000e+00, %130
>  %132 = fsub float -0.000000e+00, %130
>  %133 = fsub float -0.000000e+00, %130
>  %134 = fsub float -0.000000e+00, %130
>  call void @llvm.AMDGPU.kill(float %131)
>  call void @llvm.AMDGPU.kill(float %132)
>  call void @llvm.AMDGPU.kill(float %133)
>  call void @llvm.AMDGPU.kill(float %134)
>  %135 = insertelement <4 x float> undef, float %78, i32 0
>  %136 = insertelement <4 x float> %135, float %79, i32 1
>  %137 = insertelement <4 x float> %136, float %80, i32 2
>  %138 = insertelement <4 x float> %137, float 0.000000e+00, i32 3
>  %139 = call <4 x float> @llvm.AMDGPU.cube(<4 x float> %138)
>  %140 = extractelement <4 x float> %139, i32 0
>  %141 = extractelement <4 x float> %139, i32 1
>  %142 = extractelement <4 x float> %139, i32 2
>  %143 = extractelement <4 x float> %139, i32 3
>  %144 = call float @fabs(float %142)
>  %145 = fdiv float 1.000000e+00, %144
>  %146 = fmul float %140, %145
>  %147 = fadd float %146, 1.500000e+00
>  %148 = fmul float %141, %145
>  %149 = fadd float %148, 1.500000e+00
>  %150 = bitcast float %149 to i32
>  %151 = bitcast float %147 to i32
>  %152 = bitcast float %143 to i32
>  %153 = insertelement <4 x i32> undef, i32 %150, i32 0
>  %154 = insertelement <4 x i32> %153, i32 %151, i32 1
>  %155 = insertelement <4 x i32> %154, i32 %152, i32 2
>  %156 = insertelement <4 x i32> %155, i32 undef, i32 3
>  %157 = bitcast <8 x i32> %47 to <32 x i8>
>  %158 = bitcast <4 x i32> %49 to <16 x i8>
>  %159 = call <4 x float> @llvm.SI.sample.v4i32(<4 x i32> %156, <32 x i8> %157, <16 x i8> %158, i32 4)
>  %160 = extractelement <4 x float> %159, i32 0
>  %161 = extractelement <4 x float> %159, i32 1
>  %162 = extractelement <4 x float> %159, i32 2
>  %163 = extractelement <4 x float> %159, i32 3
>  %164 = call float @llvm.pow.f32(float %160, float 0x40019999A0000000)
>  %165 = call float @llvm.pow.f32(float %161, float 0x40019999A0000000)
>  %166 = call float @llvm.pow.f32(float %162, float 0x40019999A0000000)
>  %167 = call float @llvm.pow.f32(float %163, float 1.000000e+00)
>  %168 = fmul float %102, %164
>  %169 = fmul float %103, %165
>  %170 = fmul float %104, %166
>  %171 = fadd float %168, %168
>  %172 = fadd float %169, %169
>  %173 = fadd float %170, %170
>  %174 = fsub float -0.000000e+00, %85
>  %175 = fadd float %24, %174
>  %176 = fsub float -0.000000e+00, %86
>  %177 = fadd float %25, %176
>  %178 = fsub float -0.000000e+00, %87
>  %179 = fadd float %26, %178
>  %180 = fmul float %175, %175
>  %181 = fmul float %177, %177
>  %182 = fadd float %181, %180
>  %183 = fmul float %179, %179
>  %184 = fadd float %182, %183
>  %185 = call float @llvm.maxnum.f32(float %184, float 0x3E7AD7F2A0000000)
>  %186 = call float @llvm.AMDGPU.rsq.clamped.f32(float %185)
>  %187 = fmul float %186, %175
>  %188 = fmul float %186, %177
>  %189 = fmul float %186, %179
>  %190 = fmul float %175, %186
>  %191 = fadd float %190, 0x3FAB15B580000000
>  %192 = fmul float %177, %186
>  %193 = fadd float %192, 0x3FEB126EA0000000
>  %194 = fmul float %179, %186
>  %195 = fadd float %194, 0x3FE0ED9160000000
>  %196 = fmul float %191, %191
>  %197 = fmul float %193, %193
>  %198 = fadd float %196, %197
>  %199 = fmul float %195, %195
>  %200 = fadd float %198, %199
>  %201 = fmul float 0.000000e+00, 0.000000e+00
>  %202 = fadd float %200, %201
>  %203 = call float @llvm.AMDGPU.rsq.clamped.f32(float %202)
>  %204 = fmul float %191, %203
>  %205 = fmul float %193, %203
>  %206 = fmul float %195, %203
>  %207 = fmul float %204, %78
>  %208 = fmul float %205, %79
>  %209 = fadd float %208, %207
>  %210 = fmul float %206, %80
>  %211 = fadd float %209, %210
>  %212 = call float @llvm.AMDIL.clamp.(float %211, float 0.000000e+00, float 1.000000e+00)
>  %213 = fmul float %78, %187
>  %214 = fmul float %79, %188
>  %215 = fadd float %214, %213
>  %216 = fmul float %80, %189
>  %217 = fadd float %215, %216
>  %218 = fadd float %217, %217
>  %219 = fsub float -0.000000e+00, %187
>  %220 = fmul float %218, %78
>  %221 = fadd float %220, %219
>  %222 = fsub float -0.000000e+00, %188
>  %223 = fmul float %218, %79
>  %224 = fadd float %223, %222
>  %225 = fsub float -0.000000e+00, %189
>  %226 = fmul float %218, %80
>  %227 = fadd float %226, %225
>  %228 = insertelement <4 x float> undef, float %221, i32 0
>  %229 = insertelement <4 x float> %228, float %224, i32 1
>  %230 = insertelement <4 x float> %229, float %227, i32 2
>  %231 = insertelement <4 x float> %230, float %167, i32 3
>  %232 = call <4 x float> @llvm.AMDGPU.cube(<4 x float> %231)
>  %233 = extractelement <4 x float> %232, i32 0
>  %234 = extractelement <4 x float> %232, i32 1
>  %235 = extractelement <4 x float> %232, i32 2
>  %236 = extractelement <4 x float> %232, i32 3
>  %237 = call float @fabs(float %235)
>  %238 = fdiv float 1.000000e+00, %237
>  %239 = fmul float %233, %238
>  %240 = fadd float %239, 1.500000e+00
>  %241 = fmul float %234, %238
>  %242 = fadd float %241, 1.500000e+00
>  %243 = bitcast float %242 to i32
>  %244 = bitcast float %240 to i32
>  %245 = bitcast float %236 to i32
>  %246 = insertelement <4 x i32> undef, i32 %243, i32 0
>  %247 = insertelement <4 x i32> %246, i32 %244, i32 1
>  %248 = insertelement <4 x i32> %247, i32 %245, i32 2
>  %249 = insertelement <4 x i32> %248, i32 undef, i32 3
>  %250 = bitcast <8 x i32> %55 to <32 x i8>
>  %251 = bitcast <4 x i32> %57 to <16 x i8>
>  %252 = call <4 x float> @llvm.SI.sample.v4i32(<4 x i32> %249, <32 x i8> %250, <16 x i8> %251, i32 4)
>  %253 = extractelement <4 x float> %252, i32 0
>  %254 = extractelement <4 x float> %252, i32 1
>  %255 = extractelement <4 x float> %252, i32 2
>  %256 = call float @llvm.pow.f32(float %253, float 0x40019999A0000000)
>  %257 = call float @llvm.pow.f32(float %254, float 0x40019999A0000000)
>  %258 = call float @llvm.pow.f32(float %255, float 0x40019999A0000000)
>  %259 = bitcast float %81 to i32
>  %260 = bitcast float %82 to i32
>  %261 = insertelement <2 x i32> undef, i32 %259, i32 0
>  %262 = insertelement <2 x i32> %261, i32 %260, i32 1
>  %263 = bitcast <8 x i32> %43 to <32 x i8>
>  %264 = bitcast <4 x i32> %45 to <16 x i8>
>  %265 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %262, <32 x i8> %263, <16 x i8> %264, i32 2)
>  %266 = extractelement <4 x float> %265, i32 0
>  %267 = extractelement <4 x float> %265, i32 1
>  %268 = extractelement <4 x float> %265, i32 2
>  %269 = extractelement <4 x float> %265, i32 3
>  %270 = call float @llvm.pow.f32(float %266, float 0x40019999A0000000)
>  %271 = call float @llvm.pow.f32(float %267, float 0x40019999A0000000)
>  %272 = call float @llvm.pow.f32(float %268, float 0x40019999A0000000)
>  %273 = call float @llvm.pow.f32(float %269, float 1.000000e+00)
>  %274 = call float @fabs(float %273)
>  %275 = call float @llvm.pow.f32(float %274, float 0x40019999A0000000)
>  %276 = insertelement <4 x float> undef, float %221, i32 0
>  %277 = insertelement <4 x float> %276, float %224, i32 1
>  %278 = insertelement <4 x float> %277, float %227, i32 2
>  %279 = insertelement <4 x float> %278, float %227, i32 3
>  %280 = call <4 x float> @llvm.AMDGPU.cube(<4 x float> %279)
>  %281 = extractelement <4 x float> %280, i32 0
>  %282 = extractelement <4 x float> %280, i32 1
>  %283 = extractelement <4 x float> %280, i32 2
>  %284 = extractelement <4 x float> %280, i32 3
>  %285 = call float @fabs(float %283)
>  %286 = fdiv float 1.000000e+00, %285
>  %287 = fmul float %281, %286
>  %288 = fadd float %287, 1.500000e+00
>  %289 = fmul float %282, %286
>  %290 = fadd float %289, 1.500000e+00
>  %291 = bitcast float %290 to i32
>  %292 = bitcast float %288 to i32
>  %293 = bitcast float %284 to i32
>  %294 = insertelement <4 x i32> undef, i32 %291, i32 0
>  %295 = insertelement <4 x i32> %294, i32 %292, i32 1
>  %296 = insertelement <4 x i32> %295, i32 %293, i32 2
>  %297 = insertelement <4 x i32> %296, i32 undef, i32 3
>  %298 = bitcast <8 x i32> %51 to <32 x i8>
>  %299 = bitcast <4 x i32> %53 to <16 x i8>
>  %300 = call <4 x float> @llvm.SI.sample.v4i32(<4 x i32> %297, <32 x i8> %298, <16 x i8> %299, i32 4)
>  %301 = extractelement <4 x float> %300, i32 0
>  %302 = extractelement <4 x float> %300, i32 1
>  %303 = extractelement <4 x float> %300, i32 2
>  %304 = call float @llvm.pow.f32(float %301, float 0x40019999A0000000)
>  %305 = call float @llvm.pow.f32(float %302, float 0x40019999A0000000)
>  %306 = call float @llvm.pow.f32(float %303, float 0x40019999A0000000)
>  %307 = call float @llvm.AMDGPU.lrp(float %275, float %304, float %256)
>  %308 = call float @llvm.AMDGPU.lrp(float %275, float %305, float %257)
>  %309 = call float @llvm.AMDGPU.lrp(float %275, float %306, float %258)
>  %310 = fmul float %275, 1.990000e+02
>  %311 = fadd float %310, 1.000000e+00
>  %312 = call float @fabs(float %212)
>  %313 = call float @llvm.pow.f32(float %312, float %311)
>  %314 = fmul float %311, 0x3FB99999A0000000
>  %315 = fmul float %314, %313
>  %316 = fadd float %315, %307
>  %317 = fmul float %314, %313
>  %318 = fadd float %317, %308
>  %319 = fmul float %314, %313
>  %320 = fadd float %319, %309
>  %321 = fmul float %316, %270
>  %322 = fadd float %321, %171
>  %323 = fmul float %318, %271
>  %324 = fadd float %323, %172
>  %325 = fmul float %320, %272
>  %326 = fadd float %325, %173
>  %327 = call float @fabs(float %322)
>  %328 = call float @llvm.log2.f32(float %327)
>  %329 = call float @fabs(float %324)
>  %330 = call float @llvm.log2.f32(float %329)
>  %331 = call float @fabs(float %326)
>  %332 = call float @llvm.log2.f32(float %331)
>  %333 = fmul float %328, 0x3FDD1743E0000000
>  %334 = fmul float %330, 0x3FDD1743E0000000
>  %335 = fmul float %332, 0x3FDD1743E0000000
>  %336 = call float @llvm.AMDIL.exp.(float %333)
>  %337 = call float @llvm.AMDIL.exp.(float %334)
>  %338 = call float @llvm.AMDIL.exp.(float %335)
>  %339 = bitcast float %336 to i32
>  %340 = bitcast float %337 to i32
>  %341 = bitcast float %338 to i32
>  %342 = insertelement <4 x i32> undef, i32 %339, i32 0
>  %343 = insertelement <4 x i32> %342, i32 %340, i32 1
>  %344 = insertelement <4 x i32> %343, i32 %341, i32 2
>  %345 = insertelement <4 x i32> %344, i32 undef, i32 3
>  %346 = bitcast <8 x i32> %63 to <32 x i8>
>  %347 = bitcast <4 x i32> %65 to <16 x i8>
>  %348 = call <4 x float> @llvm.SI.sample.v4i32(<4 x i32> %345, <32 x i8> %346, <16 x i8> %347, i32 3)
>  %349 = extractelement <4 x float> %348, i32 0
>  %350 = extractelement <4 x float> %348, i32 1
>  %351 = extractelement <4 x float> %348, i32 2
>  %352 = fmul float %89, 1.000000e+00
>  %353 = fadd float %352, 0.000000e+00
>  %354 = fmul float %90, -1.000000e+00
>  %355 = fadd float %354, 1.000000e+00
>  %356 = bitcast float %353 to i32
>  %357 = bitcast float %355 to i32
>  %358 = insertelement <2 x i32> undef, i32 %356, i32 0
>  %359 = insertelement <2 x i32> %358, i32 %357, i32 1
>  %360 = bitcast <8 x i32> %67 to <32 x i8>
>  %361 = bitcast <4 x i32> %69 to <16 x i8>
>  %362 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %359, <32 x i8> %360, <16 x i8> %361, i32 2)
>  %363 = extractelement <4 x float> %362, i32 0
>  %364 = extractelement <4 x float> %362, i32 2
>  %365 = bitcast float %83 to i32
>  %366 = bitcast float %84 to i32
>  %367 = insertelement <2 x i32> undef, i32 %365, i32 0
>  %368 = insertelement <2 x i32> %367, i32 %366, i32 1
>  %369 = bitcast <8 x i32> %71 to <32 x i8>
>  %370 = bitcast <4 x i32> %73 to <16 x i8>
>  %371 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %368, <32 x i8> %369, <16 x i8> %370, i32 2)
>  %372 = extractelement <4 x float> %371, i32 0
>  %373 = extractelement <4 x float> %371, i32 1
>  %374 = extractelement <4 x float> %371, i32 2
>  %375 = call float @llvm.AMDGPU.lrp(float %363, float %349, float %372)
>  %376 = call float @llvm.AMDGPU.lrp(float %363, float %350, float %373)
>  %377 = call float @llvm.AMDGPU.lrp(float %363, float %351, float %374)
>  %378 = fsub float -0.000000e+00, %363
>  %379 = fmul float %364, %378
>  %380 = fadd float %379, %363
>  %381 = bitcast float %375 to i32
>  %382 = bitcast float %376 to i32
>  %383 = bitcast float %377 to i32
>  %384 = insertelement <4 x i32> undef, i32 %381, i32 0
>  %385 = insertelement <4 x i32> %384, i32 %382, i32 1
>  %386 = insertelement <4 x i32> %385, i32 %383, i32 2
>  %387 = insertelement <4 x i32> %386, i32 undef, i32 3
>  %388 = bitcast <8 x i32> %75 to <32 x i8>
>  %389 = bitcast <4 x i32> %77 to <16 x i8>
>  %390 = call <4 x float> @llvm.SI.sample.v4i32(<4 x i32> %387, <32 x i8> %388, <16 x i8> %389, i32 3)
>  %391 = extractelement <4 x float> %390, i32 0
>  %392 = extractelement <4 x float> %390, i32 1
>  %393 = extractelement <4 x float> %390, i32 2
>  %394 = call float @llvm.AMDGPU.lrp(float %380, float %391, float %375)
>  %395 = call float @llvm.AMDGPU.lrp(float %380, float %392, float %376)
>  %396 = call float @llvm.AMDGPU.lrp(float %380, float %393, float %377)
>  %397 = fsub float -0.000000e+00, %394
>  %398 = fadd float %397, %30
>  %399 = fsub float -0.000000e+00, %395
>  %400 = fadd float %399, %31
>  %401 = fsub float -0.000000e+00, %396
>  %402 = fadd float %401, %32
>  %403 = fmul float %37, %87
>  %404 = fsub float -0.000000e+00, %33
>  %405 = fadd float %85, %404
>  %406 = fsub float -0.000000e+00, %34
>  %407 = fadd float %86, %406
>  %408 = fsub float -0.000000e+00, %35
>  %409 = fadd float %403, %408
>  %410 = fmul float %409, %27
>  %411 = fmul float %410, 0x3FF7154CA0000000
>  %412 = call float @llvm.AMDIL.exp.(float %411)
>  %413 = fsub float -0.000000e+00, %412
>  %414 = fadd float %413, 1.000000e+00
>  %415 = fmul float %405, %405
>  %416 = fmul float %407, %407
>  %417 = fadd float %416, %415
>  %418 = fmul float %409, %409
>  %419 = fadd float %417, %418
>  %420 = fdiv float 1.000000e+00, %409
>  %421 = fmul float %419, %28
>  %422 = fmul float %414, %421
>  %423 = fmul float %420, %422
>  %424 = fmul float %423, 0x3FF7154CA0000000
>  %425 = call float @llvm.AMDIL.exp.(float %424)
>  %426 = call float @llvm.AMDIL.clamp.(float %425, float 0.000000e+00, float 1.000000e+00)
>  %427 = fsub float -0.000000e+00, %426
>  %428 = fadd float %427, 1.000000e+00
>  %429 = fmul float %88, 5.000000e-01
>  %430 = fadd float %429, 5.000000e-01
>  %431 = bitcast float %430 to i32
>  %432 = bitcast float %29 to i32
>  %433 = insertelement <2 x i32> undef, i32 %431, i32 0
>  %434 = insertelement <2 x i32> %433, i32 %432, i32 1
>  %435 = bitcast <8 x i32> %59 to <32 x i8>
>  %436 = bitcast <4 x i32> %61 to <16 x i8>
>  %437 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %434, <32 x i8> %435, <16 x i8> %436, i32 2)
>  %438 = extractelement <4 x float> %437, i32 0
>  %439 = fmul float %428, %438
>  %440 = fmul float %439, %398
>  %441 = fadd float %440, %394
>  %442 = fmul float %439, %400
>  %443 = fadd float %442, %395
>  %444 = fmul float %439, %402
>  %445 = fadd float %444, %396
>  %446 = fmul float %107, 5.000000e-01
>  %447 = fsub float -0.000000e+00, %36
>  %448 = fadd float %447, %87
>  %449 = fcmp oge float %448, 0.000000e+00
>  %450 = sext i1 %449 to i32
>  %451 = bitcast i32 %450 to float
>  %452 = bitcast float %451 to i32
>  %453 = icmp ne i32 %452, 0
>  %. = select i1 %453, float %107, float %446
>  %454 = call i32 @llvm.SI.packf16(float %441, float %443)
>  %455 = bitcast i32 %454 to float
>  %456 = call i32 @llvm.SI.packf16(float %445, float %.)
>  %457 = bitcast i32 %456 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %455, float %457, float %455, float %457)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.pow.f32(float, float) #1
>
>; Function Attrs: readnone
>declare float @fabs(float) #2
>
>declare void @llvm.AMDGPU.kill(float)
>
>; Function Attrs: readnone
>declare <4 x float> @llvm.AMDGPU.cube(<4 x float>) #2
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v4i32(<4 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.maxnum.f32(float, float) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.AMDGPU.rsq.clamped.f32(float) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.clamp.(float, float, float) #2
>
>; Function Attrs: readnone
>declare float @llvm.AMDGPU.lrp(float, float, float) #2
>
>; Function Attrs: nounwind readnone
>declare float @llvm.log2.f32(float) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.exp.(float) #2
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                         ; BEFE0A7E
>	s_mov_b64 vcc, s[6:7]                                                        ; BEEA0406
>	s_mov_b64 s[100:101], s[2:3]                                                 ; BEE40402
>	s_mov_b32 m0, s9                                                             ; BEFC0309
>	v_interp_p1_f32 v3, v0, 1, 1, [m0]                                           ; C80C0500
>	v_interp_p2_f32 v3, [v3], v1, 1, 1, [m0]                                     ; C80D0501
>	v_interp_p1_f32 v2, v0, 0, 1, [m0]                                           ; C8080400
>	v_interp_p2_f32 v2, [v2], v1, 0, 1, [m0]                                     ; C8090401
>	s_load_dwordx4 s[8:11], s[4:5], 0x0                                          ; C0840500
>	s_load_dwordx4 s[64:67], s[4:5], 0x4                                         ; C0A00504
>	s_load_dwordx4 s[88:91], s[4:5], 0x8                                         ; C0AC0508
>	s_load_dwordx4 s[52:55], s[4:5], 0xc                                         ; C09A050C
>	s_load_dwordx4 s[68:71], s[4:5], 0x10                                        ; C0A20510
>	s_load_dwordx4 s[0:3], s[4:5], 0x14                                          ; C0800514
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_writelane_b32 v47, s0, 0                                                   ; 045F0000
>	v_writelane_b32 v47, s1, 1                                                   ; 045F0201
>	v_writelane_b32 v47, s2, 2                                                   ; 045F0402
>	v_writelane_b32 v47, s3, 3                                                   ; 045F0603
>	s_load_dwordx4 s[40:43], s[4:5], 0x18                                        ; C0940518
>	s_load_dwordx4 s[28:31], s[4:5], 0x1c                                        ; C08E051C
>	s_load_dwordx4 s[24:27], s[4:5], 0x20                                        ; C08C0520
>	s_load_dwordx4 s[20:23], s[4:5], 0x24                                        ; C08A0524
>	s_load_dwordx8 s[0:7], vcc, 0x0                                              ; C0C06B00
>	s_load_dwordx8 s[72:79], vcc, 0x8                                            ; C0E46B08
>	s_load_dwordx8 s[92:99], vcc, 0x10                                           ; C0EE6B10
>	s_load_dwordx8 s[56:63], vcc, 0x18                                           ; C0DC6B18
>	s_load_dwordx8 s[80:87], vcc, 0x20                                           ; C0E86B20
>	s_load_dwordx8 s[12:19], vcc, 0x28                                           ; C0C66B28
>	s_load_dwordx8 s[44:51], vcc, 0x30                                           ; C0D66B30
>	s_load_dwordx8 s[32:39], vcc, 0x38                                           ; C0D06B38
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample v[4:7], 15, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[0:7], s[8:11]        ; F0800F00 00400402
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_log_f32_e32 v8, v7                                                         ; 7E104F07
>	v_mul_legacy_f32_e32 v8, 1.0, v8                                             ; 0E1010F2
>	v_exp_f32_e32 v8, v8                                                         ; 7E104B08
>	v_mov_b32_e32 v9, 0x7fffffff                                                 ; 7E1202FF 7FFFFFFF
>	v_and_b32_e32 v8, v8, v9                                                     ; 36101308
>	v_log_f32_e32 v8, v8                                                         ; 7E104F08
>	v_mov_b32_e32 v10, 0x400ccccd                                                ; 7E1402FF 400CCCCD
>	v_mul_legacy_f32_e32 v8, v10, v8                                             ; 0E10110A
>	v_exp_f32_e32 v8, v8                                                         ; 7E104B08
>	v_mov_b32_e32 v11, 0xbdcccccd                                                ; 7E1602FF BDCCCCCD
>	v_add_f32_e32 v11, v8, v11                                                   ; 06161708
>	v_mov_b32_e32 v15, 0                                                         ; 7E1E0280
>	v_cmp_lt_f32_e64 s[0:1], v11, 0                                              ; D0020000 0001010B
>	v_cndmask_b32_e64 v11, 0, -1, s[0:1]                                         ; D200080B 00018280
>	v_and_b32_e32 v11, 1.0, v11                                                  ; 361616F2
>	v_xor_b32_e32 v11, 0x80000000, v11                                           ; 3A1616FF 80000000
>	v_cmpx_le_f32_e32 vcc, 0, v11                                                ; 7C261680
>	v_cmpx_le_f32_e32 vcc, 0, v11                                                ; 7C261680
>	v_cmpx_le_f32_e32 vcc, 0, v11                                                ; 7C261680
>	v_cmpx_le_f32_e32 vcc, 0, v11                                                ; 7C261680
>	v_interp_p1_f32 v14, v0, 2, 0, [m0]                                          ; C8380200
>	v_interp_p2_f32 v14, [v14], v1, 2, 0, [m0]                                   ; C8390201
>	v_interp_p1_f32 v13, v0, 1, 0, [m0]                                          ; C8340100
>	v_interp_p2_f32 v13, [v13], v1, 1, 0, [m0]                                   ; C8350101
>	v_interp_p1_f32 v12, v0, 0, 0, [m0]                                          ; C8300000
>	v_interp_p2_f32 v12, [v12], v1, 0, 0, [m0]                                   ; C8310001
>	v_cubeid_f32 v19, v12, v13, v14                                              ; D2880013 043A1B0C
>	v_cubema_f32 v18, v12, v13, v14                                              ; D28E0012 043A1B0C
>	v_cubesc_f32 v17, v12, v13, v14                                              ; D28A0011 043A1B0C
>	v_cubetc_f32 v16, v12, v13, v14                                              ; D28C0010 043A1B0C
>	v_rcp_f32_e64 v11, |v18|                                                     ; D354010B 00000112
>	v_mov_b32_e32 v24, 0x3fc00000                                                ; 7E3002FF 3FC00000
>	v_mad_f32 v18, v16, v11, v24                                                 ; D2820012 04621710
>	v_mad_f32 v17, v17, v11, v24                                                 ; D2820011 04621711
>	image_sample v[16:19], 15, 0, 0, 0, 0, 0, 0, 0, v[17:20], s[92:99], s[88:91] ; F0800F00 02D71011
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_log_f32_e32 v11, v19                                                       ; 7E164F13
>	v_mul_legacy_f32_e32 v11, 1.0, v11                                           ; 0E1616F2
>	v_exp_f32_e32 v23, v11                                                       ; 7E2E4B0B
>	v_interp_p1_f32 v11, v0, 1, 2, [m0]                                          ; C82C0900
>	v_interp_p2_f32 v11, [v11], v1, 1, 2, [m0]                                   ; C82D0901
>	s_load_dwordx4 s[0:3], s[100:101], 0x0                                       ; C0806500
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	s_buffer_load_dword s4, s[0:3], 0xd                                          ; C202010D
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v25, s4, v11                                                   ; 08321604
>	v_interp_p1_f32 v26, v0, 0, 2, [m0]                                          ; C8680800
>	v_interp_p2_f32 v26, [v26], v1, 0, 2, [m0]                                   ; C8690801
>	s_buffer_load_dword s4, s[0:3], 0xc                                          ; C202010C
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v27, s4, v26                                                   ; 08363404
>	v_mul_f32_e32 v28, v27, v27                                                  ; 1038371B
>	v_mad_f32 v28, v25, v25, v28                                                 ; D282001C 04723319
>	v_interp_p1_f32 v29, v0, 2, 2, [m0]                                          ; C8740A00
>	v_interp_p2_f32 v29, [v29], v1, 2, 2, [m0]                                   ; C8750A01
>	s_buffer_load_dword s4, s[0:3], 0xe                                          ; C202010E
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v30, s4, v29                                                   ; 083C3A04
>	v_mad_f32 v28, v30, v30, v28                                                 ; D282001C 04723D1E
>	v_max_f32_e32 v28, 0x33d6bf95, v28                                           ; 203838FF 33D6BF95
>	v_rsq_clamp_f32_e32 v28, v28                                                 ; 7E38591C
>	v_mul_f32_e32 v31, v25, v28                                                  ; 103E3919
>	v_mul_f32_e32 v32, v27, v28                                                  ; 1040391B
>	v_mul_f32_e32 v33, v32, v12                                                  ; 10421920
>	v_mad_f32 v33, v13, v31, v33                                                 ; D2820021 04863F0D
>	v_mul_f32_e32 v34, v30, v28                                                  ; 1044391E
>	v_mad_f32 v33, v14, v34, v33                                                 ; D2820021 0486450E
>	v_add_f32_e32 v33, v33, v33                                                  ; 06424321
>	v_mad_f32 v22, v33, v14, -v34                                                ; D2820016 848A1D21
>	v_mad_f32 v21, v33, v13, -v31                                                ; D2820015 847E1B21
>	v_mad_f32 v20, v33, v12, -v32                                                ; D2820014 84821921
>	v_cubeid_f32 v34, v20, v21, v22                                              ; D2880022 045A2B14
>	v_cubema_f32 v33, v20, v21, v22                                              ; D28E0021 045A2B14
>	v_cubesc_f32 v32, v20, v21, v22                                              ; D28A0020 045A2B14
>	v_cubetc_f32 v31, v20, v21, v22                                              ; D28C001F 045A2B14
>	v_rcp_f32_e64 v39, |v33|                                                     ; D3540127 00000121
>	v_mad_f32 v33, v31, v39, v24                                                 ; D2820021 04624F1F
>	v_mad_f32 v32, v32, v39, v24                                                 ; D2820020 04624F20
>	image_sample v[31:33], 7, 0, 0, 0, 0, 0, 0, 0, v[32:35], s[80:87], s[68:71]  ; F0800700 02341F20
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_log_f32_e32 v34, v33                                                       ; 7E444F21
>	v_mul_legacy_f32_e32 v34, v10, v34                                           ; 0E44450A
>	v_exp_f32_e32 v34, v34                                                       ; 7E444B22
>	image_sample v[35:38], 15, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[72:79], s[64:67]   ; F0800F00 02122302
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_log_f32_e32 v2, v38                                                        ; 7E044F26
>	v_mul_legacy_f32_e32 v2, 1.0, v2                                             ; 0E0404F2
>	v_exp_f32_e32 v2, v2                                                         ; 7E044B02
>	v_and_b32_e32 v2, v2, v9                                                     ; 36041302
>	v_log_f32_e32 v2, v2                                                         ; 7E044F02
>	v_mul_legacy_f32_e32 v2, v10, v2                                             ; 0E04050A
>	v_exp_f32_e32 v2, v2                                                         ; 7E044B02
>	v_sub_f32_e32 v3, 1.0, v2                                                    ; 080604F2
>	v_mul_f32_e32 v34, v34, v3                                                   ; 10440722
>	v_mov_b32_e32 v23, v22                                                       ; 7E2E0316
>	v_cubeid_f32 v42, v20, v21, v22                                              ; D288002A 045A2B14
>	v_cubema_f32 v41, v20, v21, v22                                              ; D28E0029 045A2B14
>	v_cubesc_f32 v40, v20, v21, v22                                              ; D28A0028 045A2B14
>	v_cubetc_f32 v39, v20, v21, v22                                              ; D28C0027 045A2B14
>	v_rcp_f32_e64 v20, |v41|                                                     ; D3540114 00000129
>	v_mad_f32 v41, v39, v20, v24                                                 ; D2820029 04622927
>	v_mad_f32 v40, v40, v20, v24                                                 ; D2820028 04622928
>	image_sample v[20:22], 7, 0, 0, 0, 0, 0, 0, 0, v[40:43], s[56:63], s[52:55]  ; F0800700 01AE1428
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_log_f32_e32 v23, v22                                                       ; 7E2E4F16
>	v_mul_legacy_f32_e32 v23, v10, v23                                           ; 0E2E2F0A
>	v_exp_f32_e32 v23, v23                                                       ; 7E2E4B17
>	v_mad_f32 v23, v2, v23, v34                                                  ; D2820017 048A2F02
>	v_mov_b32_e32 v24, 0x3d58adac                                                ; 7E3002FF 3D58ADAC
>	v_mad_f32 v24, v27, v28, v24                                                 ; D2820018 0462391B
>	v_mov_b32_e32 v27, 0x3f589375                                                ; 7E3602FF 3F589375
>	v_mad_f32 v25, v25, v28, v27                                                 ; D2820019 046E3919
>	v_mul_f32_e32 v27, v25, v25                                                  ; 10363319
>	v_mad_f32 v27, v24, v24, v27                                                 ; D282001B 046E3118
>	v_mov_b32_e32 v34, 0x3f076c8b                                                ; 7E4402FF 3F076C8B
>	v_mad_f32 v28, v30, v28, v34                                                 ; D282001C 048A391E
>	v_mad_f32 v27, v28, v28, v27                                                 ; D282001B 046E391C
>	v_add_f32_e32 v27, 0, v27                                                    ; 06363680
>	v_rsq_clamp_f32_e32 v27, v27                                                 ; 7E36591B
>	v_mul_f32_e32 v25, v27, v25                                                  ; 1032331B
>	v_mul_f32_e32 v24, v27, v24                                                  ; 1030311B
>	v_mul_f32_e32 v24, v12, v24                                                  ; 1030310C
>	v_mad_f32 v24, v25, v13, v24                                                 ; D2820018 04621B19
>	v_mul_f32_e32 v25, v27, v28                                                  ; 1032391B
>	v_mad_f32 v12, v25, v14, v24                                                 ; D282000C 04621D19
>	v_add_f32_e64 v12, 0, v12 clamp                                              ; D206080C 00021880
>	v_and_b32_e32 v9, v12, v9                                                    ; 3612130C
>	v_log_f32_e32 v9, v9                                                         ; 7E124F09
>	v_mov_b32_e32 v12, 0x43470000                                                ; 7E1802FF 43470000
>	v_mad_f32 v12, v12, v2, 1.0                                                  ; D282000C 03CA050C
>	v_mul_legacy_f32_e32 v9, v12, v9                                             ; 0E12130C
>	v_exp_f32_e32 v9, v9                                                         ; 7E124B09
>	v_mul_f32_e32 v12, 0x3dcccccd, v12                                           ; 101818FF 3DCCCCCD
>	v_mad_f32 v13, v12, v9, v23                                                  ; D282000D 045E130C
>	v_log_f32_e32 v14, v37                                                       ; 7E1C4F25
>	v_mul_legacy_f32_e32 v14, v10, v14                                           ; 0E1C1D0A
>	v_exp_f32_e32 v14, v14                                                       ; 7E1C4B0E
>	v_mul_f32_e32 v13, v14, v13                                                  ; 101A1B0E
>	v_log_f32_e32 v14, v18                                                       ; 7E1C4F12
>	v_mul_legacy_f32_e32 v14, v10, v14                                           ; 0E1C1D0A
>	v_exp_f32_e32 v14, v14                                                       ; 7E1C4B0E
>	v_log_f32_e32 v15, v6                                                        ; 7E1E4F06
>	v_mul_legacy_f32_e32 v15, v10, v15                                           ; 0E1E1F0A
>	v_exp_f32_e32 v15, v15                                                       ; 7E1E4B0F
>	v_mul_f32_e32 v14, v14, v15                                                  ; 101C1F0E
>	v_mad_f32 v13, 2.0, v14, v13                                                 ; D282000D 04361CF4
>	v_log_f32_e64 v13, |v13|                                                     ; D34E010D 0000010D
>	v_mul_f32_e32 v13, 0x3ee8ba1f, v13                                           ; 101A1AFF 3EE8BA1F
>	v_exp_f32_e32 v41, v13                                                       ; 7E524B0D
>	v_log_f32_e32 v13, v32                                                       ; 7E1A4F20
>	v_mul_legacy_f32_e32 v13, v10, v13                                           ; 0E1A1B0A
>	v_exp_f32_e32 v13, v13                                                       ; 7E1A4B0D
>	v_mul_f32_e32 v13, v13, v3                                                   ; 101A070D
>	v_log_f32_e32 v14, v21                                                       ; 7E1C4F15
>	v_mul_legacy_f32_e32 v14, v10, v14                                           ; 0E1C1D0A
>	v_exp_f32_e32 v14, v14                                                       ; 7E1C4B0E
>	v_mad_f32 v13, v2, v14, v13                                                  ; D282000D 04361D02
>	v_mad_f32 v13, v12, v9, v13                                                  ; D282000D 0436130C
>	v_log_f32_e32 v14, v36                                                       ; 7E1C4F24
>	v_mul_legacy_f32_e32 v14, v10, v14                                           ; 0E1C1D0A
>	v_exp_f32_e32 v14, v14                                                       ; 7E1C4B0E
>	v_mul_f32_e32 v13, v14, v13                                                  ; 101A1B0E
>	v_log_f32_e32 v14, v17                                                       ; 7E1C4F11
>	v_mul_legacy_f32_e32 v14, v10, v14                                           ; 0E1C1D0A
>	v_exp_f32_e32 v14, v14                                                       ; 7E1C4B0E
>	v_log_f32_e32 v15, v5                                                        ; 7E1E4F05
>	v_mul_legacy_f32_e32 v15, v10, v15                                           ; 0E1E1F0A
>	v_exp_f32_e32 v15, v15                                                       ; 7E1E4B0F
>	v_mul_f32_e32 v14, v14, v15                                                  ; 101C1F0E
>	v_mad_f32 v13, 2.0, v14, v13                                                 ; D282000D 04361CF4
>	v_log_f32_e64 v13, |v13|                                                     ; D34E010D 0000010D
>	v_mul_f32_e32 v13, 0x3ee8ba1f, v13                                           ; 101A1AFF 3EE8BA1F
>	v_exp_f32_e32 v40, v13                                                       ; 7E504B0D
>	v_log_f32_e32 v13, v31                                                       ; 7E1A4F1F
>	v_mul_legacy_f32_e32 v13, v10, v13                                           ; 0E1A1B0A
>	v_exp_f32_e32 v13, v13                                                       ; 7E1A4B0D
>	v_mul_f32_e32 v3, v13, v3                                                    ; 1006070D
>	v_log_f32_e32 v13, v20                                                       ; 7E1A4F14
>	v_mul_legacy_f32_e32 v13, v10, v13                                           ; 0E1A1B0A
>	v_exp_f32_e32 v13, v13                                                       ; 7E1A4B0D
>	v_mad_f32 v2, v2, v13, v3                                                    ; D2820002 040E1B02
>	v_mad_f32 v2, v12, v9, v2                                                    ; D2820002 040A130C
>	v_log_f32_e32 v3, v35                                                        ; 7E064F23
>	v_mul_legacy_f32_e32 v3, v10, v3                                             ; 0E06070A
>	v_exp_f32_e32 v3, v3                                                         ; 7E064B03
>	v_mul_f32_e32 v2, v3, v2                                                     ; 10040503
>	v_log_f32_e32 v3, v16                                                        ; 7E064F10
>	v_mul_legacy_f32_e32 v3, v10, v3                                             ; 0E06070A
>	v_exp_f32_e32 v3, v3                                                         ; 7E064B03
>	v_log_f32_e32 v4, v4                                                         ; 7E084F04
>	v_mul_legacy_f32_e32 v4, v10, v4                                             ; 0E08090A
>	v_exp_f32_e32 v4, v4                                                         ; 7E084B04
>	v_mul_f32_e32 v3, v3, v4                                                     ; 10060903
>	v_mad_f32 v2, 2.0, v3, v2                                                    ; D2820002 040A06F4
>	v_log_f32_e64 v2, |v2|                                                       ; D34E0102 00000102
>	v_mul_f32_e32 v2, 0x3ee8ba1f, v2                                             ; 100404FF 3EE8BA1F
>	v_exp_f32_e32 v39, v2                                                        ; 7E4E4B02
>	image_sample v[2:4], 7, 0, 0, 0, 0, 0, 0, 0, v[39:42], s[44:51], s[40:43]    ; F0800700 014B0227
>	v_interp_p1_f32 v5, v0, 1, 4, [m0]                                           ; C8141100
>	v_interp_p2_f32 v5, [v5], v1, 1, 4, [m0]                                     ; C8151101
>	v_sub_f32_e32 v6, 1.0, v5                                                    ; 080C0AF2
>	v_interp_p1_f32 v7, v0, 0, 4, [m0]                                           ; C81C1000
>	v_interp_p2_f32 v7, [v7], v1, 0, 4, [m0]                                     ; C81D1001
>	v_add_f32_e32 v5, 0, v7                                                      ; 060A0E80
>	image_sample v[5:6], 5, 0, 0, 0, 0, 0, 0, 0, v[5:6], s[32:39], s[28:31]      ; F0800500 00E80505
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_sub_f32_e32 v7, 1.0, v5                                                    ; 080E0AF2
>	v_interp_p1_f32 v10, v0, 3, 1, [m0]                                          ; C8280700
>	v_interp_p2_f32 v10, [v10], v1, 3, 1, [m0]                                   ; C8290701
>	v_interp_p1_f32 v9, v0, 2, 1, [m0]                                           ; C8240600
>	v_interp_p2_f32 v9, [v9], v1, 2, 1, [m0]                                     ; C8250601
>	s_load_dwordx8 s[4:11], vcc, 0x40                                            ; C0C26B40
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample v[12:14], 7, 0, 0, 0, 0, 0, 0, 0, v[9:10], s[4:11], s[24:27]    ; F0800700 00C10C09
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mul_f32_e32 v9, v14, v7                                                    ; 10120F0E
>	v_mad_f32 v17, v5, v4, v9                                                    ; D2820011 04260905
>	v_mul_f32_e32 v9, v13, v7                                                    ; 10120F0D
>	v_mad_f32 v16, v5, v3, v9                                                    ; D2820010 04260705
>	v_mul_f32_e32 v7, v12, v7                                                    ; 100E0F0C
>	v_mad_f32 v15, v5, v2, v7                                                    ; D282000F 041E0505
>	s_load_dwordx8 s[4:11], vcc, 0x48                                            ; C0C26B48
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample v[2:4], 7, 0, 0, 0, 0, 0, 0, 0, v[15:18], s[4:11], s[20:23]     ; F0800700 00A1020F
>	v_mad_f32 v5, -v6, v5, v5                                                    ; D2820005 24160B06
>	v_sub_f32_e32 v6, 1.0, v5                                                    ; 080C0AF2
>	v_mul_f32_e32 v7, v16, v6                                                    ; 100E0D10
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v7, v5, v3, v7                                                     ; D2820007 041E0705
>	s_buffer_load_dword s4, s[0:3], 0x15                                         ; C2020115
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v9, s4, v7                                                     ; 08120E04
>	s_buffer_load_dword s4, s[0:3], 0x19                                         ; C2020119
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_subrev_f32_e32 v10, s4, v11                                                ; 0A141604
>	s_buffer_load_dword s4, s[0:3], 0x18                                         ; C2020118
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_subrev_f32_e32 v11, s4, v26                                                ; 0A163404
>	v_mul_f32_e32 v11, v11, v11                                                  ; 1016170B
>	v_mad_f32 v10, v10, v10, v11                                                 ; D282000A 042E150A
>	s_buffer_load_dword s4, s[0:3], 0x1a                                         ; C202011A
>	s_buffer_load_dword s5, s[0:3], 0x20                                         ; C2028120
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mov_b32_e32 v11, s4                                                        ; 7E160204
>	v_mad_f32 v11, v29, s5, -v11                                                 ; D282000B 842C0B1D
>	v_mad_f32 v10, v11, v11, v10                                                 ; D282000A 042A170B
>	s_buffer_load_dword s4, s[0:3], 0x11                                         ; C2020111
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v10, s4, v10                                                   ; 10141404
>	s_buffer_load_dword s4, s[0:3], 0x10                                         ; C2020110
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v12, s4, v11                                                   ; 10181604
>	v_mul_f32_e32 v12, 0x3fb8aa65, v12                                           ; 101818FF 3FB8AA65
>	v_exp_f32_e32 v12, v12                                                       ; 7E184B0C
>	v_sub_f32_e32 v12, 1.0, v12                                                  ; 081818F2
>	v_mul_f32_e32 v10, v10, v12                                                  ; 1014190A
>	v_rcp_f32_e32 v11, v11                                                       ; 7E16550B
>	v_mul_f32_e32 v10, v10, v11                                                  ; 1014170A
>	v_mul_f32_e32 v10, 0x3fb8aa65, v10                                           ; 101414FF 3FB8AA65
>	v_exp_f32_e32 v10, v10                                                       ; 7E144B0A
>	v_add_f32_e64 v10, 0, v10 clamp                                              ; D206080A 00021480
>	v_sub_f32_e32 v10, 1.0, v10                                                  ; 081414F2
>	v_interp_p1_f32 v11, v0, 1, 3, [m0]                                          ; C82C0D00
>	v_interp_p2_f32 v11, [v11], v1, 1, 3, [m0]                                   ; C82D0D01
>	v_mad_f32 v0, 0.5, v11, 0.5                                                  ; D2820000 03C216F0
>	s_buffer_load_dword s4, s[0:3], 0x13                                         ; C2020113
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mov_b32_e32 v1, s4                                                         ; 7E020204
>	v_readlane_b32 s4, v47, 0                                                    ; 0209012F
>	v_readlane_b32 s5, v47, 1                                                    ; 020B032F
>	v_readlane_b32 s6, v47, 2                                                    ; 020D052F
>	v_readlane_b32 s7, v47, 3                                                    ; 020F072F
>	s_nop 2                                                                      ; BF800002
>	image_sample v0, 1, 0, 0, 0, 0, 0, 0, 0, v[0:1], s[12:19], s[4:7]            ; F0800100 00230000
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mul_f32_e32 v0, v0, v10                                                    ; 10001500
>	v_mad_f32 v1, v0, v9, v7                                                     ; D2820001 041E1300
>	v_mul_f32_e32 v7, v15, v6                                                    ; 100E0D0F
>	v_mad_f32 v7, v5, v2, v7                                                     ; D2820007 041E0505
>	s_buffer_load_dword s4, s[0:3], 0x14                                         ; C2020114
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v9, s4, v7                                                     ; 08120E04
>	v_mad_f32 v7, v0, v9, v7                                                     ; D2820007 041E1300
>	v_cvt_pkrtz_f16_f32_e32 v1, v7, v1                                           ; 5E020307
>	v_mul_f32_e32 v6, v17, v6                                                    ; 100C0D11
>	v_mad_f32 v2, v5, v4, v6                                                     ; D2820002 041A0905
>	s_buffer_load_dword s4, s[0:3], 0x16                                         ; C2020116
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v3, s4, v2                                                     ; 08060404
>	v_mad_f32 v0, v0, v3, v2                                                     ; D2820000 040A0700
>	v_mul_f32_e32 v2, 0.5, v8                                                    ; 100410F0
>	s_buffer_load_dword s0, s[0:3], 0x1c                                         ; C200011C
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_subrev_f32_e32 v3, s0, v29                                                 ; 0A063A00
>	v_cmp_ge_f32_e64 s[0:1], v3, 0                                               ; D00C0000 00010103
>	v_cndmask_b32_e64 v3, 0, -1, s[0:1]                                          ; D2000803 00018280
>	v_cmp_ne_i32_e64 s[0:1], v3, 0                                               ; D10A0000 00010103
>	v_cndmask_b32_e64 v2, v2, v8, s[0:1]                                         ; D2000002 10021102
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v2                                           ; 5E000500
>	exp 15, 0, 1, 1, 1, v1, v0, v1, v0                                           ; F8001C0F 00010001
>	s_endpgm                                                                     ; BF810000
>
>SGPRS: 104
>VGPRS: 48
>Code Size: 1728 bytes
>LDS: 0 blocks
>Scratch: 1024 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL IN[2]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL OUT[2], GENERIC[10]
>DCL OUT[3], GENERIC[11]
>DCL OUT[4], GENERIC[12]
>DCL CONST[0..98]
>DCL TEMP[0..6], LOCAL
>DCL ADDR[0]
>IMM[0] FLT32 {    0.5000,    -0.5000,     0.0000,     1.0000}
>  0: F2I TEMP[0].x, IN[2].xxxx
>  1: UARL ADDR[0].x, TEMP[0].xxxx
>  2: UARL ADDR[0].x, TEMP[0].xxxx
>  3: MOV TEMP[1], CONST[ADDR[0].x+9]
>  4: UARL ADDR[0].x, TEMP[0].xxxx
>  5: MAD TEMP[1].xyz, IN[0], CONST[ADDR[0].x+9].wwww, TEMP[1]
>  6: MOV TEMP[2].xyz, TEMP[1].xyzx
>  7: UARL ADDR[0].x, TEMP[0].xxxx
>  8: MOV TEMP[0].xyz, CONST[ADDR[0].x+9].xyzx
>  9: MUL TEMP[3], TEMP[1].yyyy, CONST[1]
> 10: MAD TEMP[3], TEMP[1].xxxx, CONST[0], TEMP[3]
> 11: MAD TEMP[3], TEMP[1].zzzz, CONST[2], TEMP[3]
> 12: ADD TEMP[3], TEMP[3], CONST[3]
> 13: RCP TEMP[4].x, TEMP[3].wwww
> 14: MOV TEMP[2].w, TEMP[4].xxxx
> 15: MUL TEMP[4].xy, TEMP[4].xxxx, TEMP[3]
> 16: MOV TEMP[4].xy, TEMP[4].xyxx
> 17: MOV TEMP[3], TEMP[3]
> 18: ADD TEMP[5].xy, TEMP[2], CONST[8].zwzw
> 19: MUL TEMP[5].zw, TEMP[5].xyxy, CONST[8].xyxy
> 20: MOV TEMP[5].zw, TEMP[5].wwzw
> 21: MUL TEMP[6].yw, TEMP[1].yyyy, CONST[5].xxzy
> 22: MOV TEMP[2].yw, TEMP[6].wyww
> 23: MAD TEMP[6].xy, TEMP[1].xxxx, CONST[4], TEMP[2].ywzw
> 24: MOV TEMP[2].xy, TEMP[6].xyxx
> 25: MAD TEMP[1].xy, TEMP[1].zzzz, CONST[6], TEMP[2]
> 26: MOV TEMP[2].xy, TEMP[1].xyxx
> 27: ADD TEMP[1].xy, TEMP[2], CONST[7]
> 28: MOV TEMP[2].xy, TEMP[1].xyxx
> 29: MAD TEMP[1].xy, TEMP[2], IMM[0].xyzz, IMM[0].yyyy
> 30: MOV TEMP[1].xy, TEMP[1].xyxx
> 31: MOV TEMP[5].xy, IN[1].xyxx
> 32: MOV TEMP[0].w, IMM[0].wwww
> 33: MOV TEMP[1].zw, IMM[0].wwzw
> 34: MOV TEMP[4].zw, IMM[0].wwzw
> 35: MOV OUT[1], TEMP[5]
> 36: MOV OUT[0], TEMP[3]
> 37: MOV OUT[2], TEMP[0]
> 38: MOV OUT[3], TEMP[1]
> 39: MOV OUT[4], TEMP[4]
> 40: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 0)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 4)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %12, i32 8)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %12, i32 12)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %12, i32 16)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %12, i32 20)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %12, i32 24)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %12, i32 28)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %12, i32 32)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %12, i32 36)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %12, i32 40)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %12, i32 44)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %12, i32 48)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %12, i32 52)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %12, i32 56)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %12, i32 60)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %12, i32 64)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %12, i32 68)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %12, i32 80)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %12, i32 84)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %12, i32 96)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %12, i32 100)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %12, i32 112)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %12, i32 116)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %12, i32 128)
>  %38 = call float @llvm.SI.load.const(<16 x i8> %12, i32 132)
>  %39 = call float @llvm.SI.load.const(<16 x i8> %12, i32 136)
>  %40 = call float @llvm.SI.load.const(<16 x i8> %12, i32 140)
>  %41 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %42 = load <16 x i8> addrspace(2)* %41, !tbaa !0
>  %43 = add i32 %5, %7
>  %44 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %42, i32 0, i32 %43)
>  %45 = extractelement <4 x float> %44, i32 0
>  %46 = extractelement <4 x float> %44, i32 1
>  %47 = extractelement <4 x float> %44, i32 2
>  %48 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %49 = load <16 x i8> addrspace(2)* %48, !tbaa !0
>  %50 = add i32 %5, %7
>  %51 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %49, i32 0, i32 %50)
>  %52 = extractelement <4 x float> %51, i32 0
>  %53 = extractelement <4 x float> %51, i32 1
>  %54 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 2
>  %55 = load <16 x i8> addrspace(2)* %54, !tbaa !0
>  %56 = add i32 %5, %7
>  %57 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %55, i32 0, i32 %56)
>  %58 = extractelement <4 x float> %57, i32 0
>  %59 = fptosi float %58 to i32
>  %60 = bitcast i32 %59 to float
>  %61 = bitcast float %60 to i32
>  %62 = shl i32 %61, 4
>  %63 = add i32 %62, 144
>  %64 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %63)
>  %65 = shl i32 %61, 4
>  %66 = add i32 %65, 148
>  %67 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %66)
>  %68 = shl i32 %61, 4
>  %69 = add i32 %68, 152
>  %70 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %69)
>  %71 = bitcast float %60 to i32
>  %72 = shl i32 %71, 4
>  %73 = add i32 %72, 156
>  %74 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %73)
>  %75 = fmul float %45, %74
>  %76 = fadd float %75, %64
>  %77 = shl i32 %71, 4
>  %78 = add i32 %77, 156
>  %79 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %78)
>  %80 = fmul float %46, %79
>  %81 = fadd float %80, %67
>  %82 = shl i32 %71, 4
>  %83 = add i32 %82, 156
>  %84 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %83)
>  %85 = fmul float %47, %84
>  %86 = fadd float %85, %70
>  %87 = bitcast float %60 to i32
>  %88 = shl i32 %87, 4
>  %89 = add i32 %88, 144
>  %90 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %89)
>  %91 = shl i32 %87, 4
>  %92 = add i32 %91, 148
>  %93 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %92)
>  %94 = shl i32 %87, 4
>  %95 = add i32 %94, 152
>  %96 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %95)
>  %97 = fmul float %81, %17
>  %98 = fmul float %81, %18
>  %99 = fmul float %81, %19
>  %100 = fmul float %81, %20
>  %101 = fmul float %76, %13
>  %102 = fadd float %101, %97
>  %103 = fmul float %76, %14
>  %104 = fadd float %103, %98
>  %105 = fmul float %76, %15
>  %106 = fadd float %105, %99
>  %107 = fmul float %76, %16
>  %108 = fadd float %107, %100
>  %109 = fmul float %86, %21
>  %110 = fadd float %109, %102
>  %111 = fmul float %86, %22
>  %112 = fadd float %111, %104
>  %113 = fmul float %86, %23
>  %114 = fadd float %113, %106
>  %115 = fmul float %86, %24
>  %116 = fadd float %115, %108
>  %117 = fadd float %110, %25
>  %118 = fadd float %112, %26
>  %119 = fadd float %114, %27
>  %120 = fadd float %116, %28
>  %121 = fdiv float 1.000000e+00, %120
>  %122 = fmul float %121, %117
>  %123 = fmul float %121, %118
>  %124 = fadd float %76, %39
>  %125 = fadd float %81, %40
>  %126 = fmul float %124, %37
>  %127 = fmul float %125, %38
>  %128 = fmul float %81, %31
>  %129 = fmul float %81, %32
>  %130 = fmul float %76, %29
>  %131 = fadd float %130, %128
>  %132 = fmul float %76, %30
>  %133 = fadd float %132, %129
>  %134 = fmul float %86, %33
>  %135 = fadd float %134, %131
>  %136 = fmul float %86, %34
>  %137 = fadd float %136, %133
>  %138 = fadd float %135, %35
>  %139 = fadd float %137, %36
>  %140 = fmul float %138, 5.000000e-01
>  %141 = fadd float %140, -5.000000e-01
>  %142 = fmul float %139, -5.000000e-01
>  %143 = fadd float %142, -5.000000e-01
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %52, float %53, float %126, float %127)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %90, float %93, float %96, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 34, i32 0, float %141, float %143, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 35, i32 0, float %122, float %123, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %117, float %118, float %119, float %120)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                               ; 4A00000A
>	s_load_dwordx4 s[4:7], s[8:9], 0x0                      ; C0820900
>	s_load_dwordx4 s[12:15], s[8:9], 0x4                    ; C0860904
>	s_load_dwordx4 s[8:11], s[8:9], 0x8                     ; C0840908
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[8:11], 0 idxen    ; E00C2000 80020100
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	v_cvt_i32_f32_e32 v1, v1                                ; 7E021101
>	v_lshlrev_b32_e32 v1, 4, v1                             ; 34020284
>	v_add_i32_e32 v2, 0x94, v1                              ; 4A0402FF 00000094
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                      ; C0800300
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	buffer_load_dword v2, v2, s[0:3], 0 offen               ; E0301000 80000202
>	v_add_i32_e32 v3, 0x9c, v1                              ; 4A0602FF 0000009C
>	buffer_load_dword v3, v3, s[0:3], 0 offen               ; E0301000 80000303
>	buffer_load_format_xyzw v[4:7], v0, s[4:7], 0 idxen     ; E00C2000 80010400
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	v_mad_f32 v8, v5, v3, v2                                ; D2820008 040A0705
>	s_buffer_load_dword s4, s[0:3], 0x23                    ; C2020123
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	v_add_f32_e32 v9, s4, v8                                ; 06121004
>	s_buffer_load_dword s4, s[0:3], 0x21                    ; C2020121
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	v_mul_f32_e32 v9, s4, v9                                ; 10121204
>	v_add_i32_e32 v10, 0x90, v1                             ; 4A1402FF 00000090
>	buffer_load_dword v10, v10, s[0:3], 0 offen             ; E0301000 80000A0A
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	v_mad_f32 v11, v4, v3, v10                              ; D282000B 042A0704
>	s_buffer_load_dword s4, s[0:3], 0x22                    ; C2020122
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	v_add_f32_e32 v12, s4, v11                              ; 06181604
>	s_buffer_load_dword s4, s[0:3], 0x20                    ; C2020120
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	v_mul_f32_e32 v12, s4, v12                              ; 10181804
>	buffer_load_format_xyzw v[13:16], v0, s[12:15], 0 idxen ; E00C2000 80030D00
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	exp 15, 32, 0, 0, 0, v13, v14, v12, v9                  ; F800020F 090C0E0D
>	v_add_i32_e32 v0, 0x98, v1                              ; 4A0002FF 00000098
>	buffer_load_dword v0, v0, s[0:3], 0 offen               ; E0301000 80000000
>	v_mov_b32_e32 v1, 1.0                                   ; 7E0202F2
>	s_waitcnt vmcnt(0) expcnt(0)                            ; BF8C0700
>	exp 15, 33, 0, 0, 0, v10, v2, v0, v1                    ; F800021F 0100020A
>	s_buffer_load_dword s4, s[0:3], 0x14                    ; C2020114
>	s_waitcnt expcnt(0) lgkmcnt(0)                          ; BF8C000F
>	v_mul_f32_e32 v2, s4, v8                                ; 10041004
>	s_buffer_load_dword s4, s[0:3], 0x10                    ; C2020110
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	v_mad_f32 v2, v11, s4, v2                               ; D2820002 0408090B
>	v_mad_f32 v0, v6, v3, v0                                ; D2820000 04020706
>	s_buffer_load_dword s4, s[0:3], 0x18                    ; C2020118
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	v_mad_f32 v2, v0, s4, v2                                ; D2820002 04080900
>	s_buffer_load_dword s4, s[0:3], 0x1c                    ; C202011C
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	v_add_f32_e32 v2, s4, v2                                ; 06040404
>	v_mad_f32 v2, 0.5, v2, -0.5                             ; D2820002 03C604F0
>	s_buffer_load_dword s4, s[0:3], 0x15                    ; C2020115
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	v_mul_f32_e32 v3, s4, v8                                ; 10061004
>	s_buffer_load_dword s4, s[0:3], 0x11                    ; C2020111
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	v_mad_f32 v3, v11, s4, v3                               ; D2820003 040C090B
>	s_buffer_load_dword s4, s[0:3], 0x19                    ; C2020119
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	v_mad_f32 v3, v0, s4, v3                                ; D2820003 040C0900
>	s_buffer_load_dword s4, s[0:3], 0x1d                    ; C202011D
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	v_add_f32_e32 v3, s4, v3                                ; 06060604
>	v_mad_f32 v3, -0.5, v3, -0.5                            ; D2820003 03C606F1
>	v_mov_b32_e32 v4, 0                                     ; 7E080280
>	exp 15, 34, 0, 0, 0, v2, v3, v4, v1                     ; F800022F 01040302
>	s_buffer_load_dword s4, s[0:3], 0x5                     ; C2020105
>	s_waitcnt expcnt(0) lgkmcnt(0)                          ; BF8C000F
>	v_mul_f32_e32 v2, s4, v8                                ; 10041004
>	s_buffer_load_dword s4, s[0:3], 0x1                     ; C2020101
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	v_mad_f32 v2, v11, s4, v2                               ; D2820002 0408090B
>	s_buffer_load_dword s4, s[0:3], 0x9                     ; C2020109
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	v_mad_f32 v2, v0, s4, v2                                ; D2820002 04080900
>	s_buffer_load_dword s4, s[0:3], 0xd                     ; C202010D
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	v_add_f32_e32 v2, s4, v2                                ; 06040404
>	s_buffer_load_dword s4, s[0:3], 0x7                     ; C2020107
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	v_mul_f32_e32 v3, s4, v8                                ; 10061004
>	s_buffer_load_dword s4, s[0:3], 0x3                     ; C2020103
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	v_mad_f32 v3, v11, s4, v3                               ; D2820003 040C090B
>	s_buffer_load_dword s4, s[0:3], 0xb                     ; C202010B
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	v_mad_f32 v3, v0, s4, v3                                ; D2820003 040C0900
>	s_buffer_load_dword s4, s[0:3], 0xf                     ; C202010F
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	v_add_f32_e32 v3, s4, v3                                ; 06060604
>	v_rcp_f32_e32 v5, v3                                    ; 7E0A5503
>	v_mul_f32_e32 v6, v2, v5                                ; 100C0B02
>	s_buffer_load_dword s4, s[0:3], 0x4                     ; C2020104
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	v_mul_f32_e32 v7, s4, v8                                ; 100E1004
>	s_buffer_load_dword s4, s[0:3], 0x0                     ; C2020100
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	v_mad_f32 v7, v11, s4, v7                               ; D2820007 041C090B
>	s_buffer_load_dword s4, s[0:3], 0x8                     ; C2020108
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	v_mad_f32 v7, v0, s4, v7                                ; D2820007 041C0900
>	s_buffer_load_dword s4, s[0:3], 0xc                     ; C202010C
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	v_add_f32_e32 v7, s4, v7                                ; 060E0E04
>	v_mul_f32_e32 v5, v7, v5                                ; 100A0B07
>	exp 15, 35, 0, 0, 0, v5, v6, v4, v1                     ; F800023F 01040605
>	s_buffer_load_dword s4, s[0:3], 0x6                     ; C2020106
>	s_waitcnt expcnt(0) lgkmcnt(0)                          ; BF8C000F
>	v_mul_f32_e32 v1, s4, v8                                ; 10021004
>	s_buffer_load_dword s4, s[0:3], 0x2                     ; C2020102
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	v_mad_f32 v1, v11, s4, v1                               ; D2820001 0404090B
>	s_buffer_load_dword s4, s[0:3], 0xa                     ; C202010A
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	v_mad_f32 v0, v0, s4, v1                                ; D2820000 04040900
>	s_buffer_load_dword s0, s[0:3], 0xe                     ; C200010E
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	v_add_f32_e32 v0, s0, v0                                ; 06000000
>	exp 15, 12, 0, 1, 0, v7, v2, v0, v3                     ; F80008CF 03000207
>	s_endpgm                                                ; BF810000
>
>SGPRS: 24
>VGPRS: 20
>Code Size: 632 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], POSITION, LINEAR
>DCL IN[1], GENERIC[9], PERSPECTIVE
>DCL IN[2], GENERIC[10], PERSPECTIVE
>DCL IN[3], GENERIC[11], PERSPECTIVE
>DCL IN[4], GENERIC[12], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SAMP[1]
>DCL SAMP[2]
>DCL SAMP[3]
>DCL SAMP[4]
>DCL SAMP[5]
>DCL SAMP[6]
>DCL SAMP[7]
>DCL CONST[14]
>DCL CONST[0..5]
>DCL TEMP[0]
>DCL TEMP[1..10], LOCAL
>IMM[0] FLT32 {   -0.1000,     0.0000,    -1.0000,     1.0000}
>IMM[1] FLT32 {    0.0010,    -0.1471,    -0.2889,     0.4360}
>IMM[2] FLT32 {    0.6150,    -0.5150,    -0.1000,     0.5000}
>IMM[3] FLT32 {    1.0000,     1.1398,    -0.3947,    -0.5806}
>IMM[4] FLT32 {    1.0000,     2.0321,     1.4427,     0.0000}
>  0: MOV TEMP[0], IN[0]
>  1: MAD TEMP[0].y, IN[0], CONST[14].xxxx, CONST[14].yyyy
>  2: MAD TEMP[1].xyz, TEMP[0], CONST[5].xyxx, CONST[5].zwzz
>  3: MOV TEMP[2].xy, IN[1].xyyy
>  4: TEX TEMP[2], TEMP[2], SAMP[0], 2D
>  5: MOV TEMP[3].xw, TEMP[2]
>  6: ADD TEMP[4].x, TEMP[2].wwww, IMM[0].xxxx
>  7: FSGE TEMP[5].x, TEMP[4].xxxx, IMM[0].yyyy
>  8: UIF TEMP[5].xxxx :2
>  9:   MOV TEMP[5].x, IMM[0].yyyy
> 10: ELSE :2
> 11:   MOV TEMP[5].x, IMM[0].zzzz
> 12: ENDIF
> 13: MOV TEMP[5].x, TEMP[5].xxxx
> 14: FSGE TEMP[6].x, TEMP[4].xxxx, IMM[0].yyyy
> 15: UIF TEMP[6].xxxx :2
> 16:   MOV TEMP[6].x, IMM[0].yyyy
> 17: ELSE :2
> 18:   MOV TEMP[6].x, IMM[0].zzzz
> 19: ENDIF
> 20: MOV TEMP[5].y, TEMP[6].xxxx
> 21: FSGE TEMP[6].x, TEMP[4].xxxx, IMM[0].yyyy
> 22: UIF TEMP[6].xxxx :2
> 23:   MOV TEMP[6].x, IMM[0].yyyy
> 24: ELSE :2
> 25:   MOV TEMP[6].x, IMM[0].zzzz
> 26: ENDIF
> 27: MOV TEMP[5].z, TEMP[6].xxxx
> 28: FSGE TEMP[6].x, TEMP[4].xxxx, IMM[0].yyyy
> 29: UIF TEMP[6].xxxx :2
> 30: ELSE :2
> 31: ENDIF
> 32: FSLT TEMP[5].xyz, TEMP[5].xyzz, IMM[0].yyyy
> 33: OR TEMP[6].x, TEMP[5].xxxx, TEMP[5].zzzz
> 34: OR TEMP[5].x, TEMP[6].xxxx, TEMP[5].yyyy
> 35: AND TEMP[5].x, TEMP[5].xxxx, IMM[0].wwww
> 36: KILL_IF -TEMP[5].xxxx
> 37: MAD TEMP[5].xy, IN[3], IMM[0].wzww, IMM[0].ywyy
> 38: MOV TEMP[5].xy, TEMP[5].xyyy
> 39: TEX TEMP[5], TEMP[5], SAMP[5], 2D
> 40: MOV TEMP[4].z, TEMP[5]
> 41: ABS TEMP[6].x, TEMP[5]
> 42: MOV TEMP[6], -TEMP[6].xxxx
> 43: FSGE TEMP[7].x, TEMP[6].xxxx, IMM[0].yyyy
> 44: UIF TEMP[7].xxxx :2
> 45:   MOV TEMP[7].x, IMM[0].zzzz
> 46: ELSE :2
> 47:   MOV TEMP[7].x, IMM[0].yyyy
> 48: ENDIF
> 49: MOV TEMP[7].x, TEMP[7].xxxx
> 50: FSGE TEMP[8].x, TEMP[6].yyyy, IMM[0].yyyy
> 51: UIF TEMP[8].xxxx :2
> 52:   MOV TEMP[8].x, IMM[0].zzzz
> 53: ELSE :2
> 54:   MOV TEMP[8].x, IMM[0].yyyy
> 55: ENDIF
> 56: MOV TEMP[7].y, TEMP[8].xxxx
> 57: FSGE TEMP[8].x, TEMP[6].zzzz, IMM[0].yyyy
> 58: UIF TEMP[8].xxxx :2
> 59:   MOV TEMP[8].x, IMM[0].zzzz
> 60: ELSE :2
> 61:   MOV TEMP[8].x, IMM[0].yyyy
> 62: ENDIF
> 63: MOV TEMP[7].z, TEMP[8].xxxx
> 64: FSGE TEMP[6].x, TEMP[6].wwww, IMM[0].yyyy
> 65: UIF TEMP[6].xxxx :2
> 66:   MOV TEMP[6].x, IMM[0].zzzz
> 67: ELSE :2
> 68:   MOV TEMP[6].x, IMM[0].yyyy
> 69: ENDIF
> 70: MOV TEMP[7].w, TEMP[6].xxxx
> 71: MOV TEMP[6].z, TEMP[7]
> 72: FSLT TEMP[7].xyz, TEMP[7].xyzz, IMM[0].yyyy
> 73: OR TEMP[8].x, TEMP[7].xxxx, TEMP[7].zzzz
> 74: OR TEMP[7].x, TEMP[8].xxxx, TEMP[7].yyyy
> 75: AND TEMP[7].x, TEMP[7].xxxx, IMM[0].wwww
> 76: KILL_IF -TEMP[7].xxxx
> 77: RCP TEMP[6].x, CONST[0].xxxx
> 78: RCP TEMP[7].x, CONST[0].yyyy
> 79: MOV TEMP[6].y, TEMP[7].xxxx
> 80: MUL TEMP[1].yw, TEMP[6].xxzy, TEMP[1].xxzy
> 81: MOV TEMP[4].yw, TEMP[1].wyww
> 82: MUL TEMP[1].xy, IMM[1].xxxx, IN[2]
> 83: MOV TEMP[1].xy, TEMP[1].xyyy
> 84: TEX TEMP[1], TEMP[1], SAMP[1], 2D
> 85: MOV TEMP[6].xw, TEMP[1].xxxw
> 86: MAD TEMP[7].xy, TEMP[4].ywzw, IMM[0].wzww, IMM[0].ywyy
> 87: MOV TEMP[7].xy, TEMP[7].xyyy
> 88: TEX TEMP[7], TEMP[7], SAMP[2], 2D
> 89: MOV TEMP[8].w, TEMP[7].xyxw
> 90: DP3 TEMP[9].x, IMM[1].yzww, TEMP[1].xyzz
> 91: MOV TEMP[9].y, TEMP[9].xxxx
> 92: DP3 TEMP[10].x, IMM[2].xyzz, TEMP[1].xyzz
> 93: MOV TEMP[9].z, TEMP[10].xxxx
> 94: DP3 TEMP[10].x, IMM[1].yzww, TEMP[2].xyzz
> 95: MOV TEMP[6].y, TEMP[10].xxxx
> 96: DP3 TEMP[10].x, IMM[2].xyzz, TEMP[2].xyzz
> 97: MOV TEMP[6].z, TEMP[10].xxxx
> 98: LRP TEMP[1].yz, TEMP[1].wwww, TEMP[9], TEMP[6]
> 99: MOV TEMP[3].yz, TEMP[1].zyzz
>100: DP2 TEMP[6].x, IMM[3].xyyy, TEMP[3].xzzz
>101: DP3 TEMP[1].x, IMM[3].xzww, TEMP[3].xyzz
>102: MOV TEMP[6].y, TEMP[1].xxxx
>103: DP2 TEMP[1].x, IMM[4].xyyy, TEMP[3].xyyy
>104: MOV TEMP[6].z, TEMP[1].xxxx
>105: MUL TEMP[1].xyz, TEMP[7], CONST[1].xxxx
>106: MOV TEMP[3].xyz, TEMP[1].xyzx
>107: MAD TEMP[1].y, TEMP[7].wwww, -CONST[1].xxxx, -IMM[0].zzzz
>108: MOV TEMP[7].xyz, TEMP[6].xyzz
>109: TEX TEMP[7], TEMP[7], SAMP[4], 3D
>110: MAD TEMP[1].xyz, TEMP[7], TEMP[1].yyyy, TEMP[3]
>111: MOV TEMP[3].xyz, TEMP[1].xyzx
>112: MOV TEMP[1].xy, IN[1].zwww
>113: TEX TEMP[1], TEMP[1], SAMP[6], 2D
>114: LRP TEMP[1].xyz, TEMP[5].xxxx, TEMP[3], TEMP[1]
>115: MOV TEMP[8].xyz, TEMP[1].xyzx
>116: MAD TEMP[5].x, TEMP[5].zzzz, -TEMP[5].xxxx, TEMP[5].xxxx
>117: MOV TEMP[1].xyz, TEMP[1].xyzz
>118: TEX TEMP[1], TEMP[1], SAMP[7], 3D
>119: LRP TEMP[1].xyz, TEMP[5].xxxx, TEMP[1], TEMP[8]
>120: MOV TEMP[4].xyz, TEMP[1].xyzx
>121: MAD TEMP[1].x, IN[4].yyyy, IMM[2].wwww, IMM[2].wwww
>122: MOV TEMP[3].x, TEMP[1].xxxx
>123: MOV TEMP[3].y, CONST[4].wwww
>124: MOV TEMP[1].xy, TEMP[3].xyyy
>125: TEX TEMP[1].xw, TEMP[1], SAMP[3], 2D
>126: MOV TEMP[6].w, TEMP[1].wwww
>127: ADD TEMP[5].xyz, -CONST[2], IN[2]
>128: MOV TEMP[3].z, TEMP[5].xyzx
>129: DP3 TEMP[3].x, TEMP[5].xyzz, TEMP[5].xyzz
>130: MUL TEMP[7].xy, TEMP[3].xzzw, CONST[4].yxzw
>131: MUL TEMP[8].y, TEMP[7].yyyy, IMM[4].zzzz
>132: EX2 TEMP[8].x, TEMP[8].yyyy
>133: ADD TEMP[8].y, -TEMP[8].xxxx, IMM[0].wwww
>134: MUL TEMP[7].x, TEMP[8].yyyy, TEMP[7].xxxx
>135: RCP TEMP[5].x, TEMP[5].zzzz
>136: MUL TEMP[5].x, TEMP[5].xxxx, TEMP[7].xxxx
>137: MUL TEMP[5].x, TEMP[5].xxxx, IMM[4].zzzz
>138: MOV TEMP[3].x, TEMP[5].xxxx
>139: EX2 TEMP[5].x, TEMP[5].xxxx
>140: MOV_SAT TEMP[3].x, TEMP[5].xxxx
>141: ADD TEMP[3].x, -TEMP[3].xxxx, IMM[0].wwww
>142: MUL TEMP[1].x, TEMP[3].xxxx, TEMP[1].xxxx
>143: ADD TEMP[3].xyz, -TEMP[4], CONST[3]
>144: MOV TEMP[6].xyz, TEMP[3].xyzx
>145: MAD TEMP[1].xyz, TEMP[1].xxxx, TEMP[6], TEMP[4]
>146: MOV TEMP[1].xyz, TEMP[1].xyzx
>147: MOV TEMP[1].w, TEMP[2].wwww
>148: MOV OUT[0], TEMP[1]
>149: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %23 = load <16 x i8> addrspace(2)* %22, !tbaa !0
>  %24 = call float @llvm.SI.load.const(<16 x i8> %23, i32 0)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %23, i32 4)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %23, i32 16)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %23, i32 32)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %23, i32 36)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %23, i32 40)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %23, i32 48)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %23, i32 52)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %23, i32 56)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %23, i32 64)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %23, i32 68)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %23, i32 76)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %23, i32 80)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %23, i32 84)
>  %38 = call float @llvm.SI.load.const(<16 x i8> %23, i32 88)
>  %39 = call float @llvm.SI.load.const(<16 x i8> %23, i32 92)
>  %40 = call float @llvm.SI.load.const(<16 x i8> %23, i32 224)
>  %41 = call float @llvm.SI.load.const(<16 x i8> %23, i32 228)
>  %42 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %43 = load <8 x i32> addrspace(2)* %42, !tbaa !0
>  %44 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %45 = load <4 x i32> addrspace(2)* %44, !tbaa !0
>  %46 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 1
>  %47 = load <8 x i32> addrspace(2)* %46, !tbaa !0
>  %48 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 1
>  %49 = load <4 x i32> addrspace(2)* %48, !tbaa !0
>  %50 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 2
>  %51 = load <8 x i32> addrspace(2)* %50, !tbaa !0
>  %52 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 2
>  %53 = load <4 x i32> addrspace(2)* %52, !tbaa !0
>  %54 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 3
>  %55 = load <8 x i32> addrspace(2)* %54, !tbaa !0
>  %56 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 3
>  %57 = load <4 x i32> addrspace(2)* %56, !tbaa !0
>  %58 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 4
>  %59 = load <8 x i32> addrspace(2)* %58, !tbaa !0
>  %60 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 4
>  %61 = load <4 x i32> addrspace(2)* %60, !tbaa !0
>  %62 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 5
>  %63 = load <8 x i32> addrspace(2)* %62, !tbaa !0
>  %64 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 5
>  %65 = load <4 x i32> addrspace(2)* %64, !tbaa !0
>  %66 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 6
>  %67 = load <8 x i32> addrspace(2)* %66, !tbaa !0
>  %68 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 6
>  %69 = load <4 x i32> addrspace(2)* %68, !tbaa !0
>  %70 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 7
>  %71 = load <8 x i32> addrspace(2)* %70, !tbaa !0
>  %72 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 7
>  %73 = load <4 x i32> addrspace(2)* %72, !tbaa !0
>  %74 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %75 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %76 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %5, <2 x i32> %7)
>  %77 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %5, <2 x i32> %7)
>  %78 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %5, <2 x i32> %7)
>  %79 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %5, <2 x i32> %7)
>  %80 = call float @llvm.SI.fs.interp(i32 2, i32 1, i32 %5, <2 x i32> %7)
>  %81 = call float @llvm.SI.fs.interp(i32 0, i32 2, i32 %5, <2 x i32> %7)
>  %82 = call float @llvm.SI.fs.interp(i32 1, i32 2, i32 %5, <2 x i32> %7)
>  %83 = call float @llvm.SI.fs.interp(i32 1, i32 3, i32 %5, <2 x i32> %7)
>  %84 = fmul float %15, %40
>  %85 = fadd float %84, %41
>  %86 = fmul float %14, %36
>  %87 = fadd float %86, %38
>  %88 = fmul float %85, %37
>  %89 = fadd float %88, %39
>  %90 = bitcast float %74 to i32
>  %91 = bitcast float %75 to i32
>  %92 = insertelement <2 x i32> undef, i32 %90, i32 0
>  %93 = insertelement <2 x i32> %92, i32 %91, i32 1
>  %94 = bitcast <8 x i32> %43 to <32 x i8>
>  %95 = bitcast <4 x i32> %45 to <16 x i8>
>  %96 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %93, <32 x i8> %94, <16 x i8> %95, i32 2)
>  %97 = extractelement <4 x float> %96, i32 0
>  %98 = extractelement <4 x float> %96, i32 1
>  %99 = extractelement <4 x float> %96, i32 2
>  %100 = extractelement <4 x float> %96, i32 3
>  %101 = fadd float %100, 0xBFB99999A0000000
>  %102 = fcmp oge float %101, 0.000000e+00
>  %103 = sext i1 %102 to i32
>  %104 = bitcast i32 %103 to float
>  %105 = bitcast float %104 to i32
>  %106 = icmp ne i32 %105, 0
>  %. = select i1 %106, float 0.000000e+00, float -1.000000e+00
>  %107 = fcmp oge float %101, 0.000000e+00
>  %108 = sext i1 %107 to i32
>  %109 = bitcast i32 %108 to float
>  %110 = bitcast float %109 to i32
>  %111 = icmp ne i32 %110, 0
>  %temp24.0 = select i1 %111, float 0.000000e+00, float -1.000000e+00
>  %112 = fcmp oge float %101, 0.000000e+00
>  %113 = sext i1 %112 to i32
>  %114 = bitcast i32 %113 to float
>  %115 = bitcast float %114 to i32
>  %116 = icmp ne i32 %115, 0
>  %.65 = select i1 %116, float 0.000000e+00, float -1.000000e+00
>  %117 = fcmp oge float %101, 0.000000e+00
>  %118 = sext i1 %117 to i32
>  %119 = bitcast i32 %118 to float
>  %120 = bitcast float %119 to i32
>  %121 = icmp ne i32 %120, 0
>  %122 = fcmp olt float %., 0.000000e+00
>  %123 = sext i1 %122 to i32
>  %124 = fcmp olt float %temp24.0, 0.000000e+00
>  %125 = sext i1 %124 to i32
>  %126 = fcmp olt float %.65, 0.000000e+00
>  %127 = sext i1 %126 to i32
>  %128 = bitcast i32 %123 to float
>  %129 = bitcast i32 %125 to float
>  %130 = bitcast i32 %127 to float
>  %131 = bitcast float %128 to i32
>  %132 = bitcast float %130 to i32
>  %133 = or i32 %131, %132
>  %134 = bitcast i32 %133 to float
>  %135 = bitcast float %134 to i32
>  %136 = bitcast float %129 to i32
>  %137 = or i32 %135, %136
>  %138 = bitcast i32 %137 to float
>  %139 = bitcast float %138 to i32
>  %140 = and i32 %139, 1065353216
>  %141 = bitcast i32 %140 to float
>  %142 = fsub float -0.000000e+00, %141
>  %143 = fsub float -0.000000e+00, %141
>  %144 = fsub float -0.000000e+00, %141
>  %145 = fsub float -0.000000e+00, %141
>  call void @llvm.AMDGPU.kill(float %142)
>  call void @llvm.AMDGPU.kill(float %143)
>  call void @llvm.AMDGPU.kill(float %144)
>  call void @llvm.AMDGPU.kill(float %145)
>  %146 = fmul float %81, 1.000000e+00
>  %147 = fadd float %146, 0.000000e+00
>  %148 = fmul float %82, -1.000000e+00
>  %149 = fadd float %148, 1.000000e+00
>  %150 = bitcast float %147 to i32
>  %151 = bitcast float %149 to i32
>  %152 = insertelement <2 x i32> undef, i32 %150, i32 0
>  %153 = insertelement <2 x i32> %152, i32 %151, i32 1
>  %154 = bitcast <8 x i32> %63 to <32 x i8>
>  %155 = bitcast <4 x i32> %65 to <16 x i8>
>  %156 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %153, <32 x i8> %154, <16 x i8> %155, i32 2)
>  %157 = extractelement <4 x float> %156, i32 0
>  %158 = extractelement <4 x float> %156, i32 2
>  %159 = call float @fabs(float %157)
>  %160 = fsub float -0.000000e+00, %159
>  %161 = fsub float -0.000000e+00, %159
>  %162 = fsub float -0.000000e+00, %159
>  %163 = fcmp oge float %160, 0.000000e+00
>  %164 = sext i1 %163 to i32
>  %165 = bitcast i32 %164 to float
>  %166 = bitcast float %165 to i32
>  %167 = icmp ne i32 %166, 0
>  %temp28.0 = select i1 %167, float -1.000000e+00, float 0.000000e+00
>  %168 = fcmp oge float %161, 0.000000e+00
>  %169 = sext i1 %168 to i32
>  %170 = bitcast i32 %169 to float
>  %171 = bitcast float %170 to i32
>  %172 = icmp ne i32 %171, 0
>  %.66 = select i1 %172, float -1.000000e+00, float 0.000000e+00
>  %173 = fcmp oge float %162, 0.000000e+00
>  %174 = sext i1 %173 to i32
>  %175 = bitcast i32 %174 to float
>  %176 = bitcast float %175 to i32
>  %177 = icmp ne i32 %176, 0
>  %temp32.1 = select i1 %177, float -1.000000e+00, float 0.000000e+00
>  %178 = fcmp olt float %temp28.0, 0.000000e+00
>  %179 = sext i1 %178 to i32
>  %180 = fcmp olt float %.66, 0.000000e+00
>  %181 = sext i1 %180 to i32
>  %182 = fcmp olt float %temp32.1, 0.000000e+00
>  %183 = sext i1 %182 to i32
>  %184 = bitcast i32 %179 to float
>  %185 = bitcast i32 %181 to float
>  %186 = bitcast i32 %183 to float
>  %187 = bitcast float %184 to i32
>  %188 = bitcast float %186 to i32
>  %189 = or i32 %187, %188
>  %190 = bitcast i32 %189 to float
>  %191 = bitcast float %190 to i32
>  %192 = bitcast float %185 to i32
>  %193 = or i32 %191, %192
>  %194 = bitcast i32 %193 to float
>  %195 = bitcast float %194 to i32
>  %196 = and i32 %195, 1065353216
>  %197 = bitcast i32 %196 to float
>  %198 = fsub float -0.000000e+00, %197
>  %199 = fsub float -0.000000e+00, %197
>  %200 = fsub float -0.000000e+00, %197
>  %201 = fsub float -0.000000e+00, %197
>  call void @llvm.AMDGPU.kill(float %198)
>  call void @llvm.AMDGPU.kill(float %199)
>  call void @llvm.AMDGPU.kill(float %200)
>  call void @llvm.AMDGPU.kill(float %201)
>  %202 = fdiv float 1.000000e+00, %24
>  %203 = fdiv float 1.000000e+00, %25
>  %204 = fmul float %202, %87
>  %205 = fmul float %203, %89
>  %206 = fmul float 0x3F50624DE0000000, %78
>  %207 = fmul float 0x3F50624DE0000000, %79
>  %208 = bitcast float %206 to i32
>  %209 = bitcast float %207 to i32
>  %210 = insertelement <2 x i32> undef, i32 %208, i32 0
>  %211 = insertelement <2 x i32> %210, i32 %209, i32 1
>  %212 = bitcast <8 x i32> %47 to <32 x i8>
>  %213 = bitcast <4 x i32> %49 to <16 x i8>
>  %214 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %211, <32 x i8> %212, <16 x i8> %213, i32 2)
>  %215 = extractelement <4 x float> %214, i32 0
>  %216 = extractelement <4 x float> %214, i32 1
>  %217 = extractelement <4 x float> %214, i32 2
>  %218 = extractelement <4 x float> %214, i32 3
>  %219 = fmul float %204, 1.000000e+00
>  %220 = fadd float %219, 0.000000e+00
>  %221 = fmul float %205, -1.000000e+00
>  %222 = fadd float %221, 1.000000e+00
>  %223 = bitcast float %220 to i32
>  %224 = bitcast float %222 to i32
>  %225 = insertelement <2 x i32> undef, i32 %223, i32 0
>  %226 = insertelement <2 x i32> %225, i32 %224, i32 1
>  %227 = bitcast <8 x i32> %51 to <32 x i8>
>  %228 = bitcast <4 x i32> %53 to <16 x i8>
>  %229 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %226, <32 x i8> %227, <16 x i8> %228, i32 2)
>  %230 = extractelement <4 x float> %229, i32 0
>  %231 = extractelement <4 x float> %229, i32 1
>  %232 = extractelement <4 x float> %229, i32 2
>  %233 = extractelement <4 x float> %229, i32 3
>  %234 = fmul float 0xBFC2D527E0000000, %215
>  %235 = fmul float 0xBFD27CAEA0000000, %216
>  %236 = fadd float %235, %234
>  %237 = fmul float 0x3FDBE76C80000000, %217
>  %238 = fadd float %236, %237
>  %239 = fmul float 0x3FE3AE1480000000, %215
>  %240 = fmul float 0xBFE07ACC40000000, %216
>  %241 = fadd float %240, %239
>  %242 = fmul float 0xBFB99A4160000000, %217
>  %243 = fadd float %241, %242
>  %244 = fmul float 0xBFC2D527E0000000, %97
>  %245 = fmul float 0xBFD27CAEA0000000, %98
>  %246 = fadd float %245, %244
>  %247 = fmul float 0x3FDBE76C80000000, %99
>  %248 = fadd float %246, %247
>  %249 = fmul float 0x3FE3AE1480000000, %97
>  %250 = fmul float 0xBFE07ACC40000000, %98
>  %251 = fadd float %250, %249
>  %252 = fmul float 0xBFB99A4160000000, %99
>  %253 = fadd float %251, %252
>  %254 = call float @llvm.AMDGPU.lrp(float %218, float %238, float %248)
>  %255 = call float @llvm.AMDGPU.lrp(float %218, float %243, float %253)
>  %256 = fmul float 1.000000e+00, %97
>  %257 = fmul float 0x3FF23CBE60000000, %255
>  %258 = fadd float %256, %257
>  %259 = fmul float 1.000000e+00, %97
>  %260 = fmul float 0xBFD941F220000000, %254
>  %261 = fadd float %260, %259
>  %262 = fmul float 0xBFE2944680000000, %255
>  %263 = fadd float %261, %262
>  %264 = fmul float 1.000000e+00, %97
>  %265 = fmul float 0x400041C2E0000000, %254
>  %266 = fadd float %264, %265
>  %267 = fmul float %230, %26
>  %268 = fmul float %231, %26
>  %269 = fmul float %232, %26
>  %270 = fsub float -0.000000e+00, %26
>  %271 = fmul float %233, %270
>  %272 = fadd float %271, 1.000000e+00
>  %273 = bitcast float %258 to i32
>  %274 = bitcast float %263 to i32
>  %275 = bitcast float %266 to i32
>  %276 = insertelement <4 x i32> undef, i32 %273, i32 0
>  %277 = insertelement <4 x i32> %276, i32 %274, i32 1
>  %278 = insertelement <4 x i32> %277, i32 %275, i32 2
>  %279 = insertelement <4 x i32> %278, i32 undef, i32 3
>  %280 = bitcast <8 x i32> %59 to <32 x i8>
>  %281 = bitcast <4 x i32> %61 to <16 x i8>
>  %282 = call <4 x float> @llvm.SI.sample.v4i32(<4 x i32> %279, <32 x i8> %280, <16 x i8> %281, i32 3)
>  %283 = extractelement <4 x float> %282, i32 0
>  %284 = extractelement <4 x float> %282, i32 1
>  %285 = extractelement <4 x float> %282, i32 2
>  %286 = fmul float %283, %272
>  %287 = fadd float %286, %267
>  %288 = fmul float %284, %272
>  %289 = fadd float %288, %268
>  %290 = fmul float %285, %272
>  %291 = fadd float %290, %269
>  %292 = bitcast float %76 to i32
>  %293 = bitcast float %77 to i32
>  %294 = insertelement <2 x i32> undef, i32 %292, i32 0
>  %295 = insertelement <2 x i32> %294, i32 %293, i32 1
>  %296 = bitcast <8 x i32> %67 to <32 x i8>
>  %297 = bitcast <4 x i32> %69 to <16 x i8>
>  %298 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %295, <32 x i8> %296, <16 x i8> %297, i32 2)
>  %299 = extractelement <4 x float> %298, i32 0
>  %300 = extractelement <4 x float> %298, i32 1
>  %301 = extractelement <4 x float> %298, i32 2
>  %302 = call float @llvm.AMDGPU.lrp(float %157, float %287, float %299)
>  %303 = call float @llvm.AMDGPU.lrp(float %157, float %289, float %300)
>  %304 = call float @llvm.AMDGPU.lrp(float %157, float %291, float %301)
>  %305 = fsub float -0.000000e+00, %157
>  %306 = fmul float %158, %305
>  %307 = fadd float %306, %157
>  %308 = bitcast float %302 to i32
>  %309 = bitcast float %303 to i32
>  %310 = bitcast float %304 to i32
>  %311 = insertelement <4 x i32> undef, i32 %308, i32 0
>  %312 = insertelement <4 x i32> %311, i32 %309, i32 1
>  %313 = insertelement <4 x i32> %312, i32 %310, i32 2
>  %314 = insertelement <4 x i32> %313, i32 undef, i32 3
>  %315 = bitcast <8 x i32> %71 to <32 x i8>
>  %316 = bitcast <4 x i32> %73 to <16 x i8>
>  %317 = call <4 x float> @llvm.SI.sample.v4i32(<4 x i32> %314, <32 x i8> %315, <16 x i8> %316, i32 3)
>  %318 = extractelement <4 x float> %317, i32 0
>  %319 = extractelement <4 x float> %317, i32 1
>  %320 = extractelement <4 x float> %317, i32 2
>  %321 = call float @llvm.AMDGPU.lrp(float %307, float %318, float %302)
>  %322 = call float @llvm.AMDGPU.lrp(float %307, float %319, float %303)
>  %323 = call float @llvm.AMDGPU.lrp(float %307, float %320, float %304)
>  %324 = fmul float %83, 5.000000e-01
>  %325 = fadd float %324, 5.000000e-01
>  %326 = bitcast float %325 to i32
>  %327 = bitcast float %35 to i32
>  %328 = insertelement <2 x i32> undef, i32 %326, i32 0
>  %329 = insertelement <2 x i32> %328, i32 %327, i32 1
>  %330 = bitcast <8 x i32> %55 to <32 x i8>
>  %331 = bitcast <4 x i32> %57 to <16 x i8>
>  %332 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %329, <32 x i8> %330, <16 x i8> %331, i32 2)
>  %333 = extractelement <4 x float> %332, i32 0
>  %334 = fsub float -0.000000e+00, %27
>  %335 = fadd float %334, %78
>  %336 = fsub float -0.000000e+00, %28
>  %337 = fadd float %336, %79
>  %338 = fsub float -0.000000e+00, %29
>  %339 = fadd float %338, %80
>  %340 = fmul float %335, %335
>  %341 = fmul float %337, %337
>  %342 = fadd float %341, %340
>  %343 = fmul float %339, %339
>  %344 = fadd float %342, %343
>  %345 = fmul float %344, %34
>  %346 = fmul float %339, %33
>  %347 = fmul float %346, 0x3FF7154CA0000000
>  %348 = call float @llvm.AMDIL.exp.(float %347)
>  %349 = fsub float -0.000000e+00, %348
>  %350 = fadd float %349, 1.000000e+00
>  %351 = fmul float %350, %345
>  %352 = fdiv float 1.000000e+00, %339
>  %353 = fmul float %352, %351
>  %354 = fmul float %353, 0x3FF7154CA0000000
>  %355 = call float @llvm.AMDIL.exp.(float %354)
>  %356 = call float @llvm.AMDIL.clamp.(float %355, float 0.000000e+00, float 1.000000e+00)
>  %357 = fsub float -0.000000e+00, %356
>  %358 = fadd float %357, 1.000000e+00
>  %359 = fmul float %358, %333
>  %360 = fsub float -0.000000e+00, %321
>  %361 = fadd float %360, %30
>  %362 = fsub float -0.000000e+00, %322
>  %363 = fadd float %362, %31
>  %364 = fsub float -0.000000e+00, %323
>  %365 = fadd float %364, %32
>  %366 = fmul float %359, %361
>  %367 = fadd float %366, %321
>  %368 = fmul float %359, %363
>  %369 = fadd float %368, %322
>  %370 = fmul float %359, %365
>  %371 = fadd float %370, %323
>  %372 = call i32 @llvm.SI.packf16(float %367, float %369)
>  %373 = bitcast i32 %372 to float
>  %374 = call i32 @llvm.SI.packf16(float %371, float %100)
>  %375 = bitcast i32 %374 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %373, float %375, float %373, float %375)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>declare void @llvm.AMDGPU.kill(float)
>
>; Function Attrs: readnone
>declare float @fabs(float) #2
>
>; Function Attrs: readnone
>declare float @llvm.AMDGPU.lrp(float, float, float) #2
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v4i32(<4 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.exp.(float) #2
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.clamp.(float, float, float) #2
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                         ; BEFE0A7E
>	s_mov_b32 m0, s9                                                             ; BEFC0309
>	v_interp_p1_f32 v5, v0, 1, 0, [m0]                                           ; C8140100
>	v_interp_p2_f32 v5, [v5], v1, 1, 0, [m0]                                     ; C8150101
>	v_interp_p1_f32 v4, v0, 0, 0, [m0]                                           ; C8100000
>	v_interp_p2_f32 v4, [v4], v1, 0, 0, [m0]                                     ; C8110001
>	s_load_dwordx4 s[80:83], s[4:5], 0x0                                         ; C0A80500
>	s_load_dwordx4 s[68:71], s[4:5], 0x4                                         ; C0A20504
>	s_load_dwordx4 s[44:47], s[4:5], 0x8                                         ; C0960508
>	s_load_dwordx4 s[16:19], s[4:5], 0xc                                         ; C088050C
>	s_load_dwordx4 s[56:59], s[4:5], 0x10                                        ; C09C0510
>	s_load_dwordx4 s[84:87], s[4:5], 0x14                                        ; C0AA0514
>	s_load_dwordx4 s[32:35], s[4:5], 0x18                                        ; C0900518
>	s_load_dwordx4 s[20:23], s[4:5], 0x1c                                        ; C08A051C
>	s_load_dwordx8 s[88:95], s[6:7], 0x0                                         ; C0EC0700
>	s_load_dwordx8 s[72:79], s[6:7], 0x8                                         ; C0E40708
>	s_load_dwordx8 s[48:55], s[6:7], 0x10                                        ; C0D80710
>	s_load_dwordx8 s[8:15], s[6:7], 0x18                                         ; C0C40718
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_writelane_b32 v28, s8, 0                                                   ; 04390008
>	v_writelane_b32 v28, s9, 1                                                   ; 04390209
>	v_writelane_b32 v28, s10, 2                                                  ; 0439040A
>	v_writelane_b32 v28, s11, 3                                                  ; 0439060B
>	v_writelane_b32 v28, s12, 4                                                  ; 0439080C
>	v_writelane_b32 v28, s13, 5                                                  ; 04390A0D
>	v_writelane_b32 v28, s14, 6                                                  ; 04390C0E
>	v_writelane_b32 v28, s15, 7                                                  ; 04390E0F
>	s_load_dwordx8 s[60:67], s[6:7], 0x20                                        ; C0DE0720
>	s_load_dwordx8 s[8:15], s[6:7], 0x28                                         ; C0C40728
>	s_load_dwordx8 s[36:43], s[6:7], 0x30                                        ; C0D20730
>	s_load_dwordx8 s[24:31], s[6:7], 0x38                                        ; C0CC0738
>	image_sample v[4:7], 15, 0, 0, 0, 0, 0, 0, 0, v[4:5], s[88:95], s[80:83]     ; F0800F00 02960404
>	v_mov_b32_e32 v8, 0xbdcccccd                                                 ; 7E1002FF BDCCCCCD
>	s_waitcnt vmcnt(0) lgkmcnt(0)                                                ; BF8C0070
>	v_add_f32_e32 v8, v7, v8                                                     ; 06101107
>	v_cmp_ge_f32_e64 s[0:1], v8, 0                                               ; D00C0000 00010108
>	v_cndmask_b32_e64 v8, 0, -1, s[0:1]                                          ; D2000008 00018280
>	v_cmp_ne_i32_e64 s[0:1], v8, 0                                               ; D10A0000 00010108
>	v_cndmask_b32_e64 v8, -1.0, 0, s[0:1]                                        ; D2000008 180100F3
>	v_cmp_lt_f32_e64 s[0:1], v8, 0                                               ; D0020000 00010108
>	v_cndmask_b32_e64 v8, 0, -1, s[0:1]                                          ; D2000008 00018280
>	v_and_b32_e32 v8, 1.0, v8                                                    ; 361010F2
>	v_mov_b32_e32 v9, 0x80000000                                                 ; 7E1202FF 80000000
>	v_xor_b32_e32 v8, v8, v9                                                     ; 3A101308
>	v_cmpx_le_f32_e32 vcc, 0, v8                                                 ; 7C261080
>	v_cmpx_le_f32_e32 vcc, 0, v8                                                 ; 7C261080
>	v_cmpx_le_f32_e32 vcc, 0, v8                                                 ; 7C261080
>	v_cmpx_le_f32_e32 vcc, 0, v8                                                 ; 7C261080
>	v_interp_p1_f32 v8, v0, 1, 2, [m0]                                           ; C8200900
>	v_interp_p2_f32 v8, [v8], v1, 1, 2, [m0]                                     ; C8210901
>	v_sub_f32_e32 v11, 1.0, v8                                                   ; 081610F2
>	v_interp_p1_f32 v8, v0, 0, 2, [m0]                                           ; C8200800
>	v_interp_p2_f32 v8, [v8], v1, 0, 2, [m0]                                     ; C8210801
>	v_add_f32_e32 v10, 0, v8                                                     ; 06141080
>	image_sample v[10:11], 5, 0, 0, 0, 0, 0, 0, 0, v[10:11], s[8:15], s[84:87]   ; F0800500 02A20A0A
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_cmp_ge_f32_e64 s[0:1], -|v10|, 0                                           ; D00C0100 2001010A
>	v_cndmask_b32_e64 v8, 0, -1, s[0:1]                                          ; D2000008 00018280
>	v_cmp_ne_i32_e64 s[0:1], v8, 0                                               ; D10A0000 00010108
>	v_cndmask_b32_e64 v8, 0, -1.0, s[0:1]                                        ; D2000008 0001E680
>	v_cmp_lt_f32_e64 s[0:1], v8, 0                                               ; D0020000 00010108
>	v_cndmask_b32_e64 v8, 0, -1, s[0:1]                                          ; D2000008 00018280
>	v_and_b32_e32 v8, 1.0, v8                                                    ; 361010F2
>	v_xor_b32_e32 v8, v8, v9                                                     ; 3A101308
>	v_cmpx_le_f32_e32 vcc, 0, v8                                                 ; 7C261080
>	v_cmpx_le_f32_e32 vcc, 0, v8                                                 ; 7C261080
>	v_cmpx_le_f32_e32 vcc, 0, v8                                                 ; 7C261080
>	v_cmpx_le_f32_e32 vcc, 0, v8                                                 ; 7C261080
>	v_interp_p1_f32 v8, v0, 1, 1, [m0]                                           ; C8200500
>	v_interp_p2_f32 v8, [v8], v1, 1, 1, [m0]                                     ; C8210501
>	v_mul_f32_e32 v13, 0x3a83126f, v8                                            ; 101A10FF 3A83126F
>	v_interp_p1_f32 v9, v0, 0, 1, [m0]                                           ; C8240400
>	v_interp_p2_f32 v9, [v9], v1, 0, 1, [m0]                                     ; C8250401
>	v_mul_f32_e32 v12, 0x3a83126f, v9                                            ; 101812FF 3A83126F
>	image_sample v[12:15], 15, 0, 0, 0, 0, 0, 0, 0, v[12:13], s[72:79], s[68:71] ; F0800F00 02320C0C
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mul_f32_e32 v16, 0x3f1d70a4, v12                                           ; 102018FF 3F1D70A4
>	v_mov_b32_e32 v17, 0xbf03d662                                                ; 7E2202FF BF03D662
>	v_mad_f32 v16, v13, v17, v16                                                 ; D2820010 0442230D
>	v_mov_b32_e32 v18, 0xbdccd20b                                                ; 7E2402FF BDCCD20B
>	v_mad_f32 v16, v14, v18, v16                                                 ; D2820010 0442250E
>	v_mul_f32_e32 v19, 0x3f1d70a4, v4                                            ; 102608FF 3F1D70A4
>	v_mad_f32 v17, v5, v17, v19                                                  ; D2820011 044E2305
>	v_mad_f32 v17, v6, v18, v17                                                  ; D2820011 04462506
>	v_sub_f32_e32 v18, 1.0, v15                                                  ; 08241EF2
>	v_mul_f32_e32 v17, v17, v18                                                  ; 10222511
>	v_mad_f32 v16, v15, v16, v17                                                 ; D2820010 0446210F
>	v_mov_b32_e32 v17, 0xbe16a93f                                                ; 7E2202FF BE16A93F
>	v_mul_f32_e32 v19, v17, v12                                                  ; 10261911
>	v_mov_b32_e32 v20, 0xbe93e575                                                ; 7E2802FF BE93E575
>	v_mad_f32 v19, v13, v20, v19                                                 ; D2820013 044E290D
>	v_mov_b32_e32 v21, 0x3edf3b64                                                ; 7E2A02FF 3EDF3B64
>	v_mad_f32 v19, v14, v21, v19                                                 ; D2820013 044E2B0E
>	v_mul_f32_e32 v17, v17, v4                                                   ; 10220911
>	v_mad_f32 v17, v5, v20, v17                                                  ; D2820011 04462905
>	v_mad_f32 v17, v6, v21, v17                                                  ; D2820011 04462B06
>	v_mul_f32_e32 v17, v17, v18                                                  ; 10222511
>	v_mad_f32 v12, v15, v19, v17                                                 ; D282000C 0446270F
>	v_mov_b32_e32 v13, 0xbeca0f91                                                ; 7E1A02FF BECA0F91
>	v_mad_f32 v13, v13, v12, v4                                                  ; D282000D 0412190D
>	v_mov_b32_e32 v14, 0xbf14a234                                                ; 7E1C02FF BF14A234
>	v_mad_f32 v18, v14, v16, v13                                                 ; D2820012 0436210E
>	v_mov_b32_e32 v13, 0x40020e17                                                ; 7E1A02FF 40020E17
>	v_mad_f32 v19, v13, v12, v4                                                  ; D2820013 0412190D
>	v_mov_b32_e32 v12, 0x3f91e5f3                                                ; 7E1802FF 3F91E5F3
>	v_mad_f32 v17, v12, v16, v4                                                  ; D2820011 0412210C
>	image_sample v[12:14], 7, 0, 0, 0, 0, 0, 0, 0, v[17:20], s[60:67], s[56:59]  ; F0800700 01CF0C11
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                                           ; C0800300
>	s_waitcnt vmcnt(0) lgkmcnt(0)                                                ; BF8C0070
>	s_buffer_load_dword s4, s[0:3], 0x16                                         ; C2020116
>	s_buffer_load_dword s5, s[0:3], 0x14                                         ; C2028114
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mov_b32_e32 v15, s4                                                        ; 7E1E0204
>	v_mad_f32 v2, s5, v2, v15                                                    ; D2820002 043E0405
>	s_buffer_load_dword s4, s[0:3], 0x0                                          ; C2020100
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_rcp_f32_e32 v15, s4                                                        ; 7E1E5404
>	v_mad_f32 v15, v15, v2, 0                                                    ; D282000F 0202050F
>	s_buffer_load_dword s4, s[0:3], 0x39                                         ; C2020139
>	s_buffer_load_dword s5, s[0:3], 0x38                                         ; C2028138
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mov_b32_e32 v2, s4                                                         ; 7E040204
>	v_mad_f32 v2, s5, v3, v2                                                     ; D2820002 040A0605
>	s_buffer_load_dword s4, s[0:3], 0x17                                         ; C2020117
>	s_buffer_load_dword s5, s[0:3], 0x15                                         ; C2028115
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mov_b32_e32 v3, s4                                                         ; 7E060204
>	v_mad_f32 v2, s5, v2, v3                                                     ; D2820002 040E0405
>	s_buffer_load_dword s4, s[0:3], 0x1                                          ; C2020101
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_rcp_f32_e32 v3, s4                                                         ; 7E065404
>	v_mad_f32 v16, -v3, v2, 1.0                                                  ; D2820010 23CA0503
>	image_sample v[15:18], 15, 0, 0, 0, 0, 0, 0, 0, v[15:16], s[48:55], s[44:47] ; F0800F00 016C0F0F
>	s_buffer_load_dword s4, s[0:3], 0x4                                          ; C2020104
>	s_waitcnt vmcnt(0) lgkmcnt(0)                                                ; BF8C0070
>	v_mad_f32 v2, -v18, s4, 1.0                                                  ; D2820002 23C80912
>	v_mul_f32_e32 v3, s4, v17                                                    ; 10062204
>	v_mad_f32 v3, v14, v2, v3                                                    ; D2820003 040E050E
>	v_sub_f32_e32 v19, 1.0, v10                                                  ; 082614F2
>	v_interp_p1_f32 v21, v0, 3, 0, [m0]                                          ; C8540300
>	v_interp_p2_f32 v21, [v21], v1, 3, 0, [m0]                                   ; C8550301
>	v_interp_p1_f32 v20, v0, 2, 0, [m0]                                          ; C8500200
>	v_interp_p2_f32 v20, [v20], v1, 2, 0, [m0]                                   ; C8510201
>	image_sample v[20:22], 7, 0, 0, 0, 0, 0, 0, 0, v[20:21], s[36:43], s[32:35]  ; F0800700 01091414
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mul_f32_e32 v23, v22, v19                                                  ; 102E2716
>	v_mad_f32 v25, v10, v3, v23                                                  ; D2820019 045E070A
>	v_mul_f32_e32 v3, s4, v16                                                    ; 10062004
>	v_mad_f32 v3, v13, v2, v3                                                    ; D2820003 040E050D
>	v_mul_f32_e32 v27, v21, v19                                                  ; 10362715
>	v_mad_f32 v24, v10, v3, v27                                                  ; D2820018 046E070A
>	v_mul_f32_e32 v3, s4, v15                                                    ; 10061E04
>	v_mad_f32 v2, v12, v2, v3                                                    ; D2820002 040E050C
>	v_mul_f32_e32 v3, v20, v19                                                   ; 10062714
>	v_mad_f32 v23, v10, v2, v3                                                   ; D2820017 040E050A
>	image_sample v[12:14], 7, 0, 0, 0, 0, 0, 0, 0, v[23:26], s[24:31], s[20:23]  ; F0800700 00A60C17
>	v_mad_f32 v2, -v11, v10, v10                                                 ; D2820002 242A150B
>	v_sub_f32_e32 v3, 1.0, v2                                                    ; 080604F2
>	v_mul_f32_e32 v10, v24, v3                                                   ; 10140718
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v10, v2, v13, v10                                                  ; D282000A 042A1B02
>	s_buffer_load_dword s4, s[0:3], 0xd                                          ; C202010D
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v11, s4, v10                                                   ; 08161404
>	s_buffer_load_dword s4, s[0:3], 0x9                                          ; C2020109
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_subrev_f32_e32 v8, s4, v8                                                  ; 0A101004
>	s_buffer_load_dword s4, s[0:3], 0x8                                          ; C2020108
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_subrev_f32_e32 v9, s4, v9                                                  ; 0A121204
>	v_mul_f32_e32 v9, v9, v9                                                     ; 10121309
>	v_mad_f32 v8, v8, v8, v9                                                     ; D2820008 04261108
>	v_interp_p1_f32 v9, v0, 2, 1, [m0]                                           ; C8240600
>	v_interp_p2_f32 v9, [v9], v1, 2, 1, [m0]                                     ; C8250601
>	s_buffer_load_dword s4, s[0:3], 0xa                                          ; C202010A
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_subrev_f32_e32 v9, s4, v9                                                  ; 0A121204
>	v_mad_f32 v8, v9, v9, v8                                                     ; D2820008 04221309
>	s_buffer_load_dword s4, s[0:3], 0x11                                         ; C2020111
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v8, s4, v8                                                     ; 10101004
>	s_buffer_load_dword s4, s[0:3], 0x10                                         ; C2020110
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v15, s4, v9                                                    ; 101E1204
>	v_mul_f32_e32 v15, 0x3fb8aa65, v15                                           ; 101E1EFF 3FB8AA65
>	v_exp_f32_e32 v15, v15                                                       ; 7E1E4B0F
>	v_sub_f32_e32 v15, 1.0, v15                                                  ; 081E1EF2
>	v_mul_f32_e32 v8, v8, v15                                                    ; 10101F08
>	v_rcp_f32_e32 v9, v9                                                         ; 7E125509
>	v_mul_f32_e32 v8, v8, v9                                                     ; 10101308
>	v_mul_f32_e32 v8, 0x3fb8aa65, v8                                             ; 101010FF 3FB8AA65
>	v_exp_f32_e32 v8, v8                                                         ; 7E104B08
>	v_add_f32_e64 v8, 0, v8 clamp                                                ; D2060808 00021080
>	v_sub_f32_e32 v8, 1.0, v8                                                    ; 081010F2
>	v_interp_p1_f32 v9, v0, 1, 3, [m0]                                           ; C8240D00
>	v_interp_p2_f32 v9, [v9], v1, 1, 3, [m0]                                     ; C8250D01
>	v_mad_f32 v0, 0.5, v9, 0.5                                                   ; D2820000 03C212F0
>	s_buffer_load_dword s4, s[0:3], 0x13                                         ; C2020113
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mov_b32_e32 v1, s4                                                         ; 7E020204
>	v_readlane_b32 s4, v28, 0                                                    ; 0209011C
>	v_readlane_b32 s5, v28, 1                                                    ; 020B031C
>	v_readlane_b32 s6, v28, 2                                                    ; 020D051C
>	v_readlane_b32 s7, v28, 3                                                    ; 020F071C
>	v_readlane_b32 s8, v28, 4                                                    ; 0211091C
>	v_readlane_b32 s9, v28, 5                                                    ; 02130B1C
>	v_readlane_b32 s10, v28, 6                                                   ; 02150D1C
>	v_readlane_b32 s11, v28, 7                                                   ; 02170F1C
>	s_nop 2                                                                      ; BF800002
>	image_sample v0, 1, 0, 0, 0, 0, 0, 0, 0, v[0:1], s[4:11], s[16:19]           ; F0800100 00810000
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mul_f32_e32 v0, v0, v8                                                     ; 10001100
>	v_mad_f32 v1, v0, v11, v10                                                   ; D2820001 042A1700
>	v_mul_f32_e32 v8, v23, v3                                                    ; 10100717
>	v_mad_f32 v8, v2, v12, v8                                                    ; D2820008 04221902
>	s_buffer_load_dword s4, s[0:3], 0xc                                          ; C202010C
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v9, s4, v8                                                     ; 08121004
>	v_mad_f32 v8, v0, v9, v8                                                     ; D2820008 04221300
>	v_cvt_pkrtz_f16_f32_e32 v1, v8, v1                                           ; 5E020308
>	v_mul_f32_e32 v3, v25, v3                                                    ; 10060719
>	v_mad_f32 v2, v2, v14, v3                                                    ; D2820002 040E1D02
>	s_buffer_load_dword s0, s[0:3], 0xe                                          ; C200010E
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v3, s0, v2                                                     ; 08060400
>	v_mad_f32 v0, v0, v3, v2                                                     ; D2820000 040A0700
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v7                                           ; 5E000F00
>	exp 15, 0, 1, 1, 1, v1, v0, v1, v0                                           ; F8001C0F 00010001
>	s_endpgm                                                                     ; BF810000
>
>SGPRS: 104
>VGPRS: 32
>Code Size: 1216 bytes
>LDS: 0 blocks
>Scratch: 2048 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL IN[2]
>DCL IN[3]
>DCL IN[4]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL OUT[2], GENERIC[10]
>DCL OUT[3], GENERIC[11]
>DCL OUT[4], GENERIC[12]
>DCL OUT[5], GENERIC[13]
>DCL CONST[0..104]
>DCL TEMP[0..6], LOCAL
>DCL ADDR[0]
>IMM[0] FLT32 {    3.0000,     1.0000,     0.0000,     0.0000}
>IMM[1] INT32 {1, 2, 0, 0}
>IMM[2] FLT32 {    0.5000,    -0.5000,     0.0000,     0.0000}
>  0: MUL TEMP[0].x, IMM[0].xxxx, IN[2].xxxx
>  1: MAD TEMP[1], IN[0].xyzx, IMM[0].yyyz, IMM[0].zzzy
>  2: F2I TEMP[2].x, TEMP[0].xxxx
>  3: UARL ADDR[0].x, TEMP[2].xxxx
>  4: UARL ADDR[0].x, TEMP[2].xxxx
>  5: DP4 TEMP[2].x, TEMP[1], CONST[ADDR[0].x+9]
>  6: F2I TEMP[3].x, TEMP[0].xxxx
>  7: UADD TEMP[3].x, IMM[1].xxxx, TEMP[3].xxxx
>  8: UARL ADDR[0].x, TEMP[3].xxxx
>  9: DP4 TEMP[3].x, TEMP[1], CONST[ADDR[0].x+9]
> 10: MOV TEMP[2].y, TEMP[3].xxxx
> 11: F2I TEMP[3].x, TEMP[0].xxxx
> 12: UADD TEMP[3].x, IMM[1].yyyy, TEMP[3].xxxx
> 13: UARL ADDR[0].x, TEMP[3].xxxx
> 14: DP4 TEMP[3].x, TEMP[1], CONST[ADDR[0].x+9]
> 15: MOV TEMP[2].z, TEMP[3].xxxx
> 16: MUL TEMP[3].xyz, TEMP[2], IN[1].xxxx
> 17: MOV TEMP[2].xyz, TEMP[3].xyzx
> 18: F2I TEMP[3].x, TEMP[0].xxxx
> 19: UARL ADDR[0].x, TEMP[3].xxxx
> 20: UARL ADDR[0].x, TEMP[3].xxxx
> 21: DP3 TEMP[3].x, IN[3].xyzz, CONST[ADDR[0].x+9].xyzz
> 22: F2I TEMP[4].x, TEMP[0].xxxx
> 23: UADD TEMP[4].x, IMM[1].xxxx, TEMP[4].xxxx
> 24: UARL ADDR[0].x, TEMP[4].xxxx
> 25: DP3 TEMP[4].x, IN[3].xyzz, CONST[ADDR[0].x+9].xyzz
> 26: MOV TEMP[3].y, TEMP[4].xxxx
> 27: F2I TEMP[0].x, TEMP[0].xxxx
> 28: UADD TEMP[0].x, IMM[1].yyyy, TEMP[0].xxxx
> 29: UARL ADDR[0].x, TEMP[0].xxxx
> 30: DP3 TEMP[0].x, IN[3].xyzz, CONST[ADDR[0].x+9].xyzz
> 31: MOV TEMP[3].z, TEMP[0].xxxx
> 32: MUL TEMP[0].xyz, TEMP[3], IN[1].xxxx
> 33: MOV TEMP[3].xyz, TEMP[0].xyzx
> 34: FSLT TEMP[0].x, IMM[0].zzzz, IN[1].yyyy
> 35: UIF TEMP[0].xxxx :0
> 36:   MUL TEMP[0].w, IMM[0].xxxx, IN[2].yyyy
> 37:   MOV TEMP[2].w, TEMP[0].wwww
> 38:   F2I TEMP[4].x, TEMP[0].wwww
> 39:   UARL ADDR[0].x, TEMP[4].xxxx
> 40:   UARL ADDR[0].x, TEMP[4].xxxx
> 41:   DP4 TEMP[4].x, TEMP[1], CONST[ADDR[0].x+9]
> 42:   F2I TEMP[5].x, TEMP[0].wwww
> 43:   UADD TEMP[5].x, IMM[1].xxxx, TEMP[5].xxxx
> 44:   UARL ADDR[0].x, TEMP[5].xxxx
> 45:   DP4 TEMP[5].x, TEMP[1], CONST[ADDR[0].x+9]
> 46:   MOV TEMP[4].y, TEMP[5].xxxx
> 47:   F2I TEMP[5].x, TEMP[0].wwww
> 48:   UADD TEMP[5].x, IMM[1].yyyy, TEMP[5].xxxx
> 49:   UARL ADDR[0].x, TEMP[5].xxxx
> 50:   DP4 TEMP[5].x, TEMP[1], CONST[ADDR[0].x+9]
> 51:   MOV TEMP[4].z, TEMP[5].xxxx
> 52:   MAD TEMP[5].xyz, IN[1].yyyy, TEMP[4], TEMP[2]
> 53:   MOV TEMP[2].xyz, TEMP[5].xyzx
> 54:   F2I TEMP[5].x, TEMP[0].wwww
> 55:   UARL ADDR[0].x, TEMP[5].xxxx
> 56:   UARL ADDR[0].x, TEMP[5].xxxx
> 57:   DP3 TEMP[4].x, IN[3].xyzz, CONST[ADDR[0].x+9].xyzz
> 58:   F2I TEMP[5].x, TEMP[0].wwww
> 59:   UADD TEMP[5].x, IMM[1].xxxx, TEMP[5].xxxx
> 60:   UARL ADDR[0].x, TEMP[5].xxxx
> 61:   DP3 TEMP[5].x, IN[3].xyzz, CONST[ADDR[0].x+9].xyzz
> 62:   MOV TEMP[4].y, TEMP[5].xxxx
> 63:   F2I TEMP[0].x, TEMP[0].wwww
> 64:   UADD TEMP[0].x, IMM[1].yyyy, TEMP[0].xxxx
> 65:   UARL ADDR[0].x, TEMP[0].xxxx
> 66:   DP3 TEMP[0].x, IN[3].xyzz, CONST[ADDR[0].x+9].xyzz
> 67:   MOV TEMP[4].z, TEMP[0].xxxx
> 68:   MAD TEMP[0].xyz, IN[1].yyyy, TEMP[4], TEMP[3]
> 69:   MOV TEMP[3].xyz, TEMP[0].xyzx
> 70:   FSLT TEMP[0].x, IMM[0].zzzz, IN[1].zzzz
> 71:   UIF TEMP[0].xxxx :0
> 72:     MUL TEMP[0].w, IMM[0].xxxx, IN[2].zzzz
> 73:     MOV TEMP[2].w, TEMP[0].wwww
> 74:     F2I TEMP[5].x, TEMP[0].wwww
> 75:     UARL ADDR[0].x, TEMP[5].xxxx
> 76:     UARL ADDR[0].x, TEMP[5].xxxx
> 77:     DP4 TEMP[4].x, TEMP[1], CONST[ADDR[0].x+9]
> 78:     F2I TEMP[5].x, TEMP[0].wwww
> 79:     UADD TEMP[5].x, IMM[1].xxxx, TEMP[5].xxxx
> 80:     UARL ADDR[0].x, TEMP[5].xxxx
> 81:     DP4 TEMP[5].x, TEMP[1], CONST[ADDR[0].x+9]
> 82:     MOV TEMP[4].y, TEMP[5].xxxx
> 83:     F2I TEMP[5].x, TEMP[0].wwww
> 84:     UADD TEMP[5].x, IMM[1].yyyy, TEMP[5].xxxx
> 85:     UARL ADDR[0].x, TEMP[5].xxxx
> 86:     DP4 TEMP[5].x, TEMP[1], CONST[ADDR[0].x+9]
> 87:     MOV TEMP[4].z, TEMP[5].xxxx
> 88:     MAD TEMP[4].xyz, IN[1].zzzz, TEMP[4], TEMP[2]
> 89:     MOV TEMP[2].xyz, TEMP[4].xyzx
> 90:     F2I TEMP[4].x, TEMP[0].wwww
> 91:     UARL ADDR[0].x, TEMP[4].xxxx
> 92:     UARL ADDR[0].x, TEMP[4].xxxx
> 93:     DP3 TEMP[1].x, IN[3].xyzz, CONST[ADDR[0].x+9].xyzz
> 94:     F2I TEMP[4].x, TEMP[0].wwww
> 95:     UADD TEMP[4].x, IMM[1].xxxx, TEMP[4].xxxx
> 96:     UARL ADDR[0].x, TEMP[4].xxxx
> 97:     DP3 TEMP[4].x, IN[3].xyzz, CONST[ADDR[0].x+9].xyzz
> 98:     MOV TEMP[1].y, TEMP[4].xxxx
> 99:     F2I TEMP[0].x, TEMP[0].wwww
>100:     UADD TEMP[0].x, IMM[1].yyyy, TEMP[0].xxxx
>101:     UARL ADDR[0].x, TEMP[0].xxxx
>102:     DP3 TEMP[0].x, IN[3].xyzz, CONST[ADDR[0].x+9].xyzz
>103:     MOV TEMP[1].z, TEMP[0].xxxx
>104:     MAD TEMP[0].xyz, IN[1].zzzz, TEMP[1], TEMP[3]
>105:     MOV TEMP[3].xyz, TEMP[0].xyzx
>106:   ENDIF
>107: ENDIF
>108: MUL TEMP[1], TEMP[2].yyyy, CONST[1]
>109: MAD TEMP[1], TEMP[2].xxxx, CONST[0], TEMP[1]
>110: MAD TEMP[1], TEMP[2].zzzz, CONST[2], TEMP[1]
>111: ADD TEMP[1], TEMP[1], CONST[3]
>112: DP3 TEMP[0].x, TEMP[3].xyzz, TEMP[3].xyzz
>113: MAX TEMP[0].x, TEMP[0].xxxx, IMM[0].wwww
>114: RSQ TEMP[0].x, TEMP[0].xxxx
>115: MUL TEMP[0].xyz, TEMP[0].xxxx, TEMP[3]
>116: MOV TEMP[0].xyz, TEMP[0].xyzx
>117: RCP TEMP[4].x, TEMP[1].wwww
>118: MOV TEMP[2].w, TEMP[4].xxxx
>119: MUL TEMP[4].xy, TEMP[1], TEMP[4].xxxx
>120: MOV TEMP[4].xy, TEMP[4].xyxx
>121: ADD TEMP[5].xy, TEMP[2], CONST[8].zwzw
>122: MUL TEMP[5].zw, TEMP[5].xyxy, CONST[8].xyxy
>123: MOV TEMP[5].zw, TEMP[5].wwzw
>124: MUL TEMP[6].xy, TEMP[2].yyyy, CONST[5]
>125: MOV TEMP[3].xy, TEMP[6].xyxx
>126: MAD TEMP[6].xy, TEMP[2].xxxx, CONST[4], TEMP[3]
>127: MOV TEMP[3].xy, TEMP[6].xyxx
>128: MAD TEMP[6].xy, TEMP[2].zzzz, CONST[6], TEMP[3]
>129: MOV TEMP[3].xy, TEMP[6].xyxx
>130: ADD TEMP[6].xy, TEMP[3], CONST[7]
>131: MOV TEMP[3].xy, TEMP[6].xyxx
>132: MAD TEMP[3].xy, TEMP[3], IMM[2].xyzz, IMM[2].yyyy
>133: MOV TEMP[3].xy, TEMP[3].xyxx
>134: MOV TEMP[5].xy, IN[4].xyxx
>135: MOV TEMP[2].xyz, TEMP[2].xyzx
>136: MOV TEMP[2].w, IMM[0].yyyy
>137: MOV TEMP[4].zw, IMM[0].yyzy
>138: MOV TEMP[3].zw, IMM[0].yyzy
>139: MOV OUT[2], TEMP[5]
>140: MOV OUT[3], TEMP[2]
>141: MOV OUT[0], TEMP[1]
>142: MOV OUT[1], TEMP[0]
>143: MOV OUT[4], TEMP[4]
>144: MOV OUT[5], TEMP[3]
>145: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 0)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 4)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %12, i32 8)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %12, i32 12)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %12, i32 16)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %12, i32 20)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %12, i32 24)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %12, i32 28)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %12, i32 32)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %12, i32 36)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %12, i32 40)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %12, i32 44)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %12, i32 48)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %12, i32 52)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %12, i32 56)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %12, i32 60)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %12, i32 64)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %12, i32 68)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %12, i32 80)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %12, i32 84)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %12, i32 96)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %12, i32 100)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %12, i32 112)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %12, i32 116)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %12, i32 128)
>  %38 = call float @llvm.SI.load.const(<16 x i8> %12, i32 132)
>  %39 = call float @llvm.SI.load.const(<16 x i8> %12, i32 136)
>  %40 = call float @llvm.SI.load.const(<16 x i8> %12, i32 140)
>  %41 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %42 = load <16 x i8> addrspace(2)* %41, !tbaa !0
>  %43 = add i32 %5, %7
>  %44 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %42, i32 0, i32 %43)
>  %45 = extractelement <4 x float> %44, i32 0
>  %46 = extractelement <4 x float> %44, i32 1
>  %47 = extractelement <4 x float> %44, i32 2
>  %48 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %49 = load <16 x i8> addrspace(2)* %48, !tbaa !0
>  %50 = add i32 %5, %7
>  %51 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %49, i32 0, i32 %50)
>  %52 = extractelement <4 x float> %51, i32 0
>  %53 = extractelement <4 x float> %51, i32 1
>  %54 = extractelement <4 x float> %51, i32 2
>  %55 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 2
>  %56 = load <16 x i8> addrspace(2)* %55, !tbaa !0
>  %57 = add i32 %5, %7
>  %58 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %56, i32 0, i32 %57)
>  %59 = extractelement <4 x float> %58, i32 0
>  %60 = extractelement <4 x float> %58, i32 1
>  %61 = extractelement <4 x float> %58, i32 2
>  %62 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 3
>  %63 = load <16 x i8> addrspace(2)* %62, !tbaa !0
>  %64 = add i32 %5, %7
>  %65 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %63, i32 0, i32 %64)
>  %66 = extractelement <4 x float> %65, i32 0
>  %67 = extractelement <4 x float> %65, i32 1
>  %68 = extractelement <4 x float> %65, i32 2
>  %69 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 4
>  %70 = load <16 x i8> addrspace(2)* %69, !tbaa !0
>  %71 = add i32 %5, %7
>  %72 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %70, i32 0, i32 %71)
>  %73 = extractelement <4 x float> %72, i32 0
>  %74 = extractelement <4 x float> %72, i32 1
>  %75 = fmul float 3.000000e+00, %59
>  %76 = fmul float %45, 1.000000e+00
>  %77 = fadd float %76, 0.000000e+00
>  %78 = fmul float %46, 1.000000e+00
>  %79 = fadd float %78, 0.000000e+00
>  %80 = fmul float %47, 1.000000e+00
>  %81 = fadd float %80, 0.000000e+00
>  %82 = fmul float %45, 0.000000e+00
>  %83 = fadd float %82, 1.000000e+00
>  %84 = fptosi float %75 to i32
>  %85 = bitcast i32 %84 to float
>  %86 = bitcast float %85 to i32
>  %87 = shl i32 %86, 4
>  %88 = add i32 %87, 144
>  %89 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %88)
>  %90 = shl i32 %86, 4
>  %91 = add i32 %90, 148
>  %92 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %91)
>  %93 = shl i32 %86, 4
>  %94 = add i32 %93, 152
>  %95 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %94)
>  %96 = shl i32 %86, 4
>  %97 = add i32 %96, 156
>  %98 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %97)
>  %99 = fmul float %77, %89
>  %100 = fmul float %79, %92
>  %101 = fadd float %99, %100
>  %102 = fmul float %81, %95
>  %103 = fadd float %101, %102
>  %104 = fmul float %83, %98
>  %105 = fadd float %103, %104
>  %106 = fptosi float %75 to i32
>  %107 = bitcast i32 %106 to float
>  %108 = bitcast float %107 to i32
>  %109 = add i32 1, %108
>  %110 = bitcast i32 %109 to float
>  %111 = bitcast float %110 to i32
>  %112 = shl i32 %111, 4
>  %113 = add i32 %112, 144
>  %114 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %113)
>  %115 = shl i32 %111, 4
>  %116 = add i32 %115, 148
>  %117 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %116)
>  %118 = shl i32 %111, 4
>  %119 = add i32 %118, 152
>  %120 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %119)
>  %121 = shl i32 %111, 4
>  %122 = add i32 %121, 156
>  %123 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %122)
>  %124 = fmul float %77, %114
>  %125 = fmul float %79, %117
>  %126 = fadd float %124, %125
>  %127 = fmul float %81, %120
>  %128 = fadd float %126, %127
>  %129 = fmul float %83, %123
>  %130 = fadd float %128, %129
>  %131 = fptosi float %75 to i32
>  %132 = bitcast i32 %131 to float
>  %133 = bitcast float %132 to i32
>  %134 = add i32 2, %133
>  %135 = bitcast i32 %134 to float
>  %136 = bitcast float %135 to i32
>  %137 = shl i32 %136, 4
>  %138 = add i32 %137, 144
>  %139 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %138)
>  %140 = shl i32 %136, 4
>  %141 = add i32 %140, 148
>  %142 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %141)
>  %143 = shl i32 %136, 4
>  %144 = add i32 %143, 152
>  %145 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %144)
>  %146 = shl i32 %136, 4
>  %147 = add i32 %146, 156
>  %148 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %147)
>  %149 = fmul float %77, %139
>  %150 = fmul float %79, %142
>  %151 = fadd float %149, %150
>  %152 = fmul float %81, %145
>  %153 = fadd float %151, %152
>  %154 = fmul float %83, %148
>  %155 = fadd float %153, %154
>  %156 = fmul float %105, %52
>  %157 = fmul float %130, %52
>  %158 = fmul float %155, %52
>  %159 = fptosi float %75 to i32
>  %160 = bitcast i32 %159 to float
>  %161 = bitcast float %160 to i32
>  %162 = shl i32 %161, 4
>  %163 = add i32 %162, 144
>  %164 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %163)
>  %165 = shl i32 %161, 4
>  %166 = add i32 %165, 148
>  %167 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %166)
>  %168 = shl i32 %161, 4
>  %169 = add i32 %168, 152
>  %170 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %169)
>  %171 = fmul float %66, %164
>  %172 = fmul float %67, %167
>  %173 = fadd float %172, %171
>  %174 = fmul float %68, %170
>  %175 = fadd float %173, %174
>  %176 = fptosi float %75 to i32
>  %177 = bitcast i32 %176 to float
>  %178 = bitcast float %177 to i32
>  %179 = add i32 1, %178
>  %180 = bitcast i32 %179 to float
>  %181 = bitcast float %180 to i32
>  %182 = shl i32 %181, 4
>  %183 = add i32 %182, 144
>  %184 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %183)
>  %185 = shl i32 %181, 4
>  %186 = add i32 %185, 148
>  %187 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %186)
>  %188 = shl i32 %181, 4
>  %189 = add i32 %188, 152
>  %190 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %189)
>  %191 = fmul float %66, %184
>  %192 = fmul float %67, %187
>  %193 = fadd float %192, %191
>  %194 = fmul float %68, %190
>  %195 = fadd float %193, %194
>  %196 = fptosi float %75 to i32
>  %197 = bitcast i32 %196 to float
>  %198 = bitcast float %197 to i32
>  %199 = add i32 2, %198
>  %200 = bitcast i32 %199 to float
>  %201 = bitcast float %200 to i32
>  %202 = shl i32 %201, 4
>  %203 = add i32 %202, 144
>  %204 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %203)
>  %205 = shl i32 %201, 4
>  %206 = add i32 %205, 148
>  %207 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %206)
>  %208 = shl i32 %201, 4
>  %209 = add i32 %208, 152
>  %210 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %209)
>  %211 = fmul float %66, %204
>  %212 = fmul float %67, %207
>  %213 = fadd float %212, %211
>  %214 = fmul float %68, %210
>  %215 = fadd float %213, %214
>  %216 = fmul float %175, %52
>  %217 = fmul float %195, %52
>  %218 = fmul float %215, %52
>  %219 = fcmp olt float 0.000000e+00, %53
>  %220 = sext i1 %219 to i32
>  %221 = bitcast i32 %220 to float
>  %222 = bitcast float %221 to i32
>  %223 = icmp ne i32 %222, 0
>  br i1 %223, label %IF, label %ENDIF
>
>IF:                                               ; preds = %main_body
>  %224 = fmul float 3.000000e+00, %60
>  %225 = fptosi float %224 to i32
>  %226 = bitcast i32 %225 to float
>  %227 = bitcast float %226 to i32
>  %228 = shl i32 %227, 4
>  %229 = add i32 %228, 144
>  %230 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %229)
>  %231 = shl i32 %227, 4
>  %232 = add i32 %231, 148
>  %233 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %232)
>  %234 = shl i32 %227, 4
>  %235 = add i32 %234, 152
>  %236 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %235)
>  %237 = shl i32 %227, 4
>  %238 = add i32 %237, 156
>  %239 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %238)
>  %240 = fmul float %77, %230
>  %241 = fmul float %79, %233
>  %242 = fadd float %240, %241
>  %243 = fmul float %81, %236
>  %244 = fadd float %242, %243
>  %245 = fmul float %83, %239
>  %246 = fadd float %244, %245
>  %247 = fptosi float %224 to i32
>  %248 = bitcast i32 %247 to float
>  %249 = bitcast float %248 to i32
>  %250 = add i32 1, %249
>  %251 = bitcast i32 %250 to float
>  %252 = bitcast float %251 to i32
>  %253 = shl i32 %252, 4
>  %254 = add i32 %253, 144
>  %255 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %254)
>  %256 = shl i32 %252, 4
>  %257 = add i32 %256, 148
>  %258 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %257)
>  %259 = shl i32 %252, 4
>  %260 = add i32 %259, 152
>  %261 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %260)
>  %262 = shl i32 %252, 4
>  %263 = add i32 %262, 156
>  %264 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %263)
>  %265 = fmul float %77, %255
>  %266 = fmul float %79, %258
>  %267 = fadd float %265, %266
>  %268 = fmul float %81, %261
>  %269 = fadd float %267, %268
>  %270 = fmul float %83, %264
>  %271 = fadd float %269, %270
>  %272 = fptosi float %224 to i32
>  %273 = bitcast i32 %272 to float
>  %274 = bitcast float %273 to i32
>  %275 = add i32 2, %274
>  %276 = bitcast i32 %275 to float
>  %277 = bitcast float %276 to i32
>  %278 = shl i32 %277, 4
>  %279 = add i32 %278, 144
>  %280 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %279)
>  %281 = shl i32 %277, 4
>  %282 = add i32 %281, 148
>  %283 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %282)
>  %284 = shl i32 %277, 4
>  %285 = add i32 %284, 152
>  %286 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %285)
>  %287 = shl i32 %277, 4
>  %288 = add i32 %287, 156
>  %289 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %288)
>  %290 = fmul float %77, %280
>  %291 = fmul float %79, %283
>  %292 = fadd float %290, %291
>  %293 = fmul float %81, %286
>  %294 = fadd float %292, %293
>  %295 = fmul float %83, %289
>  %296 = fadd float %294, %295
>  %297 = fmul float %53, %246
>  %298 = fadd float %297, %156
>  %299 = fmul float %53, %271
>  %300 = fadd float %299, %157
>  %301 = fmul float %53, %296
>  %302 = fadd float %301, %158
>  %303 = fptosi float %224 to i32
>  %304 = bitcast i32 %303 to float
>  %305 = bitcast float %304 to i32
>  %306 = shl i32 %305, 4
>  %307 = add i32 %306, 144
>  %308 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %307)
>  %309 = shl i32 %305, 4
>  %310 = add i32 %309, 148
>  %311 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %310)
>  %312 = shl i32 %305, 4
>  %313 = add i32 %312, 152
>  %314 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %313)
>  %315 = fmul float %66, %308
>  %316 = fmul float %67, %311
>  %317 = fadd float %316, %315
>  %318 = fmul float %68, %314
>  %319 = fadd float %317, %318
>  %320 = fptosi float %224 to i32
>  %321 = bitcast i32 %320 to float
>  %322 = bitcast float %321 to i32
>  %323 = add i32 1, %322
>  %324 = bitcast i32 %323 to float
>  %325 = bitcast float %324 to i32
>  %326 = shl i32 %325, 4
>  %327 = add i32 %326, 144
>  %328 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %327)
>  %329 = shl i32 %325, 4
>  %330 = add i32 %329, 148
>  %331 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %330)
>  %332 = shl i32 %325, 4
>  %333 = add i32 %332, 152
>  %334 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %333)
>  %335 = fmul float %66, %328
>  %336 = fmul float %67, %331
>  %337 = fadd float %336, %335
>  %338 = fmul float %68, %334
>  %339 = fadd float %337, %338
>  %340 = fptosi float %224 to i32
>  %341 = bitcast i32 %340 to float
>  %342 = bitcast float %341 to i32
>  %343 = add i32 2, %342
>  %344 = bitcast i32 %343 to float
>  %345 = bitcast float %344 to i32
>  %346 = shl i32 %345, 4
>  %347 = add i32 %346, 144
>  %348 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %347)
>  %349 = shl i32 %345, 4
>  %350 = add i32 %349, 148
>  %351 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %350)
>  %352 = shl i32 %345, 4
>  %353 = add i32 %352, 152
>  %354 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %353)
>  %355 = fmul float %66, %348
>  %356 = fmul float %67, %351
>  %357 = fadd float %356, %355
>  %358 = fmul float %68, %354
>  %359 = fadd float %357, %358
>  %360 = fmul float %53, %319
>  %361 = fadd float %360, %216
>  %362 = fmul float %53, %339
>  %363 = fadd float %362, %217
>  %364 = fmul float %53, %359
>  %365 = fadd float %364, %218
>  %366 = fcmp olt float 0.000000e+00, %54
>  %367 = sext i1 %366 to i32
>  %368 = bitcast i32 %367 to float
>  %369 = bitcast float %368 to i32
>  %370 = icmp ne i32 %369, 0
>  br i1 %370, label %IF70, label %ENDIF
>
>ENDIF:                                            ; preds = %IF70, %IF, %main_body
>  %temp3.0 = phi float [ 0.000000e+00, %main_body ], [ %428, %IF70 ], [ %224, %IF ]
>  %temp8.0 = phi float [ %156, %main_body ], [ %502, %IF70 ], [ %298, %IF ]
>  %temp9.0 = phi float [ %157, %main_body ], [ %504, %IF70 ], [ %300, %IF ]
>  %temp10.0 = phi float [ %158, %main_body ], [ %506, %IF70 ], [ %302, %IF ]
>  %temp12.0 = phi float [ %216, %main_body ], [ %565, %IF70 ], [ %361, %IF ]
>  %temp13.0 = phi float [ %217, %main_body ], [ %567, %IF70 ], [ %363, %IF ]
>  %temp14.0 = phi float [ %218, %main_body ], [ %569, %IF70 ], [ %365, %IF ]
>  %371 = fmul float %temp9.0, %17
>  %372 = fmul float %temp9.0, %18
>  %373 = fmul float %temp9.0, %19
>  %374 = fmul float %temp9.0, %20
>  %375 = fmul float %temp8.0, %13
>  %376 = fadd float %375, %371
>  %377 = fmul float %temp8.0, %14
>  %378 = fadd float %377, %372
>  %379 = fmul float %temp8.0, %15
>  %380 = fadd float %379, %373
>  %381 = fmul float %temp8.0, %16
>  %382 = fadd float %381, %374
>  %383 = fmul float %temp10.0, %21
>  %384 = fadd float %383, %376
>  %385 = fmul float %temp10.0, %22
>  %386 = fadd float %385, %378
>  %387 = fmul float %temp10.0, %23
>  %388 = fadd float %387, %380
>  %389 = fmul float %temp10.0, %24
>  %390 = fadd float %389, %382
>  %391 = fadd float %384, %25
>  %392 = fadd float %386, %26
>  %393 = fadd float %388, %27
>  %394 = fadd float %390, %28
>  %395 = fmul float %temp12.0, %temp12.0
>  %396 = fmul float %temp13.0, %temp13.0
>  %397 = fadd float %396, %395
>  %398 = fmul float %temp14.0, %temp14.0
>  %399 = fadd float %397, %398
>  %400 = call float @llvm.maxnum.f32(float %399, float 0x3E7AD7F2A0000000)
>  %401 = call float @llvm.AMDGPU.rsq.clamped.f32(float %400)
>  %402 = fmul float %401, %temp12.0
>  %403 = fmul float %401, %temp13.0
>  %404 = fmul float %401, %temp14.0
>  %405 = fdiv float 1.000000e+00, %394
>  %406 = fmul float %391, %405
>  %407 = fmul float %392, %405
>  %408 = fadd float %temp8.0, %39
>  %409 = fadd float %temp9.0, %40
>  %410 = fmul float %408, %37
>  %411 = fmul float %409, %38
>  %412 = fmul float %temp9.0, %31
>  %413 = fmul float %temp9.0, %32
>  %414 = fmul float %temp8.0, %29
>  %415 = fadd float %414, %412
>  %416 = fmul float %temp8.0, %30
>  %417 = fadd float %416, %413
>  %418 = fmul float %temp10.0, %33
>  %419 = fadd float %418, %415
>  %420 = fmul float %temp10.0, %34
>  %421 = fadd float %420, %417
>  %422 = fadd float %419, %35
>  %423 = fadd float %421, %36
>  %424 = fmul float %422, 5.000000e-01
>  %425 = fadd float %424, -5.000000e-01
>  %426 = fmul float %423, -5.000000e-01
>  %427 = fadd float %426, -5.000000e-01
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %402, float %403, float %404, float %temp3.0)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %73, float %74, float %410, float %411)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 34, i32 0, float %temp8.0, float %temp9.0, float %temp10.0, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 35, i32 0, float %406, float %407, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 36, i32 0, float %425, float %427, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %391, float %392, float %393, float %394)
>  ret void
>
>IF70:                                             ; preds = %IF
>  %428 = fmul float 3.000000e+00, %61
>  %429 = fptosi float %428 to i32
>  %430 = bitcast i32 %429 to float
>  %431 = bitcast float %430 to i32
>  %432 = shl i32 %431, 4
>  %433 = add i32 %432, 144
>  %434 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %433)
>  %435 = shl i32 %431, 4
>  %436 = add i32 %435, 148
>  %437 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %436)
>  %438 = shl i32 %431, 4
>  %439 = add i32 %438, 152
>  %440 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %439)
>  %441 = shl i32 %431, 4
>  %442 = add i32 %441, 156
>  %443 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %442)
>  %444 = fmul float %77, %434
>  %445 = fmul float %79, %437
>  %446 = fadd float %444, %445
>  %447 = fmul float %81, %440
>  %448 = fadd float %446, %447
>  %449 = fmul float %83, %443
>  %450 = fadd float %448, %449
>  %451 = fptosi float %428 to i32
>  %452 = bitcast i32 %451 to float
>  %453 = bitcast float %452 to i32
>  %454 = add i32 1, %453
>  %455 = bitcast i32 %454 to float
>  %456 = bitcast float %455 to i32
>  %457 = shl i32 %456, 4
>  %458 = add i32 %457, 144
>  %459 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %458)
>  %460 = shl i32 %456, 4
>  %461 = add i32 %460, 148
>  %462 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %461)
>  %463 = shl i32 %456, 4
>  %464 = add i32 %463, 152
>  %465 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %464)
>  %466 = shl i32 %456, 4
>  %467 = add i32 %466, 156
>  %468 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %467)
>  %469 = fmul float %77, %459
>  %470 = fmul float %79, %462
>  %471 = fadd float %469, %470
>  %472 = fmul float %81, %465
>  %473 = fadd float %471, %472
>  %474 = fmul float %83, %468
>  %475 = fadd float %473, %474
>  %476 = fptosi float %428 to i32
>  %477 = bitcast i32 %476 to float
>  %478 = bitcast float %477 to i32
>  %479 = add i32 2, %478
>  %480 = bitcast i32 %479 to float
>  %481 = bitcast float %480 to i32
>  %482 = shl i32 %481, 4
>  %483 = add i32 %482, 144
>  %484 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %483)
>  %485 = shl i32 %481, 4
>  %486 = add i32 %485, 148
>  %487 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %486)
>  %488 = shl i32 %481, 4
>  %489 = add i32 %488, 152
>  %490 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %489)
>  %491 = shl i32 %481, 4
>  %492 = add i32 %491, 156
>  %493 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %492)
>  %494 = fmul float %77, %484
>  %495 = fmul float %79, %487
>  %496 = fadd float %494, %495
>  %497 = fmul float %81, %490
>  %498 = fadd float %496, %497
>  %499 = fmul float %83, %493
>  %500 = fadd float %498, %499
>  %501 = fmul float %54, %450
>  %502 = fadd float %501, %298
>  %503 = fmul float %54, %475
>  %504 = fadd float %503, %300
>  %505 = fmul float %54, %500
>  %506 = fadd float %505, %302
>  %507 = fptosi float %428 to i32
>  %508 = bitcast i32 %507 to float
>  %509 = bitcast float %508 to i32
>  %510 = shl i32 %509, 4
>  %511 = add i32 %510, 144
>  %512 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %511)
>  %513 = shl i32 %509, 4
>  %514 = add i32 %513, 148
>  %515 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %514)
>  %516 = shl i32 %509, 4
>  %517 = add i32 %516, 152
>  %518 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %517)
>  %519 = fmul float %66, %512
>  %520 = fmul float %67, %515
>  %521 = fadd float %520, %519
>  %522 = fmul float %68, %518
>  %523 = fadd float %521, %522
>  %524 = fptosi float %428 to i32
>  %525 = bitcast i32 %524 to float
>  %526 = bitcast float %525 to i32
>  %527 = add i32 1, %526
>  %528 = bitcast i32 %527 to float
>  %529 = bitcast float %528 to i32
>  %530 = shl i32 %529, 4
>  %531 = add i32 %530, 144
>  %532 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %531)
>  %533 = shl i32 %529, 4
>  %534 = add i32 %533, 148
>  %535 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %534)
>  %536 = shl i32 %529, 4
>  %537 = add i32 %536, 152
>  %538 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %537)
>  %539 = fmul float %66, %532
>  %540 = fmul float %67, %535
>  %541 = fadd float %540, %539
>  %542 = fmul float %68, %538
>  %543 = fadd float %541, %542
>  %544 = fptosi float %428 to i32
>  %545 = bitcast i32 %544 to float
>  %546 = bitcast float %545 to i32
>  %547 = add i32 2, %546
>  %548 = bitcast i32 %547 to float
>  %549 = bitcast float %548 to i32
>  %550 = shl i32 %549, 4
>  %551 = add i32 %550, 144
>  %552 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %551)
>  %553 = shl i32 %549, 4
>  %554 = add i32 %553, 148
>  %555 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %554)
>  %556 = shl i32 %549, 4
>  %557 = add i32 %556, 152
>  %558 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %557)
>  %559 = fmul float %66, %552
>  %560 = fmul float %67, %555
>  %561 = fadd float %560, %559
>  %562 = fmul float %68, %558
>  %563 = fadd float %561, %562
>  %564 = fmul float %54, %523
>  %565 = fadd float %564, %361
>  %566 = fmul float %54, %543
>  %567 = fadd float %566, %363
>  %568 = fmul float %54, %563
>  %569 = fadd float %568, %365
>  br label %ENDIF
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.maxnum.f32(float, float) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.AMDGPU.rsq.clamped.f32(float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v11, s10, v0                               ; 4A16000A
>	s_load_dwordx4 s[12:15], s[8:9], 0x0                     ; C0860900
>	s_load_dwordx4 s[16:19], s[8:9], 0x4                     ; C0880904
>	s_load_dwordx4 s[20:23], s[8:9], 0x8                     ; C08A0908
>	s_load_dwordx4 s[24:27], s[8:9], 0xc                     ; C08C090C
>	s_load_dwordx4 s[4:7], s[8:9], 0x10                      ; C0820910
>	s_waitcnt lgkmcnt(0)                                     ; BF8C007F
>	buffer_load_format_xyzw v[20:23], v11, s[20:23], 0 idxen ; E00C2000 8005140B
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v0, 0x40400000, v20                        ; 100028FF 40400000
>	v_cvt_i32_f32_e32 v0, v0                                 ; 7E001100
>	v_lshlrev_b32_e32 v3, 4, v0                              ; 34060084
>	v_add_i32_e32 v0, 0xb0, v3                               ; 4A0006FF 000000B0
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                       ; C0800300
>	s_waitcnt lgkmcnt(0)                                     ; BF8C007F
>	buffer_load_dword v0, v0, s[0:3], 0 offen                ; E0301000 80000000
>	buffer_load_format_xyzw v[15:18], v11, s[24:27], 0 idxen ; E00C2000 80060F0B
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v1, v0, v15                                ; 10021F00
>	v_add_i32_e32 v2, 0xb4, v3                               ; 4A0406FF 000000B4
>	buffer_load_dword v2, v2, s[0:3], 0 offen                ; E0301000 80000202
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v1, v16, v2, v1                                ; D2820001 04060510
>	v_add_i32_e32 v4, 0xb8, v3                               ; 4A0806FF 000000B8
>	buffer_load_dword v12, v4, s[0:3], 0 offen               ; E0301000 80000C04
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v1, v17, v12, v1                               ; D2820001 04061911
>	buffer_load_format_xyzw v[7:10], v11, s[16:19], 0 idxen  ; E00C2000 8004070B
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v1, v7, v1                                 ; 10020307
>	v_add_i32_e32 v4, 0xa0, v3                               ; 4A0806FF 000000A0
>	buffer_load_dword v13, v4, s[0:3], 0 offen               ; E0301000 80000D04
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v4, v13, v15                               ; 10081F0D
>	v_add_i32_e32 v5, 0xa4, v3                               ; 4A0A06FF 000000A4
>	buffer_load_dword v14, v5, s[0:3], 0 offen               ; E0301000 80000E05
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v4, v16, v14, v4                               ; D2820004 04121D10
>	v_add_i32_e32 v5, 0xa8, v3                               ; 4A0A06FF 000000A8
>	buffer_load_dword v27, v5, s[0:3], 0 offen               ; E0301000 80001B05
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v4, v17, v27, v4                               ; D2820004 04123711
>	v_mul_f32_e32 v4, v7, v4                                 ; 10080907
>	v_add_i32_e32 v5, 0x90, v3                               ; 4A0A06FF 00000090
>	buffer_load_dword v28, v5, s[0:3], 0 offen               ; E0301000 80001C05
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v5, v28, v15                               ; 100A1F1C
>	v_add_i32_e32 v6, 0x94, v3                               ; 4A0C06FF 00000094
>	buffer_load_dword v29, v6, s[0:3], 0 offen               ; E0301000 80001D06
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v5, v16, v29, v5                               ; D2820005 04163B10
>	v_add_i32_e32 v6, 0x98, v3                               ; 4A0C06FF 00000098
>	buffer_load_dword v30, v6, s[0:3], 0 offen               ; E0301000 80001E06
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v5, v17, v30, v5                               ; D2820005 04163D11
>	v_mul_f32_e32 v5, v7, v5                                 ; 100A0B07
>	buffer_load_format_xyzw v[31:34], v11, s[12:15], 0 idxen ; E00C2000 80031F0B
>	v_mov_b32_e32 v6, 0                                      ; 7E0C0280
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_add_f32_e32 v19, 0, v32                                ; 06264080
>	v_mul_f32_e32 v2, v2, v19                                ; 10042702
>	v_add_f32_e32 v24, 0, v31                                ; 06303E80
>	v_mad_f32 v0, v24, v0, v2                                ; D2820000 040A0118
>	v_add_f32_e32 v26, 0, v33                                ; 06344280
>	v_mad_f32 v0, v26, v12, v0                               ; D2820000 0402191A
>	v_mad_f32 v25, 0, v31, 1.0                               ; D2820019 03CA3E80
>	v_add_i32_e32 v2, 0xbc, v3                               ; 4A0406FF 000000BC
>	buffer_load_dword v2, v2, s[0:3], 0 offen                ; E0301000 80000202
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v0, v25, v2, v0                                ; D2820000 04020519
>	v_mul_f32_e32 v0, v7, v0                                 ; 10000107
>	v_mul_f32_e32 v2, v14, v19                               ; 1004270E
>	v_mad_f32 v2, v24, v13, v2                               ; D2820002 040A1B18
>	v_mad_f32 v2, v26, v27, v2                               ; D2820002 040A371A
>	v_add_i32_e32 v12, 0xac, v3                              ; 4A1806FF 000000AC
>	buffer_load_dword v12, v12, s[0:3], 0 offen              ; E0301000 80000C0C
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v2, v25, v12, v2                               ; D2820002 040A1919
>	v_mul_f32_e32 v2, v7, v2                                 ; 10040507
>	v_mul_f32_e32 v12, v29, v19                              ; 1018271D
>	v_mad_f32 v12, v24, v28, v12                             ; D282000C 04323918
>	v_mad_f32 v12, v26, v30, v12                             ; D282000C 04323D1A
>	v_add_i32_e32 v3, 0x9c, v3                               ; 4A0606FF 0000009C
>	buffer_load_dword v3, v3, s[0:3], 0 offen                ; E0301000 80000303
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v3, v25, v3, v12                               ; D2820003 04320719
>	v_mul_f32_e32 v3, v7, v3                                 ; 10060707
>	buffer_load_format_xyzw v[11:14], v11, s[4:7], 0 idxen   ; E00C2000 80010B0B
>	v_cmp_gt_f32_e64 s[4:5], v8, 0                           ; D0080004 00010108
>	v_cndmask_b32_e64 v27, 0, -1, s[4:5]                     ; D200081B 00118280
>	v_cmp_ne_i32_e64 s[32:33], v27, 0                        ; D10A0020 0001011B
>	s_buffer_load_dword s20, s[0:3], 0x23                    ; C20A0123
>	s_buffer_load_dword s19, s[0:3], 0x22                    ; C2098122
>	s_buffer_load_dword s22, s[0:3], 0x21                    ; C20B0121
>	s_buffer_load_dword s21, s[0:3], 0x20                    ; C20A8120
>	s_buffer_load_dword s4, s[0:3], 0x1d                     ; C202011D
>	s_buffer_load_dword s7, s[0:3], 0x1c                     ; C203811C
>	s_buffer_load_dword s5, s[0:3], 0x19                     ; C2028119
>	s_buffer_load_dword s11, s[0:3], 0x18                    ; C2058118
>	s_buffer_load_dword s8, s[0:3], 0x15                     ; C2040115
>	s_buffer_load_dword s12, s[0:3], 0x14                    ; C2060114
>	s_buffer_load_dword s9, s[0:3], 0x11                     ; C2048111
>	s_buffer_load_dword s13, s[0:3], 0x10                    ; C2068110
>	s_buffer_load_dword s17, s[0:3], 0xf                     ; C208810F
>	s_buffer_load_dword s6, s[0:3], 0xe                      ; C203010E
>	s_buffer_load_dword s24, s[0:3], 0xd                     ; C20C010D
>	s_buffer_load_dword s16, s[0:3], 0xc                     ; C208010C
>	s_buffer_load_dword s23, s[0:3], 0xb                     ; C20B810B
>	s_buffer_load_dword s10, s[0:3], 0xa                     ; C205010A
>	s_buffer_load_dword s26, s[0:3], 0x9                     ; C20D0109
>	s_buffer_load_dword s18, s[0:3], 0x8                     ; C2090108
>	s_buffer_load_dword s27, s[0:3], 0x7                     ; C20D8107
>	s_buffer_load_dword s14, s[0:3], 0x6                     ; C2070106
>	s_buffer_load_dword s30, s[0:3], 0x5                     ; C20F0105
>	s_buffer_load_dword s25, s[0:3], 0x4                     ; C20C8104
>	s_buffer_load_dword s29, s[0:3], 0x3                     ; C20E8103
>	s_buffer_load_dword s15, s[0:3], 0x2                     ; C2078102
>	s_buffer_load_dword s31, s[0:3], 0x1                     ; C20F8101
>	s_buffer_load_dword s28, s[0:3], 0x0                     ; C20E0100
>	s_waitcnt vmcnt(0) lgkmcnt(0)                            ; BF8C0070
>	s_and_saveexec_b64 s[32:33], s[32:33]                    ; BEA02420
>	s_xor_b64 s[32:33], exec, s[32:33]                       ; 89A0207E
>	s_cbranch_execz BB0_3                                    ; BF880000
>	v_mul_f32_e32 v6, 0x40400000, v21                        ; 100C2AFF 40400000
>	v_cvt_i32_f32_e32 v27, v6                                ; 7E361106
>	v_lshlrev_b32_e32 v27, 4, v27                            ; 34363684
>	v_add_i32_e32 v28, 0xb4, v27                             ; 4A3836FF 000000B4
>	buffer_load_dword v28, v28, s[0:3], 0 offen              ; E0301000 80001C1C
>	v_add_i32_e32 v29, 0xb0, v27                             ; 4A3A36FF 000000B0
>	buffer_load_dword v29, v29, s[0:3], 0 offen              ; E0301000 80001D1D
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v30, v29, v15                              ; 103C1F1D
>	v_mad_f32 v30, v16, v28, v30                             ; D282001E 047A3910
>	v_add_i32_e32 v31, 0xb8, v27                             ; 4A3E36FF 000000B8
>	buffer_load_dword v31, v31, s[0:3], 0 offen              ; E0301000 80001F1F
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v30, v17, v31, v30                             ; D282001E 047A3F11
>	v_mad_f32 v1, v8, v30, v1                                ; D2820001 04063D08
>	v_add_i32_e32 v30, 0xa4, v27                             ; 4A3C36FF 000000A4
>	buffer_load_dword v30, v30, s[0:3], 0 offen              ; E0301000 80001E1E
>	v_add_i32_e32 v32, 0xa0, v27                             ; 4A4036FF 000000A0
>	buffer_load_dword v32, v32, s[0:3], 0 offen              ; E0301000 80002020
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v33, v32, v15                              ; 10421F20
>	v_mad_f32 v33, v16, v30, v33                             ; D2820021 04863D10
>	v_add_i32_e32 v34, 0xa8, v27                             ; 4A4436FF 000000A8
>	buffer_load_dword v34, v34, s[0:3], 0 offen              ; E0301000 80002222
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v33, v17, v34, v33                             ; D2820021 04864511
>	v_mad_f32 v4, v8, v33, v4                                ; D2820004 04124308
>	v_add_i32_e32 v33, 0x94, v27                             ; 4A4236FF 00000094
>	buffer_load_dword v33, v33, s[0:3], 0 offen              ; E0301000 80002121
>	v_add_i32_e32 v35, 0x90, v27                             ; 4A4636FF 00000090
>	buffer_load_dword v35, v35, s[0:3], 0 offen              ; E0301000 80002323
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v36, v35, v15                              ; 10481F23
>	v_mad_f32 v36, v16, v33, v36                             ; D2820024 04924310
>	v_add_i32_e32 v37, 0x98, v27                             ; 4A4A36FF 00000098
>	buffer_load_dword v37, v37, s[0:3], 0 offen              ; E0301000 80002525
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v36, v17, v37, v36                             ; D2820024 04924B11
>	v_mad_f32 v5, v8, v36, v5                                ; D2820005 04164908
>	v_mul_f32_e32 v28, v28, v19                              ; 1038271C
>	v_mad_f32 v28, v24, v29, v28                             ; D282001C 04723B18
>	v_mad_f32 v28, v26, v31, v28                             ; D282001C 04723F1A
>	v_add_i32_e32 v29, 0xbc, v27                             ; 4A3A36FF 000000BC
>	buffer_load_dword v29, v29, s[0:3], 0 offen              ; E0301000 80001D1D
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v28, v25, v29, v28                             ; D282001C 04723B19
>	v_mad_f32 v0, v8, v28, v0                                ; D2820000 04023908
>	v_mul_f32_e32 v28, v30, v19                              ; 1038271E
>	v_mad_f32 v28, v24, v32, v28                             ; D282001C 04724118
>	v_mad_f32 v28, v26, v34, v28                             ; D282001C 0472451A
>	v_add_i32_e32 v29, 0xac, v27                             ; 4A3A36FF 000000AC
>	buffer_load_dword v29, v29, s[0:3], 0 offen              ; E0301000 80001D1D
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v28, v25, v29, v28                             ; D282001C 04723B19
>	v_mad_f32 v2, v8, v28, v2                                ; D2820002 040A3908
>	v_mul_f32_e32 v28, v33, v19                              ; 10382721
>	v_mad_f32 v28, v24, v35, v28                             ; D282001C 04724718
>	v_mad_f32 v28, v26, v37, v28                             ; D282001C 04724B1A
>	v_add_i32_e32 v27, 0x9c, v27                             ; 4A3636FF 0000009C
>	buffer_load_dword v27, v27, s[0:3], 0 offen              ; E0301000 80001B1B
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v27, v25, v27, v28                             ; D282001B 04723719
>	v_mad_f32 v3, v8, v27, v3                                ; D2820003 040E3708
>	v_cmp_gt_f32_e64 s[34:35], v9, 0                         ; D0080022 00010109
>	v_cndmask_b32_e64 v27, 0, -1, s[34:35]                   ; D200081B 00898280
>	v_cmp_ne_i32_e64 s[34:35], v27, 0                        ; D10A0022 0001011B
>	s_and_saveexec_b64 s[34:35], s[34:35]                    ; BEA22422
>	s_xor_b64 s[34:35], exec, s[34:35]                       ; 89A2227E
>	s_cbranch_execz BB0_2                                    ; BF880000
>	v_mul_f32_e32 v6, 0x40400000, v22                        ; 100C2CFF 40400000
>	v_cvt_i32_f32_e32 v20, v6                                ; 7E281106
>	v_lshlrev_b32_e32 v20, 4, v20                            ; 34282884
>	v_add_i32_e32 v21, 0xb4, v20                             ; 4A2A28FF 000000B4
>	buffer_load_dword v21, v21, s[0:3], 0 offen              ; E0301000 80001515
>	v_add_i32_e32 v22, 0xb0, v20                             ; 4A2C28FF 000000B0
>	buffer_load_dword v22, v22, s[0:3], 0 offen              ; E0301000 80001616
>	s_waitcnt vmcnt(0) expcnt(0)                             ; BF8C0700
>	v_mul_f32_e32 v23, v22, v15                              ; 102E1F16
>	v_mad_f32 v23, v16, v21, v23                             ; D2820017 045E2B10
>	v_add_i32_e32 v27, 0xb8, v20                             ; 4A3628FF 000000B8
>	buffer_load_dword v27, v27, s[0:3], 0 offen              ; E0301000 80001B1B
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v23, v17, v27, v23                             ; D2820017 045E3711
>	v_mad_f32 v1, v9, v23, v1                                ; D2820001 04062F09
>	v_add_i32_e32 v23, 0xa4, v20                             ; 4A2E28FF 000000A4
>	buffer_load_dword v23, v23, s[0:3], 0 offen              ; E0301000 80001717
>	v_add_i32_e32 v28, 0xa0, v20                             ; 4A3828FF 000000A0
>	buffer_load_dword v28, v28, s[0:3], 0 offen              ; E0301000 80001C1C
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v29, v28, v15                              ; 103A1F1C
>	v_mad_f32 v29, v16, v23, v29                             ; D282001D 04762F10
>	v_add_i32_e32 v30, 0xa8, v20                             ; 4A3C28FF 000000A8
>	buffer_load_dword v30, v30, s[0:3], 0 offen              ; E0301000 80001E1E
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v29, v17, v30, v29                             ; D282001D 04763D11
>	v_mad_f32 v4, v9, v29, v4                                ; D2820004 04123B09
>	v_add_i32_e32 v29, 0x94, v20                             ; 4A3A28FF 00000094
>	buffer_load_dword v29, v29, s[0:3], 0 offen              ; E0301000 80001D1D
>	v_add_i32_e32 v31, 0x90, v20                             ; 4A3E28FF 00000090
>	buffer_load_dword v31, v31, s[0:3], 0 offen              ; E0301000 80001F1F
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v32, v31, v15                              ; 10401F1F
>	v_mad_f32 v32, v16, v29, v32                             ; D2820020 04823B10
>	v_add_i32_e32 v33, 0x98, v20                             ; 4A4228FF 00000098
>	buffer_load_dword v33, v33, s[0:3], 0 offen              ; E0301000 80002121
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v15, v17, v33, v32                             ; D282000F 04824311
>	v_mad_f32 v5, v9, v15, v5                                ; D2820005 04161F09
>	v_mul_f32_e32 v15, v21, v19                              ; 101E2715
>	v_mad_f32 v15, v24, v22, v15                             ; D282000F 043E2D18
>	v_mad_f32 v15, v26, v27, v15                             ; D282000F 043E371A
>	v_add_i32_e32 v16, 0xbc, v20                             ; 4A2028FF 000000BC
>	buffer_load_dword v16, v16, s[0:3], 0 offen              ; E0301000 80001010
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v15, v25, v16, v15                             ; D282000F 043E2119
>	v_mad_f32 v0, v9, v15, v0                                ; D2820000 04021F09
>	v_mul_f32_e32 v15, v23, v19                              ; 101E2717
>	v_mad_f32 v15, v24, v28, v15                             ; D282000F 043E3918
>	v_mad_f32 v15, v26, v30, v15                             ; D282000F 043E3D1A
>	v_add_i32_e32 v16, 0xac, v20                             ; 4A2028FF 000000AC
>	buffer_load_dword v16, v16, s[0:3], 0 offen              ; E0301000 80001010
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v15, v25, v16, v15                             ; D282000F 043E2119
>	v_mad_f32 v2, v9, v15, v2                                ; D2820002 040A1F09
>	v_mul_f32_e32 v15, v29, v19                              ; 101E271D
>	v_mad_f32 v15, v24, v31, v15                             ; D282000F 043E3F18
>	v_mad_f32 v15, v26, v33, v15                             ; D282000F 043E431A
>	v_add_i32_e32 v16, 0x9c, v20                             ; 4A2028FF 0000009C
>	buffer_load_dword v16, v16, s[0:3], 0 offen              ; E0301000 80001010
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v15, v25, v16, v15                             ; D282000F 043E2119
>	v_mad_f32 v3, v9, v15, v3                                ; D2820003 040E1F09
>	s_or_b64 exec, exec, s[34:35]                            ; 88FE227E
>	s_or_b64 exec, exec, s[32:33]                            ; 88FE207E
>	v_mul_f32_e32 v7, v5, v5                                 ; 100E0B05
>	v_mad_f32 v7, v4, v4, v7                                 ; D2820007 041E0904
>	v_mad_f32 v7, v1, v1, v7                                 ; D2820007 041E0301
>	v_max_f32_e32 v7, 0x33d6bf95, v7                         ; 200E0EFF 33D6BF95
>	v_rsq_clamp_f32_e32 v7, v7                               ; 7E0E5907
>	v_mul_f32_e32 v1, v1, v7                                 ; 10020F01
>	v_mul_f32_e32 v4, v4, v7                                 ; 10080F04
>	v_mul_f32_e32 v5, v5, v7                                 ; 100A0F05
>	exp 15, 32, 0, 0, 0, v5, v4, v1, v6                      ; F800020F 06010405
>	s_waitcnt expcnt(0)                                      ; BF8C070F
>	v_add_f32_e32 v1, s20, v2                                ; 06020414
>	v_mul_f32_e32 v1, s22, v1                                ; 10020216
>	v_add_f32_e32 v4, s19, v3                                ; 06080613
>	v_mul_f32_e32 v4, s21, v4                                ; 10080815
>	exp 15, 33, 0, 0, 0, v11, v12, v4, v1                    ; F800021F 01040C0B
>	s_waitcnt expcnt(0)                                      ; BF8C070F
>	v_mov_b32_e32 v1, 1.0                                    ; 7E0202F2
>	exp 15, 34, 0, 0, 0, v3, v2, v0, v1                      ; F800022F 01000203
>	v_mul_f32_e32 v4, s30, v2                                ; 1008041E
>	v_mad_f32 v4, v3, s31, v4                                ; D2820004 04103F03
>	v_mad_f32 v4, v0, s26, v4                                ; D2820004 04103500
>	v_add_f32_e32 v4, s24, v4                                ; 06080818
>	v_mul_f32_e32 v5, s27, v2                                ; 100A041B
>	v_mad_f32 v5, v3, s29, v5                                ; D2820005 04143B03
>	v_mad_f32 v5, v0, s23, v5                                ; D2820005 04142F00
>	v_add_f32_e32 v5, s17, v5                                ; 060A0A11
>	v_rcp_f32_e32 v6, v5                                     ; 7E0C5505
>	v_mul_f32_e32 v7, v6, v4                                 ; 100E0906
>	v_mul_f32_e32 v8, s25, v2                                ; 10100419
>	v_mad_f32 v8, v3, s28, v8                                ; D2820008 04203903
>	v_mad_f32 v8, v0, s18, v8                                ; D2820008 04202500
>	v_add_f32_e32 v8, s16, v8                                ; 06101010
>	v_mul_f32_e32 v6, v6, v8                                 ; 100C1106
>	v_mov_b32_e32 v9, 0                                      ; 7E120280
>	exp 15, 35, 0, 0, 0, v6, v7, v9, v1                      ; F800023F 01090706
>	s_waitcnt expcnt(0)                                      ; BF8C070F
>	v_mul_f32_e32 v6, s12, v2                                ; 100C040C
>	v_mad_f32 v6, v3, s13, v6                                ; D2820006 04181B03
>	v_mad_f32 v6, v0, s11, v6                                ; D2820006 04181700
>	v_add_f32_e32 v6, s7, v6                                 ; 060C0C07
>	v_mad_f32 v6, 0.5, v6, -0.5                              ; D2820006 03C60CF0
>	v_mul_f32_e32 v7, s8, v2                                 ; 100E0408
>	v_mad_f32 v7, v3, s9, v7                                 ; D2820007 041C1303
>	v_mad_f32 v7, v0, s5, v7                                 ; D2820007 041C0B00
>	v_add_f32_e32 v7, s4, v7                                 ; 060E0E04
>	v_mad_f32 v7, -0.5, v7, -0.5                             ; D2820007 03C60EF1
>	exp 15, 36, 0, 0, 0, v6, v7, v9, v1                      ; F800024F 01090706
>	s_waitcnt expcnt(0)                                      ; BF8C070F
>	v_mul_f32_e32 v1, s14, v2                                ; 1002040E
>	v_mad_f32 v1, v3, s15, v1                                ; D2820001 04041F03
>	v_mad_f32 v0, v0, s10, v1                                ; D2820000 04041500
>	v_add_f32_e32 v0, s6, v0                                 ; 06000006
>	exp 15, 12, 0, 1, 0, v8, v4, v0, v5                      ; F80008CF 05000408
>	s_endpgm                                                 ; BF810000
>
>SGPRS: 40
>VGPRS: 40
>Code Size: 1912 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL IN[1], GENERIC[10], PERSPECTIVE
>DCL IN[2], GENERIC[11], PERSPECTIVE
>DCL IN[3], GENERIC[12], PERSPECTIVE
>DCL IN[4], GENERIC[13], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SAMP[1]
>DCL SAMP[2]
>DCL SAMP[3]
>DCL SAMP[4]
>DCL SAMP[5]
>DCL SAMP[6]
>DCL SAMP[7]
>DCL SAMP[8]
>DCL SAMP[9]
>DCL CONST[0..8]
>DCL TEMP[0..9], LOCAL
>IMM[0] FLT32 {    2.2000,     1.0000,    -0.1000,     0.0000}
>IMM[1] FLT32 {    0.0000,     0.0529,     0.8460,     0.5290}
>IMM[2] FLT32 {  199.0000,     0.1000,     0.4545,     1.4427}
>IMM[3] FLT32 {    1.0000,    -1.0000,     0.5000,     0.0000}
>  0: MOV TEMP[0].xy, IN[1].xyyy
>  1: TEX TEMP[0], TEMP[0], SAMP[0], 2D
>  2: POW TEMP[1].x, TEMP[0].xxxx, IMM[0].xxxx
>  3: POW TEMP[1].y, TEMP[0].yyyy, IMM[0].xxxx
>  4: POW TEMP[1].z, TEMP[0].zzzz, IMM[0].xxxx
>  5: POW TEMP[1].w, TEMP[0].wwww, IMM[0].yyyy
>  6: MOV TEMP[0].w, TEMP[1]
>  7: ABS TEMP[2].x, TEMP[1].wwww
>  8: POW TEMP[2].x, TEMP[2].xxxx, IMM[0].xxxx
>  9: MOV TEMP[3].x, TEMP[2].xxxx
> 10: ADD TEMP[4].xyz, TEMP[2].xxxx, IMM[0].zzzz
> 11: FSLT TEMP[5].xyz, TEMP[4].xyzz, IMM[0].wwww
> 12: OR TEMP[6].x, TEMP[5].xxxx, TEMP[5].zzzz
> 13: OR TEMP[5].x, TEMP[6].xxxx, TEMP[5].yyyy
> 14: AND TEMP[5].x, TEMP[5].xxxx, IMM[0].yyyy
> 15: KILL_IF -TEMP[5].xxxx
> 16: MOV TEMP[5].xyz, IN[0].xyzz
> 17: TEX TEMP[5], TEMP[5], SAMP[2], CUBE
> 18: POW TEMP[6].x, TEMP[5].xxxx, IMM[0].xxxx
> 19: POW TEMP[6].y, TEMP[5].yyyy, IMM[0].xxxx
> 20: POW TEMP[6].z, TEMP[5].zzzz, IMM[0].xxxx
> 21: POW TEMP[6].w, TEMP[5].wwww, IMM[0].yyyy
> 22: MUL TEMP[1].xyz, TEMP[1], TEMP[6]
> 23: MOV TEMP[0].xyz, TEMP[1].xyzx
> 24: ADD TEMP[1].xyz, TEMP[0], TEMP[0]
> 25: MOV TEMP[0].xyz, TEMP[1].xyzx
> 26: ADD TEMP[1].yzw, CONST[3].xxyz, -IN[2].xxyz
> 27: MOV TEMP[3].yzw, TEMP[1].zyzw
> 28: DP3 TEMP[5].x, TEMP[1].yzww, TEMP[1].yzww
> 29: MAX TEMP[5].x, TEMP[5].xxxx, IMM[1].xxxx
> 30: RSQ TEMP[5].x, TEMP[5].xxxx
> 31: MUL TEMP[1].xyz, TEMP[5].xxxx, TEMP[1].yzww
> 32: MAD TEMP[3].yzw, TEMP[3], TEMP[5].xxxx, IMM[1].yyzw
> 33: MOV TEMP[5].w, IMM[0].wwww
> 34: MOV TEMP[5].x, TEMP[3].yyyy
> 35: MOV TEMP[5].y, TEMP[3].zzzz
> 36: MOV TEMP[5].z, TEMP[3].wwww
> 37: DP4 TEMP[3].x, TEMP[5], TEMP[5]
> 38: RSQ TEMP[3].x, TEMP[3].xxxx
> 39: MUL TEMP[3].xyz, TEMP[5], TEMP[3].xxxx
> 40: DP3 TEMP[3].x, TEMP[3].xyzz, IN[0].xyzz
> 41: MOV_SAT TEMP[3].x, TEMP[3].xxxx
> 42: DP3 TEMP[5].x, IN[0].xyzz, TEMP[1].xyzz
> 43: ADD TEMP[5].y, TEMP[5].xxxx, TEMP[5].xxxx
> 44: MAD TEMP[5].yzw, TEMP[5].yyyy, IN[0].xxyz, -TEMP[1].xxyz
> 45: MOV TEMP[6].xyz, TEMP[5].yzww
> 46: TEX TEMP[6], TEMP[6], SAMP[4], CUBE
> 47: POW TEMP[7].x, TEMP[6].xxxx, IMM[0].xxxx
> 48: POW TEMP[7].y, TEMP[6].yyyy, IMM[0].xxxx
> 49: POW TEMP[7].z, TEMP[6].zzzz, IMM[0].xxxx
> 50: POW TEMP[7].w, TEMP[6].wwww, IMM[0].yyyy
> 51: MOV TEMP[4].w, TEMP[7].wwww
> 52: MOV TEMP[6].xy, IN[1].xyyy
> 53: TEX TEMP[6], TEMP[6], SAMP[1], 2D
> 54: POW TEMP[8].x, TEMP[6].xxxx, IMM[0].xxxx
> 55: POW TEMP[8].y, TEMP[6].yyyy, IMM[0].xxxx
> 56: POW TEMP[8].z, TEMP[6].zzzz, IMM[0].xxxx
> 57: POW TEMP[8].w, TEMP[6].wwww, IMM[0].yyyy
> 58: ABS TEMP[6].x, TEMP[8].wwww
> 59: POW TEMP[6].x, TEMP[6].xxxx, IMM[0].xxxx
> 60: MOV TEMP[5].xyz, TEMP[5].yzww
> 61: TEX TEMP[5], TEMP[5], SAMP[3], CUBE
> 62: POW TEMP[9].x, TEMP[5].xxxx, IMM[0].xxxx
> 63: POW TEMP[9].y, TEMP[5].yyyy, IMM[0].xxxx
> 64: POW TEMP[9].z, TEMP[5].zzzz, IMM[0].xxxx
> 65: POW TEMP[9].w, TEMP[5].wwww, IMM[0].yyyy
> 66: LRP TEMP[5].xyz, TEMP[6].xxxx, TEMP[9], TEMP[7]
> 67: MAD TEMP[6].y, TEMP[6].xxxx, IMM[2].xxxx, IMM[0].yyyy
> 68: ABS TEMP[3].x, TEMP[3].xxxx
> 69: POW TEMP[3].x, TEMP[3].xxxx, TEMP[6].yyyy
> 70: MUL TEMP[6].w, TEMP[6].yyyy, IMM[2].yyyy
> 71: MOV TEMP[0].w, TEMP[6].wwww
> 72: MAD TEMP[3].yzw, TEMP[6].wwww, TEMP[3].xxxx, TEMP[5].xxyz
> 73: MAD TEMP[3].xyz, TEMP[3].yzww, TEMP[8], TEMP[0]
> 74: ABS TEMP[5].x, TEMP[3].xxxx
> 75: LG2 TEMP[4].x, TEMP[5].xxxx
> 76: ABS TEMP[5].x, TEMP[3].yyyy
> 77: LG2 TEMP[5].x, TEMP[5].xxxx
> 78: MOV TEMP[4].y, TEMP[5].xxxx
> 79: ABS TEMP[3].x, TEMP[3].zzzz
> 80: LG2 TEMP[3].x, TEMP[3].xxxx
> 81: MOV TEMP[4].z, TEMP[3].xxxx
> 82: MUL TEMP[3].xyz, TEMP[4], IMM[2].zzzz
> 83: EX2 TEMP[4].x, TEMP[3].xxxx
> 84: EX2 TEMP[5].x, TEMP[3].yyyy
> 85: MOV TEMP[4].y, TEMP[5].xxxx
> 86: EX2 TEMP[3].x, TEMP[3].zzzz
> 87: MOV TEMP[4].z, TEMP[3].xxxx
> 88: MOV TEMP[3].xyz, TEMP[4].xyzz
> 89: TEX TEMP[3].xyz, TEMP[3], SAMP[6], 3D
> 90: MAD TEMP[5].xy, IN[4], IMM[3].xyxx, IMM[0].wyww
> 91: MOV TEMP[5].xy, TEMP[5].xyyy
> 92: TEX TEMP[5].xzw, TEMP[5], SAMP[7], 2D
> 93: MOV TEMP[1].w, TEMP[5].wwww
> 94: MOV TEMP[6].xy, IN[1].zwww
> 95: TEX TEMP[6].xyz, TEMP[6], SAMP[8], 2D
> 96: LRP TEMP[3].yzw, TEMP[5].xxxx, TEMP[3].xxyz, TEMP[6].xxyz
> 97: MAD TEMP[5].x, TEMP[5].zzzz, -TEMP[5].xxxx, TEMP[5].xxxx
> 98: MOV TEMP[6].xyz, TEMP[3].yzww
> 99: TEX TEMP[6], TEMP[6], SAMP[9], 3D
>100: LRP TEMP[3].xyz, TEMP[5].xxxx, TEMP[6], TEMP[3].yzww
>101: MOV TEMP[1].xyz, TEMP[3].xyzx
>102: ADD TEMP[3].xyz, -TEMP[1], CONST[5]
>103: MOV TEMP[0].xyz, TEMP[3].xyzx
>104: MUL TEMP[3].z, CONST[8].xxxx, IN[2].zzzz
>105: MOV TEMP[4].z, TEMP[3].zzzz
>106: MOV TEMP[4].xy, IN[2].xyxx
>107: ADD TEMP[3].yzw, TEMP[4].xxyz, -CONST[6].xxyz
>108: MUL TEMP[5].w, TEMP[3].wwww, CONST[4].xxxx
>109: MUL TEMP[5].w, TEMP[5].wwww, IMM[2].wwww
>110: EX2 TEMP[5].x, TEMP[5].wwww
>111: ADD TEMP[5].w, -TEMP[5].xxxx, IMM[0].yyyy
>112: DP3 TEMP[6].x, TEMP[3].yzww, TEMP[3].yzww
>113: RCP TEMP[3].x, TEMP[3].wwww
>114: MUL TEMP[6].y, TEMP[6].xxxx, CONST[4].yyyy
>115: MUL TEMP[5].w, TEMP[5].wwww, TEMP[6].yyyy
>116: MUL TEMP[3].w, TEMP[3].xxxx, TEMP[5].wwww
>117: MUL TEMP[3].w, TEMP[3].wwww, IMM[2].wwww
>118: EX2 TEMP[3].x, TEMP[3].wwww
>119: MOV_SAT TEMP[3].x, TEMP[3].xxxx
>120: ADD TEMP[3].w, -TEMP[3].xxxx, IMM[0].yyyy
>121: MAD TEMP[5].x, IN[3].yyyy, IMM[3].zzzz, IMM[3].zzzz
>122: MOV TEMP[4].x, TEMP[5].xxxx
>123: MOV TEMP[4].y, CONST[4].wwww
>124: MOV TEMP[4].xy, TEMP[4].xyyy
>125: TEX TEMP[4].x, TEMP[4], SAMP[5], 2D
>126: MUL TEMP[3].w, TEMP[3].wwww, TEMP[4].xxxx
>127: MOV TEMP[0].w, TEMP[3].wwww
>128: MAD TEMP[0].xyz, TEMP[3].wwww, TEMP[0], TEMP[1]
>129: MOV TEMP[0].xyz, TEMP[0].xyzx
>130: MUL TEMP[1].x, TEMP[2].xxxx, IMM[3].zzzz
>131: ADD TEMP[3].y, -CONST[7].xxxx, IN[2].zzzz
>132: FSGE TEMP[3].x, TEMP[3].yyyy, IMM[0].wwww
>133: UIF TEMP[3].xxxx :0
>134:   MOV TEMP[2].x, TEMP[2].xxxx
>135: ELSE :0
>136:   MOV TEMP[2].x, TEMP[1].xxxx
>137: ENDIF
>138: MOV TEMP[0].w, TEMP[2].xxxx
>139: MOV OUT[0], TEMP[0]
>140: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %23 = load <16 x i8> addrspace(2)* %22, !tbaa !0
>  %24 = call float @llvm.SI.load.const(<16 x i8> %23, i32 48)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %23, i32 52)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %23, i32 56)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %23, i32 64)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %23, i32 68)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %23, i32 76)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %23, i32 80)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %23, i32 84)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %23, i32 88)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %23, i32 96)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %23, i32 100)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %23, i32 104)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %23, i32 112)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %23, i32 128)
>  %38 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %39 = load <8 x i32> addrspace(2)* %38, !tbaa !0
>  %40 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %41 = load <4 x i32> addrspace(2)* %40, !tbaa !0
>  %42 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 1
>  %43 = load <8 x i32> addrspace(2)* %42, !tbaa !0
>  %44 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 1
>  %45 = load <4 x i32> addrspace(2)* %44, !tbaa !0
>  %46 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 2
>  %47 = load <8 x i32> addrspace(2)* %46, !tbaa !0
>  %48 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 2
>  %49 = load <4 x i32> addrspace(2)* %48, !tbaa !0
>  %50 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 3
>  %51 = load <8 x i32> addrspace(2)* %50, !tbaa !0
>  %52 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 3
>  %53 = load <4 x i32> addrspace(2)* %52, !tbaa !0
>  %54 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 4
>  %55 = load <8 x i32> addrspace(2)* %54, !tbaa !0
>  %56 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 4
>  %57 = load <4 x i32> addrspace(2)* %56, !tbaa !0
>  %58 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 5
>  %59 = load <8 x i32> addrspace(2)* %58, !tbaa !0
>  %60 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 5
>  %61 = load <4 x i32> addrspace(2)* %60, !tbaa !0
>  %62 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 6
>  %63 = load <8 x i32> addrspace(2)* %62, !tbaa !0
>  %64 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 6
>  %65 = load <4 x i32> addrspace(2)* %64, !tbaa !0
>  %66 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 7
>  %67 = load <8 x i32> addrspace(2)* %66, !tbaa !0
>  %68 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 7
>  %69 = load <4 x i32> addrspace(2)* %68, !tbaa !0
>  %70 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 8
>  %71 = load <8 x i32> addrspace(2)* %70, !tbaa !0
>  %72 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 8
>  %73 = load <4 x i32> addrspace(2)* %72, !tbaa !0
>  %74 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 9
>  %75 = load <8 x i32> addrspace(2)* %74, !tbaa !0
>  %76 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 9
>  %77 = load <4 x i32> addrspace(2)* %76, !tbaa !0
>  %78 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %79 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %80 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %5, <2 x i32> %7)
>  %81 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %5, <2 x i32> %7)
>  %82 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %5, <2 x i32> %7)
>  %83 = call float @llvm.SI.fs.interp(i32 2, i32 1, i32 %5, <2 x i32> %7)
>  %84 = call float @llvm.SI.fs.interp(i32 3, i32 1, i32 %5, <2 x i32> %7)
>  %85 = call float @llvm.SI.fs.interp(i32 0, i32 2, i32 %5, <2 x i32> %7)
>  %86 = call float @llvm.SI.fs.interp(i32 1, i32 2, i32 %5, <2 x i32> %7)
>  %87 = call float @llvm.SI.fs.interp(i32 2, i32 2, i32 %5, <2 x i32> %7)
>  %88 = call float @llvm.SI.fs.interp(i32 1, i32 3, i32 %5, <2 x i32> %7)
>  %89 = call float @llvm.SI.fs.interp(i32 0, i32 4, i32 %5, <2 x i32> %7)
>  %90 = call float @llvm.SI.fs.interp(i32 1, i32 4, i32 %5, <2 x i32> %7)
>  %91 = bitcast float %81 to i32
>  %92 = bitcast float %82 to i32
>  %93 = insertelement <2 x i32> undef, i32 %91, i32 0
>  %94 = insertelement <2 x i32> %93, i32 %92, i32 1
>  %95 = bitcast <8 x i32> %39 to <32 x i8>
>  %96 = bitcast <4 x i32> %41 to <16 x i8>
>  %97 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %94, <32 x i8> %95, <16 x i8> %96, i32 2)
>  %98 = extractelement <4 x float> %97, i32 0
>  %99 = extractelement <4 x float> %97, i32 1
>  %100 = extractelement <4 x float> %97, i32 2
>  %101 = extractelement <4 x float> %97, i32 3
>  %102 = call float @llvm.pow.f32(float %98, float 0x40019999A0000000)
>  %103 = call float @llvm.pow.f32(float %99, float 0x40019999A0000000)
>  %104 = call float @llvm.pow.f32(float %100, float 0x40019999A0000000)
>  %105 = call float @llvm.pow.f32(float %101, float 1.000000e+00)
>  %106 = call float @fabs(float %105)
>  %107 = call float @llvm.pow.f32(float %106, float 0x40019999A0000000)
>  %108 = fadd float %107, 0xBFB99999A0000000
>  %109 = fadd float %107, 0xBFB99999A0000000
>  %110 = fadd float %107, 0xBFB99999A0000000
>  %111 = fcmp olt float %108, 0.000000e+00
>  %112 = sext i1 %111 to i32
>  %113 = fcmp olt float %109, 0.000000e+00
>  %114 = sext i1 %113 to i32
>  %115 = fcmp olt float %110, 0.000000e+00
>  %116 = sext i1 %115 to i32
>  %117 = bitcast i32 %112 to float
>  %118 = bitcast i32 %114 to float
>  %119 = bitcast i32 %116 to float
>  %120 = bitcast float %117 to i32
>  %121 = bitcast float %119 to i32
>  %122 = or i32 %120, %121
>  %123 = bitcast i32 %122 to float
>  %124 = bitcast float %123 to i32
>  %125 = bitcast float %118 to i32
>  %126 = or i32 %124, %125
>  %127 = bitcast i32 %126 to float
>  %128 = bitcast float %127 to i32
>  %129 = and i32 %128, 1065353216
>  %130 = bitcast i32 %129 to float
>  %131 = fsub float -0.000000e+00, %130
>  %132 = fsub float -0.000000e+00, %130
>  %133 = fsub float -0.000000e+00, %130
>  %134 = fsub float -0.000000e+00, %130
>  call void @llvm.AMDGPU.kill(float %131)
>  call void @llvm.AMDGPU.kill(float %132)
>  call void @llvm.AMDGPU.kill(float %133)
>  call void @llvm.AMDGPU.kill(float %134)
>  %135 = insertelement <4 x float> undef, float %78, i32 0
>  %136 = insertelement <4 x float> %135, float %79, i32 1
>  %137 = insertelement <4 x float> %136, float %80, i32 2
>  %138 = insertelement <4 x float> %137, float 0.000000e+00, i32 3
>  %139 = call <4 x float> @llvm.AMDGPU.cube(<4 x float> %138)
>  %140 = extractelement <4 x float> %139, i32 0
>  %141 = extractelement <4 x float> %139, i32 1
>  %142 = extractelement <4 x float> %139, i32 2
>  %143 = extractelement <4 x float> %139, i32 3
>  %144 = call float @fabs(float %142)
>  %145 = fdiv float 1.000000e+00, %144
>  %146 = fmul float %140, %145
>  %147 = fadd float %146, 1.500000e+00
>  %148 = fmul float %141, %145
>  %149 = fadd float %148, 1.500000e+00
>  %150 = bitcast float %149 to i32
>  %151 = bitcast float %147 to i32
>  %152 = bitcast float %143 to i32
>  %153 = insertelement <4 x i32> undef, i32 %150, i32 0
>  %154 = insertelement <4 x i32> %153, i32 %151, i32 1
>  %155 = insertelement <4 x i32> %154, i32 %152, i32 2
>  %156 = insertelement <4 x i32> %155, i32 undef, i32 3
>  %157 = bitcast <8 x i32> %47 to <32 x i8>
>  %158 = bitcast <4 x i32> %49 to <16 x i8>
>  %159 = call <4 x float> @llvm.SI.sample.v4i32(<4 x i32> %156, <32 x i8> %157, <16 x i8> %158, i32 4)
>  %160 = extractelement <4 x float> %159, i32 0
>  %161 = extractelement <4 x float> %159, i32 1
>  %162 = extractelement <4 x float> %159, i32 2
>  %163 = extractelement <4 x float> %159, i32 3
>  %164 = call float @llvm.pow.f32(float %160, float 0x40019999A0000000)
>  %165 = call float @llvm.pow.f32(float %161, float 0x40019999A0000000)
>  %166 = call float @llvm.pow.f32(float %162, float 0x40019999A0000000)
>  %167 = call float @llvm.pow.f32(float %163, float 1.000000e+00)
>  %168 = fmul float %102, %164
>  %169 = fmul float %103, %165
>  %170 = fmul float %104, %166
>  %171 = fadd float %168, %168
>  %172 = fadd float %169, %169
>  %173 = fadd float %170, %170
>  %174 = fsub float -0.000000e+00, %85
>  %175 = fadd float %24, %174
>  %176 = fsub float -0.000000e+00, %86
>  %177 = fadd float %25, %176
>  %178 = fsub float -0.000000e+00, %87
>  %179 = fadd float %26, %178
>  %180 = fmul float %175, %175
>  %181 = fmul float %177, %177
>  %182 = fadd float %181, %180
>  %183 = fmul float %179, %179
>  %184 = fadd float %182, %183
>  %185 = call float @llvm.maxnum.f32(float %184, float 0x3E7AD7F2A0000000)
>  %186 = call float @llvm.AMDGPU.rsq.clamped.f32(float %185)
>  %187 = fmul float %186, %175
>  %188 = fmul float %186, %177
>  %189 = fmul float %186, %179
>  %190 = fmul float %175, %186
>  %191 = fadd float %190, 0x3FAB15B580000000
>  %192 = fmul float %177, %186
>  %193 = fadd float %192, 0x3FEB126EA0000000
>  %194 = fmul float %179, %186
>  %195 = fadd float %194, 0x3FE0ED9160000000
>  %196 = fmul float %191, %191
>  %197 = fmul float %193, %193
>  %198 = fadd float %196, %197
>  %199 = fmul float %195, %195
>  %200 = fadd float %198, %199
>  %201 = fmul float 0.000000e+00, 0.000000e+00
>  %202 = fadd float %200, %201
>  %203 = call float @llvm.AMDGPU.rsq.clamped.f32(float %202)
>  %204 = fmul float %191, %203
>  %205 = fmul float %193, %203
>  %206 = fmul float %195, %203
>  %207 = fmul float %204, %78
>  %208 = fmul float %205, %79
>  %209 = fadd float %208, %207
>  %210 = fmul float %206, %80
>  %211 = fadd float %209, %210
>  %212 = call float @llvm.AMDIL.clamp.(float %211, float 0.000000e+00, float 1.000000e+00)
>  %213 = fmul float %78, %187
>  %214 = fmul float %79, %188
>  %215 = fadd float %214, %213
>  %216 = fmul float %80, %189
>  %217 = fadd float %215, %216
>  %218 = fadd float %217, %217
>  %219 = fsub float -0.000000e+00, %187
>  %220 = fmul float %218, %78
>  %221 = fadd float %220, %219
>  %222 = fsub float -0.000000e+00, %188
>  %223 = fmul float %218, %79
>  %224 = fadd float %223, %222
>  %225 = fsub float -0.000000e+00, %189
>  %226 = fmul float %218, %80
>  %227 = fadd float %226, %225
>  %228 = insertelement <4 x float> undef, float %221, i32 0
>  %229 = insertelement <4 x float> %228, float %224, i32 1
>  %230 = insertelement <4 x float> %229, float %227, i32 2
>  %231 = insertelement <4 x float> %230, float %167, i32 3
>  %232 = call <4 x float> @llvm.AMDGPU.cube(<4 x float> %231)
>  %233 = extractelement <4 x float> %232, i32 0
>  %234 = extractelement <4 x float> %232, i32 1
>  %235 = extractelement <4 x float> %232, i32 2
>  %236 = extractelement <4 x float> %232, i32 3
>  %237 = call float @fabs(float %235)
>  %238 = fdiv float 1.000000e+00, %237
>  %239 = fmul float %233, %238
>  %240 = fadd float %239, 1.500000e+00
>  %241 = fmul float %234, %238
>  %242 = fadd float %241, 1.500000e+00
>  %243 = bitcast float %242 to i32
>  %244 = bitcast float %240 to i32
>  %245 = bitcast float %236 to i32
>  %246 = insertelement <4 x i32> undef, i32 %243, i32 0
>  %247 = insertelement <4 x i32> %246, i32 %244, i32 1
>  %248 = insertelement <4 x i32> %247, i32 %245, i32 2
>  %249 = insertelement <4 x i32> %248, i32 undef, i32 3
>  %250 = bitcast <8 x i32> %55 to <32 x i8>
>  %251 = bitcast <4 x i32> %57 to <16 x i8>
>  %252 = call <4 x float> @llvm.SI.sample.v4i32(<4 x i32> %249, <32 x i8> %250, <16 x i8> %251, i32 4)
>  %253 = extractelement <4 x float> %252, i32 0
>  %254 = extractelement <4 x float> %252, i32 1
>  %255 = extractelement <4 x float> %252, i32 2
>  %256 = call float @llvm.pow.f32(float %253, float 0x40019999A0000000)
>  %257 = call float @llvm.pow.f32(float %254, float 0x40019999A0000000)
>  %258 = call float @llvm.pow.f32(float %255, float 0x40019999A0000000)
>  %259 = bitcast float %81 to i32
>  %260 = bitcast float %82 to i32
>  %261 = insertelement <2 x i32> undef, i32 %259, i32 0
>  %262 = insertelement <2 x i32> %261, i32 %260, i32 1
>  %263 = bitcast <8 x i32> %43 to <32 x i8>
>  %264 = bitcast <4 x i32> %45 to <16 x i8>
>  %265 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %262, <32 x i8> %263, <16 x i8> %264, i32 2)
>  %266 = extractelement <4 x float> %265, i32 0
>  %267 = extractelement <4 x float> %265, i32 1
>  %268 = extractelement <4 x float> %265, i32 2
>  %269 = extractelement <4 x float> %265, i32 3
>  %270 = call float @llvm.pow.f32(float %266, float 0x40019999A0000000)
>  %271 = call float @llvm.pow.f32(float %267, float 0x40019999A0000000)
>  %272 = call float @llvm.pow.f32(float %268, float 0x40019999A0000000)
>  %273 = call float @llvm.pow.f32(float %269, float 1.000000e+00)
>  %274 = call float @fabs(float %273)
>  %275 = call float @llvm.pow.f32(float %274, float 0x40019999A0000000)
>  %276 = insertelement <4 x float> undef, float %221, i32 0
>  %277 = insertelement <4 x float> %276, float %224, i32 1
>  %278 = insertelement <4 x float> %277, float %227, i32 2
>  %279 = insertelement <4 x float> %278, float %227, i32 3
>  %280 = call <4 x float> @llvm.AMDGPU.cube(<4 x float> %279)
>  %281 = extractelement <4 x float> %280, i32 0
>  %282 = extractelement <4 x float> %280, i32 1
>  %283 = extractelement <4 x float> %280, i32 2
>  %284 = extractelement <4 x float> %280, i32 3
>  %285 = call float @fabs(float %283)
>  %286 = fdiv float 1.000000e+00, %285
>  %287 = fmul float %281, %286
>  %288 = fadd float %287, 1.500000e+00
>  %289 = fmul float %282, %286
>  %290 = fadd float %289, 1.500000e+00
>  %291 = bitcast float %290 to i32
>  %292 = bitcast float %288 to i32
>  %293 = bitcast float %284 to i32
>  %294 = insertelement <4 x i32> undef, i32 %291, i32 0
>  %295 = insertelement <4 x i32> %294, i32 %292, i32 1
>  %296 = insertelement <4 x i32> %295, i32 %293, i32 2
>  %297 = insertelement <4 x i32> %296, i32 undef, i32 3
>  %298 = bitcast <8 x i32> %51 to <32 x i8>
>  %299 = bitcast <4 x i32> %53 to <16 x i8>
>  %300 = call <4 x float> @llvm.SI.sample.v4i32(<4 x i32> %297, <32 x i8> %298, <16 x i8> %299, i32 4)
>  %301 = extractelement <4 x float> %300, i32 0
>  %302 = extractelement <4 x float> %300, i32 1
>  %303 = extractelement <4 x float> %300, i32 2
>  %304 = call float @llvm.pow.f32(float %301, float 0x40019999A0000000)
>  %305 = call float @llvm.pow.f32(float %302, float 0x40019999A0000000)
>  %306 = call float @llvm.pow.f32(float %303, float 0x40019999A0000000)
>  %307 = call float @llvm.AMDGPU.lrp(float %275, float %304, float %256)
>  %308 = call float @llvm.AMDGPU.lrp(float %275, float %305, float %257)
>  %309 = call float @llvm.AMDGPU.lrp(float %275, float %306, float %258)
>  %310 = fmul float %275, 1.990000e+02
>  %311 = fadd float %310, 1.000000e+00
>  %312 = call float @fabs(float %212)
>  %313 = call float @llvm.pow.f32(float %312, float %311)
>  %314 = fmul float %311, 0x3FB99999A0000000
>  %315 = fmul float %314, %313
>  %316 = fadd float %315, %307
>  %317 = fmul float %314, %313
>  %318 = fadd float %317, %308
>  %319 = fmul float %314, %313
>  %320 = fadd float %319, %309
>  %321 = fmul float %316, %270
>  %322 = fadd float %321, %171
>  %323 = fmul float %318, %271
>  %324 = fadd float %323, %172
>  %325 = fmul float %320, %272
>  %326 = fadd float %325, %173
>  %327 = call float @fabs(float %322)
>  %328 = call float @llvm.log2.f32(float %327)
>  %329 = call float @fabs(float %324)
>  %330 = call float @llvm.log2.f32(float %329)
>  %331 = call float @fabs(float %326)
>  %332 = call float @llvm.log2.f32(float %331)
>  %333 = fmul float %328, 0x3FDD1743E0000000
>  %334 = fmul float %330, 0x3FDD1743E0000000
>  %335 = fmul float %332, 0x3FDD1743E0000000
>  %336 = call float @llvm.AMDIL.exp.(float %333)
>  %337 = call float @llvm.AMDIL.exp.(float %334)
>  %338 = call float @llvm.AMDIL.exp.(float %335)
>  %339 = bitcast float %336 to i32
>  %340 = bitcast float %337 to i32
>  %341 = bitcast float %338 to i32
>  %342 = insertelement <4 x i32> undef, i32 %339, i32 0
>  %343 = insertelement <4 x i32> %342, i32 %340, i32 1
>  %344 = insertelement <4 x i32> %343, i32 %341, i32 2
>  %345 = insertelement <4 x i32> %344, i32 undef, i32 3
>  %346 = bitcast <8 x i32> %63 to <32 x i8>
>  %347 = bitcast <4 x i32> %65 to <16 x i8>
>  %348 = call <4 x float> @llvm.SI.sample.v4i32(<4 x i32> %345, <32 x i8> %346, <16 x i8> %347, i32 3)
>  %349 = extractelement <4 x float> %348, i32 0
>  %350 = extractelement <4 x float> %348, i32 1
>  %351 = extractelement <4 x float> %348, i32 2
>  %352 = fmul float %89, 1.000000e+00
>  %353 = fadd float %352, 0.000000e+00
>  %354 = fmul float %90, -1.000000e+00
>  %355 = fadd float %354, 1.000000e+00
>  %356 = bitcast float %353 to i32
>  %357 = bitcast float %355 to i32
>  %358 = insertelement <2 x i32> undef, i32 %356, i32 0
>  %359 = insertelement <2 x i32> %358, i32 %357, i32 1
>  %360 = bitcast <8 x i32> %67 to <32 x i8>
>  %361 = bitcast <4 x i32> %69 to <16 x i8>
>  %362 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %359, <32 x i8> %360, <16 x i8> %361, i32 2)
>  %363 = extractelement <4 x float> %362, i32 0
>  %364 = extractelement <4 x float> %362, i32 2
>  %365 = bitcast float %83 to i32
>  %366 = bitcast float %84 to i32
>  %367 = insertelement <2 x i32> undef, i32 %365, i32 0
>  %368 = insertelement <2 x i32> %367, i32 %366, i32 1
>  %369 = bitcast <8 x i32> %71 to <32 x i8>
>  %370 = bitcast <4 x i32> %73 to <16 x i8>
>  %371 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %368, <32 x i8> %369, <16 x i8> %370, i32 2)
>  %372 = extractelement <4 x float> %371, i32 0
>  %373 = extractelement <4 x float> %371, i32 1
>  %374 = extractelement <4 x float> %371, i32 2
>  %375 = call float @llvm.AMDGPU.lrp(float %363, float %349, float %372)
>  %376 = call float @llvm.AMDGPU.lrp(float %363, float %350, float %373)
>  %377 = call float @llvm.AMDGPU.lrp(float %363, float %351, float %374)
>  %378 = fsub float -0.000000e+00, %363
>  %379 = fmul float %364, %378
>  %380 = fadd float %379, %363
>  %381 = bitcast float %375 to i32
>  %382 = bitcast float %376 to i32
>  %383 = bitcast float %377 to i32
>  %384 = insertelement <4 x i32> undef, i32 %381, i32 0
>  %385 = insertelement <4 x i32> %384, i32 %382, i32 1
>  %386 = insertelement <4 x i32> %385, i32 %383, i32 2
>  %387 = insertelement <4 x i32> %386, i32 undef, i32 3
>  %388 = bitcast <8 x i32> %75 to <32 x i8>
>  %389 = bitcast <4 x i32> %77 to <16 x i8>
>  %390 = call <4 x float> @llvm.SI.sample.v4i32(<4 x i32> %387, <32 x i8> %388, <16 x i8> %389, i32 3)
>  %391 = extractelement <4 x float> %390, i32 0
>  %392 = extractelement <4 x float> %390, i32 1
>  %393 = extractelement <4 x float> %390, i32 2
>  %394 = call float @llvm.AMDGPU.lrp(float %380, float %391, float %375)
>  %395 = call float @llvm.AMDGPU.lrp(float %380, float %392, float %376)
>  %396 = call float @llvm.AMDGPU.lrp(float %380, float %393, float %377)
>  %397 = fsub float -0.000000e+00, %394
>  %398 = fadd float %397, %30
>  %399 = fsub float -0.000000e+00, %395
>  %400 = fadd float %399, %31
>  %401 = fsub float -0.000000e+00, %396
>  %402 = fadd float %401, %32
>  %403 = fmul float %37, %87
>  %404 = fsub float -0.000000e+00, %33
>  %405 = fadd float %85, %404
>  %406 = fsub float -0.000000e+00, %34
>  %407 = fadd float %86, %406
>  %408 = fsub float -0.000000e+00, %35
>  %409 = fadd float %403, %408
>  %410 = fmul float %409, %27
>  %411 = fmul float %410, 0x3FF7154CA0000000
>  %412 = call float @llvm.AMDIL.exp.(float %411)
>  %413 = fsub float -0.000000e+00, %412
>  %414 = fadd float %413, 1.000000e+00
>  %415 = fmul float %405, %405
>  %416 = fmul float %407, %407
>  %417 = fadd float %416, %415
>  %418 = fmul float %409, %409
>  %419 = fadd float %417, %418
>  %420 = fdiv float 1.000000e+00, %409
>  %421 = fmul float %419, %28
>  %422 = fmul float %414, %421
>  %423 = fmul float %420, %422
>  %424 = fmul float %423, 0x3FF7154CA0000000
>  %425 = call float @llvm.AMDIL.exp.(float %424)
>  %426 = call float @llvm.AMDIL.clamp.(float %425, float 0.000000e+00, float 1.000000e+00)
>  %427 = fsub float -0.000000e+00, %426
>  %428 = fadd float %427, 1.000000e+00
>  %429 = fmul float %88, 5.000000e-01
>  %430 = fadd float %429, 5.000000e-01
>  %431 = bitcast float %430 to i32
>  %432 = bitcast float %29 to i32
>  %433 = insertelement <2 x i32> undef, i32 %431, i32 0
>  %434 = insertelement <2 x i32> %433, i32 %432, i32 1
>  %435 = bitcast <8 x i32> %59 to <32 x i8>
>  %436 = bitcast <4 x i32> %61 to <16 x i8>
>  %437 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %434, <32 x i8> %435, <16 x i8> %436, i32 2)
>  %438 = extractelement <4 x float> %437, i32 0
>  %439 = fmul float %428, %438
>  %440 = fmul float %439, %398
>  %441 = fadd float %440, %394
>  %442 = fmul float %439, %400
>  %443 = fadd float %442, %395
>  %444 = fmul float %439, %402
>  %445 = fadd float %444, %396
>  %446 = fmul float %107, 5.000000e-01
>  %447 = fsub float -0.000000e+00, %36
>  %448 = fadd float %447, %87
>  %449 = fcmp oge float %448, 0.000000e+00
>  %450 = sext i1 %449 to i32
>  %451 = bitcast i32 %450 to float
>  %452 = bitcast float %451 to i32
>  %453 = icmp ne i32 %452, 0
>  %. = select i1 %453, float %107, float %446
>  %454 = call i32 @llvm.SI.packf16(float %441, float %443)
>  %455 = bitcast i32 %454 to float
>  %456 = call i32 @llvm.SI.packf16(float %445, float %.)
>  %457 = bitcast i32 %456 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %455, float %457, float %455, float %457)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.pow.f32(float, float) #1
>
>; Function Attrs: readnone
>declare float @fabs(float) #2
>
>declare void @llvm.AMDGPU.kill(float)
>
>; Function Attrs: readnone
>declare <4 x float> @llvm.AMDGPU.cube(<4 x float>) #2
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v4i32(<4 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.maxnum.f32(float, float) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.AMDGPU.rsq.clamped.f32(float) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.clamp.(float, float, float) #2
>
>; Function Attrs: readnone
>declare float @llvm.AMDGPU.lrp(float, float, float) #2
>
>; Function Attrs: nounwind readnone
>declare float @llvm.log2.f32(float) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.exp.(float) #2
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                         ; BEFE0A7E
>	s_mov_b64 vcc, s[6:7]                                                        ; BEEA0406
>	s_mov_b64 s[100:101], s[2:3]                                                 ; BEE40402
>	s_mov_b32 m0, s9                                                             ; BEFC0309
>	v_interp_p1_f32 v3, v0, 1, 1, [m0]                                           ; C80C0500
>	v_interp_p2_f32 v3, [v3], v1, 1, 1, [m0]                                     ; C80D0501
>	v_interp_p1_f32 v2, v0, 0, 1, [m0]                                           ; C8080400
>	v_interp_p2_f32 v2, [v2], v1, 0, 1, [m0]                                     ; C8090401
>	s_load_dwordx4 s[8:11], s[4:5], 0x0                                          ; C0840500
>	s_load_dwordx4 s[64:67], s[4:5], 0x4                                         ; C0A00504
>	s_load_dwordx4 s[88:91], s[4:5], 0x8                                         ; C0AC0508
>	s_load_dwordx4 s[52:55], s[4:5], 0xc                                         ; C09A050C
>	s_load_dwordx4 s[68:71], s[4:5], 0x10                                        ; C0A20510
>	s_load_dwordx4 s[0:3], s[4:5], 0x14                                          ; C0800514
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_writelane_b32 v47, s0, 0                                                   ; 045F0000
>	v_writelane_b32 v47, s1, 1                                                   ; 045F0201
>	v_writelane_b32 v47, s2, 2                                                   ; 045F0402
>	v_writelane_b32 v47, s3, 3                                                   ; 045F0603
>	s_load_dwordx4 s[40:43], s[4:5], 0x18                                        ; C0940518
>	s_load_dwordx4 s[28:31], s[4:5], 0x1c                                        ; C08E051C
>	s_load_dwordx4 s[24:27], s[4:5], 0x20                                        ; C08C0520
>	s_load_dwordx4 s[20:23], s[4:5], 0x24                                        ; C08A0524
>	s_load_dwordx8 s[0:7], vcc, 0x0                                              ; C0C06B00
>	s_load_dwordx8 s[72:79], vcc, 0x8                                            ; C0E46B08
>	s_load_dwordx8 s[92:99], vcc, 0x10                                           ; C0EE6B10
>	s_load_dwordx8 s[56:63], vcc, 0x18                                           ; C0DC6B18
>	s_load_dwordx8 s[80:87], vcc, 0x20                                           ; C0E86B20
>	s_load_dwordx8 s[12:19], vcc, 0x28                                           ; C0C66B28
>	s_load_dwordx8 s[44:51], vcc, 0x30                                           ; C0D66B30
>	s_load_dwordx8 s[32:39], vcc, 0x38                                           ; C0D06B38
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample v[4:7], 15, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[0:7], s[8:11]        ; F0800F00 00400402
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_log_f32_e32 v8, v7                                                         ; 7E104F07
>	v_mul_legacy_f32_e32 v8, 1.0, v8                                             ; 0E1010F2
>	v_exp_f32_e32 v8, v8                                                         ; 7E104B08
>	v_mov_b32_e32 v9, 0x7fffffff                                                 ; 7E1202FF 7FFFFFFF
>	v_and_b32_e32 v8, v8, v9                                                     ; 36101308
>	v_log_f32_e32 v8, v8                                                         ; 7E104F08
>	v_mov_b32_e32 v10, 0x400ccccd                                                ; 7E1402FF 400CCCCD
>	v_mul_legacy_f32_e32 v8, v10, v8                                             ; 0E10110A
>	v_exp_f32_e32 v8, v8                                                         ; 7E104B08
>	v_mov_b32_e32 v11, 0xbdcccccd                                                ; 7E1602FF BDCCCCCD
>	v_add_f32_e32 v11, v8, v11                                                   ; 06161708
>	v_mov_b32_e32 v15, 0                                                         ; 7E1E0280
>	v_cmp_lt_f32_e64 s[0:1], v11, 0                                              ; D0020000 0001010B
>	v_cndmask_b32_e64 v11, 0, -1, s[0:1]                                         ; D200080B 00018280
>	v_and_b32_e32 v11, 1.0, v11                                                  ; 361616F2
>	v_xor_b32_e32 v11, 0x80000000, v11                                           ; 3A1616FF 80000000
>	v_cmpx_le_f32_e32 vcc, 0, v11                                                ; 7C261680
>	v_cmpx_le_f32_e32 vcc, 0, v11                                                ; 7C261680
>	v_cmpx_le_f32_e32 vcc, 0, v11                                                ; 7C261680
>	v_cmpx_le_f32_e32 vcc, 0, v11                                                ; 7C261680
>	v_interp_p1_f32 v14, v0, 2, 0, [m0]                                          ; C8380200
>	v_interp_p2_f32 v14, [v14], v1, 2, 0, [m0]                                   ; C8390201
>	v_interp_p1_f32 v13, v0, 1, 0, [m0]                                          ; C8340100
>	v_interp_p2_f32 v13, [v13], v1, 1, 0, [m0]                                   ; C8350101
>	v_interp_p1_f32 v12, v0, 0, 0, [m0]                                          ; C8300000
>	v_interp_p2_f32 v12, [v12], v1, 0, 0, [m0]                                   ; C8310001
>	v_cubeid_f32 v19, v12, v13, v14                                              ; D2880013 043A1B0C
>	v_cubema_f32 v18, v12, v13, v14                                              ; D28E0012 043A1B0C
>	v_cubesc_f32 v17, v12, v13, v14                                              ; D28A0011 043A1B0C
>	v_cubetc_f32 v16, v12, v13, v14                                              ; D28C0010 043A1B0C
>	v_rcp_f32_e64 v11, |v18|                                                     ; D354010B 00000112
>	v_mov_b32_e32 v24, 0x3fc00000                                                ; 7E3002FF 3FC00000
>	v_mad_f32 v18, v16, v11, v24                                                 ; D2820012 04621710
>	v_mad_f32 v17, v17, v11, v24                                                 ; D2820011 04621711
>	image_sample v[16:19], 15, 0, 0, 0, 0, 0, 0, 0, v[17:20], s[92:99], s[88:91] ; F0800F00 02D71011
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_log_f32_e32 v11, v19                                                       ; 7E164F13
>	v_mul_legacy_f32_e32 v11, 1.0, v11                                           ; 0E1616F2
>	v_exp_f32_e32 v23, v11                                                       ; 7E2E4B0B
>	v_interp_p1_f32 v11, v0, 1, 2, [m0]                                          ; C82C0900
>	v_interp_p2_f32 v11, [v11], v1, 1, 2, [m0]                                   ; C82D0901
>	s_load_dwordx4 s[0:3], s[100:101], 0x0                                       ; C0806500
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	s_buffer_load_dword s4, s[0:3], 0xd                                          ; C202010D
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v25, s4, v11                                                   ; 08321604
>	v_interp_p1_f32 v26, v0, 0, 2, [m0]                                          ; C8680800
>	v_interp_p2_f32 v26, [v26], v1, 0, 2, [m0]                                   ; C8690801
>	s_buffer_load_dword s4, s[0:3], 0xc                                          ; C202010C
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v27, s4, v26                                                   ; 08363404
>	v_mul_f32_e32 v28, v27, v27                                                  ; 1038371B
>	v_mad_f32 v28, v25, v25, v28                                                 ; D282001C 04723319
>	v_interp_p1_f32 v29, v0, 2, 2, [m0]                                          ; C8740A00
>	v_interp_p2_f32 v29, [v29], v1, 2, 2, [m0]                                   ; C8750A01
>	s_buffer_load_dword s4, s[0:3], 0xe                                          ; C202010E
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v30, s4, v29                                                   ; 083C3A04
>	v_mad_f32 v28, v30, v30, v28                                                 ; D282001C 04723D1E
>	v_max_f32_e32 v28, 0x33d6bf95, v28                                           ; 203838FF 33D6BF95
>	v_rsq_clamp_f32_e32 v28, v28                                                 ; 7E38591C
>	v_mul_f32_e32 v31, v25, v28                                                  ; 103E3919
>	v_mul_f32_e32 v32, v27, v28                                                  ; 1040391B
>	v_mul_f32_e32 v33, v32, v12                                                  ; 10421920
>	v_mad_f32 v33, v13, v31, v33                                                 ; D2820021 04863F0D
>	v_mul_f32_e32 v34, v30, v28                                                  ; 1044391E
>	v_mad_f32 v33, v14, v34, v33                                                 ; D2820021 0486450E
>	v_add_f32_e32 v33, v33, v33                                                  ; 06424321
>	v_mad_f32 v22, v33, v14, -v34                                                ; D2820016 848A1D21
>	v_mad_f32 v21, v33, v13, -v31                                                ; D2820015 847E1B21
>	v_mad_f32 v20, v33, v12, -v32                                                ; D2820014 84821921
>	v_cubeid_f32 v34, v20, v21, v22                                              ; D2880022 045A2B14
>	v_cubema_f32 v33, v20, v21, v22                                              ; D28E0021 045A2B14
>	v_cubesc_f32 v32, v20, v21, v22                                              ; D28A0020 045A2B14
>	v_cubetc_f32 v31, v20, v21, v22                                              ; D28C001F 045A2B14
>	v_rcp_f32_e64 v39, |v33|                                                     ; D3540127 00000121
>	v_mad_f32 v33, v31, v39, v24                                                 ; D2820021 04624F1F
>	v_mad_f32 v32, v32, v39, v24                                                 ; D2820020 04624F20
>	image_sample v[31:33], 7, 0, 0, 0, 0, 0, 0, 0, v[32:35], s[80:87], s[68:71]  ; F0800700 02341F20
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_log_f32_e32 v34, v33                                                       ; 7E444F21
>	v_mul_legacy_f32_e32 v34, v10, v34                                           ; 0E44450A
>	v_exp_f32_e32 v34, v34                                                       ; 7E444B22
>	image_sample v[35:38], 15, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[72:79], s[64:67]   ; F0800F00 02122302
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_log_f32_e32 v2, v38                                                        ; 7E044F26
>	v_mul_legacy_f32_e32 v2, 1.0, v2                                             ; 0E0404F2
>	v_exp_f32_e32 v2, v2                                                         ; 7E044B02
>	v_and_b32_e32 v2, v2, v9                                                     ; 36041302
>	v_log_f32_e32 v2, v2                                                         ; 7E044F02
>	v_mul_legacy_f32_e32 v2, v10, v2                                             ; 0E04050A
>	v_exp_f32_e32 v2, v2                                                         ; 7E044B02
>	v_sub_f32_e32 v3, 1.0, v2                                                    ; 080604F2
>	v_mul_f32_e32 v34, v34, v3                                                   ; 10440722
>	v_mov_b32_e32 v23, v22                                                       ; 7E2E0316
>	v_cubeid_f32 v42, v20, v21, v22                                              ; D288002A 045A2B14
>	v_cubema_f32 v41, v20, v21, v22                                              ; D28E0029 045A2B14
>	v_cubesc_f32 v40, v20, v21, v22                                              ; D28A0028 045A2B14
>	v_cubetc_f32 v39, v20, v21, v22                                              ; D28C0027 045A2B14
>	v_rcp_f32_e64 v20, |v41|                                                     ; D3540114 00000129
>	v_mad_f32 v41, v39, v20, v24                                                 ; D2820029 04622927
>	v_mad_f32 v40, v40, v20, v24                                                 ; D2820028 04622928
>	image_sample v[20:22], 7, 0, 0, 0, 0, 0, 0, 0, v[40:43], s[56:63], s[52:55]  ; F0800700 01AE1428
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_log_f32_e32 v23, v22                                                       ; 7E2E4F16
>	v_mul_legacy_f32_e32 v23, v10, v23                                           ; 0E2E2F0A
>	v_exp_f32_e32 v23, v23                                                       ; 7E2E4B17
>	v_mad_f32 v23, v2, v23, v34                                                  ; D2820017 048A2F02
>	v_mov_b32_e32 v24, 0x3d58adac                                                ; 7E3002FF 3D58ADAC
>	v_mad_f32 v24, v27, v28, v24                                                 ; D2820018 0462391B
>	v_mov_b32_e32 v27, 0x3f589375                                                ; 7E3602FF 3F589375
>	v_mad_f32 v25, v25, v28, v27                                                 ; D2820019 046E3919
>	v_mul_f32_e32 v27, v25, v25                                                  ; 10363319
>	v_mad_f32 v27, v24, v24, v27                                                 ; D282001B 046E3118
>	v_mov_b32_e32 v34, 0x3f076c8b                                                ; 7E4402FF 3F076C8B
>	v_mad_f32 v28, v30, v28, v34                                                 ; D282001C 048A391E
>	v_mad_f32 v27, v28, v28, v27                                                 ; D282001B 046E391C
>	v_add_f32_e32 v27, 0, v27                                                    ; 06363680
>	v_rsq_clamp_f32_e32 v27, v27                                                 ; 7E36591B
>	v_mul_f32_e32 v25, v27, v25                                                  ; 1032331B
>	v_mul_f32_e32 v24, v27, v24                                                  ; 1030311B
>	v_mul_f32_e32 v24, v12, v24                                                  ; 1030310C
>	v_mad_f32 v24, v25, v13, v24                                                 ; D2820018 04621B19
>	v_mul_f32_e32 v25, v27, v28                                                  ; 1032391B
>	v_mad_f32 v12, v25, v14, v24                                                 ; D282000C 04621D19
>	v_add_f32_e64 v12, 0, v12 clamp                                              ; D206080C 00021880
>	v_and_b32_e32 v9, v12, v9                                                    ; 3612130C
>	v_log_f32_e32 v9, v9                                                         ; 7E124F09
>	v_mov_b32_e32 v12, 0x43470000                                                ; 7E1802FF 43470000
>	v_mad_f32 v12, v12, v2, 1.0                                                  ; D282000C 03CA050C
>	v_mul_legacy_f32_e32 v9, v12, v9                                             ; 0E12130C
>	v_exp_f32_e32 v9, v9                                                         ; 7E124B09
>	v_mul_f32_e32 v12, 0x3dcccccd, v12                                           ; 101818FF 3DCCCCCD
>	v_mad_f32 v13, v12, v9, v23                                                  ; D282000D 045E130C
>	v_log_f32_e32 v14, v37                                                       ; 7E1C4F25
>	v_mul_legacy_f32_e32 v14, v10, v14                                           ; 0E1C1D0A
>	v_exp_f32_e32 v14, v14                                                       ; 7E1C4B0E
>	v_mul_f32_e32 v13, v14, v13                                                  ; 101A1B0E
>	v_log_f32_e32 v14, v18                                                       ; 7E1C4F12
>	v_mul_legacy_f32_e32 v14, v10, v14                                           ; 0E1C1D0A
>	v_exp_f32_e32 v14, v14                                                       ; 7E1C4B0E
>	v_log_f32_e32 v15, v6                                                        ; 7E1E4F06
>	v_mul_legacy_f32_e32 v15, v10, v15                                           ; 0E1E1F0A
>	v_exp_f32_e32 v15, v15                                                       ; 7E1E4B0F
>	v_mul_f32_e32 v14, v14, v15                                                  ; 101C1F0E
>	v_mad_f32 v13, 2.0, v14, v13                                                 ; D282000D 04361CF4
>	v_log_f32_e64 v13, |v13|                                                     ; D34E010D 0000010D
>	v_mul_f32_e32 v13, 0x3ee8ba1f, v13                                           ; 101A1AFF 3EE8BA1F
>	v_exp_f32_e32 v41, v13                                                       ; 7E524B0D
>	v_log_f32_e32 v13, v32                                                       ; 7E1A4F20
>	v_mul_legacy_f32_e32 v13, v10, v13                                           ; 0E1A1B0A
>	v_exp_f32_e32 v13, v13                                                       ; 7E1A4B0D
>	v_mul_f32_e32 v13, v13, v3                                                   ; 101A070D
>	v_log_f32_e32 v14, v21                                                       ; 7E1C4F15
>	v_mul_legacy_f32_e32 v14, v10, v14                                           ; 0E1C1D0A
>	v_exp_f32_e32 v14, v14                                                       ; 7E1C4B0E
>	v_mad_f32 v13, v2, v14, v13                                                  ; D282000D 04361D02
>	v_mad_f32 v13, v12, v9, v13                                                  ; D282000D 0436130C
>	v_log_f32_e32 v14, v36                                                       ; 7E1C4F24
>	v_mul_legacy_f32_e32 v14, v10, v14                                           ; 0E1C1D0A
>	v_exp_f32_e32 v14, v14                                                       ; 7E1C4B0E
>	v_mul_f32_e32 v13, v14, v13                                                  ; 101A1B0E
>	v_log_f32_e32 v14, v17                                                       ; 7E1C4F11
>	v_mul_legacy_f32_e32 v14, v10, v14                                           ; 0E1C1D0A
>	v_exp_f32_e32 v14, v14                                                       ; 7E1C4B0E
>	v_log_f32_e32 v15, v5                                                        ; 7E1E4F05
>	v_mul_legacy_f32_e32 v15, v10, v15                                           ; 0E1E1F0A
>	v_exp_f32_e32 v15, v15                                                       ; 7E1E4B0F
>	v_mul_f32_e32 v14, v14, v15                                                  ; 101C1F0E
>	v_mad_f32 v13, 2.0, v14, v13                                                 ; D282000D 04361CF4
>	v_log_f32_e64 v13, |v13|                                                     ; D34E010D 0000010D
>	v_mul_f32_e32 v13, 0x3ee8ba1f, v13                                           ; 101A1AFF 3EE8BA1F
>	v_exp_f32_e32 v40, v13                                                       ; 7E504B0D
>	v_log_f32_e32 v13, v31                                                       ; 7E1A4F1F
>	v_mul_legacy_f32_e32 v13, v10, v13                                           ; 0E1A1B0A
>	v_exp_f32_e32 v13, v13                                                       ; 7E1A4B0D
>	v_mul_f32_e32 v3, v13, v3                                                    ; 1006070D
>	v_log_f32_e32 v13, v20                                                       ; 7E1A4F14
>	v_mul_legacy_f32_e32 v13, v10, v13                                           ; 0E1A1B0A
>	v_exp_f32_e32 v13, v13                                                       ; 7E1A4B0D
>	v_mad_f32 v2, v2, v13, v3                                                    ; D2820002 040E1B02
>	v_mad_f32 v2, v12, v9, v2                                                    ; D2820002 040A130C
>	v_log_f32_e32 v3, v35                                                        ; 7E064F23
>	v_mul_legacy_f32_e32 v3, v10, v3                                             ; 0E06070A
>	v_exp_f32_e32 v3, v3                                                         ; 7E064B03
>	v_mul_f32_e32 v2, v3, v2                                                     ; 10040503
>	v_log_f32_e32 v3, v16                                                        ; 7E064F10
>	v_mul_legacy_f32_e32 v3, v10, v3                                             ; 0E06070A
>	v_exp_f32_e32 v3, v3                                                         ; 7E064B03
>	v_log_f32_e32 v4, v4                                                         ; 7E084F04
>	v_mul_legacy_f32_e32 v4, v10, v4                                             ; 0E08090A
>	v_exp_f32_e32 v4, v4                                                         ; 7E084B04
>	v_mul_f32_e32 v3, v3, v4                                                     ; 10060903
>	v_mad_f32 v2, 2.0, v3, v2                                                    ; D2820002 040A06F4
>	v_log_f32_e64 v2, |v2|                                                       ; D34E0102 00000102
>	v_mul_f32_e32 v2, 0x3ee8ba1f, v2                                             ; 100404FF 3EE8BA1F
>	v_exp_f32_e32 v39, v2                                                        ; 7E4E4B02
>	image_sample v[2:4], 7, 0, 0, 0, 0, 0, 0, 0, v[39:42], s[44:51], s[40:43]    ; F0800700 014B0227
>	v_interp_p1_f32 v5, v0, 1, 4, [m0]                                           ; C8141100
>	v_interp_p2_f32 v5, [v5], v1, 1, 4, [m0]                                     ; C8151101
>	v_sub_f32_e32 v6, 1.0, v5                                                    ; 080C0AF2
>	v_interp_p1_f32 v7, v0, 0, 4, [m0]                                           ; C81C1000
>	v_interp_p2_f32 v7, [v7], v1, 0, 4, [m0]                                     ; C81D1001
>	v_add_f32_e32 v5, 0, v7                                                      ; 060A0E80
>	image_sample v[5:6], 5, 0, 0, 0, 0, 0, 0, 0, v[5:6], s[32:39], s[28:31]      ; F0800500 00E80505
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_sub_f32_e32 v7, 1.0, v5                                                    ; 080E0AF2
>	v_interp_p1_f32 v10, v0, 3, 1, [m0]                                          ; C8280700
>	v_interp_p2_f32 v10, [v10], v1, 3, 1, [m0]                                   ; C8290701
>	v_interp_p1_f32 v9, v0, 2, 1, [m0]                                           ; C8240600
>	v_interp_p2_f32 v9, [v9], v1, 2, 1, [m0]                                     ; C8250601
>	s_load_dwordx8 s[4:11], vcc, 0x40                                            ; C0C26B40
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample v[12:14], 7, 0, 0, 0, 0, 0, 0, 0, v[9:10], s[4:11], s[24:27]    ; F0800700 00C10C09
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mul_f32_e32 v9, v14, v7                                                    ; 10120F0E
>	v_mad_f32 v17, v5, v4, v9                                                    ; D2820011 04260905
>	v_mul_f32_e32 v9, v13, v7                                                    ; 10120F0D
>	v_mad_f32 v16, v5, v3, v9                                                    ; D2820010 04260705
>	v_mul_f32_e32 v7, v12, v7                                                    ; 100E0F0C
>	v_mad_f32 v15, v5, v2, v7                                                    ; D282000F 041E0505
>	s_load_dwordx8 s[4:11], vcc, 0x48                                            ; C0C26B48
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample v[2:4], 7, 0, 0, 0, 0, 0, 0, 0, v[15:18], s[4:11], s[20:23]     ; F0800700 00A1020F
>	v_mad_f32 v5, -v6, v5, v5                                                    ; D2820005 24160B06
>	v_sub_f32_e32 v6, 1.0, v5                                                    ; 080C0AF2
>	v_mul_f32_e32 v7, v16, v6                                                    ; 100E0D10
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v7, v5, v3, v7                                                     ; D2820007 041E0705
>	s_buffer_load_dword s4, s[0:3], 0x15                                         ; C2020115
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v9, s4, v7                                                     ; 08120E04
>	s_buffer_load_dword s4, s[0:3], 0x19                                         ; C2020119
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_subrev_f32_e32 v10, s4, v11                                                ; 0A141604
>	s_buffer_load_dword s4, s[0:3], 0x18                                         ; C2020118
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_subrev_f32_e32 v11, s4, v26                                                ; 0A163404
>	v_mul_f32_e32 v11, v11, v11                                                  ; 1016170B
>	v_mad_f32 v10, v10, v10, v11                                                 ; D282000A 042E150A
>	s_buffer_load_dword s4, s[0:3], 0x1a                                         ; C202011A
>	s_buffer_load_dword s5, s[0:3], 0x20                                         ; C2028120
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mov_b32_e32 v11, s4                                                        ; 7E160204
>	v_mad_f32 v11, v29, s5, -v11                                                 ; D282000B 842C0B1D
>	v_mad_f32 v10, v11, v11, v10                                                 ; D282000A 042A170B
>	s_buffer_load_dword s4, s[0:3], 0x11                                         ; C2020111
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v10, s4, v10                                                   ; 10141404
>	s_buffer_load_dword s4, s[0:3], 0x10                                         ; C2020110
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v12, s4, v11                                                   ; 10181604
>	v_mul_f32_e32 v12, 0x3fb8aa65, v12                                           ; 101818FF 3FB8AA65
>	v_exp_f32_e32 v12, v12                                                       ; 7E184B0C
>	v_sub_f32_e32 v12, 1.0, v12                                                  ; 081818F2
>	v_mul_f32_e32 v10, v10, v12                                                  ; 1014190A
>	v_rcp_f32_e32 v11, v11                                                       ; 7E16550B
>	v_mul_f32_e32 v10, v10, v11                                                  ; 1014170A
>	v_mul_f32_e32 v10, 0x3fb8aa65, v10                                           ; 101414FF 3FB8AA65
>	v_exp_f32_e32 v10, v10                                                       ; 7E144B0A
>	v_add_f32_e64 v10, 0, v10 clamp                                              ; D206080A 00021480
>	v_sub_f32_e32 v10, 1.0, v10                                                  ; 081414F2
>	v_interp_p1_f32 v11, v0, 1, 3, [m0]                                          ; C82C0D00
>	v_interp_p2_f32 v11, [v11], v1, 1, 3, [m0]                                   ; C82D0D01
>	v_mad_f32 v0, 0.5, v11, 0.5                                                  ; D2820000 03C216F0
>	s_buffer_load_dword s4, s[0:3], 0x13                                         ; C2020113
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mov_b32_e32 v1, s4                                                         ; 7E020204
>	v_readlane_b32 s4, v47, 0                                                    ; 0209012F
>	v_readlane_b32 s5, v47, 1                                                    ; 020B032F
>	v_readlane_b32 s6, v47, 2                                                    ; 020D052F
>	v_readlane_b32 s7, v47, 3                                                    ; 020F072F
>	s_nop 2                                                                      ; BF800002
>	image_sample v0, 1, 0, 0, 0, 0, 0, 0, 0, v[0:1], s[12:19], s[4:7]            ; F0800100 00230000
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mul_f32_e32 v0, v0, v10                                                    ; 10001500
>	v_mad_f32 v1, v0, v9, v7                                                     ; D2820001 041E1300
>	v_mul_f32_e32 v7, v15, v6                                                    ; 100E0D0F
>	v_mad_f32 v7, v5, v2, v7                                                     ; D2820007 041E0505
>	s_buffer_load_dword s4, s[0:3], 0x14                                         ; C2020114
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v9, s4, v7                                                     ; 08120E04
>	v_mad_f32 v7, v0, v9, v7                                                     ; D2820007 041E1300
>	v_cvt_pkrtz_f16_f32_e32 v1, v7, v1                                           ; 5E020307
>	v_mul_f32_e32 v6, v17, v6                                                    ; 100C0D11
>	v_mad_f32 v2, v5, v4, v6                                                     ; D2820002 041A0905
>	s_buffer_load_dword s4, s[0:3], 0x16                                         ; C2020116
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v3, s4, v2                                                     ; 08060404
>	v_mad_f32 v0, v0, v3, v2                                                     ; D2820000 040A0700
>	v_mul_f32_e32 v2, 0.5, v8                                                    ; 100410F0
>	s_buffer_load_dword s0, s[0:3], 0x1c                                         ; C200011C
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_subrev_f32_e32 v3, s0, v29                                                 ; 0A063A00
>	v_cmp_ge_f32_e64 s[0:1], v3, 0                                               ; D00C0000 00010103
>	v_cndmask_b32_e64 v3, 0, -1, s[0:1]                                          ; D2000803 00018280
>	v_cmp_ne_i32_e64 s[0:1], v3, 0                                               ; D10A0000 00010103
>	v_cndmask_b32_e64 v2, v2, v8, s[0:1]                                         ; D2000002 10021102
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v2                                           ; 5E000500
>	exp 15, 0, 1, 1, 1, v1, v0, v1, v0                                           ; F8001C0F 00010001
>	s_endpgm                                                                     ; BF810000
>
>SGPRS: 104
>VGPRS: 48
>Code Size: 1728 bytes
>LDS: 0 blocks
>Scratch: 1024 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL IN[2]
>DCL IN[3]
>DCL IN[4]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL OUT[2], GENERIC[10]
>DCL OUT[3], GENERIC[11]
>DCL OUT[4], GENERIC[12]
>DCL OUT[5], GENERIC[13]
>DCL CONST[0..104]
>DCL TEMP[0..6], LOCAL
>DCL ADDR[0]
>IMM[0] FLT32 {    3.0000,     1.0000,     0.0000,     0.0000}
>IMM[1] INT32 {1, 2, 0, 0}
>IMM[2] FLT32 {    0.5000,    -0.5000,     0.0000,     0.0000}
>  0: MUL TEMP[0].x, IMM[0].xxxx, IN[2].xxxx
>  1: MAD TEMP[1], IN[0].xyzx, IMM[0].yyyz, IMM[0].zzzy
>  2: F2I TEMP[2].x, TEMP[0].xxxx
>  3: UARL ADDR[0].x, TEMP[2].xxxx
>  4: UARL ADDR[0].x, TEMP[2].xxxx
>  5: DP4 TEMP[2].x, TEMP[1], CONST[ADDR[0].x+9]
>  6: F2I TEMP[3].x, TEMP[0].xxxx
>  7: UADD TEMP[3].x, IMM[1].xxxx, TEMP[3].xxxx
>  8: UARL ADDR[0].x, TEMP[3].xxxx
>  9: DP4 TEMP[3].x, TEMP[1], CONST[ADDR[0].x+9]
> 10: MOV TEMP[2].y, TEMP[3].xxxx
> 11: F2I TEMP[3].x, TEMP[0].xxxx
> 12: UADD TEMP[3].x, IMM[1].yyyy, TEMP[3].xxxx
> 13: UARL ADDR[0].x, TEMP[3].xxxx
> 14: DP4 TEMP[3].x, TEMP[1], CONST[ADDR[0].x+9]
> 15: MOV TEMP[2].z, TEMP[3].xxxx
> 16: MUL TEMP[3].xyz, TEMP[2], IN[1].xxxx
> 17: MOV TEMP[2].xyz, TEMP[3].xyzx
> 18: F2I TEMP[3].x, TEMP[0].xxxx
> 19: UARL ADDR[0].x, TEMP[3].xxxx
> 20: UARL ADDR[0].x, TEMP[3].xxxx
> 21: DP3 TEMP[3].x, IN[3].xyzz, CONST[ADDR[0].x+9].xyzz
> 22: F2I TEMP[4].x, TEMP[0].xxxx
> 23: UADD TEMP[4].x, IMM[1].xxxx, TEMP[4].xxxx
> 24: UARL ADDR[0].x, TEMP[4].xxxx
> 25: DP3 TEMP[4].x, IN[3].xyzz, CONST[ADDR[0].x+9].xyzz
> 26: MOV TEMP[3].y, TEMP[4].xxxx
> 27: F2I TEMP[0].x, TEMP[0].xxxx
> 28: UADD TEMP[0].x, IMM[1].yyyy, TEMP[0].xxxx
> 29: UARL ADDR[0].x, TEMP[0].xxxx
> 30: DP3 TEMP[0].x, IN[3].xyzz, CONST[ADDR[0].x+9].xyzz
> 31: MOV TEMP[3].z, TEMP[0].xxxx
> 32: MUL TEMP[0].xyz, TEMP[3], IN[1].xxxx
> 33: MOV TEMP[3].xyz, TEMP[0].xyzx
> 34: FSLT TEMP[0].x, IMM[0].zzzz, IN[1].yyyy
> 35: UIF TEMP[0].xxxx :0
> 36:   MUL TEMP[0].w, IMM[0].xxxx, IN[2].yyyy
> 37:   MOV TEMP[2].w, TEMP[0].wwww
> 38:   F2I TEMP[4].x, TEMP[0].wwww
> 39:   UARL ADDR[0].x, TEMP[4].xxxx
> 40:   UARL ADDR[0].x, TEMP[4].xxxx
> 41:   DP4 TEMP[4].x, TEMP[1], CONST[ADDR[0].x+9]
> 42:   F2I TEMP[5].x, TEMP[0].wwww
> 43:   UADD TEMP[5].x, IMM[1].xxxx, TEMP[5].xxxx
> 44:   UARL ADDR[0].x, TEMP[5].xxxx
> 45:   DP4 TEMP[5].x, TEMP[1], CONST[ADDR[0].x+9]
> 46:   MOV TEMP[4].y, TEMP[5].xxxx
> 47:   F2I TEMP[5].x, TEMP[0].wwww
> 48:   UADD TEMP[5].x, IMM[1].yyyy, TEMP[5].xxxx
> 49:   UARL ADDR[0].x, TEMP[5].xxxx
> 50:   DP4 TEMP[5].x, TEMP[1], CONST[ADDR[0].x+9]
> 51:   MOV TEMP[4].z, TEMP[5].xxxx
> 52:   MAD TEMP[5].xyz, IN[1].yyyy, TEMP[4], TEMP[2]
> 53:   MOV TEMP[2].xyz, TEMP[5].xyzx
> 54:   F2I TEMP[5].x, TEMP[0].wwww
> 55:   UARL ADDR[0].x, TEMP[5].xxxx
> 56:   UARL ADDR[0].x, TEMP[5].xxxx
> 57:   DP3 TEMP[4].x, IN[3].xyzz, CONST[ADDR[0].x+9].xyzz
> 58:   F2I TEMP[5].x, TEMP[0].wwww
> 59:   UADD TEMP[5].x, IMM[1].xxxx, TEMP[5].xxxx
> 60:   UARL ADDR[0].x, TEMP[5].xxxx
> 61:   DP3 TEMP[5].x, IN[3].xyzz, CONST[ADDR[0].x+9].xyzz
> 62:   MOV TEMP[4].y, TEMP[5].xxxx
> 63:   F2I TEMP[0].x, TEMP[0].wwww
> 64:   UADD TEMP[0].x, IMM[1].yyyy, TEMP[0].xxxx
> 65:   UARL ADDR[0].x, TEMP[0].xxxx
> 66:   DP3 TEMP[0].x, IN[3].xyzz, CONST[ADDR[0].x+9].xyzz
> 67:   MOV TEMP[4].z, TEMP[0].xxxx
> 68:   MAD TEMP[0].xyz, IN[1].yyyy, TEMP[4], TEMP[3]
> 69:   MOV TEMP[3].xyz, TEMP[0].xyzx
> 70:   FSLT TEMP[0].x, IMM[0].zzzz, IN[1].zzzz
> 71:   UIF TEMP[0].xxxx :0
> 72:     MUL TEMP[0].w, IMM[0].xxxx, IN[2].zzzz
> 73:     MOV TEMP[2].w, TEMP[0].wwww
> 74:     F2I TEMP[5].x, TEMP[0].wwww
> 75:     UARL ADDR[0].x, TEMP[5].xxxx
> 76:     UARL ADDR[0].x, TEMP[5].xxxx
> 77:     DP4 TEMP[4].x, TEMP[1], CONST[ADDR[0].x+9]
> 78:     F2I TEMP[5].x, TEMP[0].wwww
> 79:     UADD TEMP[5].x, IMM[1].xxxx, TEMP[5].xxxx
> 80:     UARL ADDR[0].x, TEMP[5].xxxx
> 81:     DP4 TEMP[5].x, TEMP[1], CONST[ADDR[0].x+9]
> 82:     MOV TEMP[4].y, TEMP[5].xxxx
> 83:     F2I TEMP[5].x, TEMP[0].wwww
> 84:     UADD TEMP[5].x, IMM[1].yyyy, TEMP[5].xxxx
> 85:     UARL ADDR[0].x, TEMP[5].xxxx
> 86:     DP4 TEMP[5].x, TEMP[1], CONST[ADDR[0].x+9]
> 87:     MOV TEMP[4].z, TEMP[5].xxxx
> 88:     MAD TEMP[4].xyz, IN[1].zzzz, TEMP[4], TEMP[2]
> 89:     MOV TEMP[2].xyz, TEMP[4].xyzx
> 90:     F2I TEMP[4].x, TEMP[0].wwww
> 91:     UARL ADDR[0].x, TEMP[4].xxxx
> 92:     UARL ADDR[0].x, TEMP[4].xxxx
> 93:     DP3 TEMP[1].x, IN[3].xyzz, CONST[ADDR[0].x+9].xyzz
> 94:     F2I TEMP[4].x, TEMP[0].wwww
> 95:     UADD TEMP[4].x, IMM[1].xxxx, TEMP[4].xxxx
> 96:     UARL ADDR[0].x, TEMP[4].xxxx
> 97:     DP3 TEMP[4].x, IN[3].xyzz, CONST[ADDR[0].x+9].xyzz
> 98:     MOV TEMP[1].y, TEMP[4].xxxx
> 99:     F2I TEMP[0].x, TEMP[0].wwww
>100:     UADD TEMP[0].x, IMM[1].yyyy, TEMP[0].xxxx
>101:     UARL ADDR[0].x, TEMP[0].xxxx
>102:     DP3 TEMP[0].x, IN[3].xyzz, CONST[ADDR[0].x+9].xyzz
>103:     MOV TEMP[1].z, TEMP[0].xxxx
>104:     MAD TEMP[0].xyz, IN[1].zzzz, TEMP[1], TEMP[3]
>105:     MOV TEMP[3].xyz, TEMP[0].xyzx
>106:   ENDIF
>107: ENDIF
>108: MUL TEMP[1], TEMP[2].yyyy, CONST[1]
>109: MAD TEMP[1], TEMP[2].xxxx, CONST[0], TEMP[1]
>110: MAD TEMP[1], TEMP[2].zzzz, CONST[2], TEMP[1]
>111: ADD TEMP[1], TEMP[1], CONST[3]
>112: DP3 TEMP[0].x, TEMP[3].xyzz, TEMP[3].xyzz
>113: MAX TEMP[0].x, TEMP[0].xxxx, IMM[0].wwww
>114: RSQ TEMP[0].x, TEMP[0].xxxx
>115: MUL TEMP[0].xyz, TEMP[0].xxxx, TEMP[3]
>116: MOV TEMP[0].xyz, TEMP[0].xyzx
>117: RCP TEMP[4].x, TEMP[1].wwww
>118: MOV TEMP[2].w, TEMP[4].xxxx
>119: MUL TEMP[4].xy, TEMP[1], TEMP[4].xxxx
>120: MOV TEMP[4].xy, TEMP[4].xyxx
>121: ADD TEMP[5].xy, TEMP[2], CONST[8].zwzw
>122: MUL TEMP[5].zw, TEMP[5].xyxy, CONST[8].xyxy
>123: MOV TEMP[5].zw, TEMP[5].wwzw
>124: MUL TEMP[6].xy, TEMP[2].yyyy, CONST[5]
>125: MOV TEMP[3].xy, TEMP[6].xyxx
>126: MAD TEMP[6].xy, TEMP[2].xxxx, CONST[4], TEMP[3]
>127: MOV TEMP[3].xy, TEMP[6].xyxx
>128: MAD TEMP[6].xy, TEMP[2].zzzz, CONST[6], TEMP[3]
>129: MOV TEMP[3].xy, TEMP[6].xyxx
>130: ADD TEMP[6].xy, TEMP[3], CONST[7]
>131: MOV TEMP[3].xy, TEMP[6].xyxx
>132: MAD TEMP[3].xy, TEMP[3], IMM[2].xyzz, IMM[2].yyyy
>133: MOV TEMP[3].xy, TEMP[3].xyxx
>134: MOV TEMP[5].xy, IN[4].xyxx
>135: MOV TEMP[2].xyz, TEMP[2].xyzx
>136: MOV TEMP[2].w, IMM[0].yyyy
>137: MOV TEMP[4].zw, IMM[0].yyzy
>138: MOV TEMP[3].zw, IMM[0].yyzy
>139: MOV OUT[2], TEMP[5]
>140: MOV OUT[3], TEMP[2]
>141: MOV OUT[0], TEMP[1]
>142: MOV OUT[1], TEMP[0]
>143: MOV OUT[4], TEMP[4]
>144: MOV OUT[5], TEMP[3]
>145: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 0)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 4)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %12, i32 8)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %12, i32 12)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %12, i32 16)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %12, i32 20)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %12, i32 24)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %12, i32 28)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %12, i32 32)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %12, i32 36)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %12, i32 40)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %12, i32 44)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %12, i32 48)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %12, i32 52)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %12, i32 56)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %12, i32 60)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %12, i32 64)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %12, i32 68)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %12, i32 80)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %12, i32 84)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %12, i32 96)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %12, i32 100)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %12, i32 112)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %12, i32 116)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %12, i32 128)
>  %38 = call float @llvm.SI.load.const(<16 x i8> %12, i32 132)
>  %39 = call float @llvm.SI.load.const(<16 x i8> %12, i32 136)
>  %40 = call float @llvm.SI.load.const(<16 x i8> %12, i32 140)
>  %41 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %42 = load <16 x i8> addrspace(2)* %41, !tbaa !0
>  %43 = add i32 %5, %7
>  %44 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %42, i32 0, i32 %43)
>  %45 = extractelement <4 x float> %44, i32 0
>  %46 = extractelement <4 x float> %44, i32 1
>  %47 = extractelement <4 x float> %44, i32 2
>  %48 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %49 = load <16 x i8> addrspace(2)* %48, !tbaa !0
>  %50 = add i32 %5, %7
>  %51 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %49, i32 0, i32 %50)
>  %52 = extractelement <4 x float> %51, i32 0
>  %53 = extractelement <4 x float> %51, i32 1
>  %54 = extractelement <4 x float> %51, i32 2
>  %55 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 2
>  %56 = load <16 x i8> addrspace(2)* %55, !tbaa !0
>  %57 = add i32 %5, %7
>  %58 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %56, i32 0, i32 %57)
>  %59 = extractelement <4 x float> %58, i32 0
>  %60 = extractelement <4 x float> %58, i32 1
>  %61 = extractelement <4 x float> %58, i32 2
>  %62 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 3
>  %63 = load <16 x i8> addrspace(2)* %62, !tbaa !0
>  %64 = add i32 %5, %7
>  %65 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %63, i32 0, i32 %64)
>  %66 = extractelement <4 x float> %65, i32 0
>  %67 = extractelement <4 x float> %65, i32 1
>  %68 = extractelement <4 x float> %65, i32 2
>  %69 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 4
>  %70 = load <16 x i8> addrspace(2)* %69, !tbaa !0
>  %71 = add i32 %5, %7
>  %72 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %70, i32 0, i32 %71)
>  %73 = extractelement <4 x float> %72, i32 0
>  %74 = extractelement <4 x float> %72, i32 1
>  %75 = fmul float 3.000000e+00, %59
>  %76 = fmul float %45, 1.000000e+00
>  %77 = fadd float %76, 0.000000e+00
>  %78 = fmul float %46, 1.000000e+00
>  %79 = fadd float %78, 0.000000e+00
>  %80 = fmul float %47, 1.000000e+00
>  %81 = fadd float %80, 0.000000e+00
>  %82 = fmul float %45, 0.000000e+00
>  %83 = fadd float %82, 1.000000e+00
>  %84 = fptosi float %75 to i32
>  %85 = bitcast i32 %84 to float
>  %86 = bitcast float %85 to i32
>  %87 = shl i32 %86, 4
>  %88 = add i32 %87, 144
>  %89 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %88)
>  %90 = shl i32 %86, 4
>  %91 = add i32 %90, 148
>  %92 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %91)
>  %93 = shl i32 %86, 4
>  %94 = add i32 %93, 152
>  %95 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %94)
>  %96 = shl i32 %86, 4
>  %97 = add i32 %96, 156
>  %98 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %97)
>  %99 = fmul float %77, %89
>  %100 = fmul float %79, %92
>  %101 = fadd float %99, %100
>  %102 = fmul float %81, %95
>  %103 = fadd float %101, %102
>  %104 = fmul float %83, %98
>  %105 = fadd float %103, %104
>  %106 = fptosi float %75 to i32
>  %107 = bitcast i32 %106 to float
>  %108 = bitcast float %107 to i32
>  %109 = add i32 1, %108
>  %110 = bitcast i32 %109 to float
>  %111 = bitcast float %110 to i32
>  %112 = shl i32 %111, 4
>  %113 = add i32 %112, 144
>  %114 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %113)
>  %115 = shl i32 %111, 4
>  %116 = add i32 %115, 148
>  %117 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %116)
>  %118 = shl i32 %111, 4
>  %119 = add i32 %118, 152
>  %120 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %119)
>  %121 = shl i32 %111, 4
>  %122 = add i32 %121, 156
>  %123 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %122)
>  %124 = fmul float %77, %114
>  %125 = fmul float %79, %117
>  %126 = fadd float %124, %125
>  %127 = fmul float %81, %120
>  %128 = fadd float %126, %127
>  %129 = fmul float %83, %123
>  %130 = fadd float %128, %129
>  %131 = fptosi float %75 to i32
>  %132 = bitcast i32 %131 to float
>  %133 = bitcast float %132 to i32
>  %134 = add i32 2, %133
>  %135 = bitcast i32 %134 to float
>  %136 = bitcast float %135 to i32
>  %137 = shl i32 %136, 4
>  %138 = add i32 %137, 144
>  %139 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %138)
>  %140 = shl i32 %136, 4
>  %141 = add i32 %140, 148
>  %142 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %141)
>  %143 = shl i32 %136, 4
>  %144 = add i32 %143, 152
>  %145 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %144)
>  %146 = shl i32 %136, 4
>  %147 = add i32 %146, 156
>  %148 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %147)
>  %149 = fmul float %77, %139
>  %150 = fmul float %79, %142
>  %151 = fadd float %149, %150
>  %152 = fmul float %81, %145
>  %153 = fadd float %151, %152
>  %154 = fmul float %83, %148
>  %155 = fadd float %153, %154
>  %156 = fmul float %105, %52
>  %157 = fmul float %130, %52
>  %158 = fmul float %155, %52
>  %159 = fptosi float %75 to i32
>  %160 = bitcast i32 %159 to float
>  %161 = bitcast float %160 to i32
>  %162 = shl i32 %161, 4
>  %163 = add i32 %162, 144
>  %164 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %163)
>  %165 = shl i32 %161, 4
>  %166 = add i32 %165, 148
>  %167 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %166)
>  %168 = shl i32 %161, 4
>  %169 = add i32 %168, 152
>  %170 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %169)
>  %171 = fmul float %66, %164
>  %172 = fmul float %67, %167
>  %173 = fadd float %172, %171
>  %174 = fmul float %68, %170
>  %175 = fadd float %173, %174
>  %176 = fptosi float %75 to i32
>  %177 = bitcast i32 %176 to float
>  %178 = bitcast float %177 to i32
>  %179 = add i32 1, %178
>  %180 = bitcast i32 %179 to float
>  %181 = bitcast float %180 to i32
>  %182 = shl i32 %181, 4
>  %183 = add i32 %182, 144
>  %184 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %183)
>  %185 = shl i32 %181, 4
>  %186 = add i32 %185, 148
>  %187 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %186)
>  %188 = shl i32 %181, 4
>  %189 = add i32 %188, 152
>  %190 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %189)
>  %191 = fmul float %66, %184
>  %192 = fmul float %67, %187
>  %193 = fadd float %192, %191
>  %194 = fmul float %68, %190
>  %195 = fadd float %193, %194
>  %196 = fptosi float %75 to i32
>  %197 = bitcast i32 %196 to float
>  %198 = bitcast float %197 to i32
>  %199 = add i32 2, %198
>  %200 = bitcast i32 %199 to float
>  %201 = bitcast float %200 to i32
>  %202 = shl i32 %201, 4
>  %203 = add i32 %202, 144
>  %204 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %203)
>  %205 = shl i32 %201, 4
>  %206 = add i32 %205, 148
>  %207 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %206)
>  %208 = shl i32 %201, 4
>  %209 = add i32 %208, 152
>  %210 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %209)
>  %211 = fmul float %66, %204
>  %212 = fmul float %67, %207
>  %213 = fadd float %212, %211
>  %214 = fmul float %68, %210
>  %215 = fadd float %213, %214
>  %216 = fmul float %175, %52
>  %217 = fmul float %195, %52
>  %218 = fmul float %215, %52
>  %219 = fcmp olt float 0.000000e+00, %53
>  %220 = sext i1 %219 to i32
>  %221 = bitcast i32 %220 to float
>  %222 = bitcast float %221 to i32
>  %223 = icmp ne i32 %222, 0
>  br i1 %223, label %IF, label %ENDIF
>
>IF:                                               ; preds = %main_body
>  %224 = fmul float 3.000000e+00, %60
>  %225 = fptosi float %224 to i32
>  %226 = bitcast i32 %225 to float
>  %227 = bitcast float %226 to i32
>  %228 = shl i32 %227, 4
>  %229 = add i32 %228, 144
>  %230 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %229)
>  %231 = shl i32 %227, 4
>  %232 = add i32 %231, 148
>  %233 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %232)
>  %234 = shl i32 %227, 4
>  %235 = add i32 %234, 152
>  %236 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %235)
>  %237 = shl i32 %227, 4
>  %238 = add i32 %237, 156
>  %239 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %238)
>  %240 = fmul float %77, %230
>  %241 = fmul float %79, %233
>  %242 = fadd float %240, %241
>  %243 = fmul float %81, %236
>  %244 = fadd float %242, %243
>  %245 = fmul float %83, %239
>  %246 = fadd float %244, %245
>  %247 = fptosi float %224 to i32
>  %248 = bitcast i32 %247 to float
>  %249 = bitcast float %248 to i32
>  %250 = add i32 1, %249
>  %251 = bitcast i32 %250 to float
>  %252 = bitcast float %251 to i32
>  %253 = shl i32 %252, 4
>  %254 = add i32 %253, 144
>  %255 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %254)
>  %256 = shl i32 %252, 4
>  %257 = add i32 %256, 148
>  %258 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %257)
>  %259 = shl i32 %252, 4
>  %260 = add i32 %259, 152
>  %261 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %260)
>  %262 = shl i32 %252, 4
>  %263 = add i32 %262, 156
>  %264 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %263)
>  %265 = fmul float %77, %255
>  %266 = fmul float %79, %258
>  %267 = fadd float %265, %266
>  %268 = fmul float %81, %261
>  %269 = fadd float %267, %268
>  %270 = fmul float %83, %264
>  %271 = fadd float %269, %270
>  %272 = fptosi float %224 to i32
>  %273 = bitcast i32 %272 to float
>  %274 = bitcast float %273 to i32
>  %275 = add i32 2, %274
>  %276 = bitcast i32 %275 to float
>  %277 = bitcast float %276 to i32
>  %278 = shl i32 %277, 4
>  %279 = add i32 %278, 144
>  %280 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %279)
>  %281 = shl i32 %277, 4
>  %282 = add i32 %281, 148
>  %283 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %282)
>  %284 = shl i32 %277, 4
>  %285 = add i32 %284, 152
>  %286 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %285)
>  %287 = shl i32 %277, 4
>  %288 = add i32 %287, 156
>  %289 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %288)
>  %290 = fmul float %77, %280
>  %291 = fmul float %79, %283
>  %292 = fadd float %290, %291
>  %293 = fmul float %81, %286
>  %294 = fadd float %292, %293
>  %295 = fmul float %83, %289
>  %296 = fadd float %294, %295
>  %297 = fmul float %53, %246
>  %298 = fadd float %297, %156
>  %299 = fmul float %53, %271
>  %300 = fadd float %299, %157
>  %301 = fmul float %53, %296
>  %302 = fadd float %301, %158
>  %303 = fptosi float %224 to i32
>  %304 = bitcast i32 %303 to float
>  %305 = bitcast float %304 to i32
>  %306 = shl i32 %305, 4
>  %307 = add i32 %306, 144
>  %308 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %307)
>  %309 = shl i32 %305, 4
>  %310 = add i32 %309, 148
>  %311 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %310)
>  %312 = shl i32 %305, 4
>  %313 = add i32 %312, 152
>  %314 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %313)
>  %315 = fmul float %66, %308
>  %316 = fmul float %67, %311
>  %317 = fadd float %316, %315
>  %318 = fmul float %68, %314
>  %319 = fadd float %317, %318
>  %320 = fptosi float %224 to i32
>  %321 = bitcast i32 %320 to float
>  %322 = bitcast float %321 to i32
>  %323 = add i32 1, %322
>  %324 = bitcast i32 %323 to float
>  %325 = bitcast float %324 to i32
>  %326 = shl i32 %325, 4
>  %327 = add i32 %326, 144
>  %328 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %327)
>  %329 = shl i32 %325, 4
>  %330 = add i32 %329, 148
>  %331 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %330)
>  %332 = shl i32 %325, 4
>  %333 = add i32 %332, 152
>  %334 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %333)
>  %335 = fmul float %66, %328
>  %336 = fmul float %67, %331
>  %337 = fadd float %336, %335
>  %338 = fmul float %68, %334
>  %339 = fadd float %337, %338
>  %340 = fptosi float %224 to i32
>  %341 = bitcast i32 %340 to float
>  %342 = bitcast float %341 to i32
>  %343 = add i32 2, %342
>  %344 = bitcast i32 %343 to float
>  %345 = bitcast float %344 to i32
>  %346 = shl i32 %345, 4
>  %347 = add i32 %346, 144
>  %348 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %347)
>  %349 = shl i32 %345, 4
>  %350 = add i32 %349, 148
>  %351 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %350)
>  %352 = shl i32 %345, 4
>  %353 = add i32 %352, 152
>  %354 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %353)
>  %355 = fmul float %66, %348
>  %356 = fmul float %67, %351
>  %357 = fadd float %356, %355
>  %358 = fmul float %68, %354
>  %359 = fadd float %357, %358
>  %360 = fmul float %53, %319
>  %361 = fadd float %360, %216
>  %362 = fmul float %53, %339
>  %363 = fadd float %362, %217
>  %364 = fmul float %53, %359
>  %365 = fadd float %364, %218
>  %366 = fcmp olt float 0.000000e+00, %54
>  %367 = sext i1 %366 to i32
>  %368 = bitcast i32 %367 to float
>  %369 = bitcast float %368 to i32
>  %370 = icmp ne i32 %369, 0
>  br i1 %370, label %IF70, label %ENDIF
>
>ENDIF:                                            ; preds = %IF70, %IF, %main_body
>  %temp3.0 = phi float [ 0.000000e+00, %main_body ], [ %428, %IF70 ], [ %224, %IF ]
>  %temp8.0 = phi float [ %156, %main_body ], [ %502, %IF70 ], [ %298, %IF ]
>  %temp9.0 = phi float [ %157, %main_body ], [ %504, %IF70 ], [ %300, %IF ]
>  %temp10.0 = phi float [ %158, %main_body ], [ %506, %IF70 ], [ %302, %IF ]
>  %temp12.0 = phi float [ %216, %main_body ], [ %565, %IF70 ], [ %361, %IF ]
>  %temp13.0 = phi float [ %217, %main_body ], [ %567, %IF70 ], [ %363, %IF ]
>  %temp14.0 = phi float [ %218, %main_body ], [ %569, %IF70 ], [ %365, %IF ]
>  %371 = fmul float %temp9.0, %17
>  %372 = fmul float %temp9.0, %18
>  %373 = fmul float %temp9.0, %19
>  %374 = fmul float %temp9.0, %20
>  %375 = fmul float %temp8.0, %13
>  %376 = fadd float %375, %371
>  %377 = fmul float %temp8.0, %14
>  %378 = fadd float %377, %372
>  %379 = fmul float %temp8.0, %15
>  %380 = fadd float %379, %373
>  %381 = fmul float %temp8.0, %16
>  %382 = fadd float %381, %374
>  %383 = fmul float %temp10.0, %21
>  %384 = fadd float %383, %376
>  %385 = fmul float %temp10.0, %22
>  %386 = fadd float %385, %378
>  %387 = fmul float %temp10.0, %23
>  %388 = fadd float %387, %380
>  %389 = fmul float %temp10.0, %24
>  %390 = fadd float %389, %382
>  %391 = fadd float %384, %25
>  %392 = fadd float %386, %26
>  %393 = fadd float %388, %27
>  %394 = fadd float %390, %28
>  %395 = fmul float %temp12.0, %temp12.0
>  %396 = fmul float %temp13.0, %temp13.0
>  %397 = fadd float %396, %395
>  %398 = fmul float %temp14.0, %temp14.0
>  %399 = fadd float %397, %398
>  %400 = call float @llvm.maxnum.f32(float %399, float 0x3E7AD7F2A0000000)
>  %401 = call float @llvm.AMDGPU.rsq.clamped.f32(float %400)
>  %402 = fmul float %401, %temp12.0
>  %403 = fmul float %401, %temp13.0
>  %404 = fmul float %401, %temp14.0
>  %405 = fdiv float 1.000000e+00, %394
>  %406 = fmul float %391, %405
>  %407 = fmul float %392, %405
>  %408 = fadd float %temp8.0, %39
>  %409 = fadd float %temp9.0, %40
>  %410 = fmul float %408, %37
>  %411 = fmul float %409, %38
>  %412 = fmul float %temp9.0, %31
>  %413 = fmul float %temp9.0, %32
>  %414 = fmul float %temp8.0, %29
>  %415 = fadd float %414, %412
>  %416 = fmul float %temp8.0, %30
>  %417 = fadd float %416, %413
>  %418 = fmul float %temp10.0, %33
>  %419 = fadd float %418, %415
>  %420 = fmul float %temp10.0, %34
>  %421 = fadd float %420, %417
>  %422 = fadd float %419, %35
>  %423 = fadd float %421, %36
>  %424 = fmul float %422, 5.000000e-01
>  %425 = fadd float %424, -5.000000e-01
>  %426 = fmul float %423, -5.000000e-01
>  %427 = fadd float %426, -5.000000e-01
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %402, float %403, float %404, float %temp3.0)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %73, float %74, float %410, float %411)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 34, i32 0, float %temp8.0, float %temp9.0, float %temp10.0, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 35, i32 0, float %406, float %407, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 36, i32 0, float %425, float %427, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %391, float %392, float %393, float %394)
>  ret void
>
>IF70:                                             ; preds = %IF
>  %428 = fmul float 3.000000e+00, %61
>  %429 = fptosi float %428 to i32
>  %430 = bitcast i32 %429 to float
>  %431 = bitcast float %430 to i32
>  %432 = shl i32 %431, 4
>  %433 = add i32 %432, 144
>  %434 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %433)
>  %435 = shl i32 %431, 4
>  %436 = add i32 %435, 148
>  %437 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %436)
>  %438 = shl i32 %431, 4
>  %439 = add i32 %438, 152
>  %440 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %439)
>  %441 = shl i32 %431, 4
>  %442 = add i32 %441, 156
>  %443 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %442)
>  %444 = fmul float %77, %434
>  %445 = fmul float %79, %437
>  %446 = fadd float %444, %445
>  %447 = fmul float %81, %440
>  %448 = fadd float %446, %447
>  %449 = fmul float %83, %443
>  %450 = fadd float %448, %449
>  %451 = fptosi float %428 to i32
>  %452 = bitcast i32 %451 to float
>  %453 = bitcast float %452 to i32
>  %454 = add i32 1, %453
>  %455 = bitcast i32 %454 to float
>  %456 = bitcast float %455 to i32
>  %457 = shl i32 %456, 4
>  %458 = add i32 %457, 144
>  %459 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %458)
>  %460 = shl i32 %456, 4
>  %461 = add i32 %460, 148
>  %462 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %461)
>  %463 = shl i32 %456, 4
>  %464 = add i32 %463, 152
>  %465 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %464)
>  %466 = shl i32 %456, 4
>  %467 = add i32 %466, 156
>  %468 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %467)
>  %469 = fmul float %77, %459
>  %470 = fmul float %79, %462
>  %471 = fadd float %469, %470
>  %472 = fmul float %81, %465
>  %473 = fadd float %471, %472
>  %474 = fmul float %83, %468
>  %475 = fadd float %473, %474
>  %476 = fptosi float %428 to i32
>  %477 = bitcast i32 %476 to float
>  %478 = bitcast float %477 to i32
>  %479 = add i32 2, %478
>  %480 = bitcast i32 %479 to float
>  %481 = bitcast float %480 to i32
>  %482 = shl i32 %481, 4
>  %483 = add i32 %482, 144
>  %484 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %483)
>  %485 = shl i32 %481, 4
>  %486 = add i32 %485, 148
>  %487 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %486)
>  %488 = shl i32 %481, 4
>  %489 = add i32 %488, 152
>  %490 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %489)
>  %491 = shl i32 %481, 4
>  %492 = add i32 %491, 156
>  %493 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %492)
>  %494 = fmul float %77, %484
>  %495 = fmul float %79, %487
>  %496 = fadd float %494, %495
>  %497 = fmul float %81, %490
>  %498 = fadd float %496, %497
>  %499 = fmul float %83, %493
>  %500 = fadd float %498, %499
>  %501 = fmul float %54, %450
>  %502 = fadd float %501, %298
>  %503 = fmul float %54, %475
>  %504 = fadd float %503, %300
>  %505 = fmul float %54, %500
>  %506 = fadd float %505, %302
>  %507 = fptosi float %428 to i32
>  %508 = bitcast i32 %507 to float
>  %509 = bitcast float %508 to i32
>  %510 = shl i32 %509, 4
>  %511 = add i32 %510, 144
>  %512 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %511)
>  %513 = shl i32 %509, 4
>  %514 = add i32 %513, 148
>  %515 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %514)
>  %516 = shl i32 %509, 4
>  %517 = add i32 %516, 152
>  %518 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %517)
>  %519 = fmul float %66, %512
>  %520 = fmul float %67, %515
>  %521 = fadd float %520, %519
>  %522 = fmul float %68, %518
>  %523 = fadd float %521, %522
>  %524 = fptosi float %428 to i32
>  %525 = bitcast i32 %524 to float
>  %526 = bitcast float %525 to i32
>  %527 = add i32 1, %526
>  %528 = bitcast i32 %527 to float
>  %529 = bitcast float %528 to i32
>  %530 = shl i32 %529, 4
>  %531 = add i32 %530, 144
>  %532 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %531)
>  %533 = shl i32 %529, 4
>  %534 = add i32 %533, 148
>  %535 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %534)
>  %536 = shl i32 %529, 4
>  %537 = add i32 %536, 152
>  %538 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %537)
>  %539 = fmul float %66, %532
>  %540 = fmul float %67, %535
>  %541 = fadd float %540, %539
>  %542 = fmul float %68, %538
>  %543 = fadd float %541, %542
>  %544 = fptosi float %428 to i32
>  %545 = bitcast i32 %544 to float
>  %546 = bitcast float %545 to i32
>  %547 = add i32 2, %546
>  %548 = bitcast i32 %547 to float
>  %549 = bitcast float %548 to i32
>  %550 = shl i32 %549, 4
>  %551 = add i32 %550, 144
>  %552 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %551)
>  %553 = shl i32 %549, 4
>  %554 = add i32 %553, 148
>  %555 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %554)
>  %556 = shl i32 %549, 4
>  %557 = add i32 %556, 152
>  %558 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %557)
>  %559 = fmul float %66, %552
>  %560 = fmul float %67, %555
>  %561 = fadd float %560, %559
>  %562 = fmul float %68, %558
>  %563 = fadd float %561, %562
>  %564 = fmul float %54, %523
>  %565 = fadd float %564, %361
>  %566 = fmul float %54, %543
>  %567 = fadd float %566, %363
>  %568 = fmul float %54, %563
>  %569 = fadd float %568, %365
>  br label %ENDIF
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.maxnum.f32(float, float) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.AMDGPU.rsq.clamped.f32(float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v11, s10, v0                               ; 4A16000A
>	s_load_dwordx4 s[12:15], s[8:9], 0x0                     ; C0860900
>	s_load_dwordx4 s[16:19], s[8:9], 0x4                     ; C0880904
>	s_load_dwordx4 s[20:23], s[8:9], 0x8                     ; C08A0908
>	s_load_dwordx4 s[24:27], s[8:9], 0xc                     ; C08C090C
>	s_load_dwordx4 s[4:7], s[8:9], 0x10                      ; C0820910
>	s_waitcnt lgkmcnt(0)                                     ; BF8C007F
>	buffer_load_format_xyzw v[20:23], v11, s[20:23], 0 idxen ; E00C2000 8005140B
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v0, 0x40400000, v20                        ; 100028FF 40400000
>	v_cvt_i32_f32_e32 v0, v0                                 ; 7E001100
>	v_lshlrev_b32_e32 v3, 4, v0                              ; 34060084
>	v_add_i32_e32 v0, 0xb0, v3                               ; 4A0006FF 000000B0
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                       ; C0800300
>	s_waitcnt lgkmcnt(0)                                     ; BF8C007F
>	buffer_load_dword v0, v0, s[0:3], 0 offen                ; E0301000 80000000
>	buffer_load_format_xyzw v[15:18], v11, s[24:27], 0 idxen ; E00C2000 80060F0B
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v1, v0, v15                                ; 10021F00
>	v_add_i32_e32 v2, 0xb4, v3                               ; 4A0406FF 000000B4
>	buffer_load_dword v2, v2, s[0:3], 0 offen                ; E0301000 80000202
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v1, v16, v2, v1                                ; D2820001 04060510
>	v_add_i32_e32 v4, 0xb8, v3                               ; 4A0806FF 000000B8
>	buffer_load_dword v12, v4, s[0:3], 0 offen               ; E0301000 80000C04
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v1, v17, v12, v1                               ; D2820001 04061911
>	buffer_load_format_xyzw v[7:10], v11, s[16:19], 0 idxen  ; E00C2000 8004070B
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v1, v7, v1                                 ; 10020307
>	v_add_i32_e32 v4, 0xa0, v3                               ; 4A0806FF 000000A0
>	buffer_load_dword v13, v4, s[0:3], 0 offen               ; E0301000 80000D04
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v4, v13, v15                               ; 10081F0D
>	v_add_i32_e32 v5, 0xa4, v3                               ; 4A0A06FF 000000A4
>	buffer_load_dword v14, v5, s[0:3], 0 offen               ; E0301000 80000E05
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v4, v16, v14, v4                               ; D2820004 04121D10
>	v_add_i32_e32 v5, 0xa8, v3                               ; 4A0A06FF 000000A8
>	buffer_load_dword v27, v5, s[0:3], 0 offen               ; E0301000 80001B05
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v4, v17, v27, v4                               ; D2820004 04123711
>	v_mul_f32_e32 v4, v7, v4                                 ; 10080907
>	v_add_i32_e32 v5, 0x90, v3                               ; 4A0A06FF 00000090
>	buffer_load_dword v28, v5, s[0:3], 0 offen               ; E0301000 80001C05
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v5, v28, v15                               ; 100A1F1C
>	v_add_i32_e32 v6, 0x94, v3                               ; 4A0C06FF 00000094
>	buffer_load_dword v29, v6, s[0:3], 0 offen               ; E0301000 80001D06
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v5, v16, v29, v5                               ; D2820005 04163B10
>	v_add_i32_e32 v6, 0x98, v3                               ; 4A0C06FF 00000098
>	buffer_load_dword v30, v6, s[0:3], 0 offen               ; E0301000 80001E06
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v5, v17, v30, v5                               ; D2820005 04163D11
>	v_mul_f32_e32 v5, v7, v5                                 ; 100A0B07
>	buffer_load_format_xyzw v[31:34], v11, s[12:15], 0 idxen ; E00C2000 80031F0B
>	v_mov_b32_e32 v6, 0                                      ; 7E0C0280
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_add_f32_e32 v19, 0, v32                                ; 06264080
>	v_mul_f32_e32 v2, v2, v19                                ; 10042702
>	v_add_f32_e32 v24, 0, v31                                ; 06303E80
>	v_mad_f32 v0, v24, v0, v2                                ; D2820000 040A0118
>	v_add_f32_e32 v26, 0, v33                                ; 06344280
>	v_mad_f32 v0, v26, v12, v0                               ; D2820000 0402191A
>	v_mad_f32 v25, 0, v31, 1.0                               ; D2820019 03CA3E80
>	v_add_i32_e32 v2, 0xbc, v3                               ; 4A0406FF 000000BC
>	buffer_load_dword v2, v2, s[0:3], 0 offen                ; E0301000 80000202
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v0, v25, v2, v0                                ; D2820000 04020519
>	v_mul_f32_e32 v0, v7, v0                                 ; 10000107
>	v_mul_f32_e32 v2, v14, v19                               ; 1004270E
>	v_mad_f32 v2, v24, v13, v2                               ; D2820002 040A1B18
>	v_mad_f32 v2, v26, v27, v2                               ; D2820002 040A371A
>	v_add_i32_e32 v12, 0xac, v3                              ; 4A1806FF 000000AC
>	buffer_load_dword v12, v12, s[0:3], 0 offen              ; E0301000 80000C0C
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v2, v25, v12, v2                               ; D2820002 040A1919
>	v_mul_f32_e32 v2, v7, v2                                 ; 10040507
>	v_mul_f32_e32 v12, v29, v19                              ; 1018271D
>	v_mad_f32 v12, v24, v28, v12                             ; D282000C 04323918
>	v_mad_f32 v12, v26, v30, v12                             ; D282000C 04323D1A
>	v_add_i32_e32 v3, 0x9c, v3                               ; 4A0606FF 0000009C
>	buffer_load_dword v3, v3, s[0:3], 0 offen                ; E0301000 80000303
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v3, v25, v3, v12                               ; D2820003 04320719
>	v_mul_f32_e32 v3, v7, v3                                 ; 10060707
>	buffer_load_format_xyzw v[11:14], v11, s[4:7], 0 idxen   ; E00C2000 80010B0B
>	v_cmp_gt_f32_e64 s[4:5], v8, 0                           ; D0080004 00010108
>	v_cndmask_b32_e64 v27, 0, -1, s[4:5]                     ; D200081B 00118280
>	v_cmp_ne_i32_e64 s[32:33], v27, 0                        ; D10A0020 0001011B
>	s_buffer_load_dword s20, s[0:3], 0x23                    ; C20A0123
>	s_buffer_load_dword s19, s[0:3], 0x22                    ; C2098122
>	s_buffer_load_dword s22, s[0:3], 0x21                    ; C20B0121
>	s_buffer_load_dword s21, s[0:3], 0x20                    ; C20A8120
>	s_buffer_load_dword s4, s[0:3], 0x1d                     ; C202011D
>	s_buffer_load_dword s7, s[0:3], 0x1c                     ; C203811C
>	s_buffer_load_dword s5, s[0:3], 0x19                     ; C2028119
>	s_buffer_load_dword s11, s[0:3], 0x18                    ; C2058118
>	s_buffer_load_dword s8, s[0:3], 0x15                     ; C2040115
>	s_buffer_load_dword s12, s[0:3], 0x14                    ; C2060114
>	s_buffer_load_dword s9, s[0:3], 0x11                     ; C2048111
>	s_buffer_load_dword s13, s[0:3], 0x10                    ; C2068110
>	s_buffer_load_dword s17, s[0:3], 0xf                     ; C208810F
>	s_buffer_load_dword s6, s[0:3], 0xe                      ; C203010E
>	s_buffer_load_dword s24, s[0:3], 0xd                     ; C20C010D
>	s_buffer_load_dword s16, s[0:3], 0xc                     ; C208010C
>	s_buffer_load_dword s23, s[0:3], 0xb                     ; C20B810B
>	s_buffer_load_dword s10, s[0:3], 0xa                     ; C205010A
>	s_buffer_load_dword s26, s[0:3], 0x9                     ; C20D0109
>	s_buffer_load_dword s18, s[0:3], 0x8                     ; C2090108
>	s_buffer_load_dword s27, s[0:3], 0x7                     ; C20D8107
>	s_buffer_load_dword s14, s[0:3], 0x6                     ; C2070106
>	s_buffer_load_dword s30, s[0:3], 0x5                     ; C20F0105
>	s_buffer_load_dword s25, s[0:3], 0x4                     ; C20C8104
>	s_buffer_load_dword s29, s[0:3], 0x3                     ; C20E8103
>	s_buffer_load_dword s15, s[0:3], 0x2                     ; C2078102
>	s_buffer_load_dword s31, s[0:3], 0x1                     ; C20F8101
>	s_buffer_load_dword s28, s[0:3], 0x0                     ; C20E0100
>	s_waitcnt vmcnt(0) lgkmcnt(0)                            ; BF8C0070
>	s_and_saveexec_b64 s[32:33], s[32:33]                    ; BEA02420
>	s_xor_b64 s[32:33], exec, s[32:33]                       ; 89A0207E
>	s_cbranch_execz BB0_3                                    ; BF880000
>	v_mul_f32_e32 v6, 0x40400000, v21                        ; 100C2AFF 40400000
>	v_cvt_i32_f32_e32 v27, v6                                ; 7E361106
>	v_lshlrev_b32_e32 v27, 4, v27                            ; 34363684
>	v_add_i32_e32 v28, 0xb4, v27                             ; 4A3836FF 000000B4
>	buffer_load_dword v28, v28, s[0:3], 0 offen              ; E0301000 80001C1C
>	v_add_i32_e32 v29, 0xb0, v27                             ; 4A3A36FF 000000B0
>	buffer_load_dword v29, v29, s[0:3], 0 offen              ; E0301000 80001D1D
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v30, v29, v15                              ; 103C1F1D
>	v_mad_f32 v30, v16, v28, v30                             ; D282001E 047A3910
>	v_add_i32_e32 v31, 0xb8, v27                             ; 4A3E36FF 000000B8
>	buffer_load_dword v31, v31, s[0:3], 0 offen              ; E0301000 80001F1F
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v30, v17, v31, v30                             ; D282001E 047A3F11
>	v_mad_f32 v1, v8, v30, v1                                ; D2820001 04063D08
>	v_add_i32_e32 v30, 0xa4, v27                             ; 4A3C36FF 000000A4
>	buffer_load_dword v30, v30, s[0:3], 0 offen              ; E0301000 80001E1E
>	v_add_i32_e32 v32, 0xa0, v27                             ; 4A4036FF 000000A0
>	buffer_load_dword v32, v32, s[0:3], 0 offen              ; E0301000 80002020
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v33, v32, v15                              ; 10421F20
>	v_mad_f32 v33, v16, v30, v33                             ; D2820021 04863D10
>	v_add_i32_e32 v34, 0xa8, v27                             ; 4A4436FF 000000A8
>	buffer_load_dword v34, v34, s[0:3], 0 offen              ; E0301000 80002222
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v33, v17, v34, v33                             ; D2820021 04864511
>	v_mad_f32 v4, v8, v33, v4                                ; D2820004 04124308
>	v_add_i32_e32 v33, 0x94, v27                             ; 4A4236FF 00000094
>	buffer_load_dword v33, v33, s[0:3], 0 offen              ; E0301000 80002121
>	v_add_i32_e32 v35, 0x90, v27                             ; 4A4636FF 00000090
>	buffer_load_dword v35, v35, s[0:3], 0 offen              ; E0301000 80002323
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v36, v35, v15                              ; 10481F23
>	v_mad_f32 v36, v16, v33, v36                             ; D2820024 04924310
>	v_add_i32_e32 v37, 0x98, v27                             ; 4A4A36FF 00000098
>	buffer_load_dword v37, v37, s[0:3], 0 offen              ; E0301000 80002525
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v36, v17, v37, v36                             ; D2820024 04924B11
>	v_mad_f32 v5, v8, v36, v5                                ; D2820005 04164908
>	v_mul_f32_e32 v28, v28, v19                              ; 1038271C
>	v_mad_f32 v28, v24, v29, v28                             ; D282001C 04723B18
>	v_mad_f32 v28, v26, v31, v28                             ; D282001C 04723F1A
>	v_add_i32_e32 v29, 0xbc, v27                             ; 4A3A36FF 000000BC
>	buffer_load_dword v29, v29, s[0:3], 0 offen              ; E0301000 80001D1D
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v28, v25, v29, v28                             ; D282001C 04723B19
>	v_mad_f32 v0, v8, v28, v0                                ; D2820000 04023908
>	v_mul_f32_e32 v28, v30, v19                              ; 1038271E
>	v_mad_f32 v28, v24, v32, v28                             ; D282001C 04724118
>	v_mad_f32 v28, v26, v34, v28                             ; D282001C 0472451A
>	v_add_i32_e32 v29, 0xac, v27                             ; 4A3A36FF 000000AC
>	buffer_load_dword v29, v29, s[0:3], 0 offen              ; E0301000 80001D1D
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v28, v25, v29, v28                             ; D282001C 04723B19
>	v_mad_f32 v2, v8, v28, v2                                ; D2820002 040A3908
>	v_mul_f32_e32 v28, v33, v19                              ; 10382721
>	v_mad_f32 v28, v24, v35, v28                             ; D282001C 04724718
>	v_mad_f32 v28, v26, v37, v28                             ; D282001C 04724B1A
>	v_add_i32_e32 v27, 0x9c, v27                             ; 4A3636FF 0000009C
>	buffer_load_dword v27, v27, s[0:3], 0 offen              ; E0301000 80001B1B
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v27, v25, v27, v28                             ; D282001B 04723719
>	v_mad_f32 v3, v8, v27, v3                                ; D2820003 040E3708
>	v_cmp_gt_f32_e64 s[34:35], v9, 0                         ; D0080022 00010109
>	v_cndmask_b32_e64 v27, 0, -1, s[34:35]                   ; D200081B 00898280
>	v_cmp_ne_i32_e64 s[34:35], v27, 0                        ; D10A0022 0001011B
>	s_and_saveexec_b64 s[34:35], s[34:35]                    ; BEA22422
>	s_xor_b64 s[34:35], exec, s[34:35]                       ; 89A2227E
>	s_cbranch_execz BB0_2                                    ; BF880000
>	v_mul_f32_e32 v6, 0x40400000, v22                        ; 100C2CFF 40400000
>	v_cvt_i32_f32_e32 v20, v6                                ; 7E281106
>	v_lshlrev_b32_e32 v20, 4, v20                            ; 34282884
>	v_add_i32_e32 v21, 0xb4, v20                             ; 4A2A28FF 000000B4
>	buffer_load_dword v21, v21, s[0:3], 0 offen              ; E0301000 80001515
>	v_add_i32_e32 v22, 0xb0, v20                             ; 4A2C28FF 000000B0
>	buffer_load_dword v22, v22, s[0:3], 0 offen              ; E0301000 80001616
>	s_waitcnt vmcnt(0) expcnt(0)                             ; BF8C0700
>	v_mul_f32_e32 v23, v22, v15                              ; 102E1F16
>	v_mad_f32 v23, v16, v21, v23                             ; D2820017 045E2B10
>	v_add_i32_e32 v27, 0xb8, v20                             ; 4A3628FF 000000B8
>	buffer_load_dword v27, v27, s[0:3], 0 offen              ; E0301000 80001B1B
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v23, v17, v27, v23                             ; D2820017 045E3711
>	v_mad_f32 v1, v9, v23, v1                                ; D2820001 04062F09
>	v_add_i32_e32 v23, 0xa4, v20                             ; 4A2E28FF 000000A4
>	buffer_load_dword v23, v23, s[0:3], 0 offen              ; E0301000 80001717
>	v_add_i32_e32 v28, 0xa0, v20                             ; 4A3828FF 000000A0
>	buffer_load_dword v28, v28, s[0:3], 0 offen              ; E0301000 80001C1C
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v29, v28, v15                              ; 103A1F1C
>	v_mad_f32 v29, v16, v23, v29                             ; D282001D 04762F10
>	v_add_i32_e32 v30, 0xa8, v20                             ; 4A3C28FF 000000A8
>	buffer_load_dword v30, v30, s[0:3], 0 offen              ; E0301000 80001E1E
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v29, v17, v30, v29                             ; D282001D 04763D11
>	v_mad_f32 v4, v9, v29, v4                                ; D2820004 04123B09
>	v_add_i32_e32 v29, 0x94, v20                             ; 4A3A28FF 00000094
>	buffer_load_dword v29, v29, s[0:3], 0 offen              ; E0301000 80001D1D
>	v_add_i32_e32 v31, 0x90, v20                             ; 4A3E28FF 00000090
>	buffer_load_dword v31, v31, s[0:3], 0 offen              ; E0301000 80001F1F
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v32, v31, v15                              ; 10401F1F
>	v_mad_f32 v32, v16, v29, v32                             ; D2820020 04823B10
>	v_add_i32_e32 v33, 0x98, v20                             ; 4A4228FF 00000098
>	buffer_load_dword v33, v33, s[0:3], 0 offen              ; E0301000 80002121
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v15, v17, v33, v32                             ; D282000F 04824311
>	v_mad_f32 v5, v9, v15, v5                                ; D2820005 04161F09
>	v_mul_f32_e32 v15, v21, v19                              ; 101E2715
>	v_mad_f32 v15, v24, v22, v15                             ; D282000F 043E2D18
>	v_mad_f32 v15, v26, v27, v15                             ; D282000F 043E371A
>	v_add_i32_e32 v16, 0xbc, v20                             ; 4A2028FF 000000BC
>	buffer_load_dword v16, v16, s[0:3], 0 offen              ; E0301000 80001010
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v15, v25, v16, v15                             ; D282000F 043E2119
>	v_mad_f32 v0, v9, v15, v0                                ; D2820000 04021F09
>	v_mul_f32_e32 v15, v23, v19                              ; 101E2717
>	v_mad_f32 v15, v24, v28, v15                             ; D282000F 043E3918
>	v_mad_f32 v15, v26, v30, v15                             ; D282000F 043E3D1A
>	v_add_i32_e32 v16, 0xac, v20                             ; 4A2028FF 000000AC
>	buffer_load_dword v16, v16, s[0:3], 0 offen              ; E0301000 80001010
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v15, v25, v16, v15                             ; D282000F 043E2119
>	v_mad_f32 v2, v9, v15, v2                                ; D2820002 040A1F09
>	v_mul_f32_e32 v15, v29, v19                              ; 101E271D
>	v_mad_f32 v15, v24, v31, v15                             ; D282000F 043E3F18
>	v_mad_f32 v15, v26, v33, v15                             ; D282000F 043E431A
>	v_add_i32_e32 v16, 0x9c, v20                             ; 4A2028FF 0000009C
>	buffer_load_dword v16, v16, s[0:3], 0 offen              ; E0301000 80001010
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v15, v25, v16, v15                             ; D282000F 043E2119
>	v_mad_f32 v3, v9, v15, v3                                ; D2820003 040E1F09
>	s_or_b64 exec, exec, s[34:35]                            ; 88FE227E
>	s_or_b64 exec, exec, s[32:33]                            ; 88FE207E
>	v_mul_f32_e32 v7, v5, v5                                 ; 100E0B05
>	v_mad_f32 v7, v4, v4, v7                                 ; D2820007 041E0904
>	v_mad_f32 v7, v1, v1, v7                                 ; D2820007 041E0301
>	v_max_f32_e32 v7, 0x33d6bf95, v7                         ; 200E0EFF 33D6BF95
>	v_rsq_clamp_f32_e32 v7, v7                               ; 7E0E5907
>	v_mul_f32_e32 v1, v1, v7                                 ; 10020F01
>	v_mul_f32_e32 v4, v4, v7                                 ; 10080F04
>	v_mul_f32_e32 v5, v5, v7                                 ; 100A0F05
>	exp 15, 32, 0, 0, 0, v5, v4, v1, v6                      ; F800020F 06010405
>	s_waitcnt expcnt(0)                                      ; BF8C070F
>	v_add_f32_e32 v1, s20, v2                                ; 06020414
>	v_mul_f32_e32 v1, s22, v1                                ; 10020216
>	v_add_f32_e32 v4, s19, v3                                ; 06080613
>	v_mul_f32_e32 v4, s21, v4                                ; 10080815
>	exp 15, 33, 0, 0, 0, v11, v12, v4, v1                    ; F800021F 01040C0B
>	s_waitcnt expcnt(0)                                      ; BF8C070F
>	v_mov_b32_e32 v1, 1.0                                    ; 7E0202F2
>	exp 15, 34, 0, 0, 0, v3, v2, v0, v1                      ; F800022F 01000203
>	v_mul_f32_e32 v4, s30, v2                                ; 1008041E
>	v_mad_f32 v4, v3, s31, v4                                ; D2820004 04103F03
>	v_mad_f32 v4, v0, s26, v4                                ; D2820004 04103500
>	v_add_f32_e32 v4, s24, v4                                ; 06080818
>	v_mul_f32_e32 v5, s27, v2                                ; 100A041B
>	v_mad_f32 v5, v3, s29, v5                                ; D2820005 04143B03
>	v_mad_f32 v5, v0, s23, v5                                ; D2820005 04142F00
>	v_add_f32_e32 v5, s17, v5                                ; 060A0A11
>	v_rcp_f32_e32 v6, v5                                     ; 7E0C5505
>	v_mul_f32_e32 v7, v6, v4                                 ; 100E0906
>	v_mul_f32_e32 v8, s25, v2                                ; 10100419
>	v_mad_f32 v8, v3, s28, v8                                ; D2820008 04203903
>	v_mad_f32 v8, v0, s18, v8                                ; D2820008 04202500
>	v_add_f32_e32 v8, s16, v8                                ; 06101010
>	v_mul_f32_e32 v6, v6, v8                                 ; 100C1106
>	v_mov_b32_e32 v9, 0                                      ; 7E120280
>	exp 15, 35, 0, 0, 0, v6, v7, v9, v1                      ; F800023F 01090706
>	s_waitcnt expcnt(0)                                      ; BF8C070F
>	v_mul_f32_e32 v6, s12, v2                                ; 100C040C
>	v_mad_f32 v6, v3, s13, v6                                ; D2820006 04181B03
>	v_mad_f32 v6, v0, s11, v6                                ; D2820006 04181700
>	v_add_f32_e32 v6, s7, v6                                 ; 060C0C07
>	v_mad_f32 v6, 0.5, v6, -0.5                              ; D2820006 03C60CF0
>	v_mul_f32_e32 v7, s8, v2                                 ; 100E0408
>	v_mad_f32 v7, v3, s9, v7                                 ; D2820007 041C1303
>	v_mad_f32 v7, v0, s5, v7                                 ; D2820007 041C0B00
>	v_add_f32_e32 v7, s4, v7                                 ; 060E0E04
>	v_mad_f32 v7, -0.5, v7, -0.5                             ; D2820007 03C60EF1
>	exp 15, 36, 0, 0, 0, v6, v7, v9, v1                      ; F800024F 01090706
>	s_waitcnt expcnt(0)                                      ; BF8C070F
>	v_mul_f32_e32 v1, s14, v2                                ; 1002040E
>	v_mad_f32 v1, v3, s15, v1                                ; D2820001 04041F03
>	v_mad_f32 v0, v0, s10, v1                                ; D2820000 04041500
>	v_add_f32_e32 v0, s6, v0                                 ; 06000006
>	exp 15, 12, 0, 1, 0, v8, v4, v0, v5                      ; F80008CF 05000408
>	s_endpgm                                                 ; BF810000
>
>SGPRS: 40
>VGPRS: 40
>Code Size: 1912 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL IN[1], GENERIC[10], PERSPECTIVE
>DCL IN[2], GENERIC[11], PERSPECTIVE
>DCL IN[3], GENERIC[12], PERSPECTIVE
>DCL IN[4], GENERIC[13], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SAMP[1]
>DCL SAMP[2]
>DCL SAMP[3]
>DCL SAMP[4]
>DCL SAMP[5]
>DCL SAMP[6]
>DCL SAMP[7]
>DCL SAMP[8]
>DCL SAMP[9]
>DCL CONST[0..8]
>DCL TEMP[0..9], LOCAL
>IMM[0] FLT32 {    2.2000,     1.0000,    -0.1000,     0.0000}
>IMM[1] FLT32 {    0.0000,     0.0529,     0.8460,     0.5290}
>IMM[2] FLT32 {  199.0000,     0.1000,     0.4545,     1.4427}
>IMM[3] FLT32 {    1.0000,    -1.0000,     0.5000,     0.0000}
>  0: MOV TEMP[0].xy, IN[1].xyyy
>  1: TEX TEMP[0], TEMP[0], SAMP[0], 2D
>  2: POW TEMP[1].x, TEMP[0].xxxx, IMM[0].xxxx
>  3: POW TEMP[1].y, TEMP[0].yyyy, IMM[0].xxxx
>  4: POW TEMP[1].z, TEMP[0].zzzz, IMM[0].xxxx
>  5: POW TEMP[1].w, TEMP[0].wwww, IMM[0].yyyy
>  6: MOV TEMP[0].w, TEMP[1]
>  7: ABS TEMP[2].x, TEMP[1].wwww
>  8: POW TEMP[2].x, TEMP[2].xxxx, IMM[0].xxxx
>  9: MOV TEMP[3].x, TEMP[2].xxxx
> 10: ADD TEMP[4].xyz, TEMP[2].xxxx, IMM[0].zzzz
> 11: FSLT TEMP[5].xyz, TEMP[4].xyzz, IMM[0].wwww
> 12: OR TEMP[6].x, TEMP[5].xxxx, TEMP[5].zzzz
> 13: OR TEMP[5].x, TEMP[6].xxxx, TEMP[5].yyyy
> 14: AND TEMP[5].x, TEMP[5].xxxx, IMM[0].yyyy
> 15: KILL_IF -TEMP[5].xxxx
> 16: MOV TEMP[5].xyz, IN[0].xyzz
> 17: TEX TEMP[5], TEMP[5], SAMP[2], CUBE
> 18: POW TEMP[6].x, TEMP[5].xxxx, IMM[0].xxxx
> 19: POW TEMP[6].y, TEMP[5].yyyy, IMM[0].xxxx
> 20: POW TEMP[6].z, TEMP[5].zzzz, IMM[0].xxxx
> 21: POW TEMP[6].w, TEMP[5].wwww, IMM[0].yyyy
> 22: MUL TEMP[1].xyz, TEMP[1], TEMP[6]
> 23: MOV TEMP[0].xyz, TEMP[1].xyzx
> 24: ADD TEMP[1].xyz, TEMP[0], TEMP[0]
> 25: MOV TEMP[0].xyz, TEMP[1].xyzx
> 26: ADD TEMP[1].yzw, CONST[3].xxyz, -IN[2].xxyz
> 27: MOV TEMP[3].yzw, TEMP[1].zyzw
> 28: DP3 TEMP[5].x, TEMP[1].yzww, TEMP[1].yzww
> 29: MAX TEMP[5].x, TEMP[5].xxxx, IMM[1].xxxx
> 30: RSQ TEMP[5].x, TEMP[5].xxxx
> 31: MUL TEMP[1].xyz, TEMP[5].xxxx, TEMP[1].yzww
> 32: MAD TEMP[3].yzw, TEMP[3], TEMP[5].xxxx, IMM[1].yyzw
> 33: MOV TEMP[5].w, IMM[0].wwww
> 34: MOV TEMP[5].x, TEMP[3].yyyy
> 35: MOV TEMP[5].y, TEMP[3].zzzz
> 36: MOV TEMP[5].z, TEMP[3].wwww
> 37: DP4 TEMP[3].x, TEMP[5], TEMP[5]
> 38: RSQ TEMP[3].x, TEMP[3].xxxx
> 39: MUL TEMP[3].xyz, TEMP[5], TEMP[3].xxxx
> 40: DP3 TEMP[3].x, TEMP[3].xyzz, IN[0].xyzz
> 41: MOV_SAT TEMP[3].x, TEMP[3].xxxx
> 42: DP3 TEMP[5].x, IN[0].xyzz, TEMP[1].xyzz
> 43: ADD TEMP[5].y, TEMP[5].xxxx, TEMP[5].xxxx
> 44: MAD TEMP[5].yzw, TEMP[5].yyyy, IN[0].xxyz, -TEMP[1].xxyz
> 45: MOV TEMP[6].xy, TEMP[5].yzzz
> 46: TEX TEMP[6], TEMP[6], SAMP[4], 2D
> 47: POW TEMP[7].x, TEMP[6].xxxx, IMM[0].xxxx
> 48: POW TEMP[7].y, TEMP[6].yyyy, IMM[0].xxxx
> 49: POW TEMP[7].z, TEMP[6].zzzz, IMM[0].xxxx
> 50: POW TEMP[7].w, TEMP[6].wwww, IMM[0].yyyy
> 51: MOV TEMP[4].w, TEMP[7].wwww
> 52: MOV TEMP[6].xy, IN[1].xyyy
> 53: TEX TEMP[6], TEMP[6], SAMP[1], 2D
> 54: POW TEMP[8].x, TEMP[6].xxxx, IMM[0].xxxx
> 55: POW TEMP[8].y, TEMP[6].yyyy, IMM[0].xxxx
> 56: POW TEMP[8].z, TEMP[6].zzzz, IMM[0].xxxx
> 57: POW TEMP[8].w, TEMP[6].wwww, IMM[0].yyyy
> 58: ABS TEMP[6].x, TEMP[8].wwww
> 59: POW TEMP[6].x, TEMP[6].xxxx, IMM[0].xxxx
> 60: MOV TEMP[5].xyz, TEMP[5].yzww
> 61: TEX TEMP[5], TEMP[5], SAMP[3], CUBE
> 62: POW TEMP[9].x, TEMP[5].xxxx, IMM[0].xxxx
> 63: POW TEMP[9].y, TEMP[5].yyyy, IMM[0].xxxx
> 64: POW TEMP[9].z, TEMP[5].zzzz, IMM[0].xxxx
> 65: POW TEMP[9].w, TEMP[5].wwww, IMM[0].yyyy
> 66: LRP TEMP[5].xyz, TEMP[6].xxxx, TEMP[9], TEMP[7]
> 67: MAD TEMP[6].y, TEMP[6].xxxx, IMM[2].xxxx, IMM[0].yyyy
> 68: ABS TEMP[3].x, TEMP[3].xxxx
> 69: POW TEMP[3].x, TEMP[3].xxxx, TEMP[6].yyyy
> 70: MUL TEMP[6].w, TEMP[6].yyyy, IMM[2].yyyy
> 71: MOV TEMP[0].w, TEMP[6].wwww
> 72: MAD TEMP[3].yzw, TEMP[6].wwww, TEMP[3].xxxx, TEMP[5].xxyz
> 73: MAD TEMP[3].xyz, TEMP[3].yzww, TEMP[8], TEMP[0]
> 74: ABS TEMP[5].x, TEMP[3].xxxx
> 75: LG2 TEMP[4].x, TEMP[5].xxxx
> 76: ABS TEMP[5].x, TEMP[3].yyyy
> 77: LG2 TEMP[5].x, TEMP[5].xxxx
> 78: MOV TEMP[4].y, TEMP[5].xxxx
> 79: ABS TEMP[3].x, TEMP[3].zzzz
> 80: LG2 TEMP[3].x, TEMP[3].xxxx
> 81: MOV TEMP[4].z, TEMP[3].xxxx
> 82: MUL TEMP[3].xyz, TEMP[4], IMM[2].zzzz
> 83: EX2 TEMP[4].x, TEMP[3].xxxx
> 84: EX2 TEMP[5].x, TEMP[3].yyyy
> 85: MOV TEMP[4].y, TEMP[5].xxxx
> 86: EX2 TEMP[3].x, TEMP[3].zzzz
> 87: MOV TEMP[4].z, TEMP[3].xxxx
> 88: MOV TEMP[3].xyz, TEMP[4].xyzz
> 89: TEX TEMP[3].xyz, TEMP[3], SAMP[6], 3D
> 90: MAD TEMP[5].xy, IN[4], IMM[3].xyxx, IMM[0].wyww
> 91: MOV TEMP[5].xy, TEMP[5].xyyy
> 92: TEX TEMP[5].xzw, TEMP[5], SAMP[7], 2D
> 93: MOV TEMP[1].w, TEMP[5].wwww
> 94: MOV TEMP[6].xy, IN[1].zwww
> 95: TEX TEMP[6].xyz, TEMP[6], SAMP[8], 2D
> 96: LRP TEMP[3].yzw, TEMP[5].xxxx, TEMP[3].xxyz, TEMP[6].xxyz
> 97: MAD TEMP[5].x, TEMP[5].zzzz, -TEMP[5].xxxx, TEMP[5].xxxx
> 98: MOV TEMP[6].xyz, TEMP[3].yzww
> 99: TEX TEMP[6], TEMP[6], SAMP[9], 3D
>100: LRP TEMP[3].xyz, TEMP[5].xxxx, TEMP[6], TEMP[3].yzww
>101: MOV TEMP[1].xyz, TEMP[3].xyzx
>102: ADD TEMP[3].xyz, -TEMP[1], CONST[5]
>103: MOV TEMP[0].xyz, TEMP[3].xyzx
>104: MUL TEMP[3].z, CONST[8].xxxx, IN[2].zzzz
>105: MOV TEMP[4].z, TEMP[3].zzzz
>106: MOV TEMP[4].xy, IN[2].xyxx
>107: ADD TEMP[3].yzw, TEMP[4].xxyz, -CONST[6].xxyz
>108: MUL TEMP[5].w, TEMP[3].wwww, CONST[4].xxxx
>109: MUL TEMP[5].w, TEMP[5].wwww, IMM[2].wwww
>110: EX2 TEMP[5].x, TEMP[5].wwww
>111: ADD TEMP[5].w, -TEMP[5].xxxx, IMM[0].yyyy
>112: DP3 TEMP[6].x, TEMP[3].yzww, TEMP[3].yzww
>113: RCP TEMP[3].x, TEMP[3].wwww
>114: MUL TEMP[6].y, TEMP[6].xxxx, CONST[4].yyyy
>115: MUL TEMP[5].w, TEMP[5].wwww, TEMP[6].yyyy
>116: MUL TEMP[3].w, TEMP[3].xxxx, TEMP[5].wwww
>117: MUL TEMP[3].w, TEMP[3].wwww, IMM[2].wwww
>118: EX2 TEMP[3].x, TEMP[3].wwww
>119: MOV_SAT TEMP[3].x, TEMP[3].xxxx
>120: ADD TEMP[3].w, -TEMP[3].xxxx, IMM[0].yyyy
>121: MAD TEMP[5].x, IN[3].yyyy, IMM[3].zzzz, IMM[3].zzzz
>122: MOV TEMP[4].x, TEMP[5].xxxx
>123: MOV TEMP[4].y, CONST[4].wwww
>124: MOV TEMP[4].xy, TEMP[4].xyyy
>125: TEX TEMP[4].x, TEMP[4], SAMP[5], 2D
>126: MUL TEMP[3].w, TEMP[3].wwww, TEMP[4].xxxx
>127: MOV TEMP[0].w, TEMP[3].wwww
>128: MAD TEMP[0].xyz, TEMP[3].wwww, TEMP[0], TEMP[1]
>129: MOV TEMP[0].xyz, TEMP[0].xyzx
>130: MUL TEMP[1].x, TEMP[2].xxxx, IMM[3].zzzz
>131: ADD TEMP[3].y, -CONST[7].xxxx, IN[2].zzzz
>132: FSGE TEMP[3].x, TEMP[3].yyyy, IMM[0].wwww
>133: UIF TEMP[3].xxxx :0
>134:   MOV TEMP[2].x, TEMP[2].xxxx
>135: ELSE :0
>136:   MOV TEMP[2].x, TEMP[1].xxxx
>137: ENDIF
>138: MOV TEMP[0].w, TEMP[2].xxxx
>139: MOV OUT[0], TEMP[0]
>140: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %23 = load <16 x i8> addrspace(2)* %22, !tbaa !0
>  %24 = call float @llvm.SI.load.const(<16 x i8> %23, i32 48)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %23, i32 52)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %23, i32 56)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %23, i32 64)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %23, i32 68)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %23, i32 76)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %23, i32 80)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %23, i32 84)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %23, i32 88)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %23, i32 96)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %23, i32 100)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %23, i32 104)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %23, i32 112)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %23, i32 128)
>  %38 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %39 = load <8 x i32> addrspace(2)* %38, !tbaa !0
>  %40 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %41 = load <4 x i32> addrspace(2)* %40, !tbaa !0
>  %42 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 1
>  %43 = load <8 x i32> addrspace(2)* %42, !tbaa !0
>  %44 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 1
>  %45 = load <4 x i32> addrspace(2)* %44, !tbaa !0
>  %46 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 2
>  %47 = load <8 x i32> addrspace(2)* %46, !tbaa !0
>  %48 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 2
>  %49 = load <4 x i32> addrspace(2)* %48, !tbaa !0
>  %50 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 3
>  %51 = load <8 x i32> addrspace(2)* %50, !tbaa !0
>  %52 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 3
>  %53 = load <4 x i32> addrspace(2)* %52, !tbaa !0
>  %54 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 4
>  %55 = load <8 x i32> addrspace(2)* %54, !tbaa !0
>  %56 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 4
>  %57 = load <4 x i32> addrspace(2)* %56, !tbaa !0
>  %58 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 5
>  %59 = load <8 x i32> addrspace(2)* %58, !tbaa !0
>  %60 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 5
>  %61 = load <4 x i32> addrspace(2)* %60, !tbaa !0
>  %62 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 6
>  %63 = load <8 x i32> addrspace(2)* %62, !tbaa !0
>  %64 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 6
>  %65 = load <4 x i32> addrspace(2)* %64, !tbaa !0
>  %66 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 7
>  %67 = load <8 x i32> addrspace(2)* %66, !tbaa !0
>  %68 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 7
>  %69 = load <4 x i32> addrspace(2)* %68, !tbaa !0
>  %70 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 8
>  %71 = load <8 x i32> addrspace(2)* %70, !tbaa !0
>  %72 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 8
>  %73 = load <4 x i32> addrspace(2)* %72, !tbaa !0
>  %74 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 9
>  %75 = load <8 x i32> addrspace(2)* %74, !tbaa !0
>  %76 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 9
>  %77 = load <4 x i32> addrspace(2)* %76, !tbaa !0
>  %78 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %79 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %80 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %5, <2 x i32> %7)
>  %81 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %5, <2 x i32> %7)
>  %82 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %5, <2 x i32> %7)
>  %83 = call float @llvm.SI.fs.interp(i32 2, i32 1, i32 %5, <2 x i32> %7)
>  %84 = call float @llvm.SI.fs.interp(i32 3, i32 1, i32 %5, <2 x i32> %7)
>  %85 = call float @llvm.SI.fs.interp(i32 0, i32 2, i32 %5, <2 x i32> %7)
>  %86 = call float @llvm.SI.fs.interp(i32 1, i32 2, i32 %5, <2 x i32> %7)
>  %87 = call float @llvm.SI.fs.interp(i32 2, i32 2, i32 %5, <2 x i32> %7)
>  %88 = call float @llvm.SI.fs.interp(i32 1, i32 3, i32 %5, <2 x i32> %7)
>  %89 = call float @llvm.SI.fs.interp(i32 0, i32 4, i32 %5, <2 x i32> %7)
>  %90 = call float @llvm.SI.fs.interp(i32 1, i32 4, i32 %5, <2 x i32> %7)
>  %91 = bitcast float %81 to i32
>  %92 = bitcast float %82 to i32
>  %93 = insertelement <2 x i32> undef, i32 %91, i32 0
>  %94 = insertelement <2 x i32> %93, i32 %92, i32 1
>  %95 = bitcast <8 x i32> %39 to <32 x i8>
>  %96 = bitcast <4 x i32> %41 to <16 x i8>
>  %97 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %94, <32 x i8> %95, <16 x i8> %96, i32 2)
>  %98 = extractelement <4 x float> %97, i32 0
>  %99 = extractelement <4 x float> %97, i32 1
>  %100 = extractelement <4 x float> %97, i32 2
>  %101 = extractelement <4 x float> %97, i32 3
>  %102 = call float @llvm.pow.f32(float %98, float 0x40019999A0000000)
>  %103 = call float @llvm.pow.f32(float %99, float 0x40019999A0000000)
>  %104 = call float @llvm.pow.f32(float %100, float 0x40019999A0000000)
>  %105 = call float @llvm.pow.f32(float %101, float 1.000000e+00)
>  %106 = call float @fabs(float %105)
>  %107 = call float @llvm.pow.f32(float %106, float 0x40019999A0000000)
>  %108 = fadd float %107, 0xBFB99999A0000000
>  %109 = fadd float %107, 0xBFB99999A0000000
>  %110 = fadd float %107, 0xBFB99999A0000000
>  %111 = fcmp olt float %108, 0.000000e+00
>  %112 = sext i1 %111 to i32
>  %113 = fcmp olt float %109, 0.000000e+00
>  %114 = sext i1 %113 to i32
>  %115 = fcmp olt float %110, 0.000000e+00
>  %116 = sext i1 %115 to i32
>  %117 = bitcast i32 %112 to float
>  %118 = bitcast i32 %114 to float
>  %119 = bitcast i32 %116 to float
>  %120 = bitcast float %117 to i32
>  %121 = bitcast float %119 to i32
>  %122 = or i32 %120, %121
>  %123 = bitcast i32 %122 to float
>  %124 = bitcast float %123 to i32
>  %125 = bitcast float %118 to i32
>  %126 = or i32 %124, %125
>  %127 = bitcast i32 %126 to float
>  %128 = bitcast float %127 to i32
>  %129 = and i32 %128, 1065353216
>  %130 = bitcast i32 %129 to float
>  %131 = fsub float -0.000000e+00, %130
>  %132 = fsub float -0.000000e+00, %130
>  %133 = fsub float -0.000000e+00, %130
>  %134 = fsub float -0.000000e+00, %130
>  call void @llvm.AMDGPU.kill(float %131)
>  call void @llvm.AMDGPU.kill(float %132)
>  call void @llvm.AMDGPU.kill(float %133)
>  call void @llvm.AMDGPU.kill(float %134)
>  %135 = insertelement <4 x float> undef, float %78, i32 0
>  %136 = insertelement <4 x float> %135, float %79, i32 1
>  %137 = insertelement <4 x float> %136, float %80, i32 2
>  %138 = insertelement <4 x float> %137, float 0.000000e+00, i32 3
>  %139 = call <4 x float> @llvm.AMDGPU.cube(<4 x float> %138)
>  %140 = extractelement <4 x float> %139, i32 0
>  %141 = extractelement <4 x float> %139, i32 1
>  %142 = extractelement <4 x float> %139, i32 2
>  %143 = extractelement <4 x float> %139, i32 3
>  %144 = call float @fabs(float %142)
>  %145 = fdiv float 1.000000e+00, %144
>  %146 = fmul float %140, %145
>  %147 = fadd float %146, 1.500000e+00
>  %148 = fmul float %141, %145
>  %149 = fadd float %148, 1.500000e+00
>  %150 = bitcast float %149 to i32
>  %151 = bitcast float %147 to i32
>  %152 = bitcast float %143 to i32
>  %153 = insertelement <4 x i32> undef, i32 %150, i32 0
>  %154 = insertelement <4 x i32> %153, i32 %151, i32 1
>  %155 = insertelement <4 x i32> %154, i32 %152, i32 2
>  %156 = insertelement <4 x i32> %155, i32 undef, i32 3
>  %157 = bitcast <8 x i32> %47 to <32 x i8>
>  %158 = bitcast <4 x i32> %49 to <16 x i8>
>  %159 = call <4 x float> @llvm.SI.sample.v4i32(<4 x i32> %156, <32 x i8> %157, <16 x i8> %158, i32 4)
>  %160 = extractelement <4 x float> %159, i32 0
>  %161 = extractelement <4 x float> %159, i32 1
>  %162 = extractelement <4 x float> %159, i32 2
>  %163 = call float @llvm.pow.f32(float %160, float 0x40019999A0000000)
>  %164 = call float @llvm.pow.f32(float %161, float 0x40019999A0000000)
>  %165 = call float @llvm.pow.f32(float %162, float 0x40019999A0000000)
>  %166 = fmul float %102, %163
>  %167 = fmul float %103, %164
>  %168 = fmul float %104, %165
>  %169 = fadd float %166, %166
>  %170 = fadd float %167, %167
>  %171 = fadd float %168, %168
>  %172 = fsub float -0.000000e+00, %85
>  %173 = fadd float %24, %172
>  %174 = fsub float -0.000000e+00, %86
>  %175 = fadd float %25, %174
>  %176 = fsub float -0.000000e+00, %87
>  %177 = fadd float %26, %176
>  %178 = fmul float %173, %173
>  %179 = fmul float %175, %175
>  %180 = fadd float %179, %178
>  %181 = fmul float %177, %177
>  %182 = fadd float %180, %181
>  %183 = call float @llvm.maxnum.f32(float %182, float 0x3E7AD7F2A0000000)
>  %184 = call float @llvm.AMDGPU.rsq.clamped.f32(float %183)
>  %185 = fmul float %184, %173
>  %186 = fmul float %184, %175
>  %187 = fmul float %184, %177
>  %188 = fmul float %173, %184
>  %189 = fadd float %188, 0x3FAB15B580000000
>  %190 = fmul float %175, %184
>  %191 = fadd float %190, 0x3FEB126EA0000000
>  %192 = fmul float %177, %184
>  %193 = fadd float %192, 0x3FE0ED9160000000
>  %194 = fmul float %189, %189
>  %195 = fmul float %191, %191
>  %196 = fadd float %194, %195
>  %197 = fmul float %193, %193
>  %198 = fadd float %196, %197
>  %199 = fmul float 0.000000e+00, 0.000000e+00
>  %200 = fadd float %198, %199
>  %201 = call float @llvm.AMDGPU.rsq.clamped.f32(float %200)
>  %202 = fmul float %189, %201
>  %203 = fmul float %191, %201
>  %204 = fmul float %193, %201
>  %205 = fmul float %202, %78
>  %206 = fmul float %203, %79
>  %207 = fadd float %206, %205
>  %208 = fmul float %204, %80
>  %209 = fadd float %207, %208
>  %210 = call float @llvm.AMDIL.clamp.(float %209, float 0.000000e+00, float 1.000000e+00)
>  %211 = fmul float %78, %185
>  %212 = fmul float %79, %186
>  %213 = fadd float %212, %211
>  %214 = fmul float %80, %187
>  %215 = fadd float %213, %214
>  %216 = fadd float %215, %215
>  %217 = fsub float -0.000000e+00, %185
>  %218 = fmul float %216, %78
>  %219 = fadd float %218, %217
>  %220 = fsub float -0.000000e+00, %186
>  %221 = fmul float %216, %79
>  %222 = fadd float %221, %220
>  %223 = fsub float -0.000000e+00, %187
>  %224 = fmul float %216, %80
>  %225 = fadd float %224, %223
>  %226 = bitcast float %219 to i32
>  %227 = bitcast float %222 to i32
>  %228 = insertelement <2 x i32> undef, i32 %226, i32 0
>  %229 = insertelement <2 x i32> %228, i32 %227, i32 1
>  %230 = bitcast <8 x i32> %55 to <32 x i8>
>  %231 = bitcast <4 x i32> %57 to <16 x i8>
>  %232 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %229, <32 x i8> %230, <16 x i8> %231, i32 2)
>  %233 = extractelement <4 x float> %232, i32 0
>  %234 = extractelement <4 x float> %232, i32 1
>  %235 = extractelement <4 x float> %232, i32 2
>  %236 = call float @llvm.pow.f32(float %233, float 0x40019999A0000000)
>  %237 = call float @llvm.pow.f32(float %234, float 0x40019999A0000000)
>  %238 = call float @llvm.pow.f32(float %235, float 0x40019999A0000000)
>  %239 = bitcast float %81 to i32
>  %240 = bitcast float %82 to i32
>  %241 = insertelement <2 x i32> undef, i32 %239, i32 0
>  %242 = insertelement <2 x i32> %241, i32 %240, i32 1
>  %243 = bitcast <8 x i32> %43 to <32 x i8>
>  %244 = bitcast <4 x i32> %45 to <16 x i8>
>  %245 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %242, <32 x i8> %243, <16 x i8> %244, i32 2)
>  %246 = extractelement <4 x float> %245, i32 0
>  %247 = extractelement <4 x float> %245, i32 1
>  %248 = extractelement <4 x float> %245, i32 2
>  %249 = extractelement <4 x float> %245, i32 3
>  %250 = call float @llvm.pow.f32(float %246, float 0x40019999A0000000)
>  %251 = call float @llvm.pow.f32(float %247, float 0x40019999A0000000)
>  %252 = call float @llvm.pow.f32(float %248, float 0x40019999A0000000)
>  %253 = call float @llvm.pow.f32(float %249, float 1.000000e+00)
>  %254 = call float @fabs(float %253)
>  %255 = call float @llvm.pow.f32(float %254, float 0x40019999A0000000)
>  %256 = insertelement <4 x float> undef, float %219, i32 0
>  %257 = insertelement <4 x float> %256, float %222, i32 1
>  %258 = insertelement <4 x float> %257, float %225, i32 2
>  %259 = insertelement <4 x float> %258, float %225, i32 3
>  %260 = call <4 x float> @llvm.AMDGPU.cube(<4 x float> %259)
>  %261 = extractelement <4 x float> %260, i32 0
>  %262 = extractelement <4 x float> %260, i32 1
>  %263 = extractelement <4 x float> %260, i32 2
>  %264 = extractelement <4 x float> %260, i32 3
>  %265 = call float @fabs(float %263)
>  %266 = fdiv float 1.000000e+00, %265
>  %267 = fmul float %261, %266
>  %268 = fadd float %267, 1.500000e+00
>  %269 = fmul float %262, %266
>  %270 = fadd float %269, 1.500000e+00
>  %271 = bitcast float %270 to i32
>  %272 = bitcast float %268 to i32
>  %273 = bitcast float %264 to i32
>  %274 = insertelement <4 x i32> undef, i32 %271, i32 0
>  %275 = insertelement <4 x i32> %274, i32 %272, i32 1
>  %276 = insertelement <4 x i32> %275, i32 %273, i32 2
>  %277 = insertelement <4 x i32> %276, i32 undef, i32 3
>  %278 = bitcast <8 x i32> %51 to <32 x i8>
>  %279 = bitcast <4 x i32> %53 to <16 x i8>
>  %280 = call <4 x float> @llvm.SI.sample.v4i32(<4 x i32> %277, <32 x i8> %278, <16 x i8> %279, i32 4)
>  %281 = extractelement <4 x float> %280, i32 0
>  %282 = extractelement <4 x float> %280, i32 1
>  %283 = extractelement <4 x float> %280, i32 2
>  %284 = call float @llvm.pow.f32(float %281, float 0x40019999A0000000)
>  %285 = call float @llvm.pow.f32(float %282, float 0x40019999A0000000)
>  %286 = call float @llvm.pow.f32(float %283, float 0x40019999A0000000)
>  %287 = call float @llvm.AMDGPU.lrp(float %255, float %284, float %236)
>  %288 = call float @llvm.AMDGPU.lrp(float %255, float %285, float %237)
>  %289 = call float @llvm.AMDGPU.lrp(float %255, float %286, float %238)
>  %290 = fmul float %255, 1.990000e+02
>  %291 = fadd float %290, 1.000000e+00
>  %292 = call float @fabs(float %210)
>  %293 = call float @llvm.pow.f32(float %292, float %291)
>  %294 = fmul float %291, 0x3FB99999A0000000
>  %295 = fmul float %294, %293
>  %296 = fadd float %295, %287
>  %297 = fmul float %294, %293
>  %298 = fadd float %297, %288
>  %299 = fmul float %294, %293
>  %300 = fadd float %299, %289
>  %301 = fmul float %296, %250
>  %302 = fadd float %301, %169
>  %303 = fmul float %298, %251
>  %304 = fadd float %303, %170
>  %305 = fmul float %300, %252
>  %306 = fadd float %305, %171
>  %307 = call float @fabs(float %302)
>  %308 = call float @llvm.log2.f32(float %307)
>  %309 = call float @fabs(float %304)
>  %310 = call float @llvm.log2.f32(float %309)
>  %311 = call float @fabs(float %306)
>  %312 = call float @llvm.log2.f32(float %311)
>  %313 = fmul float %308, 0x3FDD1743E0000000
>  %314 = fmul float %310, 0x3FDD1743E0000000
>  %315 = fmul float %312, 0x3FDD1743E0000000
>  %316 = call float @llvm.AMDIL.exp.(float %313)
>  %317 = call float @llvm.AMDIL.exp.(float %314)
>  %318 = call float @llvm.AMDIL.exp.(float %315)
>  %319 = bitcast float %316 to i32
>  %320 = bitcast float %317 to i32
>  %321 = bitcast float %318 to i32
>  %322 = insertelement <4 x i32> undef, i32 %319, i32 0
>  %323 = insertelement <4 x i32> %322, i32 %320, i32 1
>  %324 = insertelement <4 x i32> %323, i32 %321, i32 2
>  %325 = insertelement <4 x i32> %324, i32 undef, i32 3
>  %326 = bitcast <8 x i32> %63 to <32 x i8>
>  %327 = bitcast <4 x i32> %65 to <16 x i8>
>  %328 = call <4 x float> @llvm.SI.sample.v4i32(<4 x i32> %325, <32 x i8> %326, <16 x i8> %327, i32 3)
>  %329 = extractelement <4 x float> %328, i32 0
>  %330 = extractelement <4 x float> %328, i32 1
>  %331 = extractelement <4 x float> %328, i32 2
>  %332 = fmul float %89, 1.000000e+00
>  %333 = fadd float %332, 0.000000e+00
>  %334 = fmul float %90, -1.000000e+00
>  %335 = fadd float %334, 1.000000e+00
>  %336 = bitcast float %333 to i32
>  %337 = bitcast float %335 to i32
>  %338 = insertelement <2 x i32> undef, i32 %336, i32 0
>  %339 = insertelement <2 x i32> %338, i32 %337, i32 1
>  %340 = bitcast <8 x i32> %67 to <32 x i8>
>  %341 = bitcast <4 x i32> %69 to <16 x i8>
>  %342 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %339, <32 x i8> %340, <16 x i8> %341, i32 2)
>  %343 = extractelement <4 x float> %342, i32 0
>  %344 = extractelement <4 x float> %342, i32 2
>  %345 = bitcast float %83 to i32
>  %346 = bitcast float %84 to i32
>  %347 = insertelement <2 x i32> undef, i32 %345, i32 0
>  %348 = insertelement <2 x i32> %347, i32 %346, i32 1
>  %349 = bitcast <8 x i32> %71 to <32 x i8>
>  %350 = bitcast <4 x i32> %73 to <16 x i8>
>  %351 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %348, <32 x i8> %349, <16 x i8> %350, i32 2)
>  %352 = extractelement <4 x float> %351, i32 0
>  %353 = extractelement <4 x float> %351, i32 1
>  %354 = extractelement <4 x float> %351, i32 2
>  %355 = call float @llvm.AMDGPU.lrp(float %343, float %329, float %352)
>  %356 = call float @llvm.AMDGPU.lrp(float %343, float %330, float %353)
>  %357 = call float @llvm.AMDGPU.lrp(float %343, float %331, float %354)
>  %358 = fsub float -0.000000e+00, %343
>  %359 = fmul float %344, %358
>  %360 = fadd float %359, %343
>  %361 = bitcast float %355 to i32
>  %362 = bitcast float %356 to i32
>  %363 = bitcast float %357 to i32
>  %364 = insertelement <4 x i32> undef, i32 %361, i32 0
>  %365 = insertelement <4 x i32> %364, i32 %362, i32 1
>  %366 = insertelement <4 x i32> %365, i32 %363, i32 2
>  %367 = insertelement <4 x i32> %366, i32 undef, i32 3
>  %368 = bitcast <8 x i32> %75 to <32 x i8>
>  %369 = bitcast <4 x i32> %77 to <16 x i8>
>  %370 = call <4 x float> @llvm.SI.sample.v4i32(<4 x i32> %367, <32 x i8> %368, <16 x i8> %369, i32 3)
>  %371 = extractelement <4 x float> %370, i32 0
>  %372 = extractelement <4 x float> %370, i32 1
>  %373 = extractelement <4 x float> %370, i32 2
>  %374 = call float @llvm.AMDGPU.lrp(float %360, float %371, float %355)
>  %375 = call float @llvm.AMDGPU.lrp(float %360, float %372, float %356)
>  %376 = call float @llvm.AMDGPU.lrp(float %360, float %373, float %357)
>  %377 = fsub float -0.000000e+00, %374
>  %378 = fadd float %377, %30
>  %379 = fsub float -0.000000e+00, %375
>  %380 = fadd float %379, %31
>  %381 = fsub float -0.000000e+00, %376
>  %382 = fadd float %381, %32
>  %383 = fmul float %37, %87
>  %384 = fsub float -0.000000e+00, %33
>  %385 = fadd float %85, %384
>  %386 = fsub float -0.000000e+00, %34
>  %387 = fadd float %86, %386
>  %388 = fsub float -0.000000e+00, %35
>  %389 = fadd float %383, %388
>  %390 = fmul float %389, %27
>  %391 = fmul float %390, 0x3FF7154CA0000000
>  %392 = call float @llvm.AMDIL.exp.(float %391)
>  %393 = fsub float -0.000000e+00, %392
>  %394 = fadd float %393, 1.000000e+00
>  %395 = fmul float %385, %385
>  %396 = fmul float %387, %387
>  %397 = fadd float %396, %395
>  %398 = fmul float %389, %389
>  %399 = fadd float %397, %398
>  %400 = fdiv float 1.000000e+00, %389
>  %401 = fmul float %399, %28
>  %402 = fmul float %394, %401
>  %403 = fmul float %400, %402
>  %404 = fmul float %403, 0x3FF7154CA0000000
>  %405 = call float @llvm.AMDIL.exp.(float %404)
>  %406 = call float @llvm.AMDIL.clamp.(float %405, float 0.000000e+00, float 1.000000e+00)
>  %407 = fsub float -0.000000e+00, %406
>  %408 = fadd float %407, 1.000000e+00
>  %409 = fmul float %88, 5.000000e-01
>  %410 = fadd float %409, 5.000000e-01
>  %411 = bitcast float %410 to i32
>  %412 = bitcast float %29 to i32
>  %413 = insertelement <2 x i32> undef, i32 %411, i32 0
>  %414 = insertelement <2 x i32> %413, i32 %412, i32 1
>  %415 = bitcast <8 x i32> %59 to <32 x i8>
>  %416 = bitcast <4 x i32> %61 to <16 x i8>
>  %417 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %414, <32 x i8> %415, <16 x i8> %416, i32 2)
>  %418 = extractelement <4 x float> %417, i32 0
>  %419 = fmul float %408, %418
>  %420 = fmul float %419, %378
>  %421 = fadd float %420, %374
>  %422 = fmul float %419, %380
>  %423 = fadd float %422, %375
>  %424 = fmul float %419, %382
>  %425 = fadd float %424, %376
>  %426 = fmul float %107, 5.000000e-01
>  %427 = fsub float -0.000000e+00, %36
>  %428 = fadd float %427, %87
>  %429 = fcmp oge float %428, 0.000000e+00
>  %430 = sext i1 %429 to i32
>  %431 = bitcast i32 %430 to float
>  %432 = bitcast float %431 to i32
>  %433 = icmp ne i32 %432, 0
>  %. = select i1 %433, float %107, float %426
>  %434 = call i32 @llvm.SI.packf16(float %421, float %423)
>  %435 = bitcast i32 %434 to float
>  %436 = call i32 @llvm.SI.packf16(float %425, float %.)
>  %437 = bitcast i32 %436 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %435, float %437, float %435, float %437)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.pow.f32(float, float) #1
>
>; Function Attrs: readnone
>declare float @fabs(float) #2
>
>declare void @llvm.AMDGPU.kill(float)
>
>; Function Attrs: readnone
>declare <4 x float> @llvm.AMDGPU.cube(<4 x float>) #2
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v4i32(<4 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.maxnum.f32(float, float) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.AMDGPU.rsq.clamped.f32(float) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.clamp.(float, float, float) #2
>
>; Function Attrs: readnone
>declare float @llvm.AMDGPU.lrp(float, float, float) #2
>
>; Function Attrs: nounwind readnone
>declare float @llvm.log2.f32(float) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.exp.(float) #2
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                        ; BEFE0A7E
>	s_mov_b64 vcc, s[6:7]                                                       ; BEEA0406
>	s_mov_b64 s[100:101], s[2:3]                                                ; BEE40402
>	s_mov_b32 m0, s9                                                            ; BEFC0309
>	v_interp_p1_f32 v3, v0, 1, 1, [m0]                                          ; C80C0500
>	v_interp_p2_f32 v3, [v3], v1, 1, 1, [m0]                                    ; C80D0501
>	v_interp_p1_f32 v2, v0, 0, 1, [m0]                                          ; C8080400
>	v_interp_p2_f32 v2, [v2], v1, 0, 1, [m0]                                    ; C8090401
>	s_load_dwordx4 s[8:11], s[4:5], 0x0                                         ; C0840500
>	s_load_dwordx4 s[64:67], s[4:5], 0x4                                        ; C0A00504
>	s_load_dwordx4 s[52:55], s[4:5], 0x8                                        ; C09A0508
>	s_load_dwordx4 s[80:83], s[4:5], 0xc                                        ; C0A8050C
>	s_load_dwordx4 s[68:71], s[4:5], 0x10                                       ; C0A20510
>	s_load_dwordx4 s[0:3], s[4:5], 0x14                                         ; C0800514
>	s_waitcnt lgkmcnt(0)                                                        ; BF8C007F
>	v_writelane_b32 v44, s0, 0                                                  ; 04590000
>	v_writelane_b32 v44, s1, 1                                                  ; 04590201
>	v_writelane_b32 v44, s2, 2                                                  ; 04590402
>	v_writelane_b32 v44, s3, 3                                                  ; 04590603
>	s_load_dwordx4 s[40:43], s[4:5], 0x18                                       ; C0940518
>	s_load_dwordx4 s[28:31], s[4:5], 0x1c                                       ; C08E051C
>	s_load_dwordx4 s[24:27], s[4:5], 0x20                                       ; C08C0520
>	s_load_dwordx4 s[20:23], s[4:5], 0x24                                       ; C08A0524
>	s_load_dwordx8 s[0:7], vcc, 0x0                                             ; C0C06B00
>	s_load_dwordx8 s[72:79], vcc, 0x8                                           ; C0E46B08
>	s_load_dwordx8 s[56:63], vcc, 0x10                                          ; C0DC6B10
>	s_load_dwordx8 s[92:99], vcc, 0x18                                          ; C0EE6B18
>	s_load_dwordx8 s[84:91], vcc, 0x20                                          ; C0EA6B20
>	s_load_dwordx8 s[12:19], vcc, 0x28                                          ; C0C66B28
>	s_load_dwordx8 s[44:51], vcc, 0x30                                          ; C0D66B30
>	s_load_dwordx8 s[32:39], vcc, 0x38                                          ; C0D06B38
>	s_waitcnt lgkmcnt(0)                                                        ; BF8C007F
>	image_sample v[4:7], 15, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[0:7], s[8:11]       ; F0800F00 00400402
>	s_waitcnt vmcnt(0)                                                          ; BF8C0770
>	v_log_f32_e32 v8, v7                                                        ; 7E104F07
>	v_mul_legacy_f32_e32 v8, 1.0, v8                                            ; 0E1010F2
>	v_exp_f32_e32 v8, v8                                                        ; 7E104B08
>	v_mov_b32_e32 v9, 0x7fffffff                                                ; 7E1202FF 7FFFFFFF
>	v_and_b32_e32 v8, v8, v9                                                    ; 36101308
>	v_log_f32_e32 v8, v8                                                        ; 7E104F08
>	v_mov_b32_e32 v10, 0x400ccccd                                               ; 7E1402FF 400CCCCD
>	v_mul_legacy_f32_e32 v8, v10, v8                                            ; 0E10110A
>	v_exp_f32_e32 v8, v8                                                        ; 7E104B08
>	v_mov_b32_e32 v11, 0xbdcccccd                                               ; 7E1602FF BDCCCCCD
>	v_add_f32_e32 v11, v8, v11                                                  ; 06161708
>	v_mov_b32_e32 v15, 0                                                        ; 7E1E0280
>	v_cmp_lt_f32_e64 s[0:1], v11, 0                                             ; D0020000 0001010B
>	v_cndmask_b32_e64 v11, 0, -1, s[0:1]                                        ; D200080B 00018280
>	v_and_b32_e32 v11, 1.0, v11                                                 ; 361616F2
>	v_xor_b32_e32 v11, 0x80000000, v11                                          ; 3A1616FF 80000000
>	v_cmpx_le_f32_e32 vcc, 0, v11                                               ; 7C261680
>	v_cmpx_le_f32_e32 vcc, 0, v11                                               ; 7C261680
>	v_cmpx_le_f32_e32 vcc, 0, v11                                               ; 7C261680
>	v_cmpx_le_f32_e32 vcc, 0, v11                                               ; 7C261680
>	v_interp_p1_f32 v11, v0, 1, 2, [m0]                                         ; C82C0900
>	v_interp_p2_f32 v11, [v11], v1, 1, 2, [m0]                                  ; C82D0901
>	s_load_dwordx4 s[0:3], s[100:101], 0x0                                      ; C0806500
>	s_waitcnt lgkmcnt(0)                                                        ; BF8C007F
>	s_buffer_load_dword s4, s[0:3], 0xd                                         ; C202010D
>	s_waitcnt lgkmcnt(0)                                                        ; BF8C007F
>	v_sub_f32_e32 v16, s4, v11                                                  ; 08201604
>	v_interp_p1_f32 v17, v0, 0, 2, [m0]                                         ; C8440800
>	v_interp_p2_f32 v17, [v17], v1, 0, 2, [m0]                                  ; C8450801
>	s_buffer_load_dword s4, s[0:3], 0xc                                         ; C202010C
>	s_waitcnt lgkmcnt(0)                                                        ; BF8C007F
>	v_sub_f32_e32 v18, s4, v17                                                  ; 08242204
>	v_mul_f32_e32 v19, v18, v18                                                 ; 10262512
>	v_mad_f32 v19, v16, v16, v19                                                ; D2820013 044E2110
>	v_interp_p1_f32 v20, v0, 2, 2, [m0]                                         ; C8500A00
>	v_interp_p2_f32 v20, [v20], v1, 2, 2, [m0]                                  ; C8510A01
>	s_buffer_load_dword s4, s[0:3], 0xe                                         ; C202010E
>	s_waitcnt lgkmcnt(0)                                                        ; BF8C007F
>	v_sub_f32_e32 v21, s4, v20                                                  ; 082A2804
>	v_mad_f32 v19, v21, v21, v19                                                ; D2820013 044E2B15
>	v_max_f32_e32 v19, 0x33d6bf95, v19                                          ; 202626FF 33D6BF95
>	v_rsq_clamp_f32_e32 v19, v19                                                ; 7E265913
>	v_mul_f32_e32 v22, v16, v19                                                 ; 102C2710
>	v_mul_f32_e32 v23, v18, v19                                                 ; 102E2712
>	v_interp_p1_f32 v12, v0, 0, 0, [m0]                                         ; C8300000
>	v_interp_p2_f32 v12, [v12], v1, 0, 0, [m0]                                  ; C8310001
>	v_mul_f32_e32 v24, v23, v12                                                 ; 10301917
>	v_interp_p1_f32 v13, v0, 1, 0, [m0]                                         ; C8340100
>	v_interp_p2_f32 v13, [v13], v1, 1, 0, [m0]                                  ; C8350101
>	v_mad_f32 v24, v13, v22, v24                                                ; D2820018 04622D0D
>	v_mul_f32_e32 v25, v21, v19                                                 ; 10322715
>	v_interp_p1_f32 v14, v0, 2, 0, [m0]                                         ; C8380200
>	v_interp_p2_f32 v14, [v14], v1, 2, 0, [m0]                                  ; C8390201
>	v_mad_f32 v24, v14, v25, v24                                                ; D2820018 0462330E
>	v_add_f32_e32 v24, v24, v24                                                 ; 06303118
>	v_mad_f32 v27, v24, v14, -v25                                               ; D282001B 84661D18
>	v_mad_f32 v26, v24, v13, -v22                                               ; D282001A 845A1B18
>	v_mad_f32 v25, v24, v12, -v23                                               ; D2820019 845E1918
>	v_mov_b32_e32 v28, v27                                                      ; 7E38031B
>	v_cubeid_f32 v32, v25, v26, v27                                             ; D2880020 046E3519
>	v_cubema_f32 v31, v25, v26, v27                                             ; D28E001F 046E3519
>	v_cubesc_f32 v30, v25, v26, v27                                             ; D28A001E 046E3519
>	v_cubetc_f32 v29, v25, v26, v27                                             ; D28C001D 046E3519
>	v_rcp_f32_e64 v22, |v31|                                                    ; D3540116 0000011F
>	v_mov_b32_e32 v23, 0x3fc00000                                               ; 7E2E02FF 3FC00000
>	v_mad_f32 v31, v29, v22, v23                                                ; D282001F 045E2D1D
>	v_mad_f32 v30, v30, v22, v23                                                ; D282001E 045E2D1E
>	image_sample v[29:31], 7, 0, 0, 0, 0, 0, 0, 0, v[30:33], s[92:99], s[80:83] ; F0800700 02971D1E
>	s_waitcnt vmcnt(0)                                                          ; BF8C0770
>	v_log_f32_e32 v22, v31                                                      ; 7E2C4F1F
>	v_mul_legacy_f32_e32 v22, v10, v22                                          ; 0E2C2D0A
>	v_exp_f32_e32 v22, v22                                                      ; 7E2C4B16
>	image_sample v[24:26], 7, 0, 0, 0, 0, 0, 0, 0, v[25:26], s[84:91], s[68:71] ; F0800700 02351819
>	s_waitcnt vmcnt(0)                                                          ; BF8C0770
>	v_log_f32_e32 v27, v26                                                      ; 7E364F1A
>	v_mul_legacy_f32_e32 v27, v10, v27                                          ; 0E36370A
>	v_exp_f32_e32 v27, v27                                                      ; 7E364B1B
>	image_sample v[32:35], 15, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[72:79], s[64:67]  ; F0800F00 02122002
>	s_waitcnt vmcnt(0)                                                          ; BF8C0770
>	v_log_f32_e32 v2, v35                                                       ; 7E044F23
>	v_mul_legacy_f32_e32 v2, 1.0, v2                                            ; 0E0404F2
>	v_exp_f32_e32 v2, v2                                                        ; 7E044B02
>	v_and_b32_e32 v2, v2, v9                                                    ; 36041302
>	v_log_f32_e32 v2, v2                                                        ; 7E044F02
>	v_mul_legacy_f32_e32 v2, v10, v2                                            ; 0E04050A
>	v_exp_f32_e32 v2, v2                                                        ; 7E044B02
>	v_sub_f32_e32 v3, 1.0, v2                                                   ; 080604F2
>	v_mul_f32_e32 v27, v27, v3                                                  ; 1036071B
>	v_mad_f32 v22, v2, v22, v27                                                 ; D2820016 046E2D02
>	v_mov_b32_e32 v27, 0x3d58adac                                               ; 7E3602FF 3D58ADAC
>	v_mad_f32 v18, v18, v19, v27                                                ; D2820012 046E2712
>	v_mov_b32_e32 v27, 0x3f589375                                               ; 7E3602FF 3F589375
>	v_mad_f32 v16, v16, v19, v27                                                ; D2820010 046E2710
>	v_mul_f32_e32 v27, v16, v16                                                 ; 10362110
>	v_mad_f32 v27, v18, v18, v27                                                ; D282001B 046E2512
>	v_mov_b32_e32 v28, 0x3f076c8b                                               ; 7E3802FF 3F076C8B
>	v_mad_f32 v19, v21, v19, v28                                                ; D2820013 04722715
>	v_mad_f32 v21, v19, v19, v27                                                ; D2820015 046E2713
>	v_add_f32_e32 v21, 0, v21                                                   ; 062A2A80
>	v_rsq_clamp_f32_e32 v21, v21                                                ; 7E2A5915
>	v_mul_f32_e32 v16, v21, v16                                                 ; 10202115
>	v_mul_f32_e32 v18, v21, v18                                                 ; 10242515
>	v_mul_f32_e32 v18, v12, v18                                                 ; 1024250C
>	v_mad_f32 v16, v16, v13, v18                                                ; D2820010 044A1B10
>	v_mul_f32_e32 v18, v21, v19                                                 ; 10242715
>	v_mad_f32 v16, v18, v14, v16                                                ; D2820010 04421D12
>	v_add_f32_e64 v16, 0, v16 clamp                                             ; D2060810 00022080
>	v_and_b32_e32 v9, v16, v9                                                   ; 36121310
>	v_log_f32_e32 v9, v9                                                        ; 7E124F09
>	v_mov_b32_e32 v16, 0x43470000                                               ; 7E2002FF 43470000
>	v_mad_f32 v16, v16, v2, 1.0                                                 ; D2820010 03CA0510
>	v_mul_legacy_f32_e32 v9, v16, v9                                            ; 0E121310
>	v_exp_f32_e32 v9, v9                                                        ; 7E124B09
>	v_mul_f32_e32 v16, 0x3dcccccd, v16                                          ; 102020FF 3DCCCCCD
>	v_mad_f32 v18, v16, v9, v22                                                 ; D2820012 045A1310
>	v_log_f32_e32 v19, v34                                                      ; 7E264F22
>	v_mul_legacy_f32_e32 v19, v10, v19                                          ; 0E26270A
>	v_exp_f32_e32 v19, v19                                                      ; 7E264B13
>	v_mul_f32_e32 v18, v19, v18                                                 ; 10242513
>	v_cubeid_f32 v39, v12, v13, v14                                             ; D2880027 043A1B0C
>	v_cubema_f32 v38, v12, v13, v14                                             ; D28E0026 043A1B0C
>	v_cubesc_f32 v37, v12, v13, v14                                             ; D28A0025 043A1B0C
>	v_cubetc_f32 v36, v12, v13, v14                                             ; D28C0024 043A1B0C
>	v_rcp_f32_e64 v12, |v38|                                                    ; D354010C 00000126
>	v_mad_f32 v38, v36, v12, v23                                                ; D2820026 045E1924
>	v_mad_f32 v37, v37, v12, v23                                                ; D2820025 045E1925
>	image_sample v[12:14], 7, 0, 0, 0, 0, 0, 0, 0, v[37:40], s[56:63], s[52:55] ; F0800700 01AE0C25
>	s_waitcnt vmcnt(0)                                                          ; BF8C0770
>	v_log_f32_e32 v15, v14                                                      ; 7E1E4F0E
>	v_mul_legacy_f32_e32 v15, v10, v15                                          ; 0E1E1F0A
>	v_exp_f32_e32 v15, v15                                                      ; 7E1E4B0F
>	v_log_f32_e32 v19, v6                                                       ; 7E264F06
>	v_mul_legacy_f32_e32 v19, v10, v19                                          ; 0E26270A
>	v_exp_f32_e32 v19, v19                                                      ; 7E264B13
>	v_mul_f32_e32 v15, v15, v19                                                 ; 101E270F
>	v_mad_f32 v15, 2.0, v15, v18                                                ; D282000F 044A1EF4
>	v_log_f32_e64 v15, |v15|                                                    ; D34E010F 0000010F
>	v_mul_f32_e32 v15, 0x3ee8ba1f, v15                                          ; 101E1EFF 3EE8BA1F
>	v_exp_f32_e32 v38, v15                                                      ; 7E4C4B0F
>	v_log_f32_e32 v15, v30                                                      ; 7E1E4F1E
>	v_mul_legacy_f32_e32 v15, v10, v15                                          ; 0E1E1F0A
>	v_exp_f32_e32 v15, v15                                                      ; 7E1E4B0F
>	v_log_f32_e32 v18, v25                                                      ; 7E244F19
>	v_mul_legacy_f32_e32 v18, v10, v18                                          ; 0E24250A
>	v_exp_f32_e32 v18, v18                                                      ; 7E244B12
>	v_mul_f32_e32 v18, v18, v3                                                  ; 10240712
>	v_mad_f32 v15, v2, v15, v18                                                 ; D282000F 044A1F02
>	v_mad_f32 v15, v16, v9, v15                                                 ; D282000F 043E1310
>	v_log_f32_e32 v18, v33                                                      ; 7E244F21
>	v_mul_legacy_f32_e32 v18, v10, v18                                          ; 0E24250A
>	v_exp_f32_e32 v18, v18                                                      ; 7E244B12
>	v_mul_f32_e32 v15, v18, v15                                                 ; 101E1F12
>	v_log_f32_e32 v18, v13                                                      ; 7E244F0D
>	v_mul_legacy_f32_e32 v18, v10, v18                                          ; 0E24250A
>	v_exp_f32_e32 v18, v18                                                      ; 7E244B12
>	v_log_f32_e32 v19, v5                                                       ; 7E264F05
>	v_mul_legacy_f32_e32 v19, v10, v19                                          ; 0E26270A
>	v_exp_f32_e32 v19, v19                                                      ; 7E264B13
>	v_mul_f32_e32 v18, v18, v19                                                 ; 10242712
>	v_mad_f32 v15, 2.0, v18, v15                                                ; D282000F 043E24F4
>	v_log_f32_e64 v15, |v15|                                                    ; D34E010F 0000010F
>	v_mul_f32_e32 v15, 0x3ee8ba1f, v15                                          ; 101E1EFF 3EE8BA1F
>	v_exp_f32_e32 v37, v15                                                      ; 7E4A4B0F
>	v_log_f32_e32 v15, v29                                                      ; 7E1E4F1D
>	v_mul_legacy_f32_e32 v15, v10, v15                                          ; 0E1E1F0A
>	v_exp_f32_e32 v15, v15                                                      ; 7E1E4B0F
>	v_log_f32_e32 v18, v24                                                      ; 7E244F18
>	v_mul_legacy_f32_e32 v18, v10, v18                                          ; 0E24250A
>	v_exp_f32_e32 v18, v18                                                      ; 7E244B12
>	v_mul_f32_e32 v3, v18, v3                                                   ; 10060712
>	v_mad_f32 v2, v2, v15, v3                                                   ; D2820002 040E1F02
>	v_mad_f32 v2, v16, v9, v2                                                   ; D2820002 040A1310
>	v_log_f32_e32 v3, v32                                                       ; 7E064F20
>	v_mul_legacy_f32_e32 v3, v10, v3                                            ; 0E06070A
>	v_exp_f32_e32 v3, v3                                                        ; 7E064B03
>	v_mul_f32_e32 v2, v3, v2                                                    ; 10040503
>	v_log_f32_e32 v3, v12                                                       ; 7E064F0C
>	v_mul_legacy_f32_e32 v3, v10, v3                                            ; 0E06070A
>	v_exp_f32_e32 v3, v3                                                        ; 7E064B03
>	v_log_f32_e32 v4, v4                                                        ; 7E084F04
>	v_mul_legacy_f32_e32 v4, v10, v4                                            ; 0E08090A
>	v_exp_f32_e32 v4, v4                                                        ; 7E084B04
>	v_mul_f32_e32 v3, v3, v4                                                    ; 10060903
>	v_mad_f32 v2, 2.0, v3, v2                                                   ; D2820002 040A06F4
>	v_log_f32_e64 v2, |v2|                                                      ; D34E0102 00000102
>	v_mul_f32_e32 v2, 0x3ee8ba1f, v2                                            ; 100404FF 3EE8BA1F
>	v_exp_f32_e32 v36, v2                                                       ; 7E484B02
>	image_sample v[2:4], 7, 0, 0, 0, 0, 0, 0, 0, v[36:39], s[44:51], s[40:43]   ; F0800700 014B0224
>	v_interp_p1_f32 v5, v0, 1, 4, [m0]                                          ; C8141100
>	v_interp_p2_f32 v5, [v5], v1, 1, 4, [m0]                                    ; C8151101
>	v_sub_f32_e32 v6, 1.0, v5                                                   ; 080C0AF2
>	v_interp_p1_f32 v7, v0, 0, 4, [m0]                                          ; C81C1000
>	v_interp_p2_f32 v7, [v7], v1, 0, 4, [m0]                                    ; C81D1001
>	v_add_f32_e32 v5, 0, v7                                                     ; 060A0E80
>	image_sample v[5:6], 5, 0, 0, 0, 0, 0, 0, 0, v[5:6], s[32:39], s[28:31]     ; F0800500 00E80505
>	s_waitcnt vmcnt(0)                                                          ; BF8C0770
>	v_sub_f32_e32 v7, 1.0, v5                                                   ; 080E0AF2
>	v_interp_p1_f32 v10, v0, 3, 1, [m0]                                         ; C8280700
>	v_interp_p2_f32 v10, [v10], v1, 3, 1, [m0]                                  ; C8290701
>	v_interp_p1_f32 v9, v0, 2, 1, [m0]                                          ; C8240600
>	v_interp_p2_f32 v9, [v9], v1, 2, 1, [m0]                                    ; C8250601
>	s_load_dwordx8 s[4:11], vcc, 0x40                                           ; C0C26B40
>	s_waitcnt lgkmcnt(0)                                                        ; BF8C007F
>	image_sample v[12:14], 7, 0, 0, 0, 0, 0, 0, 0, v[9:10], s[4:11], s[24:27]   ; F0800700 00C10C09
>	s_waitcnt vmcnt(0)                                                          ; BF8C0770
>	v_mul_f32_e32 v9, v14, v7                                                   ; 10120F0E
>	v_mad_f32 v23, v5, v4, v9                                                   ; D2820017 04260905
>	v_mul_f32_e32 v9, v13, v7                                                   ; 10120F0D
>	v_mad_f32 v22, v5, v3, v9                                                   ; D2820016 04260705
>	v_mul_f32_e32 v7, v12, v7                                                   ; 100E0F0C
>	v_mad_f32 v21, v5, v2, v7                                                   ; D2820015 041E0505
>	s_load_dwordx8 s[4:11], vcc, 0x48                                           ; C0C26B48
>	s_waitcnt lgkmcnt(0)                                                        ; BF8C007F
>	image_sample v[2:4], 7, 0, 0, 0, 0, 0, 0, 0, v[21:24], s[4:11], s[20:23]    ; F0800700 00A10215
>	v_mad_f32 v5, -v6, v5, v5                                                   ; D2820005 24160B06
>	v_sub_f32_e32 v6, 1.0, v5                                                   ; 080C0AF2
>	v_mul_f32_e32 v7, v22, v6                                                   ; 100E0D16
>	s_waitcnt vmcnt(0)                                                          ; BF8C0770
>	v_mad_f32 v7, v5, v3, v7                                                    ; D2820007 041E0705
>	s_buffer_load_dword s4, s[0:3], 0x15                                        ; C2020115
>	s_waitcnt lgkmcnt(0)                                                        ; BF8C007F
>	v_sub_f32_e32 v9, s4, v7                                                    ; 08120E04
>	s_buffer_load_dword s4, s[0:3], 0x19                                        ; C2020119
>	s_waitcnt lgkmcnt(0)                                                        ; BF8C007F
>	v_subrev_f32_e32 v10, s4, v11                                               ; 0A141604
>	s_buffer_load_dword s4, s[0:3], 0x18                                        ; C2020118
>	s_waitcnt lgkmcnt(0)                                                        ; BF8C007F
>	v_subrev_f32_e32 v11, s4, v17                                               ; 0A162204
>	v_mul_f32_e32 v11, v11, v11                                                 ; 1016170B
>	v_mad_f32 v10, v10, v10, v11                                                ; D282000A 042E150A
>	s_buffer_load_dword s4, s[0:3], 0x1a                                        ; C202011A
>	s_buffer_load_dword s5, s[0:3], 0x20                                        ; C2028120
>	s_waitcnt lgkmcnt(0)                                                        ; BF8C007F
>	v_mov_b32_e32 v11, s4                                                       ; 7E160204
>	v_mad_f32 v11, v20, s5, -v11                                                ; D282000B 842C0B14
>	v_mad_f32 v10, v11, v11, v10                                                ; D282000A 042A170B
>	s_buffer_load_dword s4, s[0:3], 0x11                                        ; C2020111
>	s_waitcnt lgkmcnt(0)                                                        ; BF8C007F
>	v_mul_f32_e32 v10, s4, v10                                                  ; 10141404
>	s_buffer_load_dword s4, s[0:3], 0x10                                        ; C2020110
>	s_waitcnt lgkmcnt(0)                                                        ; BF8C007F
>	v_mul_f32_e32 v12, s4, v11                                                  ; 10181604
>	v_mul_f32_e32 v12, 0x3fb8aa65, v12                                          ; 101818FF 3FB8AA65
>	v_exp_f32_e32 v12, v12                                                      ; 7E184B0C
>	v_sub_f32_e32 v12, 1.0, v12                                                 ; 081818F2
>	v_mul_f32_e32 v10, v10, v12                                                 ; 1014190A
>	v_rcp_f32_e32 v11, v11                                                      ; 7E16550B
>	v_mul_f32_e32 v10, v10, v11                                                 ; 1014170A
>	v_mul_f32_e32 v10, 0x3fb8aa65, v10                                          ; 101414FF 3FB8AA65
>	v_exp_f32_e32 v10, v10                                                      ; 7E144B0A
>	v_add_f32_e64 v10, 0, v10 clamp                                             ; D206080A 00021480
>	v_sub_f32_e32 v10, 1.0, v10                                                 ; 081414F2
>	v_interp_p1_f32 v11, v0, 1, 3, [m0]                                         ; C82C0D00
>	v_interp_p2_f32 v11, [v11], v1, 1, 3, [m0]                                  ; C82D0D01
>	v_mad_f32 v0, 0.5, v11, 0.5                                                 ; D2820000 03C216F0
>	s_buffer_load_dword s4, s[0:3], 0x13                                        ; C2020113
>	s_waitcnt lgkmcnt(0)                                                        ; BF8C007F
>	v_mov_b32_e32 v1, s4                                                        ; 7E020204
>	v_readlane_b32 s4, v44, 0                                                   ; 0209012C
>	v_readlane_b32 s5, v44, 1                                                   ; 020B032C
>	v_readlane_b32 s6, v44, 2                                                   ; 020D052C
>	v_readlane_b32 s7, v44, 3                                                   ; 020F072C
>	s_nop 2                                                                     ; BF800002
>	image_sample v0, 1, 0, 0, 0, 0, 0, 0, 0, v[0:1], s[12:19], s[4:7]           ; F0800100 00230000
>	s_waitcnt vmcnt(0)                                                          ; BF8C0770
>	v_mul_f32_e32 v0, v0, v10                                                   ; 10001500
>	v_mad_f32 v1, v0, v9, v7                                                    ; D2820001 041E1300
>	v_mul_f32_e32 v7, v21, v6                                                   ; 100E0D15
>	v_mad_f32 v7, v5, v2, v7                                                    ; D2820007 041E0505
>	s_buffer_load_dword s4, s[0:3], 0x14                                        ; C2020114
>	s_waitcnt lgkmcnt(0)                                                        ; BF8C007F
>	v_sub_f32_e32 v9, s4, v7                                                    ; 08120E04
>	v_mad_f32 v7, v0, v9, v7                                                    ; D2820007 041E1300
>	v_cvt_pkrtz_f16_f32_e32 v1, v7, v1                                          ; 5E020307
>	v_mul_f32_e32 v6, v23, v6                                                   ; 100C0D17
>	v_mad_f32 v2, v5, v4, v6                                                    ; D2820002 041A0905
>	s_buffer_load_dword s4, s[0:3], 0x16                                        ; C2020116
>	s_waitcnt lgkmcnt(0)                                                        ; BF8C007F
>	v_sub_f32_e32 v3, s4, v2                                                    ; 08060404
>	v_mad_f32 v0, v0, v3, v2                                                    ; D2820000 040A0700
>	v_mul_f32_e32 v2, 0.5, v8                                                   ; 100410F0
>	s_buffer_load_dword s0, s[0:3], 0x1c                                        ; C200011C
>	s_waitcnt lgkmcnt(0)                                                        ; BF8C007F
>	v_subrev_f32_e32 v3, s0, v20                                                ; 0A062800
>	v_cmp_ge_f32_e64 s[0:1], v3, 0                                              ; D00C0000 00010103
>	v_cndmask_b32_e64 v3, 0, -1, s[0:1]                                         ; D2000803 00018280
>	v_cmp_ne_i32_e64 s[0:1], v3, 0                                              ; D10A0000 00010103
>	v_cndmask_b32_e64 v2, v2, v8, s[0:1]                                        ; D2000002 10021102
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v2                                          ; 5E000500
>	exp 15, 0, 1, 1, 1, v1, v0, v1, v0                                          ; F8001C0F 00010001
>	s_endpgm                                                                    ; BF810000
>
>SGPRS: 104
>VGPRS: 48
>Code Size: 1660 bytes
>LDS: 0 blocks
>Scratch: 1024 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL IN[2]
>DCL IN[3]
>DCL IN[4]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL OUT[2], GENERIC[10]
>DCL OUT[3], GENERIC[11]
>DCL OUT[4], GENERIC[12]
>DCL CONST[0..103]
>DCL TEMP[0..5], LOCAL
>DCL ADDR[0]
>IMM[0] FLT32 {    3.0000,     1.0000,     0.0000,     0.0000}
>IMM[1] INT32 {1, 2, 0, 0}
>  0: MUL TEMP[0].x, IMM[0].xxxx, IN[2].xxxx
>  1: MAD TEMP[1], IN[0].xyzx, IMM[0].yyyz, IMM[0].zzzy
>  2: F2I TEMP[2].x, TEMP[0].xxxx
>  3: UARL ADDR[0].x, TEMP[2].xxxx
>  4: UARL ADDR[0].x, TEMP[2].xxxx
>  5: DP4 TEMP[2].x, TEMP[1], CONST[ADDR[0].x+8]
>  6: F2I TEMP[3].x, TEMP[0].xxxx
>  7: UADD TEMP[3].x, IMM[1].xxxx, TEMP[3].xxxx
>  8: UARL ADDR[0].x, TEMP[3].xxxx
>  9: DP4 TEMP[3].x, TEMP[1], CONST[ADDR[0].x+8]
> 10: MOV TEMP[2].y, TEMP[3].xxxx
> 11: F2I TEMP[3].x, TEMP[0].xxxx
> 12: UADD TEMP[3].x, IMM[1].yyyy, TEMP[3].xxxx
> 13: UARL ADDR[0].x, TEMP[3].xxxx
> 14: DP4 TEMP[3].x, TEMP[1], CONST[ADDR[0].x+8]
> 15: MOV TEMP[2].z, TEMP[3].xxxx
> 16: MUL TEMP[3].xyz, TEMP[2], IN[1].xxxx
> 17: MOV TEMP[2].xyz, TEMP[3].xyzx
> 18: F2I TEMP[3].x, TEMP[0].xxxx
> 19: UARL ADDR[0].x, TEMP[3].xxxx
> 20: UARL ADDR[0].x, TEMP[3].xxxx
> 21: DP3 TEMP[3].x, IN[3].xyzz, CONST[ADDR[0].x+8].xyzz
> 22: F2I TEMP[4].x, TEMP[0].xxxx
> 23: UADD TEMP[4].x, IMM[1].xxxx, TEMP[4].xxxx
> 24: UARL ADDR[0].x, TEMP[4].xxxx
> 25: DP3 TEMP[4].x, IN[3].xyzz, CONST[ADDR[0].x+8].xyzz
> 26: MOV TEMP[3].y, TEMP[4].xxxx
> 27: F2I TEMP[0].x, TEMP[0].xxxx
> 28: UADD TEMP[0].x, IMM[1].yyyy, TEMP[0].xxxx
> 29: UARL ADDR[0].x, TEMP[0].xxxx
> 30: DP3 TEMP[0].x, IN[3].xyzz, CONST[ADDR[0].x+8].xyzz
> 31: MOV TEMP[3].z, TEMP[0].xxxx
> 32: MUL TEMP[0].xyz, TEMP[3], IN[1].xxxx
> 33: MOV TEMP[3].xyz, TEMP[0].xyzx
> 34: FSLT TEMP[0].x, IMM[0].zzzz, IN[1].yyyy
> 35: UIF TEMP[0].xxxx :0
> 36:   MUL TEMP[0].w, IMM[0].xxxx, IN[2].yyyy
> 37:   MOV TEMP[2].w, TEMP[0].wwww
> 38:   F2I TEMP[4].x, TEMP[0].wwww
> 39:   UARL ADDR[0].x, TEMP[4].xxxx
> 40:   UARL ADDR[0].x, TEMP[4].xxxx
> 41:   DP4 TEMP[4].x, TEMP[1], CONST[ADDR[0].x+8]
> 42:   F2I TEMP[5].x, TEMP[0].wwww
> 43:   UADD TEMP[5].x, IMM[1].xxxx, TEMP[5].xxxx
> 44:   UARL ADDR[0].x, TEMP[5].xxxx
> 45:   DP4 TEMP[5].x, TEMP[1], CONST[ADDR[0].x+8]
> 46:   MOV TEMP[4].y, TEMP[5].xxxx
> 47:   F2I TEMP[5].x, TEMP[0].wwww
> 48:   UADD TEMP[5].x, IMM[1].yyyy, TEMP[5].xxxx
> 49:   UARL ADDR[0].x, TEMP[5].xxxx
> 50:   DP4 TEMP[5].x, TEMP[1], CONST[ADDR[0].x+8]
> 51:   MOV TEMP[4].z, TEMP[5].xxxx
> 52:   MAD TEMP[5].xyz, IN[1].yyyy, TEMP[4], TEMP[2]
> 53:   MOV TEMP[2].xyz, TEMP[5].xyzx
> 54:   F2I TEMP[5].x, TEMP[0].wwww
> 55:   UARL ADDR[0].x, TEMP[5].xxxx
> 56:   UARL ADDR[0].x, TEMP[5].xxxx
> 57:   DP3 TEMP[4].x, IN[3].xyzz, CONST[ADDR[0].x+8].xyzz
> 58:   F2I TEMP[5].x, TEMP[0].wwww
> 59:   UADD TEMP[5].x, IMM[1].xxxx, TEMP[5].xxxx
> 60:   UARL ADDR[0].x, TEMP[5].xxxx
> 61:   DP3 TEMP[5].x, IN[3].xyzz, CONST[ADDR[0].x+8].xyzz
> 62:   MOV TEMP[4].y, TEMP[5].xxxx
> 63:   F2I TEMP[0].x, TEMP[0].wwww
> 64:   UADD TEMP[0].x, IMM[1].yyyy, TEMP[0].xxxx
> 65:   UARL ADDR[0].x, TEMP[0].xxxx
> 66:   DP3 TEMP[0].x, IN[3].xyzz, CONST[ADDR[0].x+8].xyzz
> 67:   MOV TEMP[4].z, TEMP[0].xxxx
> 68:   MAD TEMP[0].xyz, IN[1].yyyy, TEMP[4], TEMP[3]
> 69:   MOV TEMP[3].xyz, TEMP[0].xyzx
> 70:   FSLT TEMP[0].x, IMM[0].zzzz, IN[1].zzzz
> 71:   UIF TEMP[0].xxxx :0
> 72:     MUL TEMP[0].w, IMM[0].xxxx, IN[2].zzzz
> 73:     MOV TEMP[2].w, TEMP[0].wwww
> 74:     F2I TEMP[5].x, TEMP[0].wwww
> 75:     UARL ADDR[0].x, TEMP[5].xxxx
> 76:     UARL ADDR[0].x, TEMP[5].xxxx
> 77:     DP4 TEMP[4].x, TEMP[1], CONST[ADDR[0].x+8]
> 78:     F2I TEMP[5].x, TEMP[0].wwww
> 79:     UADD TEMP[5].x, IMM[1].xxxx, TEMP[5].xxxx
> 80:     UARL ADDR[0].x, TEMP[5].xxxx
> 81:     DP4 TEMP[5].x, TEMP[1], CONST[ADDR[0].x+8]
> 82:     MOV TEMP[4].y, TEMP[5].xxxx
> 83:     F2I TEMP[5].x, TEMP[0].wwww
> 84:     UADD TEMP[5].x, IMM[1].yyyy, TEMP[5].xxxx
> 85:     UARL ADDR[0].x, TEMP[5].xxxx
> 86:     DP4 TEMP[5].x, TEMP[1], CONST[ADDR[0].x+8]
> 87:     MOV TEMP[4].z, TEMP[5].xxxx
> 88:     MAD TEMP[4].xyz, IN[1].zzzz, TEMP[4], TEMP[2]
> 89:     MOV TEMP[2].xyz, TEMP[4].xyzx
> 90:     F2I TEMP[4].x, TEMP[0].wwww
> 91:     UARL ADDR[0].x, TEMP[4].xxxx
> 92:     UARL ADDR[0].x, TEMP[4].xxxx
> 93:     DP3 TEMP[1].x, IN[3].xyzz, CONST[ADDR[0].x+8].xyzz
> 94:     F2I TEMP[4].x, TEMP[0].wwww
> 95:     UADD TEMP[4].x, IMM[1].xxxx, TEMP[4].xxxx
> 96:     UARL ADDR[0].x, TEMP[4].xxxx
> 97:     DP3 TEMP[4].x, IN[3].xyzz, CONST[ADDR[0].x+8].xyzz
> 98:     MOV TEMP[1].y, TEMP[4].xxxx
> 99:     F2I TEMP[0].x, TEMP[0].wwww
>100:     UADD TEMP[0].x, IMM[1].yyyy, TEMP[0].xxxx
>101:     UARL ADDR[0].x, TEMP[0].xxxx
>102:     DP3 TEMP[0].x, IN[3].xyzz, CONST[ADDR[0].x+8].xyzz
>103:     MOV TEMP[1].z, TEMP[0].xxxx
>104:     MAD TEMP[0].xyz, IN[1].zzzz, TEMP[1], TEMP[3]
>105:     MOV TEMP[3].xyz, TEMP[0].xyzx
>106:   ENDIF
>107: ENDIF
>108: MUL TEMP[1], TEMP[2].yyyy, CONST[5]
>109: MAD TEMP[1], TEMP[2].xxxx, CONST[4], TEMP[1]
>110: MAD TEMP[1], TEMP[2].zzzz, CONST[6], TEMP[1]
>111: ADD TEMP[1], TEMP[1], CONST[7]
>112: ADD TEMP[0].xyz, TEMP[2], -CONST[3]
>113: MOV TEMP[0].xyz, TEMP[0].xyzx
>114: MOV TEMP[3].w, IN[4].xxxx
>115: MOV TEMP[0].w, IN[4].yyyy
>116: MOV TEMP[2].xyz, TEMP[2].xyzx
>117: MOV TEMP[4].xyz, TEMP[1].xywx
>118: MOV TEMP[2].w, IMM[0].yyyy
>119: MOV TEMP[4].w, IMM[0].yyyy
>120: MOV OUT[1], TEMP[3]
>121: MOV OUT[0], TEMP[1]
>122: MOV OUT[2], TEMP[0]
>123: MOV OUT[3], TEMP[2]
>124: MOV OUT[4], TEMP[4]
>125: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 48)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 52)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %12, i32 56)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %12, i32 64)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %12, i32 68)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %12, i32 72)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %12, i32 76)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %12, i32 80)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %12, i32 84)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %12, i32 88)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %12, i32 92)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %12, i32 96)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %12, i32 100)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %12, i32 104)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %12, i32 108)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %12, i32 112)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %12, i32 116)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %12, i32 120)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %12, i32 124)
>  %32 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %33 = load <16 x i8> addrspace(2)* %32, !tbaa !0
>  %34 = add i32 %5, %7
>  %35 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %33, i32 0, i32 %34)
>  %36 = extractelement <4 x float> %35, i32 0
>  %37 = extractelement <4 x float> %35, i32 1
>  %38 = extractelement <4 x float> %35, i32 2
>  %39 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %40 = load <16 x i8> addrspace(2)* %39, !tbaa !0
>  %41 = add i32 %5, %7
>  %42 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %40, i32 0, i32 %41)
>  %43 = extractelement <4 x float> %42, i32 0
>  %44 = extractelement <4 x float> %42, i32 1
>  %45 = extractelement <4 x float> %42, i32 2
>  %46 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 2
>  %47 = load <16 x i8> addrspace(2)* %46, !tbaa !0
>  %48 = add i32 %5, %7
>  %49 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %47, i32 0, i32 %48)
>  %50 = extractelement <4 x float> %49, i32 0
>  %51 = extractelement <4 x float> %49, i32 1
>  %52 = extractelement <4 x float> %49, i32 2
>  %53 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 3
>  %54 = load <16 x i8> addrspace(2)* %53, !tbaa !0
>  %55 = add i32 %5, %7
>  %56 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %54, i32 0, i32 %55)
>  %57 = extractelement <4 x float> %56, i32 0
>  %58 = extractelement <4 x float> %56, i32 1
>  %59 = extractelement <4 x float> %56, i32 2
>  %60 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 4
>  %61 = load <16 x i8> addrspace(2)* %60, !tbaa !0
>  %62 = add i32 %5, %7
>  %63 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %61, i32 0, i32 %62)
>  %64 = extractelement <4 x float> %63, i32 0
>  %65 = extractelement <4 x float> %63, i32 1
>  %66 = fmul float 3.000000e+00, %50
>  %67 = fmul float %36, 1.000000e+00
>  %68 = fadd float %67, 0.000000e+00
>  %69 = fmul float %37, 1.000000e+00
>  %70 = fadd float %69, 0.000000e+00
>  %71 = fmul float %38, 1.000000e+00
>  %72 = fadd float %71, 0.000000e+00
>  %73 = fmul float %36, 0.000000e+00
>  %74 = fadd float %73, 1.000000e+00
>  %75 = fptosi float %66 to i32
>  %76 = bitcast i32 %75 to float
>  %77 = bitcast float %76 to i32
>  %78 = shl i32 %77, 4
>  %79 = add i32 %78, 128
>  %80 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %79)
>  %81 = shl i32 %77, 4
>  %82 = add i32 %81, 132
>  %83 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %82)
>  %84 = shl i32 %77, 4
>  %85 = add i32 %84, 136
>  %86 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %85)
>  %87 = shl i32 %77, 4
>  %88 = add i32 %87, 140
>  %89 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %88)
>  %90 = fmul float %68, %80
>  %91 = fmul float %70, %83
>  %92 = fadd float %90, %91
>  %93 = fmul float %72, %86
>  %94 = fadd float %92, %93
>  %95 = fmul float %74, %89
>  %96 = fadd float %94, %95
>  %97 = fptosi float %66 to i32
>  %98 = bitcast i32 %97 to float
>  %99 = bitcast float %98 to i32
>  %100 = add i32 1, %99
>  %101 = bitcast i32 %100 to float
>  %102 = bitcast float %101 to i32
>  %103 = shl i32 %102, 4
>  %104 = add i32 %103, 128
>  %105 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %104)
>  %106 = shl i32 %102, 4
>  %107 = add i32 %106, 132
>  %108 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %107)
>  %109 = shl i32 %102, 4
>  %110 = add i32 %109, 136
>  %111 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %110)
>  %112 = shl i32 %102, 4
>  %113 = add i32 %112, 140
>  %114 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %113)
>  %115 = fmul float %68, %105
>  %116 = fmul float %70, %108
>  %117 = fadd float %115, %116
>  %118 = fmul float %72, %111
>  %119 = fadd float %117, %118
>  %120 = fmul float %74, %114
>  %121 = fadd float %119, %120
>  %122 = fptosi float %66 to i32
>  %123 = bitcast i32 %122 to float
>  %124 = bitcast float %123 to i32
>  %125 = add i32 2, %124
>  %126 = bitcast i32 %125 to float
>  %127 = bitcast float %126 to i32
>  %128 = shl i32 %127, 4
>  %129 = add i32 %128, 128
>  %130 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %129)
>  %131 = shl i32 %127, 4
>  %132 = add i32 %131, 132
>  %133 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %132)
>  %134 = shl i32 %127, 4
>  %135 = add i32 %134, 136
>  %136 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %135)
>  %137 = shl i32 %127, 4
>  %138 = add i32 %137, 140
>  %139 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %138)
>  %140 = fmul float %68, %130
>  %141 = fmul float %70, %133
>  %142 = fadd float %140, %141
>  %143 = fmul float %72, %136
>  %144 = fadd float %142, %143
>  %145 = fmul float %74, %139
>  %146 = fadd float %144, %145
>  %147 = fmul float %96, %43
>  %148 = fmul float %121, %43
>  %149 = fmul float %146, %43
>  %150 = fptosi float %66 to i32
>  %151 = bitcast i32 %150 to float
>  %152 = bitcast float %151 to i32
>  %153 = shl i32 %152, 4
>  %154 = add i32 %153, 128
>  %155 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %154)
>  %156 = shl i32 %152, 4
>  %157 = add i32 %156, 132
>  %158 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %157)
>  %159 = shl i32 %152, 4
>  %160 = add i32 %159, 136
>  %161 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %160)
>  %162 = fmul float %57, %155
>  %163 = fmul float %58, %158
>  %164 = fadd float %163, %162
>  %165 = fmul float %59, %161
>  %166 = fadd float %164, %165
>  %167 = fptosi float %66 to i32
>  %168 = bitcast i32 %167 to float
>  %169 = bitcast float %168 to i32
>  %170 = add i32 1, %169
>  %171 = bitcast i32 %170 to float
>  %172 = bitcast float %171 to i32
>  %173 = shl i32 %172, 4
>  %174 = add i32 %173, 128
>  %175 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %174)
>  %176 = shl i32 %172, 4
>  %177 = add i32 %176, 132
>  %178 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %177)
>  %179 = shl i32 %172, 4
>  %180 = add i32 %179, 136
>  %181 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %180)
>  %182 = fmul float %57, %175
>  %183 = fmul float %58, %178
>  %184 = fadd float %183, %182
>  %185 = fmul float %59, %181
>  %186 = fadd float %184, %185
>  %187 = fptosi float %66 to i32
>  %188 = bitcast i32 %187 to float
>  %189 = bitcast float %188 to i32
>  %190 = add i32 2, %189
>  %191 = bitcast i32 %190 to float
>  %192 = bitcast float %191 to i32
>  %193 = shl i32 %192, 4
>  %194 = add i32 %193, 128
>  %195 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %194)
>  %196 = shl i32 %192, 4
>  %197 = add i32 %196, 132
>  %198 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %197)
>  %199 = shl i32 %192, 4
>  %200 = add i32 %199, 136
>  %201 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %200)
>  %202 = fmul float %57, %195
>  %203 = fmul float %58, %198
>  %204 = fadd float %203, %202
>  %205 = fmul float %59, %201
>  %206 = fadd float %204, %205
>  %207 = fmul float %166, %43
>  %208 = fmul float %186, %43
>  %209 = fmul float %206, %43
>  %210 = fcmp olt float 0.000000e+00, %44
>  %211 = sext i1 %210 to i32
>  %212 = bitcast i32 %211 to float
>  %213 = bitcast float %212 to i32
>  %214 = icmp ne i32 %213, 0
>  br i1 %214, label %IF, label %ENDIF
>
>IF:                                               ; preds = %main_body
>  %215 = fmul float 3.000000e+00, %51
>  %216 = fptosi float %215 to i32
>  %217 = bitcast i32 %216 to float
>  %218 = bitcast float %217 to i32
>  %219 = shl i32 %218, 4
>  %220 = add i32 %219, 128
>  %221 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %220)
>  %222 = shl i32 %218, 4
>  %223 = add i32 %222, 132
>  %224 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %223)
>  %225 = shl i32 %218, 4
>  %226 = add i32 %225, 136
>  %227 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %226)
>  %228 = shl i32 %218, 4
>  %229 = add i32 %228, 140
>  %230 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %229)
>  %231 = fmul float %68, %221
>  %232 = fmul float %70, %224
>  %233 = fadd float %231, %232
>  %234 = fmul float %72, %227
>  %235 = fadd float %233, %234
>  %236 = fmul float %74, %230
>  %237 = fadd float %235, %236
>  %238 = fptosi float %215 to i32
>  %239 = bitcast i32 %238 to float
>  %240 = bitcast float %239 to i32
>  %241 = add i32 1, %240
>  %242 = bitcast i32 %241 to float
>  %243 = bitcast float %242 to i32
>  %244 = shl i32 %243, 4
>  %245 = add i32 %244, 128
>  %246 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %245)
>  %247 = shl i32 %243, 4
>  %248 = add i32 %247, 132
>  %249 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %248)
>  %250 = shl i32 %243, 4
>  %251 = add i32 %250, 136
>  %252 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %251)
>  %253 = shl i32 %243, 4
>  %254 = add i32 %253, 140
>  %255 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %254)
>  %256 = fmul float %68, %246
>  %257 = fmul float %70, %249
>  %258 = fadd float %256, %257
>  %259 = fmul float %72, %252
>  %260 = fadd float %258, %259
>  %261 = fmul float %74, %255
>  %262 = fadd float %260, %261
>  %263 = fptosi float %215 to i32
>  %264 = bitcast i32 %263 to float
>  %265 = bitcast float %264 to i32
>  %266 = add i32 2, %265
>  %267 = bitcast i32 %266 to float
>  %268 = bitcast float %267 to i32
>  %269 = shl i32 %268, 4
>  %270 = add i32 %269, 128
>  %271 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %270)
>  %272 = shl i32 %268, 4
>  %273 = add i32 %272, 132
>  %274 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %273)
>  %275 = shl i32 %268, 4
>  %276 = add i32 %275, 136
>  %277 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %276)
>  %278 = shl i32 %268, 4
>  %279 = add i32 %278, 140
>  %280 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %279)
>  %281 = fmul float %68, %271
>  %282 = fmul float %70, %274
>  %283 = fadd float %281, %282
>  %284 = fmul float %72, %277
>  %285 = fadd float %283, %284
>  %286 = fmul float %74, %280
>  %287 = fadd float %285, %286
>  %288 = fmul float %44, %237
>  %289 = fadd float %288, %147
>  %290 = fmul float %44, %262
>  %291 = fadd float %290, %148
>  %292 = fmul float %44, %287
>  %293 = fadd float %292, %149
>  %294 = fptosi float %215 to i32
>  %295 = bitcast i32 %294 to float
>  %296 = bitcast float %295 to i32
>  %297 = shl i32 %296, 4
>  %298 = add i32 %297, 128
>  %299 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %298)
>  %300 = shl i32 %296, 4
>  %301 = add i32 %300, 132
>  %302 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %301)
>  %303 = shl i32 %296, 4
>  %304 = add i32 %303, 136
>  %305 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %304)
>  %306 = fmul float %57, %299
>  %307 = fmul float %58, %302
>  %308 = fadd float %307, %306
>  %309 = fmul float %59, %305
>  %310 = fadd float %308, %309
>  %311 = fptosi float %215 to i32
>  %312 = bitcast i32 %311 to float
>  %313 = bitcast float %312 to i32
>  %314 = add i32 1, %313
>  %315 = bitcast i32 %314 to float
>  %316 = bitcast float %315 to i32
>  %317 = shl i32 %316, 4
>  %318 = add i32 %317, 128
>  %319 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %318)
>  %320 = shl i32 %316, 4
>  %321 = add i32 %320, 132
>  %322 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %321)
>  %323 = shl i32 %316, 4
>  %324 = add i32 %323, 136
>  %325 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %324)
>  %326 = fmul float %57, %319
>  %327 = fmul float %58, %322
>  %328 = fadd float %327, %326
>  %329 = fmul float %59, %325
>  %330 = fadd float %328, %329
>  %331 = fptosi float %215 to i32
>  %332 = bitcast i32 %331 to float
>  %333 = bitcast float %332 to i32
>  %334 = add i32 2, %333
>  %335 = bitcast i32 %334 to float
>  %336 = bitcast float %335 to i32
>  %337 = shl i32 %336, 4
>  %338 = add i32 %337, 128
>  %339 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %338)
>  %340 = shl i32 %336, 4
>  %341 = add i32 %340, 132
>  %342 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %341)
>  %343 = shl i32 %336, 4
>  %344 = add i32 %343, 136
>  %345 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %344)
>  %346 = fmul float %57, %339
>  %347 = fmul float %58, %342
>  %348 = fadd float %347, %346
>  %349 = fmul float %59, %345
>  %350 = fadd float %348, %349
>  %351 = fmul float %44, %310
>  %352 = fadd float %351, %207
>  %353 = fmul float %44, %330
>  %354 = fadd float %353, %208
>  %355 = fmul float %44, %350
>  %356 = fadd float %355, %209
>  %357 = fcmp olt float 0.000000e+00, %45
>  %358 = sext i1 %357 to i32
>  %359 = bitcast i32 %358 to float
>  %360 = bitcast float %359 to i32
>  %361 = icmp ne i32 %360, 0
>  br i1 %361, label %IF66, label %ENDIF
>
>ENDIF:                                            ; preds = %IF66, %IF, %main_body
>  %temp8.0 = phi float [ %147, %main_body ], [ %466, %IF66 ], [ %289, %IF ]
>  %temp9.0 = phi float [ %148, %main_body ], [ %468, %IF66 ], [ %291, %IF ]
>  %temp10.0 = phi float [ %149, %main_body ], [ %470, %IF66 ], [ %293, %IF ]
>  %temp12.0 = phi float [ %207, %main_body ], [ %529, %IF66 ], [ %352, %IF ]
>  %temp13.0 = phi float [ %208, %main_body ], [ %531, %IF66 ], [ %354, %IF ]
>  %temp14.0 = phi float [ %209, %main_body ], [ %533, %IF66 ], [ %356, %IF ]
>  %362 = fmul float %temp9.0, %20
>  %363 = fmul float %temp9.0, %21
>  %364 = fmul float %temp9.0, %22
>  %365 = fmul float %temp9.0, %23
>  %366 = fmul float %temp8.0, %16
>  %367 = fadd float %366, %362
>  %368 = fmul float %temp8.0, %17
>  %369 = fadd float %368, %363
>  %370 = fmul float %temp8.0, %18
>  %371 = fadd float %370, %364
>  %372 = fmul float %temp8.0, %19
>  %373 = fadd float %372, %365
>  %374 = fmul float %temp10.0, %24
>  %375 = fadd float %374, %367
>  %376 = fmul float %temp10.0, %25
>  %377 = fadd float %376, %369
>  %378 = fmul float %temp10.0, %26
>  %379 = fadd float %378, %371
>  %380 = fmul float %temp10.0, %27
>  %381 = fadd float %380, %373
>  %382 = fadd float %375, %28
>  %383 = fadd float %377, %29
>  %384 = fadd float %379, %30
>  %385 = fadd float %381, %31
>  %386 = fsub float -0.000000e+00, %13
>  %387 = fadd float %temp8.0, %386
>  %388 = fsub float -0.000000e+00, %14
>  %389 = fadd float %temp9.0, %388
>  %390 = fsub float -0.000000e+00, %15
>  %391 = fadd float %temp10.0, %390
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %temp12.0, float %temp13.0, float %temp14.0, float %64)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %387, float %389, float %391, float %65)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 34, i32 0, float %temp8.0, float %temp9.0, float %temp10.0, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 35, i32 0, float %382, float %383, float %385, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %382, float %383, float %384, float %385)
>  ret void
>
>IF66:                                             ; preds = %IF
>  %392 = fmul float 3.000000e+00, %52
>  %393 = fptosi float %392 to i32
>  %394 = bitcast i32 %393 to float
>  %395 = bitcast float %394 to i32
>  %396 = shl i32 %395, 4
>  %397 = add i32 %396, 128
>  %398 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %397)
>  %399 = shl i32 %395, 4
>  %400 = add i32 %399, 132
>  %401 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %400)
>  %402 = shl i32 %395, 4
>  %403 = add i32 %402, 136
>  %404 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %403)
>  %405 = shl i32 %395, 4
>  %406 = add i32 %405, 140
>  %407 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %406)
>  %408 = fmul float %68, %398
>  %409 = fmul float %70, %401
>  %410 = fadd float %408, %409
>  %411 = fmul float %72, %404
>  %412 = fadd float %410, %411
>  %413 = fmul float %74, %407
>  %414 = fadd float %412, %413
>  %415 = fptosi float %392 to i32
>  %416 = bitcast i32 %415 to float
>  %417 = bitcast float %416 to i32
>  %418 = add i32 1, %417
>  %419 = bitcast i32 %418 to float
>  %420 = bitcast float %419 to i32
>  %421 = shl i32 %420, 4
>  %422 = add i32 %421, 128
>  %423 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %422)
>  %424 = shl i32 %420, 4
>  %425 = add i32 %424, 132
>  %426 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %425)
>  %427 = shl i32 %420, 4
>  %428 = add i32 %427, 136
>  %429 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %428)
>  %430 = shl i32 %420, 4
>  %431 = add i32 %430, 140
>  %432 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %431)
>  %433 = fmul float %68, %423
>  %434 = fmul float %70, %426
>  %435 = fadd float %433, %434
>  %436 = fmul float %72, %429
>  %437 = fadd float %435, %436
>  %438 = fmul float %74, %432
>  %439 = fadd float %437, %438
>  %440 = fptosi float %392 to i32
>  %441 = bitcast i32 %440 to float
>  %442 = bitcast float %441 to i32
>  %443 = add i32 2, %442
>  %444 = bitcast i32 %443 to float
>  %445 = bitcast float %444 to i32
>  %446 = shl i32 %445, 4
>  %447 = add i32 %446, 128
>  %448 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %447)
>  %449 = shl i32 %445, 4
>  %450 = add i32 %449, 132
>  %451 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %450)
>  %452 = shl i32 %445, 4
>  %453 = add i32 %452, 136
>  %454 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %453)
>  %455 = shl i32 %445, 4
>  %456 = add i32 %455, 140
>  %457 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %456)
>  %458 = fmul float %68, %448
>  %459 = fmul float %70, %451
>  %460 = fadd float %458, %459
>  %461 = fmul float %72, %454
>  %462 = fadd float %460, %461
>  %463 = fmul float %74, %457
>  %464 = fadd float %462, %463
>  %465 = fmul float %45, %414
>  %466 = fadd float %465, %289
>  %467 = fmul float %45, %439
>  %468 = fadd float %467, %291
>  %469 = fmul float %45, %464
>  %470 = fadd float %469, %293
>  %471 = fptosi float %392 to i32
>  %472 = bitcast i32 %471 to float
>  %473 = bitcast float %472 to i32
>  %474 = shl i32 %473, 4
>  %475 = add i32 %474, 128
>  %476 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %475)
>  %477 = shl i32 %473, 4
>  %478 = add i32 %477, 132
>  %479 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %478)
>  %480 = shl i32 %473, 4
>  %481 = add i32 %480, 136
>  %482 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %481)
>  %483 = fmul float %57, %476
>  %484 = fmul float %58, %479
>  %485 = fadd float %484, %483
>  %486 = fmul float %59, %482
>  %487 = fadd float %485, %486
>  %488 = fptosi float %392 to i32
>  %489 = bitcast i32 %488 to float
>  %490 = bitcast float %489 to i32
>  %491 = add i32 1, %490
>  %492 = bitcast i32 %491 to float
>  %493 = bitcast float %492 to i32
>  %494 = shl i32 %493, 4
>  %495 = add i32 %494, 128
>  %496 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %495)
>  %497 = shl i32 %493, 4
>  %498 = add i32 %497, 132
>  %499 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %498)
>  %500 = shl i32 %493, 4
>  %501 = add i32 %500, 136
>  %502 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %501)
>  %503 = fmul float %57, %496
>  %504 = fmul float %58, %499
>  %505 = fadd float %504, %503
>  %506 = fmul float %59, %502
>  %507 = fadd float %505, %506
>  %508 = fptosi float %392 to i32
>  %509 = bitcast i32 %508 to float
>  %510 = bitcast float %509 to i32
>  %511 = add i32 2, %510
>  %512 = bitcast i32 %511 to float
>  %513 = bitcast float %512 to i32
>  %514 = shl i32 %513, 4
>  %515 = add i32 %514, 128
>  %516 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %515)
>  %517 = shl i32 %513, 4
>  %518 = add i32 %517, 132
>  %519 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %518)
>  %520 = shl i32 %513, 4
>  %521 = add i32 %520, 136
>  %522 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %521)
>  %523 = fmul float %57, %516
>  %524 = fmul float %58, %519
>  %525 = fadd float %524, %523
>  %526 = fmul float %59, %522
>  %527 = fadd float %525, %526
>  %528 = fmul float %45, %487
>  %529 = fadd float %528, %352
>  %530 = fmul float %45, %507
>  %531 = fadd float %530, %354
>  %532 = fmul float %45, %527
>  %533 = fadd float %532, %356
>  br label %ENDIF
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v14, s10, v0                               ; 4A1C000A
>	s_load_dwordx4 s[12:15], s[8:9], 0x0                     ; C0860900
>	s_load_dwordx4 s[16:19], s[8:9], 0x4                     ; C0880904
>	s_load_dwordx4 s[20:23], s[8:9], 0x8                     ; C08A0908
>	s_load_dwordx4 s[24:27], s[8:9], 0xc                     ; C08C090C
>	s_load_dwordx4 s[4:7], s[8:9], 0x10                      ; C0820910
>	s_waitcnt lgkmcnt(0)                                     ; BF8C007F
>	buffer_load_format_xyzw v[20:23], v14, s[20:23], 0 idxen ; E00C2000 8005140E
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v0, 0x40400000, v20                        ; 100028FF 40400000
>	v_cvt_i32_f32_e32 v0, v0                                 ; 7E001100
>	v_lshlrev_b32_e32 v5, 4, v0                              ; 340A0084
>	v_add_i32_e32 v0, 0xa0, v5                               ; 4A000AFF 000000A0
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                       ; C0800300
>	s_waitcnt lgkmcnt(0)                                     ; BF8C007F
>	buffer_load_dword v2, v0, s[0:3], 0 offen                ; E0301000 80000200
>	buffer_load_format_xyzw v[10:13], v14, s[24:27], 0 idxen ; E00C2000 80060A0E
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v0, v2, v10                                ; 10001502
>	v_add_i32_e32 v1, 0xa4, v5                               ; 4A020AFF 000000A4
>	buffer_load_dword v4, v1, s[0:3], 0 offen                ; E0301000 80000401
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v0, v11, v4, v0                                ; D2820000 0402090B
>	v_add_i32_e32 v1, 0xa8, v5                               ; 4A020AFF 000000A8
>	buffer_load_dword v15, v1, s[0:3], 0 offen               ; E0301000 80000F01
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v0, v12, v15, v0                               ; D2820000 04021F0C
>	buffer_load_format_xyzw v[6:9], v14, s[16:19], 0 idxen   ; E00C2000 8004060E
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v0, v6, v0                                 ; 10000106
>	v_add_i32_e32 v1, 0x90, v5                               ; 4A020AFF 00000090
>	buffer_load_dword v16, v1, s[0:3], 0 offen               ; E0301000 80001001
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v1, v16, v10                               ; 10021510
>	v_add_i32_e32 v3, 0x94, v5                               ; 4A060AFF 00000094
>	buffer_load_dword v17, v3, s[0:3], 0 offen               ; E0301000 80001103
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v1, v11, v17, v1                               ; D2820001 0406230B
>	v_add_i32_e32 v3, 0x98, v5                               ; 4A060AFF 00000098
>	buffer_load_dword v26, v3, s[0:3], 0 offen               ; E0301000 80001A03
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v1, v12, v26, v1                               ; D2820001 0406350C
>	v_mul_f32_e32 v1, v6, v1                                 ; 10020306
>	v_add_i32_e32 v3, 0x80, v5                               ; 4A060AFF 00000080
>	buffer_load_dword v27, v3, s[0:3], 0 offen               ; E0301000 80001B03
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v3, v27, v10                               ; 1006151B
>	v_add_i32_e32 v18, 0x84, v5                              ; 4A240AFF 00000084
>	buffer_load_dword v28, v18, s[0:3], 0 offen              ; E0301000 80001C12
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v3, v11, v28, v3                               ; D2820003 040E390B
>	v_add_i32_e32 v18, 0x88, v5                              ; 4A240AFF 00000088
>	buffer_load_dword v29, v18, s[0:3], 0 offen              ; E0301000 80001D12
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v3, v12, v29, v3                               ; D2820003 040E3B0C
>	v_mul_f32_e32 v3, v6, v3                                 ; 10060706
>	buffer_load_format_xyzw v[30:33], v14, s[12:15], 0 idxen ; E00C2000 80031E0E
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_add_f32_e32 v18, 0, v31                                ; 06243E80
>	v_mul_f32_e32 v4, v4, v18                                ; 10082504
>	v_add_f32_e32 v19, 0, v30                                ; 06263C80
>	v_mad_f32 v2, v19, v2, v4                                ; D2820002 04120513
>	v_add_f32_e32 v25, 0, v32                                ; 06324080
>	v_mad_f32 v2, v25, v15, v2                               ; D2820002 040A1F19
>	v_mad_f32 v24, 0, v30, 1.0                               ; D2820018 03CA3C80
>	v_add_i32_e32 v4, 0xac, v5                               ; 4A080AFF 000000AC
>	buffer_load_dword v4, v4, s[0:3], 0 offen                ; E0301000 80000404
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v2, v24, v4, v2                                ; D2820002 040A0918
>	v_mul_f32_e32 v2, v6, v2                                 ; 10040506
>	v_mul_f32_e32 v4, v17, v18                               ; 10082511
>	v_mad_f32 v4, v19, v16, v4                               ; D2820004 04122113
>	v_mad_f32 v4, v25, v26, v4                               ; D2820004 04123519
>	v_add_i32_e32 v15, 0x9c, v5                              ; 4A1E0AFF 0000009C
>	buffer_load_dword v15, v15, s[0:3], 0 offen              ; E0301000 80000F0F
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v4, v24, v15, v4                               ; D2820004 04121F18
>	v_mul_f32_e32 v4, v6, v4                                 ; 10080906
>	v_mul_f32_e32 v15, v28, v18                              ; 101E251C
>	v_mad_f32 v15, v19, v27, v15                             ; D282000F 043E3713
>	v_mad_f32 v15, v25, v29, v15                             ; D282000F 043E3B19
>	v_add_i32_e32 v5, 0x8c, v5                               ; 4A0A0AFF 0000008C
>	buffer_load_dword v5, v5, s[0:3], 0 offen                ; E0301000 80000505
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v5, v24, v5, v15                               ; D2820005 043E0B18
>	v_mul_f32_e32 v5, v6, v5                                 ; 100A0B06
>	buffer_load_format_xyzw v[14:17], v14, s[4:7], 0 idxen   ; E00C2000 80010E0E
>	v_cmp_gt_f32_e64 s[4:5], v7, 0                           ; D0080004 00010107
>	v_cndmask_b32_e64 v26, 0, -1, s[4:5]                     ; D200001A 00118280
>	v_cmp_ne_i32_e64 s[24:25], v26, 0                        ; D10A0018 0001011A
>	s_buffer_load_dword s9, s[0:3], 0x1f                     ; C204811F
>	s_buffer_load_dword s4, s[0:3], 0x1e                     ; C202011E
>	s_buffer_load_dword s7, s[0:3], 0x1d                     ; C203811D
>	s_buffer_load_dword s6, s[0:3], 0x1c                     ; C203011C
>	s_buffer_load_dword s13, s[0:3], 0x1b                    ; C206811B
>	s_buffer_load_dword s5, s[0:3], 0x1a                     ; C202811A
>	s_buffer_load_dword s12, s[0:3], 0x19                    ; C2060119
>	s_buffer_load_dword s10, s[0:3], 0x18                    ; C2050118
>	s_buffer_load_dword s17, s[0:3], 0x17                    ; C2088117
>	s_buffer_load_dword s8, s[0:3], 0x16                     ; C2040116
>	s_buffer_load_dword s15, s[0:3], 0x15                    ; C2078115
>	s_buffer_load_dword s14, s[0:3], 0x14                    ; C2070114
>	s_buffer_load_dword s19, s[0:3], 0x13                    ; C2098113
>	s_buffer_load_dword s11, s[0:3], 0x12                    ; C2058112
>	s_buffer_load_dword s18, s[0:3], 0x11                    ; C2090111
>	s_buffer_load_dword s16, s[0:3], 0x10                    ; C2080110
>	s_buffer_load_dword s20, s[0:3], 0xe                     ; C20A010E
>	s_buffer_load_dword s21, s[0:3], 0xd                     ; C20A810D
>	s_buffer_load_dword s22, s[0:3], 0xc                     ; C20B010C
>	s_waitcnt vmcnt(0) lgkmcnt(0)                            ; BF8C0070
>	s_and_saveexec_b64 s[24:25], s[24:25]                    ; BE982418
>	s_xor_b64 s[24:25], exec, s[24:25]                       ; 8998187E
>	s_cbranch_execz BB0_3                                    ; BF880000
>	v_mul_f32_e32 v26, 0x40400000, v21                       ; 10342AFF 40400000
>	v_cvt_i32_f32_e32 v26, v26                               ; 7E34111A
>	v_lshlrev_b32_e32 v26, 4, v26                            ; 34343484
>	v_add_i32_e32 v27, 0xa4, v26                             ; 4A3634FF 000000A4
>	buffer_load_dword v27, v27, s[0:3], 0 offen              ; E0301000 80001B1B
>	v_add_i32_e32 v28, 0xa0, v26                             ; 4A3834FF 000000A0
>	buffer_load_dword v28, v28, s[0:3], 0 offen              ; E0301000 80001C1C
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v29, v28, v10                              ; 103A151C
>	v_mad_f32 v29, v11, v27, v29                             ; D282001D 0476370B
>	v_add_i32_e32 v30, 0xa8, v26                             ; 4A3C34FF 000000A8
>	buffer_load_dword v30, v30, s[0:3], 0 offen              ; E0301000 80001E1E
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v29, v12, v30, v29                             ; D282001D 04763D0C
>	v_mad_f32 v0, v7, v29, v0                                ; D2820000 04023B07
>	v_add_i32_e32 v29, 0x94, v26                             ; 4A3A34FF 00000094
>	buffer_load_dword v29, v29, s[0:3], 0 offen              ; E0301000 80001D1D
>	v_add_i32_e32 v31, 0x90, v26                             ; 4A3E34FF 00000090
>	buffer_load_dword v31, v31, s[0:3], 0 offen              ; E0301000 80001F1F
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v32, v31, v10                              ; 1040151F
>	v_mad_f32 v32, v11, v29, v32                             ; D2820020 04823B0B
>	v_add_i32_e32 v33, 0x98, v26                             ; 4A4234FF 00000098
>	buffer_load_dword v33, v33, s[0:3], 0 offen              ; E0301000 80002121
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v32, v12, v33, v32                             ; D2820020 0482430C
>	v_mad_f32 v1, v7, v32, v1                                ; D2820001 04064107
>	v_add_i32_e32 v32, 0x84, v26                             ; 4A4034FF 00000084
>	buffer_load_dword v32, v32, s[0:3], 0 offen              ; E0301000 80002020
>	v_add_i32_e32 v34, 0x80, v26                             ; 4A4434FF 00000080
>	buffer_load_dword v34, v34, s[0:3], 0 offen              ; E0301000 80002222
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v35, v34, v10                              ; 10461522
>	v_mad_f32 v35, v11, v32, v35                             ; D2820023 048E410B
>	v_add_i32_e32 v36, 0x88, v26                             ; 4A4834FF 00000088
>	buffer_load_dword v36, v36, s[0:3], 0 offen              ; E0301000 80002424
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v35, v12, v36, v35                             ; D2820023 048E490C
>	v_mad_f32 v3, v7, v35, v3                                ; D2820003 040E4707
>	v_mul_f32_e32 v27, v27, v18                              ; 1036251B
>	v_mad_f32 v27, v19, v28, v27                             ; D282001B 046E3913
>	v_mad_f32 v27, v25, v30, v27                             ; D282001B 046E3D19
>	v_add_i32_e32 v28, 0xac, v26                             ; 4A3834FF 000000AC
>	buffer_load_dword v28, v28, s[0:3], 0 offen              ; E0301000 80001C1C
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v27, v24, v28, v27                             ; D282001B 046E3918
>	v_mad_f32 v2, v7, v27, v2                                ; D2820002 040A3707
>	v_mul_f32_e32 v27, v29, v18                              ; 1036251D
>	v_mad_f32 v27, v19, v31, v27                             ; D282001B 046E3F13
>	v_mad_f32 v27, v25, v33, v27                             ; D282001B 046E4319
>	v_add_i32_e32 v28, 0x9c, v26                             ; 4A3834FF 0000009C
>	buffer_load_dword v28, v28, s[0:3], 0 offen              ; E0301000 80001C1C
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v27, v24, v28, v27                             ; D282001B 046E3918
>	v_mad_f32 v4, v7, v27, v4                                ; D2820004 04123707
>	v_mul_f32_e32 v27, v32, v18                              ; 10362520
>	v_mad_f32 v27, v19, v34, v27                             ; D282001B 046E4513
>	v_mad_f32 v27, v25, v36, v27                             ; D282001B 046E4919
>	v_add_i32_e32 v26, 0x8c, v26                             ; 4A3434FF 0000008C
>	buffer_load_dword v26, v26, s[0:3], 0 offen              ; E0301000 80001A1A
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v26, v24, v26, v27                             ; D282001A 046E3518
>	v_mad_f32 v5, v7, v26, v5                                ; D2820005 04163507
>	v_cmp_gt_f32_e64 s[26:27], v8, 0                         ; D008001A 00010108
>	v_cndmask_b32_e64 v26, 0, -1, s[26:27]                   ; D200001A 00698280
>	v_cmp_ne_i32_e64 s[26:27], v26, 0                        ; D10A001A 0001011A
>	s_and_saveexec_b64 s[26:27], s[26:27]                    ; BE9A241A
>	s_xor_b64 s[26:27], exec, s[26:27]                       ; 899A1A7E
>	s_cbranch_execz BB0_2                                    ; BF880000
>	v_mul_f32_e32 v20, 0x40400000, v22                       ; 10282CFF 40400000
>	v_cvt_i32_f32_e32 v20, v20                               ; 7E281114
>	v_lshlrev_b32_e32 v20, 4, v20                            ; 34282884
>	v_add_i32_e32 v21, 0xa4, v20                             ; 4A2A28FF 000000A4
>	buffer_load_dword v21, v21, s[0:3], 0 offen              ; E0301000 80001515
>	v_add_i32_e32 v22, 0xa0, v20                             ; 4A2C28FF 000000A0
>	buffer_load_dword v22, v22, s[0:3], 0 offen              ; E0301000 80001616
>	s_waitcnt vmcnt(0) expcnt(0)                             ; BF8C0700
>	v_mul_f32_e32 v23, v22, v10                              ; 102E1516
>	v_mad_f32 v23, v11, v21, v23                             ; D2820017 045E2B0B
>	v_add_i32_e32 v26, 0xa8, v20                             ; 4A3428FF 000000A8
>	buffer_load_dword v26, v26, s[0:3], 0 offen              ; E0301000 80001A1A
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v23, v12, v26, v23                             ; D2820017 045E350C
>	v_mad_f32 v0, v8, v23, v0                                ; D2820000 04022F08
>	v_add_i32_e32 v23, 0x94, v20                             ; 4A2E28FF 00000094
>	buffer_load_dword v23, v23, s[0:3], 0 offen              ; E0301000 80001717
>	v_add_i32_e32 v27, 0x90, v20                             ; 4A3628FF 00000090
>	buffer_load_dword v27, v27, s[0:3], 0 offen              ; E0301000 80001B1B
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v28, v27, v10                              ; 1038151B
>	v_mad_f32 v28, v11, v23, v28                             ; D282001C 04722F0B
>	v_add_i32_e32 v29, 0x98, v20                             ; 4A3A28FF 00000098
>	buffer_load_dword v29, v29, s[0:3], 0 offen              ; E0301000 80001D1D
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v28, v12, v29, v28                             ; D282001C 04723B0C
>	v_mad_f32 v1, v8, v28, v1                                ; D2820001 04063908
>	v_add_i32_e32 v28, 0x84, v20                             ; 4A3828FF 00000084
>	buffer_load_dword v28, v28, s[0:3], 0 offen              ; E0301000 80001C1C
>	v_add_i32_e32 v30, 0x80, v20                             ; 4A3C28FF 00000080
>	buffer_load_dword v30, v30, s[0:3], 0 offen              ; E0301000 80001E1E
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mul_f32_e32 v31, v30, v10                              ; 103E151E
>	v_mad_f32 v31, v11, v28, v31                             ; D282001F 047E390B
>	v_add_i32_e32 v32, 0x88, v20                             ; 4A4028FF 00000088
>	buffer_load_dword v32, v32, s[0:3], 0 offen              ; E0301000 80002020
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v10, v12, v32, v31                             ; D282000A 047E410C
>	v_mad_f32 v3, v8, v10, v3                                ; D2820003 040E1508
>	v_mul_f32_e32 v10, v21, v18                              ; 10142515
>	v_mad_f32 v10, v19, v22, v10                             ; D282000A 042A2D13
>	v_mad_f32 v10, v25, v26, v10                             ; D282000A 042A3519
>	v_add_i32_e32 v11, 0xac, v20                             ; 4A1628FF 000000AC
>	buffer_load_dword v11, v11, s[0:3], 0 offen              ; E0301000 80000B0B
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v10, v24, v11, v10                             ; D282000A 042A1718
>	v_mad_f32 v2, v8, v10, v2                                ; D2820002 040A1508
>	v_mul_f32_e32 v10, v23, v18                              ; 10142517
>	v_mad_f32 v10, v19, v27, v10                             ; D282000A 042A3713
>	v_mad_f32 v10, v25, v29, v10                             ; D282000A 042A3B19
>	v_add_i32_e32 v11, 0x9c, v20                             ; 4A1628FF 0000009C
>	buffer_load_dword v11, v11, s[0:3], 0 offen              ; E0301000 80000B0B
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v10, v24, v11, v10                             ; D282000A 042A1718
>	v_mad_f32 v4, v8, v10, v4                                ; D2820004 04121508
>	v_mul_f32_e32 v10, v28, v18                              ; 1014251C
>	v_mad_f32 v10, v19, v30, v10                             ; D282000A 042A3D13
>	v_mad_f32 v10, v25, v32, v10                             ; D282000A 042A4119
>	v_add_i32_e32 v11, 0x8c, v20                             ; 4A1628FF 0000008C
>	buffer_load_dword v11, v11, s[0:3], 0 offen              ; E0301000 80000B0B
>	s_waitcnt vmcnt(0)                                       ; BF8C0770
>	v_mad_f32 v10, v24, v11, v10                             ; D282000A 042A1718
>	v_mad_f32 v5, v8, v10, v5                                ; D2820005 04161508
>	s_or_b64 exec, exec, s[26:27]                            ; 88FE1A7E
>	s_or_b64 exec, exec, s[24:25]                            ; 88FE187E
>	exp 15, 32, 0, 0, 0, v3, v1, v0, v14                     ; F800020F 0E000103
>	s_waitcnt expcnt(0)                                      ; BF8C070F
>	v_subrev_f32_e32 v0, s20, v2                             ; 0A000414
>	v_subrev_f32_e32 v1, s21, v4                             ; 0A020815
>	v_subrev_f32_e32 v3, s22, v5                             ; 0A060A16
>	exp 15, 33, 0, 0, 0, v3, v1, v0, v15                     ; F800021F 0F000103
>	s_waitcnt expcnt(0)                                      ; BF8C070F
>	v_mov_b32_e32 v0, 1.0                                    ; 7E0002F2
>	exp 15, 34, 0, 0, 0, v5, v4, v2, v0                      ; F800022F 00020405
>	v_mul_f32_e32 v1, s17, v4                                ; 10020811
>	v_mad_f32 v1, v5, s19, v1                                ; D2820001 04042705
>	v_mad_f32 v1, v2, s13, v1                                ; D2820001 04041B02
>	v_add_f32_e32 v1, s9, v1                                 ; 06020209
>	v_mul_f32_e32 v3, s15, v4                                ; 1006080F
>	v_mad_f32 v3, v5, s18, v3                                ; D2820003 040C2505
>	v_mad_f32 v3, v2, s12, v3                                ; D2820003 040C1902
>	v_add_f32_e32 v3, s7, v3                                 ; 06060607
>	v_mul_f32_e32 v6, s14, v4                                ; 100C080E
>	v_mad_f32 v6, v5, s16, v6                                ; D2820006 04182105
>	v_mad_f32 v6, v2, s10, v6                                ; D2820006 04181502
>	v_add_f32_e32 v6, s6, v6                                 ; 060C0C06
>	exp 15, 35, 0, 0, 0, v6, v3, v1, v0                      ; F800023F 00010306
>	s_waitcnt expcnt(0)                                      ; BF8C070F
>	v_mul_f32_e32 v0, s8, v4                                 ; 10000808
>	v_mad_f32 v0, v5, s11, v0                                ; D2820000 04001705
>	v_mad_f32 v0, v2, s5, v0                                 ; D2820000 04000B02
>	v_add_f32_e32 v0, s4, v0                                 ; 06000004
>	exp 15, 12, 0, 1, 0, v6, v3, v0, v1                      ; F80008CF 01000306
>	s_endpgm                                                 ; BF810000
>
>SGPRS: 32
>VGPRS: 40
>Code Size: 1732 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL IN[1], GENERIC[10], PERSPECTIVE
>DCL IN[2], GENERIC[11], PERSPECTIVE
>DCL IN[3], GENERIC[12], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SAMP[1]
>DCL SAMP[2]
>DCL SAMP[3]
>DCL SAMP[4]
>DCL SAMP[5]
>DCL CONST[0..5]
>DCL TEMP[0..8], LOCAL
>IMM[0] FLT32 {    1.4427,     1.0000,     0.5000,     0.0000}
>IMM[1] FLT32 {    0.0000,     0.3300,     0.0000,     0.0000}
>  0: ADD TEMP[0].xyz, -CONST[3], IN[2]
>  1: MUL TEMP[1].w, TEMP[0].zzzz, CONST[1].xxxx
>  2: MUL TEMP[1].w, TEMP[1].wwww, IMM[0].xxxx
>  3: EX2 TEMP[1].x, TEMP[1].wwww
>  4: ADD TEMP[1].w, -TEMP[1].xxxx, IMM[0].yyyy
>  5: DP3 TEMP[2].x, TEMP[0].xyzz, TEMP[0].xyzz
>  6: RCP TEMP[3].x, TEMP[0].zzzz
>  7: MUL TEMP[2].x, TEMP[2].xxxx, CONST[1].yyyy
>  8: MUL TEMP[1].x, TEMP[1].wwww, TEMP[2].xxxx
>  9: MUL TEMP[1].x, TEMP[3].xxxx, TEMP[1].xxxx
> 10: MUL TEMP[1].x, TEMP[1].xxxx, IMM[0].xxxx
> 11: MOV TEMP[0].x, TEMP[1].xxxx
> 12: EX2 TEMP[1].x, TEMP[1].xxxx
> 13: MOV_SAT TEMP[0].x, TEMP[1].xxxx
> 14: ADD TEMP[1].x, -TEMP[0].xxxx, IMM[0].yyyy
> 15: RCP TEMP[2].x, IN[3].zzzz
> 16: MUL TEMP[2].y, TEMP[2].xxxx, IN[3].yyyy
> 17: MAD TEMP[2].x, TEMP[2].yyyy, IMM[0].zzzz, IMM[0].zzzz
> 18: MOV TEMP[2].x, TEMP[2].xxxx
> 19: MOV TEMP[2].y, CONST[1].wwww
> 20: MOV TEMP[3].xy, TEMP[2].xyyy
> 21: TEX TEMP[3].x, TEMP[3], SAMP[5], 2D
> 22: MUL TEMP[1].x, TEMP[1].xxxx, TEMP[3].xxxx
> 23: MOV TEMP[0].x, TEMP[1].xxxx
> 24: MOV TEMP[3].w, IMM[0].wwww
> 25: MOV TEMP[3].x, IN[1].xxxx
> 26: MOV TEMP[3].y, IN[1].yyyy
> 27: MOV TEMP[3].z, IN[1].zzzz
> 28: DP4 TEMP[4].x, TEMP[3], TEMP[3]
> 29: RSQ TEMP[4].x, TEMP[4].xxxx
> 30: MUL TEMP[3].xyz, TEMP[3], TEMP[4].xxxx
> 31: MOV TEMP[4].w, IMM[0].wwww
> 32: MOV TEMP[4].x, IN[0].xxxx
> 33: MOV TEMP[4].y, IN[0].yyyy
> 34: MOV TEMP[4].z, IN[0].zzzz
> 35: DP4 TEMP[5].x, TEMP[4], TEMP[4]
> 36: RSQ TEMP[5].x, TEMP[5].xxxx
> 37: MUL TEMP[4].xyz, TEMP[4], TEMP[5].xxxx
> 38: DP3 TEMP[5].x, TEMP[3].xyzz, TEMP[4].xyzz
> 39: ADD TEMP[5].y, TEMP[5].xxxx, TEMP[5].xxxx
> 40: MAD TEMP[3].yzw, TEMP[4].xxyz, -TEMP[5].yyyy, TEMP[3].xxyz
> 41: MOV TEMP[4].xyz, TEMP[4].xyzz
> 42: TEX TEMP[4], TEMP[4], SAMP[2], CUBE
> 43: MUL TEMP[4].xyz, TEMP[4], TEMP[4]
> 44: MOV TEMP[5].xyz, TEMP[3].yzww
> 45: TEX TEMP[5], TEMP[5], SAMP[4], CUBE
> 46: MOV TEMP[6].w, TEMP[5].wwww
> 47: MOV TEMP[3].xyz, TEMP[3].yzww
> 48: TEX TEMP[3].xyz, TEMP[3], SAMP[3], CUBE
> 49: MUL TEMP[7].yzw, TEMP[3].xxyz, TEMP[3].xxyz
> 50: MOV TEMP[0].yzw, TEMP[7].zyzw
> 51: MAD TEMP[5].xyz, TEMP[5], TEMP[5], -TEMP[7].yzww
> 52: MOV TEMP[3].x, IN[0].wwww
> 53: MOV TEMP[3].y, IN[1].wwww
> 54: MOV TEMP[7].xy, TEMP[3].xyyy
> 55: TEX TEMP[7], TEMP[7], SAMP[1], 2D
> 56: MOV TEMP[8].xy, TEMP[3].xyyy
> 57: TEX TEMP[8], TEMP[8], SAMP[0], 2D
> 58: MUL TEMP[3], TEMP[8], TEMP[8]
> 59: MUL TEMP[7], TEMP[7], TEMP[7]
> 60: MUL TEMP[8].w, TEMP[7].wwww, TEMP[7].wwww
> 61: MUL TEMP[8], TEMP[8].wwww, TEMP[8].wwww
> 62: MOV_SAT TEMP[8], TEMP[8]
> 63: MOV TEMP[2].w, TEMP[8].wwww
> 64: MAD TEMP[5].yzw, TEMP[8].wwww, TEMP[5].xxyz, TEMP[0]
> 65: MOV TEMP[0].yzw, TEMP[5].zyzw
> 66: MUL TEMP[5].yzw, TEMP[0], TEMP[7].xxyz
> 67: MOV TEMP[0].yzw, TEMP[5].zyzw
> 68: MUL TEMP[5].xyz, CONST[4], CONST[4]
> 69: MOV TEMP[6].xyz, TEMP[5].xyzx
> 70: MOV TEMP[7].xyz, CONST[4].xyzx
> 71: MAD TEMP[5].xyz, TEMP[7], -TEMP[7], IMM[0].yyyy
> 72: MOV TEMP[7].xyz, TEMP[5].xyzx
> 73: MAD TEMP[5].xyz, TEMP[3].wwww, TEMP[7], TEMP[6]
> 74: MOV TEMP[6].xyz, TEMP[5].xyzx
> 75: MUL TEMP[3].xyz, TEMP[3], TEMP[6]
> 76: MAD TEMP[0].yzw, TEMP[3].xxyz, TEMP[4].xxyz, TEMP[0]
> 77: MAX TEMP[3].x, TEMP[0].yyyy, IMM[1].xxxx
> 78: RSQ TEMP[3].x, TEMP[3].xxxx
> 79: RCP TEMP[2].x, TEMP[3].xxxx
> 80: MAX TEMP[3].x, TEMP[0].zzzz, IMM[1].xxxx
> 81: RSQ TEMP[3].x, TEMP[3].xxxx
> 82: MAX TEMP[0].x, TEMP[0].wwww, IMM[1].xxxx
> 83: SQRT TEMP[0].x, TEMP[0].xxxx
> 84: MOV TEMP[2].z, TEMP[0].xxxx
> 85: RCP TEMP[0].x, TEMP[3].xxxx
> 86: MOV TEMP[2].y, TEMP[0].xxxx
> 87: ADD TEMP[0].yzw, -TEMP[2].xxyz, CONST[2].xxyz
> 88: MAD TEMP[0].xyz, TEMP[1].xxxx, TEMP[0].yzww, TEMP[2]
> 89: MOV TEMP[0].xyz, TEMP[0].xyzx
> 90: ADD TEMP[1].x, -CONST[0].xxxx, IN[2].zzzz
> 91: FSGE TEMP[1].x, TEMP[1].xxxx, IMM[0].wwww
> 92: UIF TEMP[1].xxxx :0
> 93:   MOV TEMP[1].x, IMM[0].yyyy
> 94: ELSE :0
> 95:   MOV TEMP[1].x, IMM[1].yyyy
> 96: ENDIF
> 97: ADD TEMP[2].y, -CONST[5].xxxx, CONST[5].yyyy
> 98: MAD TEMP[2].y, CONST[5].zzzz, TEMP[2].yyyy, CONST[5].xxxx
> 99: MUL TEMP[1].w, TEMP[2].yyyy, TEMP[1].xxxx
>100: MOV TEMP[0].w, TEMP[1].wwww
>101: MOV OUT[0], TEMP[0]
>102: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %23 = load <16 x i8> addrspace(2)* %22, !tbaa !0
>  %24 = call float @llvm.SI.load.const(<16 x i8> %23, i32 0)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %23, i32 16)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %23, i32 20)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %23, i32 28)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %23, i32 32)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %23, i32 36)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %23, i32 40)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %23, i32 48)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %23, i32 52)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %23, i32 56)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %23, i32 64)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %23, i32 68)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %23, i32 72)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %23, i32 80)
>  %38 = call float @llvm.SI.load.const(<16 x i8> %23, i32 84)
>  %39 = call float @llvm.SI.load.const(<16 x i8> %23, i32 88)
>  %40 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %41 = load <8 x i32> addrspace(2)* %40, !tbaa !0
>  %42 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %43 = load <4 x i32> addrspace(2)* %42, !tbaa !0
>  %44 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 1
>  %45 = load <8 x i32> addrspace(2)* %44, !tbaa !0
>  %46 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 1
>  %47 = load <4 x i32> addrspace(2)* %46, !tbaa !0
>  %48 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 2
>  %49 = load <8 x i32> addrspace(2)* %48, !tbaa !0
>  %50 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 2
>  %51 = load <4 x i32> addrspace(2)* %50, !tbaa !0
>  %52 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 3
>  %53 = load <8 x i32> addrspace(2)* %52, !tbaa !0
>  %54 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 3
>  %55 = load <4 x i32> addrspace(2)* %54, !tbaa !0
>  %56 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 4
>  %57 = load <8 x i32> addrspace(2)* %56, !tbaa !0
>  %58 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 4
>  %59 = load <4 x i32> addrspace(2)* %58, !tbaa !0
>  %60 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 5
>  %61 = load <8 x i32> addrspace(2)* %60, !tbaa !0
>  %62 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 5
>  %63 = load <4 x i32> addrspace(2)* %62, !tbaa !0
>  %64 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %65 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %66 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %5, <2 x i32> %7)
>  %67 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %5, <2 x i32> %7)
>  %68 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %5, <2 x i32> %7)
>  %69 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %5, <2 x i32> %7)
>  %70 = call float @llvm.SI.fs.interp(i32 2, i32 1, i32 %5, <2 x i32> %7)
>  %71 = call float @llvm.SI.fs.interp(i32 3, i32 1, i32 %5, <2 x i32> %7)
>  %72 = call float @llvm.SI.fs.interp(i32 0, i32 2, i32 %5, <2 x i32> %7)
>  %73 = call float @llvm.SI.fs.interp(i32 1, i32 2, i32 %5, <2 x i32> %7)
>  %74 = call float @llvm.SI.fs.interp(i32 2, i32 2, i32 %5, <2 x i32> %7)
>  %75 = call float @llvm.SI.fs.interp(i32 1, i32 3, i32 %5, <2 x i32> %7)
>  %76 = call float @llvm.SI.fs.interp(i32 2, i32 3, i32 %5, <2 x i32> %7)
>  %77 = fsub float -0.000000e+00, %31
>  %78 = fadd float %77, %72
>  %79 = fsub float -0.000000e+00, %32
>  %80 = fadd float %79, %73
>  %81 = fsub float -0.000000e+00, %33
>  %82 = fadd float %81, %74
>  %83 = fmul float %82, %25
>  %84 = fmul float %83, 0x3FF7154CA0000000
>  %85 = call float @llvm.AMDIL.exp.(float %84)
>  %86 = fsub float -0.000000e+00, %85
>  %87 = fadd float %86, 1.000000e+00
>  %88 = fmul float %78, %78
>  %89 = fmul float %80, %80
>  %90 = fadd float %89, %88
>  %91 = fmul float %82, %82
>  %92 = fadd float %90, %91
>  %93 = fdiv float 1.000000e+00, %82
>  %94 = fmul float %92, %26
>  %95 = fmul float %87, %94
>  %96 = fmul float %93, %95
>  %97 = fmul float %96, 0x3FF7154CA0000000
>  %98 = call float @llvm.AMDIL.exp.(float %97)
>  %99 = call float @llvm.AMDIL.clamp.(float %98, float 0.000000e+00, float 1.000000e+00)
>  %100 = fsub float -0.000000e+00, %99
>  %101 = fadd float %100, 1.000000e+00
>  %102 = fdiv float 1.000000e+00, %76
>  %103 = fmul float %102, %75
>  %104 = fmul float %103, 5.000000e-01
>  %105 = fadd float %104, 5.000000e-01
>  %106 = bitcast float %105 to i32
>  %107 = bitcast float %27 to i32
>  %108 = insertelement <2 x i32> undef, i32 %106, i32 0
>  %109 = insertelement <2 x i32> %108, i32 %107, i32 1
>  %110 = bitcast <8 x i32> %61 to <32 x i8>
>  %111 = bitcast <4 x i32> %63 to <16 x i8>
>  %112 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %109, <32 x i8> %110, <16 x i8> %111, i32 2)
>  %113 = extractelement <4 x float> %112, i32 0
>  %114 = fmul float %101, %113
>  %115 = fmul float %68, %68
>  %116 = fmul float %69, %69
>  %117 = fadd float %115, %116
>  %118 = fmul float %70, %70
>  %119 = fadd float %117, %118
>  %120 = fmul float 0.000000e+00, 0.000000e+00
>  %121 = fadd float %119, %120
>  %122 = call float @llvm.AMDGPU.rsq.clamped.f32(float %121)
>  %123 = fmul float %68, %122
>  %124 = fmul float %69, %122
>  %125 = fmul float %70, %122
>  %126 = fmul float %64, %64
>  %127 = fmul float %65, %65
>  %128 = fadd float %126, %127
>  %129 = fmul float %66, %66
>  %130 = fadd float %128, %129
>  %131 = fmul float 0.000000e+00, 0.000000e+00
>  %132 = fadd float %130, %131
>  %133 = call float @llvm.AMDGPU.rsq.clamped.f32(float %132)
>  %134 = fmul float %64, %133
>  %135 = fmul float %65, %133
>  %136 = fmul float %66, %133
>  %137 = fmul float %123, %134
>  %138 = fmul float %124, %135
>  %139 = fadd float %138, %137
>  %140 = fmul float %125, %136
>  %141 = fadd float %139, %140
>  %142 = fadd float %141, %141
>  %143 = fsub float -0.000000e+00, %142
>  %144 = fmul float %134, %143
>  %145 = fadd float %144, %123
>  %146 = fsub float -0.000000e+00, %142
>  %147 = fmul float %135, %146
>  %148 = fadd float %147, %124
>  %149 = fsub float -0.000000e+00, %142
>  %150 = fmul float %136, %149
>  %151 = fadd float %150, %125
>  %152 = insertelement <4 x float> undef, float %134, i32 0
>  %153 = insertelement <4 x float> %152, float %135, i32 1
>  %154 = insertelement <4 x float> %153, float %136, i32 2
>  %155 = insertelement <4 x float> %154, float 0.000000e+00, i32 3
>  %156 = call <4 x float> @llvm.AMDGPU.cube(<4 x float> %155)
>  %157 = extractelement <4 x float> %156, i32 0
>  %158 = extractelement <4 x float> %156, i32 1
>  %159 = extractelement <4 x float> %156, i32 2
>  %160 = extractelement <4 x float> %156, i32 3
>  %161 = call float @fabs(float %159)
>  %162 = fdiv float 1.000000e+00, %161
>  %163 = fmul float %157, %162
>  %164 = fadd float %163, 1.500000e+00
>  %165 = fmul float %158, %162
>  %166 = fadd float %165, 1.500000e+00
>  %167 = bitcast float %166 to i32
>  %168 = bitcast float %164 to i32
>  %169 = bitcast float %160 to i32
>  %170 = insertelement <4 x i32> undef, i32 %167, i32 0
>  %171 = insertelement <4 x i32> %170, i32 %168, i32 1
>  %172 = insertelement <4 x i32> %171, i32 %169, i32 2
>  %173 = insertelement <4 x i32> %172, i32 undef, i32 3
>  %174 = bitcast <8 x i32> %49 to <32 x i8>
>  %175 = bitcast <4 x i32> %51 to <16 x i8>
>  %176 = call <4 x float> @llvm.SI.sample.v4i32(<4 x i32> %173, <32 x i8> %174, <16 x i8> %175, i32 4)
>  %177 = extractelement <4 x float> %176, i32 0
>  %178 = extractelement <4 x float> %176, i32 1
>  %179 = extractelement <4 x float> %176, i32 2
>  %180 = fmul float %177, %177
>  %181 = fmul float %178, %178
>  %182 = fmul float %179, %179
>  %183 = insertelement <4 x float> undef, float %145, i32 0
>  %184 = insertelement <4 x float> %183, float %148, i32 1
>  %185 = insertelement <4 x float> %184, float %151, i32 2
>  %186 = insertelement <4 x float> %185, float 0.000000e+00, i32 3
>  %187 = call <4 x float> @llvm.AMDGPU.cube(<4 x float> %186)
>  %188 = extractelement <4 x float> %187, i32 0
>  %189 = extractelement <4 x float> %187, i32 1
>  %190 = extractelement <4 x float> %187, i32 2
>  %191 = extractelement <4 x float> %187, i32 3
>  %192 = call float @fabs(float %190)
>  %193 = fdiv float 1.000000e+00, %192
>  %194 = fmul float %188, %193
>  %195 = fadd float %194, 1.500000e+00
>  %196 = fmul float %189, %193
>  %197 = fadd float %196, 1.500000e+00
>  %198 = bitcast float %197 to i32
>  %199 = bitcast float %195 to i32
>  %200 = bitcast float %191 to i32
>  %201 = insertelement <4 x i32> undef, i32 %198, i32 0
>  %202 = insertelement <4 x i32> %201, i32 %199, i32 1
>  %203 = insertelement <4 x i32> %202, i32 %200, i32 2
>  %204 = insertelement <4 x i32> %203, i32 undef, i32 3
>  %205 = bitcast <8 x i32> %57 to <32 x i8>
>  %206 = bitcast <4 x i32> %59 to <16 x i8>
>  %207 = call <4 x float> @llvm.SI.sample.v4i32(<4 x i32> %204, <32 x i8> %205, <16 x i8> %206, i32 4)
>  %208 = extractelement <4 x float> %207, i32 0
>  %209 = extractelement <4 x float> %207, i32 1
>  %210 = extractelement <4 x float> %207, i32 2
>  %211 = insertelement <4 x float> undef, float %145, i32 0
>  %212 = insertelement <4 x float> %211, float %148, i32 1
>  %213 = insertelement <4 x float> %212, float %151, i32 2
>  %214 = insertelement <4 x float> %213, float %151, i32 3
>  %215 = call <4 x float> @llvm.AMDGPU.cube(<4 x float> %214)
>  %216 = extractelement <4 x float> %215, i32 0
>  %217 = extractelement <4 x float> %215, i32 1
>  %218 = extractelement <4 x float> %215, i32 2
>  %219 = extractelement <4 x float> %215, i32 3
>  %220 = call float @fabs(float %218)
>  %221 = fdiv float 1.000000e+00, %220
>  %222 = fmul float %216, %221
>  %223 = fadd float %222, 1.500000e+00
>  %224 = fmul float %217, %221
>  %225 = fadd float %224, 1.500000e+00
>  %226 = bitcast float %225 to i32
>  %227 = bitcast float %223 to i32
>  %228 = bitcast float %219 to i32
>  %229 = insertelement <4 x i32> undef, i32 %226, i32 0
>  %230 = insertelement <4 x i32> %229, i32 %227, i32 1
>  %231 = insertelement <4 x i32> %230, i32 %228, i32 2
>  %232 = insertelement <4 x i32> %231, i32 undef, i32 3
>  %233 = bitcast <8 x i32> %53 to <32 x i8>
>  %234 = bitcast <4 x i32> %55 to <16 x i8>
>  %235 = call <4 x float> @llvm.SI.sample.v4i32(<4 x i32> %232, <32 x i8> %233, <16 x i8> %234, i32 4)
>  %236 = extractelement <4 x float> %235, i32 0
>  %237 = extractelement <4 x float> %235, i32 1
>  %238 = extractelement <4 x float> %235, i32 2
>  %239 = fmul float %236, %236
>  %240 = fmul float %237, %237
>  %241 = fmul float %238, %238
>  %242 = fsub float -0.000000e+00, %239
>  %243 = fmul float %208, %208
>  %244 = fadd float %243, %242
>  %245 = fsub float -0.000000e+00, %240
>  %246 = fmul float %209, %209
>  %247 = fadd float %246, %245
>  %248 = fsub float -0.000000e+00, %241
>  %249 = fmul float %210, %210
>  %250 = fadd float %249, %248
>  %251 = bitcast float %67 to i32
>  %252 = bitcast float %71 to i32
>  %253 = insertelement <2 x i32> undef, i32 %251, i32 0
>  %254 = insertelement <2 x i32> %253, i32 %252, i32 1
>  %255 = bitcast <8 x i32> %45 to <32 x i8>
>  %256 = bitcast <4 x i32> %47 to <16 x i8>
>  %257 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %254, <32 x i8> %255, <16 x i8> %256, i32 2)
>  %258 = extractelement <4 x float> %257, i32 0
>  %259 = extractelement <4 x float> %257, i32 1
>  %260 = extractelement <4 x float> %257, i32 2
>  %261 = extractelement <4 x float> %257, i32 3
>  %262 = bitcast float %67 to i32
>  %263 = bitcast float %71 to i32
>  %264 = insertelement <2 x i32> undef, i32 %262, i32 0
>  %265 = insertelement <2 x i32> %264, i32 %263, i32 1
>  %266 = bitcast <8 x i32> %41 to <32 x i8>
>  %267 = bitcast <4 x i32> %43 to <16 x i8>
>  %268 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %265, <32 x i8> %266, <16 x i8> %267, i32 2)
>  %269 = extractelement <4 x float> %268, i32 0
>  %270 = extractelement <4 x float> %268, i32 1
>  %271 = extractelement <4 x float> %268, i32 2
>  %272 = extractelement <4 x float> %268, i32 3
>  %273 = fmul float %269, %269
>  %274 = fmul float %270, %270
>  %275 = fmul float %271, %271
>  %276 = fmul float %272, %272
>  %277 = fmul float %258, %258
>  %278 = fmul float %259, %259
>  %279 = fmul float %260, %260
>  %280 = fmul float %261, %261
>  %281 = fmul float %280, %280
>  %282 = fmul float %281, %281
>  %283 = fmul float %281, %281
>  %284 = fmul float %281, %281
>  %285 = fmul float %281, %281
>  %286 = call float @llvm.AMDIL.clamp.(float %282, float 0.000000e+00, float 1.000000e+00)
>  %287 = call float @llvm.AMDIL.clamp.(float %283, float 0.000000e+00, float 1.000000e+00)
>  %288 = call float @llvm.AMDIL.clamp.(float %284, float 0.000000e+00, float 1.000000e+00)
>  %289 = call float @llvm.AMDIL.clamp.(float %285, float 0.000000e+00, float 1.000000e+00)
>  %290 = fmul float %289, %244
>  %291 = fadd float %290, %239
>  %292 = fmul float %289, %247
>  %293 = fadd float %292, %240
>  %294 = fmul float %289, %250
>  %295 = fadd float %294, %241
>  %296 = fmul float %291, %277
>  %297 = fmul float %293, %278
>  %298 = fmul float %295, %279
>  %299 = fmul float %34, %34
>  %300 = fmul float %35, %35
>  %301 = fmul float %36, %36
>  %302 = fsub float -0.000000e+00, %34
>  %303 = fmul float %34, %302
>  %304 = fadd float %303, 1.000000e+00
>  %305 = fsub float -0.000000e+00, %35
>  %306 = fmul float %35, %305
>  %307 = fadd float %306, 1.000000e+00
>  %308 = fsub float -0.000000e+00, %36
>  %309 = fmul float %36, %308
>  %310 = fadd float %309, 1.000000e+00
>  %311 = fmul float %276, %304
>  %312 = fadd float %311, %299
>  %313 = fmul float %276, %307
>  %314 = fadd float %313, %300
>  %315 = fmul float %276, %310
>  %316 = fadd float %315, %301
>  %317 = fmul float %273, %312
>  %318 = fmul float %274, %314
>  %319 = fmul float %275, %316
>  %320 = fmul float %317, %180
>  %321 = fadd float %320, %296
>  %322 = fmul float %318, %181
>  %323 = fadd float %322, %297
>  %324 = fmul float %319, %182
>  %325 = fadd float %324, %298
>  %326 = call float @llvm.maxnum.f32(float %321, float 0x3E7AD7F2A0000000)
>  %327 = call float @llvm.AMDGPU.rsq.clamped.f32(float %326)
>  %328 = fdiv float 1.000000e+00, %327
>  %329 = call float @llvm.maxnum.f32(float %323, float 0x3E7AD7F2A0000000)
>  %330 = call float @llvm.AMDGPU.rsq.clamped.f32(float %329)
>  %331 = call float @llvm.maxnum.f32(float %325, float 0x3E7AD7F2A0000000)
>  %332 = call float @llvm.sqrt.f32(float %331)
>  %333 = fdiv float 1.000000e+00, %330
>  %334 = fsub float -0.000000e+00, %328
>  %335 = fadd float %334, %28
>  %336 = fsub float -0.000000e+00, %333
>  %337 = fadd float %336, %29
>  %338 = fsub float -0.000000e+00, %332
>  %339 = fadd float %338, %30
>  %340 = fmul float %114, %335
>  %341 = fadd float %340, %328
>  %342 = fmul float %114, %337
>  %343 = fadd float %342, %333
>  %344 = fmul float %114, %339
>  %345 = fadd float %344, %332
>  %346 = fsub float -0.000000e+00, %24
>  %347 = fadd float %346, %74
>  %348 = fcmp oge float %347, 0.000000e+00
>  %349 = sext i1 %348 to i32
>  %350 = bitcast i32 %349 to float
>  %351 = bitcast float %350 to i32
>  %352 = icmp ne i32 %351, 0
>  %. = select i1 %352, float 1.000000e+00, float 0x3FD51EB860000000
>  %353 = fsub float -0.000000e+00, %37
>  %354 = fadd float %353, %38
>  %355 = fmul float %39, %354
>  %356 = fadd float %355, %37
>  %357 = fmul float %356, %.
>  %358 = call i32 @llvm.SI.packf16(float %341, float %343)
>  %359 = bitcast i32 %358 to float
>  %360 = call i32 @llvm.SI.packf16(float %345, float %357)
>  %361 = bitcast i32 %360 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %359, float %361, float %359, float %361)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.exp.(float) #2
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.clamp.(float, float, float) #2
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.AMDGPU.rsq.clamped.f32(float) #1
>
>; Function Attrs: readnone
>declare <4 x float> @llvm.AMDGPU.cube(<4 x float>) #2
>
>; Function Attrs: readnone
>declare float @fabs(float) #2
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v4i32(<4 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.maxnum.f32(float, float) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.sqrt.f32(float) #1
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                         ; BEFE0A7E
>	s_mov_b32 m0, s9                                                             ; BEFC0309
>	v_interp_p1_f32 v2, v0, 0, 0, [m0]                                           ; C8080000
>	v_interp_p2_f32 v2, [v2], v1, 0, 0, [m0]                                     ; C8090001
>	v_interp_p1_f32 v3, v0, 1, 0, [m0]                                           ; C80C0100
>	v_interp_p2_f32 v3, [v3], v1, 1, 0, [m0]                                     ; C80D0101
>	v_mul_f32_e32 v4, v3, v3                                                     ; 10080703
>	v_mad_f32 v4, v2, v2, v4                                                     ; D2820004 04120502
>	v_interp_p1_f32 v5, v0, 2, 0, [m0]                                           ; C8140200
>	v_interp_p2_f32 v5, [v5], v1, 2, 0, [m0]                                     ; C8150201
>	v_mad_f32 v4, v5, v5, v4                                                     ; D2820004 04120B05
>	v_mov_b32_e32 v9, 0                                                          ; 7E120280
>	v_add_f32_e32 v4, 0, v4                                                      ; 06080880
>	v_rsq_clamp_f32_e32 v4, v4                                                   ; 7E085904
>	v_mul_f32_e32 v10, v4, v2                                                    ; 10140504
>	v_interp_p1_f32 v2, v0, 0, 1, [m0]                                           ; C8080400
>	v_interp_p2_f32 v2, [v2], v1, 0, 1, [m0]                                     ; C8090401
>	v_interp_p1_f32 v14, v0, 1, 1, [m0]                                          ; C8380500
>	v_interp_p2_f32 v14, [v14], v1, 1, 1, [m0]                                   ; C8390501
>	v_mul_f32_e32 v15, v14, v14                                                  ; 101E1D0E
>	v_mad_f32 v15, v2, v2, v15                                                   ; D282000F 043E0502
>	v_interp_p1_f32 v16, v0, 2, 1, [m0]                                          ; C8400600
>	v_interp_p2_f32 v16, [v16], v1, 2, 1, [m0]                                   ; C8410601
>	v_mad_f32 v15, v16, v16, v15                                                 ; D282000F 043E2110
>	v_add_f32_e32 v15, 0, v15                                                    ; 061E1E80
>	v_rsq_clamp_f32_e32 v15, v15                                                 ; 7E1E590F
>	v_mul_f32_e32 v17, v15, v2                                                   ; 1022050F
>	v_mul_f32_e32 v17, v10, v17                                                  ; 1022230A
>	v_mul_f32_e32 v11, v4, v3                                                    ; 10160704
>	v_mul_f32_e32 v3, v15, v14                                                   ; 10061D0F
>	v_mad_f32 v3, v3, v11, v17                                                   ; D2820003 04461703
>	v_mul_f32_e32 v12, v4, v5                                                    ; 10180B04
>	v_mul_f32_e32 v4, v15, v16                                                   ; 1008210F
>	v_mad_f32 v3, v4, v12, v3                                                    ; D2820003 040E1904
>	v_add_f32_e32 v3, v3, v3                                                     ; 06060703
>	v_mul_f32_e32 v4, v3, v12                                                    ; 10081903
>	v_mad_f32 v8, v16, v15, -v4                                                  ; D2820008 84121F10
>	v_mul_f32_e32 v4, v3, v11                                                    ; 10081703
>	v_mad_f32 v7, v14, v15, -v4                                                  ; D2820007 84121F0E
>	v_mul_f32_e32 v3, v3, v10                                                    ; 10061503
>	v_mad_f32 v6, v2, v15, -v3                                                   ; D2820006 840E1F02
>	v_cubeid_f32 v17, v6, v7, v8                                                 ; D2880011 04220F06
>	v_cubema_f32 v16, v6, v7, v8                                                 ; D28E0010 04220F06
>	v_cubesc_f32 v15, v6, v7, v8                                                 ; D28A000F 04220F06
>	v_cubetc_f32 v14, v6, v7, v8                                                 ; D28C000E 04220F06
>	v_rcp_f32_e64 v2, |v16|                                                      ; D3540102 00000110
>	v_mov_b32_e32 v3, 0x3fc00000                                                 ; 7E0602FF 3FC00000
>	v_mad_f32 v16, v14, v2, v3                                                   ; D2820010 040E050E
>	v_mad_f32 v15, v15, v2, v3                                                   ; D282000F 040E050F
>	s_load_dwordx4 s[20:23], s[4:5], 0x0                                         ; C08A0500
>	s_load_dwordx4 s[44:47], s[4:5], 0x4                                         ; C0960504
>	s_load_dwordx4 s[24:27], s[4:5], 0x8                                         ; C08C0508
>	s_load_dwordx4 s[48:51], s[4:5], 0xc                                         ; C098050C
>	s_load_dwordx4 s[52:55], s[4:5], 0x10                                        ; C09A0510
>	s_load_dwordx4 s[8:11], s[4:5], 0x14                                         ; C0840514
>	s_load_dwordx8 s[28:35], s[6:7], 0x0                                         ; C0CE0700
>	s_load_dwordx8 s[56:63], s[6:7], 0x8                                         ; C0DC0708
>	s_load_dwordx8 s[36:43], s[6:7], 0x10                                        ; C0D20710
>	s_load_dwordx8 s[64:71], s[6:7], 0x18                                        ; C0E00718
>	s_load_dwordx8 s[72:79], s[6:7], 0x20                                        ; C0E40720
>	s_load_dwordx8 s[12:19], s[6:7], 0x28                                        ; C0C60728
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample v[14:16], 7, 0, 0, 0, 0, 0, 0, 0, v[15:18], s[72:79], s[52:55]  ; F0800700 01B20E0F
>	v_mov_b32_e32 v17, v6                                                        ; 7E220306
>	v_mov_b32_e32 v18, v7                                                        ; 7E240307
>	v_mov_b32_e32 v19, v8                                                        ; 7E260308
>	v_mov_b32_e32 v20, v9                                                        ; 7E280309
>	v_mov_b32_e32 v20, v8                                                        ; 7E280308
>	v_cubeid_f32 v24, v17, v18, v19                                              ; D2880018 044E2511
>	v_cubema_f32 v23, v17, v18, v19                                              ; D28E0017 044E2511
>	v_cubesc_f32 v22, v17, v18, v19                                              ; D28A0016 044E2511
>	v_cubetc_f32 v21, v17, v18, v19                                              ; D28C0015 044E2511
>	v_rcp_f32_e64 v2, |v23|                                                      ; D3540102 00000117
>	v_mad_f32 v23, v21, v2, v3                                                   ; D2820017 040E0515
>	v_mad_f32 v22, v22, v2, v3                                                   ; D2820016 040E0516
>	image_sample v[17:19], 7, 0, 0, 0, 0, 0, 0, 0, v[22:25], s[64:71], s[48:51]  ; F0800700 01901116
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mul_f32_e32 v2, v18, v18                                                   ; 10042512
>	v_mad_f32 v4, v15, v15, -v2                                                  ; D2820004 840A1F0F
>	v_interp_p1_f32 v21, v0, 3, 1, [m0]                                          ; C8540700
>	v_interp_p2_f32 v21, [v21], v1, 3, 1, [m0]                                   ; C8550701
>	v_interp_p1_f32 v20, v0, 3, 0, [m0]                                          ; C8500300
>	v_interp_p2_f32 v20, [v20], v1, 3, 0, [m0]                                   ; C8510301
>	image_sample v[22:25], 15, 0, 0, 0, 0, 0, 0, 0, v[20:21], s[56:63], s[44:47] ; F0800F00 016E1614
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mul_f32_e32 v5, v25, v25                                                   ; 100A3319
>	v_mul_f32_e32 v5, v5, v5                                                     ; 100A0B05
>	v_mul_f32_e32 v5, v5, v5                                                     ; 100A0B05
>	v_add_f32_e64 v5, 0, v5 clamp                                                ; D2060805 00020A80
>	v_mad_f32 v2, v5, v4, v2                                                     ; D2820002 040A0905
>	v_mul_f32_e32 v4, v23, v23                                                   ; 10082F17
>	v_mul_f32_e32 v2, v4, v2                                                     ; 10040504
>	v_mov_b32_e32 v13, v9                                                        ; 7E1A0309
>	v_cubeid_f32 v29, v10, v11, v12                                              ; D288001D 0432170A
>	v_cubema_f32 v28, v10, v11, v12                                              ; D28E001C 0432170A
>	v_cubesc_f32 v27, v10, v11, v12                                              ; D28A001B 0432170A
>	v_cubetc_f32 v26, v10, v11, v12                                              ; D28C001A 0432170A
>	v_rcp_f32_e64 v4, |v28|                                                      ; D3540104 0000011C
>	v_mad_f32 v28, v26, v4, v3                                                   ; D282001C 040E091A
>	v_mad_f32 v27, v27, v4, v3                                                   ; D282001B 040E091B
>	image_sample v[6:8], 7, 0, 0, 0, 0, 0, 0, 0, v[27:30], s[36:43], s[24:27]    ; F0800700 00C9061B
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mul_f32_e32 v3, v7, v7                                                     ; 10060F07
>	image_sample v[9:12], 15, 0, 0, 0, 0, 0, 0, 0, v[20:21], s[28:35], s[20:23]  ; F0800F00 00A70914
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mul_f32_e32 v4, v12, v12                                                   ; 1008190C
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                                           ; C0800300
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	s_buffer_load_dword s4, s[0:3], 0x11                                         ; C2020111
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v13, -s4, s4, 1.0                                                  ; D282000D 23C80804
>	v_mul_f32_e64 v20, s4, s4                                                    ; D2100014 00000804
>	v_mad_f32 v13, v4, v13, v20                                                  ; D282000D 04521B04
>	v_mul_f32_e32 v20, v10, v10                                                  ; 1028150A
>	v_mul_f32_e32 v13, v13, v20                                                  ; 101A290D
>	v_mad_f32 v2, v13, v3, v2                                                    ; D2820002 040A070D
>	v_max_f32_e32 v2, 0x33d6bf95, v2                                             ; 200404FF 33D6BF95
>	v_rsq_clamp_f32_e32 v2, v2                                                   ; 7E045902
>	v_rcp_f32_e32 v2, v2                                                         ; 7E045502
>	s_buffer_load_dword s4, s[0:3], 0x9                                          ; C2020109
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v3, s4, v2                                                     ; 08060404
>	v_interp_p1_f32 v13, v0, 1, 2, [m0]                                          ; C8340900
>	v_interp_p2_f32 v13, [v13], v1, 1, 2, [m0]                                   ; C8350901
>	s_buffer_load_dword s4, s[0:3], 0xd                                          ; C202010D
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_subrev_f32_e32 v13, s4, v13                                                ; 0A1A1A04
>	v_interp_p1_f32 v20, v0, 0, 2, [m0]                                          ; C8500800
>	v_interp_p2_f32 v20, [v20], v1, 0, 2, [m0]                                   ; C8510801
>	s_buffer_load_dword s4, s[0:3], 0xc                                          ; C202010C
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_subrev_f32_e32 v20, s4, v20                                                ; 0A282804
>	v_mul_f32_e32 v20, v20, v20                                                  ; 10282914
>	v_mad_f32 v13, v13, v13, v20                                                 ; D282000D 04521B0D
>	v_interp_p1_f32 v20, v0, 2, 2, [m0]                                          ; C8500A00
>	v_interp_p2_f32 v20, [v20], v1, 2, 2, [m0]                                   ; C8510A01
>	s_buffer_load_dword s4, s[0:3], 0xe                                          ; C202010E
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_subrev_f32_e32 v21, s4, v20                                                ; 0A2A2804
>	v_mad_f32 v13, v21, v21, v13                                                 ; D282000D 04362B15
>	s_buffer_load_dword s4, s[0:3], 0x5                                          ; C2020105
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v13, s4, v13                                                   ; 101A1A04
>	s_buffer_load_dword s4, s[0:3], 0x4                                          ; C2020104
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v26, s4, v21                                                   ; 10342A04
>	v_mul_f32_e32 v26, 0x3fb8aa65, v26                                           ; 103434FF 3FB8AA65
>	v_exp_f32_e32 v26, v26                                                       ; 7E344B1A
>	v_sub_f32_e32 v26, 1.0, v26                                                  ; 083434F2
>	v_mul_f32_e32 v13, v13, v26                                                  ; 101A350D
>	v_rcp_f32_e32 v21, v21                                                       ; 7E2A5515
>	v_mul_f32_e32 v13, v13, v21                                                  ; 101A2B0D
>	v_mul_f32_e32 v13, 0x3fb8aa65, v13                                           ; 101A1AFF 3FB8AA65
>	v_exp_f32_e32 v13, v13                                                       ; 7E1A4B0D
>	v_add_f32_e64 v13, 0, v13 clamp                                              ; D206080D 00021A80
>	v_sub_f32_e32 v13, 1.0, v13                                                  ; 081A1AF2
>	v_interp_p1_f32 v21, v0, 1, 3, [m0]                                          ; C8540D00
>	v_interp_p2_f32 v21, [v21], v1, 1, 3, [m0]                                   ; C8550D01
>	v_interp_p1_f32 v26, v0, 2, 3, [m0]                                          ; C8680E00
>	v_interp_p2_f32 v26, [v26], v1, 2, 3, [m0]                                   ; C8690E01
>	v_rcp_f32_e32 v0, v26                                                        ; 7E00551A
>	v_mul_f32_e32 v0, v21, v0                                                    ; 10000115
>	v_mad_f32 v0, 0.5, v0, 0.5                                                   ; D2820000 03C200F0
>	s_buffer_load_dword s4, s[0:3], 0x7                                          ; C2020107
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mov_b32_e32 v1, s4                                                         ; 7E020204
>	image_sample v0, 1, 0, 0, 0, 0, 0, 0, 0, v[0:1], s[12:19], s[8:11]           ; F0800100 00430000
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mul_f32_e32 v0, v0, v13                                                    ; 10001B00
>	v_mad_f32 v1, v0, v3, v2                                                     ; D2820001 040A0700
>	v_mul_f32_e32 v2, v17, v17                                                   ; 10042311
>	v_mad_f32 v3, v14, v14, -v2                                                  ; D2820003 840A1D0E
>	v_mad_f32 v2, v5, v3, v2                                                     ; D2820002 040A0705
>	v_mul_f32_e32 v3, v22, v22                                                   ; 10062D16
>	v_mul_f32_e32 v2, v3, v2                                                     ; 10040503
>	v_mul_f32_e32 v3, v6, v6                                                     ; 10060D06
>	s_buffer_load_dword s4, s[0:3], 0x10                                         ; C2020110
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v13, -s4, s4, 1.0                                                  ; D282000D 23C80804
>	v_mul_f32_e64 v21, s4, s4                                                    ; D2100015 00000804
>	v_mad_f32 v13, v4, v13, v21                                                  ; D282000D 04561B04
>	v_mul_f32_e32 v21, v9, v9                                                    ; 102A1309
>	v_mul_f32_e32 v13, v13, v21                                                  ; 101A2B0D
>	v_mad_f32 v2, v13, v3, v2                                                    ; D2820002 040A070D
>	v_max_f32_e32 v2, 0x33d6bf95, v2                                             ; 200404FF 33D6BF95
>	v_rsq_clamp_f32_e32 v2, v2                                                   ; 7E045902
>	v_rcp_f32_e32 v2, v2                                                         ; 7E045502
>	s_buffer_load_dword s4, s[0:3], 0x8                                          ; C2020108
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v3, s4, v2                                                     ; 08060404
>	v_mad_f32 v2, v0, v3, v2                                                     ; D2820002 040A0700
>	v_cvt_pkrtz_f16_f32_e32 v1, v2, v1                                           ; 5E020302
>	v_mul_f32_e32 v2, v19, v19                                                   ; 10042713
>	v_mad_f32 v3, v16, v16, -v2                                                  ; D2820003 840A2110
>	v_mad_f32 v2, v5, v3, v2                                                     ; D2820002 040A0705
>	v_mul_f32_e32 v3, v24, v24                                                   ; 10063118
>	v_mul_f32_e32 v2, v3, v2                                                     ; 10040503
>	v_mul_f32_e32 v3, v8, v8                                                     ; 10061108
>	s_buffer_load_dword s4, s[0:3], 0x12                                         ; C2020112
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v5, -s4, s4, 1.0                                                   ; D2820005 23C80804
>	v_mul_f32_e64 v6, s4, s4                                                     ; D2100006 00000804
>	v_mad_f32 v4, v4, v5, v6                                                     ; D2820004 041A0B04
>	v_mul_f32_e32 v5, v11, v11                                                   ; 100A170B
>	v_mul_f32_e32 v4, v4, v5                                                     ; 10080B04
>	v_mad_f32 v2, v4, v3, v2                                                     ; D2820002 040A0704
>	v_max_f32_e32 v2, 0x33d6bf95, v2                                             ; 200404FF 33D6BF95
>	v_sqrt_f32_e32 v2, v2                                                        ; 7E046702
>	s_buffer_load_dword s4, s[0:3], 0xa                                          ; C202010A
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v3, s4, v2                                                     ; 08060404
>	v_mad_f32 v0, v0, v3, v2                                                     ; D2820000 040A0700
>	s_buffer_load_dword s4, s[0:3], 0x0                                          ; C2020100
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_subrev_f32_e32 v2, s4, v20                                                 ; 0A042804
>	v_cmp_ge_f32_e64 s[4:5], v2, 0                                               ; D00C0004 00010102
>	v_cndmask_b32_e64 v2, 0, -1, s[4:5]                                          ; D2000002 00118280
>	v_cmp_ne_i32_e64 s[4:5], v2, 0                                               ; D10A0004 00010102
>	v_mov_b32_e32 v2, 0x3ea8f5c3                                                 ; 7E0402FF 3EA8F5C3
>	v_cndmask_b32_e64 v2, v2, 1.0, s[4:5]                                        ; D2000002 1011E502
>	s_buffer_load_dword s4, s[0:3], 0x14                                         ; C2020114
>	s_buffer_load_dword s5, s[0:3], 0x15                                         ; C2028115
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mov_b32_e32 v3, s4                                                         ; 7E060204
>	v_sub_f32_e32 v3, s5, v3                                                     ; 08060605
>	s_buffer_load_dword s0, s[0:3], 0x16                                         ; C2000116
>	v_mov_b32_e32 v4, s4                                                         ; 7E080204
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v3, v3, s0, v4                                                     ; D2820003 04100103
>	v_mul_f32_e32 v2, v2, v3                                                     ; 10040702
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v2                                           ; 5E000500
>	exp 15, 0, 1, 1, 1, v1, v0, v1, v0                                           ; F8001C0F 00010001
>	s_endpgm                                                                     ; BF810000
>
>SGPRS: 80
>VGPRS: 36
>Code Size: 1232 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL CONST[0..95]
>DCL TEMP[0..2], LOCAL
>DCL ADDR[0]
>IMM[0] FLT32 {    4.0000,     0.0000,     0.0000,     0.0000}
>IMM[1] INT32 {1, 2, 3, 0}
>  0: FRC TEMP[0].x, IN[1].zzzz
>  1: ADD TEMP[1].x, -TEMP[0].xxxx, IN[1].zzzz
>  2: MUL TEMP[1].x, TEMP[1].xxxx, IMM[0].xxxx
>  3: F2I TEMP[2].x, TEMP[1].xxxx
>  4: UADD TEMP[2].x, IMM[1].xxxx, TEMP[2].xxxx
>  5: UARL ADDR[0].x, TEMP[2].xxxx
>  6: MUL TEMP[2].xyz, IN[0].yyyy, CONST[ADDR[0].x].xyww
>  7: MOV TEMP[0].xyz, TEMP[2].xyzx
>  8: F2I TEMP[2].x, TEMP[1].xxxx
>  9: UARL ADDR[0].x, TEMP[2].xxxx
> 10: MAD TEMP[2].xyz, IN[0].xxxx, CONST[ADDR[0].x].xyww, TEMP[0]
> 11: MOV TEMP[0].xyz, TEMP[2].xyzx
> 12: F2I TEMP[2].x, TEMP[1].xxxx
> 13: UADD TEMP[2].x, IMM[1].yyyy, TEMP[2].xxxx
> 14: UARL ADDR[0].x, TEMP[2].xxxx
> 15: MAD TEMP[0].xyz, IN[0].zzzz, CONST[ADDR[0].x].xyww, TEMP[0]
> 16: F2I TEMP[1].x, TEMP[1].xxxx
> 17: UADD TEMP[1].x, IMM[1].zzzz, TEMP[1].xxxx
> 18: UARL ADDR[0].x, TEMP[1].xxxx
> 19: ADD TEMP[0].xyw, TEMP[0].xyzz, CONST[ADDR[0].x]
> 20: MOV TEMP[0].xyw, TEMP[0].xyxw
> 21: MOV TEMP[0].z, IMM[0].yyyy
> 22: MOV OUT[0], TEMP[0]
> 23: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %14 = load <16 x i8> addrspace(2)* %13, !tbaa !0
>  %15 = add i32 %5, %7
>  %16 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %14, i32 0, i32 %15)
>  %17 = extractelement <4 x float> %16, i32 0
>  %18 = extractelement <4 x float> %16, i32 1
>  %19 = extractelement <4 x float> %16, i32 2
>  %20 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %21 = load <16 x i8> addrspace(2)* %20, !tbaa !0
>  %22 = add i32 %5, %7
>  %23 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %21, i32 0, i32 %22)
>  %24 = extractelement <4 x float> %23, i32 2
>  %25 = call float @llvm.AMDIL.fraction.(float %24)
>  %26 = fsub float -0.000000e+00, %25
>  %27 = fadd float %26, %24
>  %28 = fmul float %27, 4.000000e+00
>  %29 = fptosi float %28 to i32
>  %30 = bitcast i32 %29 to float
>  %31 = bitcast float %30 to i32
>  %32 = add i32 1, %31
>  %33 = bitcast i32 %32 to float
>  %34 = bitcast float %33 to i32
>  %35 = shl i32 %34, 4
>  %36 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %35)
>  %37 = fmul float %18, %36
>  %38 = shl i32 %34, 4
>  %39 = add i32 %38, 4
>  %40 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %39)
>  %41 = fmul float %18, %40
>  %42 = shl i32 %34, 4
>  %43 = add i32 %42, 12
>  %44 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %43)
>  %45 = fmul float %18, %44
>  %46 = fptosi float %28 to i32
>  %47 = bitcast i32 %46 to float
>  %48 = bitcast float %47 to i32
>  %49 = shl i32 %48, 4
>  %50 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %49)
>  %51 = fmul float %17, %50
>  %52 = fadd float %51, %37
>  %53 = shl i32 %48, 4
>  %54 = add i32 %53, 4
>  %55 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %54)
>  %56 = fmul float %17, %55
>  %57 = fadd float %56, %41
>  %58 = shl i32 %48, 4
>  %59 = add i32 %58, 12
>  %60 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %59)
>  %61 = fmul float %17, %60
>  %62 = fadd float %61, %45
>  %63 = fptosi float %28 to i32
>  %64 = bitcast i32 %63 to float
>  %65 = bitcast float %64 to i32
>  %66 = add i32 2, %65
>  %67 = bitcast i32 %66 to float
>  %68 = bitcast float %67 to i32
>  %69 = shl i32 %68, 4
>  %70 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %69)
>  %71 = fmul float %19, %70
>  %72 = fadd float %71, %52
>  %73 = shl i32 %68, 4
>  %74 = add i32 %73, 4
>  %75 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %74)
>  %76 = fmul float %19, %75
>  %77 = fadd float %76, %57
>  %78 = shl i32 %68, 4
>  %79 = add i32 %78, 12
>  %80 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %79)
>  %81 = fmul float %19, %80
>  %82 = fadd float %81, %62
>  %83 = fptosi float %28 to i32
>  %84 = bitcast i32 %83 to float
>  %85 = bitcast float %84 to i32
>  %86 = add i32 3, %85
>  %87 = bitcast i32 %86 to float
>  %88 = bitcast float %87 to i32
>  %89 = shl i32 %88, 4
>  %90 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %89)
>  %91 = fadd float %72, %90
>  %92 = shl i32 %88, 4
>  %93 = add i32 %92, 4
>  %94 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %93)
>  %95 = fadd float %77, %94
>  %96 = shl i32 %88, 4
>  %97 = add i32 %96, 12
>  %98 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %97)
>  %99 = fadd float %82, %98
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %91, float %95, float 0.000000e+00, float %99)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.fraction.(float) #2
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                            ; 4A00000A
>	s_load_dwordx4 s[4:7], s[8:9], 0x0                   ; C0820900
>	s_load_dwordx4 s[8:11], s[8:9], 0x4                  ; C0840904
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[8:11], 0 idxen ; E00C2000 80020100
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_fract_f32_e32 v5, v3                               ; 7E0A4103
>	v_subrev_f32_e32 v1, v5, v3                          ; 0A020705
>	v_mul_f32_e32 v1, 4.0, v1                            ; 100202F6
>	v_cvt_i32_f32_e32 v1, v1                             ; 7E021101
>	v_lshlrev_b32_e32 v1, 4, v1                          ; 34020284
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                   ; C0800300
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	buffer_load_dword v2, v1, s[0:3], 0 offen            ; E0301000 80000201
>	v_add_i32_e32 v3, 16, v1                             ; 4A060290
>	buffer_load_dword v4, v3, s[0:3], 0 offen            ; E0301000 80000403
>	buffer_load_format_xyzw v[5:8], v0, s[4:7], 0 idxen  ; E00C2000 80010500
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_mul_f32_e32 v0, v4, v6                             ; 10000D04
>	v_mad_f32 v0, v5, v2, v0                             ; D2820000 04020505
>	v_add_i32_e32 v2, 32, v1                             ; 4A0402A0
>	buffer_load_dword v4, v2, s[0:3], 0 offen            ; E0301000 80000402
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_mad_f32 v0, v7, v4, v0                             ; D2820000 04020907
>	v_add_i32_e32 v4, 48, v1                             ; 4A0802B0
>	buffer_load_dword v9, v4, s[0:3], 0 offen            ; E0301000 80000904
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_add_f32_e32 v0, v9, v0                             ; 06000109
>	v_or_b32_e32 v9, 12, v1                              ; 3812028C
>	buffer_load_dword v9, v9, s[0:3], 0 offen            ; E0301000 80000909
>	v_or_b32_e32 v10, 12, v3                             ; 3814068C
>	buffer_load_dword v10, v10, s[0:3], 0 offen          ; E0301000 80000A0A
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_mul_f32_e32 v10, v10, v6                           ; 10140D0A
>	v_mad_f32 v9, v5, v9, v10                            ; D2820009 042A1305
>	v_or_b32_e32 v10, 12, v2                             ; 3814048C
>	buffer_load_dword v10, v10, s[0:3], 0 offen          ; E0301000 80000A0A
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_mad_f32 v9, v7, v10, v9                            ; D2820009 04261507
>	v_or_b32_e32 v10, 12, v4                             ; 3814088C
>	buffer_load_dword v10, v10, s[0:3], 0 offen          ; E0301000 80000A0A
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_add_f32_e32 v9, v10, v9                            ; 0612130A
>	v_or_b32_e32 v1, 4, v1                               ; 38020284
>	buffer_load_dword v1, v1, s[0:3], 0 offen            ; E0301000 80000101
>	v_or_b32_e32 v3, 4, v3                               ; 38060684
>	buffer_load_dword v3, v3, s[0:3], 0 offen            ; E0301000 80000303
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_mul_f32_e32 v3, v3, v6                             ; 10060D03
>	v_mad_f32 v1, v5, v1, v3                             ; D2820001 040E0305
>	v_or_b32_e32 v2, 4, v2                               ; 38040484
>	buffer_load_dword v2, v2, s[0:3], 0 offen            ; E0301000 80000202
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_mad_f32 v1, v7, v2, v1                             ; D2820001 04060507
>	v_or_b32_e32 v2, 4, v4                               ; 38040884
>	buffer_load_dword v2, v2, s[0:3], 0 offen            ; E0301000 80000202
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_add_f32_e32 v1, v2, v1                             ; 06020302
>	v_mov_b32_e32 v2, 0                                  ; 7E040280
>	exp 15, 12, 0, 1, 0, v0, v1, v2, v9                  ; F80008CF 09020100
>	s_endpgm                                             ; BF810000
>
>SGPRS: 16
>VGPRS: 12
>Code Size: 328 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL OUT[0], COLOR
>IMM[0] FLT32 {    0.0000,     0.0000,     0.0000,     0.0000}
>  0: MOV OUT[0], IMM[0].xxxx
>  1: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = call i32 @llvm.SI.packf16(float 0.000000e+00, float 0.000000e+00)
>  %23 = bitcast i32 %22 to float
>  %24 = call i32 @llvm.SI.packf16(float 0.000000e+00, float 0.000000e+00)
>  %25 = bitcast i32 %24 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %23, float %25, float %23, float %25)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>Shader Disassembly:
>
>	v_cvt_pkrtz_f16_f32_e64 v0, 0, 0   ; D25E0000 00010080
>	exp 15, 0, 1, 1, 1, v0, v0, v0, v0 ; F8001C0F 00000000
>	s_endpgm                           ; BF810000
>
>SGPRS: 8
>VGPRS: 4
>Code Size: 20 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[9]
>DCL OUT[2], GENERIC[10]
>DCL CONST[0..196]
>DCL TEMP[0..4], LOCAL
>DCL ADDR[0]
>IMM[0] FLT32 {    4.0000,     0.5000,    -0.5000,     0.0000}
>IMM[1] INT32 {1, 2, 3, 0}
>IMM[2] FLT32 {    0.0000,     1.0000,     0.0000,     0.0000}
>  0: FRC TEMP[0].x, IN[1].zzzz
>  1: ADD TEMP[1].x, -TEMP[0].xxxx, IN[1].zzzz
>  2: MUL TEMP[1].x, TEMP[1].xxxx, IMM[0].xxxx
>  3: F2I TEMP[2].x, TEMP[1].xxxx
>  4: UADD TEMP[2].x, IMM[1].xxxx, TEMP[2].xxxx
>  5: UARL ADDR[0].x, TEMP[2].xxxx
>  6: UARL ADDR[0].x, TEMP[2].xxxx
>  7: MUL TEMP[0], IN[0].yyyy, CONST[ADDR[0].x]
>  8: F2I TEMP[2].x, TEMP[1].xxxx
>  9: UARL ADDR[0].x, TEMP[2].xxxx
> 10: UARL ADDR[0].x, TEMP[2].xxxx
> 11: MAD TEMP[0], IN[0].xxxx, CONST[ADDR[0].x], TEMP[0]
> 12: F2I TEMP[2].x, TEMP[1].xxxx
> 13: UADD TEMP[2].x, IMM[1].yyyy, TEMP[2].xxxx
> 14: UARL ADDR[0].x, TEMP[2].xxxx
> 15: UARL ADDR[0].x, TEMP[2].xxxx
> 16: MAD TEMP[0], IN[0].zzzz, CONST[ADDR[0].x], TEMP[0]
> 17: F2I TEMP[2].x, TEMP[1].xxxx
> 18: UADD TEMP[2].x, IMM[1].zzzz, TEMP[2].xxxx
> 19: UARL ADDR[0].x, TEMP[2].xxxx
> 20: UARL ADDR[0].x, TEMP[2].xxxx
> 21: ADD TEMP[2], TEMP[0], CONST[ADDR[0].x]
> 22: F2I TEMP[3].x, TEMP[1].xxxx
> 23: UADD TEMP[3].x, IMM[1].xxxx, TEMP[3].xxxx
> 24: UARL ADDR[0].x, TEMP[3].xxxx
> 25: MUL TEMP[3].xyz, IN[0].yyyy, CONST[ADDR[0].x+96]
> 26: MOV TEMP[0].xyz, TEMP[3].xyzx
> 27: F2I TEMP[3].x, TEMP[1].xxxx
> 28: UARL ADDR[0].x, TEMP[3].xxxx
> 29: MAD TEMP[3].xyz, IN[0].xxxx, CONST[ADDR[0].x+96], TEMP[0]
> 30: MOV TEMP[0].xyz, TEMP[3].xyzx
> 31: F2I TEMP[3].x, TEMP[1].xxxx
> 32: UADD TEMP[3].x, IMM[1].yyyy, TEMP[3].xxxx
> 33: UARL ADDR[0].x, TEMP[3].xxxx
> 34: MAD TEMP[3].xyz, IN[0].zzzz, CONST[ADDR[0].x+96], TEMP[0]
> 35: MOV TEMP[0].xyz, TEMP[3].xyzx
> 36: F2I TEMP[1].x, TEMP[1].xxxx
> 37: UADD TEMP[1].x, IMM[1].zzzz, TEMP[1].xxxx
> 38: UARL ADDR[0].x, TEMP[1].xxxx
> 39: ADD TEMP[1].xyz, TEMP[0], CONST[ADDR[0].x+96]
> 40: MOV TEMP[0].xyz, TEMP[1].xyzx
> 41: ADD TEMP[3].xy, TEMP[0], CONST[196].zwzw
> 42: MOV TEMP[3].xy, TEMP[3].xyxx
> 43: MUL TEMP[3].xy, TEMP[3], CONST[196]
> 44: MOV TEMP[3].xy, TEMP[3].xyxx
> 45: MUL TEMP[4].yw, TEMP[1].yyyy, CONST[193].xxzy
> 46: MOV TEMP[0].yw, TEMP[4].wyww
> 47: MAD TEMP[4].xy, TEMP[1].xxxx, CONST[192], TEMP[0].ywzw
> 48: MOV TEMP[0].xy, TEMP[4].xyxx
> 49: MAD TEMP[1].xy, TEMP[1].zzzz, CONST[194], TEMP[0]
> 50: MOV TEMP[0].xy, TEMP[1].xyxx
> 51: ADD TEMP[1].xy, TEMP[0], CONST[195]
> 52: MOV TEMP[0].xy, TEMP[1].xyxx
> 53: MAD TEMP[0].xy, TEMP[0], IMM[0].yzzw, IMM[0].zzzz
> 54: MOV TEMP[0].xy, TEMP[0].xyxx
> 55: MOV TEMP[3].zw, IMM[2].yyxy
> 56: MOV TEMP[0].zw, IMM[2].yyxy
> 57: MOV OUT[1], TEMP[3]
> 58: MOV OUT[2], TEMP[0]
> 59: MOV OUT[0], TEMP[2]
> 60: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)* byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %11 = getelementptr [17 x <16 x i8>] addrspace(2)* %1, i32 0, i32 0
>  %12 = load <16 x i8> addrspace(2)* %11, !tbaa !0
>  %13 = call float @llvm.SI.load.const(<16 x i8> %12, i32 3072)
>  %14 = call float @llvm.SI.load.const(<16 x i8> %12, i32 3076)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %12, i32 3088)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %12, i32 3092)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %12, i32 3104)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %12, i32 3108)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %12, i32 3120)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %12, i32 3124)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %12, i32 3136)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %12, i32 3140)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %12, i32 3144)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %12, i32 3148)
>  %25 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 0
>  %26 = load <16 x i8> addrspace(2)* %25, !tbaa !0
>  %27 = add i32 %5, %7
>  %28 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %26, i32 0, i32 %27)
>  %29 = extractelement <4 x float> %28, i32 0
>  %30 = extractelement <4 x float> %28, i32 1
>  %31 = extractelement <4 x float> %28, i32 2
>  %32 = getelementptr [16 x <16 x i8>] addrspace(2)* %4, i32 0, i32 1
>  %33 = load <16 x i8> addrspace(2)* %32, !tbaa !0
>  %34 = add i32 %5, %7
>  %35 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %33, i32 0, i32 %34)
>  %36 = extractelement <4 x float> %35, i32 2
>  %37 = call float @llvm.AMDIL.fraction.(float %36)
>  %38 = fsub float -0.000000e+00, %37
>  %39 = fadd float %38, %36
>  %40 = fmul float %39, 4.000000e+00
>  %41 = fptosi float %40 to i32
>  %42 = bitcast i32 %41 to float
>  %43 = bitcast float %42 to i32
>  %44 = add i32 1, %43
>  %45 = bitcast i32 %44 to float
>  %46 = bitcast float %45 to i32
>  %47 = shl i32 %46, 4
>  %48 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %47)
>  %49 = fmul float %30, %48
>  %50 = shl i32 %46, 4
>  %51 = add i32 %50, 4
>  %52 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %51)
>  %53 = fmul float %30, %52
>  %54 = shl i32 %46, 4
>  %55 = add i32 %54, 8
>  %56 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %55)
>  %57 = fmul float %30, %56
>  %58 = shl i32 %46, 4
>  %59 = add i32 %58, 12
>  %60 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %59)
>  %61 = fmul float %30, %60
>  %62 = fptosi float %40 to i32
>  %63 = bitcast i32 %62 to float
>  %64 = bitcast float %63 to i32
>  %65 = shl i32 %64, 4
>  %66 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %65)
>  %67 = fmul float %29, %66
>  %68 = fadd float %67, %49
>  %69 = shl i32 %64, 4
>  %70 = add i32 %69, 4
>  %71 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %70)
>  %72 = fmul float %29, %71
>  %73 = fadd float %72, %53
>  %74 = shl i32 %64, 4
>  %75 = add i32 %74, 8
>  %76 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %75)
>  %77 = fmul float %29, %76
>  %78 = fadd float %77, %57
>  %79 = shl i32 %64, 4
>  %80 = add i32 %79, 12
>  %81 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %80)
>  %82 = fmul float %29, %81
>  %83 = fadd float %82, %61
>  %84 = fptosi float %40 to i32
>  %85 = bitcast i32 %84 to float
>  %86 = bitcast float %85 to i32
>  %87 = add i32 2, %86
>  %88 = bitcast i32 %87 to float
>  %89 = bitcast float %88 to i32
>  %90 = shl i32 %89, 4
>  %91 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %90)
>  %92 = fmul float %31, %91
>  %93 = fadd float %92, %68
>  %94 = shl i32 %89, 4
>  %95 = add i32 %94, 4
>  %96 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %95)
>  %97 = fmul float %31, %96
>  %98 = fadd float %97, %73
>  %99 = shl i32 %89, 4
>  %100 = add i32 %99, 8
>  %101 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %100)
>  %102 = fmul float %31, %101
>  %103 = fadd float %102, %78
>  %104 = shl i32 %89, 4
>  %105 = add i32 %104, 12
>  %106 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %105)
>  %107 = fmul float %31, %106
>  %108 = fadd float %107, %83
>  %109 = fptosi float %40 to i32
>  %110 = bitcast i32 %109 to float
>  %111 = bitcast float %110 to i32
>  %112 = add i32 3, %111
>  %113 = bitcast i32 %112 to float
>  %114 = bitcast float %113 to i32
>  %115 = shl i32 %114, 4
>  %116 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %115)
>  %117 = fadd float %93, %116
>  %118 = shl i32 %114, 4
>  %119 = add i32 %118, 4
>  %120 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %119)
>  %121 = fadd float %98, %120
>  %122 = shl i32 %114, 4
>  %123 = add i32 %122, 8
>  %124 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %123)
>  %125 = fadd float %103, %124
>  %126 = shl i32 %114, 4
>  %127 = add i32 %126, 12
>  %128 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %127)
>  %129 = fadd float %108, %128
>  %130 = fptosi float %40 to i32
>  %131 = bitcast i32 %130 to float
>  %132 = bitcast float %131 to i32
>  %133 = add i32 1, %132
>  %134 = bitcast i32 %133 to float
>  %135 = bitcast float %134 to i32
>  %136 = shl i32 %135, 4
>  %137 = add i32 %136, 1536
>  %138 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %137)
>  %139 = fmul float %30, %138
>  %140 = shl i32 %135, 4
>  %141 = add i32 %140, 1540
>  %142 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %141)
>  %143 = fmul float %30, %142
>  %144 = shl i32 %135, 4
>  %145 = add i32 %144, 1544
>  %146 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %145)
>  %147 = fmul float %30, %146
>  %148 = fptosi float %40 to i32
>  %149 = bitcast i32 %148 to float
>  %150 = bitcast float %149 to i32
>  %151 = shl i32 %150, 4
>  %152 = add i32 %151, 1536
>  %153 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %152)
>  %154 = fmul float %29, %153
>  %155 = fadd float %154, %139
>  %156 = shl i32 %150, 4
>  %157 = add i32 %156, 1540
>  %158 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %157)
>  %159 = fmul float %29, %158
>  %160 = fadd float %159, %143
>  %161 = shl i32 %150, 4
>  %162 = add i32 %161, 1544
>  %163 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %162)
>  %164 = fmul float %29, %163
>  %165 = fadd float %164, %147
>  %166 = fptosi float %40 to i32
>  %167 = bitcast i32 %166 to float
>  %168 = bitcast float %167 to i32
>  %169 = add i32 2, %168
>  %170 = bitcast i32 %169 to float
>  %171 = bitcast float %170 to i32
>  %172 = shl i32 %171, 4
>  %173 = add i32 %172, 1536
>  %174 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %173)
>  %175 = fmul float %31, %174
>  %176 = fadd float %175, %155
>  %177 = shl i32 %171, 4
>  %178 = add i32 %177, 1540
>  %179 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %178)
>  %180 = fmul float %31, %179
>  %181 = fadd float %180, %160
>  %182 = shl i32 %171, 4
>  %183 = add i32 %182, 1544
>  %184 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %183)
>  %185 = fmul float %31, %184
>  %186 = fadd float %185, %165
>  %187 = fptosi float %40 to i32
>  %188 = bitcast i32 %187 to float
>  %189 = bitcast float %188 to i32
>  %190 = add i32 3, %189
>  %191 = bitcast i32 %190 to float
>  %192 = bitcast float %191 to i32
>  %193 = shl i32 %192, 4
>  %194 = add i32 %193, 1536
>  %195 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %194)
>  %196 = fadd float %176, %195
>  %197 = shl i32 %192, 4
>  %198 = add i32 %197, 1540
>  %199 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %198)
>  %200 = fadd float %181, %199
>  %201 = shl i32 %192, 4
>  %202 = add i32 %201, 1544
>  %203 = call float @llvm.SI.load.const(<16 x i8> %12, i32 %202)
>  %204 = fadd float %186, %203
>  %205 = fadd float %196, %23
>  %206 = fadd float %200, %24
>  %207 = fmul float %205, %21
>  %208 = fmul float %206, %22
>  %209 = fmul float %200, %15
>  %210 = fmul float %200, %16
>  %211 = fmul float %196, %13
>  %212 = fadd float %211, %209
>  %213 = fmul float %196, %14
>  %214 = fadd float %213, %210
>  %215 = fmul float %204, %17
>  %216 = fadd float %215, %212
>  %217 = fmul float %204, %18
>  %218 = fadd float %217, %214
>  %219 = fadd float %216, %19
>  %220 = fadd float %218, %20
>  %221 = fmul float %219, 5.000000e-01
>  %222 = fadd float %221, -5.000000e-01
>  %223 = fmul float %220, -5.000000e-01
>  %224 = fadd float %223, -5.000000e-01
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %207, float %208, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %222, float %224, float 0.000000e+00, float 1.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %117, float %121, float %125, float %129)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<16 x i8>, i32) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.fraction.(float) #2
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="1" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	v_add_i32_e32 v0, s10, v0                            ; 4A00000A
>	s_load_dwordx4 s[4:7], s[8:9], 0x0                   ; C0820900
>	s_load_dwordx4 s[8:11], s[8:9], 0x4                  ; C0840904
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[8:11], 0 idxen ; E00C2000 80020100
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_fract_f32_e32 v5, v3                               ; 7E0A4103
>	v_subrev_f32_e32 v1, v5, v3                          ; 0A020705
>	v_mul_f32_e32 v1, 4.0, v1                            ; 100202F6
>	v_cvt_i32_f32_e32 v1, v1                             ; 7E021101
>	v_lshlrev_b32_e32 v1, 4, v1                          ; 34020284
>	v_add_i32_e32 v2, 0x604, v1                          ; 4A0402FF 00000604
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                   ; C0800300
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	buffer_load_dword v2, v2, s[0:3], 0 offen            ; E0301000 80000202
>	v_add_i32_e32 v3, 0x614, v1                          ; 4A0602FF 00000614
>	buffer_load_dword v3, v3, s[0:3], 0 offen            ; E0301000 80000303
>	buffer_load_format_xyzw v[4:7], v0, s[4:7], 0 idxen  ; E00C2000 80010400
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_mul_f32_e32 v0, v3, v5                             ; 10000B03
>	v_mad_f32 v0, v4, v2, v0                             ; D2820000 04020504
>	v_add_i32_e32 v2, 0x624, v1                          ; 4A0402FF 00000624
>	buffer_load_dword v2, v2, s[0:3], 0 offen            ; E0301000 80000202
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_mad_f32 v0, v6, v2, v0                             ; D2820000 04020506
>	v_add_i32_e32 v2, 0x634, v1                          ; 4A0402FF 00000634
>	buffer_load_dword v2, v2, s[0:3], 0 offen            ; E0301000 80000202
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_add_f32_e32 v0, v2, v0                             ; 06000102
>	s_movk_i32 s4, 0xc4c                                 ; B0040C4C
>	s_buffer_load_dword s4, s[0:3], s4                   ; C2020004
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_add_f32_e32 v2, s4, v0                             ; 06040004
>	s_movk_i32 s4, 0xc44                                 ; B0040C44
>	s_buffer_load_dword s4, s[0:3], s4                   ; C2020004
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v2, s4, v2                             ; 10040404
>	v_add_i32_e32 v3, 0x600, v1                          ; 4A0602FF 00000600
>	buffer_load_dword v3, v3, s[0:3], 0 offen            ; E0301000 80000303
>	v_add_i32_e32 v8, 0x610, v1                          ; 4A1002FF 00000610
>	buffer_load_dword v8, v8, s[0:3], 0 offen            ; E0301000 80000808
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_mul_f32_e32 v8, v8, v5                             ; 10100B08
>	v_mad_f32 v3, v4, v3, v8                             ; D2820003 04220704
>	v_add_i32_e32 v8, 0x620, v1                          ; 4A1002FF 00000620
>	buffer_load_dword v8, v8, s[0:3], 0 offen            ; E0301000 80000808
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_mad_f32 v3, v6, v8, v3                             ; D2820003 040E1106
>	v_add_i32_e32 v8, 0x630, v1                          ; 4A1002FF 00000630
>	buffer_load_dword v8, v8, s[0:3], 0 offen            ; E0301000 80000808
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_add_f32_e32 v3, v8, v3                             ; 06060708
>	s_movk_i32 s4, 0xc48                                 ; B0040C48
>	s_buffer_load_dword s4, s[0:3], s4                   ; C2020004
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_add_f32_e32 v8, s4, v3                             ; 06100604
>	s_movk_i32 s4, 0xc40                                 ; B0040C40
>	s_buffer_load_dword s4, s[0:3], s4                   ; C2020004
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v8, s4, v8                             ; 10101004
>	v_mov_b32_e32 v9, 1.0                                ; 7E1202F2
>	v_mov_b32_e32 v10, 0                                 ; 7E140280
>	exp 15, 32, 0, 0, 0, v8, v2, v10, v9                 ; F800020F 090A0208
>	s_movk_i32 s4, 0xc10                                 ; B0040C10
>	s_buffer_load_dword s4, s[0:3], s4                   ; C2020004
>	s_waitcnt expcnt(0) lgkmcnt(0)                       ; BF8C000F
>	v_mul_f32_e32 v2, s4, v0                             ; 10040004
>	s_movk_i32 s4, 0xc00                                 ; B0040C00
>	s_buffer_load_dword s4, s[0:3], s4                   ; C2020004
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v2, v3, s4, v2                             ; D2820002 04080903
>	v_add_i32_e32 v8, 0x608, v1                          ; 4A1002FF 00000608
>	buffer_load_dword v8, v8, s[0:3], 0 offen            ; E0301000 80000808
>	v_add_i32_e32 v11, 0x618, v1                         ; 4A1602FF 00000618
>	buffer_load_dword v11, v11, s[0:3], 0 offen          ; E0301000 80000B0B
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_mul_f32_e32 v11, v11, v5                           ; 10160B0B
>	v_mad_f32 v8, v4, v8, v11                            ; D2820008 042E1104
>	v_add_i32_e32 v11, 0x628, v1                         ; 4A1602FF 00000628
>	buffer_load_dword v11, v11, s[0:3], 0 offen          ; E0301000 80000B0B
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_mad_f32 v8, v6, v11, v8                            ; D2820008 04221706
>	v_add_i32_e32 v11, 0x638, v1                         ; 4A1602FF 00000638
>	buffer_load_dword v11, v11, s[0:3], 0 offen          ; E0301000 80000B0B
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_add_f32_e32 v8, v11, v8                            ; 0610110B
>	s_movk_i32 s4, 0xc20                                 ; B0040C20
>	s_buffer_load_dword s4, s[0:3], s4                   ; C2020004
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v2, v8, s4, v2                             ; D2820002 04080908
>	s_movk_i32 s4, 0xc30                                 ; B0040C30
>	s_buffer_load_dword s4, s[0:3], s4                   ; C2020004
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_add_f32_e32 v2, s4, v2                             ; 06040404
>	v_mad_f32 v2, 0.5, v2, -0.5                          ; D2820002 03C604F0
>	s_movk_i32 s4, 0xc14                                 ; B0040C14
>	s_buffer_load_dword s4, s[0:3], s4                   ; C2020004
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mul_f32_e32 v0, s4, v0                             ; 10000004
>	s_movk_i32 s4, 0xc04                                 ; B0040C04
>	s_buffer_load_dword s4, s[0:3], s4                   ; C2020004
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v0, v3, s4, v0                             ; D2820000 04000903
>	s_movk_i32 s4, 0xc24                                 ; B0040C24
>	s_buffer_load_dword s4, s[0:3], s4                   ; C2020004
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_mad_f32 v0, v8, s4, v0                             ; D2820000 04000908
>	s_movk_i32 s4, 0xc34                                 ; B0040C34
>	s_buffer_load_dword s4, s[0:3], s4                   ; C2020004
>	s_waitcnt lgkmcnt(0)                                 ; BF8C007F
>	v_add_f32_e32 v0, s4, v0                             ; 06000004
>	v_mad_f32 v0, -0.5, v0, -0.5                         ; D2820000 03C600F1
>	exp 15, 33, 0, 0, 0, v2, v0, v10, v9                 ; F800021F 090A0002
>	s_waitcnt expcnt(0)                                  ; BF8C070F
>	buffer_load_dword v0, v1, s[0:3], 0 offen            ; E0301000 80000001
>	v_add_i32_e32 v2, 16, v1                             ; 4A040290
>	buffer_load_dword v3, v2, s[0:3], 0 offen            ; E0301000 80000302
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_mul_f32_e32 v3, v3, v5                             ; 10060B03
>	v_mad_f32 v0, v4, v0, v3                             ; D2820000 040E0104
>	v_add_i32_e32 v3, 32, v1                             ; 4A0602A0
>	buffer_load_dword v8, v3, s[0:3], 0 offen            ; E0301000 80000803
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_mad_f32 v0, v6, v8, v0                             ; D2820000 04021106
>	v_add_i32_e32 v8, 48, v1                             ; 4A1002B0
>	buffer_load_dword v9, v8, s[0:3], 0 offen            ; E0301000 80000908
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_add_f32_e32 v0, v9, v0                             ; 06000109
>	v_or_b32_e32 v9, 12, v1                              ; 3812028C
>	buffer_load_dword v9, v9, s[0:3], 0 offen            ; E0301000 80000909
>	v_or_b32_e32 v10, 12, v2                             ; 3814048C
>	buffer_load_dword v10, v10, s[0:3], 0 offen          ; E0301000 80000A0A
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_mul_f32_e32 v10, v10, v5                           ; 10140B0A
>	v_mad_f32 v9, v4, v9, v10                            ; D2820009 042A1304
>	v_or_b32_e32 v10, 12, v3                             ; 3814068C
>	buffer_load_dword v10, v10, s[0:3], 0 offen          ; E0301000 80000A0A
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_mad_f32 v9, v6, v10, v9                            ; D2820009 04261506
>	v_or_b32_e32 v10, 12, v8                             ; 3814108C
>	buffer_load_dword v10, v10, s[0:3], 0 offen          ; E0301000 80000A0A
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_add_f32_e32 v9, v10, v9                            ; 0612130A
>	v_or_b32_e32 v10, 8, v1                              ; 38140288
>	buffer_load_dword v10, v10, s[0:3], 0 offen          ; E0301000 80000A0A
>	v_or_b32_e32 v11, 8, v2                              ; 38160488
>	buffer_load_dword v11, v11, s[0:3], 0 offen          ; E0301000 80000B0B
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_mul_f32_e32 v11, v11, v5                           ; 10160B0B
>	v_mad_f32 v10, v4, v10, v11                          ; D282000A 042E1504
>	v_or_b32_e32 v11, 8, v3                              ; 38160688
>	buffer_load_dword v11, v11, s[0:3], 0 offen          ; E0301000 80000B0B
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_mad_f32 v10, v6, v11, v10                          ; D282000A 042A1706
>	v_or_b32_e32 v11, 8, v8                              ; 38161088
>	buffer_load_dword v11, v11, s[0:3], 0 offen          ; E0301000 80000B0B
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_add_f32_e32 v10, v11, v10                          ; 0614150B
>	v_or_b32_e32 v1, 4, v1                               ; 38020284
>	buffer_load_dword v1, v1, s[0:3], 0 offen            ; E0301000 80000101
>	v_or_b32_e32 v2, 4, v2                               ; 38040484
>	buffer_load_dword v2, v2, s[0:3], 0 offen            ; E0301000 80000202
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_mul_f32_e32 v2, v2, v5                             ; 10040B02
>	v_mad_f32 v1, v4, v1, v2                             ; D2820001 040A0304
>	v_or_b32_e32 v2, 4, v3                               ; 38040684
>	buffer_load_dword v2, v2, s[0:3], 0 offen            ; E0301000 80000202
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_mad_f32 v1, v6, v2, v1                             ; D2820001 04060506
>	v_or_b32_e32 v2, 4, v8                               ; 38041084
>	buffer_load_dword v2, v2, s[0:3], 0 offen            ; E0301000 80000202
>	s_waitcnt vmcnt(0)                                   ; BF8C0770
>	v_add_f32_e32 v1, v2, v1                             ; 06020302
>	exp 15, 12, 0, 1, 0, v0, v1, v10, v9                 ; F80008CF 090A0100
>	s_endpgm                                             ; BF810000
>
>SGPRS: 16
>VGPRS: 12
>Code Size: 960 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>FRAG
>DCL IN[0], GENERIC[9], PERSPECTIVE
>DCL IN[1], GENERIC[10], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SAMP[1]
>DCL TEMP[0..3], LOCAL
>IMM[0] FLT32 {    1.0000,    -1.0000,     0.0000,    -0.0100}
>  0: MAD TEMP[0].xy, IN[1], IMM[0].xyxx, IMM[0].zxzz
>  1: MOV TEMP[0].xy, TEMP[0].xyyy
>  2: TEX TEMP[0].y, TEMP[0], SAMP[1], 2D
>  3: ADD TEMP[1].x, TEMP[0].yyyy, IMM[0].wwww
>  4: FSGE TEMP[2].x, TEMP[1].xxxx, IMM[0].zzzz
>  5: UIF TEMP[2].xxxx :0
>  6:   MOV TEMP[2].x, IMM[0].zzzz
>  7: ELSE :0
>  8:   MOV TEMP[2].x, IMM[0].yyyy
>  9: ENDIF
> 10: MOV TEMP[2].x, TEMP[2].xxxx
> 11: FSGE TEMP[3].x, TEMP[1].xxxx, IMM[0].zzzz
> 12: UIF TEMP[3].xxxx :0
> 13:   MOV TEMP[3].x, IMM[0].zzzz
> 14: ELSE :0
> 15:   MOV TEMP[3].x, IMM[0].yyyy
> 16: ENDIF
> 17: MOV TEMP[2].y, TEMP[3].xxxx
> 18: FSGE TEMP[3].x, TEMP[1].xxxx, IMM[0].zzzz
> 19: UIF TEMP[3].xxxx :0
> 20:   MOV TEMP[3].x, IMM[0].zzzz
> 21: ELSE :0
> 22:   MOV TEMP[3].x, IMM[0].yyyy
> 23: ENDIF
> 24: MOV TEMP[2].z, TEMP[3].xxxx
> 25: FSGE TEMP[1].x, TEMP[1].xxxx, IMM[0].zzzz
> 26: UIF TEMP[1].xxxx :0
> 27: ELSE :0
> 28: ENDIF
> 29: FSLT TEMP[1].xyz, TEMP[2].xyzz, IMM[0].zzzz
> 30: OR TEMP[2].x, TEMP[1].xxxx, TEMP[1].zzzz
> 31: OR TEMP[1].x, TEMP[2].xxxx, TEMP[1].yyyy
> 32: AND TEMP[1].x, TEMP[1].xxxx, IMM[0].xxxx
> 33: KILL_IF -TEMP[1].xxxx
> 34: MOV TEMP[1].xy, IN[0].xyyy
> 35: TEX TEMP[1], TEMP[1], SAMP[0], 2D
> 36: MUL TEMP[0].w, TEMP[0].yyyy, TEMP[1].wwww
> 37: MOV TEMP[0].w, TEMP[0].wwww
> 38: MOV TEMP[0].xyz, TEMP[1].xyzx
> 39: MOV OUT[0], TEMP[0]
> 40: END
>; ModuleID = 'tgsi'
>
>define void @main([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) #0 {
>main_body:
>  %22 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 0
>  %23 = load <8 x i32> addrspace(2)* %22, !tbaa !0
>  %24 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 0
>  %25 = load <4 x i32> addrspace(2)* %24, !tbaa !0
>  %26 = getelementptr [34 x <8 x i32>] addrspace(2)* %3, i32 0, i32 1
>  %27 = load <8 x i32> addrspace(2)* %26, !tbaa !0
>  %28 = getelementptr [17 x <4 x i32>] addrspace(2)* %2, i32 0, i32 1
>  %29 = load <4 x i32> addrspace(2)* %28, !tbaa !0
>  %30 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %5, <2 x i32> %7)
>  %31 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
>  %32 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %5, <2 x i32> %7)
>  %33 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %5, <2 x i32> %7)
>  %34 = fmul float %32, 1.000000e+00
>  %35 = fadd float %34, 0.000000e+00
>  %36 = fmul float %33, -1.000000e+00
>  %37 = fadd float %36, 1.000000e+00
>  %38 = bitcast float %35 to i32
>  %39 = bitcast float %37 to i32
>  %40 = insertelement <2 x i32> undef, i32 %38, i32 0
>  %41 = insertelement <2 x i32> %40, i32 %39, i32 1
>  %42 = bitcast <8 x i32> %27 to <32 x i8>
>  %43 = bitcast <4 x i32> %29 to <16 x i8>
>  %44 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %41, <32 x i8> %42, <16 x i8> %43, i32 2)
>  %45 = extractelement <4 x float> %44, i32 1
>  %46 = fadd float %45, 0xBF847AE140000000
>  %47 = fcmp oge float %46, 0.000000e+00
>  %48 = sext i1 %47 to i32
>  %49 = bitcast i32 %48 to float
>  %50 = bitcast float %49 to i32
>  %51 = icmp ne i32 %50, 0
>  %. = select i1 %51, float 0.000000e+00, float -1.000000e+00
>  %52 = fcmp oge float %46, 0.000000e+00
>  %53 = sext i1 %52 to i32
>  %54 = bitcast i32 %53 to float
>  %55 = bitcast float %54 to i32
>  %56 = icmp ne i32 %55, 0
>  %temp12.0 = select i1 %56, float 0.000000e+00, float -1.000000e+00
>  %57 = fcmp oge float %46, 0.000000e+00
>  %58 = sext i1 %57 to i32
>  %59 = bitcast i32 %58 to float
>  %60 = bitcast float %59 to i32
>  %61 = icmp ne i32 %60, 0
>  %.25 = select i1 %61, float 0.000000e+00, float -1.000000e+00
>  %62 = fcmp oge float %46, 0.000000e+00
>  %63 = sext i1 %62 to i32
>  %64 = bitcast i32 %63 to float
>  %65 = bitcast float %64 to i32
>  %66 = icmp ne i32 %65, 0
>  %67 = fcmp olt float %., 0.000000e+00
>  %68 = sext i1 %67 to i32
>  %69 = fcmp olt float %temp12.0, 0.000000e+00
>  %70 = sext i1 %69 to i32
>  %71 = fcmp olt float %.25, 0.000000e+00
>  %72 = sext i1 %71 to i32
>  %73 = bitcast i32 %68 to float
>  %74 = bitcast i32 %70 to float
>  %75 = bitcast i32 %72 to float
>  %76 = bitcast float %73 to i32
>  %77 = bitcast float %75 to i32
>  %78 = or i32 %76, %77
>  %79 = bitcast i32 %78 to float
>  %80 = bitcast float %79 to i32
>  %81 = bitcast float %74 to i32
>  %82 = or i32 %80, %81
>  %83 = bitcast i32 %82 to float
>  %84 = bitcast float %83 to i32
>  %85 = and i32 %84, 1065353216
>  %86 = bitcast i32 %85 to float
>  %87 = fsub float -0.000000e+00, %86
>  %88 = fsub float -0.000000e+00, %86
>  %89 = fsub float -0.000000e+00, %86
>  %90 = fsub float -0.000000e+00, %86
>  call void @llvm.AMDGPU.kill(float %87)
>  call void @llvm.AMDGPU.kill(float %88)
>  call void @llvm.AMDGPU.kill(float %89)
>  call void @llvm.AMDGPU.kill(float %90)
>  %91 = bitcast float %30 to i32
>  %92 = bitcast float %31 to i32
>  %93 = insertelement <2 x i32> undef, i32 %91, i32 0
>  %94 = insertelement <2 x i32> %93, i32 %92, i32 1
>  %95 = bitcast <8 x i32> %23 to <32 x i8>
>  %96 = bitcast <4 x i32> %25 to <16 x i8>
>  %97 = call <4 x float> @llvm.SI.sample.v2i32(<2 x i32> %94, <32 x i8> %95, <16 x i8> %96, i32 2)
>  %98 = extractelement <4 x float> %97, i32 0
>  %99 = extractelement <4 x float> %97, i32 1
>  %100 = extractelement <4 x float> %97, i32 2
>  %101 = extractelement <4 x float> %97, i32 3
>  %102 = fmul float %45, %101
>  %103 = call i32 @llvm.SI.packf16(float %98, float %99)
>  %104 = bitcast i32 %103 to float
>  %105 = call i32 @llvm.SI.packf16(float %100, float %102)
>  %106 = bitcast i32 %105 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %104, float %106, float %104, float %106)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.sample.v2i32(<2 x i32>, <32 x i8>, <16 x i8>, i32) #1
>
>declare void @llvm.AMDGPU.kill(float)
>
>; Function Attrs: nounwind readnone
>declare i32 @llvm.SI.packf16(float, float) #1
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>attributes #1 = { nounwind readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                   ; BEFE0A7E
>	s_mov_b32 m0, s9                                                       ; BEFC0309
>	v_interp_p1_f32 v2, v0, 1, 1, [m0]                                     ; C8080500
>	v_interp_p2_f32 v2, [v2], v1, 1, 1, [m0]                               ; C8090501
>	v_sub_f32_e32 v3, 1.0, v2                                              ; 080604F2
>	v_interp_p1_f32 v4, v0, 0, 1, [m0]                                     ; C8100400
>	v_interp_p2_f32 v4, [v4], v1, 0, 1, [m0]                               ; C8110401
>	v_add_f32_e32 v2, 0, v4                                                ; 06040880
>	s_load_dwordx4 s[0:3], s[4:5], 0x0                                     ; C0800500
>	s_load_dwordx4 s[8:11], s[4:5], 0x4                                    ; C0840504
>	s_load_dwordx8 s[12:19], s[6:7], 0x0                                   ; C0C60700
>	s_load_dwordx8 s[20:27], s[6:7], 0x8                                   ; C0CA0708
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	image_sample v2, 2, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[20:27], s[8:11]     ; F0800200 00450202
>	v_mov_b32_e32 v3, 0xbc23d70a                                           ; 7E0602FF BC23D70A
>	s_waitcnt vmcnt(0)                                                     ; BF8C0770
>	v_add_f32_e32 v3, v2, v3                                               ; 06060702
>	v_cmp_ge_f32_e64 s[4:5], v3, 0                                         ; D00C0004 00010103
>	v_cndmask_b32_e64 v3, 0, -1, s[4:5]                                    ; D2000803 00118280
>	v_cmp_ne_i32_e64 s[4:5], v3, 0                                         ; D10A0004 00010103
>	v_cndmask_b32_e64 v3, -1.0, 0, s[4:5]                                  ; D2000803 181100F3
>	v_cmp_lt_f32_e64 s[4:5], v3, 0                                         ; D0020004 00010103
>	v_cndmask_b32_e64 v3, 0, -1, s[4:5]                                    ; D2000803 00118280
>	v_and_b32_e32 v3, 1.0, v3                                              ; 360606F2
>	v_xor_b32_e32 v3, 0x80000000, v3                                       ; 3A0606FF 80000000
>	v_cmpx_le_f32_e32 vcc, 0, v3                                           ; 7C260680
>	v_cmpx_le_f32_e32 vcc, 0, v3                                           ; 7C260680
>	v_cmpx_le_f32_e32 vcc, 0, v3                                           ; 7C260680
>	v_cmpx_le_f32_e32 vcc, 0, v3                                           ; 7C260680
>	v_interp_p1_f32 v4, v0, 1, 0, [m0]                                     ; C8100100
>	v_interp_p2_f32 v4, [v4], v1, 1, 0, [m0]                               ; C8110101
>	v_interp_p1_f32 v3, v0, 0, 0, [m0]                                     ; C80C0000
>	v_interp_p2_f32 v3, [v3], v1, 0, 0, [m0]                               ; C80D0001
>	image_sample v[3:6], 15, 0, 0, 0, 0, 0, 0, 0, v[3:4], s[12:19], s[0:3] ; F0800F00 00030303
>	s_waitcnt vmcnt(0)                                                     ; BF8C0770
>	v_cvt_pkrtz_f16_f32_e32 v0, v3, v4                                     ; 5E000903
>	v_mul_f32_e32 v1, v6, v2                                               ; 10020506
>	v_cvt_pkrtz_f16_f32_e32 v1, v5, v1                                     ; 5E020305
>	exp 15, 0, 1, 1, 1, v0, v1, v0, v1                                     ; F8001C0F 01000100
>	s_endpgm                                                               ; BF810000
>
>SGPRS: 32
>VGPRS: 8
>Code Size: 204 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave

Actions: View

Attachments on bug 89944: 114935 | 114936 | 114937 | 115007 | 115008

Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.