Attachment 120476 Details for Bug 93360 – Output of R600_DEBUG=vs,gs,ps

Output of R600_DEBUG=vs,gs,ps

csgo_stderr.txt (text/plain), 1.32 MB, created by Tobias Droste on 2015-12-12 03:23:08 UTC

(hide)

Description:

Filename:

MIME Type:

Creator: Tobias Droste

Created: 2015-12-12 03:23:08 UTC

Size: 1.32 MB

patch

obsolete

>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>  as_ls = 0
>  export_prim_id = 0
>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([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = add i32 %5, %8
>  %15 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %13, i32 0, i32 %14)
>  %16 = extractelement <4 x float> %15, i32 0
>  %17 = extractelement <4 x float> %15, i32 1
>  %18 = extractelement <4 x float> %15, i32 2
>  %19 = extractelement <4 x float> %15, i32 3
>  %20 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 1
>  %21 = load <16 x i8>, <16 x i8> addrspace(2)* %20, align 16, !tbaa !0
>  %22 = add i32 %5, %8
>  %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 0
>  %25 = extractelement <4 x float> %23, i32 1
>  %26 = extractelement <4 x float> %23, i32 2
>  %27 = extractelement <4 x float> %23, i32 3
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %24, float %25, float %26, float %27)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %16, float %17, float %18, float %19)
>  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:
>
>	s_load_dwordx4 s[0:3], s[8:9], 0x0                  ; C0800900
>	s_load_dwordx4 s[4:7], s[8:9], 0x4                  ; C0820904
>	v_add_i32_e32 v0, vcc, s10, v0                      ; 4A00000A
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[0:3], 0 idxen ; E00C2000 80000100
>	buffer_load_format_xyzw v[5:8], v0, s[4:7], 0 idxen ; E00C2000 80010500
>	s_waitcnt vmcnt(0)                                  ; BF8C0770
>	exp 15, 32, 0, 0, 0, v5, v6, v7, v8                 ; F800020F 08070605
>	exp 15, 12, 0, 1, 0, v1, v2, v3, v4                 ; F80008CF 04030201
>	s_endpgm                                            ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 16
>VGPRS: 12
>Code Size: 56 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
>  clamp_color = 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([9 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, 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:
>  %23 = call float @llvm.SI.fs.constant(i32 0, i32 0, i32 %6)
>  %24 = call float @llvm.SI.fs.constant(i32 1, i32 0, i32 %6)
>  %25 = call float @llvm.SI.fs.constant(i32 2, i32 0, i32 %6)
>  %26 = call float @llvm.SI.fs.constant(i32 3, i32 0, i32 %6)
>  %27 = call i32 @llvm.SI.packf16(float %23, float %24)
>  %28 = bitcast i32 %27 to float
>  %29 = call i32 @llvm.SI.packf16(float %25, float %26)
>  %30 = bitcast i32 %29 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %28, float %30, float %28, float %30)
>  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_mov_b32 m0, s10                   ; BEFC030A
>	v_interp_mov_f32 v0, P0, 0, 0, [m0] ; C8020002
>	v_interp_mov_f32 v1, P0, 1, 0, [m0] ; C8060102
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v1  ; 5E000300
>	v_interp_mov_f32 v1, P0, 2, 0, [m0] ; C8060202
>	v_interp_mov_f32 v2, P0, 3, 0, [m0] ; C80A0302
>	v_cvt_pkrtz_f16_f32_e32 v1, v1, v2  ; 5E020501
>	exp 15, 0, 1, 1, 1, v0, v1, v0, v1  ; F8001C0F 01000100
>	s_endpgm                            ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 16
>VGPRS: 4
>Code Size: 40 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>********************
>Using breakpad crash handler
>Setting breakpad minidump AppID = 730
>Forcing breakpad minidump interfaces to load
>Looking up breakpad interfaces from steamclient
>Calling BreakpadMiniDumpSystemInit
>Looking up breakpad interfaces from steamclient
>Calling BreakpadMiniDumpSystemInit
>Steam_SetMinidumpSteamID:  Caching Steam ID:  76561197960306978 [API loaded yes]
>Steam_SetMinidumpSteamID:  Setting Steam ID:  76561197960306978
>SHADER KEY
>  export_16bpc = 0x3
>  last_cbuf = 0
>  color_two_side = 0
>  alpha_func = 7
>  alpha_to_one = 0
>  poly_stipple = 0
>  clamp_color = 0
>FRAG
>DCL IN[0], GENERIC[0], LINEAR
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SVIEW[0], 2D, FLOAT
>  0: TEX OUT[0], IN[0], SAMP[0], 2D
>  1: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %24 = load <8 x i32>, <8 x i32> addrspace(2)* %23, align 32, !tbaa !0
>  %25 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %26 = load <4 x i32>, <4 x i32> addrspace(2)* %25, align 16, !tbaa !0
>  %27 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %12)
>  %28 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %12)
>  %29 = bitcast float %27 to i32
>  %30 = bitcast float %28 to i32
>  %31 = insertelement <2 x i32> undef, i32 %29, i32 0
>  %32 = insertelement <2 x i32> %31, i32 %30, i32 1
>  %33 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %32, <8 x i32> %24, <4 x i32> %26, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %34 = extractelement <4 x float> %33, i32 0
>  %35 = extractelement <4 x float> %33, i32 1
>  %36 = extractelement <4 x float> %33, i32 2
>  %37 = extractelement <4 x float> %33, i32 3
>  %38 = call i32 @llvm.SI.packf16(float %34, float %35)
>  %39 = bitcast i32 %38 to float
>  %40 = call i32 @llvm.SI.packf16(float %36, float %37)
>  %41 = bitcast i32 %40 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %39, float %41, float %39, float %41)
>  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.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, 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 }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                   ; BEFE0A7E
>	s_load_dwordx8 s[12:19], s[6:7], 0x0                                   ; C0C60700
>	s_load_dwordx4 s[0:3], s[4:5], 0x0                                     ; C0800500
>	s_mov_b32 m0, s10                                                      ; BEFC030A
>	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
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	image_sample v[0:3], 15, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[12:19], s[0:3] ; F0800F00 00030002
>	s_waitcnt vmcnt(0)                                                     ; BF8C0770
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v1                                     ; 5E000300
>	v_cvt_pkrtz_f16_f32_e32 v1, v2, v3                                     ; 5E020702
>	exp 15, 0, 1, 1, 1, v0, v1, v0, v1                                     ; F8001C0F 01000100
>	s_endpgm                                                               ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 24
>VGPRS: 4
>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
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[0]
>DCL CONST[0..3]
>DCL TEMP[0], LOCAL
>IMM[0] FLT32 {    0.0000,     0.0000,     0.0000,     0.0000}
>  0: MUL TEMP[0], CONST[0], IN[0].xxxx
>  1: MAD TEMP[0], CONST[1], IN[0].yyyy, TEMP[0]
>  2: MAD TEMP[0], CONST[2], IN[0].zzzz, TEMP[0]
>  3: MAD TEMP[0], CONST[3], IN[0].wwww, TEMP[0]
>  4: MOV OUT[1], IMM[0].xxxx
>  5: MOV OUT[0], TEMP[0]
>  6: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = call float @llvm.SI.load.const(<16 x i8> %13, i32 0)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %13, i32 4)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %13, i32 8)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %13, i32 12)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %13, i32 16)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %13, i32 20)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %13, i32 24)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %13, i32 28)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %13, i32 32)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %13, i32 36)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %13, i32 40)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %13, i32 44)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %13, i32 48)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %13, i32 52)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %13, i32 56)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %13, i32 60)
>  %30 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %31 = load <16 x i8>, <16 x i8> addrspace(2)* %30, align 16, !tbaa !0
>  %32 = add i32 %5, %8
>  %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 = extractelement <4 x float> %33, i32 2
>  %37 = extractelement <4 x float> %33, i32 3
>  %38 = fmul float %14, %34
>  %39 = fmul float %15, %34
>  %40 = fmul float %16, %34
>  %41 = fmul float %17, %34
>  %42 = fmul float %18, %35
>  %43 = fadd float %42, %38
>  %44 = fmul float %19, %35
>  %45 = fadd float %44, %39
>  %46 = fmul float %20, %35
>  %47 = fadd float %46, %40
>  %48 = fmul float %21, %35
>  %49 = fadd float %48, %41
>  %50 = fmul float %22, %36
>  %51 = fadd float %50, %43
>  %52 = fmul float %23, %36
>  %53 = fadd float %52, %45
>  %54 = fmul float %24, %36
>  %55 = fadd float %54, %47
>  %56 = fmul float %25, %36
>  %57 = fadd float %56, %49
>  %58 = fmul float %26, %37
>  %59 = fadd float %58, %51
>  %60 = fmul float %27, %37
>  %61 = fadd float %60, %53
>  %62 = fmul float %28, %37
>  %63 = fadd float %62, %55
>  %64 = fmul float %29, %37
>  %65 = fadd float %64, %57
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %59, float %61, float %63, float %65)
>  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[4:7], s[8:9], 0x0                  ; C0820900
>	v_add_i32_e32 v0, vcc, s10, v0                      ; 4A00000A
>	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 lgkmcnt(0)                                ; BF8C007F
>	s_buffer_load_dword s4, s[0:3], 0x0                 ; C2020100
>	s_buffer_load_dword s5, s[0:3], 0x1                 ; C2028101
>	s_buffer_load_dword s6, s[0:3], 0x2                 ; C2030102
>	s_buffer_load_dword s7, s[0:3], 0x3                 ; C2038103
>	s_buffer_load_dword s8, s[0:3], 0x4                 ; C2040104
>	s_buffer_load_dword s9, s[0:3], 0x5                 ; C2048105
>	s_buffer_load_dword s10, s[0:3], 0x6                ; C2050106
>	s_buffer_load_dword s11, s[0:3], 0x7                ; C2058107
>	s_buffer_load_dword s12, s[0:3], 0x8                ; C2060108
>	s_buffer_load_dword s13, s[0:3], 0x9                ; C2068109
>	s_buffer_load_dword s14, s[0:3], 0xa                ; C207010A
>	s_buffer_load_dword s15, s[0:3], 0xb                ; C207810B
>	s_buffer_load_dword s16, s[0:3], 0xc                ; C208010C
>	s_buffer_load_dword s17, s[0:3], 0xd                ; C208810D
>	s_buffer_load_dword s18, s[0:3], 0xe                ; C209010E
>	s_buffer_load_dword s0, s[0:3], 0xf                 ; C200010F
>	s_waitcnt vmcnt(0) lgkmcnt(0)                       ; BF8C0070
>	v_mul_f32_e32 v4, s4, v0                            ; 10080004
>	v_mul_f32_e32 v5, s5, v0                            ; 100A0005
>	v_mul_f32_e32 v6, s6, v0                            ; 100C0006
>	v_mul_f32_e32 v0, s7, v0                            ; 10000007
>	v_mac_f32_e32 v4, s8, v1                            ; 3E080208
>	v_mac_f32_e32 v5, s9, v1                            ; 3E0A0209
>	v_mac_f32_e32 v6, s10, v1                           ; 3E0C020A
>	v_mac_f32_e32 v0, s11, v1                           ; 3E00020B
>	v_mac_f32_e32 v4, s12, v2                           ; 3E08040C
>	v_mac_f32_e32 v5, s13, v2                           ; 3E0A040D
>	v_mac_f32_e32 v6, s14, v2                           ; 3E0C040E
>	v_mac_f32_e32 v0, s15, v2                           ; 3E00040F
>	v_mac_f32_e32 v4, s16, v3                           ; 3E080610
>	v_mac_f32_e32 v5, s17, v3                           ; 3E0A0611
>	v_mac_f32_e32 v6, s18, v3                           ; 3E0C0612
>	v_mac_f32_e32 v0, s0, v3                            ; 3E000600
>	v_mov_b32_e32 v1, 0                                 ; 7E020280
>	exp 15, 32, 0, 0, 0, v1, v1, v1, v1                 ; F800020F 01010101
>	exp 15, 12, 0, 1, 0, v4, v5, v6, v0                 ; F80008CF 00060504
>	s_endpgm                                            ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 24
>VGPRS: 8
>Code Size: 184 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
>  clamp_color = 0
>FRAG
>PROPERTY FS_COLOR0_WRITES_ALL_CBUFS 1
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SVIEW[0], 2D, FLOAT
>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([9 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, 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:
>  %23 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %24 = load <8 x i32>, <8 x i32> addrspace(2)* %23, align 32, !tbaa !0
>  %25 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %26 = load <4 x i32>, <4 x i32> addrspace(2)* %25, align 16, !tbaa !0
>  %27 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %8)
>  %28 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %8)
>  %29 = bitcast float %27 to i32
>  %30 = bitcast float %28 to i32
>  %31 = insertelement <2 x i32> undef, i32 %29, i32 0
>  %32 = insertelement <2 x i32> %31, i32 %30, i32 1
>  %33 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %32, <8 x i32> %24, <4 x i32> %26, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %34 = extractelement <4 x float> %33, i32 0
>  %35 = extractelement <4 x float> %33, i32 1
>  %36 = extractelement <4 x float> %33, i32 2
>  %37 = extractelement <4 x float> %33, i32 3
>  %38 = call i32 @llvm.SI.packf16(float %34, float %35)
>  %39 = bitcast i32 %38 to float
>  %40 = call i32 @llvm.SI.packf16(float %36, float %37)
>  %41 = bitcast i32 %40 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %39, float %41, float %39, float %41)
>  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.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, 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 }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                   ; BEFE0A7E
>	s_load_dwordx8 s[12:19], s[6:7], 0x0                                   ; C0C60700
>	s_load_dwordx4 s[0:3], s[4:5], 0x0                                     ; C0800500
>	s_mov_b32 m0, s10                                                      ; BEFC030A
>	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
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	image_sample v[0:3], 15, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[12:19], s[0:3] ; F0800F00 00030002
>	s_waitcnt vmcnt(0)                                                     ; BF8C0770
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v1                                     ; 5E000300
>	v_cvt_pkrtz_f16_f32_e32 v1, v2, v3                                     ; 5E020702
>	exp 15, 0, 1, 1, 1, v0, v1, v0, v1                                     ; F8001C0F 01000100
>	s_endpgm                                                               ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 24
>VGPRS: 4
>Code Size: 68 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
>  clamp_color = 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([9 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, 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:
>  %23 = call float @llvm.SI.fs.constant(i32 0, i32 0, i32 %6)
>  %24 = call float @llvm.SI.fs.constant(i32 1, i32 0, i32 %6)
>  %25 = call float @llvm.SI.fs.constant(i32 2, i32 0, i32 %6)
>  %26 = call float @llvm.SI.fs.constant(i32 3, i32 0, i32 %6)
>  %27 = call i32 @llvm.SI.packf16(float %23, float %24)
>  %28 = bitcast i32 %27 to float
>  %29 = call i32 @llvm.SI.packf16(float %25, float %26)
>  %30 = bitcast i32 %29 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %28, float %30, float %28, float %30)
>  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_mov_b32 m0, s10                   ; BEFC030A
>	v_interp_mov_f32 v0, P0, 0, 0, [m0] ; C8020002
>	v_interp_mov_f32 v1, P0, 1, 0, [m0] ; C8060102
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v1  ; 5E000300
>	v_interp_mov_f32 v1, P0, 2, 0, [m0] ; C8060202
>	v_interp_mov_f32 v2, P0, 3, 0, [m0] ; C80A0302
>	v_cvt_pkrtz_f16_f32_e32 v1, v1, v2  ; 5E020501
>	exp 15, 0, 1, 1, 1, v0, v1, v0, v1  ; F8001C0F 01000100
>	s_endpgm                            ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 16
>VGPRS: 4
>Code Size: 40 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
>  clamp_color = 0
>FRAG
>DCL IN[0], GENERIC[0], LINEAR
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SVIEW[0], 2D, UINT
>  0: TEX OUT[0], IN[0], SAMP[0], 2D
>  1: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %24 = load <8 x i32>, <8 x i32> addrspace(2)* %23, align 32, !tbaa !0
>  %25 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %26 = load <4 x i32>, <4 x i32> addrspace(2)* %25, align 16, !tbaa !0
>  %27 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %12)
>  %28 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %12)
>  %29 = bitcast float %27 to i32
>  %30 = bitcast float %28 to i32
>  %31 = insertelement <2 x i32> undef, i32 %29, i32 0
>  %32 = insertelement <2 x i32> %31, i32 %30, i32 1
>  %33 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %32, <8 x i32> %24, <4 x i32> %26, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %34 = extractelement <4 x float> %33, i32 0
>  %35 = extractelement <4 x float> %33, i32 1
>  %36 = extractelement <4 x float> %33, i32 2
>  %37 = extractelement <4 x float> %33, i32 3
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 0, float %34, float %35, float %36, float %37)
>  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.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, 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_load_dwordx8 s[12:19], s[6:7], 0x0                                   ; C0C60700
>	s_load_dwordx4 s[0:3], s[4:5], 0x0                                     ; C0800500
>	s_mov_b32 m0, s10                                                      ; BEFC030A
>	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
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	image_sample v[0:3], 15, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[12:19], s[0:3] ; F0800F00 00030002
>	s_waitcnt vmcnt(0)                                                     ; BF8C0770
>	exp 15, 0, 0, 1, 1, v0, v1, v2, v3                                     ; F800180F 03020100
>	s_endpgm                                                               ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 24
>VGPRS: 4
>Code Size: 60 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
>  clamp_color = 0
>FRAG
>DCL IN[0], GENERIC[0], LINEAR
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SVIEW[0], CUBE, FLOAT
>  0: TEX OUT[0], IN[0], SAMP[0], CUBE
>  1: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %24 = load <8 x i32>, <8 x i32> addrspace(2)* %23, align 32, !tbaa !0
>  %25 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %26 = load <4 x i32>, <4 x i32> addrspace(2)* %25, align 16, !tbaa !0
>  %27 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %12)
>  %28 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %12)
>  %29 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %6, <2 x i32> %12)
>  %30 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %6, <2 x i32> %12)
>  %31 = insertelement <4 x float> undef, float %27, i32 0
>  %32 = insertelement <4 x float> %31, float %28, i32 1
>  %33 = insertelement <4 x float> %32, float %29, i32 2
>  %34 = insertelement <4 x float> %33, float %30, i32 3
>  %35 = call <4 x float> @llvm.AMDGPU.cube(<4 x float> %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 = extractelement <4 x float> %35, i32 3
>  %40 = call float @llvm.fabs.f32(float %38)
>  %41 = fdiv float 1.000000e+00, %40
>  %42 = fmul float %36, %41
>  %43 = fadd float %42, 1.500000e+00
>  %44 = fmul float %37, %41
>  %45 = fadd float %44, 1.500000e+00
>  %46 = bitcast float %45 to i32
>  %47 = bitcast float %43 to i32
>  %48 = bitcast float %39 to i32
>  %49 = insertelement <4 x i32> undef, i32 %46, i32 0
>  %50 = insertelement <4 x i32> %49, i32 %47, i32 1
>  %51 = insertelement <4 x i32> %50, i32 %48, i32 2
>  %52 = call <4 x float> @llvm.SI.image.sample.v4i32(<4 x i32> %51, <8 x i32> %24, <4 x i32> %26, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %53 = extractelement <4 x float> %52, i32 0
>  %54 = extractelement <4 x float> %52, i32 1
>  %55 = extractelement <4 x float> %52, i32 2
>  %56 = extractelement <4 x float> %52, i32 3
>  %57 = call i32 @llvm.SI.packf16(float %53, float %54)
>  %58 = bitcast i32 %57 to float
>  %59 = call i32 @llvm.SI.packf16(float %55, float %56)
>  %60 = bitcast i32 %59 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %58, float %60, float %58, float %60)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: readnone
>declare <4 x float> @llvm.AMDGPU.cube(<4 x float>) #2
>
>; Function Attrs: nounwind readnone
>declare float @llvm.fabs.f32(float) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.image.sample.v4i32(<4 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, 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 }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                  ; BEFE0A7E
>	s_mov_b32 m0, s10                                                     ; BEFC030A
>	s_load_dwordx8 s[8:15], s[6:7], 0x0                                   ; C0C40700
>	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_interp_p1_f32 v4, v0, 2, 0, [m0]                                    ; C8100200
>	s_load_dwordx4 s[0:3], s[4:5], 0x0                                    ; C0800500
>	v_interp_p2_f32 v4, [v4], v1, 2, 0, [m0]                              ; C8110201
>	v_interp_p1_f32 v0, v0, 3, 0, [m0]                                    ; C8000300
>	v_interp_p2_f32 v0, [v0], v1, 3, 0, [m0]                              ; C8010301
>	v_cubeid_f32 v7, v2, v3, v4                                           ; D2880007 04120702
>	v_cubema_f32 v0, v2, v3, v4                                           ; D28E0000 04120702
>	v_rcp_f32_e64 v0, |v0|                                                ; D3540100 00000100
>	v_cubesc_f32 v1, v2, v3, v4                                           ; D28A0001 04120702
>	v_cubetc_f32 v2, v2, v3, v4                                           ; D28C0002 04120702
>	v_mov_b32_e32 v5, 0x3fc00000                                          ; 7E0A02FF 3FC00000
>	v_mad_f32 v6, v0, v2, v5                                              ; D2820006 04160500
>	v_mac_f32_e32 v5, v0, v1                                              ; 3E0A0300
>	s_waitcnt lgkmcnt(0)                                                  ; BF8C007F
>	image_sample v[0:3], 15, 0, 0, 0, 0, 0, 0, 0, v[5:8], s[8:15], s[0:3] ; F0800F00 00020005
>	s_waitcnt vmcnt(0)                                                    ; BF8C0770
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v1                                    ; 5E000300
>	v_cvt_pkrtz_f16_f32_e32 v1, v2, v3                                    ; 5E020702
>	exp 15, 0, 1, 1, 1, v0, v1, v0, v1                                    ; F8001C0F 01000100
>	s_endpgm                                                              ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 16
>VGPRS: 12
>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
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[0]
>DCL CONST[0..3]
>DCL TEMP[0], LOCAL
>IMM[0] FLT32 {    0.0000,     0.0000,     0.0000,     0.0000}
>  0: MUL TEMP[0], CONST[0], IN[0].xxxx
>  1: MAD TEMP[0], CONST[1], IN[0].yyyy, TEMP[0]
>  2: MAD TEMP[0], CONST[2], IN[0].zzzz, TEMP[0]
>  3: MAD TEMP[0], CONST[3], IN[0].wwww, TEMP[0]
>  4: MOV OUT[1], IMM[0].xxxx
>  5: MOV OUT[0], TEMP[0]
>  6: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = call float @llvm.SI.load.const(<16 x i8> %13, i32 0)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %13, i32 4)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %13, i32 8)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %13, i32 12)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %13, i32 16)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %13, i32 20)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %13, i32 24)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %13, i32 28)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %13, i32 32)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %13, i32 36)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %13, i32 40)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %13, i32 44)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %13, i32 48)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %13, i32 52)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %13, i32 56)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %13, i32 60)
>  %30 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %31 = load <16 x i8>, <16 x i8> addrspace(2)* %30, align 16, !tbaa !0
>  %32 = add i32 %5, %8
>  %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 = extractelement <4 x float> %33, i32 2
>  %37 = extractelement <4 x float> %33, i32 3
>  %38 = fmul float %14, %34
>  %39 = fmul float %15, %34
>  %40 = fmul float %16, %34
>  %41 = fmul float %17, %34
>  %42 = fmul float %18, %35
>  %43 = fadd float %42, %38
>  %44 = fmul float %19, %35
>  %45 = fadd float %44, %39
>  %46 = fmul float %20, %35
>  %47 = fadd float %46, %40
>  %48 = fmul float %21, %35
>  %49 = fadd float %48, %41
>  %50 = fmul float %22, %36
>  %51 = fadd float %50, %43
>  %52 = fmul float %23, %36
>  %53 = fadd float %52, %45
>  %54 = fmul float %24, %36
>  %55 = fadd float %54, %47
>  %56 = fmul float %25, %36
>  %57 = fadd float %56, %49
>  %58 = fmul float %26, %37
>  %59 = fadd float %58, %51
>  %60 = fmul float %27, %37
>  %61 = fadd float %60, %53
>  %62 = fmul float %28, %37
>  %63 = fadd float %62, %55
>  %64 = fmul float %29, %37
>  %65 = fadd float %64, %57
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %59, float %61, float %63, float %65)
>  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[4:7], s[8:9], 0x0                  ; C0820900
>	v_add_i32_e32 v0, vcc, s10, v0                      ; 4A00000A
>	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 lgkmcnt(0)                                ; BF8C007F
>	s_buffer_load_dword s4, s[0:3], 0x0                 ; C2020100
>	s_buffer_load_dword s5, s[0:3], 0x1                 ; C2028101
>	s_buffer_load_dword s6, s[0:3], 0x2                 ; C2030102
>	s_buffer_load_dword s7, s[0:3], 0x3                 ; C2038103
>	s_buffer_load_dword s8, s[0:3], 0x4                 ; C2040104
>	s_buffer_load_dword s9, s[0:3], 0x5                 ; C2048105
>	s_buffer_load_dword s10, s[0:3], 0x6                ; C2050106
>	s_buffer_load_dword s11, s[0:3], 0x7                ; C2058107
>	s_buffer_load_dword s12, s[0:3], 0x8                ; C2060108
>	s_buffer_load_dword s13, s[0:3], 0x9                ; C2068109
>	s_buffer_load_dword s14, s[0:3], 0xa                ; C207010A
>	s_buffer_load_dword s15, s[0:3], 0xb                ; C207810B
>	s_buffer_load_dword s16, s[0:3], 0xc                ; C208010C
>	s_buffer_load_dword s17, s[0:3], 0xd                ; C208810D
>	s_buffer_load_dword s18, s[0:3], 0xe                ; C209010E
>	s_buffer_load_dword s0, s[0:3], 0xf                 ; C200010F
>	s_waitcnt vmcnt(0) lgkmcnt(0)                       ; BF8C0070
>	v_mul_f32_e32 v4, s4, v0                            ; 10080004
>	v_mul_f32_e32 v5, s5, v0                            ; 100A0005
>	v_mul_f32_e32 v6, s6, v0                            ; 100C0006
>	v_mul_f32_e32 v0, s7, v0                            ; 10000007
>	v_mac_f32_e32 v4, s8, v1                            ; 3E080208
>	v_mac_f32_e32 v5, s9, v1                            ; 3E0A0209
>	v_mac_f32_e32 v6, s10, v1                           ; 3E0C020A
>	v_mac_f32_e32 v0, s11, v1                           ; 3E00020B
>	v_mac_f32_e32 v4, s12, v2                           ; 3E08040C
>	v_mac_f32_e32 v5, s13, v2                           ; 3E0A040D
>	v_mac_f32_e32 v6, s14, v2                           ; 3E0C040E
>	v_mac_f32_e32 v0, s15, v2                           ; 3E00040F
>	v_mac_f32_e32 v4, s16, v3                           ; 3E080610
>	v_mac_f32_e32 v5, s17, v3                           ; 3E0A0611
>	v_mac_f32_e32 v6, s18, v3                           ; 3E0C0612
>	v_mac_f32_e32 v0, s0, v3                            ; 3E000600
>	v_mov_b32_e32 v1, 0                                 ; 7E020280
>	exp 15, 32, 0, 0, 0, v1, v1, v1, v1                 ; F800020F 01010101
>	exp 15, 12, 0, 1, 0, v4, v5, v6, v0                 ; F80008CF 00060504
>	s_endpgm                                            ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 24
>VGPRS: 8
>Code Size: 184 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
>  clamp_color = 0
>FRAG
>PROPERTY FS_COLOR0_WRITES_ALL_CBUFS 1
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SVIEW[0], CUBE, FLOAT
>DCL TEMP[0], LOCAL
>  0: MOV TEMP[0].xyz, IN[0].xyzz
>  1: TEX TEMP[0], TEMP[0], SAMP[0], CUBE
>  2: MOV OUT[0], TEMP[0]
>  3: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %24 = load <8 x i32>, <8 x i32> addrspace(2)* %23, align 32, !tbaa !0
>  %25 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %26 = load <4 x i32>, <4 x i32> addrspace(2)* %25, align 16, !tbaa !0
>  %27 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %8)
>  %28 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %8)
>  %29 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %6, <2 x i32> %8)
>  %30 = insertelement <4 x float> undef, float %27, i32 0
>  %31 = insertelement <4 x float> %30, float %28, i32 1
>  %32 = insertelement <4 x float> %31, float %29, i32 2
>  %33 = call <4 x float> @llvm.AMDGPU.cube(<4 x float> %32)
>  %34 = extractelement <4 x float> %33, i32 0
>  %35 = extractelement <4 x float> %33, i32 1
>  %36 = extractelement <4 x float> %33, i32 2
>  %37 = extractelement <4 x float> %33, i32 3
>  %38 = call float @llvm.fabs.f32(float %36)
>  %39 = fdiv float 1.000000e+00, %38
>  %40 = fmul float %34, %39
>  %41 = fadd float %40, 1.500000e+00
>  %42 = fmul float %35, %39
>  %43 = fadd float %42, 1.500000e+00
>  %44 = bitcast float %43 to i32
>  %45 = bitcast float %41 to i32
>  %46 = bitcast float %37 to i32
>  %47 = insertelement <4 x i32> undef, i32 %44, i32 0
>  %48 = insertelement <4 x i32> %47, i32 %45, i32 1
>  %49 = insertelement <4 x i32> %48, i32 %46, i32 2
>  %50 = call <4 x float> @llvm.SI.image.sample.v4i32(<4 x i32> %49, <8 x i32> %24, <4 x i32> %26, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %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 = call i32 @llvm.SI.packf16(float %51, float %52)
>  %56 = bitcast i32 %55 to float
>  %57 = call i32 @llvm.SI.packf16(float %53, float %54)
>  %58 = bitcast i32 %57 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %56, float %58, float %56, float %58)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: readnone
>declare <4 x float> @llvm.AMDGPU.cube(<4 x float>) #2
>
>; Function Attrs: nounwind readnone
>declare float @llvm.fabs.f32(float) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.image.sample.v4i32(<4 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, 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 }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                  ; BEFE0A7E
>	s_mov_b32 m0, s10                                                     ; BEFC030A
>	s_load_dwordx8 s[8:15], s[6:7], 0x0                                   ; C0C40700
>	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
>	s_load_dwordx4 s[0:3], s[4:5], 0x0                                    ; C0800500
>	v_interp_p2_f32 v3, [v3], v1, 1, 0, [m0]                              ; C80D0101
>	v_interp_p1_f32 v0, v0, 2, 0, [m0]                                    ; C8000200
>	v_interp_p2_f32 v0, [v0], v1, 2, 0, [m0]                              ; C8010201
>	v_cubeid_f32 v6, v2, v3, v0                                           ; D2880006 04020702
>	v_cubema_f32 v1, v2, v3, v0                                           ; D28E0001 04020702
>	v_rcp_f32_e64 v1, |v1|                                                ; D3540101 00000101
>	v_cubesc_f32 v7, v2, v3, v0                                           ; D28A0007 04020702
>	v_cubetc_f32 v0, v2, v3, v0                                           ; D28C0000 04020702
>	v_mov_b32_e32 v4, 0x3fc00000                                          ; 7E0802FF 3FC00000
>	v_mad_f32 v5, v1, v0, v4                                              ; D2820005 04120101
>	v_mac_f32_e32 v4, v1, v7                                              ; 3E080F01
>	s_waitcnt lgkmcnt(0)                                                  ; BF8C007F
>	image_sample v[0:3], 15, 0, 0, 0, 0, 0, 0, 0, v[4:7], s[8:15], s[0:3] ; F0800F00 00020004
>	s_waitcnt vmcnt(0)                                                    ; BF8C0770
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v1                                    ; 5E000300
>	v_cvt_pkrtz_f16_f32_e32 v1, v2, v3                                    ; 5E020702
>	exp 15, 0, 1, 1, 1, v0, v1, v0, v1                                    ; F8001C0F 01000100
>	s_endpgm                                                              ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 16
>VGPRS: 8
>Code Size: 136 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
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL IN[2]
>DCL IN[3]
>DCL IN[4]
>DCL OUT[0], POSITION
>DCL OUT[1], CLIPVERTEX
>DCL OUT[2], GENERIC[0]
>DCL OUT[3], GENERIC[1]
>DCL OUT[4], GENERIC[2]
>DCL CONST[0..57]
>DCL TEMP[0..7], LOCAL
>IMM[0] FLT32 {    1.0000,     2.2000,     0.0000,     0.0000}
>  0: ABS TEMP[0].x, CONST[50].xxxx
>  1: FSLT TEMP[1].x, -TEMP[0].xxxx, TEMP[0].xxxx
>  2: AND TEMP[1].x, TEMP[1].xxxx, IMM[0].xxxx
>  3: ADD TEMP[0].xyz, IN[1].xyzz, IN[1].xyzz
>  4: LG2 TEMP[2].x, TEMP[0].xxxx
>  5: LG2 TEMP[3].x, TEMP[0].yyyy
>  6: MOV TEMP[2].y, TEMP[3].xxxx
>  7: LG2 TEMP[3].x, TEMP[0].zzzz
>  8: MOV TEMP[2].z, TEMP[3].xxxx
>  9: MUL TEMP[0].xyz, TEMP[2].xyzz, IMM[0].yyyy
> 10: EX2 TEMP[2].x, TEMP[0].xxxx
> 11: EX2 TEMP[3].x, TEMP[0].yyyy
> 12: MOV TEMP[2].y, TEMP[3].xxxx
> 13: EX2 TEMP[3].x, TEMP[0].zzzz
> 14: MOV TEMP[2].z, TEMP[3].xxxx
> 15: MUL TEMP[1].xyz, TEMP[1].xxxx, TEMP[2].xyzz
> 16: DP4 TEMP[3].x, IN[2], CONST[48]
> 17: DP4 TEMP[4].x, IN[2], CONST[49]
> 18: MOV TEMP[3].y, TEMP[4].xxxx
> 19: MOV TEMP[0].w, CONST[0].yyyy
> 20: MOV TEMP[2].w, IN[0].wwww
> 21: MAD TEMP[2].xyz, IN[3].xyzz, CONST[13].xxxx, IN[0].xyzz
> 22: DP4 TEMP[0].x, TEMP[2], CONST[54]
> 23: DP4 TEMP[4].x, TEMP[2], CONST[55]
> 24: MOV TEMP[0].y, TEMP[4].xxxx
> 25: DP4 TEMP[2].x, TEMP[2], CONST[56]
> 26: MOV TEMP[0].z, TEMP[2].xxxx
> 27: DP4 TEMP[2].x, TEMP[0], CONST[8]
> 28: DP4 TEMP[4].x, TEMP[0], CONST[9]
> 29: MOV TEMP[2].y, TEMP[4].xxxx
> 30: DP4 TEMP[5].x, TEMP[0], CONST[11]
> 31: MOV TEMP[2].w, TEMP[5].xxxx
> 32: DP4 TEMP[6].x, TEMP[0], CONST[10]
> 33: MOV TEMP[0].w, TEMP[6].xxxx
> 34: MOV TEMP[2].z, TEMP[6].xxxx
> 35: MOV TEMP[3].zw, CONST[0].xxxx
> 36: MOV TEMP[1].w, CONST[0].xxxx
> 37: MOV TEMP[7], TEMP[2]
> 38: MAD TEMP[6].x, TEMP[6].xxxx, CONST[0].zzzz, -TEMP[5].xxxx
> 39: MOV TEMP[2].z, TEMP[6].xxxx
> 40: MOV TEMP[2].y, -TEMP[4].xxxx
> 41: MAD TEMP[2].xy, CONST[57].xyyy, TEMP[5].xxxx, TEMP[2].xyyy
> 42: MOV OUT[2], TEMP[3]
> 43: MOV OUT[0], TEMP[2]
> 44: MOV OUT[1], TEMP[7]
> 45: MOV OUT[3], TEMP[1]
> 46: MOV OUT[4], TEMP[0]
> 47: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = call float @llvm.SI.load.const(<16 x i8> %13, i32 0)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %13, i32 4)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %13, i32 8)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %13, i32 128)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %13, i32 132)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %13, i32 136)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %13, i32 140)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %13, i32 144)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %13, i32 148)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %13, i32 152)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %13, i32 156)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %13, i32 160)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %13, i32 164)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %13, i32 168)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %13, i32 172)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %13, i32 176)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %13, i32 180)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %13, i32 184)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %13, i32 188)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %13, i32 208)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %13, i32 768)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %13, i32 772)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %13, i32 776)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %13, i32 780)
>  %38 = call float @llvm.SI.load.const(<16 x i8> %13, i32 784)
>  %39 = call float @llvm.SI.load.const(<16 x i8> %13, i32 788)
>  %40 = call float @llvm.SI.load.const(<16 x i8> %13, i32 792)
>  %41 = call float @llvm.SI.load.const(<16 x i8> %13, i32 796)
>  %42 = call float @llvm.SI.load.const(<16 x i8> %13, i32 800)
>  %43 = call float @llvm.SI.load.const(<16 x i8> %13, i32 864)
>  %44 = call float @llvm.SI.load.const(<16 x i8> %13, i32 868)
>  %45 = call float @llvm.SI.load.const(<16 x i8> %13, i32 872)
>  %46 = call float @llvm.SI.load.const(<16 x i8> %13, i32 876)
>  %47 = call float @llvm.SI.load.const(<16 x i8> %13, i32 880)
>  %48 = call float @llvm.SI.load.const(<16 x i8> %13, i32 884)
>  %49 = call float @llvm.SI.load.const(<16 x i8> %13, i32 888)
>  %50 = call float @llvm.SI.load.const(<16 x i8> %13, i32 892)
>  %51 = call float @llvm.SI.load.const(<16 x i8> %13, i32 896)
>  %52 = call float @llvm.SI.load.const(<16 x i8> %13, i32 900)
>  %53 = call float @llvm.SI.load.const(<16 x i8> %13, i32 904)
>  %54 = call float @llvm.SI.load.const(<16 x i8> %13, i32 908)
>  %55 = call float @llvm.SI.load.const(<16 x i8> %13, i32 912)
>  %56 = call float @llvm.SI.load.const(<16 x i8> %13, i32 916)
>  %57 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %58 = load <16 x i8>, <16 x i8> addrspace(2)* %57, align 16, !tbaa !0
>  %59 = add i32 %5, %8
>  %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>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 1
>  %66 = load <16 x i8>, <16 x i8> addrspace(2)* %65, align 16, !tbaa !0
>  %67 = add i32 %5, %8
>  %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 = extractelement <4 x float> %68, i32 2
>  %72 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 2
>  %73 = load <16 x i8>, <16 x i8> addrspace(2)* %72, align 16, !tbaa !0
>  %74 = add i32 %5, %8
>  %75 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %73, i32 0, i32 %74)
>  %76 = extractelement <4 x float> %75, i32 0
>  %77 = extractelement <4 x float> %75, i32 1
>  %78 = extractelement <4 x float> %75, i32 2
>  %79 = extractelement <4 x float> %75, i32 3
>  %80 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 3
>  %81 = load <16 x i8>, <16 x i8> addrspace(2)* %80, align 16, !tbaa !0
>  %82 = add i32 %5, %8
>  %83 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %81, i32 0, i32 %82)
>  %84 = extractelement <4 x float> %83, i32 0
>  %85 = extractelement <4 x float> %83, i32 1
>  %86 = extractelement <4 x float> %83, i32 2
>  %87 = call float @llvm.fabs.f32(float %42)
>  %88 = fsub float -0.000000e+00, %87
>  %89 = fcmp ogt float %87, %88
>  %90 = select i1 %89, float 1.000000e+00, float 0.000000e+00
>  %91 = fadd float %69, %69
>  %92 = fadd float %70, %70
>  %93 = fadd float %71, %71
>  %94 = call float @llvm.log2.f32(float %91)
>  %95 = call float @llvm.log2.f32(float %92)
>  %96 = call float @llvm.log2.f32(float %93)
>  %97 = fmul float %94, 0x40019999A0000000
>  %98 = fmul float %95, 0x40019999A0000000
>  %99 = fmul float %96, 0x40019999A0000000
>  %100 = call float @llvm.AMDIL.exp.(float %97)
>  %101 = call float @llvm.AMDIL.exp.(float %98)
>  %102 = call float @llvm.AMDIL.exp.(float %99)
>  %103 = fmul float %90, %100
>  %104 = fmul float %90, %101
>  %105 = fmul float %90, %102
>  %106 = fmul float %76, %34
>  %107 = fmul float %77, %35
>  %108 = fadd float %106, %107
>  %109 = fmul float %78, %36
>  %110 = fadd float %108, %109
>  %111 = fmul float %79, %37
>  %112 = fadd float %110, %111
>  %113 = fmul float %76, %38
>  %114 = fmul float %77, %39
>  %115 = fadd float %113, %114
>  %116 = fmul float %78, %40
>  %117 = fadd float %115, %116
>  %118 = fmul float %79, %41
>  %119 = fadd float %117, %118
>  %120 = fmul float %84, %33
>  %121 = fadd float %120, %61
>  %122 = fmul float %85, %33
>  %123 = fadd float %122, %62
>  %124 = fmul float %86, %33
>  %125 = fadd float %124, %63
>  %126 = fmul float %121, %43
>  %127 = fmul float %123, %44
>  %128 = fadd float %126, %127
>  %129 = fmul float %125, %45
>  %130 = fadd float %128, %129
>  %131 = fmul float %64, %46
>  %132 = fadd float %130, %131
>  %133 = fmul float %121, %47
>  %134 = fmul float %123, %48
>  %135 = fadd float %133, %134
>  %136 = fmul float %125, %49
>  %137 = fadd float %135, %136
>  %138 = fmul float %64, %50
>  %139 = fadd float %137, %138
>  %140 = fmul float %121, %51
>  %141 = fmul float %123, %52
>  %142 = fadd float %140, %141
>  %143 = fmul float %125, %53
>  %144 = fadd float %142, %143
>  %145 = fmul float %64, %54
>  %146 = fadd float %144, %145
>  %147 = fmul float %132, %17
>  %148 = fmul float %139, %18
>  %149 = fadd float %147, %148
>  %150 = fmul float %146, %19
>  %151 = fadd float %149, %150
>  %152 = fmul float %15, %20
>  %153 = fadd float %151, %152
>  %154 = fmul float %132, %21
>  %155 = fmul float %139, %22
>  %156 = fadd float %154, %155
>  %157 = fmul float %146, %23
>  %158 = fadd float %156, %157
>  %159 = fmul float %15, %24
>  %160 = fadd float %158, %159
>  %161 = fmul float %132, %29
>  %162 = fmul float %139, %30
>  %163 = fadd float %161, %162
>  %164 = fmul float %146, %31
>  %165 = fadd float %163, %164
>  %166 = fmul float %15, %32
>  %167 = fadd float %165, %166
>  %168 = fmul float %132, %25
>  %169 = fmul float %139, %26
>  %170 = fadd float %168, %169
>  %171 = fmul float %146, %27
>  %172 = fadd float %170, %171
>  %173 = fmul float %15, %28
>  %174 = fadd float %172, %173
>  %175 = fmul float %174, %16
>  %176 = fsub float %175, %167
>  %177 = fmul float %55, %167
>  %178 = fadd float %177, %153
>  %179 = fmul float %56, %167
>  %180 = fsub float %179, %160
>  %181 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 16
>  %182 = load <16 x i8>, <16 x i8> addrspace(2)* %181, align 16, !tbaa !0
>  %183 = call float @llvm.SI.load.const(<16 x i8> %182, i32 0)
>  %184 = fmul float %183, %153
>  %185 = call float @llvm.SI.load.const(<16 x i8> %182, i32 4)
>  %186 = fmul float %185, %160
>  %187 = fadd float %184, %186
>  %188 = call float @llvm.SI.load.const(<16 x i8> %182, i32 8)
>  %189 = fmul float %188, %174
>  %190 = fadd float %187, %189
>  %191 = call float @llvm.SI.load.const(<16 x i8> %182, i32 12)
>  %192 = fmul float %191, %167
>  %193 = fadd float %190, %192
>  %194 = call float @llvm.SI.load.const(<16 x i8> %182, i32 16)
>  %195 = fmul float %194, %153
>  %196 = call float @llvm.SI.load.const(<16 x i8> %182, i32 20)
>  %197 = fmul float %196, %160
>  %198 = fadd float %195, %197
>  %199 = call float @llvm.SI.load.const(<16 x i8> %182, i32 24)
>  %200 = fmul float %199, %174
>  %201 = fadd float %198, %200
>  %202 = call float @llvm.SI.load.const(<16 x i8> %182, i32 28)
>  %203 = fmul float %202, %167
>  %204 = fadd float %201, %203
>  %205 = call float @llvm.SI.load.const(<16 x i8> %182, i32 32)
>  %206 = fmul float %205, %153
>  %207 = call float @llvm.SI.load.const(<16 x i8> %182, i32 36)
>  %208 = fmul float %207, %160
>  %209 = fadd float %206, %208
>  %210 = call float @llvm.SI.load.const(<16 x i8> %182, i32 40)
>  %211 = fmul float %210, %174
>  %212 = fadd float %209, %211
>  %213 = call float @llvm.SI.load.const(<16 x i8> %182, i32 44)
>  %214 = fmul float %213, %167
>  %215 = fadd float %212, %214
>  %216 = call float @llvm.SI.load.const(<16 x i8> %182, i32 48)
>  %217 = fmul float %216, %153
>  %218 = call float @llvm.SI.load.const(<16 x i8> %182, i32 52)
>  %219 = fmul float %218, %160
>  %220 = fadd float %217, %219
>  %221 = call float @llvm.SI.load.const(<16 x i8> %182, i32 56)
>  %222 = fmul float %221, %174
>  %223 = fadd float %220, %222
>  %224 = call float @llvm.SI.load.const(<16 x i8> %182, i32 60)
>  %225 = fmul float %224, %167
>  %226 = fadd float %223, %225
>  %227 = call float @llvm.SI.load.const(<16 x i8> %182, i32 64)
>  %228 = fmul float %227, %153
>  %229 = call float @llvm.SI.load.const(<16 x i8> %182, i32 68)
>  %230 = fmul float %229, %160
>  %231 = fadd float %228, %230
>  %232 = call float @llvm.SI.load.const(<16 x i8> %182, i32 72)
>  %233 = fmul float %232, %174
>  %234 = fadd float %231, %233
>  %235 = call float @llvm.SI.load.const(<16 x i8> %182, i32 76)
>  %236 = fmul float %235, %167
>  %237 = fadd float %234, %236
>  %238 = call float @llvm.SI.load.const(<16 x i8> %182, i32 80)
>  %239 = fmul float %238, %153
>  %240 = call float @llvm.SI.load.const(<16 x i8> %182, i32 84)
>  %241 = fmul float %240, %160
>  %242 = fadd float %239, %241
>  %243 = call float @llvm.SI.load.const(<16 x i8> %182, i32 88)
>  %244 = fmul float %243, %174
>  %245 = fadd float %242, %244
>  %246 = call float @llvm.SI.load.const(<16 x i8> %182, i32 92)
>  %247 = fmul float %246, %167
>  %248 = fadd float %245, %247
>  %249 = call float @llvm.SI.load.const(<16 x i8> %182, i32 96)
>  %250 = fmul float %249, %153
>  %251 = call float @llvm.SI.load.const(<16 x i8> %182, i32 100)
>  %252 = fmul float %251, %160
>  %253 = fadd float %250, %252
>  %254 = call float @llvm.SI.load.const(<16 x i8> %182, i32 104)
>  %255 = fmul float %254, %174
>  %256 = fadd float %253, %255
>  %257 = call float @llvm.SI.load.const(<16 x i8> %182, i32 108)
>  %258 = fmul float %257, %167
>  %259 = fadd float %256, %258
>  %260 = call float @llvm.SI.load.const(<16 x i8> %182, i32 112)
>  %261 = fmul float %260, %153
>  %262 = call float @llvm.SI.load.const(<16 x i8> %182, i32 116)
>  %263 = fmul float %262, %160
>  %264 = fadd float %261, %263
>  %265 = call float @llvm.SI.load.const(<16 x i8> %182, i32 120)
>  %266 = fmul float %265, %174
>  %267 = fadd float %264, %266
>  %268 = call float @llvm.SI.load.const(<16 x i8> %182, i32 124)
>  %269 = fmul float %268, %167
>  %270 = fadd float %267, %269
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %112, float %119, float %14, float %14)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %103, float %104, float %105, float %14)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 34, i32 0, float %132, float %139, float %146, float %174)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 12, i32 0, float %178, float %180, float %176, float %167)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 13, i32 0, float %193, float %204, float %215, float %226)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 14, i32 0, float %237, float %248, float %259, float %270)
>  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.fabs.f32(float) #1
>
>; 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_mov_b32_e32 v1, 0x400ccccd                           ; 7E0202FF 400CCCCD
>	v_add_i32_e32 v0, vcc, 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_load_dwordx4 s[20:23], s[2:3], 0x0                   ; C08A0300
>	s_load_dwordx4 s[40:43], s[2:3], 0x40                  ; C0940340
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	buffer_load_format_xyzw v[2:5], v0, s[4:7], 0 idxen    ; E00C2000 80010200
>	buffer_load_format_xyzw v[6:9], v0, s[12:15], 0 idxen  ; E00C2000 80030600
>	s_waitcnt vmcnt(0)                                     ; BF8C0770
>	buffer_load_format_xyzw v[9:12], v0, s[16:19], 0 idxen ; E00C2000 80040900
>	buffer_load_format_xyzw v[13:16], v0, s[8:11], 0 idxen ; E00C2000 80020D00
>	s_buffer_load_dword s37, s[20:23], 0x2c                ; C212952C
>	s_buffer_load_dword s38, s[20:23], 0x2d                ; C213152D
>	s_buffer_load_dword s39, s[20:23], 0x2e                ; C213952E
>	s_buffer_load_dword s0, s[20:23], 0x2f                 ; C200152F
>	s_buffer_load_dword s44, s[20:23], 0x34                ; C2161534
>	s_buffer_load_dword s45, s[20:23], 0xd9                ; C21695D9
>	s_buffer_load_dword s46, s[20:23], 0xda                ; C21715DA
>	s_buffer_load_dword s47, s[20:23], 0xdb                ; C21795DB
>	s_buffer_load_dword s48, s[20:23], 0xdc                ; C21815DC
>	s_buffer_load_dword s49, s[20:23], 0xdd                ; C21895DD
>	s_buffer_load_dword s50, s[20:23], 0xc5                ; C21915C5
>	s_buffer_load_dword s51, s[20:23], 0xc6                ; C21995C6
>	s_buffer_load_dword s52, s[20:23], 0xc7                ; C21A15C7
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_mov_b32_e32 v0, s0                                   ; 7E000200
>	s_buffer_load_dword s0, s[20:23], 0xc8                 ; C20015C8
>	s_buffer_load_dword s53, s[20:23], 0xd8                ; C21A95D8
>	s_buffer_load_dword s54, s[20:23], 0xde                ; C21B15DE
>	s_buffer_load_dword s55, s[20:23], 0xdf                ; C21B95DF
>	s_buffer_load_dword s56, s[20:23], 0xe0                ; C21C15E0
>	s_buffer_load_dword s57, s[20:23], 0xe1                ; C21C95E1
>	s_buffer_load_dword s58, s[20:23], 0xe2                ; C21D15E2
>	s_buffer_load_dword s59, s[20:23], 0x22                ; C21D9522
>	s_buffer_load_dword s1, s[20:23], 0x23                 ; C2009523
>	s_buffer_load_dword s60, s[20:23], 0x24                ; C21E1524
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_cmp_gt_f32_e64 s[2:3], |s0|, -|s0|                   ; D0080302 40000000
>	s_waitcnt vmcnt(0)                                     ; BF8C0770
>	v_cndmask_b32_e64 v16, 0, 1.0, s[2:3]                  ; D2000010 0009E480
>	s_buffer_load_dword s61, s[20:23], 0x25                ; C21E9525
>	s_buffer_load_dword s62, s[20:23], 0x26                ; C21F1526
>	s_buffer_load_dword s63, s[20:23], 0xc0                ; C21F95C0
>	s_buffer_load_dword s64, s[20:23], 0xc1                ; C22015C1
>	s_buffer_load_dword s65, s[20:23], 0xc2                ; C22095C2
>	s_buffer_load_dword s66, s[20:23], 0xc3                ; C22115C3
>	v_mov_b32_e32 v17, s1                                  ; 7E220201
>	s_buffer_load_dword s67, s[20:23], 0xc4                ; C22195C4
>	s_buffer_load_dword s1, s[20:23], 0x0                  ; C2009500
>	s_buffer_load_dword s6, s[20:23], 0x1                  ; C2031501
>	s_buffer_load_dword s0, s[20:23], 0x2                  ; C2001502
>	s_buffer_load_dword s68, s[20:23], 0x20                ; C2221520
>	s_buffer_load_dword s69, s[20:23], 0x21                ; C2229521
>	s_buffer_load_dword s70, s[20:23], 0xe3                ; C22315E3
>	s_buffer_load_dword s12, s[20:23], 0xe4                ; C20615E4
>	s_buffer_load_dword s10, s[20:23], 0xe5                ; C20515E5
>	s_buffer_load_dword s2, s[20:23], 0x27                 ; C2011527
>	s_buffer_load_dword s71, s[20:23], 0x28                ; C2239528
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_mov_b32_e32 v18, s1                                  ; 7E240201
>	s_buffer_load_dword s72, s[20:23], 0x29                ; C2241529
>	s_buffer_load_dword s28, s[20:23], 0x2a                ; C20E152A
>	s_buffer_load_dword s3, s[20:23], 0x2b                 ; C201952B
>	s_buffer_load_dword s13, s[40:43], 0x0                 ; C206A900
>	s_buffer_load_dword s23, s[40:43], 0x1                 ; C20BA901
>	s_buffer_load_dword s4, s[40:43], 0x2                  ; C2022902
>	s_buffer_load_dword s1, s[40:43], 0x3                  ; C200A903
>	s_buffer_load_dword s16, s[40:43], 0x4                 ; C2082904
>	v_mov_b32_e32 v19, s2                                  ; 7E260202
>	s_buffer_load_dword s26, s[40:43], 0x5                 ; C20D2905
>	s_buffer_load_dword s8, s[40:43], 0x6                  ; C2042906
>	s_buffer_load_dword s2, s[40:43], 0x7                  ; C2012907
>	s_buffer_load_dword s19, s[40:43], 0x8                 ; C209A908
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_mov_b32_e32 v20, s3                                  ; 7E280203
>	s_buffer_load_dword s30, s[40:43], 0x9                 ; C20F2909
>	s_buffer_load_dword s11, s[40:43], 0xa                 ; C205A90A
>	s_buffer_load_dword s3, s[40:43], 0xb                  ; C201A90B
>	s_buffer_load_dword s22, s[40:43], 0xc                 ; C20B290C
>	s_buffer_load_dword s32, s[40:43], 0xd                 ; C210290D
>	s_buffer_load_dword s15, s[40:43], 0xe                 ; C207A90E
>	s_buffer_load_dword s5, s[40:43], 0xf                  ; C202A90F
>	s_buffer_load_dword s25, s[40:43], 0x10                ; C20CA910
>	s_buffer_load_dword s33, s[40:43], 0x11                ; C210A911
>	s_buffer_load_dword s18, s[40:43], 0x12                ; C2092912
>	s_buffer_load_dword s7, s[40:43], 0x13                 ; C203A913
>	s_buffer_load_dword s27, s[40:43], 0x14                ; C20DA914
>	s_buffer_load_dword s34, s[40:43], 0x15                ; C2112915
>	s_buffer_load_dword s20, s[40:43], 0x16                ; C20A2916
>	s_buffer_load_dword s9, s[40:43], 0x17                 ; C204A917
>	s_buffer_load_dword s29, s[40:43], 0x18                ; C20EA918
>	s_buffer_load_dword s35, s[40:43], 0x19                ; C211A919
>	s_buffer_load_dword s21, s[40:43], 0x1a                ; C20AA91A
>	s_buffer_load_dword s14, s[40:43], 0x1b                ; C207291B
>	s_buffer_load_dword s31, s[40:43], 0x1c                ; C20FA91C
>	s_buffer_load_dword s36, s[40:43], 0x1d                ; C212291D
>	s_buffer_load_dword s24, s[40:43], 0x1e                ; C20C291E
>	s_buffer_load_dword s17, s[40:43], 0x1f                ; C208A91F
>	v_add_f32_e32 v6, v6, v6                               ; 060C0D06
>	v_mul_f32_e32 v21, s64, v10                            ; 102A1440
>	v_mul_f32_e32 v10, s50, v10                            ; 10141432
>	v_mad_f32 v2, s44, v13, v2                             ; D2820002 040A1A2C
>	v_mad_f32 v3, s44, v14, v3                             ; D2820003 040E1C2C
>	v_mad_f32 v4, s44, v15, v4                             ; D2820004 04121E2C
>	v_mac_f32_e32 v21, s63, v9                             ; 3E2A123F
>	v_mac_f32_e32 v10, s67, v9                             ; 3E141243
>	v_mul_f32_e32 v9, s45, v3                              ; 1012062D
>	v_mac_f32_e32 v21, s65, v11                            ; 3E2A1641
>	v_mac_f32_e32 v10, s51, v11                            ; 3E141633
>	v_mac_f32_e32 v21, s66, v12                            ; 3E2A1842
>	v_mac_f32_e32 v10, s52, v12                            ; 3E141834
>	v_mul_f32_e32 v11, s49, v3                             ; 10160631
>	v_mul_f32_e32 v3, s57, v3                              ; 10060639
>	v_mac_f32_e32 v9, s53, v2                              ; 3E120435
>	v_mac_f32_e32 v11, s48, v2                             ; 3E160430
>	v_mac_f32_e32 v3, s56, v2                              ; 3E060438
>	v_mac_f32_e32 v9, s46, v4                              ; 3E12082E
>	v_mac_f32_e32 v11, s54, v4                             ; 3E160836
>	v_mac_f32_e32 v3, s58, v4                              ; 3E06083A
>	v_mac_f32_e32 v9, s47, v5                              ; 3E120A2F
>	v_mac_f32_e32 v11, s55, v5                             ; 3E160A37
>	v_mac_f32_e32 v3, s70, v5                              ; 3E060A46
>	v_mul_f32_e32 v2, s69, v11                             ; 10041645
>	v_mac_f32_e32 v2, s68, v9                              ; 3E041244
>	v_mul_f32_e32 v4, s61, v11                             ; 1008163D
>	v_mac_f32_e32 v4, s60, v9                              ; 3E08123C
>	v_mul_f32_e32 v5, s38, v11                             ; 100A1626
>	v_mul_f32_e32 v12, s72, v11                            ; 10181648
>	v_mac_f32_e32 v5, s37, v9                              ; 3E0A1225
>	v_mac_f32_e32 v12, s71, v9                             ; 3E181247
>	v_mac_f32_e32 v2, s59, v3                              ; 3E04063B
>	v_mac_f32_e32 v4, s62, v3                              ; 3E08063E
>	v_mac_f32_e32 v5, s39, v3                              ; 3E0A0627
>	v_add_f32_e32 v7, v7, v7                               ; 060E0F07
>	v_add_f32_e32 v8, v8, v8                               ; 06101108
>	v_log_f32_e32 v6, v6                                   ; 7E0C4F06
>	v_log_f32_e32 v7, v7                                   ; 7E0E4F07
>	v_log_f32_e32 v8, v8                                   ; 7E104F08
>	exp 15, 32, 0, 0, 0, v21, v10, v18, v18                ; F800020F 12120A15
>	v_mul_f32_e32 v6, v1, v6                               ; 100C0D01
>	v_mul_f32_e32 v7, v1, v7                               ; 100E0F01
>	v_mul_f32_e32 v1, v1, v8                               ; 10021101
>	v_exp_f32_e32 v6, v6                                   ; 7E0C4B06
>	v_exp_f32_e32 v7, v7                                   ; 7E0E4B07
>	v_exp_f32_e32 v1, v1                                   ; 7E024B01
>	v_mul_f32_e32 v6, v6, v16                              ; 100C2106
>	v_mul_f32_e32 v7, v7, v16                              ; 100E2107
>	v_mac_f32_e32 v12, s28, v3                             ; 3E18061C
>	v_mac_f32_e32 v2, s6, v17                              ; 3E042206
>	v_mac_f32_e32 v4, s6, v19                              ; 3E082606
>	v_mac_f32_e32 v5, s6, v0                               ; 3E0A0006
>	v_mac_f32_e32 v12, s6, v20                             ; 3E182806
>	v_mad_f32 v0, s12, v5, v2                              ; D2820000 040A0A0C
>	v_mad_f32 v8, s10, v5, -v4                             ; D2820008 84120A0A
>	s_waitcnt expcnt(0)                                    ; BF8C070F
>	v_mul_f32_e32 v10, s23, v4                             ; 10140817
>	v_mul_f32_e32 v13, s26, v4                             ; 101A081A
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_mul_f32_e32 v14, s30, v4                             ; 101C081E
>	v_mul_f32_e32 v15, s32, v4                             ; 101E0820
>	v_mul_f32_e32 v17, s33, v4                             ; 10220821
>	v_mul_f32_e32 v19, s34, v4                             ; 10260822
>	v_mul_f32_e32 v20, s35, v4                             ; 10280823
>	v_mul_f32_e32 v4, s36, v4                              ; 10080824
>	v_mac_f32_e32 v10, s13, v2                             ; 3E14040D
>	v_mac_f32_e32 v13, s16, v2                             ; 3E1A0410
>	v_mac_f32_e32 v14, s19, v2                             ; 3E1C0413
>	v_mac_f32_e32 v15, s22, v2                             ; 3E1E0416
>	v_mac_f32_e32 v17, s25, v2                             ; 3E220419
>	v_mac_f32_e32 v19, s27, v2                             ; 3E26041B
>	v_mac_f32_e32 v20, s29, v2                             ; 3E28041D
>	v_mac_f32_e32 v4, s31, v2                              ; 3E08041F
>	v_mac_f32_e32 v10, s4, v12                             ; 3E141804
>	v_mac_f32_e32 v13, s8, v12                             ; 3E1A1808
>	v_mac_f32_e32 v14, s11, v12                            ; 3E1C180B
>	v_mac_f32_e32 v15, s15, v12                            ; 3E1E180F
>	v_mac_f32_e32 v17, s18, v12                            ; 3E221812
>	v_mac_f32_e32 v19, s20, v12                            ; 3E261814
>	v_mac_f32_e32 v20, s21, v12                            ; 3E281815
>	v_mac_f32_e32 v4, s24, v12                             ; 3E081818
>	v_mac_f32_e32 v10, s1, v5                              ; 3E140A01
>	v_mac_f32_e32 v13, s2, v5                              ; 3E1A0A02
>	v_mac_f32_e32 v14, s3, v5                              ; 3E1C0A03
>	v_mac_f32_e32 v15, s5, v5                              ; 3E1E0A05
>	v_mac_f32_e32 v17, s7, v5                              ; 3E220A07
>	v_mac_f32_e32 v19, s9, v5                              ; 3E260A09
>	v_mac_f32_e32 v20, s14, v5                             ; 3E280A0E
>	v_mac_f32_e32 v4, s17, v5                              ; 3E080A11
>	v_mad_f32 v2, v12, s0, -v5                             ; D2820002 8414010C
>	v_mul_f32_e32 v1, v1, v16                              ; 10022101
>	exp 15, 33, 0, 0, 0, v6, v7, v1, v18                   ; F800021F 12010706
>	exp 15, 34, 0, 0, 0, v9, v11, v3, v12                  ; F800022F 0C030B09
>	exp 15, 12, 0, 0, 0, v0, v8, v2, v5                    ; F80000CF 05020800
>	exp 15, 13, 0, 0, 0, v10, v13, v14, v15                ; F80000DF 0F0E0D0A
>	exp 15, 14, 0, 1, 0, v17, v19, v20, v4                 ; F80008EF 04141311
>	s_endpgm                                               ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 80
>VGPRS: 24
>Code Size: 872 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
>  clamp_color = 0
>FRAG
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL IN[1], GENERIC[1], PERSPECTIVE
>DCL IN[2], GENERIC[2], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SVIEW[0], 2D, FLOAT
>DCL CONST[0..30]
>DCL TEMP[0..5], LOCAL
>IMM[0] FLT32 {   -1.0000,     1.0000,     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: MAD TEMP[1].x, CONST[20].wwww, TEMP[1].xxxx, IMM[0].yyyy
>  4: MUL TEMP[1].x, TEMP[1].xxxx, CONST[1].wwww
>  5: MAD TEMP[2].x, TEMP[1].xxxx, IN[1].wwww, -TEMP[1].xxxx
>  6: MAD TEMP[1].x, CONST[12].wwww, TEMP[2].xxxx, TEMP[1].xxxx
>  7: ABS TEMP[3].x, CONST[12].yyyy
>  8: MUL TEMP[4].x, CONST[29].wwww, IN[2].wwww
>  9: FSGE TEMP[5].x, -TEMP[3].xxxx, IMM[0].zzzz
> 10: UIF TEMP[5].xxxx :0
> 11:   MOV TEMP[5].x, TEMP[1].xxxx
> 12: ELSE :0
> 13:   MOV TEMP[5].x, TEMP[4].xxxx
> 14: ENDIF
> 15: MOV TEMP[3].w, TEMP[5].xxxx
> 16: ADD TEMP[4].x, TEMP[0].wwww, CONST[12].xxxx
> 17: MOV_SAT TEMP[4].x, TEMP[4].xxxx
> 18: ADD TEMP[1].xyz, IMM[0].xxxx, CONST[1].xyzz
> 19: MAD TEMP[1].xyz, TEMP[4].xxxx, TEMP[1].xyzz, IMM[0].yyyy
> 20: MUL TEMP[0].xyz, TEMP[0].xyzz, TEMP[1].xyzz
> 21: MUL TEMP[1].xyz, TEMP[0].xyzz, CONST[30].xxxx
> 22: MAD TEMP[0].xyz, TEMP[0].xyzz, -CONST[30].xxxx, CONST[29].xyzz
> 23: ADD TEMP[2].xyz, CONST[20].xyzz, -IN[2].xyzz
> 24: DP3 TEMP[2].x, TEMP[2].xyzz, TEMP[2].xyzz
> 25: SQRT TEMP[2].x, TEMP[2].xxxx
> 26: MAD TEMP[2].x, TEMP[2].xxxx, CONST[21].wwww, CONST[21].xxxx
> 27: MOV_SAT TEMP[2].x, TEMP[2].xxxx
> 28: MIN TEMP[2].x, TEMP[2].xxxx, CONST[21].zzzz
> 29: MUL TEMP[2].x, TEMP[2].xxxx, TEMP[2].xxxx
> 30: MAD TEMP[3].xyz, TEMP[2].xxxx, TEMP[0].xyzz, TEMP[1].xyzz
> 31: MOV OUT[0], TEMP[3]
> 32: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %24 = load <16 x i8>, <16 x i8> addrspace(2)* %23, align 16, !tbaa !0
>  %25 = call float @llvm.SI.load.const(<16 x i8> %24, i32 16)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %24, i32 20)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %24, i32 24)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %24, i32 28)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %24, i32 192)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %24, i32 196)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %24, i32 204)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %24, i32 320)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %24, i32 324)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %24, i32 328)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %24, i32 332)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %24, i32 336)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %24, i32 344)
>  %38 = call float @llvm.SI.load.const(<16 x i8> %24, i32 348)
>  %39 = call float @llvm.SI.load.const(<16 x i8> %24, i32 464)
>  %40 = call float @llvm.SI.load.const(<16 x i8> %24, i32 468)
>  %41 = call float @llvm.SI.load.const(<16 x i8> %24, i32 472)
>  %42 = call float @llvm.SI.load.const(<16 x i8> %24, i32 476)
>  %43 = call float @llvm.SI.load.const(<16 x i8> %24, i32 480)
>  %44 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %45 = load <8 x i32>, <8 x i32> addrspace(2)* %44, align 32, !tbaa !0
>  %46 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %47 = load <4 x i32>, <4 x i32> addrspace(2)* %46, align 16, !tbaa !0
>  %48 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %8)
>  %49 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %8)
>  %50 = call float @llvm.SI.fs.interp(i32 3, i32 1, i32 %6, <2 x i32> %8)
>  %51 = call float @llvm.SI.fs.interp(i32 0, i32 2, i32 %6, <2 x i32> %8)
>  %52 = call float @llvm.SI.fs.interp(i32 1, i32 2, i32 %6, <2 x i32> %8)
>  %53 = call float @llvm.SI.fs.interp(i32 2, i32 2, i32 %6, <2 x i32> %8)
>  %54 = call float @llvm.SI.fs.interp(i32 3, i32 2, i32 %6, <2 x i32> %8)
>  %55 = bitcast float %48 to i32
>  %56 = bitcast float %49 to i32
>  %57 = insertelement <2 x i32> undef, i32 %55, i32 0
>  %58 = insertelement <2 x i32> %57, i32 %56, i32 1
>  %59 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %58, <8 x i32> %45, <4 x i32> %47, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %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 = fadd float %63, -1.000000e+00
>  %65 = fmul float %35, %64
>  %66 = fadd float %65, 1.000000e+00
>  %67 = fmul float %66, %28
>  %68 = fmul float %67, %50
>  %69 = fsub float %68, %67
>  %70 = fmul float %31, %69
>  %71 = fadd float %70, %67
>  %72 = call float @llvm.fabs.f32(float %30)
>  %73 = fmul float %42, %54
>  %74 = fcmp ole float %72, -0.000000e+00
>  %. = select i1 %74, float %71, float %73
>  %75 = fadd float %63, %29
>  %76 = call float @llvm.AMDIL.clamp.(float %75, float 0.000000e+00, float 1.000000e+00)
>  %77 = fadd float %25, -1.000000e+00
>  %78 = fadd float %26, -1.000000e+00
>  %79 = fadd float %27, -1.000000e+00
>  %80 = fmul float %76, %77
>  %81 = fadd float %80, 1.000000e+00
>  %82 = fmul float %76, %78
>  %83 = fadd float %82, 1.000000e+00
>  %84 = fmul float %76, %79
>  %85 = fadd float %84, 1.000000e+00
>  %86 = fmul float %60, %81
>  %87 = fmul float %61, %83
>  %88 = fmul float %62, %85
>  %89 = fmul float %86, %43
>  %90 = fmul float %87, %43
>  %91 = fmul float %88, %43
>  %92 = fmul float %43, %86
>  %93 = fsub float %39, %92
>  %94 = fmul float %43, %87
>  %95 = fsub float %40, %94
>  %96 = fmul float %43, %88
>  %97 = fsub float %41, %96
>  %98 = fsub float %32, %51
>  %99 = fsub float %33, %52
>  %100 = fsub float %34, %53
>  %101 = fmul float %98, %98
>  %102 = fmul float %99, %99
>  %103 = fadd float %102, %101
>  %104 = fmul float %100, %100
>  %105 = fadd float %103, %104
>  %106 = call float @llvm.sqrt.f32(float %105)
>  %107 = fmul float %106, %38
>  %108 = fadd float %107, %36
>  %109 = call float @llvm.AMDIL.clamp.(float %108, float 0.000000e+00, float 1.000000e+00)
>  %110 = call float @llvm.minnum.f32(float %109, float %37)
>  %111 = fmul float %110, %110
>  %112 = fmul float %111, %93
>  %113 = fadd float %112, %89
>  %114 = fmul float %111, %95
>  %115 = fadd float %114, %90
>  %116 = fmul float %111, %97
>  %117 = fadd float %116, %91
>  %118 = call i32 @llvm.SI.packf16(float %113, float %115)
>  %119 = bitcast i32 %118 to float
>  %120 = call i32 @llvm.SI.packf16(float %117, float %.)
>  %121 = bitcast i32 %120 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %119, float %121, float %119, float %121)
>  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.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.fabs.f32(float) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.clamp.(float, float, float) #2
>
>; Function Attrs: nounwind readnone
>declare float @llvm.sqrt.f32(float) #1
>
>; 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[0:3], s[2:3], 0x0                                      ; C0800300
>	s_waitcnt lgkmcnt(0)                                                    ; BF8C007F
>	s_buffer_load_dword s8, s[0:3], 0x4                                     ; C2040104
>	s_buffer_load_dword s9, s[0:3], 0x5                                     ; C2048105
>	s_buffer_load_dword s11, s[0:3], 0x6                                    ; C2058106
>	s_buffer_load_dword s12, s[0:3], 0x7                                    ; C2060107
>	s_buffer_load_dword s13, s[0:3], 0x30                                   ; C2068130
>	s_buffer_load_dword s14, s[0:3], 0x31                                   ; C2070131
>	s_buffer_load_dword s15, s[0:3], 0x33                                   ; C2078133
>	s_buffer_load_dword s16, s[0:3], 0x50                                   ; C2080150
>	s_buffer_load_dword s17, s[0:3], 0x51                                   ; C2088151
>	s_buffer_load_dword s18, s[0:3], 0x52                                   ; C2090152
>	s_buffer_load_dword s19, s[0:3], 0x53                                   ; C2098153
>	s_buffer_load_dword s20, s[0:3], 0x54                                   ; C20A0154
>	s_buffer_load_dword s21, s[0:3], 0x56                                   ; C20A8156
>	s_buffer_load_dword s22, s[0:3], 0x57                                   ; C20B0157
>	s_buffer_load_dword s23, s[0:3], 0x74                                   ; C20B8174
>	s_buffer_load_dword s24, s[0:3], 0x75                                   ; C20C0175
>	s_buffer_load_dword s25, s[0:3], 0x76                                   ; C20C8176
>	s_buffer_load_dword s26, s[0:3], 0x77                                   ; C20D0177
>	s_buffer_load_dword s0, s[0:3], 0x78                                    ; C2000178
>	s_mov_b32 m0, s10                                                       ; BEFC030A
>	s_load_dwordx8 s[28:35], s[6:7], 0x0                                    ; C0CE0700
>	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_interp_p1_f32 v4, v0, 3, 1, [m0]                                      ; C8100700
>	s_load_dwordx4 s[4:7], s[4:5], 0x0                                      ; C0820500
>	v_interp_p2_f32 v4, [v4], v1, 3, 1, [m0]                                ; C8110701
>	v_interp_p1_f32 v5, v0, 0, 2, [m0]                                      ; C8140800
>	v_interp_p2_f32 v5, [v5], v1, 0, 2, [m0]                                ; C8150801
>	v_interp_p1_f32 v6, v0, 1, 2, [m0]                                      ; C8180900
>	v_interp_p2_f32 v6, [v6], v1, 1, 2, [m0]                                ; C8190901
>	v_interp_p1_f32 v7, v0, 2, 2, [m0]                                      ; C81C0A00
>	v_interp_p2_f32 v7, [v7], v1, 2, 2, [m0]                                ; C81D0A01
>	v_interp_p1_f32 v0, v0, 3, 2, [m0]                                      ; C8000B00
>	v_interp_p2_f32 v0, [v0], v1, 3, 2, [m0]                                ; C8010B01
>	s_waitcnt lgkmcnt(0)                                                    ; BF8C007F
>	image_sample v[8:11], 15, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[28:35], s[4:7] ; F0800F00 00270802
>	s_waitcnt vmcnt(0)                                                      ; BF8C0770
>	v_mad_f32 v1, v11, s19, -s19                                            ; D2820001 804C270B
>	v_mad_f32 v1, v1, s12, s12                                              ; D2820001 00301901
>	v_mad_f32 v2, v1, v4, -v1                                               ; D2820002 84060901
>	v_mac_f32_e32 v1, s15, v2                                               ; 3E02040F
>	v_mul_f32_e32 v0, s26, v0                                               ; 1000001A
>	v_mov_b32_e32 v2, 0x80000000                                            ; 7E0402FF 80000000
>	v_cmp_le_f32_e64 vcc, |s14|, v2                                         ; D006016A 0002040E
>	v_cndmask_b32_e32 v0, v0, v1                                            ; 00000300
>	v_add_f32_e32 v1, s13, v11                                              ; 0602160D
>	v_add_f32_e64 v1, 0, v1 clamp                                           ; D2060801 00020280
>	v_mad_f32 v2, s8, v1, -v1                                               ; D2820002 84060208
>	v_mad_f32 v3, s9, v1, -v1                                               ; D2820003 84060209
>	v_mad_f32 v1, s11, v1, -v1                                              ; D2820001 8406020B
>	v_mad_f32 v2, v8, v2, v8                                                ; D2820002 04220508
>	v_mad_f32 v3, v9, v3, v9                                                ; D2820003 04260709
>	v_mad_f32 v1, v10, v1, v10                                              ; D2820001 042A030A
>	v_mul_f32_e32 v4, s0, v2                                                ; 10080400
>	v_mul_f32_e32 v8, s0, v3                                                ; 10100600
>	v_mul_f32_e32 v9, s0, v1                                                ; 10120200
>	v_mov_b32_e32 v10, s23                                                  ; 7E140217
>	v_mad_f32 v2, -v2, s0, v10                                              ; D2820002 24280102
>	v_mov_b32_e32 v10, s24                                                  ; 7E140218
>	v_mad_f32 v3, -v3, s0, v10                                              ; D2820003 24280103
>	v_mov_b32_e32 v10, s25                                                  ; 7E140219
>	v_mad_f32 v1, -v1, s0, v10                                              ; D2820001 24280101
>	v_sub_f32_e32 v5, s16, v5                                               ; 080A0A10
>	v_sub_f32_e32 v6, s17, v6                                               ; 080C0C11
>	v_sub_f32_e32 v7, s18, v7                                               ; 080E0E12
>	v_mul_f32_e32 v5, v5, v5                                                ; 100A0B05
>	v_mac_f32_e32 v5, v6, v6                                                ; 3E0A0D06
>	v_mac_f32_e32 v5, v7, v7                                                ; 3E0A0F07
>	v_sqrt_f32_e32 v5, v5                                                   ; 7E0A6705
>	v_mov_b32_e32 v6, s20                                                   ; 7E0C0214
>	v_mac_f32_e32 v6, s22, v5                                               ; 3E0C0A16
>	v_add_f32_e64 v5, 0, v6 clamp                                           ; D2060805 00020C80
>	v_min_f32_e32 v5, s21, v5                                               ; 1E0A0A15
>	v_mul_f32_e32 v5, v5, v5                                                ; 100A0B05
>	v_mac_f32_e32 v4, v2, v5                                                ; 3E080B02
>	v_mac_f32_e32 v8, v3, v5                                                ; 3E100B03
>	v_mac_f32_e32 v9, v1, v5                                                ; 3E120B01
>	v_cvt_pkrtz_f16_f32_e32 v1, v4, v8                                      ; 5E021104
>	v_cvt_pkrtz_f16_f32_e32 v0, v9, v0                                      ; 5E000109
>	exp 15, 0, 1, 1, 1, v1, v0, v1, v0                                      ; F8001C0F 00010001
>	s_endpgm                                                                ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 40
>VGPRS: 12
>Code Size: 416 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
>  clamp_color = 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([9 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, 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:
>  %23 = call float @llvm.SI.fs.constant(i32 0, i32 0, i32 %6)
>  %24 = call float @llvm.SI.fs.constant(i32 1, i32 0, i32 %6)
>  %25 = call float @llvm.SI.fs.constant(i32 2, i32 0, i32 %6)
>  %26 = call float @llvm.SI.fs.constant(i32 3, i32 0, i32 %6)
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 0, float %23, float %24, float %25, float %26)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.constant(i32, i32, 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 }
>
>Shader Disassembly:
>
>	s_mov_b32 m0, s10                   ; BEFC030A
>	v_interp_mov_f32 v0, P0, 0, 0, [m0] ; C8020002
>	v_interp_mov_f32 v1, P0, 1, 0, [m0] ; C8060102
>	v_interp_mov_f32 v2, P0, 2, 0, [m0] ; C80A0202
>	v_interp_mov_f32 v3, P0, 3, 0, [m0] ; C80E0302
>	exp 15, 0, 0, 1, 1, v0, v1, v2, v3  ; F800180F 03020100
>	s_endpgm                            ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 16
>VGPRS: 4
>Code Size: 32 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
>  clamp_color = 0
>FRAG
>DCL IN[0], GENERIC[0], LINEAR
>DCL OUT[0], COLOR
>DCL OUT[1], POSITION
>DCL SAMP[0]
>DCL SVIEW[0], 2D, FLOAT
>IMM[0] FLT32 {    0.0000,     1.0000,     0.0000,     0.0000}
>  0: MOV OUT[0], IMM[0].xxxy
>  1: TEX OUT[1].z, IN[0], SAMP[0], 2D
>  2: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %24 = load <8 x i32>, <8 x i32> addrspace(2)* %23, align 32, !tbaa !0
>  %25 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %26 = load <4 x i32>, <4 x i32> addrspace(2)* %25, align 16, !tbaa !0
>  %27 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %12)
>  %28 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %12)
>  %29 = bitcast float %27 to i32
>  %30 = bitcast float %28 to i32
>  %31 = insertelement <2 x i32> undef, i32 %29, i32 0
>  %32 = insertelement <2 x i32> %31, i32 %30, i32 1
>  %33 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %32, <8 x i32> %24, <4 x i32> %26, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %34 = extractelement <4 x float> %33, i32 2
>  call void @llvm.SI.export(i32 1, i32 0, i32 0, i32 8, i32 0, float %34, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 0, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float 1.000000e+00)
>  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.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, 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_load_dwordx8 s[12:19], s[6:7], 0x0                              ; C0C60700
>	s_load_dwordx4 s[0:3], s[4:5], 0x0                                ; C0800500
>	s_mov_b32 m0, s10                                                 ; BEFC030A
>	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_mov_b32_e32 v0, 0                                               ; 7E000280
>	v_interp_p2_f32 v3, [v3], v1, 1, 0, [m0]                          ; C80D0101
>	s_waitcnt lgkmcnt(0)                                              ; BF8C007F
>	image_sample v1, 4, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[12:19], s[0:3] ; F0800400 00030102
>	s_waitcnt vmcnt(0)                                                ; BF8C0770
>	exp 1, 8, 0, 0, 0, v1, v0, v0, v0                                 ; F8000081 00000001
>	s_waitcnt expcnt(0)                                               ; BF8C070F
>	v_mov_b32_e32 v1, 1.0                                             ; 7E0202F2
>	exp 15, 0, 0, 1, 1, v0, v0, v0, v1                                ; F800180F 01000000
>	s_endpgm                                                          ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 24
>VGPRS: 4
>Code Size: 80 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
>  clamp_color = 0
>FRAG
>DCL IN[0], GENERIC[0], LINEAR
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SVIEW[0], 3D, FLOAT
>  0: TEX OUT[0], IN[0], SAMP[0], 3D
>  1: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %24 = load <8 x i32>, <8 x i32> addrspace(2)* %23, align 32, !tbaa !0
>  %25 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %26 = load <4 x i32>, <4 x i32> addrspace(2)* %25, align 16, !tbaa !0
>  %27 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %12)
>  %28 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %12)
>  %29 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %6, <2 x i32> %12)
>  %30 = bitcast float %27 to i32
>  %31 = bitcast float %28 to i32
>  %32 = bitcast float %29 to i32
>  %33 = insertelement <4 x i32> undef, i32 %30, i32 0
>  %34 = insertelement <4 x i32> %33, i32 %31, i32 1
>  %35 = insertelement <4 x i32> %34, i32 %32, i32 2
>  %36 = call <4 x float> @llvm.SI.image.sample.v4i32(<4 x i32> %35, <8 x i32> %24, <4 x i32> %26, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %37 = extractelement <4 x float> %36, i32 0
>  %38 = extractelement <4 x float> %36, i32 1
>  %39 = extractelement <4 x float> %36, i32 2
>  %40 = extractelement <4 x float> %36, i32 3
>  %41 = call i32 @llvm.SI.packf16(float %37, float %38)
>  %42 = bitcast i32 %41 to float
>  %43 = call i32 @llvm.SI.packf16(float %39, 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.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.image.sample.v4i32(<4 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, 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 }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                   ; BEFE0A7E
>	s_load_dwordx8 s[12:19], s[6:7], 0x0                                   ; C0C60700
>	s_load_dwordx4 s[0:3], s[4:5], 0x0                                     ; C0800500
>	s_mov_b32 m0, s10                                                      ; BEFC030A
>	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_interp_p1_f32 v4, v0, 2, 0, [m0]                                     ; C8100200
>	v_interp_p2_f32 v4, [v4], v1, 2, 0, [m0]                               ; C8110201
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	image_sample v[0:3], 15, 0, 0, 0, 0, 0, 0, 0, v[2:5], s[12:19], s[0:3] ; F0800F00 00030002
>	s_waitcnt vmcnt(0)                                                     ; BF8C0770
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v1                                     ; 5E000300
>	v_cvt_pkrtz_f16_f32_e32 v1, v2, v3                                     ; 5E020702
>	exp 15, 0, 1, 1, 1, v0, v1, v0, v1                                     ; F8001C0F 01000100
>	s_endpgm                                                               ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 24
>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
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[0]
>DCL CONST[0..3]
>DCL TEMP[0], LOCAL
>IMM[0] FLT32 {    0.0000,     0.0000,     0.0000,     0.0000}
>  0: MUL TEMP[0], CONST[0], IN[0].xxxx
>  1: MAD TEMP[0], CONST[1], IN[0].yyyy, TEMP[0]
>  2: MAD TEMP[0], CONST[2], IN[0].zzzz, TEMP[0]
>  3: MAD TEMP[0], CONST[3], IN[0].wwww, TEMP[0]
>  4: MOV OUT[1], IMM[0].xxxx
>  5: MOV OUT[0], TEMP[0]
>  6: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = call float @llvm.SI.load.const(<16 x i8> %13, i32 0)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %13, i32 4)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %13, i32 8)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %13, i32 12)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %13, i32 16)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %13, i32 20)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %13, i32 24)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %13, i32 28)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %13, i32 32)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %13, i32 36)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %13, i32 40)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %13, i32 44)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %13, i32 48)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %13, i32 52)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %13, i32 56)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %13, i32 60)
>  %30 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %31 = load <16 x i8>, <16 x i8> addrspace(2)* %30, align 16, !tbaa !0
>  %32 = add i32 %5, %8
>  %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 = extractelement <4 x float> %33, i32 2
>  %37 = extractelement <4 x float> %33, i32 3
>  %38 = fmul float %14, %34
>  %39 = fmul float %15, %34
>  %40 = fmul float %16, %34
>  %41 = fmul float %17, %34
>  %42 = fmul float %18, %35
>  %43 = fadd float %42, %38
>  %44 = fmul float %19, %35
>  %45 = fadd float %44, %39
>  %46 = fmul float %20, %35
>  %47 = fadd float %46, %40
>  %48 = fmul float %21, %35
>  %49 = fadd float %48, %41
>  %50 = fmul float %22, %36
>  %51 = fadd float %50, %43
>  %52 = fmul float %23, %36
>  %53 = fadd float %52, %45
>  %54 = fmul float %24, %36
>  %55 = fadd float %54, %47
>  %56 = fmul float %25, %36
>  %57 = fadd float %56, %49
>  %58 = fmul float %26, %37
>  %59 = fadd float %58, %51
>  %60 = fmul float %27, %37
>  %61 = fadd float %60, %53
>  %62 = fmul float %28, %37
>  %63 = fadd float %62, %55
>  %64 = fmul float %29, %37
>  %65 = fadd float %64, %57
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %59, float %61, float %63, float %65)
>  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[4:7], s[8:9], 0x0                  ; C0820900
>	v_add_i32_e32 v0, vcc, s10, v0                      ; 4A00000A
>	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 lgkmcnt(0)                                ; BF8C007F
>	s_buffer_load_dword s4, s[0:3], 0x0                 ; C2020100
>	s_buffer_load_dword s5, s[0:3], 0x1                 ; C2028101
>	s_buffer_load_dword s6, s[0:3], 0x2                 ; C2030102
>	s_buffer_load_dword s7, s[0:3], 0x3                 ; C2038103
>	s_buffer_load_dword s8, s[0:3], 0x4                 ; C2040104
>	s_buffer_load_dword s9, s[0:3], 0x5                 ; C2048105
>	s_buffer_load_dword s10, s[0:3], 0x6                ; C2050106
>	s_buffer_load_dword s11, s[0:3], 0x7                ; C2058107
>	s_buffer_load_dword s12, s[0:3], 0x8                ; C2060108
>	s_buffer_load_dword s13, s[0:3], 0x9                ; C2068109
>	s_buffer_load_dword s14, s[0:3], 0xa                ; C207010A
>	s_buffer_load_dword s15, s[0:3], 0xb                ; C207810B
>	s_buffer_load_dword s16, s[0:3], 0xc                ; C208010C
>	s_buffer_load_dword s17, s[0:3], 0xd                ; C208810D
>	s_buffer_load_dword s18, s[0:3], 0xe                ; C209010E
>	s_buffer_load_dword s0, s[0:3], 0xf                 ; C200010F
>	s_waitcnt vmcnt(0) lgkmcnt(0)                       ; BF8C0070
>	v_mul_f32_e32 v4, s4, v0                            ; 10080004
>	v_mul_f32_e32 v5, s5, v0                            ; 100A0005
>	v_mul_f32_e32 v6, s6, v0                            ; 100C0006
>	v_mul_f32_e32 v0, s7, v0                            ; 10000007
>	v_mac_f32_e32 v4, s8, v1                            ; 3E080208
>	v_mac_f32_e32 v5, s9, v1                            ; 3E0A0209
>	v_mac_f32_e32 v6, s10, v1                           ; 3E0C020A
>	v_mac_f32_e32 v0, s11, v1                           ; 3E00020B
>	v_mac_f32_e32 v4, s12, v2                           ; 3E08040C
>	v_mac_f32_e32 v5, s13, v2                           ; 3E0A040D
>	v_mac_f32_e32 v6, s14, v2                           ; 3E0C040E
>	v_mac_f32_e32 v0, s15, v2                           ; 3E00040F
>	v_mac_f32_e32 v4, s16, v3                           ; 3E080610
>	v_mac_f32_e32 v5, s17, v3                           ; 3E0A0611
>	v_mac_f32_e32 v6, s18, v3                           ; 3E0C0612
>	v_mac_f32_e32 v0, s0, v3                            ; 3E000600
>	v_mov_b32_e32 v1, 0                                 ; 7E020280
>	exp 15, 32, 0, 0, 0, v1, v1, v1, v1                 ; F800020F 01010101
>	exp 15, 12, 0, 1, 0, v4, v5, v6, v0                 ; F80008CF 00060504
>	s_endpgm                                            ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 24
>VGPRS: 8
>Code Size: 184 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
>  clamp_color = 0
>FRAG
>PROPERTY FS_COLOR0_WRITES_ALL_CBUFS 1
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SVIEW[0], 3D, FLOAT
>DCL TEMP[0], LOCAL
>  0: MOV TEMP[0].xyz, IN[0].xyzz
>  1: TEX TEMP[0], TEMP[0], SAMP[0], 3D
>  2: MOV OUT[0], TEMP[0]
>  3: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %24 = load <8 x i32>, <8 x i32> addrspace(2)* %23, align 32, !tbaa !0
>  %25 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %26 = load <4 x i32>, <4 x i32> addrspace(2)* %25, align 16, !tbaa !0
>  %27 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %8)
>  %28 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %8)
>  %29 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %6, <2 x i32> %8)
>  %30 = bitcast float %27 to i32
>  %31 = bitcast float %28 to i32
>  %32 = bitcast float %29 to i32
>  %33 = insertelement <4 x i32> undef, i32 %30, i32 0
>  %34 = insertelement <4 x i32> %33, i32 %31, i32 1
>  %35 = insertelement <4 x i32> %34, i32 %32, i32 2
>  %36 = call <4 x float> @llvm.SI.image.sample.v4i32(<4 x i32> %35, <8 x i32> %24, <4 x i32> %26, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %37 = extractelement <4 x float> %36, i32 0
>  %38 = extractelement <4 x float> %36, i32 1
>  %39 = extractelement <4 x float> %36, i32 2
>  %40 = extractelement <4 x float> %36, i32 3
>  %41 = call i32 @llvm.SI.packf16(float %37, float %38)
>  %42 = bitcast i32 %41 to float
>  %43 = call i32 @llvm.SI.packf16(float %39, 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.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.image.sample.v4i32(<4 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, 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 }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                   ; BEFE0A7E
>	s_load_dwordx8 s[12:19], s[6:7], 0x0                                   ; C0C60700
>	s_load_dwordx4 s[0:3], s[4:5], 0x0                                     ; C0800500
>	s_mov_b32 m0, s10                                                      ; BEFC030A
>	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_interp_p1_f32 v4, v0, 2, 0, [m0]                                     ; C8100200
>	v_interp_p2_f32 v4, [v4], v1, 2, 0, [m0]                               ; C8110201
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	image_sample v[0:3], 15, 0, 0, 0, 0, 0, 0, 0, v[2:5], s[12:19], s[0:3] ; F0800F00 00030002
>	s_waitcnt vmcnt(0)                                                     ; BF8C0770
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v1                                     ; 5E000300
>	v_cvt_pkrtz_f16_f32_e32 v1, v2, v3                                     ; 5E020702
>	exp 15, 0, 1, 1, 1, v0, v1, v0, v1                                     ; F8001C0F 01000100
>	s_endpgm                                                               ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 24
>VGPRS: 8
>Code Size: 76 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>********************
>sh: ifconfig: Kommando nicht gefunden.
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[0]
>DCL OUT[2], GENERIC[1]
>DCL CONST[0..3]
>DCL TEMP[0..2], LOCAL
>IMM[0] FLT32 {    0.0000,     1.0000,     0.0000,     0.0000}
>  0: MOV TEMP[0].zw, IMM[0].yyxy
>  1: DP4 TEMP[0].x, IN[1], CONST[0]
>  2: DP4 TEMP[1].x, IN[1], CONST[1]
>  3: MOV TEMP[0].y, TEMP[1].xxxx
>  4: DP4 TEMP[1].x, IN[1], CONST[2]
>  5: DP4 TEMP[2].x, IN[1], CONST[3]
>  6: MOV TEMP[1].y, TEMP[2].xxxx
>  7: MOV TEMP[1].xy, TEMP[1].xyxx
>  8: MOV OUT[1], IN[0]
>  9: MOV OUT[0], TEMP[0]
> 10: MOV OUT[2], TEMP[1]
> 11: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = call float @llvm.SI.load.const(<16 x i8> %13, i32 0)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %13, i32 4)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %13, i32 8)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %13, i32 12)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %13, i32 16)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %13, i32 20)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %13, i32 24)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %13, i32 28)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %13, i32 32)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %13, i32 36)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %13, i32 40)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %13, i32 44)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %13, i32 48)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %13, i32 52)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %13, i32 56)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %13, i32 60)
>  %30 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %31 = load <16 x i8>, <16 x i8> addrspace(2)* %30, align 16, !tbaa !0
>  %32 = add i32 %5, %8
>  %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 = extractelement <4 x float> %33, i32 2
>  %37 = extractelement <4 x float> %33, i32 3
>  %38 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 1
>  %39 = load <16 x i8>, <16 x i8> addrspace(2)* %38, align 16, !tbaa !0
>  %40 = add i32 %5, %8
>  %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 = extractelement <4 x float> %41, i32 2
>  %45 = extractelement <4 x float> %41, i32 3
>  %46 = fmul float %42, %14
>  %47 = fmul float %43, %15
>  %48 = fadd float %46, %47
>  %49 = fmul float %44, %16
>  %50 = fadd float %48, %49
>  %51 = fmul float %45, %17
>  %52 = fadd float %50, %51
>  %53 = fmul float %42, %18
>  %54 = fmul float %43, %19
>  %55 = fadd float %53, %54
>  %56 = fmul float %44, %20
>  %57 = fadd float %55, %56
>  %58 = fmul float %45, %21
>  %59 = fadd float %57, %58
>  %60 = fmul float %42, %22
>  %61 = fmul float %43, %23
>  %62 = fadd float %60, %61
>  %63 = fmul float %44, %24
>  %64 = fadd float %62, %63
>  %65 = fmul float %45, %25
>  %66 = fadd float %64, %65
>  %67 = fmul float %42, %26
>  %68 = fmul float %43, %27
>  %69 = fadd float %67, %68
>  %70 = fmul float %44, %28
>  %71 = fadd float %69, %70
>  %72 = fmul float %45, %29
>  %73 = fadd float %71, %72
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %34, float %35, float %36, float %37)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %66, float %73, float undef, float undef)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %52, float %59, 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_load_dwordx4 s[4:7], s[8:9], 0x0                    ; C0820900
>	s_load_dwordx4 s[12:15], s[8:9], 0x4                  ; C0860904
>	v_add_i32_e32 v0, vcc, s10, v0                        ; 4A00000A
>	v_mov_b32_e32 v1, 1.0                                 ; 7E0202F2
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	s_buffer_load_dword s8, s[0:3], 0x5                   ; C2040105
>	s_buffer_load_dword s9, s[0:3], 0x6                   ; C2048106
>	s_buffer_load_dword s10, s[0:3], 0x7                  ; C2050107
>	s_buffer_load_dword s11, s[0:3], 0x8                  ; C2058108
>	s_buffer_load_dword s16, s[0:3], 0x9                  ; C2080109
>	buffer_load_format_xyzw v[2:5], v0, s[4:7], 0 idxen   ; E00C2000 80010200
>	buffer_load_format_xyzw v[6:9], v0, s[12:15], 0 idxen ; E00C2000 80030600
>	s_buffer_load_dword s4, s[0:3], 0xa                   ; C202010A
>	s_buffer_load_dword s5, s[0:3], 0xb                   ; C202810B
>	s_buffer_load_dword s6, s[0:3], 0xc                   ; C203010C
>	s_buffer_load_dword s7, s[0:3], 0xd                   ; C203810D
>	s_buffer_load_dword s12, s[0:3], 0xe                  ; C206010E
>	s_buffer_load_dword s13, s[0:3], 0x0                  ; C2068100
>	s_buffer_load_dword s14, s[0:3], 0x1                  ; C2070101
>	s_buffer_load_dword s15, s[0:3], 0x2                  ; C2078102
>	s_buffer_load_dword s17, s[0:3], 0x3                  ; C2088103
>	s_buffer_load_dword s18, s[0:3], 0x4                  ; C2090104
>	s_buffer_load_dword s0, s[0:3], 0xf                   ; C200010F
>	s_waitcnt vmcnt(1)                                    ; BF8C0771
>	exp 15, 32, 0, 0, 0, v2, v3, v4, v5                   ; F800020F 05040302
>	s_waitcnt vmcnt(0) lgkmcnt(0)                         ; BF8C0070
>	v_mul_f32_e32 v0, s14, v7                             ; 10000E0E
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mul_f32_e32 v2, s8, v7                              ; 10040E08
>	v_mul_f32_e32 v3, s16, v7                             ; 10060E10
>	v_mul_f32_e32 v4, s7, v7                              ; 10080E07
>	v_mac_f32_e32 v0, s13, v6                             ; 3E000C0D
>	v_mac_f32_e32 v2, s18, v6                             ; 3E040C12
>	v_mac_f32_e32 v3, s11, v6                             ; 3E060C0B
>	v_mac_f32_e32 v4, s6, v6                              ; 3E080C06
>	v_mac_f32_e32 v0, s15, v8                             ; 3E00100F
>	v_mac_f32_e32 v2, s9, v8                              ; 3E041009
>	v_mac_f32_e32 v3, s4, v8                              ; 3E061004
>	v_mac_f32_e32 v4, s12, v8                             ; 3E08100C
>	v_mac_f32_e32 v0, s17, v9                             ; 3E001211
>	v_mac_f32_e32 v2, s10, v9                             ; 3E04120A
>	v_mac_f32_e32 v3, s5, v9                              ; 3E061205
>	v_mac_f32_e32 v4, s0, v9                              ; 3E081200
>	exp 15, 33, 0, 0, 0, v3, v4, v0, v0                   ; F800021F 00000403
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mov_b32_e32 v3, 0                                   ; 7E060280
>	exp 15, 12, 0, 1, 0, v0, v2, v3, v1                   ; F80008CF 01030200
>	s_endpgm                                              ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 24
>VGPRS: 12
>Code Size: 216 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
>  clamp_color = 0
>FRAG
>PROPERTY FS_COLOR0_WRITES_ALL_CBUFS 1
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL IN[1], GENERIC[1], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SVIEW[0], 2D, FLOAT
>DCL TEMP[0..1], LOCAL
>  0: MOV TEMP[0].xy, IN[1].xyyy
>  1: TEX TEMP[0], TEMP[0], SAMP[0], 2D
>  2: MOV TEMP[1].xyz, TEMP[0].xyzx
>  3: MUL TEMP[0].x, TEMP[0].wwww, IN[0].wwww
>  4: MOV TEMP[1].w, TEMP[0].xxxx
>  5: MOV OUT[0], TEMP[1]
>  6: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %24 = load <8 x i32>, <8 x i32> addrspace(2)* %23, align 32, !tbaa !0
>  %25 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %26 = load <4 x i32>, <4 x i32> addrspace(2)* %25, align 16, !tbaa !0
>  %27 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %6, <2 x i32> %8)
>  %28 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %6, <2 x i32> %8)
>  %29 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %6, <2 x i32> %8)
>  %30 = bitcast float %28 to i32
>  %31 = bitcast float %29 to i32
>  %32 = insertelement <2 x i32> undef, i32 %30, i32 0
>  %33 = insertelement <2 x i32> %32, i32 %31, i32 1
>  %34 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %33, <8 x i32> %24, <4 x i32> %26, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %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 = fmul float %38, %27
>  %40 = call i32 @llvm.SI.packf16(float %35, float %36)
>  %41 = bitcast i32 %40 to float
>  %42 = call i32 @llvm.SI.packf16(float %37, float %39)
>  %43 = bitcast i32 %42 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %41, float %43, float %41, float %43)
>  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.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, 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 }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                   ; BEFE0A7E
>	s_load_dwordx8 s[12:19], s[6:7], 0x0                                   ; C0C60700
>	s_load_dwordx4 s[0:3], s[4:5], 0x0                                     ; C0800500
>	s_mov_b32 m0, s10                                                      ; BEFC030A
>	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, 0, 1, [m0]                                     ; C80C0400
>	v_interp_p2_f32 v3, [v3], v1, 0, 1, [m0]                               ; C80D0401
>	v_interp_p1_f32 v4, v0, 1, 1, [m0]                                     ; C8100500
>	v_interp_p2_f32 v4, [v4], v1, 1, 1, [m0]                               ; C8110501
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	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_mul_f32_e32 v0, v2, v6                                               ; 10000D02
>	v_cvt_pkrtz_f16_f32_e32 v1, v3, v4                                     ; 5E020903
>	v_cvt_pkrtz_f16_f32_e32 v0, v5, v0                                     ; 5E000105
>	exp 15, 0, 1, 1, 1, v1, v0, v1, v0                                     ; F8001C0F 00010001
>	s_endpgm                                                               ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 24
>VGPRS: 8
>Code Size: 80 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
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[0]
>DCL OUT[2], GENERIC[1]
>DCL OUT[3], GENERIC[2]
>DCL OUT[4], GENERIC[3]
>DCL CONST[0..5]
>DCL TEMP[0..2], LOCAL
>IMM[0] FLT32 {    0.0000,     1.0000,     0.0000,     0.0000}
>  0: MOV TEMP[0].zw, IMM[0].yyxy
>  1: DP4 TEMP[0].x, IN[1], CONST[2]
>  2: DP4 TEMP[1].x, IN[1], CONST[3]
>  3: MOV TEMP[0].y, TEMP[1].xxxx
>  4: DP4 TEMP[1].x, IN[1], CONST[4]
>  5: DP4 TEMP[2].x, IN[1], CONST[5]
>  6: MOV TEMP[1].y, TEMP[2].xxxx
>  7: MOV TEMP[1].xy, TEMP[1].xyxx
>  8: MOV OUT[1], IN[0]
>  9: MOV OUT[2], CONST[0]
> 10: MOV OUT[3], CONST[1]
> 11: MOV OUT[0], TEMP[0]
> 12: MOV OUT[4], TEMP[1]
> 13: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = call float @llvm.SI.load.const(<16 x i8> %13, i32 0)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %13, i32 4)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %13, i32 8)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %13, i32 12)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %13, i32 16)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %13, i32 20)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %13, i32 24)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %13, i32 28)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %13, i32 32)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %13, i32 36)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %13, i32 40)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %13, i32 44)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %13, i32 48)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %13, i32 52)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %13, i32 56)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %13, i32 60)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %13, i32 64)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %13, i32 68)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %13, i32 72)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %13, i32 76)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %13, i32 80)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %13, i32 84)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %13, i32 88)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %13, i32 92)
>  %38 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %39 = load <16 x i8>, <16 x i8> addrspace(2)* %38, align 16, !tbaa !0
>  %40 = add i32 %5, %8
>  %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 = extractelement <4 x float> %41, i32 2
>  %45 = extractelement <4 x float> %41, i32 3
>  %46 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 1
>  %47 = load <16 x i8>, <16 x i8> addrspace(2)* %46, align 16, !tbaa !0
>  %48 = add i32 %5, %8
>  %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 = extractelement <4 x float> %49, i32 3
>  %54 = fmul float %50, %22
>  %55 = fmul float %51, %23
>  %56 = fadd float %54, %55
>  %57 = fmul float %52, %24
>  %58 = fadd float %56, %57
>  %59 = fmul float %53, %25
>  %60 = fadd float %58, %59
>  %61 = fmul float %50, %26
>  %62 = fmul float %51, %27
>  %63 = fadd float %61, %62
>  %64 = fmul float %52, %28
>  %65 = fadd float %63, %64
>  %66 = fmul float %53, %29
>  %67 = fadd float %65, %66
>  %68 = fmul float %50, %30
>  %69 = fmul float %51, %31
>  %70 = fadd float %68, %69
>  %71 = fmul float %52, %32
>  %72 = fadd float %70, %71
>  %73 = fmul float %53, %33
>  %74 = fadd float %72, %73
>  %75 = fmul float %50, %34
>  %76 = fmul float %51, %35
>  %77 = fadd float %75, %76
>  %78 = fmul float %52, %36
>  %79 = fadd float %77, %78
>  %80 = fmul float %53, %37
>  %81 = fadd float %79, %80
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %42, float %43, float %44, float %45)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %14, float %15, float %16, float %17)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 34, i32 0, float %18, float %19, float %20, float %21)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 35, i32 0, float %74, float %81, float undef, float undef)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %60, float %67, 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_load_dwordx4 s[4:7], s[8:9], 0x0                    ; C0820900
>	s_load_dwordx4 s[12:15], s[8:9], 0x4                  ; C0860904
>	v_add_i32_e32 v0, vcc, s10, v0                        ; 4A00000A
>	v_mov_b32_e32 v1, 1.0                                 ; 7E0202F2
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	s_buffer_load_dword s8, s[0:3], 0xf                   ; C204010F
>	s_buffer_load_dword s9, s[0:3], 0x10                  ; C2048110
>	s_buffer_load_dword s10, s[0:3], 0x11                 ; C2050111
>	s_buffer_load_dword s11, s[0:3], 0x12                 ; C2058112
>	s_buffer_load_dword s16, s[0:3], 0x13                 ; C2080113
>	buffer_load_format_xyzw v[2:5], v0, s[4:7], 0 idxen   ; E00C2000 80010200
>	buffer_load_format_xyzw v[6:9], v0, s[12:15], 0 idxen ; E00C2000 80030600
>	s_buffer_load_dword s4, s[0:3], 0x5                   ; C2020105
>	s_buffer_load_dword s5, s[0:3], 0x6                   ; C2028106
>	s_buffer_load_dword s6, s[0:3], 0x7                   ; C2030107
>	s_buffer_load_dword s7, s[0:3], 0x8                   ; C2038108
>	s_buffer_load_dword s12, s[0:3], 0x9                  ; C2060109
>	s_buffer_load_dword s13, s[0:3], 0xa                  ; C206810A
>	s_buffer_load_dword s14, s[0:3], 0xb                  ; C207010B
>	s_buffer_load_dword s15, s[0:3], 0x0                  ; C2078100
>	s_buffer_load_dword s17, s[0:3], 0x1                  ; C2088101
>	s_buffer_load_dword s18, s[0:3], 0x2                  ; C2090102
>	s_buffer_load_dword s19, s[0:3], 0x3                  ; C2098103
>	s_buffer_load_dword s20, s[0:3], 0x4                  ; C20A0104
>	s_buffer_load_dword s21, s[0:3], 0xd                  ; C20A810D
>	s_buffer_load_dword s22, s[0:3], 0xc                  ; C20B010C
>	s_buffer_load_dword s23, s[0:3], 0x14                 ; C20B8114
>	s_buffer_load_dword s24, s[0:3], 0x15                 ; C20C0115
>	s_buffer_load_dword s25, s[0:3], 0xe                  ; C20C810E
>	s_buffer_load_dword s26, s[0:3], 0x16                 ; C20D0116
>	s_buffer_load_dword s0, s[0:3], 0x17                  ; C2000117
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mov_b32_e32 v0, s15                                 ; 7E00020F
>	v_mov_b32_e32 v10, s17                                ; 7E140211
>	v_mov_b32_e32 v11, s18                                ; 7E160212
>	s_waitcnt vmcnt(1)                                    ; BF8C0771
>	exp 15, 32, 0, 0, 0, v2, v3, v4, v5                   ; F800020F 05040302
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mov_b32_e32 v2, s19                                 ; 7E040213
>	v_mov_b32_e32 v3, s20                                 ; 7E060214
>	v_mov_b32_e32 v4, s4                                  ; 7E080204
>	v_mov_b32_e32 v5, s5                                  ; 7E0A0205
>	exp 15, 33, 0, 0, 0, v0, v10, v11, v2                 ; F800021F 020B0A00
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mov_b32_e32 v0, s6                                  ; 7E000206
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mul_f32_e32 v2, s12, v7                             ; 10040E0C
>	v_mul_f32_e32 v10, s21, v7                            ; 10140E15
>	v_mul_f32_e32 v11, s10, v7                            ; 10160E0A
>	exp 15, 34, 0, 0, 0, v3, v4, v5, v0                   ; F800022F 00050403
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mul_f32_e32 v0, s24, v7                             ; 10000E18
>	v_mac_f32_e32 v2, s7, v6                              ; 3E040C07
>	v_mac_f32_e32 v10, s22, v6                            ; 3E140C16
>	v_mac_f32_e32 v11, s9, v6                             ; 3E160C09
>	v_mac_f32_e32 v0, s23, v6                             ; 3E000C17
>	v_mac_f32_e32 v2, s13, v8                             ; 3E04100D
>	v_mac_f32_e32 v10, s25, v8                            ; 3E141019
>	v_mac_f32_e32 v11, s11, v8                            ; 3E16100B
>	v_mac_f32_e32 v0, s26, v8                             ; 3E00101A
>	v_mac_f32_e32 v2, s14, v9                             ; 3E04120E
>	v_mac_f32_e32 v10, s8, v9                             ; 3E141208
>	v_mac_f32_e32 v11, s16, v9                            ; 3E161210
>	v_mac_f32_e32 v0, s0, v9                              ; 3E001200
>	exp 15, 35, 0, 0, 0, v11, v0, v0, v0                  ; F800023F 0000000B
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mov_b32_e32 v0, 0                                   ; 7E000280
>	exp 15, 12, 0, 1, 0, v2, v10, v0, v1                  ; F80008CF 01000A02
>	s_endpgm                                              ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 32
>VGPRS: 12
>Code Size: 308 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
>  clamp_color = 0
>FRAG
>PROPERTY FS_COLOR0_WRITES_ALL_CBUFS 1
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL IN[1], GENERIC[1], PERSPECTIVE
>DCL IN[2], GENERIC[2], PERSPECTIVE
>DCL IN[3], GENERIC[3], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SVIEW[0], 2D, FLOAT
>DCL TEMP[0..1], LOCAL
>  0: MOV TEMP[0].xy, IN[3].xyyy
>  1: TEX TEMP[0], TEMP[0], SAMP[0], 2D
>  2: MAD TEMP[0], TEMP[0], IN[2], IN[1]
>  3: MUL TEMP[1].x, TEMP[0].wwww, IN[0].wwww
>  4: MOV TEMP[0].w, TEMP[1].xxxx
>  5: MOV OUT[0], TEMP[0]
>  6: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %24 = load <8 x i32>, <8 x i32> addrspace(2)* %23, align 32, !tbaa !0
>  %25 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %26 = load <4 x i32>, <4 x i32> addrspace(2)* %25, align 16, !tbaa !0
>  %27 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %6, <2 x i32> %8)
>  %28 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %6, <2 x i32> %8)
>  %29 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %6, <2 x i32> %8)
>  %30 = call float @llvm.SI.fs.interp(i32 2, i32 1, i32 %6, <2 x i32> %8)
>  %31 = call float @llvm.SI.fs.interp(i32 3, i32 1, i32 %6, <2 x i32> %8)
>  %32 = call float @llvm.SI.fs.interp(i32 0, i32 2, i32 %6, <2 x i32> %8)
>  %33 = call float @llvm.SI.fs.interp(i32 1, i32 2, i32 %6, <2 x i32> %8)
>  %34 = call float @llvm.SI.fs.interp(i32 2, i32 2, i32 %6, <2 x i32> %8)
>  %35 = call float @llvm.SI.fs.interp(i32 3, i32 2, i32 %6, <2 x i32> %8)
>  %36 = call float @llvm.SI.fs.interp(i32 0, i32 3, i32 %6, <2 x i32> %8)
>  %37 = call float @llvm.SI.fs.interp(i32 1, i32 3, i32 %6, <2 x i32> %8)
>  %38 = bitcast float %36 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 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %41, <8 x i32> %24, <4 x i32> %26, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %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 = fmul float %43, %32
>  %48 = fadd float %47, %28
>  %49 = fmul float %44, %33
>  %50 = fadd float %49, %29
>  %51 = fmul float %45, %34
>  %52 = fadd float %51, %30
>  %53 = fmul float %46, %35
>  %54 = fadd float %53, %31
>  %55 = fmul float %54, %27
>  %56 = call i32 @llvm.SI.packf16(float %48, float %50)
>  %57 = bitcast i32 %56 to float
>  %58 = call i32 @llvm.SI.packf16(float %52, float %55)
>  %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.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, 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 }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                      ; BEFE0A7E
>	s_mov_b32 m0, s10                                                         ; BEFC030A
>	s_load_dwordx8 s[8:15], s[6:7], 0x0                                       ; C0C40700
>	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, 0, 1, [m0]                                        ; C80C0400
>	v_interp_p2_f32 v3, [v3], v1, 0, 1, [m0]                                  ; C80D0401
>	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 v5, v0, 2, 1, [m0]                                        ; C8140600
>	v_interp_p2_f32 v5, [v5], v1, 2, 1, [m0]                                  ; C8150601
>	v_interp_p1_f32 v6, v0, 3, 1, [m0]                                        ; C8180700
>	v_interp_p2_f32 v6, [v6], v1, 3, 1, [m0]                                  ; C8190701
>	v_interp_p1_f32 v7, v0, 0, 2, [m0]                                        ; C81C0800
>	v_interp_p2_f32 v7, [v7], v1, 0, 2, [m0]                                  ; C81D0801
>	v_interp_p1_f32 v8, v0, 1, 2, [m0]                                        ; C8200900
>	s_load_dwordx4 s[0:3], s[4:5], 0x0                                        ; C0800500
>	v_interp_p2_f32 v8, [v8], v1, 1, 2, [m0]                                  ; C8210901
>	v_interp_p1_f32 v9, v0, 2, 2, [m0]                                        ; C8240A00
>	v_interp_p2_f32 v9, [v9], v1, 2, 2, [m0]                                  ; C8250A01
>	v_interp_p1_f32 v10, v0, 3, 2, [m0]                                       ; C8280B00
>	v_interp_p2_f32 v10, [v10], v1, 3, 2, [m0]                                ; C8290B01
>	v_interp_p1_f32 v11, v0, 0, 3, [m0]                                       ; C82C0C00
>	v_interp_p2_f32 v11, [v11], v1, 0, 3, [m0]                                ; C82D0C01
>	v_interp_p1_f32 v12, v0, 1, 3, [m0]                                       ; C8300D00
>	v_interp_p2_f32 v12, [v12], v1, 1, 3, [m0]                                ; C8310D01
>	s_waitcnt lgkmcnt(0)                                                      ; BF8C007F
>	image_sample v[11:14], 15, 0, 0, 0, 0, 0, 0, 0, v[11:12], s[8:15], s[0:3] ; F0800F00 00020B0B
>	s_waitcnt vmcnt(0)                                                        ; BF8C0770
>	v_mac_f32_e32 v3, v7, v11                                                 ; 3E061707
>	v_mac_f32_e32 v4, v8, v12                                                 ; 3E081908
>	v_mac_f32_e32 v5, v9, v13                                                 ; 3E0A1B09
>	v_mac_f32_e32 v6, v10, v14                                                ; 3E0C1D0A
>	v_mul_f32_e32 v0, v2, v6                                                  ; 10000D02
>	v_cvt_pkrtz_f16_f32_e32 v1, v3, v4                                        ; 5E020903
>	v_cvt_pkrtz_f16_f32_e32 v0, v5, v0                                        ; 5E000105
>	exp 15, 0, 1, 1, 1, v1, v0, v1, v0                                        ; F8001C0F 00010001
>	s_endpgm                                                                  ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 16
>VGPRS: 16
>Code Size: 160 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
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL IN[2]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[0]
>DCL OUT[2], GENERIC[1]
>DCL OUT[3], GENERIC[2]
>DCL OUT[4], GENERIC[3]
>DCL CONST[0..3]
>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: DP4 TEMP[0].x, IN[2], CONST[2]
>  2: DP4 TEMP[1].x, IN[2], CONST[3]
>  3: MOV TEMP[0].y, TEMP[1].xxxx
>  4: MOV OUT[2], IN[1]
>  5: MOV OUT[3], CONST[0]
>  6: MOV OUT[1], IN[0]
>  7: MOV OUT[4], CONST[1]
>  8: MOV OUT[0], TEMP[0]
>  9: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = call float @llvm.SI.load.const(<16 x i8> %13, i32 0)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %13, i32 4)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %13, i32 8)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %13, i32 12)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %13, i32 16)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %13, i32 20)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %13, i32 24)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %13, i32 28)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %13, i32 32)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %13, i32 36)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %13, i32 40)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %13, i32 44)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %13, i32 48)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %13, i32 52)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %13, i32 56)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %13, i32 60)
>  %30 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %31 = load <16 x i8>, <16 x i8> addrspace(2)* %30, align 16, !tbaa !0
>  %32 = add i32 %5, %8
>  %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 = extractelement <4 x float> %33, i32 2
>  %37 = extractelement <4 x float> %33, i32 3
>  %38 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 1
>  %39 = load <16 x i8>, <16 x i8> addrspace(2)* %38, align 16, !tbaa !0
>  %40 = add i32 %5, %8
>  %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 = extractelement <4 x float> %41, i32 2
>  %45 = extractelement <4 x float> %41, i32 3
>  %46 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 2
>  %47 = load <16 x i8>, <16 x i8> addrspace(2)* %46, align 16, !tbaa !0
>  %48 = add i32 %5, %8
>  %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 = extractelement <4 x float> %49, i32 3
>  %54 = fmul float %50, %22
>  %55 = fmul float %51, %23
>  %56 = fadd float %54, %55
>  %57 = fmul float %52, %24
>  %58 = fadd float %56, %57
>  %59 = fmul float %53, %25
>  %60 = fadd float %58, %59
>  %61 = fmul float %50, %26
>  %62 = fmul float %51, %27
>  %63 = fadd float %61, %62
>  %64 = fmul float %52, %28
>  %65 = fadd float %63, %64
>  %66 = fmul float %53, %29
>  %67 = fadd float %65, %66
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %34, float %35, float %36, float %37)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %42, float %43, float %44, float %45)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 34, i32 0, float %14, float %15, float %16, float %17)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 35, i32 0, float %18, float %19, float %20, float %21)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %60, float %67, 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_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
>	v_add_i32_e32 v0, vcc, s10, v0                          ; 4A00000A
>	v_mov_b32_e32 v1, 1.0                                   ; 7E0202F2
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	s_buffer_load_dword s8, s[0:3], 0x5                     ; C2040105
>	s_buffer_load_dword s9, s[0:3], 0x6                     ; C2048106
>	s_buffer_load_dword s10, s[0:3], 0x7                    ; C2050107
>	s_buffer_load_dword s11, s[0:3], 0x8                    ; C2058108
>	s_buffer_load_dword s20, s[0:3], 0x9                    ; C20A0109
>	buffer_load_format_xyzw v[2:5], v0, s[4:7], 0 idxen     ; E00C2000 80010200
>	buffer_load_format_xyzw v[6:9], v0, s[12:15], 0 idxen   ; E00C2000 80030600
>	buffer_load_format_xyzw v[10:13], v0, s[16:19], 0 idxen ; E00C2000 80040A00
>	s_buffer_load_dword s4, s[0:3], 0xa                     ; C202010A
>	s_buffer_load_dword s5, s[0:3], 0xb                     ; C202810B
>	s_buffer_load_dword s6, s[0:3], 0xc                     ; C203010C
>	s_buffer_load_dword s7, s[0:3], 0xd                     ; C203810D
>	s_buffer_load_dword s12, s[0:3], 0xe                    ; C206010E
>	s_buffer_load_dword s13, s[0:3], 0x0                    ; C2068100
>	s_buffer_load_dword s14, s[0:3], 0x1                    ; C2070101
>	s_buffer_load_dword s15, s[0:3], 0x2                    ; C2078102
>	s_buffer_load_dword s16, s[0:3], 0x3                    ; C2080103
>	s_buffer_load_dword s17, s[0:3], 0x4                    ; C2088104
>	s_buffer_load_dword s0, s[0:3], 0xf                     ; C200010F
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	v_mov_b32_e32 v0, s8                                    ; 7E000208
>	v_mov_b32_e32 v14, s13                                  ; 7E1C020D
>	v_mov_b32_e32 v15, s14                                  ; 7E1E020E
>	v_mov_b32_e32 v16, s15                                  ; 7E20020F
>	v_mov_b32_e32 v17, s16                                  ; 7E220210
>	v_mov_b32_e32 v18, s17                                  ; 7E240211
>	s_waitcnt vmcnt(2)                                      ; BF8C0772
>	exp 15, 32, 0, 0, 0, v2, v3, v4, v5                     ; F800020F 05040302
>	s_waitcnt vmcnt(1)                                      ; BF8C0771
>	exp 15, 33, 0, 0, 0, v6, v7, v8, v9                     ; F800021F 09080706
>	s_waitcnt expcnt(0)                                     ; BF8C070F
>	v_mov_b32_e32 v2, s9                                    ; 7E040209
>	v_mov_b32_e32 v3, s10                                   ; 7E06020A
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	v_mul_f32_e32 v4, s20, v11                              ; 10081614
>	exp 15, 34, 0, 0, 0, v14, v15, v16, v17                 ; F800022F 11100F0E
>	v_mul_f32_e32 v5, s7, v11                               ; 100A1607
>	v_mac_f32_e32 v4, s11, v10                              ; 3E08140B
>	v_mac_f32_e32 v5, s6, v10                               ; 3E0A1406
>	v_mac_f32_e32 v4, s4, v12                               ; 3E081804
>	v_mac_f32_e32 v5, s12, v12                              ; 3E0A180C
>	v_mac_f32_e32 v4, s5, v13                               ; 3E081A05
>	v_mac_f32_e32 v5, s0, v13                               ; 3E0A1A00
>	exp 15, 35, 0, 0, 0, v18, v0, v2, v3                    ; F800023F 03020012
>	s_waitcnt expcnt(0)                                     ; BF8C070F
>	v_mov_b32_e32 v0, 0                                     ; 7E000280
>	exp 15, 12, 0, 1, 0, v4, v5, v0, v1                     ; F80008CF 01000504
>	s_endpgm                                                ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 24
>VGPRS: 20
>Code Size: 252 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
>  clamp_color = 0
>FRAG
>PROPERTY FS_COLOR0_WRITES_ALL_CBUFS 1
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL IN[1], GENERIC[1], PERSPECTIVE
>DCL IN[2], GENERIC[2], PERSPECTIVE
>DCL IN[3], GENERIC[3], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL TEMP[0..1], LOCAL
>  0: MAD TEMP[0], IN[0], IN[3], IN[2]
>  1: MUL TEMP[1].x, TEMP[0].wwww, IN[1].wwww
>  2: MOV TEMP[0].w, TEMP[1].xxxx
>  3: MOV OUT[0], TEMP[0]
>  4: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %8)
>  %24 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %8)
>  %25 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %6, <2 x i32> %8)
>  %26 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %6, <2 x i32> %8)
>  %27 = call float @llvm.SI.fs.interp(i32 3, i32 1, i32 %6, <2 x i32> %8)
>  %28 = call float @llvm.SI.fs.interp(i32 0, i32 2, i32 %6, <2 x i32> %8)
>  %29 = call float @llvm.SI.fs.interp(i32 1, i32 2, i32 %6, <2 x i32> %8)
>  %30 = call float @llvm.SI.fs.interp(i32 2, i32 2, i32 %6, <2 x i32> %8)
>  %31 = call float @llvm.SI.fs.interp(i32 3, i32 2, i32 %6, <2 x i32> %8)
>  %32 = call float @llvm.SI.fs.interp(i32 0, i32 3, i32 %6, <2 x i32> %8)
>  %33 = call float @llvm.SI.fs.interp(i32 1, i32 3, i32 %6, <2 x i32> %8)
>  %34 = call float @llvm.SI.fs.interp(i32 2, i32 3, i32 %6, <2 x i32> %8)
>  %35 = call float @llvm.SI.fs.interp(i32 3, i32 3, i32 %6, <2 x i32> %8)
>  %36 = fmul float %23, %32
>  %37 = fadd float %36, %28
>  %38 = fmul float %24, %33
>  %39 = fadd float %38, %29
>  %40 = fmul float %25, %34
>  %41 = fadd float %40, %30
>  %42 = fmul float %26, %35
>  %43 = fadd float %42, %31
>  %44 = fmul float %43, %27
>  %45 = call i32 @llvm.SI.packf16(float %37, float %39)
>  %46 = bitcast i32 %45 to float
>  %47 = call i32 @llvm.SI.packf16(float %41, float %44)
>  %48 = bitcast i32 %47 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %46, float %48, float %46, float %48)
>  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_mov_b32 m0, s10                          ; BEFC030A
>	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_interp_p1_f32 v4, v0, 2, 0, [m0]         ; C8100200
>	v_interp_p2_f32 v4, [v4], v1, 2, 0, [m0]   ; C8110201
>	v_interp_p1_f32 v5, v0, 3, 0, [m0]         ; C8140300
>	v_interp_p2_f32 v5, [v5], v1, 3, 0, [m0]   ; C8150301
>	v_interp_p1_f32 v6, v0, 3, 1, [m0]         ; C8180700
>	v_interp_p2_f32 v6, [v6], v1, 3, 1, [m0]   ; C8190701
>	v_interp_p1_f32 v7, v0, 0, 2, [m0]         ; C81C0800
>	v_interp_p2_f32 v7, [v7], v1, 0, 2, [m0]   ; C81D0801
>	v_interp_p1_f32 v8, v0, 1, 2, [m0]         ; C8200900
>	v_interp_p2_f32 v8, [v8], v1, 1, 2, [m0]   ; C8210901
>	v_interp_p1_f32 v9, v0, 2, 2, [m0]         ; C8240A00
>	v_interp_p2_f32 v9, [v9], v1, 2, 2, [m0]   ; C8250A01
>	v_interp_p1_f32 v10, v0, 3, 2, [m0]        ; C8280B00
>	v_interp_p2_f32 v10, [v10], v1, 3, 2, [m0] ; C8290B01
>	v_interp_p1_f32 v11, v0, 0, 3, [m0]        ; C82C0C00
>	v_interp_p2_f32 v11, [v11], v1, 0, 3, [m0] ; C82D0C01
>	v_interp_p1_f32 v12, v0, 1, 3, [m0]        ; C8300D00
>	v_interp_p2_f32 v12, [v12], v1, 1, 3, [m0] ; C8310D01
>	v_interp_p1_f32 v13, v0, 2, 3, [m0]        ; C8340E00
>	v_interp_p2_f32 v13, [v13], v1, 2, 3, [m0] ; C8350E01
>	v_interp_p1_f32 v0, v0, 3, 3, [m0]         ; C8000F00
>	v_interp_p2_f32 v0, [v0], v1, 3, 3, [m0]   ; C8010F01
>	v_mac_f32_e32 v7, v11, v2                  ; 3E0E050B
>	v_mac_f32_e32 v8, v12, v3                  ; 3E10070C
>	v_mac_f32_e32 v9, v13, v4                  ; 3E12090D
>	v_mac_f32_e32 v10, v0, v5                  ; 3E140B00
>	v_mul_f32_e32 v0, v6, v10                  ; 10001506
>	v_cvt_pkrtz_f16_f32_e32 v1, v7, v8         ; 5E021107
>	v_cvt_pkrtz_f16_f32_e32 v0, v9, v0         ; 5E000109
>	exp 15, 0, 1, 1, 1, v1, v0, v1, v0         ; F8001C0F 00010001
>	s_endpgm                                   ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 16
>VGPRS: 16
>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
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL IN[2]
>DCL IN[3]
>DCL OUT[0], POSITION
>DCL OUT[1], CLIPVERTEX
>DCL OUT[2], GENERIC[0]
>DCL OUT[3], GENERIC[1]
>DCL OUT[4], GENERIC[2]
>DCL OUT[5], GENERIC[3]
>DCL CONST[0..57]
>DCL TEMP[0..7], LOCAL
>IMM[0] FLT32 {    2.2000,     0.0000,     0.0000,     0.0000}
>  0: LG2 TEMP[0].x, IN[1].xxxx
>  1: LG2 TEMP[1].x, IN[1].yyyy
>  2: MOV TEMP[0].y, TEMP[1].xxxx
>  3: LG2 TEMP[1].x, IN[1].zzzz
>  4: MOV TEMP[0].z, TEMP[1].xxxx
>  5: MUL TEMP[0].xyz, TEMP[0].xyzz, IMM[0].xxxx
>  6: EX2 TEMP[1].x, TEMP[0].xxxx
>  7: EX2 TEMP[2].x, TEMP[0].yyyy
>  8: MOV TEMP[1].y, TEMP[2].xxxx
>  9: EX2 TEMP[2].x, TEMP[0].zzzz
> 10: MOV TEMP[1].z, TEMP[2].xxxx
> 11: DP4 TEMP[2].x, IN[2], CONST[48]
> 12: DP4 TEMP[3].x, IN[2], CONST[49]
> 13: MOV TEMP[2].y, TEMP[3].xxxx
> 14: MOV TEMP[0].w, CONST[0].yyyy
> 15: MOV TEMP[3].w, IN[0].wwww
> 16: MAD TEMP[3].xyz, IN[3].xyzz, CONST[13].xxxx, IN[0].xyzz
> 17: DP4 TEMP[0].x, TEMP[3], CONST[54]
> 18: DP4 TEMP[4].x, TEMP[3], CONST[55]
> 19: MOV TEMP[0].y, TEMP[4].xxxx
> 20: DP4 TEMP[3].x, TEMP[3], CONST[56]
> 21: MOV TEMP[0].z, TEMP[3].xxxx
> 22: DP4 TEMP[3].x, TEMP[0], CONST[8]
> 23: DP4 TEMP[4].x, TEMP[0], CONST[9]
> 24: MOV TEMP[3].y, TEMP[4].xxxx
> 25: DP4 TEMP[5].x, TEMP[0], CONST[11]
> 26: MOV TEMP[3].w, TEMP[5].xxxx
> 27: DP4 TEMP[6].x, TEMP[0], CONST[10]
> 28: MOV TEMP[0].w, TEMP[6].xxxx
> 29: MOV TEMP[3].z, TEMP[6].xxxx
> 30: MOV TEMP[2].zw, CONST[0].xxxx
> 31: MOV TEMP[1].w, IN[1].wwww
> 32: MOV TEMP[7], TEMP[3]
> 33: MAD TEMP[6].x, TEMP[6].xxxx, CONST[0].zzzz, -TEMP[5].xxxx
> 34: MOV TEMP[3].z, TEMP[6].xxxx
> 35: MOV TEMP[3].y, -TEMP[4].xxxx
> 36: MAD TEMP[3].xy, CONST[57].xyyy, TEMP[5].xxxx, TEMP[3].xyyy
> 37: MOV OUT[2], TEMP[2]
> 38: MOV OUT[3], CONST[0].xxxx
> 39: MOV OUT[0], TEMP[3]
> 40: MOV OUT[1], TEMP[7]
> 41: MOV OUT[4], TEMP[1]
> 42: MOV OUT[5], TEMP[0]
> 43: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = call float @llvm.SI.load.const(<16 x i8> %13, i32 0)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %13, i32 4)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %13, i32 8)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %13, i32 128)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %13, i32 132)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %13, i32 136)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %13, i32 140)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %13, i32 144)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %13, i32 148)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %13, i32 152)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %13, i32 156)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %13, i32 160)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %13, i32 164)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %13, i32 168)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %13, i32 172)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %13, i32 176)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %13, i32 180)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %13, i32 184)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %13, i32 188)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %13, i32 208)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %13, i32 768)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %13, i32 772)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %13, i32 776)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %13, i32 780)
>  %38 = call float @llvm.SI.load.const(<16 x i8> %13, i32 784)
>  %39 = call float @llvm.SI.load.const(<16 x i8> %13, i32 788)
>  %40 = call float @llvm.SI.load.const(<16 x i8> %13, i32 792)
>  %41 = call float @llvm.SI.load.const(<16 x i8> %13, i32 796)
>  %42 = call float @llvm.SI.load.const(<16 x i8> %13, i32 864)
>  %43 = call float @llvm.SI.load.const(<16 x i8> %13, i32 868)
>  %44 = call float @llvm.SI.load.const(<16 x i8> %13, i32 872)
>  %45 = call float @llvm.SI.load.const(<16 x i8> %13, i32 876)
>  %46 = call float @llvm.SI.load.const(<16 x i8> %13, i32 880)
>  %47 = call float @llvm.SI.load.const(<16 x i8> %13, i32 884)
>  %48 = call float @llvm.SI.load.const(<16 x i8> %13, i32 888)
>  %49 = call float @llvm.SI.load.const(<16 x i8> %13, i32 892)
>  %50 = call float @llvm.SI.load.const(<16 x i8> %13, i32 896)
>  %51 = call float @llvm.SI.load.const(<16 x i8> %13, i32 900)
>  %52 = call float @llvm.SI.load.const(<16 x i8> %13, i32 904)
>  %53 = call float @llvm.SI.load.const(<16 x i8> %13, i32 908)
>  %54 = call float @llvm.SI.load.const(<16 x i8> %13, i32 912)
>  %55 = call float @llvm.SI.load.const(<16 x i8> %13, i32 916)
>  %56 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %57 = load <16 x i8>, <16 x i8> addrspace(2)* %56, align 16, !tbaa !0
>  %58 = add i32 %5, %8
>  %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 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 1
>  %65 = load <16 x i8>, <16 x i8> addrspace(2)* %64, align 16, !tbaa !0
>  %66 = add i32 %5, %8
>  %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 = extractelement <4 x float> %67, i32 2
>  %71 = extractelement <4 x float> %67, i32 3
>  %72 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 2
>  %73 = load <16 x i8>, <16 x i8> addrspace(2)* %72, align 16, !tbaa !0
>  %74 = add i32 %5, %8
>  %75 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %73, i32 0, i32 %74)
>  %76 = extractelement <4 x float> %75, i32 0
>  %77 = extractelement <4 x float> %75, i32 1
>  %78 = extractelement <4 x float> %75, i32 2
>  %79 = extractelement <4 x float> %75, i32 3
>  %80 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 3
>  %81 = load <16 x i8>, <16 x i8> addrspace(2)* %80, align 16, !tbaa !0
>  %82 = add i32 %5, %8
>  %83 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %81, i32 0, i32 %82)
>  %84 = extractelement <4 x float> %83, i32 0
>  %85 = extractelement <4 x float> %83, i32 1
>  %86 = extractelement <4 x float> %83, i32 2
>  %87 = call float @llvm.log2.f32(float %68)
>  %88 = call float @llvm.log2.f32(float %69)
>  %89 = call float @llvm.log2.f32(float %70)
>  %90 = fmul float %87, 0x40019999A0000000
>  %91 = fmul float %88, 0x40019999A0000000
>  %92 = fmul float %89, 0x40019999A0000000
>  %93 = call float @llvm.AMDIL.exp.(float %90)
>  %94 = call float @llvm.AMDIL.exp.(float %91)
>  %95 = call float @llvm.AMDIL.exp.(float %92)
>  %96 = fmul float %76, %34
>  %97 = fmul float %77, %35
>  %98 = fadd float %96, %97
>  %99 = fmul float %78, %36
>  %100 = fadd float %98, %99
>  %101 = fmul float %79, %37
>  %102 = fadd float %100, %101
>  %103 = fmul float %76, %38
>  %104 = fmul float %77, %39
>  %105 = fadd float %103, %104
>  %106 = fmul float %78, %40
>  %107 = fadd float %105, %106
>  %108 = fmul float %79, %41
>  %109 = fadd float %107, %108
>  %110 = fmul float %84, %33
>  %111 = fadd float %110, %60
>  %112 = fmul float %85, %33
>  %113 = fadd float %112, %61
>  %114 = fmul float %86, %33
>  %115 = fadd float %114, %62
>  %116 = fmul float %111, %42
>  %117 = fmul float %113, %43
>  %118 = fadd float %116, %117
>  %119 = fmul float %115, %44
>  %120 = fadd float %118, %119
>  %121 = fmul float %63, %45
>  %122 = fadd float %120, %121
>  %123 = fmul float %111, %46
>  %124 = fmul float %113, %47
>  %125 = fadd float %123, %124
>  %126 = fmul float %115, %48
>  %127 = fadd float %125, %126
>  %128 = fmul float %63, %49
>  %129 = fadd float %127, %128
>  %130 = fmul float %111, %50
>  %131 = fmul float %113, %51
>  %132 = fadd float %130, %131
>  %133 = fmul float %115, %52
>  %134 = fadd float %132, %133
>  %135 = fmul float %63, %53
>  %136 = fadd float %134, %135
>  %137 = fmul float %122, %17
>  %138 = fmul float %129, %18
>  %139 = fadd float %137, %138
>  %140 = fmul float %136, %19
>  %141 = fadd float %139, %140
>  %142 = fmul float %15, %20
>  %143 = fadd float %141, %142
>  %144 = fmul float %122, %21
>  %145 = fmul float %129, %22
>  %146 = fadd float %144, %145
>  %147 = fmul float %136, %23
>  %148 = fadd float %146, %147
>  %149 = fmul float %15, %24
>  %150 = fadd float %148, %149
>  %151 = fmul float %122, %29
>  %152 = fmul float %129, %30
>  %153 = fadd float %151, %152
>  %154 = fmul float %136, %31
>  %155 = fadd float %153, %154
>  %156 = fmul float %15, %32
>  %157 = fadd float %155, %156
>  %158 = fmul float %122, %25
>  %159 = fmul float %129, %26
>  %160 = fadd float %158, %159
>  %161 = fmul float %136, %27
>  %162 = fadd float %160, %161
>  %163 = fmul float %15, %28
>  %164 = fadd float %162, %163
>  %165 = fmul float %164, %16
>  %166 = fsub float %165, %157
>  %167 = fmul float %54, %157
>  %168 = fadd float %167, %143
>  %169 = fmul float %55, %157
>  %170 = fsub float %169, %150
>  %171 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 16
>  %172 = load <16 x i8>, <16 x i8> addrspace(2)* %171, align 16, !tbaa !0
>  %173 = call float @llvm.SI.load.const(<16 x i8> %172, i32 0)
>  %174 = fmul float %173, %143
>  %175 = call float @llvm.SI.load.const(<16 x i8> %172, i32 4)
>  %176 = fmul float %175, %150
>  %177 = fadd float %174, %176
>  %178 = call float @llvm.SI.load.const(<16 x i8> %172, i32 8)
>  %179 = fmul float %178, %164
>  %180 = fadd float %177, %179
>  %181 = call float @llvm.SI.load.const(<16 x i8> %172, i32 12)
>  %182 = fmul float %181, %157
>  %183 = fadd float %180, %182
>  %184 = call float @llvm.SI.load.const(<16 x i8> %172, i32 16)
>  %185 = fmul float %184, %143
>  %186 = call float @llvm.SI.load.const(<16 x i8> %172, i32 20)
>  %187 = fmul float %186, %150
>  %188 = fadd float %185, %187
>  %189 = call float @llvm.SI.load.const(<16 x i8> %172, i32 24)
>  %190 = fmul float %189, %164
>  %191 = fadd float %188, %190
>  %192 = call float @llvm.SI.load.const(<16 x i8> %172, i32 28)
>  %193 = fmul float %192, %157
>  %194 = fadd float %191, %193
>  %195 = call float @llvm.SI.load.const(<16 x i8> %172, i32 32)
>  %196 = fmul float %195, %143
>  %197 = call float @llvm.SI.load.const(<16 x i8> %172, i32 36)
>  %198 = fmul float %197, %150
>  %199 = fadd float %196, %198
>  %200 = call float @llvm.SI.load.const(<16 x i8> %172, i32 40)
>  %201 = fmul float %200, %164
>  %202 = fadd float %199, %201
>  %203 = call float @llvm.SI.load.const(<16 x i8> %172, i32 44)
>  %204 = fmul float %203, %157
>  %205 = fadd float %202, %204
>  %206 = call float @llvm.SI.load.const(<16 x i8> %172, i32 48)
>  %207 = fmul float %206, %143
>  %208 = call float @llvm.SI.load.const(<16 x i8> %172, i32 52)
>  %209 = fmul float %208, %150
>  %210 = fadd float %207, %209
>  %211 = call float @llvm.SI.load.const(<16 x i8> %172, i32 56)
>  %212 = fmul float %211, %164
>  %213 = fadd float %210, %212
>  %214 = call float @llvm.SI.load.const(<16 x i8> %172, i32 60)
>  %215 = fmul float %214, %157
>  %216 = fadd float %213, %215
>  %217 = call float @llvm.SI.load.const(<16 x i8> %172, i32 64)
>  %218 = fmul float %217, %143
>  %219 = call float @llvm.SI.load.const(<16 x i8> %172, i32 68)
>  %220 = fmul float %219, %150
>  %221 = fadd float %218, %220
>  %222 = call float @llvm.SI.load.const(<16 x i8> %172, i32 72)
>  %223 = fmul float %222, %164
>  %224 = fadd float %221, %223
>  %225 = call float @llvm.SI.load.const(<16 x i8> %172, i32 76)
>  %226 = fmul float %225, %157
>  %227 = fadd float %224, %226
>  %228 = call float @llvm.SI.load.const(<16 x i8> %172, i32 80)
>  %229 = fmul float %228, %143
>  %230 = call float @llvm.SI.load.const(<16 x i8> %172, i32 84)
>  %231 = fmul float %230, %150
>  %232 = fadd float %229, %231
>  %233 = call float @llvm.SI.load.const(<16 x i8> %172, i32 88)
>  %234 = fmul float %233, %164
>  %235 = fadd float %232, %234
>  %236 = call float @llvm.SI.load.const(<16 x i8> %172, i32 92)
>  %237 = fmul float %236, %157
>  %238 = fadd float %235, %237
>  %239 = call float @llvm.SI.load.const(<16 x i8> %172, i32 96)
>  %240 = fmul float %239, %143
>  %241 = call float @llvm.SI.load.const(<16 x i8> %172, i32 100)
>  %242 = fmul float %241, %150
>  %243 = fadd float %240, %242
>  %244 = call float @llvm.SI.load.const(<16 x i8> %172, i32 104)
>  %245 = fmul float %244, %164
>  %246 = fadd float %243, %245
>  %247 = call float @llvm.SI.load.const(<16 x i8> %172, i32 108)
>  %248 = fmul float %247, %157
>  %249 = fadd float %246, %248
>  %250 = call float @llvm.SI.load.const(<16 x i8> %172, i32 112)
>  %251 = fmul float %250, %143
>  %252 = call float @llvm.SI.load.const(<16 x i8> %172, i32 116)
>  %253 = fmul float %252, %150
>  %254 = fadd float %251, %253
>  %255 = call float @llvm.SI.load.const(<16 x i8> %172, i32 120)
>  %256 = fmul float %255, %164
>  %257 = fadd float %254, %256
>  %258 = call float @llvm.SI.load.const(<16 x i8> %172, i32 124)
>  %259 = fmul float %258, %157
>  %260 = fadd float %257, %259
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %102, float %109, float %14, float %14)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %14, float %14, float %14, float %14)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 34, i32 0, float %93, float %94, float %95, float %71)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 35, i32 0, float %122, float %129, float %136, float %164)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 12, i32 0, float %168, float %170, float %166, float %157)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 13, i32 0, float %183, float %194, float %205, float %216)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 14, i32 0, float %227, float %238, float %249, float %260)
>  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.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:
>
>	s_load_dwordx4 s[52:55], s[2:3], 0x0                    ; C09A0300
>	s_load_dwordx4 s[24:27], s[2:3], 0x40                   ; C08C0340
>	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[32:35], s[8:9], 0xc                    ; C090090C
>	v_add_i32_e32 v0, vcc, s10, v0                          ; 4A00000A
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	s_buffer_load_dword s3, s[52:55], 0x2c                  ; C201B52C
>	s_buffer_load_dword s5, s[52:55], 0x2d                  ; C202B52D
>	s_buffer_load_dword s1, s[52:55], 0x2e                  ; C200B52E
>	s_buffer_load_dword s0, s[52:55], 0x2f                  ; C200352F
>	s_buffer_load_dword s28, s[52:55], 0x34                 ; C20E3534
>	buffer_load_format_xyzw v[1:4], v0, s[12:15], 0 idxen   ; E00C2000 80030100
>	buffer_load_format_xyzw v[5:8], v0, s[16:19], 0 idxen   ; E00C2000 80040500
>	buffer_load_format_xyzw v[9:12], v0, s[20:23], 0 idxen  ; E00C2000 80050900
>	buffer_load_format_xyzw v[13:16], v0, s[32:35], 0 idxen ; E00C2000 80080D00
>	s_buffer_load_dword s14, s[52:55], 0xda                 ; C20735DA
>	s_buffer_load_dword s11, s[52:55], 0xdb                 ; C205B5DB
>	s_buffer_load_dword s17, s[52:55], 0xdc                 ; C208B5DC
>	s_buffer_load_dword s33, s[52:55], 0xdd                 ; C210B5DD
>	s_buffer_load_dword s15, s[52:55], 0xde                 ; C207B5DE
>	s_buffer_load_dword s45, s[52:55], 0xc5                 ; C216B5C5
>	s_buffer_load_dword s32, s[52:55], 0xc6                 ; C21035C6
>	s_buffer_load_dword s29, s[52:55], 0xc7                 ; C20EB5C7
>	s_buffer_load_dword s22, s[52:55], 0xd8                 ; C20B35D8
>	s_buffer_load_dword s40, s[52:55], 0xd9                 ; C21435D9
>	s_buffer_load_dword s18, s[52:55], 0xdf                 ; C20935DF
>	s_buffer_load_dword s30, s[52:55], 0xe0                 ; C20F35E0
>	s_buffer_load_dword s41, s[52:55], 0xe1                 ; C214B5E1
>	s_buffer_load_dword s21, s[52:55], 0xe2                 ; C20AB5E2
>	s_buffer_load_dword s20, s[52:55], 0xe3                 ; C20A35E3
>	s_buffer_load_dword s48, s[52:55], 0xc0                 ; C21835C0
>	s_buffer_load_dword s51, s[52:55], 0xc1                 ; C219B5C1
>	s_buffer_load_dword s44, s[52:55], 0xc2                 ; C21635C2
>	s_buffer_load_dword s42, s[52:55], 0xc3                 ; C21535C3
>	s_buffer_load_dword s49, s[52:55], 0xc4                 ; C218B5C4
>	s_buffer_load_dword s13, s[52:55], 0x22                 ; C206B522
>	s_buffer_load_dword s2, s[52:55], 0x23                  ; C2013523
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	v_mov_b32_e32 v0, s0                                    ; 7E000200
>	s_buffer_load_dword s19, s[52:55], 0x24                 ; C209B524
>	s_buffer_load_dword s31, s[52:55], 0x25                 ; C20FB525
>	s_buffer_load_dword s16, s[52:55], 0x26                 ; C2083526
>	s_buffer_load_dword s0, s[52:55], 0x0                   ; C2003500
>	s_buffer_load_dword s12, s[52:55], 0x1                  ; C2063501
>	s_buffer_load_dword s9, s[52:55], 0x2                   ; C204B502
>	s_buffer_load_dword s34, s[52:55], 0x20                 ; C2113520
>	s_buffer_load_dword s38, s[52:55], 0x21                 ; C2133521
>	s_buffer_load_dword s6, s[52:55], 0x27                  ; C2033527
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	v_mov_b32_e32 v16, s2                                   ; 7E200202
>	s_buffer_load_dword s39, s[52:55], 0x28                 ; C213B528
>	s_buffer_load_dword s47, s[52:55], 0x29                 ; C217B529
>	s_buffer_load_dword s37, s[52:55], 0x2a                 ; C212B52A
>	s_buffer_load_dword s50, s[52:55], 0x2b                 ; C219352B
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	v_mov_b32_e32 v17, s0                                   ; 7E220200
>	s_buffer_load_dword s7, s[24:27], 0x0                   ; C2039900
>	s_buffer_load_dword s23, s[24:27], 0x1                  ; C20B9901
>	s_buffer_load_dword s4, s[24:27], 0x2                   ; C2021902
>	s_buffer_load_dword s0, s[24:27], 0x3                   ; C2001903
>	s_buffer_load_dword s8, s[24:27], 0x4                   ; C2041904
>	v_mov_b32_e32 v18, s6                                   ; 7E240206
>	s_buffer_load_dword s35, s[24:27], 0x5                  ; C2119905
>	s_buffer_load_dword s6, s[24:27], 0x6                   ; C2031906
>	s_buffer_load_dword s2, s[24:27], 0x7                   ; C2011907
>	s_buffer_load_dword s10, s[24:27], 0x8                  ; C2051908
>	s_buffer_load_dword s36, s[24:27], 0x9                  ; C2121909
>	s_buffer_load_dword s43, s[52:55], 0xe4                 ; C215B5E4
>	s_buffer_load_dword s46, s[52:55], 0xe5                 ; C21735E5
>	s_buffer_load_dword s52, s[24:27], 0xa                  ; C21A190A
>	s_buffer_load_dword s53, s[24:27], 0xb                  ; C21A990B
>	s_buffer_load_dword s54, s[24:27], 0xc                  ; C21B190C
>	s_buffer_load_dword s55, s[24:27], 0xd                  ; C21B990D
>	s_buffer_load_dword s56, s[24:27], 0xe                  ; C21C190E
>	s_buffer_load_dword s57, s[24:27], 0xf                  ; C21C990F
>	s_buffer_load_dword s58, s[24:27], 0x10                 ; C21D1910
>	s_buffer_load_dword s59, s[24:27], 0x11                 ; C21D9911
>	s_buffer_load_dword s60, s[24:27], 0x12                 ; C21E1912
>	s_buffer_load_dword s61, s[24:27], 0x13                 ; C21E9913
>	s_buffer_load_dword s62, s[24:27], 0x14                 ; C21F1914
>	s_buffer_load_dword s63, s[24:27], 0x15                 ; C21F9915
>	s_buffer_load_dword s64, s[24:27], 0x16                 ; C2201916
>	s_buffer_load_dword s65, s[24:27], 0x17                 ; C2209917
>	s_buffer_load_dword s66, s[24:27], 0x18                 ; C2211918
>	s_buffer_load_dword s67, s[24:27], 0x19                 ; C2219919
>	s_buffer_load_dword s68, s[24:27], 0x1a                 ; C222191A
>	s_buffer_load_dword s69, s[24:27], 0x1b                 ; C222991B
>	s_buffer_load_dword s70, s[24:27], 0x1c                 ; C223191C
>	s_buffer_load_dword s71, s[24:27], 0x1d                 ; C223991D
>	s_buffer_load_dword s72, s[24:27], 0x1e                 ; C224191E
>	s_buffer_load_dword s24, s[24:27], 0x1f                 ; C20C191F
>	v_mul_f32_e32 v19, s51, v10                             ; 10261433
>	v_mul_f32_e32 v10, s45, v10                             ; 1014142D
>	v_mad_f32 v1, s28, v13, v1                              ; D2820001 04061A1C
>	v_mad_f32 v2, s28, v14, v2                              ; D2820002 040A1C1C
>	v_mad_f32 v3, s28, v15, v3                              ; D2820003 040E1E1C
>	v_mac_f32_e32 v19, s48, v9                              ; 3E261230
>	v_mac_f32_e32 v10, s49, v9                              ; 3E141231
>	v_mul_f32_e32 v9, s40, v2                               ; 10120428
>	v_mul_f32_e32 v13, s33, v2                              ; 101A0421
>	v_mul_f32_e32 v2, s41, v2                               ; 10040429
>	v_mac_f32_e32 v19, s44, v11                             ; 3E26162C
>	v_mac_f32_e32 v10, s32, v11                             ; 3E141620
>	v_mac_f32_e32 v19, s42, v12                             ; 3E26182A
>	v_mac_f32_e32 v10, s29, v12                             ; 3E14181D
>	v_mac_f32_e32 v9, s22, v1                               ; 3E120216
>	v_mac_f32_e32 v13, s17, v1                              ; 3E1A0211
>	v_mac_f32_e32 v2, s30, v1                               ; 3E04021E
>	v_mac_f32_e32 v9, s14, v3                               ; 3E12060E
>	v_mac_f32_e32 v13, s15, v3                              ; 3E1A060F
>	v_mac_f32_e32 v2, s21, v3                               ; 3E040615
>	v_mac_f32_e32 v9, s11, v4                               ; 3E12080B
>	v_mac_f32_e32 v13, s18, v4                              ; 3E1A0812
>	v_mac_f32_e32 v2, s20, v4                               ; 3E040814
>	v_mov_b32_e32 v1, s50                                   ; 7E020232
>	v_mul_f32_e32 v3, s38, v13                              ; 10061A26
>	v_mul_f32_e32 v4, s31, v13                              ; 10081A1F
>	v_mul_f32_e32 v11, s5, v13                              ; 10161A05
>	v_mul_f32_e32 v12, s47, v13                             ; 10181A2F
>	v_mac_f32_e32 v3, s34, v9                               ; 3E061222
>	v_mac_f32_e32 v4, s19, v9                               ; 3E081213
>	v_mac_f32_e32 v11, s3, v9                               ; 3E161203
>	v_mac_f32_e32 v12, s39, v9                              ; 3E181227
>	v_mac_f32_e32 v3, s13, v2                               ; 3E06040D
>	v_mac_f32_e32 v4, s16, v2                               ; 3E080410
>	v_mac_f32_e32 v11, s1, v2                               ; 3E160401
>	v_mac_f32_e32 v12, s37, v2                              ; 3E180425
>	v_mac_f32_e32 v3, s12, v16                              ; 3E06200C
>	v_mac_f32_e32 v4, s12, v18                              ; 3E08240C
>	v_mac_f32_e32 v11, s12, v0                              ; 3E16000C
>	v_mac_f32_e32 v12, s12, v1                              ; 3E18020C
>	v_mad_f32 v0, v12, s9, -v11                             ; D2820000 842C130C
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	v_mul_f32_e32 v1, s23, v4                               ; 10020817
>	v_mul_f32_e32 v14, s35, v4                              ; 101C0823
>	v_mul_f32_e32 v15, s36, v4                              ; 101E0824
>	v_mul_f32_e32 v16, s55, v4                              ; 10200837
>	v_mul_f32_e32 v18, s59, v4                              ; 1024083B
>	v_mul_f32_e32 v20, s63, v4                              ; 1028083F
>	v_mul_f32_e32 v21, s67, v4                              ; 102A0843
>	v_log_f32_e32 v5, v5                                    ; 7E0A4F05
>	v_log_f32_e32 v6, v6                                    ; 7E0C4F06
>	exp 15, 32, 0, 0, 0, v19, v10, v17, v17                 ; F800020F 11110A13
>	v_log_f32_e32 v7, v7                                    ; 7E0E4F07
>	s_waitcnt expcnt(0)                                     ; BF8C070F
>	v_mov_b32_e32 v10, 0x400ccccd                           ; 7E1402FF 400CCCCD
>	v_mul_f32_e32 v5, v10, v5                               ; 100A0B0A
>	v_mul_f32_e32 v6, v10, v6                               ; 100C0D0A
>	v_mul_f32_e32 v7, v10, v7                               ; 100E0F0A
>	v_mul_f32_e32 v10, s71, v4                              ; 10140847
>	v_mac_f32_e32 v1, s7, v3                                ; 3E020607
>	v_mac_f32_e32 v14, s8, v3                               ; 3E1C0608
>	v_mac_f32_e32 v15, s10, v3                              ; 3E1E060A
>	v_mac_f32_e32 v16, s54, v3                              ; 3E200636
>	v_mac_f32_e32 v18, s58, v3                              ; 3E24063A
>	v_mac_f32_e32 v20, s62, v3                              ; 3E28063E
>	v_mac_f32_e32 v21, s66, v3                              ; 3E2A0642
>	v_mac_f32_e32 v10, s70, v3                              ; 3E140646
>	v_mad_f32 v3, s43, v11, v3                              ; D2820003 040E162B
>	v_mad_f32 v4, s46, v11, -v4                             ; D2820004 8412162E
>	v_mac_f32_e32 v1, s4, v12                               ; 3E021804
>	v_mac_f32_e32 v14, s6, v12                              ; 3E1C1806
>	v_mac_f32_e32 v15, s52, v12                             ; 3E1E1834
>	v_mac_f32_e32 v16, s56, v12                             ; 3E201838
>	v_mac_f32_e32 v18, s60, v12                             ; 3E24183C
>	v_mac_f32_e32 v20, s64, v12                             ; 3E281840
>	v_mac_f32_e32 v21, s68, v12                             ; 3E2A1844
>	v_mac_f32_e32 v10, s72, v12                             ; 3E141848
>	v_mac_f32_e32 v1, s0, v11                               ; 3E021600
>	v_mac_f32_e32 v14, s2, v11                              ; 3E1C1602
>	v_mac_f32_e32 v15, s53, v11                             ; 3E1E1635
>	v_mac_f32_e32 v16, s57, v11                             ; 3E201639
>	v_mac_f32_e32 v18, s61, v11                             ; 3E24163D
>	v_mac_f32_e32 v20, s65, v11                             ; 3E281641
>	v_mac_f32_e32 v21, s69, v11                             ; 3E2A1645
>	v_mac_f32_e32 v10, s24, v11                             ; 3E141618
>	exp 15, 33, 0, 0, 0, v17, v17, v17, v17                 ; F800021F 11111111
>	v_exp_f32_e32 v5, v5                                    ; 7E0A4B05
>	v_exp_f32_e32 v6, v6                                    ; 7E0C4B06
>	v_exp_f32_e32 v7, v7                                    ; 7E0E4B07
>	exp 15, 34, 0, 0, 0, v5, v6, v7, v8                     ; F800022F 08070605
>	exp 15, 35, 0, 0, 0, v9, v13, v2, v12                   ; F800023F 0C020D09
>	exp 15, 12, 0, 0, 0, v3, v4, v0, v11                    ; F80000CF 0B000403
>	exp 15, 13, 0, 0, 0, v1, v14, v15, v16                  ; F80000DF 100F0E01
>	exp 15, 14, 0, 1, 0, v18, v20, v21, v10                 ; F80008EF 0A151412
>	s_endpgm                                                ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 80
>VGPRS: 24
>Code Size: 824 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
>  clamp_color = 0
>FRAG
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL IN[1], GENERIC[1], PERSPECTIVE
>DCL IN[2], GENERIC[2], PERSPECTIVE
>DCL IN[3], GENERIC[3], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SAMP[1]
>DCL SVIEW[0], 2D, FLOAT
>DCL SVIEW[1], SHADOW2D, FLOAT
>DCL CONST[0..90]
>DCL TEMP[0..15], LOCAL
>IMM[0] FLT32 {    1.0000,     0.0000,     2.0000,    -0.5000}
>IMM[1] FLT32 {   -0.0000,    -1.0000,    -2.0000,     0.0625}
>IMM[2] FLT32 {    0.0005,     0.0000,    -0.0005,     0.1250}
>IMM[3] FLT32 {    0.2500,     0.0000,    -1.0000,    -2.0000}
>  0: MOV TEMP[0].xy, IN[0].xyyy
>  1: TEX TEMP[0], TEMP[0], SAMP[0], 2D
>  2: UIF CONST[90].xxxx :0
>  3:   MAD TEMP[1], IN[3].xyzx, IMM[0].xxxy, IMM[0].yyyx
>  4:   DP4 TEMP[2].x, TEMP[1], CONST[69]
>  5:   DP4 TEMP[3].x, TEMP[1], CONST[70]
>  6:   MOV TEMP[2].y, TEMP[3].xxxx
>  7:   MOV_SAT TEMP[4].xy, TEMP[2].xyyy
>  8:   ADD TEMP[4].xy, -TEMP[2].xyyy, TEMP[4].xyyy
>  9:   DP2 TEMP[5].x, TEMP[4].xyyy, IMM[0].xxxx
> 10:   DP4 TEMP[4].x, TEMP[1], CONST[73]
> 11:   DP4 TEMP[6].x, TEMP[1], CONST[74]
> 12:   MOV TEMP[4].y, TEMP[6].xxxx
> 13:   MOV_SAT TEMP[7].xy, TEMP[4].xyyy
> 14:   ADD TEMP[7].xy, -TEMP[4].xyyy, TEMP[7].xyyy
> 15:   DP2 TEMP[8].x, TEMP[7].xyyy, IMM[0].xxxx
> 16:   MOV TEMP[4].w, TEMP[8].xxxx
> 17:   DP4 TEMP[7].x, TEMP[1], CONST[77]
> 18:   DP4 TEMP[9].x, TEMP[1], CONST[78]
> 19:   MOV TEMP[4].z, IMM[0].xxxx
> 20:   MOV TEMP[10].w, TEMP[4]
> 21:   ABS TEMP[11].x, TEMP[8].xxxx
> 22:   FSGE TEMP[11].x, -TEMP[11].xxxx, IMM[0].yyyy
> 23:   UIF TEMP[11].xxxx :0
> 24:     MOV TEMP[11].x, TEMP[4].xxxx
> 25:   ELSE :0
> 26:     MOV TEMP[11].x, TEMP[7].xxxx
> 27:   ENDIF
> 28:   MOV TEMP[10].x, TEMP[11].xxxx
> 29:   ABS TEMP[11].x, TEMP[8].xxxx
> 30:   FSGE TEMP[11].x, -TEMP[11].xxxx, IMM[0].yyyy
> 31:   UIF TEMP[11].xxxx :0
> 32:     MOV TEMP[6].x, TEMP[6].xxxx
> 33:   ELSE :0
> 34:     MOV TEMP[6].x, TEMP[9].xxxx
> 35:   ENDIF
> 36:   MOV TEMP[10].y, TEMP[6].xxxx
> 37:   ABS TEMP[6].x, TEMP[8].xxxx
> 38:   FSGE TEMP[6].x, -TEMP[6].xxxx, IMM[0].yyyy
> 39:   UIF TEMP[6].xxxx :0
> 40:     MOV TEMP[6].x, IMM[0].xxxx
> 41:   ELSE :0
> 42:     MOV TEMP[6].x, IMM[0].zzzz
> 43:   ENDIF
> 44:   MOV TEMP[10].z, TEMP[6].xxxx
> 45:   MOV TEMP[4].xyz, TEMP[10]
> 46:   ABS TEMP[6].x, TEMP[5].xxxx
> 47:   FSGE TEMP[6].x, -TEMP[6].xxxx, IMM[0].yyyy
> 48:   UIF TEMP[6].xxxx :0
> 49:     MOV TEMP[6].x, TEMP[2].xxxx
> 50:   ELSE :0
> 51:     MOV TEMP[6].x, TEMP[4].xxxx
> 52:   ENDIF
> 53:   MOV TEMP[10].x, TEMP[6].xxxx
> 54:   ABS TEMP[6].x, TEMP[5].xxxx
> 55:   FSGE TEMP[6].x, -TEMP[6].xxxx, IMM[0].yyyy
> 56:   UIF TEMP[6].xxxx :0
> 57:     MOV TEMP[3].x, TEMP[3].xxxx
> 58:   ELSE :0
> 59:     MOV TEMP[3].x, TEMP[4].yyyy
> 60:   ENDIF
> 61:   MOV TEMP[10].y, TEMP[3].xxxx
> 62:   ABS TEMP[3].x, TEMP[5].xxxx
> 63:   FSGE TEMP[3].x, -TEMP[3].xxxx, IMM[0].yyyy
> 64:   UIF TEMP[3].xxxx :0
> 65:     MOV TEMP[3].x, IMM[0].yyyy
> 66:   ELSE :0
> 67:     MOV TEMP[3].x, TEMP[4].zzzz
> 68:   ENDIF
> 69:   MOV TEMP[10].z, TEMP[3].xxxx
> 70:   MOV TEMP[2].z, TEMP[10].xyzx
> 71:   DP4 TEMP[5].x, TEMP[1], CONST[71]
> 72:   MOV TEMP[4].z, TEMP[5].xxxx
> 73:   ADD TEMP[7].xy, TEMP[10].xyyy, IMM[0].wwww
> 74:   ABS TEMP[6].xy, TEMP[7].xyyy
> 75:   ADD TEMP[7].xy, TEMP[6].xyyy, -CONST[67].zzzz
> 76:   MUL TEMP[7].xy, TEMP[7].xyyy, CONST[67].wwww
> 77:   MOV_SAT TEMP[6].xy, TEMP[7].xyyy
> 78:   ADD TEMP[7].xy, -TEMP[6].xyyy, IMM[0].xxxx
> 79:   MUL TEMP[6].x, TEMP[7].yyyy, TEMP[7].xxxx
> 80:   MOV_SAT TEMP[8].xy, TEMP[10].xyyy
> 81:   ADD TEMP[7].xyz, TEMP[3].xxxx, IMM[1].xyzz
> 82:   MOV TEMP[3].y, IMM[0].yyyy
> 83:   ABS TEMP[11].x, TEMP[7].xxxx
> 84:   FSGE TEMP[11].x, -TEMP[11].xxxx, IMM[0].yyyy
> 85:   UIF TEMP[11].xxxx :0
> 86:     MOV TEMP[11].x, CONST[85].zzzz
> 87:   ELSE :0
> 88:     MOV TEMP[11].x, IMM[0].yyyy
> 89:   ENDIF
> 90:   ABS TEMP[12].x, TEMP[7].xxxx
> 91:   FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
> 92:   UIF TEMP[12].xxxx :0
> 93:     MOV TEMP[12].x, CONST[85].wwww
> 94:   ELSE :0
> 95:     MOV TEMP[12].x, IMM[0].yyyy
> 96:   ENDIF
> 97:   MOV TEMP[10].y, TEMP[12].xxxx
> 98:   ABS TEMP[12].x, TEMP[7].xxxx
> 99:   FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>100:   UIF TEMP[12].xxxx :0
>101:     MOV TEMP[12].x, CONST[85].xxxx
>102:   ELSE :0
>103:     MOV TEMP[12].x, IMM[0].yyyy
>104:   ENDIF
>105:   MOV TEMP[10].z, TEMP[12].xxxx
>106:   ABS TEMP[12].x, TEMP[7].xxxx
>107:   FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>108:   UIF TEMP[12].xxxx :0
>109:     MOV TEMP[12].x, CONST[85].yyyy
>110:   ELSE :0
>111:     MOV TEMP[12].x, IMM[0].yyyy
>112:   ENDIF
>113:   MOV TEMP[10].w, TEMP[12].xxxx
>114:   ABS TEMP[12].x, TEMP[7].yyyy
>115:   FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>116:   UIF TEMP[12].xxxx :0
>117:     MOV TEMP[12].x, CONST[86].zzzz
>118:   ELSE :0
>119:     MOV TEMP[12].x, TEMP[11].xxxx
>120:   ENDIF
>121:   MOV TEMP[10].x, TEMP[12].xxxx
>122:   ABS TEMP[11].x, TEMP[7].yyyy
>123:   FSGE TEMP[11].x, -TEMP[11].xxxx, IMM[0].yyyy
>124:   UIF TEMP[11].xxxx :0
>125:     MOV TEMP[11].x, CONST[86].wwww
>126:   ELSE :0
>127:     MOV TEMP[11].x, TEMP[10].yyyy
>128:   ENDIF
>129:   MOV TEMP[10].y, TEMP[11].xxxx
>130:   ABS TEMP[11].x, TEMP[7].yyyy
>131:   FSGE TEMP[11].x, -TEMP[11].xxxx, IMM[0].yyyy
>132:   UIF TEMP[11].xxxx :0
>133:     MOV TEMP[11].x, CONST[86].xxxx
>134:   ELSE :0
>135:     MOV TEMP[11].x, TEMP[10].zzzz
>136:   ENDIF
>137:   MOV TEMP[10].z, TEMP[11].xxxx
>138:   ABS TEMP[11].x, TEMP[7].yyyy
>139:   FSGE TEMP[11].x, -TEMP[11].xxxx, IMM[0].yyyy
>140:   UIF TEMP[11].xxxx :0
>141:     MOV TEMP[11].x, CONST[86].yyyy
>142:   ELSE :0
>143:     MOV TEMP[11].x, TEMP[10].wwww
>144:   ENDIF
>145:   MOV TEMP[10].w, TEMP[11].xxxx
>146:   MOV TEMP[9], TEMP[10]
>147:   ABS TEMP[11].x, TEMP[7].zzzz
>148:   FSGE TEMP[11].x, -TEMP[11].xxxx, IMM[0].yyyy
>149:   UIF TEMP[11].xxxx :0
>150:     MOV TEMP[11].x, CONST[87].zzzz
>151:   ELSE :0
>152:     MOV TEMP[11].x, TEMP[9].xxxx
>153:   ENDIF
>154:   MOV TEMP[10].x, TEMP[11].xxxx
>155:   ABS TEMP[11].x, TEMP[7].zzzz
>156:   FSGE TEMP[11].x, -TEMP[11].xxxx, IMM[0].yyyy
>157:   UIF TEMP[11].xxxx :0
>158:     MOV TEMP[11].x, CONST[87].wwww
>159:   ELSE :0
>160:     MOV TEMP[11].x, TEMP[9].yyyy
>161:   ENDIF
>162:   MOV TEMP[10].y, TEMP[11].xxxx
>163:   ABS TEMP[11].x, TEMP[7].zzzz
>164:   FSGE TEMP[11].x, -TEMP[11].xxxx, IMM[0].yyyy
>165:   UIF TEMP[11].xxxx :0
>166:     MOV TEMP[11].x, CONST[87].xxxx
>167:   ELSE :0
>168:     MOV TEMP[11].x, TEMP[9].zzzz
>169:   ENDIF
>170:   MOV TEMP[10].z, TEMP[11].xxxx
>171:   ABS TEMP[11].x, TEMP[7].zzzz
>172:   FSGE TEMP[11].x, -TEMP[11].xxxx, IMM[0].yyyy
>173:   UIF TEMP[11].xxxx :0
>174:     MOV TEMP[11].x, CONST[87].yyyy
>175:   ELSE :0
>176:     MOV TEMP[11].x, TEMP[9].wwww
>177:   ENDIF
>178:   MOV TEMP[10].w, TEMP[11].xxxx
>179:   MAD TEMP[4].xy, TEMP[8].xyyy, TEMP[10].xyyy, TEMP[10].zwww
>180:   MOV TEMP[4].w, IMM[0].yyyy
>181:   ADD TEMP[7], TEMP[4], IMM[2].xxyy
>182:   TXL TEMP[8].x, TEMP[7], SAMP[1], SHADOW2D
>183:   MOV TEMP[7].x, TEMP[8].xxxx
>184:   ADD TEMP[9], TEMP[4], IMM[2].zxyy
>185:   ADD TEMP[8], TEMP[4], IMM[2].xzyy
>186:   ADD TEMP[11], TEMP[4], IMM[2].zzyy
>187:   TXL TEMP[12].x, TEMP[9], SAMP[1], SHADOW2D
>188:   MOV TEMP[7].y, TEMP[12].xxxx
>189:   TXL TEMP[12].x, TEMP[8], SAMP[1], SHADOW2D
>190:   MOV TEMP[7].z, TEMP[12].xxxx
>191:   TXL TEMP[12].x, TEMP[11], SAMP[1], SHADOW2D
>192:   MOV TEMP[7].w, TEMP[12].xxxx
>193:   DP4 TEMP[12].x, TEMP[7], IMM[1].wwww
>194:   ADD TEMP[7], TEMP[4], IMM[2].xyyy
>195:   TXL TEMP[13].x, TEMP[7], SAMP[1], SHADOW2D
>196:   MOV TEMP[7].x, TEMP[13].xxxx
>197:   ADD TEMP[9], TEMP[4], IMM[2].zyyy
>198:   TXL TEMP[13], TEMP[9], SAMP[1], SHADOW2D
>199:   MOV TEMP[9], TEMP[13]
>200:   ADD TEMP[8], TEMP[4], IMM[2].yzyy
>201:   TXL TEMP[14], TEMP[8], SAMP[1], SHADOW2D
>202:   MOV TEMP[8], TEMP[14]
>203:   ADD TEMP[11], TEMP[4], IMM[2].yxyy
>204:   TXL TEMP[15], TEMP[11], SAMP[1], SHADOW2D
>205:   MOV TEMP[11], TEMP[15]
>206:   MOV TEMP[7].y, TEMP[13].xxxx
>207:   MOV TEMP[7].z, TEMP[14].xxxx
>208:   MOV TEMP[7].w, TEMP[15].xxxx
>209:   DP4 TEMP[13].x, TEMP[7], IMM[2].wwww
>210:   MOV TEMP[14].xy, TEMP[4].xyyy
>211:   MOV TEMP[14].z, TEMP[5].xxxx
>212:   MOV TEMP[14].w, IMM[0].yyyy
>213:   TXL TEMP[14], TEMP[14], SAMP[1], SHADOW2D
>214:   MOV TEMP[7].xyz, TEMP[14]
>215:   ADD TEMP[2].x, TEMP[13].xxxx, TEMP[12].xxxx
>216:   MAD TEMP[2].x, TEMP[14].xxxx, IMM[3].xxxx, TEMP[2].xxxx
>217:   FSLT TEMP[12].x, TEMP[6].xxxx, IMM[0].xxxx
>218:   UIF TEMP[12].xxxx :0
>219:     ADD TEMP[7].xyz, TEMP[2].zzzz, IMM[3].yzww
>220:     ABS TEMP[12].x, TEMP[7].xxxx
>221:     FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>222:     UIF TEMP[12].xxxx :0
>223:       MOV TEMP[12].x, CONST[73].xxxx
>224:     ELSE :0
>225:       MOV TEMP[12].x, IMM[0].yyyy
>226:     ENDIF
>227:     MOV TEMP[10].x, TEMP[12].xxxx
>228:     ABS TEMP[12].x, TEMP[7].xxxx
>229:     FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>230:     UIF TEMP[12].xxxx :0
>231:       MOV TEMP[12].x, CONST[73].yyyy
>232:     ELSE :0
>233:       MOV TEMP[12].x, IMM[0].yyyy
>234:     ENDIF
>235:     MOV TEMP[10].y, TEMP[12].xxxx
>236:     ABS TEMP[12].x, TEMP[7].xxxx
>237:     FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>238:     UIF TEMP[12].xxxx :0
>239:       MOV TEMP[12].x, CONST[73].zzzz
>240:     ELSE :0
>241:       MOV TEMP[12].x, IMM[0].yyyy
>242:     ENDIF
>243:     MOV TEMP[10].z, TEMP[12].xxxx
>244:     ABS TEMP[12].x, TEMP[7].xxxx
>245:     FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>246:     UIF TEMP[12].xxxx :0
>247:       MOV TEMP[12].x, CONST[73].wwww
>248:     ELSE :0
>249:       MOV TEMP[12].x, IMM[0].yyyy
>250:     ENDIF
>251:     MOV TEMP[10].w, TEMP[12].xxxx
>252:     MOV TEMP[9], TEMP[10]
>253:     ABS TEMP[12].x, TEMP[7].xxxx
>254:     FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>255:     UIF TEMP[12].xxxx :0
>256:       MOV TEMP[12].x, CONST[74].xxxx
>257:     ELSE :0
>258:       MOV TEMP[12].x, IMM[0].yyyy
>259:     ENDIF
>260:     MOV TEMP[10].x, TEMP[12].xxxx
>261:     ABS TEMP[12].x, TEMP[7].xxxx
>262:     FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>263:     UIF TEMP[12].xxxx :0
>264:       MOV TEMP[12].x, CONST[74].yyyy
>265:     ELSE :0
>266:       MOV TEMP[12].x, IMM[0].yyyy
>267:     ENDIF
>268:     MOV TEMP[10].y, TEMP[12].xxxx
>269:     ABS TEMP[12].x, TEMP[7].xxxx
>270:     FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>271:     UIF TEMP[12].xxxx :0
>272:       MOV TEMP[12].x, CONST[74].zzzz
>273:     ELSE :0
>274:       MOV TEMP[12].x, IMM[0].yyyy
>275:     ENDIF
>276:     MOV TEMP[10].z, TEMP[12].xxxx
>277:     ABS TEMP[12].x, TEMP[7].xxxx
>278:     FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>279:     UIF TEMP[12].xxxx :0
>280:       MOV TEMP[12].x, CONST[74].wwww
>281:     ELSE :0
>282:       MOV TEMP[12].x, IMM[0].yyyy
>283:     ENDIF
>284:     MOV TEMP[10].w, TEMP[12].xxxx
>285:     MOV TEMP[8], TEMP[10]
>286:     ABS TEMP[12].x, TEMP[7].yyyy
>287:     FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>288:     UIF TEMP[12].xxxx :0
>289:       MOV TEMP[12].x, CONST[77].xxxx
>290:     ELSE :0
>291:       MOV TEMP[12].x, TEMP[9].xxxx
>292:     ENDIF
>293:     MOV TEMP[10].x, TEMP[12].xxxx
>294:     ABS TEMP[12].x, TEMP[7].yyyy
>295:     FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>296:     UIF TEMP[12].xxxx :0
>297:       MOV TEMP[12].x, CONST[77].yyyy
>298:     ELSE :0
>299:       MOV TEMP[12].x, TEMP[9].yyyy
>300:     ENDIF
>301:     MOV TEMP[10].y, TEMP[12].xxxx
>302:     ABS TEMP[12].x, TEMP[7].yyyy
>303:     FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>304:     UIF TEMP[12].xxxx :0
>305:       MOV TEMP[12].x, CONST[77].zzzz
>306:     ELSE :0
>307:       MOV TEMP[12].x, TEMP[9].zzzz
>308:     ENDIF
>309:     MOV TEMP[10].z, TEMP[12].xxxx
>310:     ABS TEMP[12].x, TEMP[7].yyyy
>311:     FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>312:     UIF TEMP[12].xxxx :0
>313:       MOV TEMP[12].x, CONST[77].wwww
>314:     ELSE :0
>315:       MOV TEMP[12].x, TEMP[9].wwww
>316:     ENDIF
>317:     MOV TEMP[10].w, TEMP[12].xxxx
>318:     MOV TEMP[9], TEMP[10]
>319:     ABS TEMP[12].x, TEMP[7].yyyy
>320:     FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>321:     UIF TEMP[12].xxxx :0
>322:       MOV TEMP[12].x, CONST[78].xxxx
>323:     ELSE :0
>324:       MOV TEMP[12].x, TEMP[8].xxxx
>325:     ENDIF
>326:     MOV TEMP[10].x, TEMP[12].xxxx
>327:     ABS TEMP[12].x, TEMP[7].yyyy
>328:     FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>329:     UIF TEMP[12].xxxx :0
>330:       MOV TEMP[12].x, CONST[78].yyyy
>331:     ELSE :0
>332:       MOV TEMP[12].x, TEMP[8].yyyy
>333:     ENDIF
>334:     MOV TEMP[10].y, TEMP[12].xxxx
>335:     ABS TEMP[12].x, TEMP[7].yyyy
>336:     FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>337:     UIF TEMP[12].xxxx :0
>338:       MOV TEMP[12].x, CONST[78].zzzz
>339:     ELSE :0
>340:       MOV TEMP[12].x, TEMP[8].zzzz
>341:     ENDIF
>342:     MOV TEMP[10].z, TEMP[12].xxxx
>343:     ABS TEMP[12].x, TEMP[7].yyyy
>344:     FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>345:     UIF TEMP[12].xxxx :0
>346:       MOV TEMP[12].x, CONST[78].wwww
>347:     ELSE :0
>348:       MOV TEMP[12].x, TEMP[8].wwww
>349:     ENDIF
>350:     MOV TEMP[10].w, TEMP[12].xxxx
>351:     MOV TEMP[8], TEMP[10]
>352:     ABS TEMP[12].x, TEMP[7].zzzz
>353:     FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>354:     UIF TEMP[12].xxxx :0
>355:       MOV TEMP[12].x, CONST[81].xxxx
>356:     ELSE :0
>357:       MOV TEMP[12].x, TEMP[9].xxxx
>358:     ENDIF
>359:     MOV TEMP[10].x, TEMP[12].xxxx
>360:     ABS TEMP[12].x, TEMP[7].zzzz
>361:     FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>362:     UIF TEMP[12].xxxx :0
>363:       MOV TEMP[12].x, CONST[81].yyyy
>364:     ELSE :0
>365:       MOV TEMP[12].x, TEMP[9].yyyy
>366:     ENDIF
>367:     MOV TEMP[10].y, TEMP[12].xxxx
>368:     ABS TEMP[12].x, TEMP[7].zzzz
>369:     FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>370:     UIF TEMP[12].xxxx :0
>371:       MOV TEMP[12].x, CONST[81].zzzz
>372:     ELSE :0
>373:       MOV TEMP[12].x, TEMP[9].zzzz
>374:     ENDIF
>375:     MOV TEMP[10].z, TEMP[12].xxxx
>376:     ABS TEMP[12].x, TEMP[7].zzzz
>377:     FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>378:     UIF TEMP[12].xxxx :0
>379:       MOV TEMP[12].x, CONST[81].wwww
>380:     ELSE :0
>381:       MOV TEMP[12].x, TEMP[9].wwww
>382:     ENDIF
>383:     MOV TEMP[10].w, TEMP[12].xxxx
>384:     MOV TEMP[9], TEMP[10]
>385:     ABS TEMP[12].x, TEMP[7].zzzz
>386:     FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>387:     UIF TEMP[12].xxxx :0
>388:       MOV TEMP[12].x, CONST[82].xxxx
>389:     ELSE :0
>390:       MOV TEMP[12].x, TEMP[8].xxxx
>391:     ENDIF
>392:     MOV TEMP[10].x, TEMP[12].xxxx
>393:     ABS TEMP[12].x, TEMP[7].zzzz
>394:     FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>395:     UIF TEMP[12].xxxx :0
>396:       MOV TEMP[12].x, CONST[82].yyyy
>397:     ELSE :0
>398:       MOV TEMP[12].x, TEMP[8].yyyy
>399:     ENDIF
>400:     MOV TEMP[10].y, TEMP[12].xxxx
>401:     ABS TEMP[12].x, TEMP[7].zzzz
>402:     FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>403:     UIF TEMP[12].xxxx :0
>404:       MOV TEMP[12].x, CONST[82].zzzz
>405:     ELSE :0
>406:       MOV TEMP[12].x, TEMP[8].zzzz
>407:     ENDIF
>408:     MOV TEMP[10].z, TEMP[12].xxxx
>409:     ABS TEMP[12].x, TEMP[7].zzzz
>410:     FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>411:     UIF TEMP[12].xxxx :0
>412:       MOV TEMP[12].x, CONST[82].wwww
>413:     ELSE :0
>414:       MOV TEMP[12].x, TEMP[8].wwww
>415:     ENDIF
>416:     MOV TEMP[10].w, TEMP[12].xxxx
>417:     DP4 TEMP[9].x, TEMP[1], TEMP[9]
>418:     MOV_SAT TEMP[9].x, TEMP[9].xxxx
>419:     DP4 TEMP[12].x, TEMP[1], TEMP[10]
>420:     MOV_SAT TEMP[12].x, TEMP[12].xxxx
>421:     MOV TEMP[9].y, TEMP[12].xxxx
>422:     ABS TEMP[12].x, TEMP[7].xxxx
>423:     FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>424:     UIF TEMP[12].xxxx :0
>425:       MOV TEMP[12].x, CONST[86].zzzz
>426:     ELSE :0
>427:       MOV TEMP[12].x, IMM[0].yyyy
>428:     ENDIF
>429:     ABS TEMP[13].x, TEMP[7].xxxx
>430:     FSGE TEMP[13].x, -TEMP[13].xxxx, IMM[0].yyyy
>431:     UIF TEMP[13].xxxx :0
>432:       MOV TEMP[13].x, CONST[86].wwww
>433:     ELSE :0
>434:       MOV TEMP[13].x, IMM[0].yyyy
>435:     ENDIF
>436:     MOV TEMP[10].y, TEMP[13].xxxx
>437:     ABS TEMP[13].x, TEMP[7].xxxx
>438:     FSGE TEMP[13].x, -TEMP[13].xxxx, IMM[0].yyyy
>439:     UIF TEMP[13].xxxx :0
>440:       MOV TEMP[13].x, CONST[86].xxxx
>441:     ELSE :0
>442:       MOV TEMP[13].x, IMM[0].yyyy
>443:     ENDIF
>444:     MOV TEMP[10].z, TEMP[13].xxxx
>445:     ABS TEMP[13].x, TEMP[7].xxxx
>446:     FSGE TEMP[13].x, -TEMP[13].xxxx, IMM[0].yyyy
>447:     UIF TEMP[13].xxxx :0
>448:       MOV TEMP[13].x, CONST[86].yyyy
>449:     ELSE :0
>450:       MOV TEMP[13].x, IMM[0].yyyy
>451:     ENDIF
>452:     MOV TEMP[10].w, TEMP[13].xxxx
>453:     ABS TEMP[13].x, TEMP[7].yyyy
>454:     FSGE TEMP[13].x, -TEMP[13].xxxx, IMM[0].yyyy
>455:     UIF TEMP[13].xxxx :0
>456:       MOV TEMP[13].x, CONST[87].zzzz
>457:     ELSE :0
>458:       MOV TEMP[13].x, TEMP[12].xxxx
>459:     ENDIF
>460:     ABS TEMP[12].x, TEMP[7].yyyy
>461:     FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>462:     UIF TEMP[12].xxxx :0
>463:       MOV TEMP[12].x, CONST[87].wwww
>464:     ELSE :0
>465:       MOV TEMP[12].x, TEMP[10].yyyy
>466:     ENDIF
>467:     MOV TEMP[10].y, TEMP[12].xxxx
>468:     ABS TEMP[12].x, TEMP[7].yyyy
>469:     FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>470:     UIF TEMP[12].xxxx :0
>471:       MOV TEMP[12].x, CONST[87].xxxx
>472:     ELSE :0
>473:       MOV TEMP[12].x, TEMP[10].zzzz
>474:     ENDIF
>475:     MOV TEMP[10].z, TEMP[12].xxxx
>476:     ABS TEMP[12].x, TEMP[7].yyyy
>477:     FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>478:     UIF TEMP[12].xxxx :0
>479:       MOV TEMP[12].x, CONST[87].yyyy
>480:     ELSE :0
>481:       MOV TEMP[12].x, TEMP[10].wwww
>482:     ENDIF
>483:     MOV TEMP[10].w, TEMP[12].xxxx
>484:     ABS TEMP[12].x, TEMP[7].zzzz
>485:     FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>486:     UIF TEMP[12].xxxx :0
>487:       MOV TEMP[12].x, CONST[88].zzzz
>488:     ELSE :0
>489:       MOV TEMP[12].x, TEMP[13].xxxx
>490:     ENDIF
>491:     MOV TEMP[10].x, TEMP[12].xxxx
>492:     ABS TEMP[12].x, TEMP[7].zzzz
>493:     FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>494:     UIF TEMP[12].xxxx :0
>495:       MOV TEMP[12].x, CONST[88].wwww
>496:     ELSE :0
>497:       MOV TEMP[12].x, TEMP[10].yyyy
>498:     ENDIF
>499:     MOV TEMP[10].y, TEMP[12].xxxx
>500:     ABS TEMP[12].x, TEMP[7].zzzz
>501:     FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>502:     UIF TEMP[12].xxxx :0
>503:       MOV TEMP[12].x, CONST[88].xxxx
>504:     ELSE :0
>505:       MOV TEMP[12].x, TEMP[10].zzzz
>506:     ENDIF
>507:     MOV TEMP[10].z, TEMP[12].xxxx
>508:     ABS TEMP[12].x, TEMP[7].zzzz
>509:     FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>510:     UIF TEMP[12].xxxx :0
>511:       MOV TEMP[12].x, CONST[88].yyyy
>512:     ELSE :0
>513:       MOV TEMP[12].x, TEMP[10].wwww
>514:     ENDIF
>515:     MOV TEMP[10].w, TEMP[12].xxxx
>516:     MAD TEMP[4].xy, TEMP[9].xyyy, TEMP[10].xyyy, TEMP[10].zwww
>517:     ADD TEMP[1], TEMP[4], IMM[2].xxyy
>518:     TXL TEMP[10].x, TEMP[1], SAMP[1], SHADOW2D
>519:     MOV TEMP[1].x, TEMP[10].xxxx
>520:     ADD TEMP[3], TEMP[4], IMM[2].zxyy
>521:     ADD TEMP[9], TEMP[4], IMM[2].xzyy
>522:     ADD TEMP[8], TEMP[4], IMM[2].zzyy
>523:     TXL TEMP[10].x, TEMP[3], SAMP[1], SHADOW2D
>524:     MOV TEMP[1].y, TEMP[10].xxxx
>525:     TXL TEMP[10].x, TEMP[9], SAMP[1], SHADOW2D
>526:     MOV TEMP[1].z, TEMP[10].xxxx
>527:     TXL TEMP[10].x, TEMP[8], SAMP[1], SHADOW2D
>528:     MOV TEMP[1].w, TEMP[10].xxxx
>529:     DP4 TEMP[10].x, TEMP[1], IMM[1].wwww
>530:     ADD TEMP[3], TEMP[4], IMM[2].xyyy
>531:     TXL TEMP[12].x, TEMP[3], SAMP[1], SHADOW2D
>532:     MOV TEMP[3].x, TEMP[12].xxxx
>533:     ADD TEMP[9], TEMP[4], IMM[2].zyyy
>534:     TXL TEMP[9].x, TEMP[9], SAMP[1], SHADOW2D
>535:     ADD TEMP[8], TEMP[4], IMM[2].yzyy
>536:     TXL TEMP[8].x, TEMP[8], SAMP[1], SHADOW2D
>537:     ADD TEMP[11], TEMP[4], IMM[2].yxyy
>538:     TXL TEMP[11].x, TEMP[11], SAMP[1], SHADOW2D
>539:     MOV TEMP[3].y, TEMP[9].xxxx
>540:     MOV TEMP[3].z, TEMP[8].xxxx
>541:     MOV TEMP[3].w, TEMP[11].xxxx
>542:     DP4 TEMP[3].x, TEMP[3], IMM[2].wwww
>543:     MOV TEMP[8].xy, TEMP[4].xyyy
>544:     MOV TEMP[8].z, TEMP[5].xxxx
>545:     MOV TEMP[8].w, IMM[0].yyyy
>546:     TXL TEMP[5].x, TEMP[8], SAMP[1], SHADOW2D
>547:     ADD TEMP[1].x, TEMP[3].xxxx, TEMP[10].xxxx
>548:     MAD TEMP[1].x, TEMP[5].xxxx, IMM[3].xxxx, TEMP[1].xxxx
>549:     FSGE TEMP[3].x, TEMP[7].zzzz, IMM[0].yyyy
>550:     UIF TEMP[3].xxxx :0
>551:       MOV TEMP[3].x, IMM[0].xxxx
>552:     ELSE :0
>553:       MOV TEMP[3].x, TEMP[1].xxxx
>554:     ENDIF
>555:     LRP TEMP[4].x, TEMP[6].xxxx, TEMP[2].xxxx, TEMP[3].xxxx
>556:     MOV TEMP[2].x, TEMP[4].xxxx
>557:   ENDIF
>558:   ADD TEMP[1].xyz, -CONST[89].xyzz, IN[3].xyzz
>559:   DP3 TEMP[3].x, TEMP[1].xyzz, TEMP[1].xyzz
>560:   MAD TEMP[1].x, TEMP[3].xxxx, CONST[68].yyyy, CONST[68].xxxx
>561:   MOV_SAT TEMP[3].x, TEMP[1].xxxx
>562:   LRP TEMP[4].x, TEMP[3].xxxx, IMM[0].xxxx, TEMP[2].xxxx
>563: ELSE :0
>564:   MOV TEMP[4].x, IMM[0].xxxx
>565: ENDIF
>566: MAD TEMP[1].xyz, IN[1].xyzz, TEMP[4].xxxx, IN[2].xyzz
>567: ADD TEMP[3].x, TEMP[0].wwww, IMM[1].yyyy
>568: MAD TEMP[3].x, CONST[20].wwww, TEMP[3].xxxx, IMM[0].xxxx
>569: ADD TEMP[4].x, TEMP[0].wwww, CONST[12].xxxx
>570: ADD TEMP[2].xyz, IMM[1].yyyy, CONST[1].xyzz
>571: MOV_SAT TEMP[4].x, TEMP[4].xxxx
>572: MAD TEMP[2].xyz, TEMP[4].xxxx, TEMP[2].xyzz, IMM[0].xxxx
>573: MUL TEMP[1].xyz, TEMP[1].xyzz, TEMP[2].xyzz
>574: MUL TEMP[2].x, TEMP[3].xxxx, CONST[1].wwww
>575: MAD TEMP[3].x, TEMP[2].xxxx, IN[2].wwww, -TEMP[2].xxxx
>576: MAD TEMP[2].x, CONST[12].wwww, TEMP[3].xxxx, TEMP[2].xxxx
>577: MUL TEMP[0].xyz, TEMP[0].xyzz, TEMP[1].xyzz
>578: ADD TEMP[1].xyz, CONST[20].xyzz, -IN[3].xyzz
>579: DP3 TEMP[1].x, TEMP[1].xyzz, TEMP[1].xyzz
>580: SQRT TEMP[1].x, TEMP[1].xxxx
>581: MAD TEMP[1].x, TEMP[1].xxxx, CONST[21].wwww, CONST[21].xxxx
>582: MOV_SAT TEMP[1].x, TEMP[1].xxxx
>583: MIN TEMP[1].x, TEMP[1].xxxx, CONST[21].zzzz
>584: ABS TEMP[3].x, CONST[12].yyyy
>585: MUL TEMP[4].xyz, TEMP[0].xyzz, CONST[30].xxxx
>586: MUL TEMP[5].x, CONST[29].wwww, IN[3].wwww
>587: FSGE TEMP[3].x, -TEMP[3].xxxx, IMM[0].yyyy
>588: UIF TEMP[3].xxxx :0
>589:   MOV TEMP[3].x, TEMP[2].xxxx
>590: ELSE :0
>591:   MOV TEMP[3].x, TEMP[5].xxxx
>592: ENDIF
>593: MOV TEMP[2].w, TEMP[3].xxxx
>594: MUL TEMP[1].x, TEMP[1].xxxx, TEMP[1].xxxx
>595: MAD TEMP[0].xyz, TEMP[0].xyzz, -CONST[30].xxxx, CONST[29].xyzz
>596: MAD TEMP[2].xyz, TEMP[1].xxxx, TEMP[0].xyzz, TEMP[4].xyzz
>597: MOV OUT[0], TEMP[2]
>598: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %24 = load <16 x i8>, <16 x i8> addrspace(2)* %23, align 16, !tbaa !0
>  %25 = call float @llvm.SI.load.const(<16 x i8> %24, i32 16)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %24, i32 20)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %24, i32 24)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %24, i32 28)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %24, i32 192)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %24, i32 196)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %24, i32 204)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %24, i32 320)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %24, i32 324)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %24, i32 328)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %24, i32 332)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %24, i32 336)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %24, i32 344)
>  %38 = call float @llvm.SI.load.const(<16 x i8> %24, i32 348)
>  %39 = call float @llvm.SI.load.const(<16 x i8> %24, i32 464)
>  %40 = call float @llvm.SI.load.const(<16 x i8> %24, i32 468)
>  %41 = call float @llvm.SI.load.const(<16 x i8> %24, i32 472)
>  %42 = call float @llvm.SI.load.const(<16 x i8> %24, i32 476)
>  %43 = call float @llvm.SI.load.const(<16 x i8> %24, i32 480)
>  %44 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1080)
>  %45 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1084)
>  %46 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1088)
>  %47 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1092)
>  %48 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1168)
>  %49 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1172)
>  %50 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1176)
>  %51 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1180)
>  %52 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1184)
>  %53 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1188)
>  %54 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1192)
>  %55 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1196)
>  %56 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1232)
>  %57 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1236)
>  %58 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1240)
>  %59 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1244)
>  %60 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1248)
>  %61 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1252)
>  %62 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1256)
>  %63 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1260)
>  %64 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1296)
>  %65 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1300)
>  %66 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1304)
>  %67 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1308)
>  %68 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1312)
>  %69 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1316)
>  %70 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1320)
>  %71 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1324)
>  %72 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1376)
>  %73 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1380)
>  %74 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1384)
>  %75 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1388)
>  %76 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1392)
>  %77 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1396)
>  %78 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1400)
>  %79 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1404)
>  %80 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1408)
>  %81 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1412)
>  %82 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1416)
>  %83 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1420)
>  %84 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1424)
>  %85 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1428)
>  %86 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1432)
>  %87 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1440)
>  %88 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %89 = load <8 x i32>, <8 x i32> addrspace(2)* %88, align 32, !tbaa !0
>  %90 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %91 = load <4 x i32>, <4 x i32> addrspace(2)* %90, align 16, !tbaa !0
>  %92 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 1
>  %93 = load <8 x i32>, <8 x i32> addrspace(2)* %92, align 32, !tbaa !0
>  %94 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 1
>  %95 = load <4 x i32>, <4 x i32> addrspace(2)* %94, align 16, !tbaa !0
>  %96 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %8)
>  %97 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %8)
>  %98 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %6, <2 x i32> %8)
>  %99 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %6, <2 x i32> %8)
>  %100 = call float @llvm.SI.fs.interp(i32 2, i32 1, i32 %6, <2 x i32> %8)
>  %101 = call float @llvm.SI.fs.interp(i32 0, i32 2, i32 %6, <2 x i32> %8)
>  %102 = call float @llvm.SI.fs.interp(i32 1, i32 2, i32 %6, <2 x i32> %8)
>  %103 = call float @llvm.SI.fs.interp(i32 2, i32 2, i32 %6, <2 x i32> %8)
>  %104 = call float @llvm.SI.fs.interp(i32 3, i32 2, i32 %6, <2 x i32> %8)
>  %105 = call float @llvm.SI.fs.interp(i32 0, i32 3, i32 %6, <2 x i32> %8)
>  %106 = call float @llvm.SI.fs.interp(i32 1, i32 3, i32 %6, <2 x i32> %8)
>  %107 = call float @llvm.SI.fs.interp(i32 2, i32 3, i32 %6, <2 x i32> %8)
>  %108 = call float @llvm.SI.fs.interp(i32 3, i32 3, i32 %6, <2 x i32> %8)
>  %109 = bitcast float %96 to i32
>  %110 = bitcast float %97 to i32
>  %111 = insertelement <2 x i32> undef, i32 %109, i32 0
>  %112 = insertelement <2 x i32> %111, i32 %110, i32 1
>  %113 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %112, <8 x i32> %89, <4 x i32> %91, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %114 = extractelement <4 x float> %113, i32 0
>  %115 = extractelement <4 x float> %113, i32 1
>  %116 = extractelement <4 x float> %113, i32 2
>  %117 = extractelement <4 x float> %113, i32 3
>  %118 = bitcast float %87 to i32
>  %119 = icmp eq i32 %118, 0
>  br i1 %119, label %ENDIF, label %IF
>
>IF:                                               ; preds = %main_body
>  %120 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1372)
>  %121 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1368)
>  %122 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1364)
>  %123 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1360)
>  %124 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1148)
>  %125 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1144)
>  %126 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1140)
>  %127 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1136)
>  %128 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1132)
>  %129 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1128)
>  %130 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1124)
>  %131 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1120)
>  %132 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1116)
>  %133 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1112)
>  %134 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1108)
>  %135 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1104)
>  %136 = fadd float %105, 0.000000e+00
>  %137 = fadd float %106, 0.000000e+00
>  %138 = fadd float %107, 0.000000e+00
>  %139 = fmul float %105, 0.000000e+00
>  %140 = fadd float %139, 1.000000e+00
>  %141 = fmul float %136, %135
>  %142 = fmul float %137, %134
>  %143 = fadd float %141, %142
>  %144 = fmul float %138, %133
>  %145 = fadd float %143, %144
>  %146 = fmul float %140, %132
>  %147 = fadd float %145, %146
>  %148 = fmul float %136, %131
>  %149 = fmul float %137, %130
>  %150 = fadd float %148, %149
>  %151 = fmul float %138, %129
>  %152 = fadd float %150, %151
>  %153 = fmul float %140, %128
>  %154 = fadd float %152, %153
>  %155 = call float @llvm.AMDIL.clamp.(float %147, float 0.000000e+00, float 1.000000e+00)
>  %156 = call float @llvm.AMDIL.clamp.(float %154, float 0.000000e+00, float 1.000000e+00)
>  %157 = fsub float %155, %147
>  %158 = fsub float %156, %154
>  %159 = fadd float %157, %158
>  %160 = fmul float %136, %48
>  %161 = fmul float %137, %49
>  %162 = fadd float %160, %161
>  %163 = fmul float %138, %50
>  %164 = fadd float %162, %163
>  %165 = fmul float %140, %51
>  %166 = fadd float %164, %165
>  %167 = fmul float %136, %52
>  %168 = fmul float %137, %53
>  %169 = fadd float %167, %168
>  %170 = fmul float %138, %54
>  %171 = fadd float %169, %170
>  %172 = fmul float %140, %55
>  %173 = fadd float %171, %172
>  %174 = call float @llvm.AMDIL.clamp.(float %166, float 0.000000e+00, float 1.000000e+00)
>  %175 = call float @llvm.AMDIL.clamp.(float %173, float 0.000000e+00, float 1.000000e+00)
>  %176 = fsub float %174, %166
>  %177 = fsub float %175, %173
>  %178 = fadd float %176, %177
>  %179 = fmul float %136, %56
>  %180 = fmul float %137, %57
>  %181 = fadd float %179, %180
>  %182 = fmul float %138, %58
>  %183 = fadd float %181, %182
>  %184 = fmul float %140, %59
>  %185 = fadd float %183, %184
>  %186 = fmul float %136, %60
>  %187 = fmul float %137, %61
>  %188 = fadd float %186, %187
>  %189 = fmul float %138, %62
>  %190 = fadd float %188, %189
>  %191 = fmul float %140, %63
>  %192 = fadd float %190, %191
>  %193 = call float @llvm.fabs.f32(float %178)
>  %194 = fcmp ole float %193, -0.000000e+00
>  %. = select i1 %194, float %166, float %185
>  %195 = call float @llvm.fabs.f32(float %178)
>  %196 = fcmp ole float %195, -0.000000e+00
>  %temp24.0 = select i1 %196, float %173, float %192
>  %197 = call float @llvm.fabs.f32(float %178)
>  %198 = fcmp ole float %197, -0.000000e+00
>  %.236 = select i1 %198, float 1.000000e+00, float 2.000000e+00
>  %199 = call float @llvm.fabs.f32(float %159)
>  %200 = fcmp ole float %199, -0.000000e+00
>  %temp24.2 = select i1 %200, float %147, float %.
>  %201 = call float @llvm.fabs.f32(float %159)
>  %202 = fcmp ole float %201, -0.000000e+00
>  %.temp24.0 = select i1 %202, float %154, float %temp24.0
>  %203 = call float @llvm.fabs.f32(float %159)
>  %204 = fcmp ole float %203, -0.000000e+00
>  %temp12.1 = select i1 %204, float 0.000000e+00, float %.236
>  %205 = fmul float %136, %127
>  %206 = fmul float %137, %126
>  %207 = fadd float %205, %206
>  %208 = fmul float %138, %125
>  %209 = fadd float %207, %208
>  %210 = fmul float %140, %124
>  %211 = fadd float %209, %210
>  %212 = fadd float %temp24.2, -5.000000e-01
>  %213 = fadd float %.temp24.0, -5.000000e-01
>  %214 = call float @llvm.fabs.f32(float %212)
>  %215 = call float @llvm.fabs.f32(float %213)
>  %216 = fsub float %214, %44
>  %217 = fsub float %215, %44
>  %218 = fmul float %216, %45
>  %219 = fmul float %217, %45
>  %220 = call float @llvm.AMDIL.clamp.(float %218, float 0.000000e+00, float 1.000000e+00)
>  %221 = call float @llvm.AMDIL.clamp.(float %219, float 0.000000e+00, float 1.000000e+00)
>  %222 = fsub float 1.000000e+00, %220
>  %223 = fsub float 1.000000e+00, %221
>  %224 = fmul float %223, %222
>  %225 = call float @llvm.AMDIL.clamp.(float %temp24.2, float 0.000000e+00, float 1.000000e+00)
>  %226 = call float @llvm.AMDIL.clamp.(float %.temp24.0, float 0.000000e+00, float 1.000000e+00)
>  %227 = fadd float %temp12.1, -1.000000e+00
>  %228 = fadd float %temp12.1, -2.000000e+00
>  %229 = call float @llvm.fabs.f32(float %temp12.1)
>  %230 = fcmp ole float %229, -0.000000e+00
>  %.237 = select i1 %230, float %121, float 0.000000e+00
>  %231 = call float @llvm.fabs.f32(float %temp12.1)
>  %232 = fcmp ole float %231, -0.000000e+00
>  %temp48.0 = select i1 %232, float %120, float 0.000000e+00
>  %233 = call float @llvm.fabs.f32(float %temp12.1)
>  %234 = fcmp ole float %233, -0.000000e+00
>  %.238 = select i1 %234, float %123, float 0.000000e+00
>  %235 = call float @llvm.fabs.f32(float %temp12.1)
>  %236 = fcmp ole float %235, -0.000000e+00
>  %temp48.2 = select i1 %236, float %122, float 0.000000e+00
>  %237 = call float @llvm.fabs.f32(float %227)
>  %238 = fcmp ole float %237, -0.000000e+00
>  %..237 = select i1 %238, float %74, float %.237
>  %239 = call float @llvm.fabs.f32(float %227)
>  %240 = fcmp ole float %239, -0.000000e+00
>  %temp44.2 = select i1 %240, float %75, float %temp48.0
>  %241 = call float @llvm.fabs.f32(float %227)
>  %242 = fcmp ole float %241, -0.000000e+00
>  %..238 = select i1 %242, float %72, float %.238
>  %243 = call float @llvm.fabs.f32(float %227)
>  %244 = fcmp ole float %243, -0.000000e+00
>  %temp44.4 = select i1 %244, float %73, float %temp48.2
>  %245 = call float @llvm.fabs.f32(float %228)
>  %246 = fcmp ole float %245, -0.000000e+00
>  %...237 = select i1 %246, float %78, float %..237
>  %247 = call float @llvm.fabs.f32(float %228)
>  %248 = fcmp ole float %247, -0.000000e+00
>  %temp44.6 = select i1 %248, float %79, float %temp44.2
>  %249 = call float @llvm.fabs.f32(float %228)
>  %250 = fcmp ole float %249, -0.000000e+00
>  %...238 = select i1 %250, float %76, float %..238
>  %251 = call float @llvm.fabs.f32(float %228)
>  %252 = fcmp ole float %251, -0.000000e+00
>  %temp44.8 = select i1 %252, float %77, float %temp44.4
>  %253 = fmul float %225, %...237
>  %254 = fadd float %253, %...238
>  %255 = fmul float %226, %temp44.6
>  %256 = fadd float %255, %temp44.8
>  %257 = fadd float %254, 0x3F40000000000000
>  %258 = fadd float %256, 0x3F40000000000000
>  %259 = fadd float %211, 0.000000e+00
>  %260 = bitcast float %259 to i32
>  %261 = bitcast float %257 to i32
>  %262 = bitcast float %258 to i32
>  %263 = insertelement <4 x i32> undef, i32 %260, i32 0
>  %264 = insertelement <4 x i32> %263, i32 %261, i32 1
>  %265 = insertelement <4 x i32> %264, i32 %262, i32 2
>  %266 = insertelement <4 x i32> %265, i32 0, i32 3
>  %267 = call <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32> %266, <8 x i32> %93, <4 x i32> %95, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %268 = extractelement <4 x float> %267, i32 0
>  %269 = fadd float %254, 0xBF40000000000000
>  %270 = fadd float %256, 0x3F40000000000000
>  %271 = fadd float %211, 0.000000e+00
>  %272 = fadd float %254, 0x3F40000000000000
>  %273 = fadd float %256, 0xBF40000000000000
>  %274 = fadd float %211, 0.000000e+00
>  %275 = fadd float %254, 0xBF40000000000000
>  %276 = fadd float %256, 0xBF40000000000000
>  %277 = fadd float %211, 0.000000e+00
>  %278 = bitcast float %271 to i32
>  %279 = bitcast float %269 to i32
>  %280 = bitcast float %270 to i32
>  %281 = insertelement <4 x i32> undef, i32 %278, i32 0
>  %282 = insertelement <4 x i32> %281, i32 %279, i32 1
>  %283 = insertelement <4 x i32> %282, i32 %280, i32 2
>  %284 = insertelement <4 x i32> %283, i32 0, i32 3
>  %285 = call <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32> %284, <8 x i32> %93, <4 x i32> %95, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %286 = extractelement <4 x float> %285, i32 0
>  %287 = bitcast float %274 to i32
>  %288 = bitcast float %272 to i32
>  %289 = bitcast float %273 to i32
>  %290 = insertelement <4 x i32> undef, i32 %287, i32 0
>  %291 = insertelement <4 x i32> %290, i32 %288, i32 1
>  %292 = insertelement <4 x i32> %291, i32 %289, i32 2
>  %293 = insertelement <4 x i32> %292, i32 0, i32 3
>  %294 = call <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32> %293, <8 x i32> %93, <4 x i32> %95, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %295 = extractelement <4 x float> %294, i32 0
>  %296 = bitcast float %277 to i32
>  %297 = bitcast float %275 to i32
>  %298 = bitcast float %276 to i32
>  %299 = insertelement <4 x i32> undef, i32 %296, i32 0
>  %300 = insertelement <4 x i32> %299, i32 %297, i32 1
>  %301 = insertelement <4 x i32> %300, i32 %298, i32 2
>  %302 = insertelement <4 x i32> %301, i32 0, i32 3
>  %303 = call <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32> %302, <8 x i32> %93, <4 x i32> %95, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %304 = extractelement <4 x float> %303, i32 0
>  %305 = fmul float %268, 6.250000e-02
>  %306 = fmul float %286, 6.250000e-02
>  %307 = fadd float %305, %306
>  %308 = fmul float %295, 6.250000e-02
>  %309 = fadd float %307, %308
>  %310 = fmul float %304, 6.250000e-02
>  %311 = fadd float %309, %310
>  %312 = fadd float %254, 0x3F40000000000000
>  %313 = fadd float %256, 0.000000e+00
>  %314 = fadd float %211, 0.000000e+00
>  %315 = bitcast float %314 to i32
>  %316 = bitcast float %312 to i32
>  %317 = bitcast float %313 to i32
>  %318 = insertelement <4 x i32> undef, i32 %315, i32 0
>  %319 = insertelement <4 x i32> %318, i32 %316, i32 1
>  %320 = insertelement <4 x i32> %319, i32 %317, i32 2
>  %321 = insertelement <4 x i32> %320, i32 0, i32 3
>  %322 = call <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32> %321, <8 x i32> %93, <4 x i32> %95, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %323 = extractelement <4 x float> %322, i32 0
>  %324 = fadd float %254, 0xBF40000000000000
>  %325 = fadd float %256, 0.000000e+00
>  %326 = fadd float %211, 0.000000e+00
>  %327 = bitcast float %326 to i32
>  %328 = bitcast float %324 to i32
>  %329 = bitcast float %325 to i32
>  %330 = insertelement <4 x i32> undef, i32 %327, i32 0
>  %331 = insertelement <4 x i32> %330, i32 %328, i32 1
>  %332 = insertelement <4 x i32> %331, i32 %329, i32 2
>  %333 = insertelement <4 x i32> %332, i32 0, i32 3
>  %334 = call <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32> %333, <8 x i32> %93, <4 x i32> %95, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %335 = extractelement <4 x float> %334, i32 0
>  %336 = fadd float %254, 0.000000e+00
>  %337 = fadd float %256, 0xBF40000000000000
>  %338 = fadd float %211, 0.000000e+00
>  %339 = bitcast float %338 to i32
>  %340 = bitcast float %336 to i32
>  %341 = bitcast float %337 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 0, i32 3
>  %346 = call <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32> %345, <8 x i32> %93, <4 x i32> %95, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %347 = extractelement <4 x float> %346, i32 0
>  %348 = fadd float %254, 0.000000e+00
>  %349 = fadd float %256, 0x3F40000000000000
>  %350 = fadd float %211, 0.000000e+00
>  %351 = bitcast float %350 to i32
>  %352 = bitcast float %348 to i32
>  %353 = bitcast float %349 to i32
>  %354 = insertelement <4 x i32> undef, i32 %351, i32 0
>  %355 = insertelement <4 x i32> %354, i32 %352, i32 1
>  %356 = insertelement <4 x i32> %355, i32 %353, i32 2
>  %357 = insertelement <4 x i32> %356, i32 0, i32 3
>  %358 = call <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32> %357, <8 x i32> %93, <4 x i32> %95, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %359 = extractelement <4 x float> %358, i32 0
>  %360 = fmul float %323, 1.250000e-01
>  %361 = fmul float %335, 1.250000e-01
>  %362 = fadd float %360, %361
>  %363 = fmul float %347, 1.250000e-01
>  %364 = fadd float %362, %363
>  %365 = fmul float %359, 1.250000e-01
>  %366 = fadd float %364, %365
>  %367 = bitcast float %211 to i32
>  %368 = bitcast float %254 to i32
>  %369 = bitcast float %256 to i32
>  %370 = insertelement <4 x i32> undef, i32 %367, i32 0
>  %371 = insertelement <4 x i32> %370, i32 %368, i32 1
>  %372 = insertelement <4 x i32> %371, i32 %369, i32 2
>  %373 = insertelement <4 x i32> %372, i32 0, i32 3
>  %374 = call <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32> %373, <8 x i32> %93, <4 x i32> %95, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %375 = extractelement <4 x float> %374, i32 0
>  %376 = fadd float %366, %311
>  %377 = fmul float %375, 2.500000e-01
>  %378 = fadd float %377, %376
>  %379 = fcmp olt float %224, 1.000000e+00
>  br i1 %379, label %IF119, label %ENDIF118
>
>ENDIF:                                            ; preds = %main_body, %ENDIF118
>  %temp16.0 = phi float [ %682, %ENDIF118 ], [ 1.000000e+00, %main_body ]
>  %380 = fmul float %98, %temp16.0
>  %381 = fadd float %380, %101
>  %382 = fmul float %99, %temp16.0
>  %383 = fadd float %382, %102
>  %384 = fmul float %100, %temp16.0
>  %385 = fadd float %384, %103
>  %386 = fadd float %117, -1.000000e+00
>  %387 = fmul float %35, %386
>  %388 = fadd float %387, 1.000000e+00
>  %389 = fadd float %117, %29
>  %390 = fadd float %25, -1.000000e+00
>  %391 = fadd float %26, -1.000000e+00
>  %392 = fadd float %27, -1.000000e+00
>  %393 = call float @llvm.AMDIL.clamp.(float %389, float 0.000000e+00, float 1.000000e+00)
>  %394 = fmul float %393, %390
>  %395 = fadd float %394, 1.000000e+00
>  %396 = fmul float %393, %391
>  %397 = fadd float %396, 1.000000e+00
>  %398 = fmul float %393, %392
>  %399 = fadd float %398, 1.000000e+00
>  %400 = fmul float %381, %395
>  %401 = fmul float %383, %397
>  %402 = fmul float %385, %399
>  %403 = fmul float %388, %28
>  %404 = fmul float %403, %104
>  %405 = fsub float %404, %403
>  %406 = fmul float %31, %405
>  %407 = fadd float %406, %403
>  %408 = fmul float %114, %400
>  %409 = fmul float %115, %401
>  %410 = fmul float %116, %402
>  %411 = fsub float %32, %105
>  %412 = fsub float %33, %106
>  %413 = fsub float %34, %107
>  %414 = fmul float %411, %411
>  %415 = fmul float %412, %412
>  %416 = fadd float %415, %414
>  %417 = fmul float %413, %413
>  %418 = fadd float %416, %417
>  %419 = call float @llvm.sqrt.f32(float %418)
>  %420 = fmul float %419, %38
>  %421 = fadd float %420, %36
>  %422 = call float @llvm.AMDIL.clamp.(float %421, float 0.000000e+00, float 1.000000e+00)
>  %423 = call float @llvm.minnum.f32(float %422, float %37)
>  %424 = call float @llvm.fabs.f32(float %30)
>  %425 = fmul float %408, %43
>  %426 = fmul float %409, %43
>  %427 = fmul float %410, %43
>  %428 = fmul float %42, %108
>  %429 = fcmp ole float %424, -0.000000e+00
>  %.235 = select i1 %429, float %407, float %428
>  %430 = fmul float %423, %423
>  %431 = fmul float %43, %408
>  %432 = fsub float %39, %431
>  %433 = fmul float %43, %409
>  %434 = fsub float %40, %433
>  %435 = fmul float %43, %410
>  %436 = fsub float %41, %435
>  %437 = fmul float %430, %432
>  %438 = fadd float %437, %425
>  %439 = fmul float %430, %434
>  %440 = fadd float %439, %426
>  %441 = fmul float %430, %436
>  %442 = fadd float %441, %427
>  %443 = call i32 @llvm.SI.packf16(float %438, float %440)
>  %444 = bitcast i32 %443 to float
>  %445 = call i32 @llvm.SI.packf16(float %442, float %.235)
>  %446 = bitcast i32 %445 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %444, float %446, float %444, float %446)
>  ret void
>
>IF119:                                            ; preds = %IF
>  %447 = fadd float %temp12.1, 0.000000e+00
>  %448 = fadd float %temp12.1, -1.000000e+00
>  %449 = fadd float %temp12.1, -2.000000e+00
>  %450 = call float @llvm.fabs.f32(float %447)
>  %451 = fcmp ole float %450, -0.000000e+00
>  %.239 = select i1 %451, float %48, float 0.000000e+00
>  %452 = call float @llvm.fabs.f32(float %447)
>  %453 = fcmp ole float %452, -0.000000e+00
>  %temp48.5 = select i1 %453, float %49, float 0.000000e+00
>  %454 = call float @llvm.fabs.f32(float %447)
>  %455 = fcmp ole float %454, -0.000000e+00
>  %.240 = select i1 %455, float %50, float 0.000000e+00
>  %456 = call float @llvm.fabs.f32(float %447)
>  %457 = fcmp ole float %456, -0.000000e+00
>  %temp48.7 = select i1 %457, float %51, float 0.000000e+00
>  %458 = call float @llvm.fabs.f32(float %447)
>  %459 = fcmp ole float %458, -0.000000e+00
>  %.241 = select i1 %459, float %52, float 0.000000e+00
>  %460 = call float @llvm.fabs.f32(float %447)
>  %461 = fcmp ole float %460, -0.000000e+00
>  %temp48.9 = select i1 %461, float %53, float 0.000000e+00
>  %462 = call float @llvm.fabs.f32(float %447)
>  %463 = fcmp ole float %462, -0.000000e+00
>  %.242 = select i1 %463, float %54, float 0.000000e+00
>  %464 = call float @llvm.fabs.f32(float %447)
>  %465 = fcmp ole float %464, -0.000000e+00
>  %temp48.11 = select i1 %465, float %55, float 0.000000e+00
>  %466 = call float @llvm.fabs.f32(float %448)
>  %467 = fcmp ole float %466, -0.000000e+00
>  %..239 = select i1 %467, float %56, float %.239
>  %468 = call float @llvm.fabs.f32(float %448)
>  %469 = fcmp ole float %468, -0.000000e+00
>  %temp48.13 = select i1 %469, float %57, float %temp48.5
>  %470 = call float @llvm.fabs.f32(float %448)
>  %471 = fcmp ole float %470, -0.000000e+00
>  %..240 = select i1 %471, float %58, float %.240
>  %472 = call float @llvm.fabs.f32(float %448)
>  %473 = fcmp ole float %472, -0.000000e+00
>  %temp48.15 = select i1 %473, float %59, float %temp48.7
>  %474 = call float @llvm.fabs.f32(float %448)
>  %475 = fcmp ole float %474, -0.000000e+00
>  %..241 = select i1 %475, float %60, float %.241
>  %476 = call float @llvm.fabs.f32(float %448)
>  %477 = fcmp ole float %476, -0.000000e+00
>  %temp48.17 = select i1 %477, float %61, float %temp48.9
>  %478 = call float @llvm.fabs.f32(float %448)
>  %479 = fcmp ole float %478, -0.000000e+00
>  %..242 = select i1 %479, float %62, float %.242
>  %480 = call float @llvm.fabs.f32(float %448)
>  %481 = fcmp ole float %480, -0.000000e+00
>  %temp48.19 = select i1 %481, float %63, float %temp48.11
>  %482 = call float @llvm.fabs.f32(float %449)
>  %483 = fcmp ole float %482, -0.000000e+00
>  %...239 = select i1 %483, float %64, float %..239
>  %484 = call float @llvm.fabs.f32(float %449)
>  %485 = fcmp ole float %484, -0.000000e+00
>  %temp48.21 = select i1 %485, float %65, float %temp48.13
>  %486 = call float @llvm.fabs.f32(float %449)
>  %487 = fcmp ole float %486, -0.000000e+00
>  %...240 = select i1 %487, float %66, float %..240
>  %488 = call float @llvm.fabs.f32(float %449)
>  %489 = fcmp ole float %488, -0.000000e+00
>  %temp48.23 = select i1 %489, float %67, float %temp48.15
>  %490 = call float @llvm.fabs.f32(float %449)
>  %491 = fcmp ole float %490, -0.000000e+00
>  %...241 = select i1 %491, float %68, float %..241
>  %492 = call float @llvm.fabs.f32(float %449)
>  %493 = fcmp ole float %492, -0.000000e+00
>  %temp48.25 = select i1 %493, float %69, float %temp48.17
>  %494 = call float @llvm.fabs.f32(float %449)
>  %495 = fcmp ole float %494, -0.000000e+00
>  %...242 = select i1 %495, float %70, float %..242
>  %496 = call float @llvm.fabs.f32(float %449)
>  %497 = fcmp ole float %496, -0.000000e+00
>  %temp48.27 = select i1 %497, float %71, float %temp48.19
>  %498 = fmul float %136, %...239
>  %499 = fmul float %137, %temp48.21
>  %500 = fadd float %498, %499
>  %501 = fmul float %138, %...240
>  %502 = fadd float %500, %501
>  %503 = fmul float %140, %temp48.23
>  %504 = fadd float %502, %503
>  %505 = call float @llvm.AMDIL.clamp.(float %504, float 0.000000e+00, float 1.000000e+00)
>  %506 = fmul float %136, %...241
>  %507 = fmul float %137, %temp48.25
>  %508 = fadd float %506, %507
>  %509 = fmul float %138, %...242
>  %510 = fadd float %508, %509
>  %511 = fmul float %140, %temp48.27
>  %512 = fadd float %510, %511
>  %513 = call float @llvm.AMDIL.clamp.(float %512, float 0.000000e+00, float 1.000000e+00)
>  %514 = call float @llvm.fabs.f32(float %447)
>  %515 = fcmp ole float %514, -0.000000e+00
>  %.243 = select i1 %515, float %74, float 0.000000e+00
>  %516 = call float @llvm.fabs.f32(float %447)
>  %517 = fcmp ole float %516, -0.000000e+00
>  %temp52.0 = select i1 %517, float %75, float 0.000000e+00
>  %518 = call float @llvm.fabs.f32(float %447)
>  %519 = fcmp ole float %518, -0.000000e+00
>  %.244 = select i1 %519, float %72, float 0.000000e+00
>  %520 = call float @llvm.fabs.f32(float %447)
>  %521 = fcmp ole float %520, -0.000000e+00
>  %temp52.2 = select i1 %521, float %73, float 0.000000e+00
>  %522 = call float @llvm.fabs.f32(float %448)
>  %523 = fcmp ole float %522, -0.000000e+00
>  %..243 = select i1 %523, float %78, float %.243
>  %524 = call float @llvm.fabs.f32(float %448)
>  %525 = fcmp ole float %524, -0.000000e+00
>  %temp48.29 = select i1 %525, float %79, float %temp52.0
>  %526 = call float @llvm.fabs.f32(float %448)
>  %527 = fcmp ole float %526, -0.000000e+00
>  %..244 = select i1 %527, float %76, float %.244
>  %528 = call float @llvm.fabs.f32(float %448)
>  %529 = fcmp ole float %528, -0.000000e+00
>  %temp48.31 = select i1 %529, float %77, float %temp52.2
>  %530 = call float @llvm.fabs.f32(float %449)
>  %531 = fcmp ole float %530, -0.000000e+00
>  %...243 = select i1 %531, float %82, float %..243
>  %532 = call float @llvm.fabs.f32(float %449)
>  %533 = fcmp ole float %532, -0.000000e+00
>  %temp48.33 = select i1 %533, float %83, float %temp48.29
>  %534 = call float @llvm.fabs.f32(float %449)
>  %535 = fcmp ole float %534, -0.000000e+00
>  %...244 = select i1 %535, float %80, float %..244
>  %536 = call float @llvm.fabs.f32(float %449)
>  %537 = fcmp ole float %536, -0.000000e+00
>  %temp48.35 = select i1 %537, float %81, float %temp48.31
>  %538 = fmul float %505, %...243
>  %539 = fadd float %538, %...244
>  %540 = fmul float %513, %temp48.33
>  %541 = fadd float %540, %temp48.35
>  %542 = fadd float %539, 0x3F40000000000000
>  %543 = fadd float %541, 0x3F40000000000000
>  %544 = fadd float %211, 0.000000e+00
>  %545 = bitcast float %544 to i32
>  %546 = bitcast float %542 to i32
>  %547 = bitcast float %543 to i32
>  %548 = insertelement <4 x i32> undef, i32 %545, i32 0
>  %549 = insertelement <4 x i32> %548, i32 %546, i32 1
>  %550 = insertelement <4 x i32> %549, i32 %547, i32 2
>  %551 = insertelement <4 x i32> %550, i32 0, i32 3
>  %552 = call <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32> %551, <8 x i32> %93, <4 x i32> %95, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %553 = extractelement <4 x float> %552, i32 0
>  %554 = fadd float %539, 0xBF40000000000000
>  %555 = fadd float %541, 0x3F40000000000000
>  %556 = fadd float %211, 0.000000e+00
>  %557 = fadd float %539, 0x3F40000000000000
>  %558 = fadd float %541, 0xBF40000000000000
>  %559 = fadd float %211, 0.000000e+00
>  %560 = fadd float %539, 0xBF40000000000000
>  %561 = fadd float %541, 0xBF40000000000000
>  %562 = fadd float %211, 0.000000e+00
>  %563 = bitcast float %556 to i32
>  %564 = bitcast float %554 to i32
>  %565 = bitcast float %555 to i32
>  %566 = insertelement <4 x i32> undef, i32 %563, i32 0
>  %567 = insertelement <4 x i32> %566, i32 %564, i32 1
>  %568 = insertelement <4 x i32> %567, i32 %565, i32 2
>  %569 = insertelement <4 x i32> %568, i32 0, i32 3
>  %570 = call <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32> %569, <8 x i32> %93, <4 x i32> %95, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %571 = extractelement <4 x float> %570, i32 0
>  %572 = bitcast float %559 to i32
>  %573 = bitcast float %557 to i32
>  %574 = bitcast float %558 to i32
>  %575 = insertelement <4 x i32> undef, i32 %572, i32 0
>  %576 = insertelement <4 x i32> %575, i32 %573, i32 1
>  %577 = insertelement <4 x i32> %576, i32 %574, i32 2
>  %578 = insertelement <4 x i32> %577, i32 0, i32 3
>  %579 = call <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32> %578, <8 x i32> %93, <4 x i32> %95, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %580 = extractelement <4 x float> %579, i32 0
>  %581 = bitcast float %562 to i32
>  %582 = bitcast float %560 to i32
>  %583 = bitcast float %561 to i32
>  %584 = insertelement <4 x i32> undef, i32 %581, i32 0
>  %585 = insertelement <4 x i32> %584, i32 %582, i32 1
>  %586 = insertelement <4 x i32> %585, i32 %583, i32 2
>  %587 = insertelement <4 x i32> %586, i32 0, i32 3
>  %588 = call <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32> %587, <8 x i32> %93, <4 x i32> %95, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %589 = extractelement <4 x float> %588, i32 0
>  %590 = fmul float %553, 6.250000e-02
>  %591 = fmul float %571, 6.250000e-02
>  %592 = fadd float %590, %591
>  %593 = fmul float %580, 6.250000e-02
>  %594 = fadd float %592, %593
>  %595 = fmul float %589, 6.250000e-02
>  %596 = fadd float %594, %595
>  %597 = fadd float %539, 0x3F40000000000000
>  %598 = fadd float %541, 0.000000e+00
>  %599 = fadd float %211, 0.000000e+00
>  %600 = bitcast float %599 to i32
>  %601 = bitcast float %597 to i32
>  %602 = bitcast float %598 to i32
>  %603 = insertelement <4 x i32> undef, i32 %600, i32 0
>  %604 = insertelement <4 x i32> %603, i32 %601, i32 1
>  %605 = insertelement <4 x i32> %604, i32 %602, i32 2
>  %606 = insertelement <4 x i32> %605, i32 0, i32 3
>  %607 = call <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32> %606, <8 x i32> %93, <4 x i32> %95, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %608 = extractelement <4 x float> %607, i32 0
>  %609 = fadd float %539, 0xBF40000000000000
>  %610 = fadd float %541, 0.000000e+00
>  %611 = fadd float %211, 0.000000e+00
>  %612 = bitcast float %611 to i32
>  %613 = bitcast float %609 to i32
>  %614 = bitcast float %610 to i32
>  %615 = insertelement <4 x i32> undef, i32 %612, i32 0
>  %616 = insertelement <4 x i32> %615, i32 %613, i32 1
>  %617 = insertelement <4 x i32> %616, i32 %614, i32 2
>  %618 = insertelement <4 x i32> %617, i32 0, i32 3
>  %619 = call <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32> %618, <8 x i32> %93, <4 x i32> %95, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %620 = extractelement <4 x float> %619, i32 0
>  %621 = fadd float %539, 0.000000e+00
>  %622 = fadd float %541, 0xBF40000000000000
>  %623 = fadd float %211, 0.000000e+00
>  %624 = bitcast float %623 to i32
>  %625 = bitcast float %621 to i32
>  %626 = bitcast float %622 to i32
>  %627 = insertelement <4 x i32> undef, i32 %624, i32 0
>  %628 = insertelement <4 x i32> %627, i32 %625, i32 1
>  %629 = insertelement <4 x i32> %628, i32 %626, i32 2
>  %630 = insertelement <4 x i32> %629, i32 0, i32 3
>  %631 = call <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32> %630, <8 x i32> %93, <4 x i32> %95, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %632 = extractelement <4 x float> %631, i32 0
>  %633 = fadd float %539, 0.000000e+00
>  %634 = fadd float %541, 0x3F40000000000000
>  %635 = fadd float %211, 0.000000e+00
>  %636 = bitcast float %635 to i32
>  %637 = bitcast float %633 to i32
>  %638 = bitcast float %634 to i32
>  %639 = insertelement <4 x i32> undef, i32 %636, i32 0
>  %640 = insertelement <4 x i32> %639, i32 %637, i32 1
>  %641 = insertelement <4 x i32> %640, i32 %638, i32 2
>  %642 = insertelement <4 x i32> %641, i32 0, i32 3
>  %643 = call <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32> %642, <8 x i32> %93, <4 x i32> %95, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %644 = extractelement <4 x float> %643, i32 0
>  %645 = fmul float %608, 1.250000e-01
>  %646 = fmul float %620, 1.250000e-01
>  %647 = fadd float %645, %646
>  %648 = fmul float %632, 1.250000e-01
>  %649 = fadd float %647, %648
>  %650 = fmul float %644, 1.250000e-01
>  %651 = fadd float %649, %650
>  %652 = bitcast float %211 to i32
>  %653 = bitcast float %539 to i32
>  %654 = bitcast float %541 to i32
>  %655 = insertelement <4 x i32> undef, i32 %652, i32 0
>  %656 = insertelement <4 x i32> %655, i32 %653, i32 1
>  %657 = insertelement <4 x i32> %656, i32 %654, i32 2
>  %658 = insertelement <4 x i32> %657, i32 0, i32 3
>  %659 = call <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32> %658, <8 x i32> %93, <4 x i32> %95, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %660 = extractelement <4 x float> %659, i32 0
>  %661 = fadd float %651, %596
>  %662 = fmul float %660, 2.500000e-01
>  %663 = fadd float %662, %661
>  %664 = fcmp oge float %449, 0.000000e+00
>  %.245 = select i1 %664, float 1.000000e+00, float %663
>  %665 = fsub float 1.000000e+00, %224
>  %666 = fmul float %378, %224
>  %667 = fmul float %.245, %665
>  %668 = fadd float %666, %667
>  br label %ENDIF118
>
>ENDIF118:                                         ; preds = %IF, %IF119
>  %temp8.0 = phi float [ %668, %IF119 ], [ %378, %IF ]
>  %669 = fsub float %105, %84
>  %670 = fsub float %106, %85
>  %671 = fsub float %107, %86
>  %672 = fmul float %669, %669
>  %673 = fmul float %670, %670
>  %674 = fadd float %673, %672
>  %675 = fmul float %671, %671
>  %676 = fadd float %674, %675
>  %677 = fmul float %676, %47
>  %678 = fadd float %677, %46
>  %679 = call float @llvm.AMDIL.clamp.(float %678, float 0.000000e+00, float 1.000000e+00)
>  %680 = fsub float 1.000000e+00, %679
>  %681 = fmul float %temp8.0, %680
>  %682 = fadd float %679, %681
>  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: nounwind readnone
>declare <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.clamp.(float, float, float) #2
>
>; Function Attrs: nounwind readnone
>declare float @llvm.fabs.f32(float) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.sqrt.f32(float) #1
>
>; 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_mov_b32 m0, s10                                                            ; BEFC030A
>	s_load_dwordx8 s[16:23], s[6:7], 0x0                                         ; C0C80700
>	v_interp_p1_f32 v12, v0, 0, 0, [m0]                                          ; C8300000
>	v_interp_p2_f32 v12, [v12], v1, 0, 0, [m0]                                   ; C8310001
>	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 v5, v0, 0, 1, [m0]                                           ; C8140400
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	s_buffer_load_dword s9, s[12:15], 0x57                                       ; C2048D57
>	s_buffer_load_dword s8, s[12:15], 0x78                                       ; C2040D78
>	s_buffer_load_dword s0, s[12:15], 0x168                                      ; C2000CFF 00000168
>	v_interp_p2_f32 v5, [v5], v1, 0, 1, [m0]                                     ; C8150401
>	v_interp_p1_f32 v6, v0, 1, 1, [m0]                                           ; C8180500
>	v_interp_p2_f32 v6, [v6], v1, 1, 1, [m0]                                     ; C8190501
>	v_interp_p1_f32 v7, v0, 2, 1, [m0]                                           ; C81C0600
>	v_interp_p2_f32 v7, [v7], v1, 2, 1, [m0]                                     ; C81D0601
>	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
>	v_interp_p1_f32 v4, v0, 2, 2, [m0]                                           ; C8100A00
>	v_interp_p2_f32 v4, [v4], v1, 2, 2, [m0]                                     ; C8110A01
>	v_interp_p1_f32 v9, v0, 3, 2, [m0]                                           ; C8240B00
>	s_load_dwordx4 s[24:27], s[4:5], 0x0                                         ; C08C0500
>	v_interp_p2_f32 v9, [v9], v1, 3, 2, [m0]                                     ; C8250B01
>	v_interp_p1_f32 v8, v0, 0, 3, [m0]                                           ; C8200C00
>	v_interp_p2_f32 v8, [v8], v1, 0, 3, [m0]                                     ; C8210C01
>	v_interp_p1_f32 v10, v0, 1, 3, [m0]                                          ; C8280D00
>	v_interp_p2_f32 v10, [v10], v1, 1, 3, [m0]                                   ; C8290D01
>	v_interp_p1_f32 v11, v0, 2, 3, [m0]                                          ; C82C0E00
>	v_interp_p2_f32 v11, [v11], v1, 2, 3, [m0]                                   ; C82D0E01
>	v_interp_p1_f32 v0, v0, 3, 3, [m0]                                           ; C8000F00
>	v_interp_p2_f32 v0, [v0], v1, 3, 3, [m0]                                     ; C8010F01
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample v[12:15], 15, 0, 0, 0, 0, 0, 0, 0, v[12:13], s[16:23], s[24:27] ; F0800F00 00C40C0C
>	v_cmp_ne_i32_e64 s[0:1], 0, s0                                               ; D10A0000 00000080
>	v_mov_b32_e32 v16, 1.0                                                       ; 7E2002F2
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	s_and_saveexec_b64 s[10:11], s[0:1]                                          ; BE8A2400
>	s_xor_b64 s[10:11], exec, s[10:11]                                           ; 898A0A7E
>	s_cbranch_execz BB0_4                                                        ; BF880000
>	v_add_f32_e32 v16, 0, v8                                                     ; 06201080
>	s_buffer_load_dword s1, s[12:15], 0x10e                                      ; C2008CFF 0000010E
>	s_buffer_load_dword s0, s[12:15], 0x10f                                      ; C2000CFF 0000010F
>	s_buffer_load_dword s16, s[12:15], 0x110                                     ; C2080CFF 00000110
>	s_buffer_load_dword s2, s[12:15], 0x111                                      ; C2010CFF 00000111
>	s_buffer_load_dword s3, s[12:15], 0x114                                      ; C2018CFF 00000114
>	s_buffer_load_dword s17, s[12:15], 0x128                                     ; C2088CFF 00000128
>	s_buffer_load_dword s18, s[12:15], 0x129                                     ; C2090CFF 00000129
>	s_buffer_load_dword s20, s[12:15], 0x12a                                     ; C20A0CFF 0000012A
>	s_buffer_load_dword s21, s[12:15], 0x12b                                     ; C20A8CFF 0000012B
>	s_buffer_load_dword s22, s[12:15], 0x134                                     ; C20B0CFF 00000134
>	s_buffer_load_dword s23, s[12:15], 0x135                                     ; C20B8CFF 00000135
>	s_buffer_load_dword s24, s[12:15], 0x136                                     ; C20C0CFF 00000136
>	s_buffer_load_dword s25, s[12:15], 0x137                                     ; C20C8CFF 00000137
>	s_buffer_load_dword s26, s[12:15], 0x138                                     ; C20D0CFF 00000138
>	s_buffer_load_dword s27, s[12:15], 0x139                                     ; C20D8CFF 00000139
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mov_b32_e32 v1, s2                                                         ; 7E020202
>	s_buffer_load_dword s28, s[12:15], 0x13a                                     ; C20E0CFF 0000013A
>	s_buffer_load_dword s30, s[12:15], 0x13b                                     ; C20F0CFF 0000013B
>	s_buffer_load_dword s2, s[12:15], 0x154                                      ; C2010CFF 00000154
>	s_buffer_load_dword s50, s[12:15], 0x155                                     ; C2190CFF 00000155
>	s_buffer_load_dword s29, s[12:15], 0x156                                     ; C20E8CFF 00000156
>	s_buffer_load_dword s33, s[12:15], 0x15c                                     ; C2108CFF 0000015C
>	s_buffer_load_dword s35, s[12:15], 0x15d                                     ; C2118CFF 0000015D
>	s_buffer_load_dword s37, s[12:15], 0x15e                                     ; C2128CFF 0000015E
>	s_buffer_load_dword s38, s[12:15], 0x15f                                     ; C2130CFF 0000015F
>	s_buffer_load_dword s19, s[12:15], 0x164                                     ; C2098CFF 00000164
>	s_load_dwordx8 s[40:47], s[6:7], 0x8                                         ; C0D40708
>	s_buffer_load_dword s51, s[12:15], 0x157                                     ; C2198CFF 00000157
>	s_buffer_load_dword s31, s[12:15], 0x158                                     ; C20F8CFF 00000158
>	s_buffer_load_dword s32, s[12:15], 0x159                                     ; C2100CFF 00000159
>	s_buffer_load_dword s34, s[12:15], 0x15a                                     ; C2110CFF 0000015A
>	s_buffer_load_dword s36, s[12:15], 0x15b                                     ; C2120CFF 0000015B
>	v_add_f32_e32 v19, 0, v10                                                    ; 06261480
>	v_add_f32_e32 v18, 0, v11                                                    ; 06241680
>	v_mad_f32 v17, 0, v8, 1.0                                                    ; D2820011 03CA1080
>	v_mul_f32_e32 v20, s18, v19                                                  ; 10282612
>	v_mac_f32_e32 v20, s17, v16                                                  ; 3E282011
>	v_mac_f32_e32 v20, s20, v18                                                  ; 3E282414
>	v_mac_f32_e32 v20, s21, v17                                                  ; 3E282215
>	v_add_f32_e64 v21, 0, v20 clamp                                              ; D2060815 00022880
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mov_b32_e32 v22, s29                                                       ; 7E2C021D
>	s_buffer_load_dword s52, s[12:15], 0x11a                                     ; C21A0CFF 0000011A
>	s_buffer_load_dword s53, s[12:15], 0x11b                                     ; C21A8CFF 0000011B
>	s_buffer_load_dword s54, s[12:15], 0x11c                                     ; C21B0CFF 0000011C
>	s_buffer_load_dword s55, s[12:15], 0x11d                                     ; C21B8CFF 0000011D
>	s_buffer_load_dword s56, s[12:15], 0x11e                                     ; C21C0CFF 0000011E
>	s_buffer_load_dword s39, s[12:15], 0x115                                     ; C2138CFF 00000115
>	s_buffer_load_dword s57, s[12:15], 0x116                                     ; C21C8CFF 00000116
>	s_buffer_load_dword s58, s[12:15], 0x117                                     ; C21D0CFF 00000117
>	s_buffer_load_dword s59, s[12:15], 0x118                                     ; C21D8CFF 00000118
>	s_buffer_load_dword s60, s[12:15], 0x119                                     ; C21E0CFF 00000119
>	s_buffer_load_dword s29, s[12:15], 0x125                                     ; C20E8CFF 00000125
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v23, s39, v19                                                  ; 102E2627
>	v_mac_f32_e32 v23, s3, v16                                                   ; 3E2E2003
>	s_buffer_load_dword s39, s[12:15], 0x124                                     ; C2138CFF 00000124
>	s_buffer_load_dword s48, s[12:15], 0x126                                     ; C2180CFF 00000126
>	s_buffer_load_dword s49, s[12:15], 0x127                                     ; C2188CFF 00000127
>	v_mac_f32_e32 v23, s57, v18                                                  ; 3E2E2439
>	v_mac_f32_e32 v23, s58, v17                                                  ; 3E2E223A
>	v_mul_f32_e32 v24, s60, v19                                                  ; 1030263C
>	v_mac_f32_e32 v24, s59, v16                                                  ; 3E30203B
>	v_mac_f32_e32 v24, s52, v18                                                  ; 3E302434
>	v_mac_f32_e32 v24, s53, v17                                                  ; 3E302235
>	v_mov_b32_e32 v25, s51                                                       ; 7E320233
>	v_mul_f32_e32 v26, s29, v19                                                  ; 1034261D
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mac_f32_e32 v26, s39, v16                                                  ; 3E342027
>	v_mac_f32_e32 v26, s48, v18                                                  ; 3E342430
>	v_mac_f32_e32 v26, s49, v17                                                  ; 3E342231
>	v_add_f32_e64 v27, 0, v26 clamp                                              ; D206081B 00023480
>	v_subrev_f32_e32 v27, v26, v27                                               ; 0A36371A
>	v_subrev_f32_e32 v21, v20, v21                                               ; 0A2A2B14
>	v_add_f32_e32 v21, v21, v27                                                  ; 062A3715
>	v_mul_f32_e32 v27, s23, v19                                                  ; 10362617
>	v_mac_f32_e32 v27, s22, v16                                                  ; 3E362016
>	v_mac_f32_e32 v27, s24, v18                                                  ; 3E362418
>	v_mac_f32_e32 v27, s25, v17                                                  ; 3E362219
>	v_mul_f32_e32 v28, s27, v19                                                  ; 1038261B
>	v_mac_f32_e32 v28, s26, v16                                                  ; 3E38201A
>	v_mac_f32_e32 v28, s28, v18                                                  ; 3E38241C
>	v_mac_f32_e32 v28, s30, v17                                                  ; 3E38221E
>	v_mov_b32_e32 v29, 0x80000000                                                ; 7E3A02FF 80000000
>	v_cmp_le_f32_e64 vcc, |v21|, v29                                             ; D006016A 00023B15
>	v_cndmask_b32_e32 v21, v27, v26                                              ; 002A351B
>	v_cndmask_b32_e32 v20, v28, v20                                              ; 0028291C
>	v_cndmask_b32_e64 v26, 2.0, 1.0, vcc                                         ; D200001A 01A9E4F4
>	v_add_f32_e64 v27, 0, v23 clamp                                              ; D206081B 00022E80
>	v_add_f32_e64 v28, 0, v24 clamp                                              ; D206081C 00023080
>	v_subrev_f32_e32 v27, v23, v27                                               ; 0A363717
>	v_subrev_f32_e32 v28, v24, v28                                               ; 0A383918
>	v_add_f32_e32 v27, v28, v27                                                  ; 0636371C
>	v_cmp_le_f32_e64 vcc, |v27|, v29                                             ; D006016A 00023B1B
>	v_cndmask_b32_e32 v23, v21, v23                                              ; 002E2F15
>	v_cndmask_b32_e32 v24, v20, v24                                              ; 00303114
>	v_cndmask_b32_e64 v26, v26, 0, vcc                                           ; D200001A 01A9011A
>	v_mov_b32_e32 v20, s2                                                        ; 7E280202
>	v_mov_b32_e32 v21, s50                                                       ; 7E2A0232
>	v_cmp_le_f32_e64 vcc, |v26|, v29                                             ; D006016A 00023B1A
>	v_cndmask_b32_e32 v22, 0, v22                                                ; 002C2C80
>	v_cndmask_b32_e32 v25, 0, v25                                                ; 00323280
>	v_cndmask_b32_e32 v20, 0, v20                                                ; 00282880
>	v_cndmask_b32_e32 v21, 0, v21                                                ; 002A2A80
>	s_buffer_load_dword s2, s[12:15], 0x11f                                      ; C2010CFF 0000011F
>	v_add_f32_e32 v27, -1.0, v26                                                 ; 063634F3
>	v_mov_b32_e32 v28, s34                                                       ; 7E380222
>	v_cmp_le_f32_e64 vcc, |v27|, v29                                             ; D006016A 00023B1B
>	v_cndmask_b32_e32 v22, v22, v28                                              ; 002C3916
>	v_mov_b32_e32 v27, s36                                                       ; 7E360224
>	v_cndmask_b32_e32 v25, v25, v27                                              ; 00323719
>	v_mov_b32_e32 v27, s31                                                       ; 7E36021F
>	v_cndmask_b32_e32 v27, v20, v27                                              ; 00363714
>	v_mov_b32_e32 v20, s32                                                       ; 7E280220
>	v_cndmask_b32_e32 v28, v21, v20                                              ; 00382915
>	v_mul_f32_e32 v20, s55, v19                                                  ; 10282637
>	v_mac_f32_e32 v20, s54, v16                                                  ; 3E282036
>	v_mac_f32_e32 v20, s56, v18                                                  ; 3E282438
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mac_f32_e32 v20, s2, v17                                                   ; 3E282202
>	v_add_f32_e32 v21, -2.0, v26                                                 ; 062A34F5
>	v_mov_b32_e32 v30, s37                                                       ; 7E3C0225
>	v_cmp_le_f32_e64 vcc, |v21|, v29                                             ; D006016A 00023B15
>	v_cndmask_b32_e32 v29, v22, v30                                              ; 003A3D16
>	v_mov_b32_e32 v21, s38                                                       ; 7E2A0226
>	v_cndmask_b32_e32 v25, v25, v21                                              ; 00322B19
>	v_mov_b32_e32 v21, s33                                                       ; 7E2A0221
>	v_cndmask_b32_e32 v21, v27, v21                                              ; 002A2B1B
>	v_mov_b32_e32 v22, s35                                                       ; 7E2C0223
>	v_cndmask_b32_e32 v22, v28, v22                                              ; 002C2D1C
>	v_add_f32_e64 v27, 0, v23 clamp                                              ; D206081B 00022E80
>	v_add_f32_e64 v28, 0, v24 clamp                                              ; D206081C 00023080
>	v_mac_f32_e32 v21, v29, v27                                                  ; 3E2A371D
>	v_mac_f32_e32 v22, v25, v28                                                  ; 3E2C3919
>	v_mov_b32_e32 v25, 0x3a000000                                                ; 7E3202FF 3A000000
>	v_add_f32_e32 v28, v25, v21                                                  ; 06382B19
>	v_add_f32_e32 v29, v25, v22                                                  ; 063A2D19
>	v_add_f32_e32 v27, 0, v20                                                    ; 06362880
>	s_mov_b32 s50, 0                                                             ; BEB20380
>	s_load_dwordx4 s[52:55], s[4:5], 0x4                                         ; C09A0504
>	v_mov_b32_e32 v30, s50                                                       ; 7E3C0232
>	v_mov_b32_e32 v25, 0xba000000                                                ; 7E3202FF BA000000
>	v_add_f32_e32 v31, v25, v21                                                  ; 063E2B19
>	v_mov_b32_e32 v32, v27                                                       ; 7E40031B
>	v_mov_b32_e32 v33, v28                                                       ; 7E42031C
>	v_mov_b32_e32 v34, v29                                                       ; 7E44031D
>	v_mov_b32_e32 v35, v30                                                       ; 7E46031E
>	v_mov_b32_e32 v33, v31                                                       ; 7E42031F
>	v_add_f32_e32 v23, -0.5, v23                                                 ; 062E2EF1
>	v_add_f32_e32 v24, -0.5, v24                                                 ; 063030F1
>	v_add_f32_e32 v25, v25, v22                                                  ; 06322D19
>	v_mov_b32_e32 v36, v27                                                       ; 7E48031B
>	v_mov_b32_e32 v37, v28                                                       ; 7E4A031C
>	v_mov_b32_e32 v38, v29                                                       ; 7E4C031D
>	v_mov_b32_e32 v39, v30                                                       ; 7E4E031E
>	v_sub_f32_e64 v23, |v23|, s1                                                 ; D2080117 00000317
>	v_mov_b32_e32 v34, v29                                                       ; 7E44031D
>	v_mov_b32_e32 v38, v25                                                       ; 7E4C0319
>	v_sub_f32_e64 v24, |v24|, s1                                                 ; D2080118 00000318
>	v_mov_b32_e32 v35, s50                                                       ; 7E460232
>	v_mov_b32_e32 v39, s50                                                       ; 7E4E0232
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample_c_l v31, 1, 0, 0, 0, 0, 0, 0, 0, v[27:30], s[40:47], s[52:55]   ; F0B00100 01AA1F1B
>	image_sample_c_l v40, 1, 0, 0, 0, 0, 0, 0, 0, v[32:35], s[40:47], s[52:55]   ; F0B00100 01AA2820
>	v_mov_b32_e32 v34, v25                                                       ; 7E440319
>	image_sample_c_l v36, 1, 0, 0, 0, 0, 0, 0, 0, v[36:39], s[40:47], s[52:55]   ; F0B00100 01AA2424
>	v_mov_b32_e32 v35, s50                                                       ; 7E460232
>	image_sample_c_l v37, 1, 0, 0, 0, 0, 0, 0, 0, v[32:35], s[40:47], s[52:55]   ; F0B00100 01AA2520
>	v_add_f32_e32 v34, 0, v22                                                    ; 06442C80
>	v_mov_b32_e32 v41, v27                                                       ; 7E52031B
>	v_mov_b32_e32 v42, v28                                                       ; 7E54031C
>	v_mov_b32_e32 v43, v29                                                       ; 7E56031D
>	v_mov_b32_e32 v44, v30                                                       ; 7E58031E
>	v_mul_f32_e32 v23, s0, v23                                                   ; 102E2E00
>	v_mov_b32_e32 v43, v34                                                       ; 7E560322
>	v_add_f32_e32 v28, 0, v21                                                    ; 06382A80
>	v_mul_f32_e32 v24, s0, v24                                                   ; 10303000
>	v_mov_b32_e32 v44, s50                                                       ; 7E580232
>	v_mov_b32_e32 v45, v27                                                       ; 7E5A031B
>	v_mov_b32_e32 v46, v28                                                       ; 7E5C031C
>	v_mov_b32_e32 v47, v29                                                       ; 7E5E031D
>	v_mov_b32_e32 v48, v30                                                       ; 7E60031E
>	image_sample_c_l v38, 1, 0, 0, 0, 0, 0, 0, 0, v[41:44], s[40:47], s[52:55]   ; F0B00100 01AA2629
>	v_mov_b32_e32 v47, v25                                                       ; 7E5E0319
>	v_mov_b32_e32 v35, s50                                                       ; 7E460232
>	image_sample_c_l v25, 1, 0, 0, 0, 0, 0, 0, 0, v[32:35], s[40:47], s[52:55]   ; F0B00100 01AA1920
>	v_mov_b32_e32 v48, s50                                                       ; 7E600232
>	image_sample_c_l v32, 1, 0, 0, 0, 0, 0, 0, 0, v[45:48], s[40:47], s[52:55]   ; F0B00100 01AA202D
>	v_mov_b32_e32 v30, s50                                                       ; 7E3C0232
>	image_sample_c_l v27, 1, 0, 0, 0, 0, 0, 0, 0, v[27:30], s[40:47], s[52:55]   ; F0B00100 01AA1B1B
>	v_mov_b32_e32 v28, 0x3d800000                                                ; 7E3802FF 3D800000
>	s_waitcnt vmcnt(6)                                                           ; BF8C0776
>	v_mul_f32_e32 v29, v28, v40                                                  ; 103A511C
>	v_mac_f32_e32 v29, v28, v31                                                  ; 3E3A3F1C
>	s_waitcnt vmcnt(5)                                                           ; BF8C0775
>	v_mac_f32_e32 v29, v28, v36                                                  ; 3E3A491C
>	s_waitcnt vmcnt(4)                                                           ; BF8C0774
>	v_mac_f32_e32 v29, v28, v37                                                  ; 3E3A4B1C
>	v_mov_b32_e32 v28, 0x3e000000                                                ; 7E3802FF 3E000000
>	s_waitcnt vmcnt(2)                                                           ; BF8C0772
>	v_mul_f32_e32 v30, v28, v25                                                  ; 103C331C
>	v_mac_f32_e32 v30, v28, v38                                                  ; 3E3C4D1C
>	s_waitcnt vmcnt(1)                                                           ; BF8C0771
>	v_mac_f32_e32 v30, v28, v32                                                  ; 3E3C411C
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mac_f32_e32 v30, v28, v27                                                  ; 3E3C371C
>	s_buffer_load_dword s56, s[12:15], 0x165                                     ; C21C0CFF 00000165
>	s_buffer_load_dword s51, s[12:15], 0x166                                     ; C2198CFF 00000166
>	v_add_f32_e64 v23, 0, v23 clamp                                              ; D2060817 00022E80
>	v_add_f32_e64 v24, 0, v24 clamp                                              ; D2060818 00023080
>	v_sub_f32_e32 v23, 1.0, v23                                                  ; 082E2EF2
>	v_mad_f32 v25, -v24, v23, v23                                                ; D2820019 245E2F18
>	v_mov_b32_e32 v23, 0                                                         ; 7E2E0280
>	v_add_f32_e32 v24, v29, v30                                                  ; 06303D1D
>	image_sample_c_l v21, 1, 0, 0, 0, 0, 0, 0, 0, v[20:23], s[40:47], s[52:55]   ; F0B00100 01AA1514
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_madmk_f32_e32 v24, v21, v24, 0x3e800000                                    ; 40303115 3E800000
>	v_cmp_gt_f32_e32 vcc, 1.0, v25                                               ; 7C0832F2
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	s_and_saveexec_b64 s[58:59], vcc                                             ; BEBA246A
>	s_xor_b64 s[58:59], exec, s[58:59]                                           ; 89BA3A7E
>	s_cbranch_execz BB0_5                                                        ; BF880000
>	v_mov_b32_e32 v21, s39                                                       ; 7E2A0227
>	v_mov_b32_e32 v22, s29                                                       ; 7E2C021D
>	v_mov_b32_e32 v23, s48                                                       ; 7E2E0230
>	v_mov_b32_e32 v27, s49                                                       ; 7E360231
>	v_mov_b32_e32 v28, s17                                                       ; 7E380211
>	v_mov_b32_e32 v29, s18                                                       ; 7E3A0212
>	v_mov_b32_e32 v30, s20                                                       ; 7E3C0214
>	v_mov_b32_e32 v31, s21                                                       ; 7E3E0215
>	v_mov_b32_e32 v32, s22                                                       ; 7E400216
>	v_mov_b32_e32 v33, s23                                                       ; 7E420217
>	v_mov_b32_e32 v34, s24                                                       ; 7E440218
>	v_mov_b32_e32 v35, s25                                                       ; 7E460219
>	v_mov_b32_e32 v36, s26                                                       ; 7E48021A
>	s_buffer_load_dword s0, s[12:15], 0x144                                      ; C2000CFF 00000144
>	s_buffer_load_dword s2, s[12:15], 0x145                                      ; C2010CFF 00000145
>	s_buffer_load_dword s3, s[12:15], 0x146                                      ; C2018CFF 00000146
>	s_buffer_load_dword s57, s[12:15], 0x147                                     ; C21C8CFF 00000147
>	s_buffer_load_dword s60, s[12:15], 0x148                                     ; C21E0CFF 00000148
>	s_buffer_load_dword s61, s[12:15], 0x149                                     ; C21E8CFF 00000149
>	s_buffer_load_dword s62, s[12:15], 0x14a                                     ; C21F0CFF 0000014A
>	s_buffer_load_dword s63, s[12:15], 0x14b                                     ; C21F8CFF 0000014B
>	s_buffer_load_dword s64, s[12:15], 0x160                                     ; C2200CFF 00000160
>	s_buffer_load_dword s65, s[12:15], 0x161                                     ; C2208CFF 00000161
>	s_buffer_load_dword s66, s[12:15], 0x162                                     ; C2210CFF 00000162
>	s_buffer_load_dword s67, s[12:15], 0x163                                     ; C2218CFF 00000163
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mov_b32_e32 v37, s0                                                        ; 7E4A0200
>	v_add_f32_e32 v38, 0, v26                                                    ; 064C3480
>	v_mov_b32_e32 v39, 0x80000000                                                ; 7E4E02FF 80000000
>	v_cmp_le_f32_e64 vcc, |v38|, v39                                             ; D006016A 00024F26
>	v_add_f32_e32 v38, -1.0, v26                                                 ; 064C34F3
>	v_cmp_le_f32_e64 s[0:1], |v38|, v39                                          ; D0060100 00024F26
>	v_cndmask_b32_e32 v21, 0, v21                                                ; 002A2A80
>	v_cndmask_b32_e64 v21, v21, v32, s[0:1]                                      ; D2000015 00024115
>	v_cndmask_b32_e32 v22, 0, v22                                                ; 002C2C80
>	v_cndmask_b32_e64 v22, v22, v33, s[0:1]                                      ; D2000016 00024316
>	v_cndmask_b32_e32 v23, 0, v23                                                ; 002E2E80
>	v_cndmask_b32_e64 v23, v23, v34, s[0:1]                                      ; D2000017 00024517
>	v_mov_b32_e32 v32, s2                                                        ; 7E400202
>	v_cndmask_b32_e32 v27, 0, v27                                                ; 00363680
>	v_cndmask_b32_e64 v27, v27, v35, s[0:1]                                      ; D200001B 0002471B
>	v_cndmask_b32_e32 v28, 0, v28                                                ; 00383880
>	v_cndmask_b32_e64 v28, v28, v36, s[0:1]                                      ; D200001C 0002491C
>	v_mov_b32_e32 v33, s27                                                       ; 7E42021B
>	v_cndmask_b32_e32 v29, 0, v29                                                ; 003A3A80
>	v_cndmask_b32_e64 v29, v29, v33, s[0:1]                                      ; D200001D 0002431D
>	v_mov_b32_e32 v33, s28                                                       ; 7E42021C
>	v_cndmask_b32_e32 v30, 0, v30                                                ; 003C3C80
>	v_cndmask_b32_e64 v30, v30, v33, s[0:1]                                      ; D200001E 0002431E
>	v_mov_b32_e32 v33, s30                                                       ; 7E42021E
>	v_add_f32_e32 v26, -2.0, v26                                                 ; 063434F5
>	v_cndmask_b32_e32 v31, 0, v31                                                ; 003E3E80
>	v_cndmask_b32_e64 v31, v31, v33, s[0:1]                                      ; D200001F 0002431F
>	v_mov_b32_e32 v33, s3                                                        ; 7E420203
>	v_mov_b32_e32 v34, s57                                                       ; 7E440239
>	v_cmp_le_f32_e64 s[2:3], |v26|, v39                                          ; D0060102 00024F1A
>	v_mov_b32_e32 v35, s60                                                       ; 7E46023C
>	v_cndmask_b32_e64 v21, v21, v37, s[2:3]                                      ; D2000015 000A4B15
>	v_mov_b32_e32 v36, s61                                                       ; 7E48023D
>	v_cndmask_b32_e64 v22, v22, v32, s[2:3]                                      ; D2000016 000A4116
>	v_mov_b32_e32 v32, s31                                                       ; 7E40021F
>	v_cndmask_b32_e64 v23, v23, v33, s[2:3]                                      ; D2000017 000A4317
>	v_mov_b32_e32 v33, s32                                                       ; 7E420220
>	v_cndmask_b32_e64 v27, v27, v34, s[2:3]                                      ; D200001B 000A451B
>	v_mov_b32_e32 v34, s34                                                       ; 7E440222
>	v_cndmask_b32_e64 v28, v28, v35, s[2:3]                                      ; D200001C 000A471C
>	v_mov_b32_e32 v35, s36                                                       ; 7E460224
>	v_cndmask_b32_e64 v29, v29, v36, s[2:3]                                      ; D200001D 000A491D
>	v_mov_b32_e32 v36, s62                                                       ; 7E48023E
>	v_cndmask_b32_e64 v30, v30, v36, s[2:3]                                      ; D200001E 000A491E
>	v_mov_b32_e32 v36, s63                                                       ; 7E48023F
>	v_cndmask_b32_e64 v31, v31, v36, s[2:3]                                      ; D200001F 000A491F
>	v_mov_b32_e32 v36, s33                                                       ; 7E480221
>	v_mul_f32_e32 v22, v22, v19                                                  ; 102C2716
>	v_mac_f32_e32 v22, v21, v16                                                  ; 3E2C2115
>	v_mov_b32_e32 v21, s35                                                       ; 7E2A0223
>	v_mac_f32_e32 v22, v23, v18                                                  ; 3E2C2517
>	v_mov_b32_e32 v23, s37                                                       ; 7E2E0225
>	v_mac_f32_e32 v22, v27, v17                                                  ; 3E2C231B
>	v_mov_b32_e32 v27, s38                                                       ; 7E360226
>	v_add_f32_e64 v37, 0, v22 clamp                                              ; D2060825 00022C80
>	v_mul_f32_e32 v19, v29, v19                                                  ; 1026271D
>	v_mac_f32_e32 v19, v28, v16                                                  ; 3E26211C
>	v_mac_f32_e32 v19, v30, v18                                                  ; 3E26251E
>	v_mac_f32_e32 v19, v31, v17                                                  ; 3E26231F
>	v_add_f32_e64 v16, 0, v19 clamp                                              ; D2060810 00022680
>	v_cndmask_b32_e32 v17, 0, v34                                                ; 00224480
>	v_cndmask_b32_e32 v18, 0, v35                                                ; 00244680
>	v_cndmask_b32_e32 v19, 0, v32                                                ; 00264080
>	v_cndmask_b32_e32 v22, 0, v33                                                ; 002C4280
>	v_cndmask_b32_e64 v17, v17, v23, s[0:1]                                      ; D2000011 00022F11
>	v_cndmask_b32_e64 v18, v18, v27, s[0:1]                                      ; D2000012 00023712
>	v_cndmask_b32_e64 v19, v19, v36, s[0:1]                                      ; D2000013 00024913
>	v_cndmask_b32_e64 v22, v22, v21, s[0:1]                                      ; D2000016 00022B16
>	v_mov_b32_e32 v21, s66                                                       ; 7E2A0242
>	v_cndmask_b32_e64 v17, v17, v21, s[2:3]                                      ; D2000011 000A2B11
>	v_mov_b32_e32 v21, s67                                                       ; 7E2A0243
>	v_cndmask_b32_e64 v18, v18, v21, s[2:3]                                      ; D2000012 000A2B12
>	v_mov_b32_e32 v21, s64                                                       ; 7E2A0240
>	v_cndmask_b32_e64 v21, v19, v21, s[2:3]                                      ; D2000015 000A2B13
>	v_mov_b32_e32 v19, s65                                                       ; 7E260241
>	v_cndmask_b32_e64 v22, v22, v19, s[2:3]                                      ; D2000016 000A2716
>	v_mac_f32_e32 v21, v17, v37                                                  ; 3E2A4B11
>	v_mac_f32_e32 v22, v18, v16                                                  ; 3E2C2112
>	v_mov_b32_e32 v16, 0x3a000000                                                ; 7E2002FF 3A000000
>	v_add_f32_e32 v28, v16, v21                                                  ; 06382B10
>	v_add_f32_e32 v29, v16, v22                                                  ; 063A2D10
>	v_add_f32_e32 v27, 0, v20                                                    ; 06362880
>	v_mov_b32_e32 v30, s50                                                       ; 7E3C0232
>	v_mov_b32_e32 v16, 0xba000000                                                ; 7E2002FF BA000000
>	v_add_f32_e32 v17, v16, v21                                                  ; 06222B10
>	v_mov_b32_e32 v31, v27                                                       ; 7E3E031B
>	v_mov_b32_e32 v32, v28                                                       ; 7E40031C
>	v_mov_b32_e32 v33, v29                                                       ; 7E42031D
>	v_mov_b32_e32 v34, v30                                                       ; 7E44031E
>	v_mov_b32_e32 v32, v17                                                       ; 7E400311
>	v_add_f32_e32 v16, v16, v22                                                  ; 06202D10
>	v_mov_b32_e32 v33, v29                                                       ; 7E42031D
>	v_mov_b32_e32 v35, v27                                                       ; 7E46031B
>	v_mov_b32_e32 v36, v28                                                       ; 7E48031C
>	v_mov_b32_e32 v37, v29                                                       ; 7E4A031D
>	v_mov_b32_e32 v38, v30                                                       ; 7E4C031E
>	image_sample_c_l v17, 1, 0, 0, 0, 0, 0, 0, 0, v[27:30], s[40:47], s[52:55]   ; F0B00100 01AA111B
>	v_mov_b32_e32 v34, s50                                                       ; 7E440232
>	v_mov_b32_e32 v37, v16                                                       ; 7E4A0310
>	image_sample_c_l v18, 1, 0, 0, 0, 0, 0, 0, 0, v[31:34], s[40:47], s[52:55]   ; F0B00100 01AA121F
>	v_mov_b32_e32 v38, s50                                                       ; 7E4C0232
>	v_mov_b32_e32 v33, v16                                                       ; 7E420310
>	image_sample_c_l v19, 1, 0, 0, 0, 0, 0, 0, 0, v[35:38], s[40:47], s[52:55]   ; F0B00100 01AA1323
>	v_mov_b32_e32 v34, s50                                                       ; 7E440232
>	image_sample_c_l v23, 1, 0, 0, 0, 0, 0, 0, 0, v[31:34], s[40:47], s[52:55]   ; F0B00100 01AA171F
>	v_mov_b32_e32 v30, 0x3d800000                                                ; 7E3C02FF 3D800000
>	s_waitcnt vmcnt(2)                                                           ; BF8C0772
>	v_mul_f32_e32 v18, v30, v18                                                  ; 1024251E
>	v_add_f32_e32 v33, 0, v22                                                    ; 06422C80
>	v_mov_b32_e32 v34, v27                                                       ; 7E44031B
>	v_mov_b32_e32 v35, v28                                                       ; 7E46031C
>	v_mov_b32_e32 v36, v29                                                       ; 7E48031D
>	v_mov_b32_e32 v37, v30                                                       ; 7E4A031E
>	v_mac_f32_e32 v18, v30, v17                                                  ; 3E24231E
>	v_mov_b32_e32 v36, v33                                                       ; 7E480321
>	s_waitcnt vmcnt(1)                                                           ; BF8C0771
>	v_mac_f32_e32 v18, v30, v19                                                  ; 3E24271E
>	v_mov_b32_e32 v37, s50                                                       ; 7E4A0232
>	v_add_f32_e32 v28, 0, v21                                                    ; 06382A80
>	image_sample_c_l v17, 1, 0, 0, 0, 0, 0, 0, 0, v[34:37], s[40:47], s[52:55]   ; F0B00100 01AA1122
>	v_mov_b32_e32 v35, v27                                                       ; 7E46031B
>	v_mov_b32_e32 v36, v28                                                       ; 7E48031C
>	v_mov_b32_e32 v37, v29                                                       ; 7E4A031D
>	v_mov_b32_e32 v38, v30                                                       ; 7E4C031E
>	s_waitcnt vmcnt(1)                                                           ; BF8C0771
>	v_mac_f32_e32 v18, v30, v23                                                  ; 3E242F1E
>	v_mov_b32_e32 v34, s50                                                       ; 7E440232
>	v_mov_b32_e32 v37, v16                                                       ; 7E4A0310
>	image_sample_c_l v16, 1, 0, 0, 0, 0, 0, 0, 0, v[31:34], s[40:47], s[52:55]   ; F0B00100 01AA101F
>	v_mov_b32_e32 v38, s50                                                       ; 7E4C0232
>	image_sample_c_l v19, 1, 0, 0, 0, 0, 0, 0, 0, v[35:38], s[40:47], s[52:55]   ; F0B00100 01AA1323
>	v_mov_b32_e32 v30, s50                                                       ; 7E3C0232
>	image_sample_c_l v23, 1, 0, 0, 0, 0, 0, 0, 0, v[27:30], s[40:47], s[52:55]   ; F0B00100 01AA171B
>	v_mov_b32_e32 v27, 0x3e000000                                                ; 7E3602FF 3E000000
>	s_waitcnt vmcnt(2)                                                           ; BF8C0772
>	v_mul_f32_e32 v16, v27, v16                                                  ; 1020211B
>	v_mac_f32_e32 v16, v27, v17                                                  ; 3E20231B
>	s_waitcnt vmcnt(1)                                                           ; BF8C0771
>	v_mac_f32_e32 v16, v27, v19                                                  ; 3E20271B
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mac_f32_e32 v16, v27, v23                                                  ; 3E202F1B
>	v_mov_b32_e32 v23, 0                                                         ; 7E2E0280
>	image_sample_c_l v17, 1, 0, 0, 0, 0, 0, 0, 0, v[20:23], s[40:47], s[52:55]   ; F0B00100 01AA1114
>	v_add_f32_e32 v16, v18, v16                                                  ; 06202112
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_madmk_f32_e32 v16, v17, v16, 0x3e800000                                    ; 40202111 3E800000
>	v_cmp_le_f32_e32 vcc, 0, v26                                                 ; 7C063480
>	v_cndmask_b32_e64 v16, v16, 1.0, vcc                                         ; D2000010 01A9E510
>	v_mad_f32 v16, -v25, v16, v16                                                ; D2820010 24422119
>	v_mac_f32_e32 v16, v25, v24                                                  ; 3E203119
>	v_mov_b32_e32 v24, v16                                                       ; 7E300310
>	s_or_b64 exec, exec, s[58:59]                                                ; 88FE3A7E
>	v_subrev_f32_e32 v16, s19, v8                                                ; 0A201013
>	v_subrev_f32_e32 v17, s56, v10                                               ; 0A221438
>	v_subrev_f32_e32 v18, s51, v11                                               ; 0A241633
>	v_mul_f32_e32 v16, v16, v16                                                  ; 10202110
>	v_mac_f32_e32 v16, v17, v17                                                  ; 3E202311
>	v_mac_f32_e32 v16, v18, v18                                                  ; 3E202512
>	v_mad_f32 v1, v1, v16, s16                                                   ; D2820001 00422101
>	v_add_f32_e64 v16, 0, v1 clamp                                               ; D2060810 00020280
>	v_sub_f32_e32 v1, 1.0, v16                                                   ; 080220F2
>	v_mac_f32_e32 v16, v1, v24                                                   ; 3E203101
>	s_or_b64 exec, exec, s[10:11]                                                ; 88FE0A7E
>	s_buffer_load_dword s20, s[12:15], 0x4                                       ; C20A0D04
>	s_buffer_load_dword s21, s[12:15], 0x5                                       ; C20A8D05
>	s_buffer_load_dword s22, s[12:15], 0x6                                       ; C20B0D06
>	s_buffer_load_dword s18, s[12:15], 0x7                                       ; C2090D07
>	s_buffer_load_dword s23, s[12:15], 0x30                                      ; C20B8D30
>	s_buffer_load_dword s0, s[12:15], 0x31                                       ; C2000D31
>	s_buffer_load_dword s19, s[12:15], 0x33                                      ; C2098D33
>	s_buffer_load_dword s7, s[12:15], 0x50                                       ; C2038D50
>	s_buffer_load_dword s16, s[12:15], 0x51                                      ; C2080D51
>	s_buffer_load_dword s17, s[12:15], 0x52                                      ; C2088D52
>	s_buffer_load_dword s24, s[12:15], 0x53                                      ; C20C0D53
>	s_buffer_load_dword s5, s[12:15], 0x54                                       ; C2028D54
>	s_buffer_load_dword s4, s[12:15], 0x56                                       ; C2020D56
>	s_buffer_load_dword s1, s[12:15], 0x74                                       ; C2008D74
>	s_buffer_load_dword s2, s[12:15], 0x75                                       ; C2010D75
>	s_buffer_load_dword s3, s[12:15], 0x76                                       ; C2018D76
>	s_buffer_load_dword s6, s[12:15], 0x77                                       ; C2030D77
>	v_mov_b32_e32 v17, s9                                                        ; 7E220209
>	v_mov_b32_e32 v1, s8                                                         ; 7E020208
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mac_f32_e32 v2, v16, v5                                                    ; 3E040B10
>	v_mac_f32_e32 v3, v16, v6                                                    ; 3E060D10
>	v_mac_f32_e32 v4, v16, v7                                                    ; 3E080F10
>	v_mad_f32 v5, v15, s24, -s24                                                 ; D2820005 8060310F
>	v_add_f32_e32 v6, s23, v15                                                   ; 060C1E17
>	v_add_f32_e64 v6, 0, v6 clamp                                                ; D2060806 00020C80
>	v_mad_f32 v7, s20, v6, -v6                                                   ; D2820007 841A0C14
>	v_mad_f32 v15, s21, v6, -v6                                                  ; D282000F 841A0C15
>	v_mad_f32 v6, s22, v6, -v6                                                   ; D2820006 841A0C16
>	v_mac_f32_e32 v2, v2, v7                                                     ; 3E040F02
>	v_mac_f32_e32 v3, v3, v15                                                    ; 3E061F03
>	v_mac_f32_e32 v4, v4, v6                                                     ; 3E080D04
>	v_mad_f32 v5, v5, s18, s18                                                   ; D2820005 00482505
>	v_mad_f32 v6, v5, v9, -v5                                                    ; D2820006 84161305
>	v_mac_f32_e32 v5, s19, v6                                                    ; 3E0A0C13
>	v_mul_f32_e32 v2, v2, v12                                                    ; 10041902
>	v_mul_f32_e32 v3, v3, v13                                                    ; 10061B03
>	v_mul_f32_e32 v4, v4, v14                                                    ; 10081D04
>	v_sub_f32_e32 v6, s7, v8                                                     ; 080C1007
>	v_sub_f32_e32 v7, s16, v10                                                   ; 080E1410
>	v_sub_f32_e32 v8, s17, v11                                                   ; 08101611
>	v_mul_f32_e32 v6, v6, v6                                                     ; 100C0D06
>	v_mac_f32_e32 v6, v7, v7                                                     ; 3E0C0F07
>	v_mac_f32_e32 v6, v8, v8                                                     ; 3E0C1108
>	v_sqrt_f32_e32 v6, v6                                                        ; 7E0C6706
>	v_mad_f32 v6, v17, v6, s5                                                    ; D2820006 00160D11
>	v_add_f32_e64 v6, 0, v6 clamp                                                ; D2060806 00020C80
>	v_min_f32_e32 v6, s4, v6                                                     ; 1E0C0C04
>	v_mul_f32_e32 v7, s8, v2                                                     ; 100E0408
>	v_mul_f32_e32 v8, s8, v3                                                     ; 10100608
>	v_mul_f32_e32 v9, s8, v4                                                     ; 10120808
>	v_mul_f32_e32 v0, s6, v0                                                     ; 10000006
>	v_mov_b32_e32 v10, 0x80000000                                                ; 7E1402FF 80000000
>	v_cmp_le_f32_e64 vcc, |s0|, v10                                              ; D006016A 00021400
>	v_cndmask_b32_e32 v0, v0, v5                                                 ; 00000B00
>	v_mul_f32_e32 v5, v6, v6                                                     ; 100A0D06
>	v_mad_f32 v2, -v2, v1, s1                                                    ; D2820002 20060302
>	v_mad_f32 v3, -v3, v1, s2                                                    ; D2820003 200A0303
>	v_mad_f32 v1, -v4, v1, s3                                                    ; D2820001 200E0304
>	v_mac_f32_e32 v7, v2, v5                                                     ; 3E0E0B02
>	v_mac_f32_e32 v8, v3, v5                                                     ; 3E100B03
>	v_mac_f32_e32 v9, v1, v5                                                     ; 3E120B01
>	v_cvt_pkrtz_f16_f32_e32 v1, v7, v8                                           ; 5E021107
>	v_cvt_pkrtz_f16_f32_e32 v0, v9, v0                                           ; 5E000109
>	exp 15, 0, 1, 1, 1, v1, v0, v1, v0                                           ; F8001C0F 00010001
>	s_endpgm                                                                     ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 72
>VGPRS: 52
>Code Size: 2724 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
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL IN[2]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[0]
>DCL OUT[2], GENERIC[1]
>DCL OUT[3], GENERIC[2]
>DCL OUT[4], GENERIC[3]
>DCL CONST[0..95]
>DCL TEMP[0..2], LOCAL
>DCL ADDR[0]
>IMM[0] FLT32 {    0.0000,     1.0000,  1020.0400,     2.1000}
>IMM[1] FLT32 {    3.1000,     0.1000,     1.1000,     0.0000}
>  0: MOV TEMP[0].zw, IMM[0].yyxy
>  1: MAD TEMP[1].x, IMM[0].zzzz, IN[1].zzzz, IMM[0].wwww
>  2: F2I TEMP[1].x, TEMP[1].xxxx
>  3: UARL ADDR[0].x, TEMP[1].xxxx
>  4: UARL ADDR[0].x, TEMP[1].xxxx
>  5: DP4 TEMP[0].x, IN[2], CONST[ADDR[0].x]
>  6: MAD TEMP[1].x, IMM[0].zzzz, IN[1].zzzz, IMM[1].xxxx
>  7: F2I TEMP[1].x, TEMP[1].xxxx
>  8: UARL ADDR[0].x, TEMP[1].xxxx
>  9: DP4 TEMP[1].x, IN[2], CONST[ADDR[0].x]
> 10: MOV TEMP[0].y, TEMP[1].xxxx
> 11: MAD TEMP[1].x, IMM[0].zzzz, IN[1].zzzz, IMM[1].yyyy
> 12: F2I TEMP[1].x, TEMP[1].xxxx
> 13: UARL ADDR[0].x, TEMP[1].xxxx
> 14: MOV TEMP[1], CONST[ADDR[0].x]
> 15: MAD TEMP[2].x, IMM[0].zzzz, IN[1].zzzz, IMM[1].zzzz
> 16: F2I TEMP[2].x, TEMP[2].xxxx
> 17: UARL ADDR[0].x, TEMP[2].xxxx
> 18: MOV TEMP[2], CONST[ADDR[0].x]
> 19: MOV OUT[2], IN[1]
> 20: MOV OUT[3], TEMP[1]
> 21: MOV OUT[1], IN[0]
> 22: MOV OUT[4], TEMP[2]
> 23: MOV OUT[0], TEMP[0]
> 24: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %15 = load <16 x i8>, <16 x i8> addrspace(2)* %14, align 16, !tbaa !0
>  %16 = add i32 %5, %8
>  %17 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %15, i32 0, i32 %16)
>  %18 = extractelement <4 x float> %17, i32 0
>  %19 = extractelement <4 x float> %17, i32 1
>  %20 = extractelement <4 x float> %17, i32 2
>  %21 = extractelement <4 x float> %17, i32 3
>  %22 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 1
>  %23 = load <16 x i8>, <16 x i8> addrspace(2)* %22, align 16, !tbaa !0
>  %24 = add i32 %5, %8
>  %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 = extractelement <4 x float> %25, i32 3
>  %30 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 2
>  %31 = load <16 x i8>, <16 x i8> addrspace(2)* %30, align 16, !tbaa !0
>  %32 = add i32 %5, %8
>  %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 = extractelement <4 x float> %33, i32 2
>  %37 = extractelement <4 x float> %33, i32 3
>  %38 = fmul float %28, 0x408FE051E0000000
>  %39 = fadd float %38, 0x4000CCCCC0000000
>  %40 = fptosi float %39 to i32
>  %41 = shl i32 %40, 4
>  %42 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %41)
>  %43 = shl i32 %40, 4
>  %44 = or i32 %43, 4
>  %45 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %44)
>  %46 = shl i32 %40, 4
>  %47 = or i32 %46, 8
>  %48 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %47)
>  %49 = shl i32 %40, 4
>  %50 = or i32 %49, 12
>  %51 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %50)
>  %52 = fmul float %34, %42
>  %53 = fmul float %35, %45
>  %54 = fadd float %52, %53
>  %55 = fmul float %36, %48
>  %56 = fadd float %54, %55
>  %57 = fmul float %37, %51
>  %58 = fadd float %56, %57
>  %59 = fmul float %28, 0x408FE051E0000000
>  %60 = fadd float %59, 0x4008CCCCC0000000
>  %61 = fptosi float %60 to i32
>  %62 = shl i32 %61, 4
>  %63 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %62)
>  %64 = shl i32 %61, 4
>  %65 = or i32 %64, 4
>  %66 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %65)
>  %67 = shl i32 %61, 4
>  %68 = or i32 %67, 8
>  %69 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %68)
>  %70 = shl i32 %61, 4
>  %71 = or i32 %70, 12
>  %72 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %71)
>  %73 = fmul float %34, %63
>  %74 = fmul float %35, %66
>  %75 = fadd float %73, %74
>  %76 = fmul float %36, %69
>  %77 = fadd float %75, %76
>  %78 = fmul float %37, %72
>  %79 = fadd float %77, %78
>  %80 = fmul float %28, 0x408FE051E0000000
>  %81 = fadd float %80, 0x3FB99999A0000000
>  %82 = fptosi float %81 to i32
>  %83 = shl i32 %82, 4
>  %84 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %83)
>  %85 = shl i32 %82, 4
>  %86 = or i32 %85, 4
>  %87 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %86)
>  %88 = shl i32 %82, 4
>  %89 = or i32 %88, 8
>  %90 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %89)
>  %91 = shl i32 %82, 4
>  %92 = or i32 %91, 12
>  %93 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %92)
>  %94 = fmul float %28, 0x408FE051E0000000
>  %95 = fadd float %94, 0x3FF19999A0000000
>  %96 = fptosi float %95 to i32
>  %97 = shl i32 %96, 4
>  %98 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %97)
>  %99 = shl i32 %96, 4
>  %100 = or i32 %99, 4
>  %101 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %100)
>  %102 = shl i32 %96, 4
>  %103 = or i32 %102, 8
>  %104 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %103)
>  %105 = shl i32 %96, 4
>  %106 = or i32 %105, 12
>  %107 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %106)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %18, float %19, float %20, float %21)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %26, float %27, float %28, float %29)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 34, i32 0, float %84, float %87, float %90, float %93)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 35, i32 0, float %98, float %101, float %104, float %107)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %58, float %79, 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_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
>	v_mov_b32_e32 v1, 0x447f028f                            ; 7E0202FF 447F028F
>	v_mov_b32_e32 v2, 1.0                                   ; 7E0402F2
>	v_add_i32_e32 v0, vcc, s10, v0                          ; 4A00000A
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	buffer_load_format_xyzw v[3:6], v0, s[4:7], 0 idxen     ; E00C2000 80010300
>	buffer_load_format_xyzw v[7:10], v0, s[12:15], 0 idxen  ; E00C2000 80030700
>	buffer_load_format_xyzw v[11:14], v0, s[16:19], 0 idxen ; E00C2000 80040B00
>	s_waitcnt vmcnt(2)                                      ; BF8C0772
>	exp 15, 32, 0, 0, 0, v3, v4, v5, v6                     ; F800020F 06050403
>	s_waitcnt vmcnt(1)                                      ; BF8C0771
>	v_madak_f32_e32 v0, v9, v1, 0x40066666                  ; 42000309 40066666
>	s_waitcnt expcnt(0)                                     ; BF8C070F
>	v_madak_f32_e32 v3, v9, v1, 0x40466666                  ; 42060309 40466666
>	v_cvt_i32_f32_e32 v0, v0                                ; 7E001100
>	v_cvt_i32_f32_e32 v3, v3                                ; 7E061103
>	v_madak_f32_e32 v4, v9, v1, 0x3dcccccd                  ; 42080309 3DCCCCCD
>	v_madak_f32_e32 v1, v9, v1, 0x3f8ccccd                  ; 42020309 3F8CCCCD
>	v_cvt_i32_f32_e32 v4, v4                                ; 7E081104
>	v_cvt_i32_f32_e32 v1, v1                                ; 7E021101
>	v_lshlrev_b32_e32 v0, 4, v0                             ; 34000084
>	v_lshlrev_b32_e32 v3, 4, v3                             ; 34060684
>	v_lshlrev_b32_e32 v4, 4, v4                             ; 34080884
>	v_lshlrev_b32_e32 v1, 4, v1                             ; 34020284
>	buffer_load_dword v5, v0, s[0:3], 0 offen               ; E0301000 80000500
>	v_or_b32_e32 v6, 4, v0                                  ; 380C0084
>	v_or_b32_e32 v15, 8, v0                                 ; 381E0088
>	v_or_b32_e32 v0, 12, v0                                 ; 3800008C
>	buffer_load_dword v16, v3, s[0:3], 0 offen              ; E0301000 80001003
>	v_or_b32_e32 v17, 4, v3                                 ; 38220684
>	v_or_b32_e32 v18, 8, v3                                 ; 38240688
>	v_or_b32_e32 v3, 12, v3                                 ; 3806068C
>	buffer_load_dword v19, v4, s[0:3], 0 offen              ; E0301000 80001304
>	v_or_b32_e32 v20, 4, v4                                 ; 38280884
>	v_or_b32_e32 v21, 8, v4                                 ; 382A0888
>	v_or_b32_e32 v4, 12, v4                                 ; 3808088C
>	buffer_load_dword v22, v1, s[0:3], 0 offen              ; E0301000 80001601
>	v_or_b32_e32 v23, 4, v1                                 ; 382E0284
>	v_or_b32_e32 v24, 8, v1                                 ; 38300288
>	v_or_b32_e32 v1, 12, v1                                 ; 3802028C
>	buffer_load_dword v20, v20, s[0:3], 0 offen             ; E0301000 80001414
>	buffer_load_dword v21, v21, s[0:3], 0 offen             ; E0301000 80001515
>	buffer_load_dword v4, v4, s[0:3], 0 offen               ; E0301000 80000404
>	buffer_load_dword v6, v6, s[0:3], 0 offen               ; E0301000 80000606
>	buffer_load_dword v17, v17, s[0:3], 0 offen             ; E0301000 80001111
>	buffer_load_dword v23, v23, s[0:3], 0 offen             ; E0301000 80001717
>	buffer_load_dword v24, v24, s[0:3], 0 offen             ; E0301000 80001818
>	buffer_load_dword v1, v1, s[0:3], 0 offen               ; E0301000 80000101
>	buffer_load_dword v15, v15, s[0:3], 0 offen             ; E0301000 80000F0F
>	buffer_load_dword v18, v18, s[0:3], 0 offen             ; E0301000 80001212
>	buffer_load_dword v0, v0, s[0:3], 0 offen               ; E0301000 80000000
>	buffer_load_dword v3, v3, s[0:3], 0 offen               ; E0301000 80000303
>	exp 15, 33, 0, 0, 0, v7, v8, v9, v10                    ; F800021F 0A090807
>	s_waitcnt vmcnt(9)                                      ; BF8C0779
>	exp 15, 34, 0, 0, 0, v19, v20, v21, v4                  ; F800022F 04151413
>	s_waitcnt vmcnt(8) expcnt(0)                            ; BF8C0708
>	v_mul_f32_e32 v4, v6, v12                               ; 10081906
>	s_waitcnt vmcnt(7)                                      ; BF8C0777
>	v_mul_f32_e32 v6, v17, v12                              ; 100C1911
>	v_mac_f32_e32 v4, v5, v11                               ; 3E081705
>	v_mac_f32_e32 v6, v16, v11                              ; 3E0C1710
>	s_waitcnt vmcnt(4)                                      ; BF8C0774
>	exp 15, 35, 0, 0, 0, v22, v23, v24, v1                  ; F800023F 01181716
>	s_waitcnt vmcnt(3)                                      ; BF8C0773
>	v_mac_f32_e32 v4, v15, v13                              ; 3E081B0F
>	s_waitcnt vmcnt(2)                                      ; BF8C0772
>	v_mac_f32_e32 v6, v18, v13                              ; 3E0C1B12
>	s_waitcnt vmcnt(1)                                      ; BF8C0771
>	v_mac_f32_e32 v4, v0, v14                               ; 3E081D00
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	v_mac_f32_e32 v6, v3, v14                               ; 3E0C1D03
>	v_mov_b32_e32 v0, 0                                     ; 7E000280
>	exp 15, 12, 0, 1, 0, v4, v6, v0, v2                     ; F80008CF 02000604
>	s_endpgm                                                ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 24
>VGPRS: 28
>Code Size: 424 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
>  clamp_color = 0
>FRAG
>PROPERTY FS_COLOR0_WRITES_ALL_CBUFS 1
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL IN[1], GENERIC[1], PERSPECTIVE
>DCL IN[2], GENERIC[2], PERSPECTIVE
>DCL IN[3], GENERIC[3], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL TEMP[0..1], LOCAL
>  0: MAD TEMP[0], IN[0], IN[3], IN[2]
>  1: MUL TEMP[1].x, TEMP[0].wwww, IN[1].wwww
>  2: MOV TEMP[0].w, TEMP[1].xxxx
>  3: MOV OUT[0], TEMP[0]
>  4: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %8)
>  %24 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %8)
>  %25 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %6, <2 x i32> %8)
>  %26 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %6, <2 x i32> %8)
>  %27 = call float @llvm.SI.fs.interp(i32 3, i32 1, i32 %6, <2 x i32> %8)
>  %28 = call float @llvm.SI.fs.interp(i32 0, i32 2, i32 %6, <2 x i32> %8)
>  %29 = call float @llvm.SI.fs.interp(i32 1, i32 2, i32 %6, <2 x i32> %8)
>  %30 = call float @llvm.SI.fs.interp(i32 2, i32 2, i32 %6, <2 x i32> %8)
>  %31 = call float @llvm.SI.fs.interp(i32 3, i32 2, i32 %6, <2 x i32> %8)
>  %32 = call float @llvm.SI.fs.interp(i32 0, i32 3, i32 %6, <2 x i32> %8)
>  %33 = call float @llvm.SI.fs.interp(i32 1, i32 3, i32 %6, <2 x i32> %8)
>  %34 = call float @llvm.SI.fs.interp(i32 2, i32 3, i32 %6, <2 x i32> %8)
>  %35 = call float @llvm.SI.fs.interp(i32 3, i32 3, i32 %6, <2 x i32> %8)
>  %36 = fmul float %23, %32
>  %37 = fadd float %36, %28
>  %38 = fmul float %24, %33
>  %39 = fadd float %38, %29
>  %40 = fmul float %25, %34
>  %41 = fadd float %40, %30
>  %42 = fmul float %26, %35
>  %43 = fadd float %42, %31
>  %44 = fmul float %43, %27
>  %45 = call i32 @llvm.SI.packf16(float %37, float %39)
>  %46 = bitcast i32 %45 to float
>  %47 = call i32 @llvm.SI.packf16(float %41, float %44)
>  %48 = bitcast i32 %47 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %46, float %48, float %46, float %48)
>  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_mov_b32 m0, s10                          ; BEFC030A
>	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_interp_p1_f32 v4, v0, 2, 0, [m0]         ; C8100200
>	v_interp_p2_f32 v4, [v4], v1, 2, 0, [m0]   ; C8110201
>	v_interp_p1_f32 v5, v0, 3, 0, [m0]         ; C8140300
>	v_interp_p2_f32 v5, [v5], v1, 3, 0, [m0]   ; C8150301
>	v_interp_p1_f32 v6, v0, 3, 1, [m0]         ; C8180700
>	v_interp_p2_f32 v6, [v6], v1, 3, 1, [m0]   ; C8190701
>	v_interp_p1_f32 v7, v0, 0, 2, [m0]         ; C81C0800
>	v_interp_p2_f32 v7, [v7], v1, 0, 2, [m0]   ; C81D0801
>	v_interp_p1_f32 v8, v0, 1, 2, [m0]         ; C8200900
>	v_interp_p2_f32 v8, [v8], v1, 1, 2, [m0]   ; C8210901
>	v_interp_p1_f32 v9, v0, 2, 2, [m0]         ; C8240A00
>	v_interp_p2_f32 v9, [v9], v1, 2, 2, [m0]   ; C8250A01
>	v_interp_p1_f32 v10, v0, 3, 2, [m0]        ; C8280B00
>	v_interp_p2_f32 v10, [v10], v1, 3, 2, [m0] ; C8290B01
>	v_interp_p1_f32 v11, v0, 0, 3, [m0]        ; C82C0C00
>	v_interp_p2_f32 v11, [v11], v1, 0, 3, [m0] ; C82D0C01
>	v_interp_p1_f32 v12, v0, 1, 3, [m0]        ; C8300D00
>	v_interp_p2_f32 v12, [v12], v1, 1, 3, [m0] ; C8310D01
>	v_interp_p1_f32 v13, v0, 2, 3, [m0]        ; C8340E00
>	v_interp_p2_f32 v13, [v13], v1, 2, 3, [m0] ; C8350E01
>	v_interp_p1_f32 v0, v0, 3, 3, [m0]         ; C8000F00
>	v_interp_p2_f32 v0, [v0], v1, 3, 3, [m0]   ; C8010F01
>	v_mac_f32_e32 v7, v11, v2                  ; 3E0E050B
>	v_mac_f32_e32 v8, v12, v3                  ; 3E10070C
>	v_mac_f32_e32 v9, v13, v4                  ; 3E12090D
>	v_mac_f32_e32 v10, v0, v5                  ; 3E140B00
>	v_mul_f32_e32 v0, v6, v10                  ; 10001506
>	v_cvt_pkrtz_f16_f32_e32 v1, v7, v8         ; 5E021107
>	v_cvt_pkrtz_f16_f32_e32 v0, v9, v0         ; 5E000109
>	exp 15, 0, 1, 1, 1, v1, v0, v1, v0         ; F8001C0F 00010001
>	s_endpgm                                   ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 16
>VGPRS: 16
>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
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL IN[2]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[0]
>DCL OUT[2], GENERIC[1]
>DCL CONST[0..3]
>DCL TEMP[0..2], LOCAL
>IMM[0] FLT32 {    0.0000,     1.0000,     0.0000,     0.0000}
>  0: MOV TEMP[0].zw, IMM[0].yyxy
>  1: DP4 TEMP[0].x, IN[2], CONST[2]
>  2: DP4 TEMP[1].x, IN[2], CONST[3]
>  3: MOV TEMP[0].y, TEMP[1].xxxx
>  4: MAD TEMP[1], IN[0], CONST[1], CONST[0]
>  5: MOV TEMP[2].xy, IN[1].xyxx
>  6: MOV OUT[1], TEMP[1]
>  7: MOV OUT[0], TEMP[0]
>  8: MOV OUT[2], TEMP[2]
>  9: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = call float @llvm.SI.load.const(<16 x i8> %13, i32 0)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %13, i32 4)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %13, i32 8)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %13, i32 12)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %13, i32 16)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %13, i32 20)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %13, i32 24)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %13, i32 28)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %13, i32 32)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %13, i32 36)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %13, i32 40)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %13, i32 44)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %13, i32 48)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %13, i32 52)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %13, i32 56)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %13, i32 60)
>  %30 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %31 = load <16 x i8>, <16 x i8> addrspace(2)* %30, align 16, !tbaa !0
>  %32 = add i32 %5, %8
>  %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 = extractelement <4 x float> %33, i32 2
>  %37 = extractelement <4 x float> %33, i32 3
>  %38 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 1
>  %39 = load <16 x i8>, <16 x i8> addrspace(2)* %38, align 16, !tbaa !0
>  %40 = add i32 %5, %8
>  %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>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 2
>  %45 = load <16 x i8>, <16 x i8> addrspace(2)* %44, align 16, !tbaa !0
>  %46 = add i32 %5, %8
>  %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 = fmul float %48, %22
>  %53 = fmul float %49, %23
>  %54 = fadd float %52, %53
>  %55 = fmul float %50, %24
>  %56 = fadd float %54, %55
>  %57 = fmul float %51, %25
>  %58 = fadd float %56, %57
>  %59 = fmul float %48, %26
>  %60 = fmul float %49, %27
>  %61 = fadd float %59, %60
>  %62 = fmul float %50, %28
>  %63 = fadd float %61, %62
>  %64 = fmul float %51, %29
>  %65 = fadd float %63, %64
>  %66 = fmul float %34, %18
>  %67 = fadd float %66, %14
>  %68 = fmul float %35, %19
>  %69 = fadd float %68, %15
>  %70 = fmul float %36, %20
>  %71 = fadd float %70, %16
>  %72 = fmul float %37, %21
>  %73 = fadd float %72, %17
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %67, float %69, float %71, float %73)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %42, float %43, float undef, float undef)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %58, float %65, 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_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
>	v_add_i32_e32 v0, vcc, s10, v0                         ; 4A00000A
>	v_mov_b32_e32 v1, 1.0                                  ; 7E0202F2
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	s_buffer_load_dword s8, s[0:3], 0x0                    ; C2040100
>	s_buffer_load_dword s9, s[0:3], 0x1                    ; C2048101
>	s_buffer_load_dword s10, s[0:3], 0x2                   ; C2050102
>	s_buffer_load_dword s11, s[0:3], 0x3                   ; C2058103
>	s_buffer_load_dword s20, s[0:3], 0x4                   ; C20A0104
>	buffer_load_format_xyzw v[2:5], v0, s[4:7], 0 idxen    ; E00C2000 80010200
>	buffer_load_format_xyzw v[6:9], v0, s[12:15], 0 idxen  ; E00C2000 80030600
>	s_waitcnt vmcnt(0)                                     ; BF8C0770
>	buffer_load_format_xyzw v[8:11], v0, s[16:19], 0 idxen ; E00C2000 80040800
>	s_buffer_load_dword s4, s[0:3], 0x5                    ; C2020105
>	s_buffer_load_dword s5, s[0:3], 0x6                    ; C2028106
>	s_buffer_load_dword s6, s[0:3], 0x7                    ; C2030107
>	s_buffer_load_dword s7, s[0:3], 0x8                    ; C2038108
>	s_buffer_load_dword s12, s[0:3], 0x9                   ; C2060109
>	s_buffer_load_dword s13, s[0:3], 0xa                   ; C206810A
>	s_buffer_load_dword s14, s[0:3], 0xb                   ; C207010B
>	s_buffer_load_dword s15, s[0:3], 0xc                   ; C207810C
>	s_buffer_load_dword s16, s[0:3], 0xd                   ; C208010D
>	s_buffer_load_dword s17, s[0:3], 0xe                   ; C208810E
>	s_buffer_load_dword s0, s[0:3], 0xf                    ; C200010F
>	s_waitcnt lgkmcnt(0)                                   ; BF8C007F
>	v_mov_b32_e32 v0, s8                                   ; 7E000208
>	v_mov_b32_e32 v12, s9                                  ; 7E180209
>	v_mov_b32_e32 v13, s10                                 ; 7E1A020A
>	v_mov_b32_e32 v14, s11                                 ; 7E1C020B
>	v_mac_f32_e32 v0, s20, v2                              ; 3E000414
>	v_mac_f32_e32 v12, s4, v3                              ; 3E180604
>	v_mac_f32_e32 v13, s5, v4                              ; 3E1A0805
>	v_mac_f32_e32 v14, s6, v5                              ; 3E1C0A06
>	s_waitcnt vmcnt(0)                                     ; BF8C0770
>	v_mul_f32_e32 v2, s12, v9                              ; 1004120C
>	exp 15, 32, 0, 0, 0, v0, v12, v13, v14                 ; F800020F 0E0D0C00
>	s_waitcnt expcnt(0)                                    ; BF8C070F
>	v_mul_f32_e32 v0, s16, v9                              ; 10001210
>	v_mac_f32_e32 v2, s7, v8                               ; 3E041007
>	v_mac_f32_e32 v0, s15, v8                              ; 3E00100F
>	v_mac_f32_e32 v2, s13, v10                             ; 3E04140D
>	v_mac_f32_e32 v0, s17, v10                             ; 3E001411
>	exp 15, 33, 0, 0, 0, v6, v7, v0, v0                    ; F800021F 00000706
>	v_mac_f32_e32 v2, s14, v11                             ; 3E04160E
>	s_waitcnt expcnt(0)                                    ; BF8C070F
>	v_mac_f32_e32 v0, s0, v11                              ; 3E001600
>	v_mov_b32_e32 v3, 0                                    ; 7E060280
>	exp 15, 12, 0, 1, 0, v2, v0, v3, v1                    ; F80008CF 01030002
>	s_endpgm                                               ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 24
>VGPRS: 16
>Code Size: 232 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
>  clamp_color = 0
>FRAG
>PROPERTY FS_COLOR0_WRITES_ALL_CBUFS 1
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL IN[1], GENERIC[1], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SVIEW[0], 2D, FLOAT
>DCL TEMP[0..1], LOCAL
>  0: MOV TEMP[0].xyz, IN[0].xyzx
>  1: MOV TEMP[1].xy, IN[1].xyyy
>  2: TEX TEMP[1].w, TEMP[1], SAMP[0], 2D
>  3: MUL TEMP[1].x, IN[0].wwww, TEMP[1].wwww
>  4: MOV TEMP[0].w, TEMP[1].xxxx
>  5: MOV OUT[0], TEMP[0]
>  6: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %24 = load <8 x i32>, <8 x i32> addrspace(2)* %23, align 32, !tbaa !0
>  %25 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %26 = load <4 x i32>, <4 x i32> addrspace(2)* %25, align 16, !tbaa !0
>  %27 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %8)
>  %28 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %8)
>  %29 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %6, <2 x i32> %8)
>  %30 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %6, <2 x i32> %8)
>  %31 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %6, <2 x i32> %8)
>  %32 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %6, <2 x i32> %8)
>  %33 = bitcast float %31 to i32
>  %34 = bitcast float %32 to i32
>  %35 = insertelement <2 x i32> undef, i32 %33, i32 0
>  %36 = insertelement <2 x i32> %35, i32 %34, i32 1
>  %37 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %36, <8 x i32> %24, <4 x i32> %26, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %38 = extractelement <4 x float> %37, i32 3
>  %39 = fmul float %30, %38
>  %40 = call i32 @llvm.SI.packf16(float %27, float %28)
>  %41 = bitcast i32 %40 to float
>  %42 = call i32 @llvm.SI.packf16(float %29, float %39)
>  %43 = bitcast i32 %42 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %41, float %43, float %41, float %43)
>  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.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, 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 }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                             ; BEFE0A7E
>	s_mov_b32 m0, s10                                                ; BEFC030A
>	s_load_dwordx8 s[8:15], s[6:7], 0x0                              ; C0C40700
>	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
>	s_load_dwordx4 s[0:3], s[4:5], 0x0                               ; C0800500
>	v_interp_p2_f32 v3, [v3], v1, 1, 0, [m0]                         ; C80D0101
>	v_interp_p1_f32 v4, v0, 2, 0, [m0]                               ; C8100200
>	v_interp_p2_f32 v4, [v4], v1, 2, 0, [m0]                         ; C8110201
>	v_interp_p1_f32 v5, v0, 3, 0, [m0]                               ; C8140300
>	v_interp_p2_f32 v5, [v5], v1, 3, 0, [m0]                         ; C8150301
>	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 v7, v0, 1, 1, [m0]                               ; C81C0500
>	v_interp_p2_f32 v7, [v7], v1, 1, 1, [m0]                         ; C81D0501
>	s_waitcnt lgkmcnt(0)                                             ; BF8C007F
>	image_sample v0, 8, 0, 0, 0, 0, 0, 0, 0, v[6:7], s[8:15], s[0:3] ; F0800800 00020006
>	s_waitcnt vmcnt(0)                                               ; BF8C0770
>	v_mul_f32_e32 v0, v0, v5                                         ; 10000B00
>	v_cvt_pkrtz_f16_f32_e32 v1, v2, v3                               ; 5E020702
>	v_cvt_pkrtz_f16_f32_e32 v0, v4, v0                               ; 5E000104
>	exp 15, 0, 1, 1, 1, v1, v0, v1, v0                               ; F8001C0F 00010001
>	s_endpgm                                                         ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 16
>VGPRS: 8
>Code Size: 104 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
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL IN[2]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[0]
>DCL OUT[2], GENERIC[1]
>DCL CONST[0..47]
>DCL TEMP[0..1], LOCAL
>DCL ADDR[0]
>IMM[0] FLT32 {    0.0000,     1.0000,   510.0200,     0.1000}
>IMM[1] FLT32 {    1.1000,     0.0000,     0.0000,     0.0000}
>  0: MOV TEMP[0].zw, IMM[0].yyxy
>  1: MAD TEMP[1].x, IMM[0].zzzz, IN[1].zzzz, IMM[0].wwww
>  2: F2I TEMP[1].x, TEMP[1].xxxx
>  3: UARL ADDR[0].x, TEMP[1].xxxx
>  4: UARL ADDR[0].x, TEMP[1].xxxx
>  5: DP4 TEMP[0].x, IN[2], CONST[ADDR[0].x]
>  6: MAD TEMP[1].x, IMM[0].zzzz, IN[1].zzzz, IMM[1].xxxx
>  7: F2I TEMP[1].x, TEMP[1].xxxx
>  8: UARL ADDR[0].x, TEMP[1].xxxx
>  9: DP4 TEMP[1].x, IN[2], CONST[ADDR[0].x]
> 10: MOV TEMP[0].y, TEMP[1].xxxx
> 11: MOV OUT[2], IN[1]
> 12: MOV OUT[1], IN[0]
> 13: MOV OUT[0], TEMP[0]
> 14: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %15 = load <16 x i8>, <16 x i8> addrspace(2)* %14, align 16, !tbaa !0
>  %16 = add i32 %5, %8
>  %17 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %15, i32 0, i32 %16)
>  %18 = extractelement <4 x float> %17, i32 0
>  %19 = extractelement <4 x float> %17, i32 1
>  %20 = extractelement <4 x float> %17, i32 2
>  %21 = extractelement <4 x float> %17, i32 3
>  %22 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 1
>  %23 = load <16 x i8>, <16 x i8> addrspace(2)* %22, align 16, !tbaa !0
>  %24 = add i32 %5, %8
>  %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 = extractelement <4 x float> %25, i32 3
>  %30 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 2
>  %31 = load <16 x i8>, <16 x i8> addrspace(2)* %30, align 16, !tbaa !0
>  %32 = add i32 %5, %8
>  %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 = extractelement <4 x float> %33, i32 2
>  %37 = extractelement <4 x float> %33, i32 3
>  %38 = fmul float %28, 0x407FE051E0000000
>  %39 = fadd float %38, 0x3FB99999A0000000
>  %40 = fptosi float %39 to i32
>  %41 = shl i32 %40, 4
>  %42 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %41)
>  %43 = shl i32 %40, 4
>  %44 = or i32 %43, 4
>  %45 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %44)
>  %46 = shl i32 %40, 4
>  %47 = or i32 %46, 8
>  %48 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %47)
>  %49 = shl i32 %40, 4
>  %50 = or i32 %49, 12
>  %51 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %50)
>  %52 = fmul float %34, %42
>  %53 = fmul float %35, %45
>  %54 = fadd float %52, %53
>  %55 = fmul float %36, %48
>  %56 = fadd float %54, %55
>  %57 = fmul float %37, %51
>  %58 = fadd float %56, %57
>  %59 = fmul float %28, 0x407FE051E0000000
>  %60 = fadd float %59, 0x3FF19999A0000000
>  %61 = fptosi float %60 to i32
>  %62 = shl i32 %61, 4
>  %63 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %62)
>  %64 = shl i32 %61, 4
>  %65 = or i32 %64, 4
>  %66 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %65)
>  %67 = shl i32 %61, 4
>  %68 = or i32 %67, 8
>  %69 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %68)
>  %70 = shl i32 %61, 4
>  %71 = or i32 %70, 12
>  %72 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %71)
>  %73 = fmul float %34, %63
>  %74 = fmul float %35, %66
>  %75 = fadd float %73, %74
>  %76 = fmul float %36, %69
>  %77 = fadd float %75, %76
>  %78 = fmul float %37, %72
>  %79 = fadd float %77, %78
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %18, float %19, float %20, float %21)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %26, float %27, float %28, float %29)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %58, float %79, 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_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
>	v_mov_b32_e32 v1, 0x43ff028f                            ; 7E0202FF 43FF028F
>	v_mov_b32_e32 v2, 1.0                                   ; 7E0402F2
>	v_add_i32_e32 v0, vcc, s10, v0                          ; 4A00000A
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	buffer_load_format_xyzw v[3:6], v0, s[4:7], 0 idxen     ; E00C2000 80010300
>	buffer_load_format_xyzw v[7:10], v0, s[12:15], 0 idxen  ; E00C2000 80030700
>	buffer_load_format_xyzw v[11:14], v0, s[16:19], 0 idxen ; E00C2000 80040B00
>	s_waitcnt vmcnt(2)                                      ; BF8C0772
>	exp 15, 32, 0, 0, 0, v3, v4, v5, v6                     ; F800020F 06050403
>	s_waitcnt vmcnt(1)                                      ; BF8C0771
>	v_madak_f32_e32 v0, v9, v1, 0x3dcccccd                  ; 42000309 3DCCCCCD
>	v_madak_f32_e32 v1, v9, v1, 0x3f8ccccd                  ; 42020309 3F8CCCCD
>	v_cvt_i32_f32_e32 v0, v0                                ; 7E001100
>	v_cvt_i32_f32_e32 v1, v1                                ; 7E021101
>	v_lshlrev_b32_e32 v0, 4, v0                             ; 34000084
>	v_lshlrev_b32_e32 v1, 4, v1                             ; 34020284
>	s_waitcnt expcnt(0)                                     ; BF8C070F
>	buffer_load_dword v3, v0, s[0:3], 0 offen               ; E0301000 80000300
>	v_or_b32_e32 v4, 4, v0                                  ; 38080084
>	v_or_b32_e32 v5, 8, v0                                  ; 380A0088
>	v_or_b32_e32 v0, 12, v0                                 ; 3800008C
>	buffer_load_dword v6, v1, s[0:3], 0 offen               ; E0301000 80000601
>	v_or_b32_e32 v15, 4, v1                                 ; 381E0284
>	buffer_load_dword v4, v4, s[0:3], 0 offen               ; E0301000 80000404
>	buffer_load_dword v15, v15, s[0:3], 0 offen             ; E0301000 80000F0F
>	v_or_b32_e32 v16, 8, v1                                 ; 38200288
>	v_or_b32_e32 v1, 12, v1                                 ; 3802028C
>	buffer_load_dword v5, v5, s[0:3], 0 offen               ; E0301000 80000505
>	buffer_load_dword v16, v16, s[0:3], 0 offen             ; E0301000 80001010
>	buffer_load_dword v0, v0, s[0:3], 0 offen               ; E0301000 80000000
>	buffer_load_dword v1, v1, s[0:3], 0 offen               ; E0301000 80000101
>	exp 15, 33, 0, 0, 0, v7, v8, v9, v10                    ; F800021F 0A090807
>	s_waitcnt vmcnt(5)                                      ; BF8C0775
>	v_mul_f32_e32 v4, v4, v12                               ; 10081904
>	s_waitcnt vmcnt(4) expcnt(0)                            ; BF8C0704
>	v_mul_f32_e32 v7, v15, v12                              ; 100E190F
>	v_mac_f32_e32 v4, v3, v11                               ; 3E081703
>	v_mac_f32_e32 v7, v6, v11                               ; 3E0E1706
>	s_waitcnt vmcnt(3)                                      ; BF8C0773
>	v_mac_f32_e32 v4, v5, v13                               ; 3E081B05
>	s_waitcnt vmcnt(2)                                      ; BF8C0772
>	v_mac_f32_e32 v7, v16, v13                              ; 3E0E1B10
>	s_waitcnt vmcnt(1)                                      ; BF8C0771
>	v_mac_f32_e32 v4, v0, v14                               ; 3E081D00
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	v_mac_f32_e32 v7, v1, v14                               ; 3E0E1D01
>	v_mov_b32_e32 v0, 0                                     ; 7E000280
>	exp 15, 12, 0, 1, 0, v4, v7, v0, v2                     ; F80008CF 02000704
>	s_endpgm                                                ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 24
>VGPRS: 20
>Code Size: 280 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
>  clamp_color = 0
>FRAG
>PROPERTY FS_COLOR0_WRITES_ALL_CBUFS 1
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL IN[1], GENERIC[1], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL TEMP[0..1], LOCAL
>  0: MOV TEMP[0].xyz, IN[0].xyzx
>  1: MUL TEMP[1].x, IN[0].wwww, IN[1].wwww
>  2: MOV TEMP[0].w, TEMP[1].xxxx
>  3: MOV OUT[0], TEMP[0]
>  4: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %8)
>  %24 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %8)
>  %25 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %6, <2 x i32> %8)
>  %26 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %6, <2 x i32> %8)
>  %27 = call float @llvm.SI.fs.interp(i32 3, i32 1, i32 %6, <2 x i32> %8)
>  %28 = fmul float %26, %27
>  %29 = call i32 @llvm.SI.packf16(float %23, float %24)
>  %30 = bitcast i32 %29 to float
>  %31 = call i32 @llvm.SI.packf16(float %25, float %28)
>  %32 = bitcast i32 %31 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %30, float %32, float %30, float %32)
>  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_mov_b32 m0, s10                        ; BEFC030A
>	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_interp_p1_f32 v4, v0, 2, 0, [m0]       ; C8100200
>	v_interp_p2_f32 v4, [v4], v1, 2, 0, [m0] ; C8110201
>	v_interp_p1_f32 v5, v0, 3, 0, [m0]       ; C8140300
>	v_interp_p2_f32 v5, [v5], v1, 3, 0, [m0] ; C8150301
>	v_interp_p1_f32 v0, v0, 3, 1, [m0]       ; C8000700
>	v_interp_p2_f32 v0, [v0], v1, 3, 1, [m0] ; C8010701
>	v_mul_f32_e32 v0, v0, v5                 ; 10000B00
>	v_cvt_pkrtz_f16_f32_e32 v1, v2, v3       ; 5E020702
>	v_cvt_pkrtz_f16_f32_e32 v0, v4, v0       ; 5E000104
>	exp 15, 0, 1, 1, 1, v1, v0, v1, v0       ; F8001C0F 00010001
>	s_endpgm                                 ; BF810000
>
>*** SHADER STATS ***
>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
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL OUT[1], CLIPVERTEX
>DCL OUT[2], GENERIC[0]
>DCL OUT[3], GENERIC[1]
>DCL CONST[0..54]
>DCL TEMP[0..3], LOCAL
>  0: MUL TEMP[0].xy, CONST[48].xyyy, IN[1].xyyy
>  1: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
>  2: ADD TEMP[1].x, TEMP[0].xxxx, CONST[48].wwww
>  3: MOV TEMP[1].z, TEMP[1].xxxx
>  4: MUL TEMP[0].xy, CONST[49].xyyy, IN[1].xyyy
>  5: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
>  6: ADD TEMP[2].x, TEMP[0].xxxx, CONST[49].wwww
>  7: MOV TEMP[1].w, TEMP[2].xxxx
>  8: MUL TEMP[0].xy, CONST[50].xyyy, IN[1].xyyy
>  9: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 10: ADD TEMP[2].x, TEMP[0].xxxx, CONST[50].wwww
> 11: MUL TEMP[0].xy, CONST[51].xyyy, IN[1].xyyy
> 12: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 13: ADD TEMP[3].x, TEMP[0].xxxx, CONST[51].wwww
> 14: MOV TEMP[2].y, TEMP[3].xxxx
> 15: MUL TEMP[0].xy, CONST[52].xyyy, IN[1].xyyy
> 16: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 17: ADD TEMP[3].x, TEMP[0].xxxx, CONST[52].wwww
> 18: MOV TEMP[2].z, TEMP[3].xxxx
> 19: MUL TEMP[0].xy, CONST[53].xyyy, IN[1].xyyy
> 20: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 21: ADD TEMP[0].x, TEMP[0].xxxx, CONST[53].wwww
> 22: MOV TEMP[2].w, TEMP[0].xxxx
> 23: MOV TEMP[0].xw, IN[0].xxxw
> 24: MOV TEMP[1].xy, IN[1].xyxx
> 25: MAD TEMP[3].x, IN[0].zzzz, CONST[0].zzzz, -IN[0].wwww
> 26: MOV TEMP[0].z, TEMP[3].xxxx
> 27: MOV TEMP[0].y, -IN[0].yyyy
> 28: MAD TEMP[0].xy, CONST[54].xyyy, IN[0].wwww, TEMP[0].xyyy
> 29: MOV OUT[2], TEMP[1]
> 30: MOV OUT[3], TEMP[2]
> 31: MOV OUT[0], TEMP[0]
> 32: MOV OUT[1], IN[0]
> 33: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = call float @llvm.SI.load.const(<16 x i8> %13, i32 8)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %13, i32 768)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %13, i32 772)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %13, i32 780)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %13, i32 784)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %13, i32 788)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %13, i32 796)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %13, i32 800)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %13, i32 804)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %13, i32 812)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %13, i32 816)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %13, i32 820)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %13, i32 828)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %13, i32 832)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %13, i32 836)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %13, i32 844)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %13, i32 848)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %13, i32 852)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %13, i32 860)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %13, i32 864)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %13, i32 868)
>  %35 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %36 = load <16 x i8>, <16 x i8> addrspace(2)* %35, align 16, !tbaa !0
>  %37 = add i32 %5, %8
>  %38 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %36, i32 0, i32 %37)
>  %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 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 1
>  %44 = load <16 x i8>, <16 x i8> addrspace(2)* %43, align 16, !tbaa !0
>  %45 = add i32 %5, %8
>  %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 = fmul float %15, %47
>  %50 = fmul float %16, %48
>  %51 = fadd float %50, %49
>  %52 = fadd float %51, %17
>  %53 = fmul float %18, %47
>  %54 = fmul float %19, %48
>  %55 = fadd float %54, %53
>  %56 = fadd float %55, %20
>  %57 = fmul float %21, %47
>  %58 = fmul float %22, %48
>  %59 = fadd float %58, %57
>  %60 = fadd float %59, %23
>  %61 = fmul float %24, %47
>  %62 = fmul float %25, %48
>  %63 = fadd float %62, %61
>  %64 = fadd float %63, %26
>  %65 = fmul float %27, %47
>  %66 = fmul float %28, %48
>  %67 = fadd float %66, %65
>  %68 = fadd float %67, %29
>  %69 = fmul float %30, %47
>  %70 = fmul float %31, %48
>  %71 = fadd float %70, %69
>  %72 = fadd float %71, %32
>  %73 = fmul float %41, %14
>  %74 = fsub float %73, %42
>  %75 = fmul float %33, %42
>  %76 = fadd float %75, %39
>  %77 = fmul float %34, %42
>  %78 = fsub float %77, %40
>  %79 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 16
>  %80 = load <16 x i8>, <16 x i8> addrspace(2)* %79, align 16, !tbaa !0
>  %81 = call float @llvm.SI.load.const(<16 x i8> %80, i32 0)
>  %82 = fmul float %81, %39
>  %83 = call float @llvm.SI.load.const(<16 x i8> %80, i32 4)
>  %84 = fmul float %83, %40
>  %85 = fadd float %82, %84
>  %86 = call float @llvm.SI.load.const(<16 x i8> %80, i32 8)
>  %87 = fmul float %86, %41
>  %88 = fadd float %85, %87
>  %89 = call float @llvm.SI.load.const(<16 x i8> %80, i32 12)
>  %90 = fmul float %89, %42
>  %91 = fadd float %88, %90
>  %92 = call float @llvm.SI.load.const(<16 x i8> %80, i32 16)
>  %93 = fmul float %92, %39
>  %94 = call float @llvm.SI.load.const(<16 x i8> %80, i32 20)
>  %95 = fmul float %94, %40
>  %96 = fadd float %93, %95
>  %97 = call float @llvm.SI.load.const(<16 x i8> %80, i32 24)
>  %98 = fmul float %97, %41
>  %99 = fadd float %96, %98
>  %100 = call float @llvm.SI.load.const(<16 x i8> %80, i32 28)
>  %101 = fmul float %100, %42
>  %102 = fadd float %99, %101
>  %103 = call float @llvm.SI.load.const(<16 x i8> %80, i32 32)
>  %104 = fmul float %103, %39
>  %105 = call float @llvm.SI.load.const(<16 x i8> %80, i32 36)
>  %106 = fmul float %105, %40
>  %107 = fadd float %104, %106
>  %108 = call float @llvm.SI.load.const(<16 x i8> %80, i32 40)
>  %109 = fmul float %108, %41
>  %110 = fadd float %107, %109
>  %111 = call float @llvm.SI.load.const(<16 x i8> %80, i32 44)
>  %112 = fmul float %111, %42
>  %113 = fadd float %110, %112
>  %114 = call float @llvm.SI.load.const(<16 x i8> %80, i32 48)
>  %115 = fmul float %114, %39
>  %116 = call float @llvm.SI.load.const(<16 x i8> %80, i32 52)
>  %117 = fmul float %116, %40
>  %118 = fadd float %115, %117
>  %119 = call float @llvm.SI.load.const(<16 x i8> %80, i32 56)
>  %120 = fmul float %119, %41
>  %121 = fadd float %118, %120
>  %122 = call float @llvm.SI.load.const(<16 x i8> %80, i32 60)
>  %123 = fmul float %122, %42
>  %124 = fadd float %121, %123
>  %125 = call float @llvm.SI.load.const(<16 x i8> %80, i32 64)
>  %126 = fmul float %125, %39
>  %127 = call float @llvm.SI.load.const(<16 x i8> %80, i32 68)
>  %128 = fmul float %127, %40
>  %129 = fadd float %126, %128
>  %130 = call float @llvm.SI.load.const(<16 x i8> %80, i32 72)
>  %131 = fmul float %130, %41
>  %132 = fadd float %129, %131
>  %133 = call float @llvm.SI.load.const(<16 x i8> %80, i32 76)
>  %134 = fmul float %133, %42
>  %135 = fadd float %132, %134
>  %136 = call float @llvm.SI.load.const(<16 x i8> %80, i32 80)
>  %137 = fmul float %136, %39
>  %138 = call float @llvm.SI.load.const(<16 x i8> %80, i32 84)
>  %139 = fmul float %138, %40
>  %140 = fadd float %137, %139
>  %141 = call float @llvm.SI.load.const(<16 x i8> %80, i32 88)
>  %142 = fmul float %141, %41
>  %143 = fadd float %140, %142
>  %144 = call float @llvm.SI.load.const(<16 x i8> %80, i32 92)
>  %145 = fmul float %144, %42
>  %146 = fadd float %143, %145
>  %147 = call float @llvm.SI.load.const(<16 x i8> %80, i32 96)
>  %148 = fmul float %147, %39
>  %149 = call float @llvm.SI.load.const(<16 x i8> %80, i32 100)
>  %150 = fmul float %149, %40
>  %151 = fadd float %148, %150
>  %152 = call float @llvm.SI.load.const(<16 x i8> %80, i32 104)
>  %153 = fmul float %152, %41
>  %154 = fadd float %151, %153
>  %155 = call float @llvm.SI.load.const(<16 x i8> %80, i32 108)
>  %156 = fmul float %155, %42
>  %157 = fadd float %154, %156
>  %158 = call float @llvm.SI.load.const(<16 x i8> %80, i32 112)
>  %159 = fmul float %158, %39
>  %160 = call float @llvm.SI.load.const(<16 x i8> %80, i32 116)
>  %161 = fmul float %160, %40
>  %162 = fadd float %159, %161
>  %163 = call float @llvm.SI.load.const(<16 x i8> %80, i32 120)
>  %164 = fmul float %163, %41
>  %165 = fadd float %162, %164
>  %166 = call float @llvm.SI.load.const(<16 x i8> %80, i32 124)
>  %167 = fmul float %166, %42
>  %168 = fadd float %165, %167
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %47, float %48, float %52, float %56)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %60, float %64, float %68, float %72)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 12, i32 0, float %76, float %78, float %74, float %42)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 13, i32 0, float %91, float %102, float %113, float %124)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 14, i32 0, float %135, float %146, float %157, 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
>
>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[4:7], s[8:9], 0x0                    ; C0820900
>	s_load_dwordx4 s[12:15], s[8:9], 0x4                  ; C0860904
>	v_add_i32_e32 v0, vcc, s10, v0                        ; 4A00000A
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[4:7], 0 idxen   ; E00C2000 80010100
>	buffer_load_format_xyzw v[5:8], v0, s[12:15], 0 idxen ; E00C2000 80030500
>	s_load_dwordx4 s[4:7], s[2:3], 0x0                    ; C0820300
>	s_load_dwordx4 s[0:3], s[2:3], 0x40                   ; C0800340
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	s_buffer_load_dword s8, s[4:7], 0x2                   ; C2040502
>	s_buffer_load_dword s9, s[4:7], 0xc0                  ; C20485C0
>	s_buffer_load_dword s10, s[4:7], 0xc1                 ; C20505C1
>	s_buffer_load_dword s11, s[4:7], 0xc3                 ; C20585C3
>	s_buffer_load_dword s12, s[4:7], 0xc4                 ; C20605C4
>	s_buffer_load_dword s13, s[4:7], 0xc5                 ; C20685C5
>	s_buffer_load_dword s14, s[4:7], 0xc7                 ; C20705C7
>	s_buffer_load_dword s15, s[4:7], 0xc8                 ; C20785C8
>	s_buffer_load_dword s16, s[4:7], 0xc9                 ; C20805C9
>	s_buffer_load_dword s17, s[4:7], 0xcb                 ; C20885CB
>	s_buffer_load_dword s18, s[4:7], 0xcc                 ; C20905CC
>	s_buffer_load_dword s19, s[4:7], 0xcd                 ; C20985CD
>	s_buffer_load_dword s20, s[4:7], 0xcf                 ; C20A05CF
>	s_buffer_load_dword s21, s[4:7], 0xd0                 ; C20A85D0
>	s_buffer_load_dword s22, s[4:7], 0xd1                 ; C20B05D1
>	s_buffer_load_dword s23, s[4:7], 0xd3                 ; C20B85D3
>	s_buffer_load_dword s24, s[4:7], 0xd4                 ; C20C05D4
>	s_buffer_load_dword s25, s[4:7], 0xd5                 ; C20C85D5
>	s_buffer_load_dword s26, s[4:7], 0xd7                 ; C20D05D7
>	s_buffer_load_dword s27, s[4:7], 0xd8                 ; C20D85D8
>	s_buffer_load_dword s4, s[4:7], 0xd9                  ; C20205D9
>	s_buffer_load_dword s5, s[0:3], 0x0                   ; C2028100
>	s_buffer_load_dword s6, s[0:3], 0x1                   ; C2030101
>	s_buffer_load_dword s7, s[0:3], 0x2                   ; C2038102
>	s_buffer_load_dword s28, s[0:3], 0x3                  ; C20E0103
>	s_buffer_load_dword s29, s[0:3], 0x4                  ; C20E8104
>	s_buffer_load_dword s30, s[0:3], 0x5                  ; C20F0105
>	s_buffer_load_dword s31, s[0:3], 0x6                  ; C20F8106
>	s_buffer_load_dword s32, s[0:3], 0x7                  ; C2100107
>	s_buffer_load_dword s33, s[0:3], 0x8                  ; C2108108
>	s_buffer_load_dword s34, s[0:3], 0x9                  ; C2110109
>	s_buffer_load_dword s35, s[0:3], 0xa                  ; C211810A
>	s_buffer_load_dword s36, s[0:3], 0xb                  ; C212010B
>	s_buffer_load_dword s37, s[0:3], 0xc                  ; C212810C
>	s_buffer_load_dword s38, s[0:3], 0xd                  ; C213010D
>	s_buffer_load_dword s39, s[0:3], 0xe                  ; C213810E
>	s_buffer_load_dword s40, s[0:3], 0xf                  ; C214010F
>	s_buffer_load_dword s41, s[0:3], 0x10                 ; C2148110
>	s_buffer_load_dword s42, s[0:3], 0x11                 ; C2150111
>	s_buffer_load_dword s43, s[0:3], 0x12                 ; C2158112
>	s_buffer_load_dword s44, s[0:3], 0x13                 ; C2160113
>	s_buffer_load_dword s45, s[0:3], 0x14                 ; C2168114
>	s_buffer_load_dword s46, s[0:3], 0x15                 ; C2170115
>	s_buffer_load_dword s47, s[0:3], 0x16                 ; C2178116
>	s_buffer_load_dword s48, s[0:3], 0x17                 ; C2180117
>	s_buffer_load_dword s49, s[0:3], 0x18                 ; C2188118
>	s_buffer_load_dword s50, s[0:3], 0x19                 ; C2190119
>	s_buffer_load_dword s51, s[0:3], 0x1a                 ; C219811A
>	s_buffer_load_dword s52, s[0:3], 0x1b                 ; C21A011B
>	s_buffer_load_dword s53, s[0:3], 0x1c                 ; C21A811C
>	s_buffer_load_dword s54, s[0:3], 0x1d                 ; C21B011D
>	s_buffer_load_dword s55, s[0:3], 0x1e                 ; C21B811E
>	s_buffer_load_dword s0, s[0:3], 0x1f                  ; C200011F
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mov_b32_e32 v0, s11                                 ; 7E00020B
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mov_b32_e32 v7, s14                                 ; 7E0E020E
>	v_mov_b32_e32 v8, s17                                 ; 7E100211
>	v_mov_b32_e32 v9, s20                                 ; 7E120214
>	v_mov_b32_e32 v10, s23                                ; 7E140217
>	v_mov_b32_e32 v11, s26                                ; 7E16021A
>	v_mac_f32_e32 v0, s9, v5                              ; 3E000A09
>	v_mac_f32_e32 v7, s12, v5                             ; 3E0E0A0C
>	v_mac_f32_e32 v8, s15, v5                             ; 3E100A0F
>	v_mac_f32_e32 v9, s18, v5                             ; 3E120A12
>	v_mac_f32_e32 v10, s21, v5                            ; 3E140A15
>	v_mac_f32_e32 v11, s24, v5                            ; 3E160A18
>	v_mac_f32_e32 v0, s10, v6                             ; 3E000C0A
>	v_mac_f32_e32 v7, s13, v6                             ; 3E0E0C0D
>	v_mac_f32_e32 v8, s16, v6                             ; 3E100C10
>	v_mac_f32_e32 v9, s19, v6                             ; 3E120C13
>	exp 15, 32, 0, 0, 0, v5, v6, v0, v7                   ; F800020F 07000605
>	v_mac_f32_e32 v10, s22, v6                            ; 3E140C16
>	v_mac_f32_e32 v11, s25, v6                            ; 3E160C19
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mad_f32 v0, v3, s8, -v4                             ; D2820000 84101103
>	v_mad_f32 v5, s27, v4, v1                             ; D2820005 0406081B
>	v_mad_f32 v6, s4, v4, -v2                             ; D2820006 840A0804
>	v_mul_f32_e32 v7, s6, v2                              ; 100E0406
>	exp 15, 33, 0, 0, 0, v8, v9, v10, v11                 ; F800021F 0B0A0908
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mul_f32_e32 v8, s30, v2                             ; 1010041E
>	v_mac_f32_e32 v7, s5, v1                              ; 3E0E0205
>	exp 15, 12, 0, 0, 0, v5, v6, v0, v4                   ; F80000CF 04000605
>	v_mac_f32_e32 v8, s29, v1                             ; 3E10021D
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mul_f32_e32 v0, s34, v2                             ; 10000422
>	v_mac_f32_e32 v0, s33, v1                             ; 3E000221
>	v_mul_f32_e32 v5, s38, v2                             ; 100A0426
>	v_mac_f32_e32 v5, s37, v1                             ; 3E0A0225
>	v_mul_f32_e32 v6, s42, v2                             ; 100C042A
>	v_mul_f32_e32 v9, s46, v2                             ; 1012042E
>	v_mul_f32_e32 v10, s50, v2                            ; 10140432
>	v_mul_f32_e32 v2, s54, v2                             ; 10040436
>	v_mac_f32_e32 v6, s41, v1                             ; 3E0C0229
>	v_mac_f32_e32 v9, s45, v1                             ; 3E12022D
>	v_mac_f32_e32 v10, s49, v1                            ; 3E140231
>	v_mac_f32_e32 v2, s53, v1                             ; 3E040235
>	v_mac_f32_e32 v7, s7, v3                              ; 3E0E0607
>	v_mac_f32_e32 v8, s31, v3                             ; 3E10061F
>	v_mac_f32_e32 v0, s35, v3                             ; 3E000623
>	v_mac_f32_e32 v5, s39, v3                             ; 3E0A0627
>	v_mac_f32_e32 v6, s43, v3                             ; 3E0C062B
>	v_mac_f32_e32 v9, s47, v3                             ; 3E12062F
>	v_mac_f32_e32 v10, s51, v3                            ; 3E140633
>	v_mac_f32_e32 v2, s55, v3                             ; 3E040637
>	v_mac_f32_e32 v7, s28, v4                             ; 3E0E081C
>	v_mac_f32_e32 v8, s32, v4                             ; 3E100820
>	v_mac_f32_e32 v0, s36, v4                             ; 3E000824
>	v_mac_f32_e32 v5, s40, v4                             ; 3E0A0828
>	v_mac_f32_e32 v6, s44, v4                             ; 3E0C082C
>	v_mac_f32_e32 v9, s48, v4                             ; 3E120830
>	v_mac_f32_e32 v10, s52, v4                            ; 3E140834
>	v_mac_f32_e32 v2, s0, v4                              ; 3E040800
>	exp 15, 13, 0, 0, 0, v7, v8, v0, v5                   ; F80000DF 05000807
>	exp 15, 14, 0, 1, 0, v6, v9, v10, v2                  ; F80008EF 020A0906
>	s_endpgm                                              ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 64
>VGPRS: 12
>Code Size: 544 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
>  clamp_color = 0
>FRAG
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL IN[1], GENERIC[1], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SAMP[1]
>DCL SAMP[2]
>DCL SVIEW[0], 2D, FLOAT
>DCL SVIEW[1], 2D, FLOAT
>DCL SVIEW[2], 2D, FLOAT
>DCL CONST[0..3]
>DCL TEMP[0..3], LOCAL
>IMM[0] FLT32 {    6.0000,     1.0000,    -0.5800,    10.0000}
>IMM[1] FLT32 {   -2.0000,     3.0000,    -1.0000,    -0.0000}
>IMM[2] FLT32 {    1.0000,     0.0000,     0.0000,     0.0000}
>  0: MOV TEMP[0].xy, IN[0].xyyy
>  1: TEX TEMP[0].xw, TEMP[0], SAMP[0], 2D
>  2: MOV TEMP[1].xy, IN[1].xyyy
>  3: TEX TEMP[1].y, TEMP[1], SAMP[1], 2D
>  4: MAD TEMP[0].x, TEMP[1].yyyy, TEMP[0].xxxx, TEMP[0].wwww
>  5: MAD TEMP[2].x, CONST[3].wwww, IMM[0].xxxx, IMM[0].yyyy
>  6: MAD TEMP[0].x, TEMP[0].xxxx, TEMP[2].xxxx, IMM[0].zzzz
>  7: MUL TEMP[0].x, TEMP[0].xxxx, IMM[0].wwww
>  8: MOV_SAT TEMP[2].x, TEMP[0].xxxx
>  9: MAD TEMP[3].x, TEMP[2].xxxx, IMM[1].xxxx, IMM[1].yyyy
> 10: MUL TEMP[0].x, TEMP[2].xxxx, TEMP[2].xxxx
> 11: MUL TEMP[0].x, TEMP[0].xxxx, TEMP[3].xxxx
> 12: MOV TEMP[2].xy, IN[0].xyyy
> 13: TEX TEMP[2].xyz, TEMP[2], SAMP[2], 2D
> 14: MAD TEMP[2].xyz, CONST[3].yyyy, IMM[1].zwww, TEMP[2].xyzz
> 15: MUL TEMP[1].xyz, IMM[2].xyyy, CONST[3].yyyy
> 16: MAD TEMP[0].xyz, TEMP[0].xxxx, TEMP[2].xyzz, TEMP[1].xyzz
> 17: MOV TEMP[0].w, IMM[0].yyyy
> 18: MOV OUT[0], TEMP[0]
> 19: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %24 = load <16 x i8>, <16 x i8> addrspace(2)* %23, align 16, !tbaa !0
>  %25 = call float @llvm.SI.load.const(<16 x i8> %24, i32 52)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %24, i32 60)
>  %27 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %28 = load <8 x i32>, <8 x i32> addrspace(2)* %27, align 32, !tbaa !0
>  %29 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %30 = load <4 x i32>, <4 x i32> addrspace(2)* %29, align 16, !tbaa !0
>  %31 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 1
>  %32 = load <8 x i32>, <8 x i32> addrspace(2)* %31, align 32, !tbaa !0
>  %33 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 1
>  %34 = load <4 x i32>, <4 x i32> addrspace(2)* %33, align 16, !tbaa !0
>  %35 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 2
>  %36 = load <8 x i32>, <8 x i32> addrspace(2)* %35, align 32, !tbaa !0
>  %37 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 2
>  %38 = load <4 x i32>, <4 x i32> addrspace(2)* %37, align 16, !tbaa !0
>  %39 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %8)
>  %40 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %8)
>  %41 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %6, <2 x i32> %8)
>  %42 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %6, <2 x i32> %8)
>  %43 = bitcast float %39 to i32
>  %44 = bitcast float %40 to i32
>  %45 = insertelement <2 x i32> undef, i32 %43, i32 0
>  %46 = insertelement <2 x i32> %45, i32 %44, i32 1
>  %47 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %46, <8 x i32> %28, <4 x i32> %30, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %48 = extractelement <4 x float> %47, i32 0
>  %49 = extractelement <4 x float> %47, i32 3
>  %50 = bitcast float %41 to i32
>  %51 = bitcast float %42 to i32
>  %52 = insertelement <2 x i32> undef, i32 %50, i32 0
>  %53 = insertelement <2 x i32> %52, i32 %51, i32 1
>  %54 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %53, <8 x i32> %32, <4 x i32> %34, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %55 = extractelement <4 x float> %54, i32 1
>  %56 = fmul float %55, %48
>  %57 = fadd float %56, %49
>  %58 = fmul float %26, 6.000000e+00
>  %59 = fadd float %58, 1.000000e+00
>  %60 = fmul float %57, %59
>  %61 = fadd float %60, 0xBFE28F5C20000000
>  %62 = fmul float %61, 1.000000e+01
>  %63 = call float @llvm.AMDIL.clamp.(float %62, float 0.000000e+00, float 1.000000e+00)
>  %64 = fmul float %63, -2.000000e+00
>  %65 = fadd float %64, 3.000000e+00
>  %66 = fmul float %63, %63
>  %67 = fmul float %66, %65
>  %68 = bitcast float %39 to i32
>  %69 = bitcast float %40 to i32
>  %70 = insertelement <2 x i32> undef, i32 %68, i32 0
>  %71 = insertelement <2 x i32> %70, i32 %69, i32 1
>  %72 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %71, <8 x i32> %36, <4 x i32> %38, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %73 = extractelement <4 x float> %72, i32 0
>  %74 = extractelement <4 x float> %72, i32 1
>  %75 = extractelement <4 x float> %72, i32 2
>  %76 = fsub float %73, %25
>  %77 = fmul float %25, -0.000000e+00
>  %78 = fadd float %77, %74
>  %79 = fmul float %25, -0.000000e+00
>  %80 = fadd float %79, %75
>  %81 = fmul float %25, 0.000000e+00
>  %82 = fmul float %25, 0.000000e+00
>  %83 = fmul float %67, %76
>  %84 = fadd float %83, %25
>  %85 = fmul float %67, %78
>  %86 = fadd float %85, %81
>  %87 = fmul float %67, %80
>  %88 = fadd float %87, %82
>  %89 = call i32 @llvm.SI.packf16(float %84, float %86)
>  %90 = bitcast i32 %89 to float
>  %91 = call i32 @llvm.SI.packf16(float %88, float 1.000000e+00)
>  %92 = bitcast i32 %91 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %90, float %92, float %90, float %92)
>  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.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #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, s10                                                       ; BEFC030A
>	v_interp_p1_f32 v2, v0, 0, 0, [m0]                                      ; C8080000
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                                      ; C0800300
>	s_load_dwordx4 s[8:11], s[4:5], 0x0                                     ; C0840500
>	s_load_dwordx4 s[12:15], s[4:5], 0x4                                    ; C0860504
>	s_load_dwordx4 s[16:19], s[4:5], 0x8                                    ; C0880508
>	v_interp_p2_f32 v2, [v2], v1, 0, 0, [m0]                                ; C8090001
>	s_load_dwordx8 s[20:27], s[6:7], 0x0                                    ; C0CA0700
>	s_load_dwordx8 s[28:35], s[6:7], 0x8                                    ; C0CE0708
>	s_load_dwordx8 s[36:43], s[6:7], 0x10                                   ; C0D20710
>	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, 1, [m0]                                      ; C8100400
>	s_waitcnt lgkmcnt(0)                                                    ; BF8C007F
>	s_buffer_load_dword s4, s[0:3], 0xf                                     ; C202010F
>	v_interp_p2_f32 v4, [v4], v1, 0, 1, [m0]                                ; C8110401
>	v_interp_p1_f32 v5, v0, 1, 1, [m0]                                      ; C8140500
>	v_interp_p2_f32 v5, [v5], v1, 1, 1, [m0]                                ; C8150501
>	image_sample v[0:1], 9, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[20:27], s[8:11]  ; F0800900 00450002
>	image_sample v4, 2, 0, 0, 0, 0, 0, 0, 0, v[4:5], s[28:35], s[12:15]     ; F0800200 00670404
>	image_sample v[5:7], 7, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[36:43], s[16:19] ; F0800700 00890502
>	s_buffer_load_dword s0, s[0:3], 0xd                                     ; C200010D
>	s_waitcnt vmcnt(1)                                                      ; BF8C0771
>	v_mac_f32_e32 v1, v0, v4                                                ; 3E020900
>	v_mov_b32_e32 v0, 0x40c00000                                            ; 7E0002FF 40C00000
>	s_waitcnt lgkmcnt(0)                                                    ; BF8C007F
>	v_mad_f32 v0, v0, s4, 1.0                                               ; D2820000 03C80900
>	v_madak_f32_e32 v0, v1, v0, 0xbf147ae1                                  ; 42000101 BF147AE1
>	v_mul_f32_e32 v0, 0x41200000, v0                                        ; 100000FF 41200000
>	v_add_f32_e64 v0, 0, v0 clamp                                           ; D2060800 00020080
>	v_madak_f32_e32 v1, -2.0, v0, 0x40400000                                ; 420200F5 40400000
>	v_mul_f32_e32 v0, v0, v0                                                ; 10000100
>	v_mul_f32_e32 v0, v1, v0                                                ; 10000101
>	s_waitcnt vmcnt(0)                                                      ; BF8C0770
>	v_subrev_f32_e32 v1, s0, v5                                             ; 0A020A00
>	v_mov_b32_e32 v2, 0x80000000                                            ; 7E0402FF 80000000
>	v_mad_f32 v3, s0, v2, v6                                                ; D2820003 041A0400
>	v_mac_f32_e32 v7, s0, v2                                                ; 3E0E0400
>	v_mul_f32_e64 v2, 0, s0                                                 ; D2100002 00000080
>	v_mad_f32 v1, v0, v1, s0                                                ; D2820001 00020300
>	v_mad_f32 v3, v3, v0, v2                                                ; D2820003 040A0103
>	v_mac_f32_e32 v2, v7, v0                                                ; 3E040107
>	v_cvt_pkrtz_f16_f32_e32 v0, v1, v3                                      ; 5E000701
>	v_cvt_pkrtz_f16_f32_e64 v1, v2, 1.0                                     ; D25E0001 0001E502
>	exp 15, 0, 1, 1, 1, v0, v1, v0, v1                                      ; F8001C0F 01000100
>	s_endpgm                                                                ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 48
>VGPRS: 8
>Code Size: 252 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
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL OUT[1], CLIPVERTEX
>DCL OUT[2], GENERIC[0]
>DCL OUT[3], GENERIC[1]
>DCL CONST[0..54]
>DCL TEMP[0..3], LOCAL
>  0: MUL TEMP[0].xy, CONST[48].xyyy, IN[1].xyyy
>  1: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
>  2: ADD TEMP[1].x, TEMP[0].xxxx, CONST[48].wwww
>  3: MOV TEMP[1].z, TEMP[1].xxxx
>  4: MUL TEMP[0].xy, CONST[49].xyyy, IN[1].xyyy
>  5: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
>  6: ADD TEMP[2].x, TEMP[0].xxxx, CONST[49].wwww
>  7: MOV TEMP[1].w, TEMP[2].xxxx
>  8: MUL TEMP[0].xy, CONST[50].xyyy, IN[1].xyyy
>  9: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 10: ADD TEMP[2].x, TEMP[0].xxxx, CONST[50].wwww
> 11: MUL TEMP[0].xy, CONST[51].xyyy, IN[1].xyyy
> 12: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 13: ADD TEMP[3].x, TEMP[0].xxxx, CONST[51].wwww
> 14: MOV TEMP[2].y, TEMP[3].xxxx
> 15: MUL TEMP[0].xy, CONST[52].xyyy, IN[1].xyyy
> 16: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 17: ADD TEMP[3].x, TEMP[0].xxxx, CONST[52].wwww
> 18: MOV TEMP[2].z, TEMP[3].xxxx
> 19: MUL TEMP[0].xy, CONST[53].xyyy, IN[1].xyyy
> 20: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 21: ADD TEMP[0].x, TEMP[0].xxxx, CONST[53].wwww
> 22: MOV TEMP[2].w, TEMP[0].xxxx
> 23: MOV TEMP[0].xw, IN[0].xxxw
> 24: MOV TEMP[1].xy, IN[1].xyxx
> 25: MAD TEMP[3].x, IN[0].zzzz, CONST[0].zzzz, -IN[0].wwww
> 26: MOV TEMP[0].z, TEMP[3].xxxx
> 27: MOV TEMP[0].y, -IN[0].yyyy
> 28: MAD TEMP[0].xy, CONST[54].xyyy, IN[0].wwww, TEMP[0].xyyy
> 29: MOV OUT[2], TEMP[1]
> 30: MOV OUT[3], TEMP[2]
> 31: MOV OUT[0], TEMP[0]
> 32: MOV OUT[1], IN[0]
> 33: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = call float @llvm.SI.load.const(<16 x i8> %13, i32 8)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %13, i32 768)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %13, i32 772)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %13, i32 780)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %13, i32 784)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %13, i32 788)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %13, i32 796)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %13, i32 800)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %13, i32 804)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %13, i32 812)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %13, i32 816)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %13, i32 820)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %13, i32 828)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %13, i32 832)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %13, i32 836)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %13, i32 844)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %13, i32 848)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %13, i32 852)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %13, i32 860)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %13, i32 864)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %13, i32 868)
>  %35 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %36 = load <16 x i8>, <16 x i8> addrspace(2)* %35, align 16, !tbaa !0
>  %37 = add i32 %5, %8
>  %38 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %36, i32 0, i32 %37)
>  %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 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 1
>  %44 = load <16 x i8>, <16 x i8> addrspace(2)* %43, align 16, !tbaa !0
>  %45 = add i32 %5, %8
>  %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 = fmul float %15, %47
>  %50 = fmul float %16, %48
>  %51 = fadd float %50, %49
>  %52 = fadd float %51, %17
>  %53 = fmul float %18, %47
>  %54 = fmul float %19, %48
>  %55 = fadd float %54, %53
>  %56 = fadd float %55, %20
>  %57 = fmul float %21, %47
>  %58 = fmul float %22, %48
>  %59 = fadd float %58, %57
>  %60 = fadd float %59, %23
>  %61 = fmul float %24, %47
>  %62 = fmul float %25, %48
>  %63 = fadd float %62, %61
>  %64 = fadd float %63, %26
>  %65 = fmul float %27, %47
>  %66 = fmul float %28, %48
>  %67 = fadd float %66, %65
>  %68 = fadd float %67, %29
>  %69 = fmul float %30, %47
>  %70 = fmul float %31, %48
>  %71 = fadd float %70, %69
>  %72 = fadd float %71, %32
>  %73 = fmul float %41, %14
>  %74 = fsub float %73, %42
>  %75 = fmul float %33, %42
>  %76 = fadd float %75, %39
>  %77 = fmul float %34, %42
>  %78 = fsub float %77, %40
>  %79 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 16
>  %80 = load <16 x i8>, <16 x i8> addrspace(2)* %79, align 16, !tbaa !0
>  %81 = call float @llvm.SI.load.const(<16 x i8> %80, i32 0)
>  %82 = fmul float %81, %39
>  %83 = call float @llvm.SI.load.const(<16 x i8> %80, i32 4)
>  %84 = fmul float %83, %40
>  %85 = fadd float %82, %84
>  %86 = call float @llvm.SI.load.const(<16 x i8> %80, i32 8)
>  %87 = fmul float %86, %41
>  %88 = fadd float %85, %87
>  %89 = call float @llvm.SI.load.const(<16 x i8> %80, i32 12)
>  %90 = fmul float %89, %42
>  %91 = fadd float %88, %90
>  %92 = call float @llvm.SI.load.const(<16 x i8> %80, i32 16)
>  %93 = fmul float %92, %39
>  %94 = call float @llvm.SI.load.const(<16 x i8> %80, i32 20)
>  %95 = fmul float %94, %40
>  %96 = fadd float %93, %95
>  %97 = call float @llvm.SI.load.const(<16 x i8> %80, i32 24)
>  %98 = fmul float %97, %41
>  %99 = fadd float %96, %98
>  %100 = call float @llvm.SI.load.const(<16 x i8> %80, i32 28)
>  %101 = fmul float %100, %42
>  %102 = fadd float %99, %101
>  %103 = call float @llvm.SI.load.const(<16 x i8> %80, i32 32)
>  %104 = fmul float %103, %39
>  %105 = call float @llvm.SI.load.const(<16 x i8> %80, i32 36)
>  %106 = fmul float %105, %40
>  %107 = fadd float %104, %106
>  %108 = call float @llvm.SI.load.const(<16 x i8> %80, i32 40)
>  %109 = fmul float %108, %41
>  %110 = fadd float %107, %109
>  %111 = call float @llvm.SI.load.const(<16 x i8> %80, i32 44)
>  %112 = fmul float %111, %42
>  %113 = fadd float %110, %112
>  %114 = call float @llvm.SI.load.const(<16 x i8> %80, i32 48)
>  %115 = fmul float %114, %39
>  %116 = call float @llvm.SI.load.const(<16 x i8> %80, i32 52)
>  %117 = fmul float %116, %40
>  %118 = fadd float %115, %117
>  %119 = call float @llvm.SI.load.const(<16 x i8> %80, i32 56)
>  %120 = fmul float %119, %41
>  %121 = fadd float %118, %120
>  %122 = call float @llvm.SI.load.const(<16 x i8> %80, i32 60)
>  %123 = fmul float %122, %42
>  %124 = fadd float %121, %123
>  %125 = call float @llvm.SI.load.const(<16 x i8> %80, i32 64)
>  %126 = fmul float %125, %39
>  %127 = call float @llvm.SI.load.const(<16 x i8> %80, i32 68)
>  %128 = fmul float %127, %40
>  %129 = fadd float %126, %128
>  %130 = call float @llvm.SI.load.const(<16 x i8> %80, i32 72)
>  %131 = fmul float %130, %41
>  %132 = fadd float %129, %131
>  %133 = call float @llvm.SI.load.const(<16 x i8> %80, i32 76)
>  %134 = fmul float %133, %42
>  %135 = fadd float %132, %134
>  %136 = call float @llvm.SI.load.const(<16 x i8> %80, i32 80)
>  %137 = fmul float %136, %39
>  %138 = call float @llvm.SI.load.const(<16 x i8> %80, i32 84)
>  %139 = fmul float %138, %40
>  %140 = fadd float %137, %139
>  %141 = call float @llvm.SI.load.const(<16 x i8> %80, i32 88)
>  %142 = fmul float %141, %41
>  %143 = fadd float %140, %142
>  %144 = call float @llvm.SI.load.const(<16 x i8> %80, i32 92)
>  %145 = fmul float %144, %42
>  %146 = fadd float %143, %145
>  %147 = call float @llvm.SI.load.const(<16 x i8> %80, i32 96)
>  %148 = fmul float %147, %39
>  %149 = call float @llvm.SI.load.const(<16 x i8> %80, i32 100)
>  %150 = fmul float %149, %40
>  %151 = fadd float %148, %150
>  %152 = call float @llvm.SI.load.const(<16 x i8> %80, i32 104)
>  %153 = fmul float %152, %41
>  %154 = fadd float %151, %153
>  %155 = call float @llvm.SI.load.const(<16 x i8> %80, i32 108)
>  %156 = fmul float %155, %42
>  %157 = fadd float %154, %156
>  %158 = call float @llvm.SI.load.const(<16 x i8> %80, i32 112)
>  %159 = fmul float %158, %39
>  %160 = call float @llvm.SI.load.const(<16 x i8> %80, i32 116)
>  %161 = fmul float %160, %40
>  %162 = fadd float %159, %161
>  %163 = call float @llvm.SI.load.const(<16 x i8> %80, i32 120)
>  %164 = fmul float %163, %41
>  %165 = fadd float %162, %164
>  %166 = call float @llvm.SI.load.const(<16 x i8> %80, i32 124)
>  %167 = fmul float %166, %42
>  %168 = fadd float %165, %167
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %47, float %48, float %52, float %56)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %60, float %64, float %68, float %72)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 12, i32 0, float %76, float %78, float %74, float %42)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 13, i32 0, float %91, float %102, float %113, float %124)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 14, i32 0, float %135, float %146, float %157, 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
>
>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[4:7], s[8:9], 0x0                    ; C0820900
>	s_load_dwordx4 s[12:15], s[8:9], 0x4                  ; C0860904
>	v_add_i32_e32 v0, vcc, s10, v0                        ; 4A00000A
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[4:7], 0 idxen   ; E00C2000 80010100
>	buffer_load_format_xyzw v[5:8], v0, s[12:15], 0 idxen ; E00C2000 80030500
>	s_load_dwordx4 s[4:7], s[2:3], 0x0                    ; C0820300
>	s_load_dwordx4 s[0:3], s[2:3], 0x40                   ; C0800340
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	s_buffer_load_dword s8, s[4:7], 0x2                   ; C2040502
>	s_buffer_load_dword s9, s[4:7], 0xc0                  ; C20485C0
>	s_buffer_load_dword s10, s[4:7], 0xc1                 ; C20505C1
>	s_buffer_load_dword s11, s[4:7], 0xc3                 ; C20585C3
>	s_buffer_load_dword s12, s[4:7], 0xc4                 ; C20605C4
>	s_buffer_load_dword s13, s[4:7], 0xc5                 ; C20685C5
>	s_buffer_load_dword s14, s[4:7], 0xc7                 ; C20705C7
>	s_buffer_load_dword s15, s[4:7], 0xc8                 ; C20785C8
>	s_buffer_load_dword s16, s[4:7], 0xc9                 ; C20805C9
>	s_buffer_load_dword s17, s[4:7], 0xcb                 ; C20885CB
>	s_buffer_load_dword s18, s[4:7], 0xcc                 ; C20905CC
>	s_buffer_load_dword s19, s[4:7], 0xcd                 ; C20985CD
>	s_buffer_load_dword s20, s[4:7], 0xcf                 ; C20A05CF
>	s_buffer_load_dword s21, s[4:7], 0xd0                 ; C20A85D0
>	s_buffer_load_dword s22, s[4:7], 0xd1                 ; C20B05D1
>	s_buffer_load_dword s23, s[4:7], 0xd3                 ; C20B85D3
>	s_buffer_load_dword s24, s[4:7], 0xd4                 ; C20C05D4
>	s_buffer_load_dword s25, s[4:7], 0xd5                 ; C20C85D5
>	s_buffer_load_dword s26, s[4:7], 0xd7                 ; C20D05D7
>	s_buffer_load_dword s27, s[4:7], 0xd8                 ; C20D85D8
>	s_buffer_load_dword s4, s[4:7], 0xd9                  ; C20205D9
>	s_buffer_load_dword s5, s[0:3], 0x0                   ; C2028100
>	s_buffer_load_dword s6, s[0:3], 0x1                   ; C2030101
>	s_buffer_load_dword s7, s[0:3], 0x2                   ; C2038102
>	s_buffer_load_dword s28, s[0:3], 0x3                  ; C20E0103
>	s_buffer_load_dword s29, s[0:3], 0x4                  ; C20E8104
>	s_buffer_load_dword s30, s[0:3], 0x5                  ; C20F0105
>	s_buffer_load_dword s31, s[0:3], 0x6                  ; C20F8106
>	s_buffer_load_dword s32, s[0:3], 0x7                  ; C2100107
>	s_buffer_load_dword s33, s[0:3], 0x8                  ; C2108108
>	s_buffer_load_dword s34, s[0:3], 0x9                  ; C2110109
>	s_buffer_load_dword s35, s[0:3], 0xa                  ; C211810A
>	s_buffer_load_dword s36, s[0:3], 0xb                  ; C212010B
>	s_buffer_load_dword s37, s[0:3], 0xc                  ; C212810C
>	s_buffer_load_dword s38, s[0:3], 0xd                  ; C213010D
>	s_buffer_load_dword s39, s[0:3], 0xe                  ; C213810E
>	s_buffer_load_dword s40, s[0:3], 0xf                  ; C214010F
>	s_buffer_load_dword s41, s[0:3], 0x10                 ; C2148110
>	s_buffer_load_dword s42, s[0:3], 0x11                 ; C2150111
>	s_buffer_load_dword s43, s[0:3], 0x12                 ; C2158112
>	s_buffer_load_dword s44, s[0:3], 0x13                 ; C2160113
>	s_buffer_load_dword s45, s[0:3], 0x14                 ; C2168114
>	s_buffer_load_dword s46, s[0:3], 0x15                 ; C2170115
>	s_buffer_load_dword s47, s[0:3], 0x16                 ; C2178116
>	s_buffer_load_dword s48, s[0:3], 0x17                 ; C2180117
>	s_buffer_load_dword s49, s[0:3], 0x18                 ; C2188118
>	s_buffer_load_dword s50, s[0:3], 0x19                 ; C2190119
>	s_buffer_load_dword s51, s[0:3], 0x1a                 ; C219811A
>	s_buffer_load_dword s52, s[0:3], 0x1b                 ; C21A011B
>	s_buffer_load_dword s53, s[0:3], 0x1c                 ; C21A811C
>	s_buffer_load_dword s54, s[0:3], 0x1d                 ; C21B011D
>	s_buffer_load_dword s55, s[0:3], 0x1e                 ; C21B811E
>	s_buffer_load_dword s0, s[0:3], 0x1f                  ; C200011F
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mov_b32_e32 v0, s11                                 ; 7E00020B
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mov_b32_e32 v7, s14                                 ; 7E0E020E
>	v_mov_b32_e32 v8, s17                                 ; 7E100211
>	v_mov_b32_e32 v9, s20                                 ; 7E120214
>	v_mov_b32_e32 v10, s23                                ; 7E140217
>	v_mov_b32_e32 v11, s26                                ; 7E16021A
>	v_mac_f32_e32 v0, s9, v5                              ; 3E000A09
>	v_mac_f32_e32 v7, s12, v5                             ; 3E0E0A0C
>	v_mac_f32_e32 v8, s15, v5                             ; 3E100A0F
>	v_mac_f32_e32 v9, s18, v5                             ; 3E120A12
>	v_mac_f32_e32 v10, s21, v5                            ; 3E140A15
>	v_mac_f32_e32 v11, s24, v5                            ; 3E160A18
>	v_mac_f32_e32 v0, s10, v6                             ; 3E000C0A
>	v_mac_f32_e32 v7, s13, v6                             ; 3E0E0C0D
>	v_mac_f32_e32 v8, s16, v6                             ; 3E100C10
>	v_mac_f32_e32 v9, s19, v6                             ; 3E120C13
>	exp 15, 32, 0, 0, 0, v5, v6, v0, v7                   ; F800020F 07000605
>	v_mac_f32_e32 v10, s22, v6                            ; 3E140C16
>	v_mac_f32_e32 v11, s25, v6                            ; 3E160C19
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mad_f32 v0, v3, s8, -v4                             ; D2820000 84101103
>	v_mad_f32 v5, s27, v4, v1                             ; D2820005 0406081B
>	v_mad_f32 v6, s4, v4, -v2                             ; D2820006 840A0804
>	v_mul_f32_e32 v7, s6, v2                              ; 100E0406
>	exp 15, 33, 0, 0, 0, v8, v9, v10, v11                 ; F800021F 0B0A0908
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mul_f32_e32 v8, s30, v2                             ; 1010041E
>	v_mac_f32_e32 v7, s5, v1                              ; 3E0E0205
>	exp 15, 12, 0, 0, 0, v5, v6, v0, v4                   ; F80000CF 04000605
>	v_mac_f32_e32 v8, s29, v1                             ; 3E10021D
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mul_f32_e32 v0, s34, v2                             ; 10000422
>	v_mac_f32_e32 v0, s33, v1                             ; 3E000221
>	v_mul_f32_e32 v5, s38, v2                             ; 100A0426
>	v_mac_f32_e32 v5, s37, v1                             ; 3E0A0225
>	v_mul_f32_e32 v6, s42, v2                             ; 100C042A
>	v_mul_f32_e32 v9, s46, v2                             ; 1012042E
>	v_mul_f32_e32 v10, s50, v2                            ; 10140432
>	v_mul_f32_e32 v2, s54, v2                             ; 10040436
>	v_mac_f32_e32 v6, s41, v1                             ; 3E0C0229
>	v_mac_f32_e32 v9, s45, v1                             ; 3E12022D
>	v_mac_f32_e32 v10, s49, v1                            ; 3E140231
>	v_mac_f32_e32 v2, s53, v1                             ; 3E040235
>	v_mac_f32_e32 v7, s7, v3                              ; 3E0E0607
>	v_mac_f32_e32 v8, s31, v3                             ; 3E10061F
>	v_mac_f32_e32 v0, s35, v3                             ; 3E000623
>	v_mac_f32_e32 v5, s39, v3                             ; 3E0A0627
>	v_mac_f32_e32 v6, s43, v3                             ; 3E0C062B
>	v_mac_f32_e32 v9, s47, v3                             ; 3E12062F
>	v_mac_f32_e32 v10, s51, v3                            ; 3E140633
>	v_mac_f32_e32 v2, s55, v3                             ; 3E040637
>	v_mac_f32_e32 v7, s28, v4                             ; 3E0E081C
>	v_mac_f32_e32 v8, s32, v4                             ; 3E100820
>	v_mac_f32_e32 v0, s36, v4                             ; 3E000824
>	v_mac_f32_e32 v5, s40, v4                             ; 3E0A0828
>	v_mac_f32_e32 v6, s44, v4                             ; 3E0C082C
>	v_mac_f32_e32 v9, s48, v4                             ; 3E120830
>	v_mac_f32_e32 v10, s52, v4                            ; 3E140834
>	v_mac_f32_e32 v2, s0, v4                              ; 3E040800
>	exp 15, 13, 0, 0, 0, v7, v8, v0, v5                   ; F80000DF 05000807
>	exp 15, 14, 0, 1, 0, v6, v9, v10, v2                  ; F80008EF 020A0906
>	s_endpgm                                              ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 64
>VGPRS: 12
>Code Size: 544 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
>  clamp_color = 0
>FRAG
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL IN[1], GENERIC[1], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SAMP[1]
>DCL SAMP[2]
>DCL SAMP[3]
>DCL SAMP[4]
>DCL SVIEW[0], 2D, FLOAT
>DCL SVIEW[1], 2D, FLOAT
>DCL SVIEW[2], 2D, FLOAT
>DCL SVIEW[3], 2D, FLOAT
>DCL SVIEW[4], 2D, FLOAT
>DCL CONST[0..3]
>DCL TEMP[0..4], LOCAL
>IMM[0] FLT32 {   -0.5000,    -1.0000,    10.0000,   -10.0000}
>IMM[1] FLT32 {   -2.0000,     3.0000,     6.0000,     1.0000}
>IMM[2] FLT32 {   -0.9900,   100.0000,     1.0000,     0.0000}
>  0: MOV TEMP[0].xy, IN[0].xyyy
>  1: TEX TEMP[0].xw, TEMP[0], SAMP[0], 2D
>  2: MOV TEMP[1].xy, IN[1].xyyy
>  3: TEX TEMP[1].y, TEMP[1], SAMP[1], 2D
>  4: MAD TEMP[0].x, TEMP[1].yyyy, TEMP[0].xxxx, TEMP[0].wwww
>  5: MOV TEMP[1].xy, IN[0].zwww
>  6: TEX TEMP[1].w, TEMP[1], SAMP[4], 2D
>  7: ADD TEMP[2].xy, TEMP[1].wwww, IMM[0].xyyy
>  8: ADD TEMP[3].x, TEMP[1].wwww, TEMP[1].wwww
>  9: MOV_SAT TEMP[3].x, TEMP[3].xxxx
> 10: MUL TEMP[2].xy, TEMP[2].xyyy, IMM[0].zwww
> 11: MOV_SAT TEMP[2].xy, TEMP[2].xyyy
> 12: MAD TEMP[1].xy, TEMP[2].xyyy, IMM[1].xxxx, IMM[1].yyyy
> 13: MUL TEMP[2].xy, TEMP[2].xyyy, TEMP[2].xyyy
> 14: MUL TEMP[1].xy, TEMP[2].xyyy, TEMP[1].xyyy
> 15: MUL TEMP[1].x, TEMP[1].yyyy, TEMP[1].xxxx
> 16: MAD TEMP[2].x, CONST[3].wwww, IMM[1].zzzz, IMM[1].wwww
> 17: MAD TEMP[0].x, TEMP[0].xxxx, TEMP[2].xxxx, TEMP[1].xxxx
> 18: ADD TEMP[0].x, TEMP[0].xxxx, IMM[2].xxxx
> 19: MUL TEMP[0].x, TEMP[0].xxxx, IMM[2].yyyy
> 20: MOV_SAT TEMP[1].x, TEMP[0].xxxx
> 21: MAD TEMP[2].x, TEMP[1].xxxx, IMM[1].xxxx, IMM[1].yyyy
> 22: MUL TEMP[0].x, TEMP[1].xxxx, TEMP[1].xxxx
> 23: MAD TEMP[0].x, TEMP[2].xxxx, TEMP[0].xxxx, IMM[1].wwww
> 24: MOV TEMP[1].xy, IN[0].xyyy
> 25: TEX TEMP[1].xyz, TEMP[1], SAMP[3], 2D
> 26: ADD TEMP[0].x, TEMP[0].xxxx, -TEMP[1].xxxx
> 27: MOV_SAT TEMP[0].x, TEMP[0].xxxx
> 28: MAX TEMP[1].x, TEMP[1].yyyy, TEMP[1].zzzz
> 29: MAX TEMP[1].x, TEMP[1].xxxx, TEMP[0].xxxx
> 30: MOV TEMP[2].xy, IN[0].xyyy
> 31: TEX TEMP[2].xyz, TEMP[2], SAMP[2], 2D
> 32: ADD TEMP[4].x, -TEMP[3].xxxx, TEMP[2].xxxx
> 33: MAD TEMP[1].x, TEMP[1].xxxx, TEMP[4].xxxx, TEMP[3].xxxx
> 34: ADD TEMP[3].x, TEMP[2].yyyy, IMM[0].yyyy
> 35: MAD TEMP[2].xy, TEMP[2].zzzz, IMM[2].zwww, IMM[2].wzzz
> 36: MOV TEMP[1].zw, TEMP[2].yyxy
> 37: MAD TEMP[0].x, TEMP[0].xxxx, TEMP[3].xxxx, IMM[1].wwww
> 38: MOV TEMP[1].y, TEMP[0].xxxx
> 39: MOV OUT[0], TEMP[1]
> 40: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %24 = load <16 x i8>, <16 x i8> addrspace(2)* %23, align 16, !tbaa !0
>  %25 = call float @llvm.SI.load.const(<16 x i8> %24, i32 60)
>  %26 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %27 = load <8 x i32>, <8 x i32> addrspace(2)* %26, align 32, !tbaa !0
>  %28 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %29 = load <4 x i32>, <4 x i32> addrspace(2)* %28, align 16, !tbaa !0
>  %30 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 1
>  %31 = load <8 x i32>, <8 x i32> addrspace(2)* %30, align 32, !tbaa !0
>  %32 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 1
>  %33 = load <4 x i32>, <4 x i32> addrspace(2)* %32, align 16, !tbaa !0
>  %34 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 2
>  %35 = load <8 x i32>, <8 x i32> addrspace(2)* %34, align 32, !tbaa !0
>  %36 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 2
>  %37 = load <4 x i32>, <4 x i32> addrspace(2)* %36, align 16, !tbaa !0
>  %38 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 3
>  %39 = load <8 x i32>, <8 x i32> addrspace(2)* %38, align 32, !tbaa !0
>  %40 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 3
>  %41 = load <4 x i32>, <4 x i32> addrspace(2)* %40, align 16, !tbaa !0
>  %42 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 4
>  %43 = load <8 x i32>, <8 x i32> addrspace(2)* %42, align 32, !tbaa !0
>  %44 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 4
>  %45 = load <4 x i32>, <4 x i32> addrspace(2)* %44, align 16, !tbaa !0
>  %46 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %8)
>  %47 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %8)
>  %48 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %6, <2 x i32> %8)
>  %49 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %6, <2 x i32> %8)
>  %50 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %6, <2 x i32> %8)
>  %51 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %6, <2 x i32> %8)
>  %52 = bitcast float %46 to i32
>  %53 = bitcast float %47 to i32
>  %54 = insertelement <2 x i32> undef, i32 %52, i32 0
>  %55 = insertelement <2 x i32> %54, i32 %53, i32 1
>  %56 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %55, <8 x i32> %27, <4 x i32> %29, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %57 = extractelement <4 x float> %56, i32 0
>  %58 = extractelement <4 x float> %56, i32 3
>  %59 = bitcast float %50 to i32
>  %60 = bitcast float %51 to i32
>  %61 = insertelement <2 x i32> undef, i32 %59, i32 0
>  %62 = insertelement <2 x i32> %61, i32 %60, i32 1
>  %63 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %62, <8 x i32> %31, <4 x i32> %33, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %64 = extractelement <4 x float> %63, i32 1
>  %65 = fmul float %64, %57
>  %66 = fadd float %65, %58
>  %67 = bitcast float %48 to i32
>  %68 = bitcast float %49 to i32
>  %69 = insertelement <2 x i32> undef, i32 %67, i32 0
>  %70 = insertelement <2 x i32> %69, i32 %68, i32 1
>  %71 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %70, <8 x i32> %43, <4 x i32> %45, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %72 = extractelement <4 x float> %71, i32 3
>  %73 = fadd float %72, -5.000000e-01
>  %74 = fadd float %72, -1.000000e+00
>  %75 = fadd float %72, %72
>  %76 = call float @llvm.AMDIL.clamp.(float %75, float 0.000000e+00, float 1.000000e+00)
>  %77 = fmul float %73, 1.000000e+01
>  %78 = fmul float %74, -1.000000e+01
>  %79 = call float @llvm.AMDIL.clamp.(float %77, float 0.000000e+00, float 1.000000e+00)
>  %80 = call float @llvm.AMDIL.clamp.(float %78, float 0.000000e+00, float 1.000000e+00)
>  %81 = fmul float %79, -2.000000e+00
>  %82 = fadd float %81, 3.000000e+00
>  %83 = fmul float %80, -2.000000e+00
>  %84 = fadd float %83, 3.000000e+00
>  %85 = fmul float %79, %79
>  %86 = fmul float %80, %80
>  %87 = fmul float %85, %82
>  %88 = fmul float %86, %84
>  %89 = fmul float %88, %87
>  %90 = fmul float %25, 6.000000e+00
>  %91 = fadd float %90, 1.000000e+00
>  %92 = fmul float %66, %91
>  %93 = fadd float %92, %89
>  %94 = fadd float %93, 0xBFEFAE1480000000
>  %95 = fmul float %94, 1.000000e+02
>  %96 = call float @llvm.AMDIL.clamp.(float %95, float 0.000000e+00, float 1.000000e+00)
>  %97 = fmul float %96, -2.000000e+00
>  %98 = fadd float %97, 3.000000e+00
>  %99 = fmul float %96, %96
>  %100 = fmul float %98, %99
>  %101 = fadd float %100, 1.000000e+00
>  %102 = bitcast float %46 to i32
>  %103 = bitcast float %47 to i32
>  %104 = insertelement <2 x i32> undef, i32 %102, i32 0
>  %105 = insertelement <2 x i32> %104, i32 %103, i32 1
>  %106 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %105, <8 x i32> %39, <4 x i32> %41, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %107 = extractelement <4 x float> %106, i32 0
>  %108 = extractelement <4 x float> %106, i32 1
>  %109 = extractelement <4 x float> %106, i32 2
>  %110 = fsub float %101, %107
>  %111 = call float @llvm.AMDIL.clamp.(float %110, float 0.000000e+00, float 1.000000e+00)
>  %112 = call float @llvm.maxnum.f32(float %108, float %109)
>  %113 = call float @llvm.maxnum.f32(float %112, float %111)
>  %114 = bitcast float %46 to i32
>  %115 = bitcast float %47 to i32
>  %116 = insertelement <2 x i32> undef, i32 %114, i32 0
>  %117 = insertelement <2 x i32> %116, i32 %115, i32 1
>  %118 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %117, <8 x i32> %35, <4 x i32> %37, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %119 = extractelement <4 x float> %118, i32 0
>  %120 = extractelement <4 x float> %118, i32 1
>  %121 = extractelement <4 x float> %118, i32 2
>  %122 = fsub float %119, %76
>  %123 = fmul float %113, %122
>  %124 = fadd float %123, %76
>  %125 = fadd float %120, -1.000000e+00
>  %126 = fadd float %121, 0.000000e+00
>  %127 = fmul float %121, 0.000000e+00
>  %128 = fadd float %127, 1.000000e+00
>  %129 = fmul float %111, %125
>  %130 = fadd float %129, 1.000000e+00
>  %131 = call i32 @llvm.SI.packf16(float %124, float %130)
>  %132 = bitcast i32 %131 to float
>  %133 = call i32 @llvm.SI.packf16(float %126, float %128)
>  %134 = bitcast i32 %133 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %132, float %134, float %132, float %134)
>  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.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #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 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, s10                                                       ; BEFC030A
>	v_mov_b32_e32 v2, 0x41200000                                            ; 7E0402FF 41200000
>	v_mov_b32_e32 v3, 0x40400000                                            ; 7E0602FF 40400000
>	v_mov_b32_e32 v4, 0x40c00000                                            ; 7E0802FF 40C00000
>	v_interp_p1_f32 v5, v0, 0, 0, [m0]                                      ; C8140000
>	v_interp_p2_f32 v5, [v5], v1, 0, 0, [m0]                                ; C8150001
>	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, 2, 0, [m0]                                      ; C81C0200
>	v_interp_p2_f32 v7, [v7], v1, 2, 0, [m0]                                ; C81D0201
>	v_interp_p1_f32 v8, v0, 3, 0, [m0]                                      ; C8200300
>	v_interp_p2_f32 v8, [v8], v1, 3, 0, [m0]                                ; C8210301
>	v_interp_p1_f32 v9, v0, 0, 1, [m0]                                      ; C8240400
>	v_interp_p2_f32 v9, [v9], v1, 0, 1, [m0]                                ; C8250401
>	v_interp_p1_f32 v10, v0, 1, 1, [m0]                                     ; C8280500
>	v_interp_p2_f32 v10, [v10], v1, 1, 1, [m0]                              ; C8290501
>	s_load_dwordx4 s[28:31], s[2:3], 0x0                                    ; C08E0300
>	s_load_dwordx4 s[32:35], s[4:5], 0x0                                    ; C0900500
>	s_load_dwordx4 s[36:39], s[4:5], 0x4                                    ; C0920504
>	s_load_dwordx4 s[0:3], s[4:5], 0x8                                      ; C0800508
>	s_load_dwordx4 s[16:19], s[4:5], 0xc                                    ; C088050C
>	s_load_dwordx4 s[40:43], s[4:5], 0x10                                   ; C0940510
>	s_load_dwordx8 s[44:51], s[6:7], 0x0                                    ; C0D60700
>	s_load_dwordx8 s[52:59], s[6:7], 0x8                                    ; C0DA0708
>	s_load_dwordx8 s[8:15], s[6:7], 0x10                                    ; C0C40710
>	s_load_dwordx8 s[20:27], s[6:7], 0x18                                   ; C0CA0718
>	s_waitcnt lgkmcnt(0)                                                    ; BF8C007F
>	s_buffer_load_dword s4, s[28:31], 0xf                                   ; C2021D0F
>	s_load_dwordx8 s[60:67], s[6:7], 0x20                                   ; C0DE0720
>	image_sample v[0:1], 9, 0, 0, 0, 0, 0, 0, 0, v[5:6], s[44:51], s[32:35] ; F0800900 010B0005
>	image_sample v9, 2, 0, 0, 0, 0, 0, 0, 0, v[9:10], s[52:59], s[36:39]    ; F0800200 012D0909
>	s_waitcnt vmcnt(0)                                                      ; BF8C0770
>	v_mac_f32_e32 v1, v0, v9                                                ; 3E021300
>	s_waitcnt lgkmcnt(0)                                                    ; BF8C007F
>	v_mad_f32 v0, v4, s4, 1.0                                               ; D2820000 03C80904
>	image_sample v4, 8, 0, 0, 0, 0, 0, 0, 0, v[7:8], s[60:67], s[40:43]     ; F0800800 014F0407
>	s_waitcnt vmcnt(0)                                                      ; BF8C0770
>	v_add_f32_e32 v7, -0.5, v4                                              ; 060E08F1
>	v_add_f32_e32 v8, v4, v4                                                ; 06100904
>	v_add_f32_e64 v8, 0, v8 clamp                                           ; D2060808 00021080
>	v_mul_f32_e32 v7, v2, v7                                                ; 100E0F02
>	v_madmk_f32_e32 v2, v4, v2, 0xc1200000                                  ; 40040504 C1200000
>	v_add_f32_e64 v4, 0, v7 clamp                                           ; D2060804 00020E80
>	v_add_f32_e64 v2, 0, v2 clamp                                           ; D2060802 00020480
>	v_mad_f32 v7, -2.0, v4, v3                                              ; D2820007 040E08F5
>	v_mul_f32_e32 v4, v4, v4                                                ; 10080904
>	v_mul_f32_e32 v4, v7, v4                                                ; 10080907
>	v_mad_f32 v7, -2.0, v2, v3                                              ; D2820007 040E04F5
>	v_mul_f32_e32 v2, v2, v2                                                ; 10040502
>	v_mul_f32_e32 v2, v7, v2                                                ; 10040507
>	v_madak_f32_e32 v2, v2, v4, 0xbf7d70a4                                  ; 42040902 BF7D70A4
>	v_mac_f32_e32 v2, v0, v1                                                ; 3E040300
>	v_mul_f32_e32 v0, 0x42c80000, v2                                        ; 100004FF 42C80000
>	v_add_f32_e64 v0, 0, v0 clamp                                           ; D2060800 00020080
>	v_mac_f32_e32 v3, -2.0, v0                                              ; 3E0600F5
>	v_mul_f32_e32 v0, v0, v0                                                ; 10000100
>	v_mad_f32 v0, v3, v0, 1.0                                               ; D2820000 03CA0103
>	image_sample v[1:3], 7, 0, 0, 0, 0, 0, 0, 0, v[5:6], s[20:27], s[16:19] ; F0800700 00850105
>	s_waitcnt vmcnt(0)                                                      ; BF8C0770
>	v_subrev_f32_e32 v0, v1, v0                                             ; 0A000101
>	v_add_f32_e64 v0, 0, v0 clamp                                           ; D2060800 00020080
>	v_max3_f32 v1, v2, v3, v0                                               ; D2A80001 04020702
>	image_sample v[2:4], 7, 0, 0, 0, 0, 0, 0, 0, v[5:6], s[8:15], s[0:3]    ; F0800700 00020205
>	s_waitcnt vmcnt(0)                                                      ; BF8C0770
>	v_subrev_f32_e32 v2, v8, v2                                             ; 0A040508
>	v_mac_f32_e32 v8, v2, v1                                                ; 3E100302
>	v_add_f32_e32 v1, -1.0, v3                                              ; 060206F3
>	v_add_f32_e32 v2, 0, v4                                                 ; 06040880
>	v_mad_f32 v3, 0, v4, 1.0                                                ; D2820003 03CA0880
>	v_mad_f32 v0, v0, v1, 1.0                                               ; D2820000 03CA0300
>	v_cvt_pkrtz_f16_f32_e32 v0, v8, v0                                      ; 5E000108
>	v_cvt_pkrtz_f16_f32_e32 v1, v2, v3                                      ; 5E020702
>	exp 15, 0, 1, 1, 1, v0, v1, v0, v1                                      ; F8001C0F 01000100
>	s_endpgm                                                                ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 72
>VGPRS: 12
>Code Size: 396 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
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL OUT[1], CLIPVERTEX
>DCL OUT[2], GENERIC[0]
>DCL OUT[3], GENERIC[1]
>DCL CONST[0..54]
>DCL TEMP[0..3], LOCAL
>  0: MUL TEMP[0].xy, CONST[48].xyyy, IN[1].xyyy
>  1: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
>  2: ADD TEMP[1].x, TEMP[0].xxxx, CONST[48].wwww
>  3: MOV TEMP[1].z, TEMP[1].xxxx
>  4: MUL TEMP[0].xy, CONST[49].xyyy, IN[1].xyyy
>  5: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
>  6: ADD TEMP[2].x, TEMP[0].xxxx, CONST[49].wwww
>  7: MOV TEMP[1].w, TEMP[2].xxxx
>  8: MUL TEMP[0].xy, CONST[50].xyyy, IN[1].xyyy
>  9: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 10: ADD TEMP[2].x, TEMP[0].xxxx, CONST[50].wwww
> 11: MUL TEMP[0].xy, CONST[51].xyyy, IN[1].xyyy
> 12: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 13: ADD TEMP[3].x, TEMP[0].xxxx, CONST[51].wwww
> 14: MOV TEMP[2].y, TEMP[3].xxxx
> 15: MUL TEMP[0].xy, CONST[52].xyyy, IN[1].xyyy
> 16: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 17: ADD TEMP[3].x, TEMP[0].xxxx, CONST[52].wwww
> 18: MOV TEMP[2].z, TEMP[3].xxxx
> 19: MUL TEMP[0].xy, CONST[53].xyyy, IN[1].xyyy
> 20: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 21: ADD TEMP[0].x, TEMP[0].xxxx, CONST[53].wwww
> 22: MOV TEMP[2].w, TEMP[0].xxxx
> 23: MOV TEMP[0].xw, IN[0].xxxw
> 24: MOV TEMP[1].xy, IN[1].xyxx
> 25: MAD TEMP[3].x, IN[0].zzzz, CONST[0].zzzz, -IN[0].wwww
> 26: MOV TEMP[0].z, TEMP[3].xxxx
> 27: MOV TEMP[0].y, -IN[0].yyyy
> 28: MAD TEMP[0].xy, CONST[54].xyyy, IN[0].wwww, TEMP[0].xyyy
> 29: MOV OUT[2], TEMP[1]
> 30: MOV OUT[3], TEMP[2]
> 31: MOV OUT[0], TEMP[0]
> 32: MOV OUT[1], IN[0]
> 33: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = call float @llvm.SI.load.const(<16 x i8> %13, i32 8)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %13, i32 768)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %13, i32 772)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %13, i32 780)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %13, i32 784)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %13, i32 788)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %13, i32 796)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %13, i32 800)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %13, i32 804)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %13, i32 812)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %13, i32 816)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %13, i32 820)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %13, i32 828)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %13, i32 832)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %13, i32 836)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %13, i32 844)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %13, i32 848)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %13, i32 852)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %13, i32 860)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %13, i32 864)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %13, i32 868)
>  %35 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %36 = load <16 x i8>, <16 x i8> addrspace(2)* %35, align 16, !tbaa !0
>  %37 = add i32 %5, %8
>  %38 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %36, i32 0, i32 %37)
>  %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 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 1
>  %44 = load <16 x i8>, <16 x i8> addrspace(2)* %43, align 16, !tbaa !0
>  %45 = add i32 %5, %8
>  %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 = fmul float %15, %47
>  %50 = fmul float %16, %48
>  %51 = fadd float %50, %49
>  %52 = fadd float %51, %17
>  %53 = fmul float %18, %47
>  %54 = fmul float %19, %48
>  %55 = fadd float %54, %53
>  %56 = fadd float %55, %20
>  %57 = fmul float %21, %47
>  %58 = fmul float %22, %48
>  %59 = fadd float %58, %57
>  %60 = fadd float %59, %23
>  %61 = fmul float %24, %47
>  %62 = fmul float %25, %48
>  %63 = fadd float %62, %61
>  %64 = fadd float %63, %26
>  %65 = fmul float %27, %47
>  %66 = fmul float %28, %48
>  %67 = fadd float %66, %65
>  %68 = fadd float %67, %29
>  %69 = fmul float %30, %47
>  %70 = fmul float %31, %48
>  %71 = fadd float %70, %69
>  %72 = fadd float %71, %32
>  %73 = fmul float %41, %14
>  %74 = fsub float %73, %42
>  %75 = fmul float %33, %42
>  %76 = fadd float %75, %39
>  %77 = fmul float %34, %42
>  %78 = fsub float %77, %40
>  %79 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 16
>  %80 = load <16 x i8>, <16 x i8> addrspace(2)* %79, align 16, !tbaa !0
>  %81 = call float @llvm.SI.load.const(<16 x i8> %80, i32 0)
>  %82 = fmul float %81, %39
>  %83 = call float @llvm.SI.load.const(<16 x i8> %80, i32 4)
>  %84 = fmul float %83, %40
>  %85 = fadd float %82, %84
>  %86 = call float @llvm.SI.load.const(<16 x i8> %80, i32 8)
>  %87 = fmul float %86, %41
>  %88 = fadd float %85, %87
>  %89 = call float @llvm.SI.load.const(<16 x i8> %80, i32 12)
>  %90 = fmul float %89, %42
>  %91 = fadd float %88, %90
>  %92 = call float @llvm.SI.load.const(<16 x i8> %80, i32 16)
>  %93 = fmul float %92, %39
>  %94 = call float @llvm.SI.load.const(<16 x i8> %80, i32 20)
>  %95 = fmul float %94, %40
>  %96 = fadd float %93, %95
>  %97 = call float @llvm.SI.load.const(<16 x i8> %80, i32 24)
>  %98 = fmul float %97, %41
>  %99 = fadd float %96, %98
>  %100 = call float @llvm.SI.load.const(<16 x i8> %80, i32 28)
>  %101 = fmul float %100, %42
>  %102 = fadd float %99, %101
>  %103 = call float @llvm.SI.load.const(<16 x i8> %80, i32 32)
>  %104 = fmul float %103, %39
>  %105 = call float @llvm.SI.load.const(<16 x i8> %80, i32 36)
>  %106 = fmul float %105, %40
>  %107 = fadd float %104, %106
>  %108 = call float @llvm.SI.load.const(<16 x i8> %80, i32 40)
>  %109 = fmul float %108, %41
>  %110 = fadd float %107, %109
>  %111 = call float @llvm.SI.load.const(<16 x i8> %80, i32 44)
>  %112 = fmul float %111, %42
>  %113 = fadd float %110, %112
>  %114 = call float @llvm.SI.load.const(<16 x i8> %80, i32 48)
>  %115 = fmul float %114, %39
>  %116 = call float @llvm.SI.load.const(<16 x i8> %80, i32 52)
>  %117 = fmul float %116, %40
>  %118 = fadd float %115, %117
>  %119 = call float @llvm.SI.load.const(<16 x i8> %80, i32 56)
>  %120 = fmul float %119, %41
>  %121 = fadd float %118, %120
>  %122 = call float @llvm.SI.load.const(<16 x i8> %80, i32 60)
>  %123 = fmul float %122, %42
>  %124 = fadd float %121, %123
>  %125 = call float @llvm.SI.load.const(<16 x i8> %80, i32 64)
>  %126 = fmul float %125, %39
>  %127 = call float @llvm.SI.load.const(<16 x i8> %80, i32 68)
>  %128 = fmul float %127, %40
>  %129 = fadd float %126, %128
>  %130 = call float @llvm.SI.load.const(<16 x i8> %80, i32 72)
>  %131 = fmul float %130, %41
>  %132 = fadd float %129, %131
>  %133 = call float @llvm.SI.load.const(<16 x i8> %80, i32 76)
>  %134 = fmul float %133, %42
>  %135 = fadd float %132, %134
>  %136 = call float @llvm.SI.load.const(<16 x i8> %80, i32 80)
>  %137 = fmul float %136, %39
>  %138 = call float @llvm.SI.load.const(<16 x i8> %80, i32 84)
>  %139 = fmul float %138, %40
>  %140 = fadd float %137, %139
>  %141 = call float @llvm.SI.load.const(<16 x i8> %80, i32 88)
>  %142 = fmul float %141, %41
>  %143 = fadd float %140, %142
>  %144 = call float @llvm.SI.load.const(<16 x i8> %80, i32 92)
>  %145 = fmul float %144, %42
>  %146 = fadd float %143, %145
>  %147 = call float @llvm.SI.load.const(<16 x i8> %80, i32 96)
>  %148 = fmul float %147, %39
>  %149 = call float @llvm.SI.load.const(<16 x i8> %80, i32 100)
>  %150 = fmul float %149, %40
>  %151 = fadd float %148, %150
>  %152 = call float @llvm.SI.load.const(<16 x i8> %80, i32 104)
>  %153 = fmul float %152, %41
>  %154 = fadd float %151, %153
>  %155 = call float @llvm.SI.load.const(<16 x i8> %80, i32 108)
>  %156 = fmul float %155, %42
>  %157 = fadd float %154, %156
>  %158 = call float @llvm.SI.load.const(<16 x i8> %80, i32 112)
>  %159 = fmul float %158, %39
>  %160 = call float @llvm.SI.load.const(<16 x i8> %80, i32 116)
>  %161 = fmul float %160, %40
>  %162 = fadd float %159, %161
>  %163 = call float @llvm.SI.load.const(<16 x i8> %80, i32 120)
>  %164 = fmul float %163, %41
>  %165 = fadd float %162, %164
>  %166 = call float @llvm.SI.load.const(<16 x i8> %80, i32 124)
>  %167 = fmul float %166, %42
>  %168 = fadd float %165, %167
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %47, float %48, float %52, float %56)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %60, float %64, float %68, float %72)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 12, i32 0, float %76, float %78, float %74, float %42)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 13, i32 0, float %91, float %102, float %113, float %124)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 14, i32 0, float %135, float %146, float %157, 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
>
>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[4:7], s[8:9], 0x0                    ; C0820900
>	s_load_dwordx4 s[12:15], s[8:9], 0x4                  ; C0860904
>	v_add_i32_e32 v0, vcc, s10, v0                        ; 4A00000A
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[4:7], 0 idxen   ; E00C2000 80010100
>	buffer_load_format_xyzw v[5:8], v0, s[12:15], 0 idxen ; E00C2000 80030500
>	s_load_dwordx4 s[4:7], s[2:3], 0x0                    ; C0820300
>	s_load_dwordx4 s[0:3], s[2:3], 0x40                   ; C0800340
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	s_buffer_load_dword s8, s[4:7], 0x2                   ; C2040502
>	s_buffer_load_dword s9, s[4:7], 0xc0                  ; C20485C0
>	s_buffer_load_dword s10, s[4:7], 0xc1                 ; C20505C1
>	s_buffer_load_dword s11, s[4:7], 0xc3                 ; C20585C3
>	s_buffer_load_dword s12, s[4:7], 0xc4                 ; C20605C4
>	s_buffer_load_dword s13, s[4:7], 0xc5                 ; C20685C5
>	s_buffer_load_dword s14, s[4:7], 0xc7                 ; C20705C7
>	s_buffer_load_dword s15, s[4:7], 0xc8                 ; C20785C8
>	s_buffer_load_dword s16, s[4:7], 0xc9                 ; C20805C9
>	s_buffer_load_dword s17, s[4:7], 0xcb                 ; C20885CB
>	s_buffer_load_dword s18, s[4:7], 0xcc                 ; C20905CC
>	s_buffer_load_dword s19, s[4:7], 0xcd                 ; C20985CD
>	s_buffer_load_dword s20, s[4:7], 0xcf                 ; C20A05CF
>	s_buffer_load_dword s21, s[4:7], 0xd0                 ; C20A85D0
>	s_buffer_load_dword s22, s[4:7], 0xd1                 ; C20B05D1
>	s_buffer_load_dword s23, s[4:7], 0xd3                 ; C20B85D3
>	s_buffer_load_dword s24, s[4:7], 0xd4                 ; C20C05D4
>	s_buffer_load_dword s25, s[4:7], 0xd5                 ; C20C85D5
>	s_buffer_load_dword s26, s[4:7], 0xd7                 ; C20D05D7
>	s_buffer_load_dword s27, s[4:7], 0xd8                 ; C20D85D8
>	s_buffer_load_dword s4, s[4:7], 0xd9                  ; C20205D9
>	s_buffer_load_dword s5, s[0:3], 0x0                   ; C2028100
>	s_buffer_load_dword s6, s[0:3], 0x1                   ; C2030101
>	s_buffer_load_dword s7, s[0:3], 0x2                   ; C2038102
>	s_buffer_load_dword s28, s[0:3], 0x3                  ; C20E0103
>	s_buffer_load_dword s29, s[0:3], 0x4                  ; C20E8104
>	s_buffer_load_dword s30, s[0:3], 0x5                  ; C20F0105
>	s_buffer_load_dword s31, s[0:3], 0x6                  ; C20F8106
>	s_buffer_load_dword s32, s[0:3], 0x7                  ; C2100107
>	s_buffer_load_dword s33, s[0:3], 0x8                  ; C2108108
>	s_buffer_load_dword s34, s[0:3], 0x9                  ; C2110109
>	s_buffer_load_dword s35, s[0:3], 0xa                  ; C211810A
>	s_buffer_load_dword s36, s[0:3], 0xb                  ; C212010B
>	s_buffer_load_dword s37, s[0:3], 0xc                  ; C212810C
>	s_buffer_load_dword s38, s[0:3], 0xd                  ; C213010D
>	s_buffer_load_dword s39, s[0:3], 0xe                  ; C213810E
>	s_buffer_load_dword s40, s[0:3], 0xf                  ; C214010F
>	s_buffer_load_dword s41, s[0:3], 0x10                 ; C2148110
>	s_buffer_load_dword s42, s[0:3], 0x11                 ; C2150111
>	s_buffer_load_dword s43, s[0:3], 0x12                 ; C2158112
>	s_buffer_load_dword s44, s[0:3], 0x13                 ; C2160113
>	s_buffer_load_dword s45, s[0:3], 0x14                 ; C2168114
>	s_buffer_load_dword s46, s[0:3], 0x15                 ; C2170115
>	s_buffer_load_dword s47, s[0:3], 0x16                 ; C2178116
>	s_buffer_load_dword s48, s[0:3], 0x17                 ; C2180117
>	s_buffer_load_dword s49, s[0:3], 0x18                 ; C2188118
>	s_buffer_load_dword s50, s[0:3], 0x19                 ; C2190119
>	s_buffer_load_dword s51, s[0:3], 0x1a                 ; C219811A
>	s_buffer_load_dword s52, s[0:3], 0x1b                 ; C21A011B
>	s_buffer_load_dword s53, s[0:3], 0x1c                 ; C21A811C
>	s_buffer_load_dword s54, s[0:3], 0x1d                 ; C21B011D
>	s_buffer_load_dword s55, s[0:3], 0x1e                 ; C21B811E
>	s_buffer_load_dword s0, s[0:3], 0x1f                  ; C200011F
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mov_b32_e32 v0, s11                                 ; 7E00020B
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mov_b32_e32 v7, s14                                 ; 7E0E020E
>	v_mov_b32_e32 v8, s17                                 ; 7E100211
>	v_mov_b32_e32 v9, s20                                 ; 7E120214
>	v_mov_b32_e32 v10, s23                                ; 7E140217
>	v_mov_b32_e32 v11, s26                                ; 7E16021A
>	v_mac_f32_e32 v0, s9, v5                              ; 3E000A09
>	v_mac_f32_e32 v7, s12, v5                             ; 3E0E0A0C
>	v_mac_f32_e32 v8, s15, v5                             ; 3E100A0F
>	v_mac_f32_e32 v9, s18, v5                             ; 3E120A12
>	v_mac_f32_e32 v10, s21, v5                            ; 3E140A15
>	v_mac_f32_e32 v11, s24, v5                            ; 3E160A18
>	v_mac_f32_e32 v0, s10, v6                             ; 3E000C0A
>	v_mac_f32_e32 v7, s13, v6                             ; 3E0E0C0D
>	v_mac_f32_e32 v8, s16, v6                             ; 3E100C10
>	v_mac_f32_e32 v9, s19, v6                             ; 3E120C13
>	exp 15, 32, 0, 0, 0, v5, v6, v0, v7                   ; F800020F 07000605
>	v_mac_f32_e32 v10, s22, v6                            ; 3E140C16
>	v_mac_f32_e32 v11, s25, v6                            ; 3E160C19
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mad_f32 v0, v3, s8, -v4                             ; D2820000 84101103
>	v_mad_f32 v5, s27, v4, v1                             ; D2820005 0406081B
>	v_mad_f32 v6, s4, v4, -v2                             ; D2820006 840A0804
>	v_mul_f32_e32 v7, s6, v2                              ; 100E0406
>	exp 15, 33, 0, 0, 0, v8, v9, v10, v11                 ; F800021F 0B0A0908
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mul_f32_e32 v8, s30, v2                             ; 1010041E
>	v_mac_f32_e32 v7, s5, v1                              ; 3E0E0205
>	exp 15, 12, 0, 0, 0, v5, v6, v0, v4                   ; F80000CF 04000605
>	v_mac_f32_e32 v8, s29, v1                             ; 3E10021D
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mul_f32_e32 v0, s34, v2                             ; 10000422
>	v_mac_f32_e32 v0, s33, v1                             ; 3E000221
>	v_mul_f32_e32 v5, s38, v2                             ; 100A0426
>	v_mac_f32_e32 v5, s37, v1                             ; 3E0A0225
>	v_mul_f32_e32 v6, s42, v2                             ; 100C042A
>	v_mul_f32_e32 v9, s46, v2                             ; 1012042E
>	v_mul_f32_e32 v10, s50, v2                            ; 10140432
>	v_mul_f32_e32 v2, s54, v2                             ; 10040436
>	v_mac_f32_e32 v6, s41, v1                             ; 3E0C0229
>	v_mac_f32_e32 v9, s45, v1                             ; 3E12022D
>	v_mac_f32_e32 v10, s49, v1                            ; 3E140231
>	v_mac_f32_e32 v2, s53, v1                             ; 3E040235
>	v_mac_f32_e32 v7, s7, v3                              ; 3E0E0607
>	v_mac_f32_e32 v8, s31, v3                             ; 3E10061F
>	v_mac_f32_e32 v0, s35, v3                             ; 3E000623
>	v_mac_f32_e32 v5, s39, v3                             ; 3E0A0627
>	v_mac_f32_e32 v6, s43, v3                             ; 3E0C062B
>	v_mac_f32_e32 v9, s47, v3                             ; 3E12062F
>	v_mac_f32_e32 v10, s51, v3                            ; 3E140633
>	v_mac_f32_e32 v2, s55, v3                             ; 3E040637
>	v_mac_f32_e32 v7, s28, v4                             ; 3E0E081C
>	v_mac_f32_e32 v8, s32, v4                             ; 3E100820
>	v_mac_f32_e32 v0, s36, v4                             ; 3E000824
>	v_mac_f32_e32 v5, s40, v4                             ; 3E0A0828
>	v_mac_f32_e32 v6, s44, v4                             ; 3E0C082C
>	v_mac_f32_e32 v9, s48, v4                             ; 3E120830
>	v_mac_f32_e32 v10, s52, v4                            ; 3E140834
>	v_mac_f32_e32 v2, s0, v4                              ; 3E040800
>	exp 15, 13, 0, 0, 0, v7, v8, v0, v5                   ; F80000DF 05000807
>	exp 15, 14, 0, 1, 0, v6, v9, v10, v2                  ; F80008EF 020A0906
>	s_endpgm                                              ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 64
>VGPRS: 12
>Code Size: 544 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
>  clamp_color = 0
>FRAG
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL IN[1], GENERIC[1], 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 SVIEW[0], 2D, FLOAT
>DCL SVIEW[1], 2D, FLOAT
>DCL SVIEW[2], 2D, FLOAT
>DCL SVIEW[3], 2D, FLOAT
>DCL SVIEW[4], 2D, FLOAT
>DCL SVIEW[5], 2D, FLOAT
>DCL SVIEW[6], 2D, FLOAT
>DCL CONST[0..3]
>DCL TEMP[0..7], LOCAL
>IMM[0] FLT32 {   -0.5000,    -1.0000,    10.0000,   -10.0000}
>IMM[1] FLT32 {   -2.0000,     3.0000,     1.0000,     6.0000}
>IMM[2] FLT32 {   -0.5800,   100.0000,     0.0500,     0.7500}
>IMM[3] FLT32 {    0.2500,     0.3000,     0.5900,     0.1100}
>IMM[4] FLT32 {   -0.0800,    14.2857,     0.0050,     2.0000}
>  0: ADD TEMP[0].xyz, -CONST[0].xyzz, CONST[1].xyzz
>  1: MOV TEMP[1].xy, IN[0].zwww
>  2: TEX TEMP[1], TEMP[1], SAMP[6], 2D
>  3: MAD TEMP[0].xyz, TEMP[1].xxxx, TEMP[0].xyzz, CONST[0].xyzz
>  4: LRP TEMP[2].xyz, TEMP[1].yyyy, CONST[2].xyzz, TEMP[0].xyzz
>  5: MOV TEMP[0].x, CONST[0].wwww
>  6: MOV TEMP[0].y, CONST[1].wwww
>  7: MOV TEMP[0].z, CONST[2].wwww
>  8: LRP TEMP[3].xyz, TEMP[1].zzzz, TEMP[0].xyzz, TEMP[2].xyzz
>  9: ADD TEMP[1].xy, TEMP[1].wwww, IMM[0].xyyy
> 10: MUL TEMP[1].xy, TEMP[1].xyyy, IMM[0].zwww
> 11: MOV_SAT TEMP[1].xy, TEMP[1].xyyy
> 12: MOV TEMP[4].xy, IN[0].xyyy
> 13: TEX TEMP[4].xyz, TEMP[4], SAMP[4], 2D
> 14: LRP TEMP[5].xyz, TEMP[4].yyyy, CONST[2].xyzz, TEMP[3].xyzz
> 15: LRP TEMP[3].xyz, TEMP[4].zzzz, TEMP[0].xyzz, TEMP[5].xyzz
> 16: MAD TEMP[0].xy, TEMP[1].xyyy, IMM[1].xxxx, IMM[1].yyyy
> 17: MUL TEMP[1].xy, TEMP[1].xyyy, TEMP[1].xyyy
> 18: MUL TEMP[0].xy, TEMP[1].xyyy, TEMP[0].xyyy
> 19: MUL TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 20: MAD TEMP[1].x, CONST[3].wwww, IMM[1].wwww, IMM[1].zzzz
> 21: MOV TEMP[6].xy, IN[0].xyyy
> 22: TEX TEMP[6], TEMP[6], SAMP[0], 2D
> 23: MOV TEMP[7].xy, IN[1].xyyy
> 24: TEX TEMP[7].y, TEMP[7], SAMP[1], 2D
> 25: MAD TEMP[7].x, TEMP[7].yyyy, TEMP[6].xxxx, TEMP[6].wwww
> 26: MAD TEMP[0].x, TEMP[7].xxxx, TEMP[1].xxxx, TEMP[0].xxxx
> 27: ADD TEMP[0].x, TEMP[0].xxxx, IMM[2].xxxx
> 28: MUL TEMP[0].x, TEMP[0].xxxx, IMM[0].zzzz
> 29: MOV_SAT TEMP[1].x, TEMP[0].xxxx
> 30: MAD TEMP[7].x, TEMP[1].xxxx, IMM[1].xxxx, IMM[1].yyyy
> 31: MUL TEMP[0].x, TEMP[1].xxxx, TEMP[1].xxxx
> 32: MUL TEMP[1].x, TEMP[0].xxxx, TEMP[7].xxxx
> 33: MAD TEMP[0].x, TEMP[7].xxxx, TEMP[0].xxxx, IMM[1].zzzz
> 34: ADD TEMP[0].x, -TEMP[4].xxxx, TEMP[0].xxxx
> 35: MOV_SAT TEMP[0].x, TEMP[0].xxxx
> 36: MUL TEMP[1].x, TEMP[1].xxxx, IMM[2].yyyy
> 37: MOV_SAT TEMP[1].x, TEMP[1].xxxx
> 38: MAD TEMP[4].x, TEMP[1].xxxx, IMM[1].xxxx, IMM[1].yyyy
> 39: MUL TEMP[1].x, TEMP[1].xxxx, TEMP[1].xxxx
> 40: MUL TEMP[1].x, TEMP[1].xxxx, TEMP[4].xxxx
> 41: LRP TEMP[2].xyz, TEMP[1].xxxx, IMM[2].zzzz, TEMP[3].xyzz
> 42: MUL TEMP[3].x, IMM[2].wwww, CONST[3].wwww
> 43: ADD TEMP[4].x, -TEMP[6].xxxx, IMM[1].zzzz
> 44: MUL TEMP[4].x, TEMP[4].xxxx, TEMP[4].xxxx
> 45: MUL TEMP[4].x, TEMP[4].xxxx, TEMP[4].xxxx
> 46: MAD TEMP[3].x, TEMP[4].xxxx, IMM[3].xxxx, TEMP[3].xxxx
> 47: MOV TEMP[4].xy, IN[1].zwww
> 48: TEX TEMP[4], TEMP[4], SAMP[5], 2D
> 49: ADD TEMP[4], TEMP[4], IMM[0].yyyy
> 50: MAD TEMP[5], TEMP[3].xxxx, TEMP[4], IMM[1].zzzz
> 51: LRP TEMP[3].xyz, TEMP[1].xxxx, IMM[1].zzzz, TEMP[5].xyzz
> 52: MUL TEMP[3].xyz, TEMP[3].xyzz, TEMP[2].xyzz
> 53: DP3 TEMP[4].x, TEMP[3].xyzz, IMM[3].yzww
> 54: ADD TEMP[4].x, TEMP[4].xxxx, IMM[4].xxxx
> 55: MUL TEMP[4].x, TEMP[4].xxxx, IMM[4].yyyy
> 56: MOV_SAT TEMP[4].x, TEMP[4].xxxx
> 57: MAD TEMP[7].x, TEMP[4].xxxx, IMM[1].xxxx, IMM[1].yyyy
> 58: MUL TEMP[4].x, TEMP[4].xxxx, TEMP[4].xxxx
> 59: MAD TEMP[4].x, TEMP[7].xxxx, -TEMP[4].xxxx, IMM[1].zzzz
> 60: MAD TEMP[2].xyz, TEMP[4].xxxx, IMM[4].zzzz, TEMP[3].xyzz
> 61: MUL TEMP[2].xyz, TEMP[2].xyzz, TEMP[6].zzzz
> 62: MAD TEMP[3].xyz, TEMP[2].xyzz, IMM[4].wwww, TEMP[3].xyzz
> 63: MOV_SAT TEMP[3].xyz, TEMP[3].xyzz
> 64: MUL TEMP[2].xyz, TEMP[6].yyyy, TEMP[3].xyzz
> 65: MOV TEMP[4].xy, IN[0].xyyy
> 66: TEX TEMP[4], TEMP[4], SAMP[3], 2D
> 67: MAD TEMP[3].xyz, TEMP[3].xyzz, -TEMP[6].yyyy, TEMP[4].xyzz
> 68: MUL TEMP[6].x, TEMP[6].yyyy, CONST[3].zzzz
> 69: MUL TEMP[5].x, TEMP[5].wwww, TEMP[6].xxxx
> 70: MAD TEMP[3].xyz, TEMP[0].xxxx, TEMP[3].xyzz, TEMP[2].xyzz
> 71: LRP TEMP[2].x, TEMP[1].xxxx, IMM[1].zzzz, CONST[3].zzzz
> 72: MUL TEMP[1].x, TEMP[5].xxxx, TEMP[2].xxxx
> 73: MOV TEMP[2].xy, IN[0].xyyy
> 74: TEX TEMP[2].y, TEMP[2], SAMP[2], 2D
> 75: MUL TEMP[2].x, TEMP[0].xxxx, TEMP[2].yyyy
> 76: ADD TEMP[5].x, IMM[0].yyyy, CONST[3].xxxx
> 77: MAD TEMP[2].x, TEMP[2].xxxx, TEMP[5].xxxx, IMM[1].zzzz
> 78: MAD TEMP[2].x, TEMP[4].wwww, TEMP[2].xxxx, -TEMP[1].xxxx
> 79: MAD TEMP[0].x, TEMP[0].xxxx, TEMP[2].xxxx, TEMP[1].xxxx
> 80: MOV TEMP[3].w, TEMP[0].xxxx
> 81: MOV OUT[0], TEMP[3]
> 82: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %24 = load <16 x i8>, <16 x i8> addrspace(2)* %23, align 16, !tbaa !0
>  %25 = call float @llvm.SI.load.const(<16 x i8> %24, i32 0)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %24, i32 4)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %24, i32 8)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %24, i32 12)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %24, i32 16)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %24, i32 20)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %24, i32 24)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %24, i32 28)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %24, i32 32)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %24, i32 36)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %24, i32 40)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %24, i32 44)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %24, i32 48)
>  %38 = call float @llvm.SI.load.const(<16 x i8> %24, i32 56)
>  %39 = call float @llvm.SI.load.const(<16 x i8> %24, i32 60)
>  %40 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %41 = load <8 x i32>, <8 x i32> addrspace(2)* %40, align 32, !tbaa !0
>  %42 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %43 = load <4 x i32>, <4 x i32> addrspace(2)* %42, align 16, !tbaa !0
>  %44 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 1
>  %45 = load <8 x i32>, <8 x i32> addrspace(2)* %44, align 32, !tbaa !0
>  %46 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 1
>  %47 = load <4 x i32>, <4 x i32> addrspace(2)* %46, align 16, !tbaa !0
>  %48 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 2
>  %49 = load <8 x i32>, <8 x i32> addrspace(2)* %48, align 32, !tbaa !0
>  %50 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 2
>  %51 = load <4 x i32>, <4 x i32> addrspace(2)* %50, align 16, !tbaa !0
>  %52 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 3
>  %53 = load <8 x i32>, <8 x i32> addrspace(2)* %52, align 32, !tbaa !0
>  %54 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 3
>  %55 = load <4 x i32>, <4 x i32> addrspace(2)* %54, align 16, !tbaa !0
>  %56 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 4
>  %57 = load <8 x i32>, <8 x i32> addrspace(2)* %56, align 32, !tbaa !0
>  %58 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 4
>  %59 = load <4 x i32>, <4 x i32> addrspace(2)* %58, align 16, !tbaa !0
>  %60 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 5
>  %61 = load <8 x i32>, <8 x i32> addrspace(2)* %60, align 32, !tbaa !0
>  %62 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 5
>  %63 = load <4 x i32>, <4 x i32> addrspace(2)* %62, align 16, !tbaa !0
>  %64 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 6
>  %65 = load <8 x i32>, <8 x i32> addrspace(2)* %64, align 32, !tbaa !0
>  %66 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 6
>  %67 = load <4 x i32>, <4 x i32> addrspace(2)* %66, align 16, !tbaa !0
>  %68 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %8)
>  %69 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %8)
>  %70 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %6, <2 x i32> %8)
>  %71 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %6, <2 x i32> %8)
>  %72 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %6, <2 x i32> %8)
>  %73 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %6, <2 x i32> %8)
>  %74 = call float @llvm.SI.fs.interp(i32 2, i32 1, i32 %6, <2 x i32> %8)
>  %75 = call float @llvm.SI.fs.interp(i32 3, i32 1, i32 %6, <2 x i32> %8)
>  %76 = fsub float %29, %25
>  %77 = fsub float %30, %26
>  %78 = fsub float %31, %27
>  %79 = bitcast float %70 to i32
>  %80 = bitcast float %71 to i32
>  %81 = insertelement <2 x i32> undef, i32 %79, i32 0
>  %82 = insertelement <2 x i32> %81, i32 %80, i32 1
>  %83 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %82, <8 x i32> %65, <4 x i32> %67, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %84 = extractelement <4 x float> %83, i32 0
>  %85 = extractelement <4 x float> %83, i32 1
>  %86 = extractelement <4 x float> %83, i32 2
>  %87 = extractelement <4 x float> %83, i32 3
>  %88 = fmul float %84, %76
>  %89 = fadd float %88, %25
>  %90 = fmul float %84, %77
>  %91 = fadd float %90, %26
>  %92 = fmul float %84, %78
>  %93 = fadd float %92, %27
>  %94 = fsub float 1.000000e+00, %85
>  %95 = fmul float %33, %85
>  %96 = fmul float %89, %94
>  %97 = fadd float %95, %96
>  %98 = fsub float 1.000000e+00, %85
>  %99 = fmul float %34, %85
>  %100 = fmul float %91, %98
>  %101 = fadd float %99, %100
>  %102 = fsub float 1.000000e+00, %85
>  %103 = fmul float %35, %85
>  %104 = fmul float %93, %102
>  %105 = fadd float %103, %104
>  %106 = fsub float 1.000000e+00, %86
>  %107 = fmul float %28, %86
>  %108 = fmul float %97, %106
>  %109 = fadd float %107, %108
>  %110 = fsub float 1.000000e+00, %86
>  %111 = fmul float %32, %86
>  %112 = fmul float %101, %110
>  %113 = fadd float %111, %112
>  %114 = fsub float 1.000000e+00, %86
>  %115 = fmul float %36, %86
>  %116 = fmul float %105, %114
>  %117 = fadd float %115, %116
>  %118 = fadd float %87, -5.000000e-01
>  %119 = fadd float %87, -1.000000e+00
>  %120 = fmul float %118, 1.000000e+01
>  %121 = fmul float %119, -1.000000e+01
>  %122 = call float @llvm.AMDIL.clamp.(float %120, float 0.000000e+00, float 1.000000e+00)
>  %123 = call float @llvm.AMDIL.clamp.(float %121, float 0.000000e+00, float 1.000000e+00)
>  %124 = bitcast float %68 to i32
>  %125 = bitcast float %69 to i32
>  %126 = insertelement <2 x i32> undef, i32 %124, i32 0
>  %127 = insertelement <2 x i32> %126, i32 %125, i32 1
>  %128 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %127, <8 x i32> %57, <4 x i32> %59, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %129 = extractelement <4 x float> %128, i32 0
>  %130 = extractelement <4 x float> %128, i32 1
>  %131 = extractelement <4 x float> %128, i32 2
>  %132 = fsub float 1.000000e+00, %130
>  %133 = fmul float %33, %130
>  %134 = fmul float %109, %132
>  %135 = fadd float %133, %134
>  %136 = fsub float 1.000000e+00, %130
>  %137 = fmul float %34, %130
>  %138 = fmul float %113, %136
>  %139 = fadd float %137, %138
>  %140 = fsub float 1.000000e+00, %130
>  %141 = fmul float %35, %130
>  %142 = fmul float %117, %140
>  %143 = fadd float %141, %142
>  %144 = fsub float 1.000000e+00, %131
>  %145 = fmul float %28, %131
>  %146 = fmul float %135, %144
>  %147 = fadd float %145, %146
>  %148 = fsub float 1.000000e+00, %131
>  %149 = fmul float %32, %131
>  %150 = fmul float %139, %148
>  %151 = fadd float %149, %150
>  %152 = fsub float 1.000000e+00, %131
>  %153 = fmul float %36, %131
>  %154 = fmul float %143, %152
>  %155 = fadd float %153, %154
>  %156 = fmul float %122, -2.000000e+00
>  %157 = fadd float %156, 3.000000e+00
>  %158 = fmul float %123, -2.000000e+00
>  %159 = fadd float %158, 3.000000e+00
>  %160 = fmul float %122, %122
>  %161 = fmul float %123, %123
>  %162 = fmul float %160, %157
>  %163 = fmul float %161, %159
>  %164 = fmul float %163, %162
>  %165 = fmul float %39, 6.000000e+00
>  %166 = fadd float %165, 1.000000e+00
>  %167 = bitcast float %68 to i32
>  %168 = bitcast float %69 to i32
>  %169 = insertelement <2 x i32> undef, i32 %167, i32 0
>  %170 = insertelement <2 x i32> %169, i32 %168, i32 1
>  %171 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %170, <8 x i32> %41, <4 x i32> %43, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %172 = extractelement <4 x float> %171, i32 0
>  %173 = extractelement <4 x float> %171, i32 1
>  %174 = extractelement <4 x float> %171, i32 2
>  %175 = extractelement <4 x float> %171, i32 3
>  %176 = bitcast float %72 to i32
>  %177 = bitcast float %73 to i32
>  %178 = insertelement <2 x i32> undef, i32 %176, i32 0
>  %179 = insertelement <2 x i32> %178, i32 %177, i32 1
>  %180 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %179, <8 x i32> %45, <4 x i32> %47, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %181 = extractelement <4 x float> %180, i32 1
>  %182 = fmul float %181, %172
>  %183 = fadd float %182, %175
>  %184 = fmul float %183, %166
>  %185 = fadd float %184, %164
>  %186 = fadd float %185, 0xBFE28F5C20000000
>  %187 = fmul float %186, 1.000000e+01
>  %188 = call float @llvm.AMDIL.clamp.(float %187, float 0.000000e+00, float 1.000000e+00)
>  %189 = fmul float %188, -2.000000e+00
>  %190 = fadd float %189, 3.000000e+00
>  %191 = fmul float %188, %188
>  %192 = fmul float %191, %190
>  %193 = fmul float %190, %191
>  %194 = fadd float %193, 1.000000e+00
>  %195 = fsub float %194, %129
>  %196 = call float @llvm.AMDIL.clamp.(float %195, float 0.000000e+00, float 1.000000e+00)
>  %197 = fmul float %192, 1.000000e+02
>  %198 = call float @llvm.AMDIL.clamp.(float %197, float 0.000000e+00, float 1.000000e+00)
>  %199 = fmul float %198, -2.000000e+00
>  %200 = fadd float %199, 3.000000e+00
>  %201 = fmul float %198, %198
>  %202 = fmul float %201, %200
>  %203 = fsub float 1.000000e+00, %202
>  %204 = fmul float %202, 0x3FA99999A0000000
>  %205 = fmul float %147, %203
>  %206 = fadd float %204, %205
>  %207 = fsub float 1.000000e+00, %202
>  %208 = fmul float %202, 0x3FA99999A0000000
>  %209 = fmul float %151, %207
>  %210 = fadd float %208, %209
>  %211 = fsub float 1.000000e+00, %202
>  %212 = fmul float %202, 0x3FA99999A0000000
>  %213 = fmul float %155, %211
>  %214 = fadd float %212, %213
>  %215 = fmul float %39, 7.500000e-01
>  %216 = fsub float 1.000000e+00, %172
>  %217 = fmul float %216, %216
>  %218 = fmul float %217, %217
>  %219 = fmul float %218, 2.500000e-01
>  %220 = fadd float %219, %215
>  %221 = bitcast float %74 to i32
>  %222 = bitcast float %75 to i32
>  %223 = insertelement <2 x i32> undef, i32 %221, i32 0
>  %224 = insertelement <2 x i32> %223, i32 %222, i32 1
>  %225 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %224, <8 x i32> %61, <4 x i32> %63, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %226 = extractelement <4 x float> %225, i32 0
>  %227 = extractelement <4 x float> %225, i32 1
>  %228 = extractelement <4 x float> %225, i32 2
>  %229 = extractelement <4 x float> %225, i32 3
>  %230 = fadd float %226, -1.000000e+00
>  %231 = fadd float %227, -1.000000e+00
>  %232 = fadd float %228, -1.000000e+00
>  %233 = fadd float %229, -1.000000e+00
>  %234 = fmul float %220, %230
>  %235 = fadd float %234, 1.000000e+00
>  %236 = fmul float %220, %231
>  %237 = fadd float %236, 1.000000e+00
>  %238 = fmul float %220, %232
>  %239 = fadd float %238, 1.000000e+00
>  %240 = fmul float %220, %233
>  %241 = fadd float %240, 1.000000e+00
>  %242 = fsub float 1.000000e+00, %202
>  %243 = fmul float %235, %242
>  %244 = fadd float %202, %243
>  %245 = fsub float 1.000000e+00, %202
>  %246 = fmul float %237, %245
>  %247 = fadd float %202, %246
>  %248 = fsub float 1.000000e+00, %202
>  %249 = fmul float %239, %248
>  %250 = fadd float %202, %249
>  %251 = fmul float %244, %206
>  %252 = fmul float %247, %210
>  %253 = fmul float %250, %214
>  %254 = fmul float %251, 0x3FD3333340000000
>  %255 = fmul float %252, 0x3FE2E147A0000000
>  %256 = fadd float %255, %254
>  %257 = fmul float %253, 0x3FBC28F5C0000000
>  %258 = fadd float %256, %257
>  %259 = fadd float %258, 0xBFB47AE140000000
>  %260 = fmul float %259, 0x402C924920000000
>  %261 = call float @llvm.AMDIL.clamp.(float %260, float 0.000000e+00, float 1.000000e+00)
>  %262 = fmul float %261, -2.000000e+00
>  %263 = fadd float %262, 3.000000e+00
>  %264 = fmul float %261, %261
>  %265 = fmul float %264, %263
>  %266 = fsub float 1.000000e+00, %265
>  %267 = fmul float %266, 0x3F747AE140000000
>  %268 = fadd float %267, %251
>  %269 = fmul float %266, 0x3F747AE140000000
>  %270 = fadd float %269, %252
>  %271 = fmul float %266, 0x3F747AE140000000
>  %272 = fadd float %271, %253
>  %273 = fmul float %268, %174
>  %274 = fmul float %270, %174
>  %275 = fmul float %272, %174
>  %276 = fmul float %273, 2.000000e+00
>  %277 = fadd float %276, %251
>  %278 = fmul float %274, 2.000000e+00
>  %279 = fadd float %278, %252
>  %280 = fmul float %275, 2.000000e+00
>  %281 = fadd float %280, %253
>  %282 = call float @llvm.AMDIL.clamp.(float %277, float 0.000000e+00, float 1.000000e+00)
>  %283 = call float @llvm.AMDIL.clamp.(float %279, float 0.000000e+00, float 1.000000e+00)
>  %284 = call float @llvm.AMDIL.clamp.(float %281, float 0.000000e+00, float 1.000000e+00)
>  %285 = fmul float %173, %282
>  %286 = fmul float %173, %283
>  %287 = fmul float %173, %284
>  %288 = bitcast float %68 to i32
>  %289 = bitcast float %69 to i32
>  %290 = insertelement <2 x i32> undef, i32 %288, i32 0
>  %291 = insertelement <2 x i32> %290, i32 %289, i32 1
>  %292 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %291, <8 x i32> %53, <4 x i32> %55, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %293 = extractelement <4 x float> %292, i32 0
>  %294 = extractelement <4 x float> %292, i32 1
>  %295 = extractelement <4 x float> %292, i32 2
>  %296 = extractelement <4 x float> %292, i32 3
>  %297 = fmul float %173, %282
>  %298 = fsub float %293, %297
>  %299 = fmul float %173, %283
>  %300 = fsub float %294, %299
>  %301 = fmul float %173, %284
>  %302 = fsub float %295, %301
>  %303 = fmul float %173, %38
>  %304 = fmul float %241, %303
>  %305 = fmul float %196, %298
>  %306 = fadd float %305, %285
>  %307 = fmul float %196, %300
>  %308 = fadd float %307, %286
>  %309 = fmul float %196, %302
>  %310 = fadd float %309, %287
>  %311 = fsub float 1.000000e+00, %202
>  %312 = fmul float %38, %311
>  %313 = fadd float %202, %312
>  %314 = fmul float %304, %313
>  %315 = bitcast float %68 to i32
>  %316 = bitcast float %69 to i32
>  %317 = insertelement <2 x i32> undef, i32 %315, i32 0
>  %318 = insertelement <2 x i32> %317, i32 %316, i32 1
>  %319 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %318, <8 x i32> %49, <4 x i32> %51, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %320 = extractelement <4 x float> %319, i32 1
>  %321 = fmul float %196, %320
>  %322 = fadd float %37, -1.000000e+00
>  %323 = fmul float %321, %322
>  %324 = fadd float %323, 1.000000e+00
>  %325 = fmul float %296, %324
>  %326 = fsub float %325, %314
>  %327 = fmul float %196, %326
>  %328 = fadd float %327, %314
>  %329 = call i32 @llvm.SI.packf16(float %306, float %308)
>  %330 = bitcast i32 %329 to float
>  %331 = call i32 @llvm.SI.packf16(float %310, float %328)
>  %332 = bitcast i32 %331 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %330, float %332, float %330, float %332)
>  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.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #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, s10                                                            ; BEFC030A
>	v_mov_b32_e32 v2, 0x41200000                                                 ; 7E0402FF 41200000
>	v_mov_b32_e32 v3, 0x40400000                                                 ; 7E0602FF 40400000
>	v_mov_b32_e32 v4, 0x40c00000                                                 ; 7E0802FF 40C00000
>	v_mov_b32_e32 v5, 0x3f400000                                                 ; 7E0A02FF 3F400000
>	v_mov_b32_e32 v6, 0xbda3d70a                                                 ; 7E0C02FF BDA3D70A
>	v_mov_b32_e32 v7, 0x3ba3d70a                                                 ; 7E0E02FF 3BA3D70A
>	v_interp_p1_f32 v8, v0, 0, 0, [m0]                                           ; C8200000
>	v_interp_p2_f32 v8, [v8], v1, 0, 0, [m0]                                     ; C8210001
>	v_interp_p1_f32 v9, v0, 1, 0, [m0]                                           ; C8240100
>	v_interp_p2_f32 v9, [v9], v1, 1, 0, [m0]                                     ; C8250101
>	v_interp_p1_f32 v10, v0, 2, 0, [m0]                                          ; C8280200
>	v_interp_p2_f32 v10, [v10], v1, 2, 0, [m0]                                   ; C8290201
>	v_interp_p1_f32 v11, v0, 3, 0, [m0]                                          ; C82C0300
>	v_interp_p2_f32 v11, [v11], v1, 3, 0, [m0]                                   ; C82D0301
>	v_interp_p1_f32 v12, v0, 0, 1, [m0]                                          ; C8300400
>	v_interp_p2_f32 v12, [v12], v1, 0, 1, [m0]                                   ; C8310401
>	v_interp_p1_f32 v13, v0, 1, 1, [m0]                                          ; C8340500
>	v_interp_p2_f32 v13, [v13], v1, 1, 1, [m0]                                   ; C8350501
>	v_interp_p1_f32 v14, v0, 2, 1, [m0]                                          ; C8380600
>	v_interp_p2_f32 v14, [v14], v1, 2, 1, [m0]                                   ; C8390601
>	v_interp_p1_f32 v15, v0, 3, 1, [m0]                                          ; C83C0700
>	v_interp_p2_f32 v15, [v15], v1, 3, 1, [m0]                                   ; C83D0701
>	s_load_dwordx4 s[8:11], s[2:3], 0x0                                          ; C0840300
>	s_load_dwordx4 s[40:43], s[4:5], 0x0                                         ; C0940500
>	s_load_dwordx4 s[44:47], s[4:5], 0x4                                         ; C0960504
>	s_load_dwordx4 s[0:3], s[4:5], 0x8                                           ; C0800508
>	s_load_dwordx4 s[16:19], s[4:5], 0xc                                         ; C088050C
>	s_load_dwordx4 s[48:51], s[4:5], 0x10                                        ; C0980510
>	s_load_dwordx4 s[28:31], s[4:5], 0x14                                        ; C08E0514
>	s_load_dwordx4 s[68:71], s[4:5], 0x18                                        ; C0A20518
>	s_load_dwordx8 s[60:67], s[6:7], 0x0                                         ; C0DE0700
>	s_load_dwordx8 s[52:59], s[6:7], 0x8                                         ; C0DA0708
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	s_buffer_load_dword s5, s[8:11], 0x0                                         ; C2028900
>	s_buffer_load_dword s72, s[8:11], 0x1                                        ; C2240901
>	s_buffer_load_dword s73, s[8:11], 0x2                                        ; C2248902
>	s_buffer_load_dword s74, s[8:11], 0x3                                        ; C2250903
>	s_buffer_load_dword s12, s[8:11], 0x4                                        ; C2060904
>	s_buffer_load_dword s13, s[8:11], 0x5                                        ; C2068905
>	s_buffer_load_dword s14, s[8:11], 0x6                                        ; C2070906
>	s_buffer_load_dword s75, s[8:11], 0x7                                        ; C2258907
>	s_buffer_load_dword s76, s[8:11], 0x8                                        ; C2260908
>	s_buffer_load_dword s77, s[8:11], 0x9                                        ; C2268909
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mov_b32_e32 v0, s5                                                         ; 7E000205
>	s_buffer_load_dword s78, s[8:11], 0xa                                        ; C227090A
>	v_mov_b32_e32 v1, s72                                                        ; 7E020248
>	s_buffer_load_dword s79, s[8:11], 0xb                                        ; C227890B
>	v_mov_b32_e32 v16, s73                                                       ; 7E200249
>	s_buffer_load_dword s80, s[8:11], 0xc                                        ; C228090C
>	s_buffer_load_dword s4, s[8:11], 0xe                                         ; C202090E
>	s_buffer_load_dword s81, s[8:11], 0xf                                        ; C228890F
>	v_sub_f32_e32 v0, s12, v0                                                    ; 0800000C
>	v_sub_f32_e32 v1, s13, v1                                                    ; 0802020D
>	v_sub_f32_e32 v16, s14, v16                                                  ; 0820200E
>	s_load_dwordx8 s[8:15], s[6:7], 0x10                                         ; C0C40710
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_writelane_b32 v29, s8, 0                                                   ; 043B0008
>	v_writelane_b32 v29, s9, 1                                                   ; 043B0209
>	v_writelane_b32 v29, s10, 2                                                  ; 043B040A
>	v_writelane_b32 v29, s11, 3                                                  ; 043B060B
>	v_writelane_b32 v29, s12, 4                                                  ; 043B080C
>	v_writelane_b32 v29, s13, 5                                                  ; 043B0A0D
>	v_writelane_b32 v29, s14, 6                                                  ; 043B0C0E
>	v_writelane_b32 v29, s15, 7                                                  ; 043B0E0F
>	s_load_dwordx8 s[20:27], s[6:7], 0x18                                        ; C0CA0718
>	s_load_dwordx8 s[84:91], s[6:7], 0x20                                        ; C0EA0720
>	s_load_dwordx8 s[32:39], s[6:7], 0x28                                        ; C0D00728
>	v_add_f32_e64 v17, -1.0, s80                                                 ; D2060011 0000A0F3
>	s_load_dwordx8 s[8:15], s[6:7], 0x30                                         ; C0C40730
>	v_mad_f32 v4, v4, s81, 1.0                                                   ; D2820004 03C8A304
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample v[18:21], 15, 0, 0, 0, 0, 0, 0, 0, v[10:11], s[8:15], s[68:71]  ; F0800F00 0222120A
>	image_sample v[22:24], 7, 0, 0, 0, 0, 0, 0, 0, v[8:9], s[84:91], s[48:51]    ; F0800700 01951608
>	image_sample v[25:28], 15, 0, 0, 0, 0, 0, 0, 0, v[8:9], s[60:67], s[40:43]   ; F0800F00 014F1908
>	image_sample v10, 2, 0, 0, 0, 0, 0, 0, 0, v[12:13], s[52:59], s[44:47]       ; F0800200 016D0A0C
>	s_waitcnt vmcnt(3)                                                           ; BF8C0773
>	v_mad_f32 v0, v18, v0, s5                                                    ; D2820000 00160112
>	v_mad_f32 v1, v18, v1, s72                                                   ; D2820001 01220312
>	v_mad_f32 v11, v18, v16, s73                                                 ; D282000B 01262112
>	v_mad_f32 v0, -v19, v0, v0                                                   ; D2820000 24020113
>	v_mac_f32_e32 v0, s76, v19                                                   ; 3E00264C
>	v_mad_f32 v0, -v20, v0, v0                                                   ; D2820000 24020114
>	v_mac_f32_e32 v0, s74, v20                                                   ; 3E00284A
>	s_waitcnt vmcnt(2)                                                           ; BF8C0772
>	v_mad_f32 v0, -v23, v0, v0                                                   ; D2820000 24020117
>	v_mac_f32_e32 v0, s76, v23                                                   ; 3E002E4C
>	v_mad_f32 v1, -v19, v1, v1                                                   ; D2820001 24060313
>	v_mac_f32_e32 v1, s77, v19                                                   ; 3E02264D
>	v_mad_f32 v1, -v20, v1, v1                                                   ; D2820001 24060314
>	v_mac_f32_e32 v1, s75, v20                                                   ; 3E02284B
>	v_mad_f32 v1, -v23, v1, v1                                                   ; D2820001 24060317
>	v_mac_f32_e32 v1, s77, v23                                                   ; 3E022E4D
>	v_mad_f32 v11, -v19, v11, v11                                                ; D282000B 242E1713
>	v_mac_f32_e32 v11, s78, v19                                                  ; 3E16264E
>	v_mad_f32 v11, -v20, v11, v11                                                ; D282000B 242E1714
>	v_mac_f32_e32 v11, s79, v20                                                  ; 3E16284F
>	v_mad_f32 v11, -v23, v11, v11                                                ; D282000B 242E1717
>	v_mac_f32_e32 v11, s78, v23                                                  ; 3E162E4E
>	v_mad_f32 v0, -v24, v0, v0                                                   ; D2820000 24020118
>	v_mac_f32_e32 v0, s74, v24                                                   ; 3E00304A
>	v_mad_f32 v1, -v24, v1, v1                                                   ; D2820001 24060318
>	v_mac_f32_e32 v1, s75, v24                                                   ; 3E02304B
>	v_mad_f32 v11, -v24, v11, v11                                                ; D282000B 242E1718
>	v_mac_f32_e32 v11, s79, v24                                                  ; 3E16304F
>	v_mul_f32_e32 v5, s81, v5                                                    ; 100A0A51
>	v_add_f32_e32 v12, -0.5, v21                                                 ; 06182AF1
>	v_madmk_f32_e32 v13, v21, v2, 0xc1200000                                     ; 401A0515 C1200000
>	v_mul_f32_e32 v12, v2, v12                                                   ; 10181902
>	v_add_f32_e64 v12, 0, v12 clamp                                              ; D206080C 00021880
>	v_mad_f32 v16, -2.0, v12, v3                                                 ; D2820010 040E18F5
>	v_mul_f32_e32 v12, v12, v12                                                  ; 1018190C
>	v_mul_f32_e32 v12, v16, v12                                                  ; 10181910
>	v_add_f32_e64 v13, 0, v13 clamp                                              ; D206080D 00021A80
>	v_mad_f32 v16, -2.0, v13, v3                                                 ; D2820010 040E1AF5
>	v_mul_f32_e32 v13, v13, v13                                                  ; 101A1B0D
>	v_mul_f32_e32 v13, v16, v13                                                  ; 101A1B10
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mac_f32_e32 v28, v25, v10                                                  ; 3E381519
>	v_madak_f32_e32 v10, v13, v12, 0xbf147ae1                                    ; 4214190D BF147AE1
>	v_mac_f32_e32 v10, v4, v28                                                   ; 3E143904
>	v_mul_f32_e32 v2, v2, v10                                                    ; 10041502
>	v_add_f32_e64 v2, 0, v2 clamp                                                ; D2060802 00020480
>	v_mad_f32 v4, -2.0, v2, v3                                                   ; D2820004 040E04F5
>	v_mul_f32_e32 v2, v2, v2                                                     ; 10040502
>	v_mad_f32 v10, v2, v4, 1.0                                                   ; D282000A 03CA0902
>	v_subrev_f32_e32 v10, v22, v10                                               ; 0A141516
>	v_sub_f32_e32 v12, 1.0, v25                                                  ; 081832F2
>	v_mad_f32 v12, -v25, v12, v12                                                ; D282000C 24321919
>	v_mul_f32_e32 v12, v12, v12                                                  ; 1018190C
>	v_madmk_f32_e32 v5, v12, v5, 0x3e800000                                      ; 400A0B0C 3E800000
>	image_sample v[12:15], 15, 0, 0, 0, 0, 0, 0, 0, v[14:15], s[32:39], s[28:31] ; F0800F00 00E80C0E
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_add_f32_e32 v12, -1.0, v12                                                 ; 061818F3
>	v_add_f32_e32 v13, -1.0, v13                                                 ; 061A1AF3
>	v_add_f32_e32 v14, -1.0, v14                                                 ; 061C1CF3
>	v_mad_f32 v15, v15, v5, -v5                                                  ; D282000F 84160B0F
>	v_mul_f32_e32 v2, v4, v2                                                     ; 10040504
>	v_add_f32_e64 v4, 0, v10 clamp                                               ; D2060804 00021480
>	v_mul_f32_e32 v2, 0x42c80000, v2                                             ; 100404FF 42C80000
>	v_add_f32_e64 v2, 0, v2 clamp                                                ; D2060802 00020480
>	v_mad_f32 v10, -2.0, v2, v3                                                  ; D282000A 040E04F5
>	v_mul_f32_e32 v2, v2, v2                                                     ; 10040502
>	v_mul_f32_e32 v16, v10, v2                                                   ; 1020050A
>	v_mad_f32 v2, -v2, v10, 1.0                                                  ; D2820002 23CA1502
>	v_mul_f32_e32 v10, 0x3d4ccccd, v16                                           ; 101420FF 3D4CCCCD
>	v_mad_f32 v0, v2, v0, v10                                                    ; D2820000 042A0102
>	v_mad_f32 v1, v2, v1, v10                                                    ; D2820001 042A0302
>	v_mad_f32 v10, v2, v11, v10                                                  ; D282000A 042A1702
>	v_mad_f32 v11, v5, v12, 1.0                                                  ; D282000B 03CA1905
>	v_mad_f32 v12, v5, v13, 1.0                                                  ; D282000C 03CA1B05
>	v_mad_f32 v5, v5, v14, 1.0                                                   ; D2820005 03CA1D05
>	v_mad_f32 v11, v2, v11, v16                                                  ; D282000B 04421702
>	v_mad_f32 v12, v2, v12, v16                                                  ; D282000C 04421902
>	v_mad_f32 v5, v2, v5, v16                                                    ; D2820005 04420B02
>	v_mul_f32_e32 v13, v0, v11                                                   ; 101A1700
>	v_mul_f32_e32 v14, v1, v12                                                   ; 101C1901
>	v_mul_f32_e32 v13, 0x3e99999a, v13                                           ; 101A1AFF 3E99999A
>	v_madmk_f32_e32 v13, v14, v13, 0x3f170a3d                                    ; 401A1B0E 3F170A3D
>	v_mul_f32_e32 v14, v10, v5                                                   ; 101C0B0A
>	v_madmk_f32_e32 v13, v14, v13, 0x3de147ae                                    ; 401A1B0E 3DE147AE
>	v_add_f32_e32 v6, v13, v6                                                    ; 060C0D0D
>	v_mul_f32_e32 v6, 0x41649249, v6                                             ; 100C0CFF 41649249
>	v_add_f32_e64 v6, 0, v6 clamp                                                ; D2060806 00020C80
>	v_mac_f32_e32 v3, -2.0, v6                                                   ; 3E060CF5
>	v_mul_f32_e32 v6, v6, v6                                                     ; 100C0D06
>	v_mul_f32_e32 v3, v3, v6                                                     ; 10060D03
>	v_mad_f32 v3, -v3, v7, v7                                                    ; D2820003 241E0F03
>	v_mad_f32 v6, v0, v11, v3                                                    ; D2820006 040E1700
>	v_mad_f32 v7, v1, v12, v3                                                    ; D2820007 040E1901
>	v_mad_f32 v3, v10, v5, v3                                                    ; D2820003 040E0B0A
>	v_mul_f32_e32 v13, v27, v6                                                   ; 101A0D1B
>	v_mul_f32_e32 v14, v27, v7                                                   ; 101C0F1B
>	v_mac_f32_e32 v13, v27, v6                                                   ; 3E1A0D1B
>	v_mul_f32_e32 v6, v27, v3                                                    ; 100C071B
>	v_mac_f32_e32 v13, v0, v11                                                   ; 3E1A1700
>	v_mac_f32_e32 v14, v27, v7                                                   ; 3E1C0F1B
>	v_mac_f32_e32 v14, v1, v12                                                   ; 3E1C1901
>	v_mac_f32_e32 v6, v27, v3                                                    ; 3E0C071B
>	v_mac_f32_e32 v6, v10, v5                                                    ; 3E0C0B0A
>	v_add_f32_e64 v0, 0, v13 clamp                                               ; D2060800 00021A80
>	v_add_f32_e64 v1, 0, v14 clamp                                               ; D2060801 00021C80
>	v_add_f32_e64 v3, 0, v6 clamp                                                ; D2060803 00020C80
>	v_mul_f32_e32 v5, v0, v26                                                    ; 100A3500
>	v_mul_f32_e32 v6, v1, v26                                                    ; 100C3501
>	v_mul_f32_e32 v7, v3, v26                                                    ; 100E3503
>	image_sample v[10:13], 15, 0, 0, 0, 0, 0, 0, 0, v[8:9], s[20:27], s[16:19]   ; F0800F00 00850A08
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v0, -v26, v0, v10                                                  ; D2820000 242A011A
>	v_mad_f32 v1, -v26, v1, v11                                                  ; D2820001 242E031A
>	v_mad_f32 v3, -v26, v3, v12                                                  ; D2820003 2432071A
>	v_mul_f32_e32 v10, s4, v26                                                   ; 10143404
>	v_mac_f32_e32 v10, v10, v15                                                  ; 3E141F0A
>	v_mac_f32_e32 v5, v0, v4                                                     ; 3E0A0900
>	v_mac_f32_e32 v6, v1, v4                                                     ; 3E0C0901
>	v_mac_f32_e32 v7, v3, v4                                                     ; 3E0E0903
>	v_mac_f32_e32 v16, s4, v2                                                    ; 3E200404
>	v_mul_f32_e32 v0, v16, v10                                                   ; 10001510
>	v_readlane_b32 s4, v29, 0                                                    ; 0209011D
>	v_readlane_b32 s5, v29, 1                                                    ; 020B031D
>	v_readlane_b32 s6, v29, 2                                                    ; 020D051D
>	v_readlane_b32 s7, v29, 3                                                    ; 020F071D
>	v_readlane_b32 s8, v29, 4                                                    ; 0211091D
>	v_readlane_b32 s9, v29, 5                                                    ; 02130B1D
>	v_readlane_b32 s10, v29, 6                                                   ; 02150D1D
>	v_readlane_b32 s11, v29, 7                                                   ; 02170F1D
>	image_sample v1, 2, 0, 0, 0, 0, 0, 0, 0, v[8:9], s[4:11], s[0:3]             ; F0800200 00010108
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mul_f32_e32 v1, v1, v4                                                     ; 10020901
>	v_mad_f32 v1, v1, v17, 1.0                                                   ; D2820001 03CA2301
>	v_mad_f32 v1, v13, v1, -v0                                                   ; D2820001 8402030D
>	v_mac_f32_e32 v0, v1, v4                                                     ; 3E000901
>	v_cvt_pkrtz_f16_f32_e32 v1, v5, v6                                           ; 5E020D05
>	v_cvt_pkrtz_f16_f32_e32 v0, v7, v0                                           ; 5E000107
>	exp 15, 0, 1, 1, 1, v1, v0, v1, v0                                           ; F8001C0F 00010001
>	s_endpgm                                                                     ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 96
>VGPRS: 32
>Code Size: 1172 bytes
>LDS: 0 blocks
>Scratch: 3072 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
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL OUT[1], CLIPVERTEX
>DCL OUT[2], GENERIC[0]
>DCL CONST[0..54]
>DCL TEMP[0..2], LOCAL
>  0: MUL TEMP[0].xy, CONST[48].xyyy, IN[1].xyyy
>  1: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
>  2: ADD TEMP[1].x, TEMP[0].xxxx, CONST[48].wwww
>  3: MOV TEMP[1].z, TEMP[1].xxxx
>  4: MUL TEMP[0].xy, CONST[49].xyyy, IN[1].xyyy
>  5: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
>  6: ADD TEMP[0].x, TEMP[0].xxxx, CONST[49].wwww
>  7: MOV TEMP[1].w, TEMP[0].xxxx
>  8: MOV TEMP[0].xw, IN[0].xxxw
>  9: MOV TEMP[1].xy, IN[1].xyxx
> 10: MAD TEMP[2].x, IN[0].zzzz, CONST[0].zzzz, -IN[0].wwww
> 11: MOV TEMP[0].z, TEMP[2].xxxx
> 12: MOV TEMP[0].y, -IN[0].yyyy
> 13: MAD TEMP[0].xy, CONST[54].xyyy, IN[0].wwww, TEMP[0].xyyy
> 14: MOV OUT[2], TEMP[1]
> 15: MOV OUT[0], TEMP[0]
> 16: MOV OUT[1], IN[0]
> 17: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = call float @llvm.SI.load.const(<16 x i8> %13, i32 8)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %13, i32 768)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %13, i32 772)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %13, i32 780)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %13, i32 784)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %13, i32 788)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %13, i32 796)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %13, i32 864)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %13, i32 868)
>  %23 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %24 = load <16 x i8>, <16 x i8> addrspace(2)* %23, align 16, !tbaa !0
>  %25 = add i32 %5, %8
>  %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 = extractelement <4 x float> %26, i32 2
>  %30 = extractelement <4 x float> %26, i32 3
>  %31 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 1
>  %32 = load <16 x i8>, <16 x i8> addrspace(2)* %31, align 16, !tbaa !0
>  %33 = add i32 %5, %8
>  %34 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %32, i32 0, i32 %33)
>  %35 = extractelement <4 x float> %34, i32 0
>  %36 = extractelement <4 x float> %34, i32 1
>  %37 = fmul float %15, %35
>  %38 = fmul float %16, %36
>  %39 = fadd float %38, %37
>  %40 = fadd float %39, %17
>  %41 = fmul float %18, %35
>  %42 = fmul float %19, %36
>  %43 = fadd float %42, %41
>  %44 = fadd float %43, %20
>  %45 = fmul float %29, %14
>  %46 = fsub float %45, %30
>  %47 = fmul float %21, %30
>  %48 = fadd float %47, %27
>  %49 = fmul float %22, %30
>  %50 = fsub float %49, %28
>  %51 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 16
>  %52 = load <16 x i8>, <16 x i8> addrspace(2)* %51, align 16, !tbaa !0
>  %53 = call float @llvm.SI.load.const(<16 x i8> %52, i32 0)
>  %54 = fmul float %53, %27
>  %55 = call float @llvm.SI.load.const(<16 x i8> %52, i32 4)
>  %56 = fmul float %55, %28
>  %57 = fadd float %54, %56
>  %58 = call float @llvm.SI.load.const(<16 x i8> %52, i32 8)
>  %59 = fmul float %58, %29
>  %60 = fadd float %57, %59
>  %61 = call float @llvm.SI.load.const(<16 x i8> %52, i32 12)
>  %62 = fmul float %61, %30
>  %63 = fadd float %60, %62
>  %64 = call float @llvm.SI.load.const(<16 x i8> %52, i32 16)
>  %65 = fmul float %64, %27
>  %66 = call float @llvm.SI.load.const(<16 x i8> %52, i32 20)
>  %67 = fmul float %66, %28
>  %68 = fadd float %65, %67
>  %69 = call float @llvm.SI.load.const(<16 x i8> %52, i32 24)
>  %70 = fmul float %69, %29
>  %71 = fadd float %68, %70
>  %72 = call float @llvm.SI.load.const(<16 x i8> %52, i32 28)
>  %73 = fmul float %72, %30
>  %74 = fadd float %71, %73
>  %75 = call float @llvm.SI.load.const(<16 x i8> %52, i32 32)
>  %76 = fmul float %75, %27
>  %77 = call float @llvm.SI.load.const(<16 x i8> %52, i32 36)
>  %78 = fmul float %77, %28
>  %79 = fadd float %76, %78
>  %80 = call float @llvm.SI.load.const(<16 x i8> %52, i32 40)
>  %81 = fmul float %80, %29
>  %82 = fadd float %79, %81
>  %83 = call float @llvm.SI.load.const(<16 x i8> %52, i32 44)
>  %84 = fmul float %83, %30
>  %85 = fadd float %82, %84
>  %86 = call float @llvm.SI.load.const(<16 x i8> %52, i32 48)
>  %87 = fmul float %86, %27
>  %88 = call float @llvm.SI.load.const(<16 x i8> %52, i32 52)
>  %89 = fmul float %88, %28
>  %90 = fadd float %87, %89
>  %91 = call float @llvm.SI.load.const(<16 x i8> %52, i32 56)
>  %92 = fmul float %91, %29
>  %93 = fadd float %90, %92
>  %94 = call float @llvm.SI.load.const(<16 x i8> %52, i32 60)
>  %95 = fmul float %94, %30
>  %96 = fadd float %93, %95
>  %97 = call float @llvm.SI.load.const(<16 x i8> %52, i32 64)
>  %98 = fmul float %97, %27
>  %99 = call float @llvm.SI.load.const(<16 x i8> %52, i32 68)
>  %100 = fmul float %99, %28
>  %101 = fadd float %98, %100
>  %102 = call float @llvm.SI.load.const(<16 x i8> %52, i32 72)
>  %103 = fmul float %102, %29
>  %104 = fadd float %101, %103
>  %105 = call float @llvm.SI.load.const(<16 x i8> %52, i32 76)
>  %106 = fmul float %105, %30
>  %107 = fadd float %104, %106
>  %108 = call float @llvm.SI.load.const(<16 x i8> %52, i32 80)
>  %109 = fmul float %108, %27
>  %110 = call float @llvm.SI.load.const(<16 x i8> %52, i32 84)
>  %111 = fmul float %110, %28
>  %112 = fadd float %109, %111
>  %113 = call float @llvm.SI.load.const(<16 x i8> %52, i32 88)
>  %114 = fmul float %113, %29
>  %115 = fadd float %112, %114
>  %116 = call float @llvm.SI.load.const(<16 x i8> %52, i32 92)
>  %117 = fmul float %116, %30
>  %118 = fadd float %115, %117
>  %119 = call float @llvm.SI.load.const(<16 x i8> %52, i32 96)
>  %120 = fmul float %119, %27
>  %121 = call float @llvm.SI.load.const(<16 x i8> %52, i32 100)
>  %122 = fmul float %121, %28
>  %123 = fadd float %120, %122
>  %124 = call float @llvm.SI.load.const(<16 x i8> %52, i32 104)
>  %125 = fmul float %124, %29
>  %126 = fadd float %123, %125
>  %127 = call float @llvm.SI.load.const(<16 x i8> %52, i32 108)
>  %128 = fmul float %127, %30
>  %129 = fadd float %126, %128
>  %130 = call float @llvm.SI.load.const(<16 x i8> %52, i32 112)
>  %131 = fmul float %130, %27
>  %132 = call float @llvm.SI.load.const(<16 x i8> %52, i32 116)
>  %133 = fmul float %132, %28
>  %134 = fadd float %131, %133
>  %135 = call float @llvm.SI.load.const(<16 x i8> %52, i32 120)
>  %136 = fmul float %135, %29
>  %137 = fadd float %134, %136
>  %138 = call float @llvm.SI.load.const(<16 x i8> %52, i32 124)
>  %139 = fmul float %138, %30
>  %140 = fadd float %137, %139
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %35, float %36, float %40, float %44)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 12, i32 0, float %48, float %50, float %46, float %30)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 13, i32 0, float %63, float %74, float %85, float %96)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 14, i32 0, float %107, float %118, float %129, float %140)
>  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[4:7], s[8:9], 0x0                    ; C0820900
>	s_load_dwordx4 s[12:15], s[8:9], 0x4                  ; C0860904
>	v_add_i32_e32 v0, vcc, s10, v0                        ; 4A00000A
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[4:7], 0 idxen   ; E00C2000 80010100
>	buffer_load_format_xyzw v[5:8], v0, s[12:15], 0 idxen ; E00C2000 80030500
>	s_load_dwordx4 s[4:7], s[2:3], 0x0                    ; C0820300
>	s_load_dwordx4 s[0:3], s[2:3], 0x40                   ; C0800340
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	s_buffer_load_dword s8, s[4:7], 0x2                   ; C2040502
>	s_buffer_load_dword s9, s[4:7], 0xc0                  ; C20485C0
>	s_buffer_load_dword s10, s[4:7], 0xc1                 ; C20505C1
>	s_buffer_load_dword s11, s[4:7], 0xc3                 ; C20585C3
>	s_buffer_load_dword s12, s[4:7], 0xc4                 ; C20605C4
>	s_buffer_load_dword s13, s[4:7], 0xc5                 ; C20685C5
>	s_buffer_load_dword s14, s[4:7], 0xc7                 ; C20705C7
>	s_buffer_load_dword s15, s[4:7], 0xd8                 ; C20785D8
>	s_buffer_load_dword s4, s[4:7], 0xd9                  ; C20205D9
>	s_buffer_load_dword s5, s[0:3], 0x0                   ; C2028100
>	s_buffer_load_dword s6, s[0:3], 0x1                   ; C2030101
>	s_buffer_load_dword s7, s[0:3], 0x2                   ; C2038102
>	s_buffer_load_dword s16, s[0:3], 0x3                  ; C2080103
>	s_buffer_load_dword s17, s[0:3], 0x4                  ; C2088104
>	s_buffer_load_dword s18, s[0:3], 0x5                  ; C2090105
>	s_buffer_load_dword s19, s[0:3], 0x6                  ; C2098106
>	s_buffer_load_dword s20, s[0:3], 0x7                  ; C20A0107
>	s_buffer_load_dword s21, s[0:3], 0x8                  ; C20A8108
>	s_buffer_load_dword s22, s[0:3], 0x9                  ; C20B0109
>	s_buffer_load_dword s23, s[0:3], 0xa                  ; C20B810A
>	s_buffer_load_dword s24, s[0:3], 0xb                  ; C20C010B
>	s_buffer_load_dword s25, s[0:3], 0xc                  ; C20C810C
>	s_buffer_load_dword s26, s[0:3], 0xd                  ; C20D010D
>	s_buffer_load_dword s27, s[0:3], 0xe                  ; C20D810E
>	s_buffer_load_dword s28, s[0:3], 0xf                  ; C20E010F
>	s_buffer_load_dword s29, s[0:3], 0x10                 ; C20E8110
>	s_buffer_load_dword s30, s[0:3], 0x11                 ; C20F0111
>	s_buffer_load_dword s31, s[0:3], 0x12                 ; C20F8112
>	s_buffer_load_dword s32, s[0:3], 0x13                 ; C2100113
>	s_buffer_load_dword s33, s[0:3], 0x14                 ; C2108114
>	s_buffer_load_dword s34, s[0:3], 0x15                 ; C2110115
>	s_buffer_load_dword s35, s[0:3], 0x16                 ; C2118116
>	s_buffer_load_dword s36, s[0:3], 0x17                 ; C2120117
>	s_buffer_load_dword s37, s[0:3], 0x18                 ; C2128118
>	s_buffer_load_dword s38, s[0:3], 0x19                 ; C2130119
>	s_buffer_load_dword s39, s[0:3], 0x1a                 ; C213811A
>	s_buffer_load_dword s40, s[0:3], 0x1b                 ; C214011B
>	s_buffer_load_dword s41, s[0:3], 0x1c                 ; C214811C
>	s_buffer_load_dword s42, s[0:3], 0x1d                 ; C215011D
>	s_buffer_load_dword s43, s[0:3], 0x1e                 ; C215811E
>	s_buffer_load_dword s0, s[0:3], 0x1f                  ; C200011F
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mov_b32_e32 v0, s11                                 ; 7E00020B
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mov_b32_e32 v7, s14                                 ; 7E0E020E
>	v_mac_f32_e32 v0, s9, v5                              ; 3E000A09
>	v_mac_f32_e32 v7, s12, v5                             ; 3E0E0A0C
>	v_mul_f32_e32 v8, s6, v2                              ; 10100406
>	v_mul_f32_e32 v9, s18, v2                             ; 10120412
>	v_mul_f32_e32 v10, s22, v2                            ; 10140416
>	v_mul_f32_e32 v11, s26, v2                            ; 1016041A
>	v_mul_f32_e32 v12, s30, v2                            ; 1018041E
>	v_mul_f32_e32 v13, s34, v2                            ; 101A0422
>	v_mul_f32_e32 v14, s38, v2                            ; 101C0426
>	v_mul_f32_e32 v15, s42, v2                            ; 101E042A
>	v_mac_f32_e32 v0, s10, v6                             ; 3E000C0A
>	v_mac_f32_e32 v7, s13, v6                             ; 3E0E0C0D
>	v_mac_f32_e32 v8, s5, v1                              ; 3E100205
>	v_mac_f32_e32 v9, s17, v1                             ; 3E120211
>	v_mac_f32_e32 v10, s21, v1                            ; 3E140215
>	v_mac_f32_e32 v11, s25, v1                            ; 3E160219
>	v_mac_f32_e32 v12, s29, v1                            ; 3E18021D
>	v_mac_f32_e32 v13, s33, v1                            ; 3E1A0221
>	v_mac_f32_e32 v14, s37, v1                            ; 3E1C0225
>	v_mac_f32_e32 v15, s41, v1                            ; 3E1E0229
>	v_mac_f32_e32 v8, s7, v3                              ; 3E100607
>	v_mac_f32_e32 v9, s19, v3                             ; 3E120613
>	exp 15, 32, 0, 0, 0, v5, v6, v0, v7                   ; F800020F 07000605
>	v_mac_f32_e32 v10, s23, v3                            ; 3E140617
>	v_mac_f32_e32 v11, s27, v3                            ; 3E16061B
>	v_mac_f32_e32 v12, s31, v3                            ; 3E18061F
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mad_f32 v0, v3, s8, -v4                             ; D2820000 84101103
>	v_mad_f32 v1, s15, v4, v1                             ; D2820001 0406080F
>	v_mad_f32 v2, s4, v4, -v2                             ; D2820002 840A0804
>	v_mac_f32_e32 v13, s35, v3                            ; 3E1A0623
>	v_mac_f32_e32 v14, s39, v3                            ; 3E1C0627
>	v_mac_f32_e32 v15, s43, v3                            ; 3E1E062B
>	v_mac_f32_e32 v8, s16, v4                             ; 3E100810
>	v_mac_f32_e32 v9, s20, v4                             ; 3E120814
>	v_mac_f32_e32 v10, s24, v4                            ; 3E140818
>	v_mac_f32_e32 v11, s28, v4                            ; 3E16081C
>	v_mac_f32_e32 v12, s32, v4                            ; 3E180820
>	v_mac_f32_e32 v13, s36, v4                            ; 3E1A0824
>	v_mac_f32_e32 v14, s40, v4                            ; 3E1C0828
>	v_mac_f32_e32 v15, s0, v4                             ; 3E1E0800
>	exp 15, 12, 0, 0, 0, v1, v2, v0, v4                   ; F80000CF 04000201
>	exp 15, 13, 0, 0, 0, v8, v9, v10, v11                 ; F80000DF 0B0A0908
>	exp 15, 14, 0, 1, 0, v12, v13, v14, v15               ; F80008EF 0F0E0D0C
>	s_endpgm                                              ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 48
>VGPRS: 16
>Code Size: 432 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
>  clamp_color = 0
>FRAG
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SAMP[1]
>DCL SAMP[2]
>DCL SVIEW[0], 2D, FLOAT
>DCL SVIEW[1], 2D, FLOAT
>DCL SVIEW[2], 2D, FLOAT
>DCL CONST[0..3]
>DCL TEMP[0..3], LOCAL
>IMM[0] FLT32 {    0.9960,     1.0000,     0.0000,    -1.0000}
>  0: MOV TEMP[0].xy, IN[0].xyyy
>  1: TEX TEMP[0].w, TEMP[0], SAMP[0], 2D
>  2: ADD TEMP[0].x, -TEMP[0].wwww, IMM[0].xxxx
>  3: MOV TEMP[1].xy, IN[0].xyyy
>  4: TEX TEMP[1].x, TEMP[1], SAMP[2], 2D
>  5: ADD TEMP[1].x, -TEMP[1].xxxx, IMM[0].yyyy
>  6: FSGE TEMP[0].x, TEMP[0].xxxx, IMM[0].zzzz
>  7: UIF TEMP[0].xxxx :0
>  8:   MOV TEMP[0].x, TEMP[1].xxxx
>  9: ELSE :0
> 10:   MOV TEMP[0].x, IMM[0].yyyy
> 11: ENDIF
> 12: MOV TEMP[1].xy, IN[0].xyyy
> 13: TEX TEMP[1].xyz, TEMP[1], SAMP[1], 2D
> 14: ADD TEMP[2].x, TEMP[1].xxxx, -CONST[3].yyyy
> 15: MAD TEMP[2].x, TEMP[0].xxxx, TEMP[2].xxxx, CONST[3].yyyy
> 16: ADD TEMP[3].x, TEMP[1].yyyy, IMM[0].wwww
> 17: MAD TEMP[1].xy, TEMP[1].zzzz, IMM[0].yzzz, IMM[0].zyyy
> 18: MOV TEMP[2].zw, TEMP[1].yyxy
> 19: MAD TEMP[0].x, TEMP[0].xxxx, TEMP[3].xxxx, IMM[0].yyyy
> 20: MOV TEMP[2].y, TEMP[0].xxxx
> 21: MOV OUT[0], TEMP[2]
> 22: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %24 = load <16 x i8>, <16 x i8> addrspace(2)* %23, align 16, !tbaa !0
>  %25 = call float @llvm.SI.load.const(<16 x i8> %24, i32 52)
>  %26 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %27 = load <8 x i32>, <8 x i32> addrspace(2)* %26, align 32, !tbaa !0
>  %28 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %29 = load <4 x i32>, <4 x i32> addrspace(2)* %28, align 16, !tbaa !0
>  %30 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 1
>  %31 = load <8 x i32>, <8 x i32> addrspace(2)* %30, align 32, !tbaa !0
>  %32 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 1
>  %33 = load <4 x i32>, <4 x i32> addrspace(2)* %32, align 16, !tbaa !0
>  %34 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 2
>  %35 = load <8 x i32>, <8 x i32> addrspace(2)* %34, align 32, !tbaa !0
>  %36 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 2
>  %37 = load <4 x i32>, <4 x i32> addrspace(2)* %36, align 16, !tbaa !0
>  %38 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %8)
>  %39 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %8)
>  %40 = bitcast float %38 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 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %43, <8 x i32> %27, <4 x i32> %29, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %45 = extractelement <4 x float> %44, i32 3
>  %46 = fsub float 0x3FEFDF3B60000000, %45
>  %47 = bitcast float %38 to i32
>  %48 = bitcast float %39 to i32
>  %49 = insertelement <2 x i32> undef, i32 %47, i32 0
>  %50 = insertelement <2 x i32> %49, i32 %48, i32 1
>  %51 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %50, <8 x i32> %35, <4 x i32> %37, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %52 = extractelement <4 x float> %51, i32 0
>  %53 = fsub float 1.000000e+00, %52
>  %54 = fcmp oge float %46, 0.000000e+00
>  %. = select i1 %54, float %53, float 1.000000e+00
>  %55 = bitcast float %38 to i32
>  %56 = bitcast float %39 to i32
>  %57 = insertelement <2 x i32> undef, i32 %55, i32 0
>  %58 = insertelement <2 x i32> %57, i32 %56, i32 1
>  %59 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %58, <8 x i32> %31, <4 x i32> %33, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %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 = fsub float %60, %25
>  %64 = fmul float %., %63
>  %65 = fadd float %64, %25
>  %66 = fadd float %61, -1.000000e+00
>  %67 = fadd float %62, 0.000000e+00
>  %68 = fmul float %62, 0.000000e+00
>  %69 = fadd float %68, 1.000000e+00
>  %70 = fmul float %., %66
>  %71 = fadd float %70, 1.000000e+00
>  %72 = call i32 @llvm.SI.packf16(float %65, float %71)
>  %73 = bitcast i32 %72 to float
>  %74 = call i32 @llvm.SI.packf16(float %67, float %69)
>  %75 = bitcast i32 %74 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %73, float %75, float %73, float %75)
>  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.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, 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 }
>
>!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_load_dwordx8 s[12:19], s[6:7], 0x0                                    ; C0C60700
>	s_load_dwordx8 s[20:27], s[6:7], 0x10                                   ; C0CA0710
>	s_load_dwordx4 s[28:31], s[4:5], 0x0                                    ; C08E0500
>	s_load_dwordx4 s[32:35], s[4:5], 0x4                                    ; C0900504
>	s_load_dwordx4 s[36:39], s[4:5], 0x8                                    ; C0920508
>	s_load_dwordx8 s[40:47], s[6:7], 0x8                                    ; C0D40708
>	s_mov_b32 m0, s10                                                       ; BEFC030A
>	s_waitcnt lgkmcnt(0)                                                    ; BF8C007F
>	s_buffer_load_dword s0, s[0:3], 0xd                                     ; C200010D
>	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
>	image_sample v0, 8, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[12:19], s[28:31]     ; F0800800 00E30002
>	image_sample v1, 1, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[20:27], s[36:39]     ; F0800100 01250102
>	image_sample v[2:4], 7, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[40:47], s[32:35] ; F0800700 010A0202
>	s_waitcnt vmcnt(2)                                                      ; BF8C0772
>	v_sub_f32_e32 v0, 0x3f7ef9db, v0                                        ; 080000FF 3F7EF9DB
>	s_waitcnt vmcnt(1)                                                      ; BF8C0771
>	v_sub_f32_e32 v1, 1.0, v1                                               ; 080202F2
>	v_cmp_le_f32_e32 vcc, 0, v0                                             ; 7C060080
>	v_cndmask_b32_e32 v0, 1.0, v1                                           ; 000002F2
>	s_waitcnt vmcnt(0) lgkmcnt(0)                                           ; BF8C0070
>	v_subrev_f32_e32 v1, s0, v2                                             ; 0A020400
>	v_mad_f32 v1, v0, v1, s0                                                ; D2820001 00020300
>	v_add_f32_e32 v2, -1.0, v3                                              ; 060406F3
>	v_add_f32_e32 v3, 0, v4                                                 ; 06060880
>	v_mad_f32 v4, 0, v4, 1.0                                                ; D2820004 03CA0880
>	v_mad_f32 v0, v0, v2, 1.0                                               ; D2820000 03CA0500
>	v_cvt_pkrtz_f16_f32_e32 v0, v1, v0                                      ; 5E000101
>	v_cvt_pkrtz_f16_f32_e32 v1, v3, v4                                      ; 5E020903
>	exp 15, 0, 1, 1, 1, v0, v1, v0, v1                                      ; F8001C0F 01000100
>	s_endpgm                                                                ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 56
>VGPRS: 8
>Code Size: 172 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
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL OUT[1], CLIPVERTEX
>DCL OUT[2], GENERIC[0]
>DCL OUT[3], GENERIC[1]
>DCL CONST[0..54]
>DCL TEMP[0..3], LOCAL
>  0: MUL TEMP[0].xy, CONST[48].xyyy, IN[1].xyyy
>  1: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
>  2: ADD TEMP[1].x, TEMP[0].xxxx, CONST[48].wwww
>  3: MOV TEMP[1].z, TEMP[1].xxxx
>  4: MUL TEMP[0].xy, CONST[49].xyyy, IN[1].xyyy
>  5: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
>  6: ADD TEMP[2].x, TEMP[0].xxxx, CONST[49].wwww
>  7: MOV TEMP[1].w, TEMP[2].xxxx
>  8: MUL TEMP[0].xy, CONST[50].xyyy, IN[1].xyyy
>  9: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 10: ADD TEMP[2].x, TEMP[0].xxxx, CONST[50].wwww
> 11: MUL TEMP[0].xy, CONST[51].xyyy, IN[1].xyyy
> 12: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 13: ADD TEMP[3].x, TEMP[0].xxxx, CONST[51].wwww
> 14: MOV TEMP[2].y, TEMP[3].xxxx
> 15: MUL TEMP[0].xy, CONST[52].xyyy, IN[1].xyyy
> 16: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 17: ADD TEMP[3].x, TEMP[0].xxxx, CONST[52].wwww
> 18: MOV TEMP[2].z, TEMP[3].xxxx
> 19: MUL TEMP[0].xy, CONST[53].xyyy, IN[1].xyyy
> 20: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 21: ADD TEMP[0].x, TEMP[0].xxxx, CONST[53].wwww
> 22: MOV TEMP[2].w, TEMP[0].xxxx
> 23: MOV TEMP[0].xw, IN[0].xxxw
> 24: MOV TEMP[1].xy, IN[1].xyxx
> 25: MAD TEMP[3].x, IN[0].zzzz, CONST[0].zzzz, -IN[0].wwww
> 26: MOV TEMP[0].z, TEMP[3].xxxx
> 27: MOV TEMP[0].y, -IN[0].yyyy
> 28: MAD TEMP[0].xy, CONST[54].xyyy, IN[0].wwww, TEMP[0].xyyy
> 29: MOV OUT[2], TEMP[1]
> 30: MOV OUT[3], TEMP[2]
> 31: MOV OUT[0], TEMP[0]
> 32: MOV OUT[1], IN[0]
> 33: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = call float @llvm.SI.load.const(<16 x i8> %13, i32 8)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %13, i32 768)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %13, i32 772)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %13, i32 780)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %13, i32 784)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %13, i32 788)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %13, i32 796)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %13, i32 800)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %13, i32 804)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %13, i32 812)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %13, i32 816)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %13, i32 820)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %13, i32 828)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %13, i32 832)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %13, i32 836)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %13, i32 844)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %13, i32 848)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %13, i32 852)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %13, i32 860)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %13, i32 864)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %13, i32 868)
>  %35 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %36 = load <16 x i8>, <16 x i8> addrspace(2)* %35, align 16, !tbaa !0
>  %37 = add i32 %5, %8
>  %38 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %36, i32 0, i32 %37)
>  %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 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 1
>  %44 = load <16 x i8>, <16 x i8> addrspace(2)* %43, align 16, !tbaa !0
>  %45 = add i32 %5, %8
>  %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 = fmul float %15, %47
>  %50 = fmul float %16, %48
>  %51 = fadd float %50, %49
>  %52 = fadd float %51, %17
>  %53 = fmul float %18, %47
>  %54 = fmul float %19, %48
>  %55 = fadd float %54, %53
>  %56 = fadd float %55, %20
>  %57 = fmul float %21, %47
>  %58 = fmul float %22, %48
>  %59 = fadd float %58, %57
>  %60 = fadd float %59, %23
>  %61 = fmul float %24, %47
>  %62 = fmul float %25, %48
>  %63 = fadd float %62, %61
>  %64 = fadd float %63, %26
>  %65 = fmul float %27, %47
>  %66 = fmul float %28, %48
>  %67 = fadd float %66, %65
>  %68 = fadd float %67, %29
>  %69 = fmul float %30, %47
>  %70 = fmul float %31, %48
>  %71 = fadd float %70, %69
>  %72 = fadd float %71, %32
>  %73 = fmul float %41, %14
>  %74 = fsub float %73, %42
>  %75 = fmul float %33, %42
>  %76 = fadd float %75, %39
>  %77 = fmul float %34, %42
>  %78 = fsub float %77, %40
>  %79 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 16
>  %80 = load <16 x i8>, <16 x i8> addrspace(2)* %79, align 16, !tbaa !0
>  %81 = call float @llvm.SI.load.const(<16 x i8> %80, i32 0)
>  %82 = fmul float %81, %39
>  %83 = call float @llvm.SI.load.const(<16 x i8> %80, i32 4)
>  %84 = fmul float %83, %40
>  %85 = fadd float %82, %84
>  %86 = call float @llvm.SI.load.const(<16 x i8> %80, i32 8)
>  %87 = fmul float %86, %41
>  %88 = fadd float %85, %87
>  %89 = call float @llvm.SI.load.const(<16 x i8> %80, i32 12)
>  %90 = fmul float %89, %42
>  %91 = fadd float %88, %90
>  %92 = call float @llvm.SI.load.const(<16 x i8> %80, i32 16)
>  %93 = fmul float %92, %39
>  %94 = call float @llvm.SI.load.const(<16 x i8> %80, i32 20)
>  %95 = fmul float %94, %40
>  %96 = fadd float %93, %95
>  %97 = call float @llvm.SI.load.const(<16 x i8> %80, i32 24)
>  %98 = fmul float %97, %41
>  %99 = fadd float %96, %98
>  %100 = call float @llvm.SI.load.const(<16 x i8> %80, i32 28)
>  %101 = fmul float %100, %42
>  %102 = fadd float %99, %101
>  %103 = call float @llvm.SI.load.const(<16 x i8> %80, i32 32)
>  %104 = fmul float %103, %39
>  %105 = call float @llvm.SI.load.const(<16 x i8> %80, i32 36)
>  %106 = fmul float %105, %40
>  %107 = fadd float %104, %106
>  %108 = call float @llvm.SI.load.const(<16 x i8> %80, i32 40)
>  %109 = fmul float %108, %41
>  %110 = fadd float %107, %109
>  %111 = call float @llvm.SI.load.const(<16 x i8> %80, i32 44)
>  %112 = fmul float %111, %42
>  %113 = fadd float %110, %112
>  %114 = call float @llvm.SI.load.const(<16 x i8> %80, i32 48)
>  %115 = fmul float %114, %39
>  %116 = call float @llvm.SI.load.const(<16 x i8> %80, i32 52)
>  %117 = fmul float %116, %40
>  %118 = fadd float %115, %117
>  %119 = call float @llvm.SI.load.const(<16 x i8> %80, i32 56)
>  %120 = fmul float %119, %41
>  %121 = fadd float %118, %120
>  %122 = call float @llvm.SI.load.const(<16 x i8> %80, i32 60)
>  %123 = fmul float %122, %42
>  %124 = fadd float %121, %123
>  %125 = call float @llvm.SI.load.const(<16 x i8> %80, i32 64)
>  %126 = fmul float %125, %39
>  %127 = call float @llvm.SI.load.const(<16 x i8> %80, i32 68)
>  %128 = fmul float %127, %40
>  %129 = fadd float %126, %128
>  %130 = call float @llvm.SI.load.const(<16 x i8> %80, i32 72)
>  %131 = fmul float %130, %41
>  %132 = fadd float %129, %131
>  %133 = call float @llvm.SI.load.const(<16 x i8> %80, i32 76)
>  %134 = fmul float %133, %42
>  %135 = fadd float %132, %134
>  %136 = call float @llvm.SI.load.const(<16 x i8> %80, i32 80)
>  %137 = fmul float %136, %39
>  %138 = call float @llvm.SI.load.const(<16 x i8> %80, i32 84)
>  %139 = fmul float %138, %40
>  %140 = fadd float %137, %139
>  %141 = call float @llvm.SI.load.const(<16 x i8> %80, i32 88)
>  %142 = fmul float %141, %41
>  %143 = fadd float %140, %142
>  %144 = call float @llvm.SI.load.const(<16 x i8> %80, i32 92)
>  %145 = fmul float %144, %42
>  %146 = fadd float %143, %145
>  %147 = call float @llvm.SI.load.const(<16 x i8> %80, i32 96)
>  %148 = fmul float %147, %39
>  %149 = call float @llvm.SI.load.const(<16 x i8> %80, i32 100)
>  %150 = fmul float %149, %40
>  %151 = fadd float %148, %150
>  %152 = call float @llvm.SI.load.const(<16 x i8> %80, i32 104)
>  %153 = fmul float %152, %41
>  %154 = fadd float %151, %153
>  %155 = call float @llvm.SI.load.const(<16 x i8> %80, i32 108)
>  %156 = fmul float %155, %42
>  %157 = fadd float %154, %156
>  %158 = call float @llvm.SI.load.const(<16 x i8> %80, i32 112)
>  %159 = fmul float %158, %39
>  %160 = call float @llvm.SI.load.const(<16 x i8> %80, i32 116)
>  %161 = fmul float %160, %40
>  %162 = fadd float %159, %161
>  %163 = call float @llvm.SI.load.const(<16 x i8> %80, i32 120)
>  %164 = fmul float %163, %41
>  %165 = fadd float %162, %164
>  %166 = call float @llvm.SI.load.const(<16 x i8> %80, i32 124)
>  %167 = fmul float %166, %42
>  %168 = fadd float %165, %167
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %47, float %48, float %52, float %56)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %60, float %64, float %68, float %72)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 12, i32 0, float %76, float %78, float %74, float %42)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 13, i32 0, float %91, float %102, float %113, float %124)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 14, i32 0, float %135, float %146, float %157, 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
>
>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[4:7], s[8:9], 0x0                    ; C0820900
>	s_load_dwordx4 s[12:15], s[8:9], 0x4                  ; C0860904
>	v_add_i32_e32 v0, vcc, s10, v0                        ; 4A00000A
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[4:7], 0 idxen   ; E00C2000 80010100
>	buffer_load_format_xyzw v[5:8], v0, s[12:15], 0 idxen ; E00C2000 80030500
>	s_load_dwordx4 s[4:7], s[2:3], 0x0                    ; C0820300
>	s_load_dwordx4 s[0:3], s[2:3], 0x40                   ; C0800340
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	s_buffer_load_dword s8, s[4:7], 0x2                   ; C2040502
>	s_buffer_load_dword s9, s[4:7], 0xc0                  ; C20485C0
>	s_buffer_load_dword s10, s[4:7], 0xc1                 ; C20505C1
>	s_buffer_load_dword s11, s[4:7], 0xc3                 ; C20585C3
>	s_buffer_load_dword s12, s[4:7], 0xc4                 ; C20605C4
>	s_buffer_load_dword s13, s[4:7], 0xc5                 ; C20685C5
>	s_buffer_load_dword s14, s[4:7], 0xc7                 ; C20705C7
>	s_buffer_load_dword s15, s[4:7], 0xc8                 ; C20785C8
>	s_buffer_load_dword s16, s[4:7], 0xc9                 ; C20805C9
>	s_buffer_load_dword s17, s[4:7], 0xcb                 ; C20885CB
>	s_buffer_load_dword s18, s[4:7], 0xcc                 ; C20905CC
>	s_buffer_load_dword s19, s[4:7], 0xcd                 ; C20985CD
>	s_buffer_load_dword s20, s[4:7], 0xcf                 ; C20A05CF
>	s_buffer_load_dword s21, s[4:7], 0xd0                 ; C20A85D0
>	s_buffer_load_dword s22, s[4:7], 0xd1                 ; C20B05D1
>	s_buffer_load_dword s23, s[4:7], 0xd3                 ; C20B85D3
>	s_buffer_load_dword s24, s[4:7], 0xd4                 ; C20C05D4
>	s_buffer_load_dword s25, s[4:7], 0xd5                 ; C20C85D5
>	s_buffer_load_dword s26, s[4:7], 0xd7                 ; C20D05D7
>	s_buffer_load_dword s27, s[4:7], 0xd8                 ; C20D85D8
>	s_buffer_load_dword s4, s[4:7], 0xd9                  ; C20205D9
>	s_buffer_load_dword s5, s[0:3], 0x0                   ; C2028100
>	s_buffer_load_dword s6, s[0:3], 0x1                   ; C2030101
>	s_buffer_load_dword s7, s[0:3], 0x2                   ; C2038102
>	s_buffer_load_dword s28, s[0:3], 0x3                  ; C20E0103
>	s_buffer_load_dword s29, s[0:3], 0x4                  ; C20E8104
>	s_buffer_load_dword s30, s[0:3], 0x5                  ; C20F0105
>	s_buffer_load_dword s31, s[0:3], 0x6                  ; C20F8106
>	s_buffer_load_dword s32, s[0:3], 0x7                  ; C2100107
>	s_buffer_load_dword s33, s[0:3], 0x8                  ; C2108108
>	s_buffer_load_dword s34, s[0:3], 0x9                  ; C2110109
>	s_buffer_load_dword s35, s[0:3], 0xa                  ; C211810A
>	s_buffer_load_dword s36, s[0:3], 0xb                  ; C212010B
>	s_buffer_load_dword s37, s[0:3], 0xc                  ; C212810C
>	s_buffer_load_dword s38, s[0:3], 0xd                  ; C213010D
>	s_buffer_load_dword s39, s[0:3], 0xe                  ; C213810E
>	s_buffer_load_dword s40, s[0:3], 0xf                  ; C214010F
>	s_buffer_load_dword s41, s[0:3], 0x10                 ; C2148110
>	s_buffer_load_dword s42, s[0:3], 0x11                 ; C2150111
>	s_buffer_load_dword s43, s[0:3], 0x12                 ; C2158112
>	s_buffer_load_dword s44, s[0:3], 0x13                 ; C2160113
>	s_buffer_load_dword s45, s[0:3], 0x14                 ; C2168114
>	s_buffer_load_dword s46, s[0:3], 0x15                 ; C2170115
>	s_buffer_load_dword s47, s[0:3], 0x16                 ; C2178116
>	s_buffer_load_dword s48, s[0:3], 0x17                 ; C2180117
>	s_buffer_load_dword s49, s[0:3], 0x18                 ; C2188118
>	s_buffer_load_dword s50, s[0:3], 0x19                 ; C2190119
>	s_buffer_load_dword s51, s[0:3], 0x1a                 ; C219811A
>	s_buffer_load_dword s52, s[0:3], 0x1b                 ; C21A011B
>	s_buffer_load_dword s53, s[0:3], 0x1c                 ; C21A811C
>	s_buffer_load_dword s54, s[0:3], 0x1d                 ; C21B011D
>	s_buffer_load_dword s55, s[0:3], 0x1e                 ; C21B811E
>	s_buffer_load_dword s0, s[0:3], 0x1f                  ; C200011F
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mov_b32_e32 v0, s11                                 ; 7E00020B
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mov_b32_e32 v7, s14                                 ; 7E0E020E
>	v_mov_b32_e32 v8, s17                                 ; 7E100211
>	v_mov_b32_e32 v9, s20                                 ; 7E120214
>	v_mov_b32_e32 v10, s23                                ; 7E140217
>	v_mov_b32_e32 v11, s26                                ; 7E16021A
>	v_mac_f32_e32 v0, s9, v5                              ; 3E000A09
>	v_mac_f32_e32 v7, s12, v5                             ; 3E0E0A0C
>	v_mac_f32_e32 v8, s15, v5                             ; 3E100A0F
>	v_mac_f32_e32 v9, s18, v5                             ; 3E120A12
>	v_mac_f32_e32 v10, s21, v5                            ; 3E140A15
>	v_mac_f32_e32 v11, s24, v5                            ; 3E160A18
>	v_mac_f32_e32 v0, s10, v6                             ; 3E000C0A
>	v_mac_f32_e32 v7, s13, v6                             ; 3E0E0C0D
>	v_mac_f32_e32 v8, s16, v6                             ; 3E100C10
>	v_mac_f32_e32 v9, s19, v6                             ; 3E120C13
>	exp 15, 32, 0, 0, 0, v5, v6, v0, v7                   ; F800020F 07000605
>	v_mac_f32_e32 v10, s22, v6                            ; 3E140C16
>	v_mac_f32_e32 v11, s25, v6                            ; 3E160C19
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mad_f32 v0, v3, s8, -v4                             ; D2820000 84101103
>	v_mad_f32 v5, s27, v4, v1                             ; D2820005 0406081B
>	v_mad_f32 v6, s4, v4, -v2                             ; D2820006 840A0804
>	v_mul_f32_e32 v7, s6, v2                              ; 100E0406
>	exp 15, 33, 0, 0, 0, v8, v9, v10, v11                 ; F800021F 0B0A0908
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mul_f32_e32 v8, s30, v2                             ; 1010041E
>	v_mac_f32_e32 v7, s5, v1                              ; 3E0E0205
>	exp 15, 12, 0, 0, 0, v5, v6, v0, v4                   ; F80000CF 04000605
>	v_mac_f32_e32 v8, s29, v1                             ; 3E10021D
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mul_f32_e32 v0, s34, v2                             ; 10000422
>	v_mac_f32_e32 v0, s33, v1                             ; 3E000221
>	v_mul_f32_e32 v5, s38, v2                             ; 100A0426
>	v_mac_f32_e32 v5, s37, v1                             ; 3E0A0225
>	v_mul_f32_e32 v6, s42, v2                             ; 100C042A
>	v_mul_f32_e32 v9, s46, v2                             ; 1012042E
>	v_mul_f32_e32 v10, s50, v2                            ; 10140432
>	v_mul_f32_e32 v2, s54, v2                             ; 10040436
>	v_mac_f32_e32 v6, s41, v1                             ; 3E0C0229
>	v_mac_f32_e32 v9, s45, v1                             ; 3E12022D
>	v_mac_f32_e32 v10, s49, v1                            ; 3E140231
>	v_mac_f32_e32 v2, s53, v1                             ; 3E040235
>	v_mac_f32_e32 v7, s7, v3                              ; 3E0E0607
>	v_mac_f32_e32 v8, s31, v3                             ; 3E10061F
>	v_mac_f32_e32 v0, s35, v3                             ; 3E000623
>	v_mac_f32_e32 v5, s39, v3                             ; 3E0A0627
>	v_mac_f32_e32 v6, s43, v3                             ; 3E0C062B
>	v_mac_f32_e32 v9, s47, v3                             ; 3E12062F
>	v_mac_f32_e32 v10, s51, v3                            ; 3E140633
>	v_mac_f32_e32 v2, s55, v3                             ; 3E040637
>	v_mac_f32_e32 v7, s28, v4                             ; 3E0E081C
>	v_mac_f32_e32 v8, s32, v4                             ; 3E100820
>	v_mac_f32_e32 v0, s36, v4                             ; 3E000824
>	v_mac_f32_e32 v5, s40, v4                             ; 3E0A0828
>	v_mac_f32_e32 v6, s44, v4                             ; 3E0C082C
>	v_mac_f32_e32 v9, s48, v4                             ; 3E120830
>	v_mac_f32_e32 v10, s52, v4                            ; 3E140834
>	v_mac_f32_e32 v2, s0, v4                              ; 3E040800
>	exp 15, 13, 0, 0, 0, v7, v8, v0, v5                   ; F80000DF 05000807
>	exp 15, 14, 0, 1, 0, v6, v9, v10, v2                  ; F80008EF 020A0906
>	s_endpgm                                              ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 64
>VGPRS: 12
>Code Size: 544 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
>  clamp_color = 0
>FRAG
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL IN[1], GENERIC[1], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SAMP[1]
>DCL SAMP[2]
>DCL SAMP[3]
>DCL SAMP[4]
>DCL SAMP[5]
>DCL SVIEW[0], 2D, FLOAT
>DCL SVIEW[1], 2D, FLOAT
>DCL SVIEW[2], 2D, FLOAT
>DCL SVIEW[3], 2D, FLOAT
>DCL SVIEW[4], 2D, FLOAT
>DCL SVIEW[5], 2D, FLOAT
>DCL CONST[0..3]
>DCL TEMP[0..9], LOCAL
>IMM[0] FLT32 {    1.0000,     0.1000,     0.7500,     0.0800}
>IMM[1] FLT32 {    6.6667,    -2.0000,     3.0000,    -0.1000}
>IMM[2] FLT32 {    0.3000,     0.5900,     0.1100,    10.0000}
>IMM[3] FLT32 {    0.2500,    -1.0000,    -0.0800,    14.2857}
>IMM[4] FLT32 {    0.0050,     2.0000,     0.0000,     0.0000}
>  0: SQRT TEMP[0].x, CONST[3].wwww
>  1: ADD TEMP[1].x, -CONST[1].xxxx, CONST[0].wwww
>  2: ADD TEMP[2].x, -CONST[1].yyyy, CONST[1].wwww
>  3: MOV TEMP[1].y, TEMP[2].xxxx
>  4: ADD TEMP[3].x, -CONST[1].zzzz, CONST[2].wwww
>  5: MOV TEMP[1].z, TEMP[3].xxxx
>  6: MAD TEMP[0].xyz, TEMP[0].xxxx, TEMP[1].xyzz, CONST[1].xyzz
>  7: ADD TEMP[1].xyz, -CONST[1].xyzz, CONST[2].xyzz
>  8: MAD TEMP[2].xyz, CONST[3].wwww, TEMP[1].xyzz, CONST[1].xyzz
>  9: MOV TEMP[3].xy, IN[1].zwww
> 10: TEX TEMP[3], TEMP[3], SAMP[4], 2D
> 11: MUL TEMP[4].x, TEMP[3].yyyy, TEMP[3].xxxx
> 12: MUL TEMP[5].xy, IMM[0].zyyy, CONST[3].wwww
> 13: MOV TEMP[6].xy, IN[0].xyyy
> 14: TEX TEMP[6].xyz, TEMP[6], SAMP[0], 2D
> 15: MAD TEMP[7].x, TEMP[6].xxxx, TEMP[6].yyyy, -TEMP[5].yyyy
> 16: MOV_SAT TEMP[7].x, TEMP[7].xxxx
> 17: MAD TEMP[4].x, TEMP[4].xxxx, -TEMP[3].zzzz, TEMP[7].xxxx
> 18: ADD TEMP[4].x, TEMP[4].xxxx, IMM[0].wwww
> 19: MUL TEMP[4].x, TEMP[4].xxxx, IMM[1].xxxx
> 20: MOV_SAT TEMP[4].x, TEMP[4].xxxx
> 21: MAD TEMP[7].x, TEMP[4].xxxx, IMM[1].yyyy, IMM[1].zzzz
> 22: MUL TEMP[4].x, TEMP[4].xxxx, TEMP[4].xxxx
> 23: MUL TEMP[4].x, TEMP[4].xxxx, TEMP[7].xxxx
> 24: LRP TEMP[7].xyz, TEMP[4].xxxx, TEMP[2].xyzz, TEMP[0].xyzz
> 25: MOV TEMP[8].xy, IN[0].zwww
> 26: TEX TEMP[8].xyz, TEMP[8], SAMP[5], 2D
> 27: MUL TEMP[0].xyz, TEMP[8].xyzz, TEMP[7].xyzz
> 28: DP3 TEMP[8].x, TEMP[8].xyzz, IMM[2].xyzz
> 29: MAD TEMP[2].xyz, CONST[0].xyzz, TEMP[8].xxxx, -TEMP[0].xyzz
> 30: MOV TEMP[8].xy, IN[1].xyyy
> 31: TEX TEMP[8].y, TEMP[8], SAMP[1], 2D
> 32: MUL TEMP[8].x, TEMP[6].yyyy, TEMP[8].yyyy
> 33: MUL TEMP[9].x, TEMP[6].xxxx, TEMP[6].xxxx
> 34: MUL TEMP[8].x, TEMP[8].xxxx, TEMP[9].xxxx
> 35: MAD TEMP[8].x, TEMP[8].xxxx, CONST[3].wwww, IMM[1].wwww
> 36: MUL TEMP[8].x, TEMP[8].xxxx, IMM[2].wwww
> 37: MOV_SAT TEMP[8].x, TEMP[8].xxxx
> 38: MAD TEMP[9].x, TEMP[8].xxxx, IMM[1].yyyy, IMM[1].zzzz
> 39: MUL TEMP[8].x, TEMP[8].xxxx, TEMP[8].xxxx
> 40: MUL TEMP[8].x, TEMP[8].xxxx, TEMP[9].xxxx
> 41: MAD TEMP[0].xyz, TEMP[8].xxxx, TEMP[2].xyzz, TEMP[0].xyzz
> 42: ADD TEMP[2].x, -TEMP[6].xxxx, IMM[0].xxxx
> 43: MUL TEMP[2].x, TEMP[2].xxxx, TEMP[2].xxxx
> 44: MUL TEMP[2].x, TEMP[2].xxxx, TEMP[2].xxxx
> 45: MAD TEMP[2].x, TEMP[2].xxxx, IMM[3].xxxx, TEMP[5].xxxx
> 46: ADD TEMP[3], TEMP[3], IMM[3].yyyy
> 47: MAD TEMP[7], TEMP[2].xxxx, TEMP[3], IMM[0].xxxx
> 48: MUL TEMP[0].xyz, TEMP[0].xyzz, TEMP[7].xyzz
> 49: DP3 TEMP[3].x, TEMP[0].xyzz, IMM[2].xyzz
> 50: ADD TEMP[2].x, TEMP[3].xxxx, IMM[3].zzzz
> 51: MUL TEMP[2].x, TEMP[2].xxxx, IMM[3].wwww
> 52: MOV_SAT TEMP[3].x, TEMP[2].xxxx
> 53: MAD TEMP[5].x, TEMP[3].xxxx, IMM[1].yyyy, IMM[1].zzzz
> 54: MUL TEMP[2].x, TEMP[3].xxxx, TEMP[3].xxxx
> 55: MAD TEMP[2].x, TEMP[5].xxxx, -TEMP[2].xxxx, IMM[0].xxxx
> 56: MAD TEMP[2].xyz, TEMP[2].xxxx, IMM[4].xxxx, TEMP[0].xyzz
> 57: MUL TEMP[2].xyz, TEMP[2].xyzz, TEMP[6].zzzz
> 58: MAD TEMP[0].xyz, TEMP[2].xyzz, IMM[4].yyyy, TEMP[0].xyzz
> 59: MOV_SAT TEMP[3].xyz, TEMP[0].xyzz
> 60: MUL TEMP[2].xyz, TEMP[6].yyyy, TEMP[3].xyzz
> 61: MOV TEMP[5].xy, IN[0].xyyy
> 62: TEX TEMP[5], TEMP[5], SAMP[2], 2D
> 63: MAD TEMP[0].xyz, TEMP[3].xyzz, -TEMP[6].yyyy, TEMP[5].xyzz
> 64: MUL TEMP[1].x, TEMP[6].yyyy, CONST[3].zzzz
> 65: MUL TEMP[1].x, TEMP[7].wwww, TEMP[1].xxxx
> 66: MOV TEMP[3].xy, IN[0].xyyy
> 67: TEX TEMP[3].x, TEMP[3], SAMP[3], 2D
> 68: ADD TEMP[3].x, -TEMP[3].xxxx, IMM[0].xxxx
> 69: MAD TEMP[2].xyz, TEMP[3].xxxx, TEMP[0].xyzz, TEMP[2].xyzz
> 70: MAD TEMP[0].x, TEMP[8].xxxx, -CONST[3].wwww, IMM[0].xxxx
> 71: MUL TEMP[6].x, TEMP[0].xxxx, TEMP[4].xxxx
> 72: MAD TEMP[0].x, TEMP[4].xxxx, -TEMP[0].xxxx, IMM[0].xxxx
> 73: MAD TEMP[0].x, TEMP[8].xxxx, TEMP[0].xxxx, TEMP[6].xxxx
> 74: MUL TEMP[0].x, TEMP[0].xxxx, TEMP[1].xxxx
> 75: MAD TEMP[1].x, TEMP[0].xxxx, -CONST[3].xxxx, TEMP[5].wwww
> 76: MUL TEMP[0].x, TEMP[0].xxxx, CONST[3].xxxx
> 77: MAD TEMP[0].x, TEMP[3].xxxx, TEMP[1].xxxx, TEMP[0].xxxx
> 78: MOV TEMP[2].w, TEMP[0].xxxx
> 79: MOV OUT[0], TEMP[2]
> 80: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %24 = load <16 x i8>, <16 x i8> addrspace(2)* %23, align 16, !tbaa !0
>  %25 = call float @llvm.SI.load.const(<16 x i8> %24, i32 0)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %24, i32 4)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %24, i32 8)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %24, i32 12)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %24, i32 16)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %24, i32 20)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %24, i32 24)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %24, i32 28)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %24, i32 32)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %24, i32 36)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %24, i32 40)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %24, i32 44)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %24, i32 48)
>  %38 = call float @llvm.SI.load.const(<16 x i8> %24, i32 56)
>  %39 = call float @llvm.SI.load.const(<16 x i8> %24, i32 60)
>  %40 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %41 = load <8 x i32>, <8 x i32> addrspace(2)* %40, align 32, !tbaa !0
>  %42 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %43 = load <4 x i32>, <4 x i32> addrspace(2)* %42, align 16, !tbaa !0
>  %44 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 1
>  %45 = load <8 x i32>, <8 x i32> addrspace(2)* %44, align 32, !tbaa !0
>  %46 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 1
>  %47 = load <4 x i32>, <4 x i32> addrspace(2)* %46, align 16, !tbaa !0
>  %48 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 2
>  %49 = load <8 x i32>, <8 x i32> addrspace(2)* %48, align 32, !tbaa !0
>  %50 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 2
>  %51 = load <4 x i32>, <4 x i32> addrspace(2)* %50, align 16, !tbaa !0
>  %52 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 3
>  %53 = load <8 x i32>, <8 x i32> addrspace(2)* %52, align 32, !tbaa !0
>  %54 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 3
>  %55 = load <4 x i32>, <4 x i32> addrspace(2)* %54, align 16, !tbaa !0
>  %56 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 4
>  %57 = load <8 x i32>, <8 x i32> addrspace(2)* %56, align 32, !tbaa !0
>  %58 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 4
>  %59 = load <4 x i32>, <4 x i32> addrspace(2)* %58, align 16, !tbaa !0
>  %60 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 5
>  %61 = load <8 x i32>, <8 x i32> addrspace(2)* %60, align 32, !tbaa !0
>  %62 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 5
>  %63 = load <4 x i32>, <4 x i32> addrspace(2)* %62, align 16, !tbaa !0
>  %64 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %8)
>  %65 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %8)
>  %66 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %6, <2 x i32> %8)
>  %67 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %6, <2 x i32> %8)
>  %68 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %6, <2 x i32> %8)
>  %69 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %6, <2 x i32> %8)
>  %70 = call float @llvm.SI.fs.interp(i32 2, i32 1, i32 %6, <2 x i32> %8)
>  %71 = call float @llvm.SI.fs.interp(i32 3, i32 1, i32 %6, <2 x i32> %8)
>  %72 = call float @llvm.sqrt.f32(float %39)
>  %73 = fsub float %28, %29
>  %74 = fsub float %32, %30
>  %75 = fsub float %36, %31
>  %76 = fmul float %72, %73
>  %77 = fadd float %76, %29
>  %78 = fmul float %72, %74
>  %79 = fadd float %78, %30
>  %80 = fmul float %72, %75
>  %81 = fadd float %80, %31
>  %82 = fsub float %33, %29
>  %83 = fsub float %34, %30
>  %84 = fsub float %35, %31
>  %85 = fmul float %39, %82
>  %86 = fadd float %85, %29
>  %87 = fmul float %39, %83
>  %88 = fadd float %87, %30
>  %89 = fmul float %39, %84
>  %90 = fadd float %89, %31
>  %91 = bitcast float %70 to i32
>  %92 = bitcast float %71 to i32
>  %93 = insertelement <2 x i32> undef, i32 %91, i32 0
>  %94 = insertelement <2 x i32> %93, i32 %92, i32 1
>  %95 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %94, <8 x i32> %57, <4 x i32> %59, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %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 %97, %96
>  %101 = fmul float %39, 7.500000e-01
>  %102 = fmul float %39, 0x3FB99999A0000000
>  %103 = bitcast float %64 to i32
>  %104 = bitcast float %65 to i32
>  %105 = insertelement <2 x i32> undef, i32 %103, i32 0
>  %106 = insertelement <2 x i32> %105, i32 %104, i32 1
>  %107 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %106, <8 x i32> %41, <4 x i32> %43, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %108 = extractelement <4 x float> %107, i32 0
>  %109 = extractelement <4 x float> %107, i32 1
>  %110 = extractelement <4 x float> %107, i32 2
>  %111 = fmul float %108, %109
>  %112 = fsub float %111, %102
>  %113 = call float @llvm.AMDIL.clamp.(float %112, float 0.000000e+00, float 1.000000e+00)
>  %114 = fmul float %98, %100
>  %115 = fsub float %113, %114
>  %116 = fadd float %115, 0x3FB47AE140000000
>  %117 = fmul float %116, 0x401AAAAAA0000000
>  %118 = call float @llvm.AMDIL.clamp.(float %117, float 0.000000e+00, float 1.000000e+00)
>  %119 = fmul float %118, -2.000000e+00
>  %120 = fadd float %119, 3.000000e+00
>  %121 = fmul float %118, %118
>  %122 = fmul float %121, %120
>  %123 = fsub float 1.000000e+00, %122
>  %124 = fmul float %86, %122
>  %125 = fmul float %77, %123
>  %126 = fadd float %124, %125
>  %127 = fsub float 1.000000e+00, %122
>  %128 = fmul float %88, %122
>  %129 = fmul float %79, %127
>  %130 = fadd float %128, %129
>  %131 = fsub float 1.000000e+00, %122
>  %132 = fmul float %90, %122
>  %133 = fmul float %81, %131
>  %134 = fadd float %132, %133
>  %135 = bitcast float %66 to i32
>  %136 = bitcast float %67 to i32
>  %137 = insertelement <2 x i32> undef, i32 %135, i32 0
>  %138 = insertelement <2 x i32> %137, i32 %136, i32 1
>  %139 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %138, <8 x i32> %61, <4 x i32> %63, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %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 = fmul float %140, %126
>  %144 = fmul float %141, %130
>  %145 = fmul float %142, %134
>  %146 = fmul float %140, 0x3FD3333340000000
>  %147 = fmul float %141, 0x3FE2E147A0000000
>  %148 = fadd float %147, %146
>  %149 = fmul float %142, 0x3FBC28F5C0000000
>  %150 = fadd float %148, %149
>  %151 = fmul float %25, %150
>  %152 = fsub float %151, %143
>  %153 = fmul float %26, %150
>  %154 = fsub float %153, %144
>  %155 = fmul float %27, %150
>  %156 = fsub float %155, %145
>  %157 = bitcast float %68 to i32
>  %158 = bitcast float %69 to i32
>  %159 = insertelement <2 x i32> undef, i32 %157, i32 0
>  %160 = insertelement <2 x i32> %159, i32 %158, i32 1
>  %161 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %160, <8 x i32> %45, <4 x i32> %47, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %162 = extractelement <4 x float> %161, i32 1
>  %163 = fmul float %109, %162
>  %164 = fmul float %108, %108
>  %165 = fmul float %163, %164
>  %166 = fmul float %165, %39
>  %167 = fadd float %166, 0xBFB99999A0000000
>  %168 = fmul float %167, 1.000000e+01
>  %169 = call float @llvm.AMDIL.clamp.(float %168, float 0.000000e+00, float 1.000000e+00)
>  %170 = fmul float %169, -2.000000e+00
>  %171 = fadd float %170, 3.000000e+00
>  %172 = fmul float %169, %169
>  %173 = fmul float %172, %171
>  %174 = fmul float %173, %152
>  %175 = fadd float %174, %143
>  %176 = fmul float %173, %154
>  %177 = fadd float %176, %144
>  %178 = fmul float %173, %156
>  %179 = fadd float %178, %145
>  %180 = fsub float 1.000000e+00, %108
>  %181 = fmul float %180, %180
>  %182 = fmul float %181, %181
>  %183 = fmul float %182, 2.500000e-01
>  %184 = fadd float %183, %101
>  %185 = fadd float %96, -1.000000e+00
>  %186 = fadd float %97, -1.000000e+00
>  %187 = fadd float %98, -1.000000e+00
>  %188 = fadd float %99, -1.000000e+00
>  %189 = fmul float %184, %185
>  %190 = fadd float %189, 1.000000e+00
>  %191 = fmul float %184, %186
>  %192 = fadd float %191, 1.000000e+00
>  %193 = fmul float %184, %187
>  %194 = fadd float %193, 1.000000e+00
>  %195 = fmul float %184, %188
>  %196 = fadd float %195, 1.000000e+00
>  %197 = fmul float %175, %190
>  %198 = fmul float %177, %192
>  %199 = fmul float %179, %194
>  %200 = fmul float %197, 0x3FD3333340000000
>  %201 = fmul float %198, 0x3FE2E147A0000000
>  %202 = fadd float %201, %200
>  %203 = fmul float %199, 0x3FBC28F5C0000000
>  %204 = fadd float %202, %203
>  %205 = fadd float %204, 0xBFB47AE140000000
>  %206 = fmul float %205, 0x402C924920000000
>  %207 = call float @llvm.AMDIL.clamp.(float %206, float 0.000000e+00, float 1.000000e+00)
>  %208 = fmul float %207, -2.000000e+00
>  %209 = fadd float %208, 3.000000e+00
>  %210 = fmul float %207, %207
>  %211 = fmul float %210, %209
>  %212 = fsub float 1.000000e+00, %211
>  %213 = fmul float %212, 0x3F747AE140000000
>  %214 = fadd float %213, %197
>  %215 = fmul float %212, 0x3F747AE140000000
>  %216 = fadd float %215, %198
>  %217 = fmul float %212, 0x3F747AE140000000
>  %218 = fadd float %217, %199
>  %219 = fmul float %214, %110
>  %220 = fmul float %216, %110
>  %221 = fmul float %218, %110
>  %222 = fmul float %219, 2.000000e+00
>  %223 = fadd float %222, %197
>  %224 = fmul float %220, 2.000000e+00
>  %225 = fadd float %224, %198
>  %226 = fmul float %221, 2.000000e+00
>  %227 = fadd float %226, %199
>  %228 = call float @llvm.AMDIL.clamp.(float %223, float 0.000000e+00, float 1.000000e+00)
>  %229 = call float @llvm.AMDIL.clamp.(float %225, float 0.000000e+00, float 1.000000e+00)
>  %230 = call float @llvm.AMDIL.clamp.(float %227, float 0.000000e+00, float 1.000000e+00)
>  %231 = fmul float %109, %228
>  %232 = fmul float %109, %229
>  %233 = fmul float %109, %230
>  %234 = bitcast float %64 to i32
>  %235 = bitcast float %65 to i32
>  %236 = insertelement <2 x i32> undef, i32 %234, i32 0
>  %237 = insertelement <2 x i32> %236, i32 %235, i32 1
>  %238 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %237, <8 x i32> %49, <4 x i32> %51, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %239 = extractelement <4 x float> %238, i32 0
>  %240 = extractelement <4 x float> %238, i32 1
>  %241 = extractelement <4 x float> %238, i32 2
>  %242 = extractelement <4 x float> %238, i32 3
>  %243 = fmul float %109, %228
>  %244 = fsub float %239, %243
>  %245 = fmul float %109, %229
>  %246 = fsub float %240, %245
>  %247 = fmul float %109, %230
>  %248 = fsub float %241, %247
>  %249 = fmul float %109, %38
>  %250 = fmul float %196, %249
>  %251 = bitcast float %64 to i32
>  %252 = bitcast float %65 to i32
>  %253 = insertelement <2 x i32> undef, i32 %251, i32 0
>  %254 = insertelement <2 x i32> %253, i32 %252, i32 1
>  %255 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %254, <8 x i32> %53, <4 x i32> %55, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %256 = extractelement <4 x float> %255, i32 0
>  %257 = fsub float 1.000000e+00, %256
>  %258 = fmul float %257, %244
>  %259 = fadd float %258, %231
>  %260 = fmul float %257, %246
>  %261 = fadd float %260, %232
>  %262 = fmul float %257, %248
>  %263 = fadd float %262, %233
>  %264 = fmul float %39, %173
>  %265 = fsub float 1.000000e+00, %264
>  %266 = fmul float %265, %122
>  %267 = fmul float %265, %122
>  %268 = fsub float 1.000000e+00, %267
>  %269 = fmul float %173, %268
>  %270 = fadd float %269, %266
>  %271 = fmul float %270, %250
>  %272 = fmul float %37, %271
>  %273 = fsub float %242, %272
>  %274 = fmul float %271, %37
>  %275 = fmul float %257, %273
>  %276 = fadd float %275, %274
>  %277 = call i32 @llvm.SI.packf16(float %259, float %261)
>  %278 = bitcast i32 %277 to float
>  %279 = call i32 @llvm.SI.packf16(float %263, float %276)
>  %280 = bitcast i32 %279 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %278, float %280, float %278, float %280)
>  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.sqrt.f32(float) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #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, s10                                                            ; BEFC030A
>	v_mov_b32_e32 v2, 0x3f400000                                                 ; 7E0402FF 3F400000
>	v_mov_b32_e32 v3, 0xbdcccccd                                                 ; 7E0602FF BDCCCCCD
>	v_mov_b32_e32 v4, 0x40400000                                                 ; 7E0802FF 40400000
>	v_mov_b32_e32 v5, 0x3e99999a                                                 ; 7E0A02FF 3E99999A
>	v_mov_b32_e32 v6, 0x3f170a3d                                                 ; 7E0C02FF 3F170A3D
>	v_mov_b32_e32 v7, 0x3de147ae                                                 ; 7E0E02FF 3DE147AE
>	v_mov_b32_e32 v8, 0xbda3d70a                                                 ; 7E1002FF BDA3D70A
>	v_mov_b32_e32 v9, 0x3ba3d70a                                                 ; 7E1202FF 3BA3D70A
>	v_interp_p1_f32 v10, v0, 0, 0, [m0]                                          ; C8280000
>	v_interp_p2_f32 v10, [v10], v1, 0, 0, [m0]                                   ; C8290001
>	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 v12, v0, 2, 0, [m0]                                          ; C8300200
>	v_interp_p2_f32 v12, [v12], v1, 2, 0, [m0]                                   ; C8310201
>	v_interp_p1_f32 v13, v0, 3, 0, [m0]                                          ; C8340300
>	v_interp_p2_f32 v13, [v13], v1, 3, 0, [m0]                                   ; C8350301
>	v_interp_p1_f32 v14, v0, 0, 1, [m0]                                          ; C8380400
>	v_interp_p2_f32 v14, [v14], v1, 0, 1, [m0]                                   ; C8390401
>	v_interp_p1_f32 v15, v0, 1, 1, [m0]                                          ; C83C0500
>	v_interp_p2_f32 v15, [v15], v1, 1, 1, [m0]                                   ; C83D0501
>	v_interp_p1_f32 v16, v0, 2, 1, [m0]                                          ; C8400600
>	v_interp_p2_f32 v16, [v16], v1, 2, 1, [m0]                                   ; C8410601
>	v_interp_p1_f32 v17, v0, 3, 1, [m0]                                          ; C8440700
>	v_interp_p2_f32 v17, [v17], v1, 3, 1, [m0]                                   ; C8450701
>	s_load_dwordx4 s[28:31], s[2:3], 0x0                                         ; C08E0300
>	s_load_dwordx4 s[56:59], s[4:5], 0x0                                         ; C09C0500
>	s_load_dwordx4 s[32:35], s[4:5], 0x4                                         ; C0900504
>	s_load_dwordx4 s[12:15], s[4:5], 0x8                                         ; C0860508
>	s_load_dwordx4 s[8:11], s[4:5], 0xc                                          ; C084050C
>	s_load_dwordx4 s[24:27], s[4:5], 0x10                                        ; C08C0510
>	s_load_dwordx4 s[44:47], s[4:5], 0x14                                        ; C0960514
>	s_load_dwordx8 s[60:67], s[6:7], 0x0                                         ; C0DE0700
>	s_load_dwordx8 s[36:43], s[6:7], 0x8                                         ; C0D20708
>	s_load_dwordx8 s[16:23], s[6:7], 0x10                                        ; C0C80710
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	s_buffer_load_dword s1, s[28:31], 0x0                                        ; C2009D00
>	s_buffer_load_dword s2, s[28:31], 0x1                                        ; C2011D01
>	s_buffer_load_dword s3, s[28:31], 0x2                                        ; C2019D02
>	s_buffer_load_dword s4, s[28:31], 0x3                                        ; C2021D03
>	s_buffer_load_dword s5, s[28:31], 0x4                                        ; C2029D04
>	s_buffer_load_dword s76, s[28:31], 0x5                                       ; C2261D05
>	s_buffer_load_dword s77, s[28:31], 0x6                                       ; C2269D06
>	s_buffer_load_dword s78, s[28:31], 0x7                                       ; C2271D07
>	s_buffer_load_dword s79, s[28:31], 0x8                                       ; C2279D08
>	s_buffer_load_dword s80, s[28:31], 0x9                                       ; C2281D09
>	s_buffer_load_dword s81, s[28:31], 0xa                                       ; C2289D0A
>	s_buffer_load_dword s82, s[28:31], 0xb                                       ; C2291D0B
>	s_buffer_load_dword s0, s[28:31], 0xc                                        ; C2001D0C
>	s_buffer_load_dword s83, s[28:31], 0xe                                       ; C2299D0E
>	s_buffer_load_dword s84, s[28:31], 0xf                                       ; C22A1D0F
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mov_b32_e32 v0, s5                                                         ; 7E000205
>	v_sub_f32_e32 v1, s4, v0                                                     ; 08020004
>	v_mov_b32_e32 v18, s76                                                       ; 7E24024C
>	v_mov_b32_e32 v19, s77                                                       ; 7E26024D
>	s_load_dwordx8 s[68:75], s[6:7], 0x20                                        ; C0E20720
>	s_load_dwordx8 s[48:55], s[6:7], 0x28                                        ; C0D80728
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample v[20:23], 15, 0, 0, 0, 0, 0, 0, 0, v[16:17], s[68:75], s[24:27] ; F0800F00 00D11410
>	s_load_dwordx8 s[24:31], s[6:7], 0x18                                        ; C0CC0718
>	image_sample v[24:26], 7, 0, 0, 0, 0, 0, 0, 0, v[10:11], s[60:67], s[56:59]  ; F0800700 01CF180A
>	image_sample v[27:29], 7, 0, 0, 0, 0, 0, 0, 0, v[12:13], s[48:55], s[44:47]  ; F0800700 016C1B0C
>	image_sample v12, 2, 0, 0, 0, 0, 0, 0, 0, v[14:15], s[36:43], s[32:35]       ; F0800200 01090C0E
>	v_sqrt_f32_e32 v13, s84                                                      ; 7E1A6654
>	v_mad_f32 v1, v13, v1, s5                                                    ; D2820001 0016030D
>	v_sub_f32_e32 v14, s78, v18                                                  ; 081C244E
>	v_mad_f32 v14, v13, v14, s76                                                 ; D282000E 01321D0D
>	v_sub_f32_e32 v15, s82, v19                                                  ; 081E2652
>	v_mad_f32 v13, v13, v15, s77                                                 ; D282000D 01361F0D
>	v_sub_f32_e32 v15, s79, v0                                                   ; 081E004F
>	v_mad_f32 v0, s84, v15, v0                                                   ; D2820000 04021E54
>	v_sub_f32_e32 v15, s80, v18                                                  ; 081E2450
>	v_mad_f32 v15, s84, v15, v18                                                 ; D282000F 044A1E54
>	v_sub_f32_e32 v16, s81, v19                                                  ; 08202651
>	v_mad_f32 v16, s84, v16, v19                                                 ; D2820010 044E2054
>	v_mul_f32_e32 v17, s84, v3                                                   ; 10220654
>	s_waitcnt vmcnt(2)                                                           ; BF8C0772
>	v_mac_f32_e32 v17, v25, v24                                                  ; 3E223119
>	v_add_f32_e64 v17, 0, v17 clamp                                              ; D2060811 00022280
>	v_mul_f32_e32 v18, v20, v21                                                  ; 10242B14
>	v_mad_f32 v17, -v22, v18, v17                                                ; D2820011 24462516
>	v_add_f32_e32 v17, 0x3da3d70a, v17                                           ; 062222FF 3DA3D70A
>	v_mul_f32_e32 v17, 0x40d55555, v17                                           ; 102222FF 40D55555
>	v_add_f32_e64 v17, 0, v17 clamp                                              ; D2060811 00022280
>	v_mad_f32 v18, -2.0, v17, v4                                                 ; D2820012 041222F5
>	v_mul_f32_e32 v17, v17, v17                                                  ; 10222311
>	v_mul_f32_e32 v17, v18, v17                                                  ; 10222312
>	v_mad_f32 v1, -v17, v1, v1                                                   ; D2820001 24060311
>	v_mac_f32_e32 v1, v17, v0                                                    ; 3E020111
>	v_mad_f32 v0, -v17, v14, v14                                                 ; D2820000 243A1D11
>	v_mac_f32_e32 v0, v17, v15                                                   ; 3E001F11
>	v_mad_f32 v13, -v17, v13, v13                                                ; D282000D 24361B11
>	v_mac_f32_e32 v13, v17, v16                                                  ; 3E1A2111
>	s_waitcnt vmcnt(1)                                                           ; BF8C0771
>	v_mul_f32_e32 v1, v1, v27                                                    ; 10023701
>	v_mul_f32_e32 v0, v0, v28                                                    ; 10003900
>	v_mul_f32_e32 v13, v13, v29                                                  ; 101A3B0D
>	v_mul_f32_e32 v14, v5, v27                                                   ; 101C3705
>	v_mac_f32_e32 v14, v6, v28                                                   ; 3E1C3906
>	v_mac_f32_e32 v14, v7, v29                                                   ; 3E1C3B07
>	v_mul_f32_e32 v2, s84, v2                                                    ; 10040454
>	v_sub_f32_e32 v15, 1.0, v24                                                  ; 081E30F2
>	v_mad_f32 v15, -v24, v15, v15                                                ; D282000F 243E1F18
>	v_mul_f32_e32 v15, v15, v15                                                  ; 101E1F0F
>	v_madmk_f32_e32 v2, v15, v2, 0x3e800000                                      ; 4004050F 3E800000
>	v_add_f32_e32 v15, -1.0, v20                                                 ; 061E28F3
>	v_add_f32_e32 v16, -1.0, v21                                                 ; 06202AF3
>	v_add_f32_e32 v18, -1.0, v22                                                 ; 06242CF3
>	v_mad_f32 v19, v20, v2, -v2                                                  ; D2820013 840A0514
>	v_mad_f32 v20, v21, v2, -v2                                                  ; D2820014 840A0515
>	v_mad_f32 v21, v22, v2, -v2                                                  ; D2820015 840A0516
>	v_mad_f32 v22, v23, v2, -v2                                                  ; D2820016 840A0517
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mul_f32_e32 v12, v12, v25                                                  ; 1018330C
>	v_mul_f32_e32 v23, v24, v24                                                  ; 102E3118
>	v_mul_f32_e32 v12, v23, v12                                                  ; 10181917
>	v_mac_f32_e32 v3, s84, v12                                                   ; 3E061854
>	v_mul_f32_e32 v3, 0x41200000, v3                                             ; 100606FF 41200000
>	v_add_f32_e64 v3, 0, v3 clamp                                                ; D2060803 00020680
>	v_mad_f32 v12, -2.0, v3, v4                                                  ; D282000C 041206F5
>	v_mul_f32_e32 v3, v3, v3                                                     ; 10060703
>	v_mul_f32_e32 v3, v12, v3                                                    ; 1006070C
>	v_mad_f32 v12, s1, v14, -v1                                                  ; D282000C 84061C01
>	v_mac_f32_e32 v1, v12, v3                                                    ; 3E02070C
>	v_mad_f32 v12, s2, v14, -v0                                                  ; D282000C 84021C02
>	v_mac_f32_e32 v0, v12, v3                                                    ; 3E00070C
>	v_mad_f32 v12, v1, v19, v1                                                   ; D282000C 04062701
>	v_mul_f32_e32 v5, v5, v12                                                    ; 100A1905
>	v_mad_f32 v12, v0, v20, v0                                                   ; D282000C 04022900
>	v_mac_f32_e32 v5, v6, v12                                                    ; 3E0A1906
>	v_mad_f32 v6, s3, v14, -v13                                                  ; D2820006 84361C03
>	v_mac_f32_e32 v13, v6, v3                                                    ; 3E1A0706
>	v_mad_f32 v6, v2, v15, 1.0                                                   ; D2820006 03CA1F02
>	v_mad_f32 v12, v2, v16, 1.0                                                  ; D282000C 03CA2102
>	v_mad_f32 v2, v2, v18, 1.0                                                   ; D2820002 03CA2502
>	v_mad_f32 v14, v13, v21, v13                                                 ; D282000E 04362B0D
>	v_mac_f32_e32 v5, v7, v14                                                    ; 3E0A1D07
>	v_add_f32_e32 v5, v5, v8                                                     ; 060A1105
>	v_mul_f32_e32 v5, 0x41649249, v5                                             ; 100A0AFF 41649249
>	v_add_f32_e64 v5, 0, v5 clamp                                                ; D2060805 00020A80
>	v_mac_f32_e32 v4, -2.0, v5                                                   ; 3E080AF5
>	v_mul_f32_e32 v5, v5, v5                                                     ; 100A0B05
>	v_mul_f32_e32 v4, v4, v5                                                     ; 10080B04
>	v_mad_f32 v4, -v4, v9, v9                                                    ; D2820004 24261304
>	v_mad_f32 v5, v6, v1, v4                                                     ; D2820005 04120306
>	v_mad_f32 v7, v12, v0, v4                                                    ; D2820007 0412010C
>	v_mad_f32 v4, v2, v13, v4                                                    ; D2820004 04121B02
>	v_mul_f32_e32 v8, v26, v5                                                    ; 10100B1A
>	v_mul_f32_e32 v9, v26, v7                                                    ; 10120F1A
>	v_mul_f32_e32 v14, v26, v4                                                   ; 101C091A
>	v_mac_f32_e32 v8, v26, v5                                                    ; 3E100B1A
>	v_mac_f32_e32 v8, v6, v1                                                     ; 3E100306
>	v_mac_f32_e32 v9, v26, v7                                                    ; 3E120F1A
>	v_mac_f32_e32 v9, v12, v0                                                    ; 3E12010C
>	v_mac_f32_e32 v14, v26, v4                                                   ; 3E1C091A
>	v_mac_f32_e32 v14, v2, v13                                                   ; 3E1C1B02
>	v_add_f32_e64 v0, 0, v8 clamp                                                ; D2060800 00021080
>	v_add_f32_e64 v1, 0, v9 clamp                                                ; D2060801 00021280
>	v_add_f32_e64 v2, 0, v14 clamp                                               ; D2060802 00021C80
>	v_mul_f32_e32 v4, v0, v25                                                    ; 10083300
>	v_mul_f32_e32 v5, v1, v25                                                    ; 100A3301
>	v_mul_f32_e32 v6, v2, v25                                                    ; 100C3302
>	image_sample v[12:15], 15, 0, 0, 0, 0, 0, 0, 0, v[10:11], s[16:23], s[12:15] ; F0800F00 00640C0A
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v0, -v25, v0, v12                                                  ; D2820000 24320119
>	v_mad_f32 v1, -v25, v1, v13                                                  ; D2820001 24360319
>	v_mad_f32 v2, -v25, v2, v14                                                  ; D2820002 243A0519
>	v_mul_f32_e32 v7, s83, v25                                                   ; 100E3253
>	v_mac_f32_e32 v7, v7, v22                                                    ; 3E0E2D07
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample v8, 1, 0, 0, 0, 0, 0, 0, 0, v[10:11], s[24:31], s[8:11]         ; F0800100 0046080A
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_sub_f32_e32 v8, 1.0, v8                                                    ; 081010F2
>	v_mac_f32_e32 v4, v0, v8                                                     ; 3E081100
>	v_mac_f32_e32 v5, v1, v8                                                     ; 3E0A1101
>	v_mac_f32_e32 v6, v2, v8                                                     ; 3E0C1102
>	v_mul_f32_e32 v0, s84, v3                                                    ; 10000654
>	v_mad_f32 v1, -s84, v3, 1.0                                                  ; D2820001 23CA0654
>	v_mad_f32 v0, -v0, v17, v17                                                  ; D2820000 24462300
>	v_mad_f32 v1, -v1, v17, 1.0                                                  ; D2820001 23CA2301
>	v_mac_f32_e32 v0, v1, v3                                                     ; 3E000701
>	v_mul_f32_e32 v0, v7, v0                                                     ; 10000107
>	v_mul_f32_e32 v1, s0, v0                                                     ; 10020000
>	v_mad_f32 v0, -s0, v0, v15                                                   ; D2820000 243E0000
>	v_mac_f32_e32 v1, v0, v8                                                     ; 3E021100
>	v_cvt_pkrtz_f16_f32_e32 v0, v4, v5                                           ; 5E000B04
>	v_cvt_pkrtz_f16_f32_e32 v1, v6, v1                                           ; 5E020306
>	exp 15, 0, 1, 1, 1, v0, v1, v0, v1                                           ; F8001C0F 01000100
>	s_endpgm                                                                     ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 88
>VGPRS: 32
>Code Size: 1020 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
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL OUT[1], CLIPVERTEX
>DCL OUT[2], GENERIC[0]
>DCL OUT[3], GENERIC[1]
>DCL CONST[0..54]
>DCL TEMP[0..3], LOCAL
>  0: MUL TEMP[0].xy, CONST[48].xyyy, IN[1].xyyy
>  1: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
>  2: ADD TEMP[1].x, TEMP[0].xxxx, CONST[48].wwww
>  3: MOV TEMP[1].z, TEMP[1].xxxx
>  4: MUL TEMP[0].xy, CONST[49].xyyy, IN[1].xyyy
>  5: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
>  6: ADD TEMP[2].x, TEMP[0].xxxx, CONST[49].wwww
>  7: MOV TEMP[1].w, TEMP[2].xxxx
>  8: MUL TEMP[0].xy, CONST[50].xyyy, IN[1].xyyy
>  9: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 10: ADD TEMP[2].x, TEMP[0].xxxx, CONST[50].wwww
> 11: MUL TEMP[0].xy, CONST[51].xyyy, IN[1].xyyy
> 12: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 13: ADD TEMP[3].x, TEMP[0].xxxx, CONST[51].wwww
> 14: MOV TEMP[2].y, TEMP[3].xxxx
> 15: MUL TEMP[0].xy, CONST[52].xyyy, IN[1].xyyy
> 16: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 17: ADD TEMP[3].x, TEMP[0].xxxx, CONST[52].wwww
> 18: MOV TEMP[2].z, TEMP[3].xxxx
> 19: MUL TEMP[0].xy, CONST[53].xyyy, IN[1].xyyy
> 20: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 21: ADD TEMP[0].x, TEMP[0].xxxx, CONST[53].wwww
> 22: MOV TEMP[2].w, TEMP[0].xxxx
> 23: MOV TEMP[0].xw, IN[0].xxxw
> 24: MOV TEMP[1].xy, IN[1].xyxx
> 25: MAD TEMP[3].x, IN[0].zzzz, CONST[0].zzzz, -IN[0].wwww
> 26: MOV TEMP[0].z, TEMP[3].xxxx
> 27: MOV TEMP[0].y, -IN[0].yyyy
> 28: MAD TEMP[0].xy, CONST[54].xyyy, IN[0].wwww, TEMP[0].xyyy
> 29: MOV OUT[2], TEMP[1]
> 30: MOV OUT[3], TEMP[2]
> 31: MOV OUT[0], TEMP[0]
> 32: MOV OUT[1], IN[0]
> 33: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = call float @llvm.SI.load.const(<16 x i8> %13, i32 8)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %13, i32 768)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %13, i32 772)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %13, i32 780)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %13, i32 784)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %13, i32 788)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %13, i32 796)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %13, i32 800)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %13, i32 804)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %13, i32 812)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %13, i32 816)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %13, i32 820)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %13, i32 828)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %13, i32 832)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %13, i32 836)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %13, i32 844)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %13, i32 848)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %13, i32 852)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %13, i32 860)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %13, i32 864)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %13, i32 868)
>  %35 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %36 = load <16 x i8>, <16 x i8> addrspace(2)* %35, align 16, !tbaa !0
>  %37 = add i32 %5, %8
>  %38 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %36, i32 0, i32 %37)
>  %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 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 1
>  %44 = load <16 x i8>, <16 x i8> addrspace(2)* %43, align 16, !tbaa !0
>  %45 = add i32 %5, %8
>  %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 = fmul float %15, %47
>  %50 = fmul float %16, %48
>  %51 = fadd float %50, %49
>  %52 = fadd float %51, %17
>  %53 = fmul float %18, %47
>  %54 = fmul float %19, %48
>  %55 = fadd float %54, %53
>  %56 = fadd float %55, %20
>  %57 = fmul float %21, %47
>  %58 = fmul float %22, %48
>  %59 = fadd float %58, %57
>  %60 = fadd float %59, %23
>  %61 = fmul float %24, %47
>  %62 = fmul float %25, %48
>  %63 = fadd float %62, %61
>  %64 = fadd float %63, %26
>  %65 = fmul float %27, %47
>  %66 = fmul float %28, %48
>  %67 = fadd float %66, %65
>  %68 = fadd float %67, %29
>  %69 = fmul float %30, %47
>  %70 = fmul float %31, %48
>  %71 = fadd float %70, %69
>  %72 = fadd float %71, %32
>  %73 = fmul float %41, %14
>  %74 = fsub float %73, %42
>  %75 = fmul float %33, %42
>  %76 = fadd float %75, %39
>  %77 = fmul float %34, %42
>  %78 = fsub float %77, %40
>  %79 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 16
>  %80 = load <16 x i8>, <16 x i8> addrspace(2)* %79, align 16, !tbaa !0
>  %81 = call float @llvm.SI.load.const(<16 x i8> %80, i32 0)
>  %82 = fmul float %81, %39
>  %83 = call float @llvm.SI.load.const(<16 x i8> %80, i32 4)
>  %84 = fmul float %83, %40
>  %85 = fadd float %82, %84
>  %86 = call float @llvm.SI.load.const(<16 x i8> %80, i32 8)
>  %87 = fmul float %86, %41
>  %88 = fadd float %85, %87
>  %89 = call float @llvm.SI.load.const(<16 x i8> %80, i32 12)
>  %90 = fmul float %89, %42
>  %91 = fadd float %88, %90
>  %92 = call float @llvm.SI.load.const(<16 x i8> %80, i32 16)
>  %93 = fmul float %92, %39
>  %94 = call float @llvm.SI.load.const(<16 x i8> %80, i32 20)
>  %95 = fmul float %94, %40
>  %96 = fadd float %93, %95
>  %97 = call float @llvm.SI.load.const(<16 x i8> %80, i32 24)
>  %98 = fmul float %97, %41
>  %99 = fadd float %96, %98
>  %100 = call float @llvm.SI.load.const(<16 x i8> %80, i32 28)
>  %101 = fmul float %100, %42
>  %102 = fadd float %99, %101
>  %103 = call float @llvm.SI.load.const(<16 x i8> %80, i32 32)
>  %104 = fmul float %103, %39
>  %105 = call float @llvm.SI.load.const(<16 x i8> %80, i32 36)
>  %106 = fmul float %105, %40
>  %107 = fadd float %104, %106
>  %108 = call float @llvm.SI.load.const(<16 x i8> %80, i32 40)
>  %109 = fmul float %108, %41
>  %110 = fadd float %107, %109
>  %111 = call float @llvm.SI.load.const(<16 x i8> %80, i32 44)
>  %112 = fmul float %111, %42
>  %113 = fadd float %110, %112
>  %114 = call float @llvm.SI.load.const(<16 x i8> %80, i32 48)
>  %115 = fmul float %114, %39
>  %116 = call float @llvm.SI.load.const(<16 x i8> %80, i32 52)
>  %117 = fmul float %116, %40
>  %118 = fadd float %115, %117
>  %119 = call float @llvm.SI.load.const(<16 x i8> %80, i32 56)
>  %120 = fmul float %119, %41
>  %121 = fadd float %118, %120
>  %122 = call float @llvm.SI.load.const(<16 x i8> %80, i32 60)
>  %123 = fmul float %122, %42
>  %124 = fadd float %121, %123
>  %125 = call float @llvm.SI.load.const(<16 x i8> %80, i32 64)
>  %126 = fmul float %125, %39
>  %127 = call float @llvm.SI.load.const(<16 x i8> %80, i32 68)
>  %128 = fmul float %127, %40
>  %129 = fadd float %126, %128
>  %130 = call float @llvm.SI.load.const(<16 x i8> %80, i32 72)
>  %131 = fmul float %130, %41
>  %132 = fadd float %129, %131
>  %133 = call float @llvm.SI.load.const(<16 x i8> %80, i32 76)
>  %134 = fmul float %133, %42
>  %135 = fadd float %132, %134
>  %136 = call float @llvm.SI.load.const(<16 x i8> %80, i32 80)
>  %137 = fmul float %136, %39
>  %138 = call float @llvm.SI.load.const(<16 x i8> %80, i32 84)
>  %139 = fmul float %138, %40
>  %140 = fadd float %137, %139
>  %141 = call float @llvm.SI.load.const(<16 x i8> %80, i32 88)
>  %142 = fmul float %141, %41
>  %143 = fadd float %140, %142
>  %144 = call float @llvm.SI.load.const(<16 x i8> %80, i32 92)
>  %145 = fmul float %144, %42
>  %146 = fadd float %143, %145
>  %147 = call float @llvm.SI.load.const(<16 x i8> %80, i32 96)
>  %148 = fmul float %147, %39
>  %149 = call float @llvm.SI.load.const(<16 x i8> %80, i32 100)
>  %150 = fmul float %149, %40
>  %151 = fadd float %148, %150
>  %152 = call float @llvm.SI.load.const(<16 x i8> %80, i32 104)
>  %153 = fmul float %152, %41
>  %154 = fadd float %151, %153
>  %155 = call float @llvm.SI.load.const(<16 x i8> %80, i32 108)
>  %156 = fmul float %155, %42
>  %157 = fadd float %154, %156
>  %158 = call float @llvm.SI.load.const(<16 x i8> %80, i32 112)
>  %159 = fmul float %158, %39
>  %160 = call float @llvm.SI.load.const(<16 x i8> %80, i32 116)
>  %161 = fmul float %160, %40
>  %162 = fadd float %159, %161
>  %163 = call float @llvm.SI.load.const(<16 x i8> %80, i32 120)
>  %164 = fmul float %163, %41
>  %165 = fadd float %162, %164
>  %166 = call float @llvm.SI.load.const(<16 x i8> %80, i32 124)
>  %167 = fmul float %166, %42
>  %168 = fadd float %165, %167
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %47, float %48, float %52, float %56)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %60, float %64, float %68, float %72)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 12, i32 0, float %76, float %78, float %74, float %42)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 13, i32 0, float %91, float %102, float %113, float %124)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 14, i32 0, float %135, float %146, float %157, 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
>
>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[4:7], s[8:9], 0x0                    ; C0820900
>	s_load_dwordx4 s[12:15], s[8:9], 0x4                  ; C0860904
>	v_add_i32_e32 v0, vcc, s10, v0                        ; 4A00000A
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[4:7], 0 idxen   ; E00C2000 80010100
>	buffer_load_format_xyzw v[5:8], v0, s[12:15], 0 idxen ; E00C2000 80030500
>	s_load_dwordx4 s[4:7], s[2:3], 0x0                    ; C0820300
>	s_load_dwordx4 s[0:3], s[2:3], 0x40                   ; C0800340
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	s_buffer_load_dword s8, s[4:7], 0x2                   ; C2040502
>	s_buffer_load_dword s9, s[4:7], 0xc0                  ; C20485C0
>	s_buffer_load_dword s10, s[4:7], 0xc1                 ; C20505C1
>	s_buffer_load_dword s11, s[4:7], 0xc3                 ; C20585C3
>	s_buffer_load_dword s12, s[4:7], 0xc4                 ; C20605C4
>	s_buffer_load_dword s13, s[4:7], 0xc5                 ; C20685C5
>	s_buffer_load_dword s14, s[4:7], 0xc7                 ; C20705C7
>	s_buffer_load_dword s15, s[4:7], 0xc8                 ; C20785C8
>	s_buffer_load_dword s16, s[4:7], 0xc9                 ; C20805C9
>	s_buffer_load_dword s17, s[4:7], 0xcb                 ; C20885CB
>	s_buffer_load_dword s18, s[4:7], 0xcc                 ; C20905CC
>	s_buffer_load_dword s19, s[4:7], 0xcd                 ; C20985CD
>	s_buffer_load_dword s20, s[4:7], 0xcf                 ; C20A05CF
>	s_buffer_load_dword s21, s[4:7], 0xd0                 ; C20A85D0
>	s_buffer_load_dword s22, s[4:7], 0xd1                 ; C20B05D1
>	s_buffer_load_dword s23, s[4:7], 0xd3                 ; C20B85D3
>	s_buffer_load_dword s24, s[4:7], 0xd4                 ; C20C05D4
>	s_buffer_load_dword s25, s[4:7], 0xd5                 ; C20C85D5
>	s_buffer_load_dword s26, s[4:7], 0xd7                 ; C20D05D7
>	s_buffer_load_dword s27, s[4:7], 0xd8                 ; C20D85D8
>	s_buffer_load_dword s4, s[4:7], 0xd9                  ; C20205D9
>	s_buffer_load_dword s5, s[0:3], 0x0                   ; C2028100
>	s_buffer_load_dword s6, s[0:3], 0x1                   ; C2030101
>	s_buffer_load_dword s7, s[0:3], 0x2                   ; C2038102
>	s_buffer_load_dword s28, s[0:3], 0x3                  ; C20E0103
>	s_buffer_load_dword s29, s[0:3], 0x4                  ; C20E8104
>	s_buffer_load_dword s30, s[0:3], 0x5                  ; C20F0105
>	s_buffer_load_dword s31, s[0:3], 0x6                  ; C20F8106
>	s_buffer_load_dword s32, s[0:3], 0x7                  ; C2100107
>	s_buffer_load_dword s33, s[0:3], 0x8                  ; C2108108
>	s_buffer_load_dword s34, s[0:3], 0x9                  ; C2110109
>	s_buffer_load_dword s35, s[0:3], 0xa                  ; C211810A
>	s_buffer_load_dword s36, s[0:3], 0xb                  ; C212010B
>	s_buffer_load_dword s37, s[0:3], 0xc                  ; C212810C
>	s_buffer_load_dword s38, s[0:3], 0xd                  ; C213010D
>	s_buffer_load_dword s39, s[0:3], 0xe                  ; C213810E
>	s_buffer_load_dword s40, s[0:3], 0xf                  ; C214010F
>	s_buffer_load_dword s41, s[0:3], 0x10                 ; C2148110
>	s_buffer_load_dword s42, s[0:3], 0x11                 ; C2150111
>	s_buffer_load_dword s43, s[0:3], 0x12                 ; C2158112
>	s_buffer_load_dword s44, s[0:3], 0x13                 ; C2160113
>	s_buffer_load_dword s45, s[0:3], 0x14                 ; C2168114
>	s_buffer_load_dword s46, s[0:3], 0x15                 ; C2170115
>	s_buffer_load_dword s47, s[0:3], 0x16                 ; C2178116
>	s_buffer_load_dword s48, s[0:3], 0x17                 ; C2180117
>	s_buffer_load_dword s49, s[0:3], 0x18                 ; C2188118
>	s_buffer_load_dword s50, s[0:3], 0x19                 ; C2190119
>	s_buffer_load_dword s51, s[0:3], 0x1a                 ; C219811A
>	s_buffer_load_dword s52, s[0:3], 0x1b                 ; C21A011B
>	s_buffer_load_dword s53, s[0:3], 0x1c                 ; C21A811C
>	s_buffer_load_dword s54, s[0:3], 0x1d                 ; C21B011D
>	s_buffer_load_dword s55, s[0:3], 0x1e                 ; C21B811E
>	s_buffer_load_dword s0, s[0:3], 0x1f                  ; C200011F
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mov_b32_e32 v0, s11                                 ; 7E00020B
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mov_b32_e32 v7, s14                                 ; 7E0E020E
>	v_mov_b32_e32 v8, s17                                 ; 7E100211
>	v_mov_b32_e32 v9, s20                                 ; 7E120214
>	v_mov_b32_e32 v10, s23                                ; 7E140217
>	v_mov_b32_e32 v11, s26                                ; 7E16021A
>	v_mac_f32_e32 v0, s9, v5                              ; 3E000A09
>	v_mac_f32_e32 v7, s12, v5                             ; 3E0E0A0C
>	v_mac_f32_e32 v8, s15, v5                             ; 3E100A0F
>	v_mac_f32_e32 v9, s18, v5                             ; 3E120A12
>	v_mac_f32_e32 v10, s21, v5                            ; 3E140A15
>	v_mac_f32_e32 v11, s24, v5                            ; 3E160A18
>	v_mac_f32_e32 v0, s10, v6                             ; 3E000C0A
>	v_mac_f32_e32 v7, s13, v6                             ; 3E0E0C0D
>	v_mac_f32_e32 v8, s16, v6                             ; 3E100C10
>	v_mac_f32_e32 v9, s19, v6                             ; 3E120C13
>	exp 15, 32, 0, 0, 0, v5, v6, v0, v7                   ; F800020F 07000605
>	v_mac_f32_e32 v10, s22, v6                            ; 3E140C16
>	v_mac_f32_e32 v11, s25, v6                            ; 3E160C19
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mad_f32 v0, v3, s8, -v4                             ; D2820000 84101103
>	v_mad_f32 v5, s27, v4, v1                             ; D2820005 0406081B
>	v_mad_f32 v6, s4, v4, -v2                             ; D2820006 840A0804
>	v_mul_f32_e32 v7, s6, v2                              ; 100E0406
>	exp 15, 33, 0, 0, 0, v8, v9, v10, v11                 ; F800021F 0B0A0908
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mul_f32_e32 v8, s30, v2                             ; 1010041E
>	v_mac_f32_e32 v7, s5, v1                              ; 3E0E0205
>	exp 15, 12, 0, 0, 0, v5, v6, v0, v4                   ; F80000CF 04000605
>	v_mac_f32_e32 v8, s29, v1                             ; 3E10021D
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mul_f32_e32 v0, s34, v2                             ; 10000422
>	v_mac_f32_e32 v0, s33, v1                             ; 3E000221
>	v_mul_f32_e32 v5, s38, v2                             ; 100A0426
>	v_mac_f32_e32 v5, s37, v1                             ; 3E0A0225
>	v_mul_f32_e32 v6, s42, v2                             ; 100C042A
>	v_mul_f32_e32 v9, s46, v2                             ; 1012042E
>	v_mul_f32_e32 v10, s50, v2                            ; 10140432
>	v_mul_f32_e32 v2, s54, v2                             ; 10040436
>	v_mac_f32_e32 v6, s41, v1                             ; 3E0C0229
>	v_mac_f32_e32 v9, s45, v1                             ; 3E12022D
>	v_mac_f32_e32 v10, s49, v1                            ; 3E140231
>	v_mac_f32_e32 v2, s53, v1                             ; 3E040235
>	v_mac_f32_e32 v7, s7, v3                              ; 3E0E0607
>	v_mac_f32_e32 v8, s31, v3                             ; 3E10061F
>	v_mac_f32_e32 v0, s35, v3                             ; 3E000623
>	v_mac_f32_e32 v5, s39, v3                             ; 3E0A0627
>	v_mac_f32_e32 v6, s43, v3                             ; 3E0C062B
>	v_mac_f32_e32 v9, s47, v3                             ; 3E12062F
>	v_mac_f32_e32 v10, s51, v3                            ; 3E140633
>	v_mac_f32_e32 v2, s55, v3                             ; 3E040637
>	v_mac_f32_e32 v7, s28, v4                             ; 3E0E081C
>	v_mac_f32_e32 v8, s32, v4                             ; 3E100820
>	v_mac_f32_e32 v0, s36, v4                             ; 3E000824
>	v_mac_f32_e32 v5, s40, v4                             ; 3E0A0828
>	v_mac_f32_e32 v6, s44, v4                             ; 3E0C082C
>	v_mac_f32_e32 v9, s48, v4                             ; 3E120830
>	v_mac_f32_e32 v10, s52, v4                            ; 3E140834
>	v_mac_f32_e32 v2, s0, v4                              ; 3E040800
>	exp 15, 13, 0, 0, 0, v7, v8, v0, v5                   ; F80000DF 05000807
>	exp 15, 14, 0, 1, 0, v6, v9, v10, v2                  ; F80008EF 020A0906
>	s_endpgm                                              ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 64
>VGPRS: 12
>Code Size: 544 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
>  clamp_color = 0
>FRAG
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL IN[1], GENERIC[1], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SAMP[1]
>DCL SAMP[2]
>DCL SAMP[3]
>DCL SVIEW[0], 2D, FLOAT
>DCL SVIEW[1], 2D, FLOAT
>DCL SVIEW[2], 2D, FLOAT
>DCL SVIEW[3], 2D, FLOAT
>DCL CONST[0..3]
>DCL TEMP[0..4], LOCAL
>IMM[0] FLT32 {   -0.5000,    -1.0000,    10.0000,   -10.0000}
>IMM[1] FLT32 {   -2.0000,     3.0000,     6.0000,     1.0000}
>IMM[2] FLT32 {   -0.5800,    -1.0000,    -0.0000,     1.0000}
>IMM[3] FLT32 {    1.0000,     0.0000,     0.0000,     0.0000}
>  0: MOV TEMP[0].xy, IN[0].xyyy
>  1: TEX TEMP[0].xw, TEMP[0], SAMP[0], 2D
>  2: MOV TEMP[1].xy, IN[1].xyyy
>  3: TEX TEMP[1].y, TEMP[1], SAMP[1], 2D
>  4: MAD TEMP[0].x, TEMP[1].yyyy, TEMP[0].xxxx, TEMP[0].wwww
>  5: MOV TEMP[1].xy, IN[0].zwww
>  6: TEX TEMP[1].w, TEMP[1], SAMP[3], 2D
>  7: ADD TEMP[2].xy, TEMP[1].wwww, IMM[0].xyyy
>  8: ADD TEMP[3].x, TEMP[1].wwww, TEMP[1].wwww
>  9: MOV_SAT TEMP[3].x, TEMP[3].xxxx
> 10: MUL TEMP[2].xy, TEMP[2].xyyy, IMM[0].zwww
> 11: MOV_SAT TEMP[2].xy, TEMP[2].xyyy
> 12: MAD TEMP[1].xy, TEMP[2].xyyy, IMM[1].xxxx, IMM[1].yyyy
> 13: MUL TEMP[2].xy, TEMP[2].xyyy, TEMP[2].xyyy
> 14: MUL TEMP[2].xy, TEMP[2].xyyy, TEMP[1].xyyy
> 15: MUL TEMP[2].x, TEMP[2].yyyy, TEMP[2].xxxx
> 16: MAD TEMP[4].x, CONST[3].wwww, IMM[1].zzzz, IMM[1].wwww
> 17: MAD TEMP[0].x, TEMP[0].xxxx, TEMP[4].xxxx, TEMP[2].xxxx
> 18: MAD TEMP[2].x, TEMP[3].xxxx, IMM[1].xxxx, IMM[1].yyyy
> 19: MUL TEMP[3].x, TEMP[3].xxxx, TEMP[3].xxxx
> 20: MUL TEMP[2].x, TEMP[3].xxxx, TEMP[2].xxxx
> 21: MAD TEMP[0].x, TEMP[0].xxxx, TEMP[2].xxxx, IMM[2].xxxx
> 22: MUL TEMP[0].x, TEMP[0].xxxx, IMM[0].zzzz
> 23: MOV_SAT TEMP[2].x, TEMP[0].xxxx
> 24: MAD TEMP[3].x, TEMP[2].xxxx, IMM[1].xxxx, IMM[1].yyyy
> 25: MUL TEMP[0].x, TEMP[2].xxxx, TEMP[2].xxxx
> 26: MUL TEMP[0].x, TEMP[0].xxxx, TEMP[3].xxxx
> 27: MOV TEMP[2].xy, IN[0].xyyy
> 28: TEX TEMP[2].xyz, TEMP[2], SAMP[2], 2D
> 29: MAD TEMP[2].xyz, CONST[3].yyyy, IMM[2].yzzz, TEMP[2].xyzz
> 30: MUL TEMP[1].xyz, IMM[3].xyyy, CONST[3].yyyy
> 31: MAD TEMP[0].xyz, TEMP[0].xxxx, TEMP[2].xyzz, TEMP[1].xyzz
> 32: MOV TEMP[0].w, IMM[1].wwww
> 33: MOV OUT[0], TEMP[0]
> 34: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %24 = load <16 x i8>, <16 x i8> addrspace(2)* %23, align 16, !tbaa !0
>  %25 = call float @llvm.SI.load.const(<16 x i8> %24, i32 52)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %24, i32 60)
>  %27 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %28 = load <8 x i32>, <8 x i32> addrspace(2)* %27, align 32, !tbaa !0
>  %29 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %30 = load <4 x i32>, <4 x i32> addrspace(2)* %29, align 16, !tbaa !0
>  %31 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 1
>  %32 = load <8 x i32>, <8 x i32> addrspace(2)* %31, align 32, !tbaa !0
>  %33 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 1
>  %34 = load <4 x i32>, <4 x i32> addrspace(2)* %33, align 16, !tbaa !0
>  %35 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 2
>  %36 = load <8 x i32>, <8 x i32> addrspace(2)* %35, align 32, !tbaa !0
>  %37 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 2
>  %38 = load <4 x i32>, <4 x i32> addrspace(2)* %37, align 16, !tbaa !0
>  %39 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 3
>  %40 = load <8 x i32>, <8 x i32> addrspace(2)* %39, align 32, !tbaa !0
>  %41 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 3
>  %42 = load <4 x i32>, <4 x i32> addrspace(2)* %41, align 16, !tbaa !0
>  %43 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %8)
>  %44 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %8)
>  %45 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %6, <2 x i32> %8)
>  %46 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %6, <2 x i32> %8)
>  %47 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %6, <2 x i32> %8)
>  %48 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %6, <2 x i32> %8)
>  %49 = bitcast float %43 to i32
>  %50 = bitcast float %44 to i32
>  %51 = insertelement <2 x i32> undef, i32 %49, i32 0
>  %52 = insertelement <2 x i32> %51, i32 %50, i32 1
>  %53 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %52, <8 x i32> %28, <4 x i32> %30, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %54 = extractelement <4 x float> %53, i32 0
>  %55 = extractelement <4 x float> %53, i32 3
>  %56 = bitcast float %47 to i32
>  %57 = bitcast float %48 to i32
>  %58 = insertelement <2 x i32> undef, i32 %56, i32 0
>  %59 = insertelement <2 x i32> %58, i32 %57, i32 1
>  %60 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %59, <8 x i32> %32, <4 x i32> %34, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %61 = extractelement <4 x float> %60, i32 1
>  %62 = fmul float %61, %54
>  %63 = fadd float %62, %55
>  %64 = bitcast float %45 to i32
>  %65 = bitcast float %46 to i32
>  %66 = insertelement <2 x i32> undef, i32 %64, i32 0
>  %67 = insertelement <2 x i32> %66, i32 %65, i32 1
>  %68 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %67, <8 x i32> %40, <4 x i32> %42, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %69 = extractelement <4 x float> %68, i32 3
>  %70 = fadd float %69, -5.000000e-01
>  %71 = fadd float %69, -1.000000e+00
>  %72 = fadd float %69, %69
>  %73 = call float @llvm.AMDIL.clamp.(float %72, float 0.000000e+00, float 1.000000e+00)
>  %74 = fmul float %70, 1.000000e+01
>  %75 = fmul float %71, -1.000000e+01
>  %76 = call float @llvm.AMDIL.clamp.(float %74, float 0.000000e+00, float 1.000000e+00)
>  %77 = call float @llvm.AMDIL.clamp.(float %75, float 0.000000e+00, float 1.000000e+00)
>  %78 = fmul float %76, -2.000000e+00
>  %79 = fadd float %78, 3.000000e+00
>  %80 = fmul float %77, -2.000000e+00
>  %81 = fadd float %80, 3.000000e+00
>  %82 = fmul float %76, %76
>  %83 = fmul float %77, %77
>  %84 = fmul float %82, %79
>  %85 = fmul float %83, %81
>  %86 = fmul float %85, %84
>  %87 = fmul float %26, 6.000000e+00
>  %88 = fadd float %87, 1.000000e+00
>  %89 = fmul float %63, %88
>  %90 = fadd float %89, %86
>  %91 = fmul float %73, -2.000000e+00
>  %92 = fadd float %91, 3.000000e+00
>  %93 = fmul float %73, %73
>  %94 = fmul float %93, %92
>  %95 = fmul float %90, %94
>  %96 = fadd float %95, 0xBFE28F5C20000000
>  %97 = fmul float %96, 1.000000e+01
>  %98 = call float @llvm.AMDIL.clamp.(float %97, float 0.000000e+00, float 1.000000e+00)
>  %99 = fmul float %98, -2.000000e+00
>  %100 = fadd float %99, 3.000000e+00
>  %101 = fmul float %98, %98
>  %102 = fmul float %101, %100
>  %103 = bitcast float %43 to i32
>  %104 = bitcast float %44 to i32
>  %105 = insertelement <2 x i32> undef, i32 %103, i32 0
>  %106 = insertelement <2 x i32> %105, i32 %104, i32 1
>  %107 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %106, <8 x i32> %36, <4 x i32> %38, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %108 = extractelement <4 x float> %107, i32 0
>  %109 = extractelement <4 x float> %107, i32 1
>  %110 = extractelement <4 x float> %107, i32 2
>  %111 = fsub float %108, %25
>  %112 = fmul float %25, -0.000000e+00
>  %113 = fadd float %112, %109
>  %114 = fmul float %25, -0.000000e+00
>  %115 = fadd float %114, %110
>  %116 = fmul float %25, 0.000000e+00
>  %117 = fmul float %25, 0.000000e+00
>  %118 = fmul float %102, %111
>  %119 = fadd float %118, %25
>  %120 = fmul float %102, %113
>  %121 = fadd float %120, %116
>  %122 = fmul float %102, %115
>  %123 = fadd float %122, %117
>  %124 = call i32 @llvm.SI.packf16(float %119, float %121)
>  %125 = bitcast i32 %124 to float
>  %126 = call i32 @llvm.SI.packf16(float %123, float 1.000000e+00)
>  %127 = bitcast i32 %126 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %125, float %127, float %125, float %127)
>  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.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #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, s10                                                       ; BEFC030A
>	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_interp_p1_f32 v4, v0, 2, 0, [m0]                                      ; C8100200
>	v_interp_p2_f32 v4, [v4], v1, 2, 0, [m0]                                ; C8110201
>	v_interp_p1_f32 v5, v0, 3, 0, [m0]                                      ; C8140300
>	v_interp_p2_f32 v5, [v5], v1, 3, 0, [m0]                                ; C8150301
>	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 v7, v0, 1, 1, [m0]                                      ; C81C0500
>	v_interp_p2_f32 v7, [v7], v1, 1, 1, [m0]                                ; C81D0501
>	s_load_dwordx4 s[16:19], s[2:3], 0x0                                    ; C0880300
>	s_load_dwordx4 s[20:23], s[4:5], 0x0                                    ; C08A0500
>	s_load_dwordx4 s[24:27], s[4:5], 0x4                                    ; C08C0504
>	s_load_dwordx4 s[0:3], s[4:5], 0x8                                      ; C0800508
>	s_load_dwordx4 s[28:31], s[4:5], 0xc                                    ; C08E050C
>	s_load_dwordx8 s[32:39], s[6:7], 0x0                                    ; C0D00700
>	s_load_dwordx8 s[40:47], s[6:7], 0x8                                    ; C0D40708
>	s_load_dwordx8 s[8:15], s[6:7], 0x10                                    ; C0C40710
>	s_load_dwordx8 s[48:55], s[6:7], 0x18                                   ; C0D80718
>	s_waitcnt lgkmcnt(0)                                                    ; BF8C007F
>	s_buffer_load_dword s4, s[16:19], 0xd                                   ; C202110D
>	s_buffer_load_dword s5, s[16:19], 0xf                                   ; C202910F
>	image_sample v[0:1], 9, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[32:39], s[20:23] ; F0800900 00A80002
>	image_sample v6, 2, 0, 0, 0, 0, 0, 0, 0, v[6:7], s[40:47], s[24:27]     ; F0800200 00CA0606
>	s_waitcnt vmcnt(0)                                                      ; BF8C0770
>	v_mac_f32_e32 v1, v0, v6                                                ; 3E020D00
>	image_sample v0, 8, 0, 0, 0, 0, 0, 0, 0, v[4:5], s[48:55], s[28:31]     ; F0800800 00EC0004
>	s_waitcnt vmcnt(0)                                                      ; BF8C0770
>	v_add_f32_e32 v4, -0.5, v0                                              ; 060800F1
>	v_add_f32_e32 v5, v0, v0                                                ; 060A0100
>	v_add_f32_e64 v5, 0, v5 clamp                                           ; D2060805 00020A80
>	image_sample v[6:8], 7, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[8:15], s[0:3]    ; F0800700 00020602
>	v_mov_b32_e32 v2, 0x41200000                                            ; 7E0402FF 41200000
>	v_mul_f32_e32 v3, v2, v4                                                ; 10060902
>	v_madmk_f32_e32 v0, v0, v2, 0xc1200000                                  ; 40000500 C1200000
>	v_add_f32_e64 v3, 0, v3 clamp                                           ; D2060803 00020680
>	v_add_f32_e64 v0, 0, v0 clamp                                           ; D2060800 00020080
>	v_mov_b32_e32 v4, 0x40400000                                            ; 7E0802FF 40400000
>	v_mad_f32 v9, -2.0, v3, v4                                              ; D2820009 041206F5
>	v_mul_f32_e32 v3, v3, v3                                                ; 10060703
>	v_mul_f32_e32 v3, v9, v3                                                ; 10060709
>	v_mad_f32 v9, -2.0, v0, v4                                              ; D2820009 041200F5
>	v_mul_f32_e32 v0, v0, v0                                                ; 10000100
>	v_mul_f32_e32 v0, v9, v0                                                ; 10000109
>	v_mul_f32_e32 v0, v3, v0                                                ; 10000103
>	v_mov_b32_e32 v3, 0x40c00000                                            ; 7E0602FF 40C00000
>	s_waitcnt lgkmcnt(0)                                                    ; BF8C007F
>	v_mad_f32 v3, v3, s5, 1.0                                               ; D2820003 03C80B03
>	v_mac_f32_e32 v0, v3, v1                                                ; 3E000303
>	v_mad_f32 v1, -2.0, v5, v4                                              ; D2820001 04120AF5
>	v_mul_f32_e32 v3, v5, v5                                                ; 10060B05
>	v_mul_f32_e32 v1, v1, v3                                                ; 10020701
>	v_madak_f32_e32 v0, v0, v1, 0xbf147ae1                                  ; 42000300 BF147AE1
>	v_mul_f32_e32 v0, v2, v0                                                ; 10000102
>	v_add_f32_e64 v0, 0, v0 clamp                                           ; D2060800 00020080
>	v_mac_f32_e32 v4, -2.0, v0                                              ; 3E0800F5
>	v_mul_f32_e32 v0, v0, v0                                                ; 10000100
>	v_mul_f32_e32 v0, v4, v0                                                ; 10000104
>	s_waitcnt vmcnt(0)                                                      ; BF8C0770
>	v_subrev_f32_e32 v1, s4, v6                                             ; 0A020C04
>	v_mov_b32_e32 v2, 0x80000000                                            ; 7E0402FF 80000000
>	v_mad_f32 v3, s4, v2, v7                                                ; D2820003 041E0404
>	v_mac_f32_e32 v8, s4, v2                                                ; 3E100404
>	v_mul_f32_e64 v2, 0, s4                                                 ; D2100002 00000880
>	v_mad_f32 v1, v0, v1, s4                                                ; D2820001 00120300
>	v_mad_f32 v3, v3, v0, v2                                                ; D2820003 040A0103
>	v_mac_f32_e32 v2, v8, v0                                                ; 3E040108
>	v_cvt_pkrtz_f16_f32_e32 v0, v1, v3                                      ; 5E000701
>	v_cvt_pkrtz_f16_f32_e64 v1, v2, 1.0                                     ; D25E0001 0001E502
>	exp 15, 0, 1, 1, 1, v0, v1, v0, v1                                      ; F8001C0F 01000100
>	s_endpgm                                                                ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 56
>VGPRS: 12
>Code Size: 396 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
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL OUT[1], CLIPVERTEX
>DCL OUT[2], GENERIC[0]
>DCL OUT[3], GENERIC[1]
>DCL CONST[0..54]
>DCL TEMP[0..3], LOCAL
>  0: MUL TEMP[0].xy, CONST[48].xyyy, IN[1].xyyy
>  1: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
>  2: ADD TEMP[1].x, TEMP[0].xxxx, CONST[48].wwww
>  3: MOV TEMP[1].z, TEMP[1].xxxx
>  4: MUL TEMP[0].xy, CONST[49].xyyy, IN[1].xyyy
>  5: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
>  6: ADD TEMP[2].x, TEMP[0].xxxx, CONST[49].wwww
>  7: MOV TEMP[1].w, TEMP[2].xxxx
>  8: MUL TEMP[0].xy, CONST[50].xyyy, IN[1].xyyy
>  9: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 10: ADD TEMP[2].x, TEMP[0].xxxx, CONST[50].wwww
> 11: MUL TEMP[0].xy, CONST[51].xyyy, IN[1].xyyy
> 12: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 13: ADD TEMP[3].x, TEMP[0].xxxx, CONST[51].wwww
> 14: MOV TEMP[2].y, TEMP[3].xxxx
> 15: MUL TEMP[0].xy, CONST[52].xyyy, IN[1].xyyy
> 16: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 17: ADD TEMP[3].x, TEMP[0].xxxx, CONST[52].wwww
> 18: MOV TEMP[2].z, TEMP[3].xxxx
> 19: MUL TEMP[0].xy, CONST[53].xyyy, IN[1].xyyy
> 20: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 21: ADD TEMP[0].x, TEMP[0].xxxx, CONST[53].wwww
> 22: MOV TEMP[2].w, TEMP[0].xxxx
> 23: MOV TEMP[0].xw, IN[0].xxxw
> 24: MOV TEMP[1].xy, IN[1].xyxx
> 25: MAD TEMP[3].x, IN[0].zzzz, CONST[0].zzzz, -IN[0].wwww
> 26: MOV TEMP[0].z, TEMP[3].xxxx
> 27: MOV TEMP[0].y, -IN[0].yyyy
> 28: MAD TEMP[0].xy, CONST[54].xyyy, IN[0].wwww, TEMP[0].xyyy
> 29: MOV OUT[2], TEMP[1]
> 30: MOV OUT[3], TEMP[2]
> 31: MOV OUT[0], TEMP[0]
> 32: MOV OUT[1], IN[0]
> 33: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = call float @llvm.SI.load.const(<16 x i8> %13, i32 8)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %13, i32 768)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %13, i32 772)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %13, i32 780)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %13, i32 784)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %13, i32 788)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %13, i32 796)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %13, i32 800)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %13, i32 804)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %13, i32 812)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %13, i32 816)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %13, i32 820)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %13, i32 828)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %13, i32 832)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %13, i32 836)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %13, i32 844)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %13, i32 848)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %13, i32 852)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %13, i32 860)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %13, i32 864)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %13, i32 868)
>  %35 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %36 = load <16 x i8>, <16 x i8> addrspace(2)* %35, align 16, !tbaa !0
>  %37 = add i32 %5, %8
>  %38 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %36, i32 0, i32 %37)
>  %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 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 1
>  %44 = load <16 x i8>, <16 x i8> addrspace(2)* %43, align 16, !tbaa !0
>  %45 = add i32 %5, %8
>  %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 = fmul float %15, %47
>  %50 = fmul float %16, %48
>  %51 = fadd float %50, %49
>  %52 = fadd float %51, %17
>  %53 = fmul float %18, %47
>  %54 = fmul float %19, %48
>  %55 = fadd float %54, %53
>  %56 = fadd float %55, %20
>  %57 = fmul float %21, %47
>  %58 = fmul float %22, %48
>  %59 = fadd float %58, %57
>  %60 = fadd float %59, %23
>  %61 = fmul float %24, %47
>  %62 = fmul float %25, %48
>  %63 = fadd float %62, %61
>  %64 = fadd float %63, %26
>  %65 = fmul float %27, %47
>  %66 = fmul float %28, %48
>  %67 = fadd float %66, %65
>  %68 = fadd float %67, %29
>  %69 = fmul float %30, %47
>  %70 = fmul float %31, %48
>  %71 = fadd float %70, %69
>  %72 = fadd float %71, %32
>  %73 = fmul float %41, %14
>  %74 = fsub float %73, %42
>  %75 = fmul float %33, %42
>  %76 = fadd float %75, %39
>  %77 = fmul float %34, %42
>  %78 = fsub float %77, %40
>  %79 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 16
>  %80 = load <16 x i8>, <16 x i8> addrspace(2)* %79, align 16, !tbaa !0
>  %81 = call float @llvm.SI.load.const(<16 x i8> %80, i32 0)
>  %82 = fmul float %81, %39
>  %83 = call float @llvm.SI.load.const(<16 x i8> %80, i32 4)
>  %84 = fmul float %83, %40
>  %85 = fadd float %82, %84
>  %86 = call float @llvm.SI.load.const(<16 x i8> %80, i32 8)
>  %87 = fmul float %86, %41
>  %88 = fadd float %85, %87
>  %89 = call float @llvm.SI.load.const(<16 x i8> %80, i32 12)
>  %90 = fmul float %89, %42
>  %91 = fadd float %88, %90
>  %92 = call float @llvm.SI.load.const(<16 x i8> %80, i32 16)
>  %93 = fmul float %92, %39
>  %94 = call float @llvm.SI.load.const(<16 x i8> %80, i32 20)
>  %95 = fmul float %94, %40
>  %96 = fadd float %93, %95
>  %97 = call float @llvm.SI.load.const(<16 x i8> %80, i32 24)
>  %98 = fmul float %97, %41
>  %99 = fadd float %96, %98
>  %100 = call float @llvm.SI.load.const(<16 x i8> %80, i32 28)
>  %101 = fmul float %100, %42
>  %102 = fadd float %99, %101
>  %103 = call float @llvm.SI.load.const(<16 x i8> %80, i32 32)
>  %104 = fmul float %103, %39
>  %105 = call float @llvm.SI.load.const(<16 x i8> %80, i32 36)
>  %106 = fmul float %105, %40
>  %107 = fadd float %104, %106
>  %108 = call float @llvm.SI.load.const(<16 x i8> %80, i32 40)
>  %109 = fmul float %108, %41
>  %110 = fadd float %107, %109
>  %111 = call float @llvm.SI.load.const(<16 x i8> %80, i32 44)
>  %112 = fmul float %111, %42
>  %113 = fadd float %110, %112
>  %114 = call float @llvm.SI.load.const(<16 x i8> %80, i32 48)
>  %115 = fmul float %114, %39
>  %116 = call float @llvm.SI.load.const(<16 x i8> %80, i32 52)
>  %117 = fmul float %116, %40
>  %118 = fadd float %115, %117
>  %119 = call float @llvm.SI.load.const(<16 x i8> %80, i32 56)
>  %120 = fmul float %119, %41
>  %121 = fadd float %118, %120
>  %122 = call float @llvm.SI.load.const(<16 x i8> %80, i32 60)
>  %123 = fmul float %122, %42
>  %124 = fadd float %121, %123
>  %125 = call float @llvm.SI.load.const(<16 x i8> %80, i32 64)
>  %126 = fmul float %125, %39
>  %127 = call float @llvm.SI.load.const(<16 x i8> %80, i32 68)
>  %128 = fmul float %127, %40
>  %129 = fadd float %126, %128
>  %130 = call float @llvm.SI.load.const(<16 x i8> %80, i32 72)
>  %131 = fmul float %130, %41
>  %132 = fadd float %129, %131
>  %133 = call float @llvm.SI.load.const(<16 x i8> %80, i32 76)
>  %134 = fmul float %133, %42
>  %135 = fadd float %132, %134
>  %136 = call float @llvm.SI.load.const(<16 x i8> %80, i32 80)
>  %137 = fmul float %136, %39
>  %138 = call float @llvm.SI.load.const(<16 x i8> %80, i32 84)
>  %139 = fmul float %138, %40
>  %140 = fadd float %137, %139
>  %141 = call float @llvm.SI.load.const(<16 x i8> %80, i32 88)
>  %142 = fmul float %141, %41
>  %143 = fadd float %140, %142
>  %144 = call float @llvm.SI.load.const(<16 x i8> %80, i32 92)
>  %145 = fmul float %144, %42
>  %146 = fadd float %143, %145
>  %147 = call float @llvm.SI.load.const(<16 x i8> %80, i32 96)
>  %148 = fmul float %147, %39
>  %149 = call float @llvm.SI.load.const(<16 x i8> %80, i32 100)
>  %150 = fmul float %149, %40
>  %151 = fadd float %148, %150
>  %152 = call float @llvm.SI.load.const(<16 x i8> %80, i32 104)
>  %153 = fmul float %152, %41
>  %154 = fadd float %151, %153
>  %155 = call float @llvm.SI.load.const(<16 x i8> %80, i32 108)
>  %156 = fmul float %155, %42
>  %157 = fadd float %154, %156
>  %158 = call float @llvm.SI.load.const(<16 x i8> %80, i32 112)
>  %159 = fmul float %158, %39
>  %160 = call float @llvm.SI.load.const(<16 x i8> %80, i32 116)
>  %161 = fmul float %160, %40
>  %162 = fadd float %159, %161
>  %163 = call float @llvm.SI.load.const(<16 x i8> %80, i32 120)
>  %164 = fmul float %163, %41
>  %165 = fadd float %162, %164
>  %166 = call float @llvm.SI.load.const(<16 x i8> %80, i32 124)
>  %167 = fmul float %166, %42
>  %168 = fadd float %165, %167
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %47, float %48, float %52, float %56)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %60, float %64, float %68, float %72)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 12, i32 0, float %76, float %78, float %74, float %42)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 13, i32 0, float %91, float %102, float %113, float %124)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 14, i32 0, float %135, float %146, float %157, 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
>
>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[4:7], s[8:9], 0x0                    ; C0820900
>	s_load_dwordx4 s[12:15], s[8:9], 0x4                  ; C0860904
>	v_add_i32_e32 v0, vcc, s10, v0                        ; 4A00000A
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[4:7], 0 idxen   ; E00C2000 80010100
>	buffer_load_format_xyzw v[5:8], v0, s[12:15], 0 idxen ; E00C2000 80030500
>	s_load_dwordx4 s[4:7], s[2:3], 0x0                    ; C0820300
>	s_load_dwordx4 s[0:3], s[2:3], 0x40                   ; C0800340
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	s_buffer_load_dword s8, s[4:7], 0x2                   ; C2040502
>	s_buffer_load_dword s9, s[4:7], 0xc0                  ; C20485C0
>	s_buffer_load_dword s10, s[4:7], 0xc1                 ; C20505C1
>	s_buffer_load_dword s11, s[4:7], 0xc3                 ; C20585C3
>	s_buffer_load_dword s12, s[4:7], 0xc4                 ; C20605C4
>	s_buffer_load_dword s13, s[4:7], 0xc5                 ; C20685C5
>	s_buffer_load_dword s14, s[4:7], 0xc7                 ; C20705C7
>	s_buffer_load_dword s15, s[4:7], 0xc8                 ; C20785C8
>	s_buffer_load_dword s16, s[4:7], 0xc9                 ; C20805C9
>	s_buffer_load_dword s17, s[4:7], 0xcb                 ; C20885CB
>	s_buffer_load_dword s18, s[4:7], 0xcc                 ; C20905CC
>	s_buffer_load_dword s19, s[4:7], 0xcd                 ; C20985CD
>	s_buffer_load_dword s20, s[4:7], 0xcf                 ; C20A05CF
>	s_buffer_load_dword s21, s[4:7], 0xd0                 ; C20A85D0
>	s_buffer_load_dword s22, s[4:7], 0xd1                 ; C20B05D1
>	s_buffer_load_dword s23, s[4:7], 0xd3                 ; C20B85D3
>	s_buffer_load_dword s24, s[4:7], 0xd4                 ; C20C05D4
>	s_buffer_load_dword s25, s[4:7], 0xd5                 ; C20C85D5
>	s_buffer_load_dword s26, s[4:7], 0xd7                 ; C20D05D7
>	s_buffer_load_dword s27, s[4:7], 0xd8                 ; C20D85D8
>	s_buffer_load_dword s4, s[4:7], 0xd9                  ; C20205D9
>	s_buffer_load_dword s5, s[0:3], 0x0                   ; C2028100
>	s_buffer_load_dword s6, s[0:3], 0x1                   ; C2030101
>	s_buffer_load_dword s7, s[0:3], 0x2                   ; C2038102
>	s_buffer_load_dword s28, s[0:3], 0x3                  ; C20E0103
>	s_buffer_load_dword s29, s[0:3], 0x4                  ; C20E8104
>	s_buffer_load_dword s30, s[0:3], 0x5                  ; C20F0105
>	s_buffer_load_dword s31, s[0:3], 0x6                  ; C20F8106
>	s_buffer_load_dword s32, s[0:3], 0x7                  ; C2100107
>	s_buffer_load_dword s33, s[0:3], 0x8                  ; C2108108
>	s_buffer_load_dword s34, s[0:3], 0x9                  ; C2110109
>	s_buffer_load_dword s35, s[0:3], 0xa                  ; C211810A
>	s_buffer_load_dword s36, s[0:3], 0xb                  ; C212010B
>	s_buffer_load_dword s37, s[0:3], 0xc                  ; C212810C
>	s_buffer_load_dword s38, s[0:3], 0xd                  ; C213010D
>	s_buffer_load_dword s39, s[0:3], 0xe                  ; C213810E
>	s_buffer_load_dword s40, s[0:3], 0xf                  ; C214010F
>	s_buffer_load_dword s41, s[0:3], 0x10                 ; C2148110
>	s_buffer_load_dword s42, s[0:3], 0x11                 ; C2150111
>	s_buffer_load_dword s43, s[0:3], 0x12                 ; C2158112
>	s_buffer_load_dword s44, s[0:3], 0x13                 ; C2160113
>	s_buffer_load_dword s45, s[0:3], 0x14                 ; C2168114
>	s_buffer_load_dword s46, s[0:3], 0x15                 ; C2170115
>	s_buffer_load_dword s47, s[0:3], 0x16                 ; C2178116
>	s_buffer_load_dword s48, s[0:3], 0x17                 ; C2180117
>	s_buffer_load_dword s49, s[0:3], 0x18                 ; C2188118
>	s_buffer_load_dword s50, s[0:3], 0x19                 ; C2190119
>	s_buffer_load_dword s51, s[0:3], 0x1a                 ; C219811A
>	s_buffer_load_dword s52, s[0:3], 0x1b                 ; C21A011B
>	s_buffer_load_dword s53, s[0:3], 0x1c                 ; C21A811C
>	s_buffer_load_dword s54, s[0:3], 0x1d                 ; C21B011D
>	s_buffer_load_dword s55, s[0:3], 0x1e                 ; C21B811E
>	s_buffer_load_dword s0, s[0:3], 0x1f                  ; C200011F
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mov_b32_e32 v0, s11                                 ; 7E00020B
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mov_b32_e32 v7, s14                                 ; 7E0E020E
>	v_mov_b32_e32 v8, s17                                 ; 7E100211
>	v_mov_b32_e32 v9, s20                                 ; 7E120214
>	v_mov_b32_e32 v10, s23                                ; 7E140217
>	v_mov_b32_e32 v11, s26                                ; 7E16021A
>	v_mac_f32_e32 v0, s9, v5                              ; 3E000A09
>	v_mac_f32_e32 v7, s12, v5                             ; 3E0E0A0C
>	v_mac_f32_e32 v8, s15, v5                             ; 3E100A0F
>	v_mac_f32_e32 v9, s18, v5                             ; 3E120A12
>	v_mac_f32_e32 v10, s21, v5                            ; 3E140A15
>	v_mac_f32_e32 v11, s24, v5                            ; 3E160A18
>	v_mac_f32_e32 v0, s10, v6                             ; 3E000C0A
>	v_mac_f32_e32 v7, s13, v6                             ; 3E0E0C0D
>	v_mac_f32_e32 v8, s16, v6                             ; 3E100C10
>	v_mac_f32_e32 v9, s19, v6                             ; 3E120C13
>	exp 15, 32, 0, 0, 0, v5, v6, v0, v7                   ; F800020F 07000605
>	v_mac_f32_e32 v10, s22, v6                            ; 3E140C16
>	v_mac_f32_e32 v11, s25, v6                            ; 3E160C19
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mad_f32 v0, v3, s8, -v4                             ; D2820000 84101103
>	v_mad_f32 v5, s27, v4, v1                             ; D2820005 0406081B
>	v_mad_f32 v6, s4, v4, -v2                             ; D2820006 840A0804
>	v_mul_f32_e32 v7, s6, v2                              ; 100E0406
>	exp 15, 33, 0, 0, 0, v8, v9, v10, v11                 ; F800021F 0B0A0908
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mul_f32_e32 v8, s30, v2                             ; 1010041E
>	v_mac_f32_e32 v7, s5, v1                              ; 3E0E0205
>	exp 15, 12, 0, 0, 0, v5, v6, v0, v4                   ; F80000CF 04000605
>	v_mac_f32_e32 v8, s29, v1                             ; 3E10021D
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mul_f32_e32 v0, s34, v2                             ; 10000422
>	v_mac_f32_e32 v0, s33, v1                             ; 3E000221
>	v_mul_f32_e32 v5, s38, v2                             ; 100A0426
>	v_mac_f32_e32 v5, s37, v1                             ; 3E0A0225
>	v_mul_f32_e32 v6, s42, v2                             ; 100C042A
>	v_mul_f32_e32 v9, s46, v2                             ; 1012042E
>	v_mul_f32_e32 v10, s50, v2                            ; 10140432
>	v_mul_f32_e32 v2, s54, v2                             ; 10040436
>	v_mac_f32_e32 v6, s41, v1                             ; 3E0C0229
>	v_mac_f32_e32 v9, s45, v1                             ; 3E12022D
>	v_mac_f32_e32 v10, s49, v1                            ; 3E140231
>	v_mac_f32_e32 v2, s53, v1                             ; 3E040235
>	v_mac_f32_e32 v7, s7, v3                              ; 3E0E0607
>	v_mac_f32_e32 v8, s31, v3                             ; 3E10061F
>	v_mac_f32_e32 v0, s35, v3                             ; 3E000623
>	v_mac_f32_e32 v5, s39, v3                             ; 3E0A0627
>	v_mac_f32_e32 v6, s43, v3                             ; 3E0C062B
>	v_mac_f32_e32 v9, s47, v3                             ; 3E12062F
>	v_mac_f32_e32 v10, s51, v3                            ; 3E140633
>	v_mac_f32_e32 v2, s55, v3                             ; 3E040637
>	v_mac_f32_e32 v7, s28, v4                             ; 3E0E081C
>	v_mac_f32_e32 v8, s32, v4                             ; 3E100820
>	v_mac_f32_e32 v0, s36, v4                             ; 3E000824
>	v_mac_f32_e32 v5, s40, v4                             ; 3E0A0828
>	v_mac_f32_e32 v6, s44, v4                             ; 3E0C082C
>	v_mac_f32_e32 v9, s48, v4                             ; 3E120830
>	v_mac_f32_e32 v10, s52, v4                            ; 3E140834
>	v_mac_f32_e32 v2, s0, v4                              ; 3E040800
>	exp 15, 13, 0, 0, 0, v7, v8, v0, v5                   ; F80000DF 05000807
>	exp 15, 14, 0, 1, 0, v6, v9, v10, v2                  ; F80008EF 020A0906
>	s_endpgm                                              ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 64
>VGPRS: 12
>Code Size: 544 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
>  clamp_color = 0
>FRAG
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL IN[1], GENERIC[1], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SAMP[1]
>DCL SAMP[2]
>DCL SAMP[3]
>DCL SAMP[4]
>DCL SVIEW[0], 2D, FLOAT
>DCL SVIEW[1], 2D, FLOAT
>DCL SVIEW[2], 2D, FLOAT
>DCL SVIEW[3], 2D, FLOAT
>DCL SVIEW[4], 2D, FLOAT
>DCL CONST[0..3]
>DCL TEMP[0..7], LOCAL
>IMM[0] FLT32 {    6.0000,     1.0000,     0.7500,     0.2500}
>IMM[1] FLT32 {   -1.0000,     0.3000,     0.5900,     0.1100}
>IMM[2] FLT32 {   -0.0800,    14.2857,    -2.0000,     3.0000}
>IMM[3] FLT32 {    0.0300,     0.5000,    -0.5000,    -1.0000}
>IMM[4] FLT32 {   10.0000,   -10.0000,    -0.5800,  -100.0000}
>IMM[5] FLT32 {   -0.9000,    -0.0100,     0.0000,     0.0000}
>  0: MUL TEMP[0].x, IMM[0].zzzz, CONST[3].wwww
>  1: MOV TEMP[1].xy, IN[0].xyyy
>  2: TEX TEMP[1], TEMP[1], SAMP[0], 2D
>  3: ADD TEMP[2].x, -TEMP[1].xxxx, IMM[0].yyyy
>  4: MUL TEMP[2].x, TEMP[2].xxxx, TEMP[2].xxxx
>  5: MUL TEMP[2].x, TEMP[2].xxxx, TEMP[2].xxxx
>  6: MAD TEMP[0].x, TEMP[2].xxxx, IMM[0].wwww, TEMP[0].xxxx
>  7: MOV TEMP[2].xy, IN[1].zwww
>  8: TEX TEMP[2], TEMP[2], SAMP[3], 2D
>  9: ADD TEMP[2], TEMP[2], IMM[1].xxxx
> 10: MAD TEMP[2], TEMP[0].xxxx, TEMP[2], IMM[0].yyyy
> 11: MOV TEMP[3].xy, IN[0].zwww
> 12: TEX TEMP[3], TEMP[3], SAMP[4], 2D
> 13: MUL TEMP[4].xyz, TEMP[2].xyzz, TEMP[3].xyzz
> 14: DP3 TEMP[5].x, TEMP[4].xyzz, IMM[1].yzww
> 15: ADD TEMP[0].x, TEMP[5].xxxx, IMM[2].xxxx
> 16: MUL TEMP[0].x, TEMP[0].xxxx, IMM[2].yyyy
> 17: MOV_SAT TEMP[5].x, TEMP[0].xxxx
> 18: MAD TEMP[6].x, TEMP[5].xxxx, IMM[2].zzzz, IMM[2].wwww
> 19: MUL TEMP[0].x, TEMP[5].xxxx, TEMP[5].xxxx
> 20: MAD TEMP[0].x, TEMP[6].xxxx, -TEMP[0].xxxx, IMM[0].yyyy
> 21: MAD TEMP[2].xyz, TEMP[0].xxxx, IMM[3].xxxx, TEMP[4].xyzz
> 22: MUL TEMP[2].xyz, TEMP[1].zzzz, TEMP[2].xyzz
> 23: MAD TEMP[4].xyz, TEMP[2].xyzz, IMM[3].yyyy, TEMP[4].xyzz
> 24: MOV_SAT TEMP[5].xyz, TEMP[4].xyzz
> 25: MUL TEMP[2].xyz, TEMP[1].yyyy, TEMP[5].xyzz
> 26: MOV TEMP[6].xy, IN[0].xyyy
> 27: TEX TEMP[6], TEMP[6], SAMP[2], 2D
> 28: MAD TEMP[4].xyz, TEMP[5].xyzz, -TEMP[1].yyyy, TEMP[6].xyzz
> 29: ADD TEMP[5].xy, TEMP[3].wwww, IMM[3].zwww
> 30: ADD TEMP[7].x, TEMP[3].wwww, TEMP[3].wwww
> 31: MOV_SAT TEMP[7].x, TEMP[7].xxxx
> 32: MUL TEMP[5].xy, TEMP[5].xyyy, IMM[4].xyyy
> 33: MOV_SAT TEMP[5].xy, TEMP[5].xyyy
> 34: MAD TEMP[3].xy, TEMP[5].xyyy, IMM[2].zzzz, IMM[2].wwww
> 35: MUL TEMP[5].xy, TEMP[5].xyyy, TEMP[5].xyyy
> 36: MUL TEMP[3].xy, TEMP[5].xyyy, TEMP[3].xyyy
> 37: MUL TEMP[0].x, TEMP[3].yyyy, TEMP[3].xxxx
> 38: MOV TEMP[3].xy, IN[1].xyyy
> 39: TEX TEMP[3].y, TEMP[3], SAMP[1], 2D
> 40: MAD TEMP[3].x, TEMP[3].yyyy, TEMP[1].xxxx, TEMP[1].wwww
> 41: MUL TEMP[1].x, TEMP[1].yyyy, CONST[3].zzzz
> 42: MUL TEMP[1].x, TEMP[2].wwww, TEMP[1].xxxx
> 43: MAD TEMP[5].x, CONST[3].wwww, IMM[0].xxxx, IMM[0].yyyy
> 44: MAD TEMP[0].x, TEMP[3].xxxx, TEMP[5].xxxx, TEMP[0].xxxx
> 45: MAD TEMP[3].x, TEMP[7].xxxx, IMM[2].zzzz, IMM[2].wwww
> 46: MUL TEMP[5].x, TEMP[7].xxxx, TEMP[7].xxxx
> 47: MUL TEMP[3].x, TEMP[5].xxxx, TEMP[3].xxxx
> 48: MAD TEMP[3].x, TEMP[0].xxxx, TEMP[3].xxxx, IMM[4].zzzz
> 49: MUL TEMP[3].x, TEMP[3].xxxx, IMM[4].xxxx
> 50: MOV_SAT TEMP[3].x, TEMP[3].xxxx
> 51: MAD TEMP[0].x, TEMP[3].xxxx, IMM[2].zzzz, IMM[2].wwww
> 52: MUL TEMP[3].x, TEMP[3].xxxx, TEMP[3].xxxx
> 53: MUL TEMP[5].x, TEMP[3].xxxx, TEMP[0].xxxx
> 54: MAD TEMP[0].xy, TEMP[0].xxxx, TEMP[3].xxxx, IMM[5].xyyy
> 55: MUL TEMP[0].xy, TEMP[0].xyyy, IMM[4].xwww
> 56: MAD TEMP[2].xyz, TEMP[5].xxxx, TEMP[4].xyzz, TEMP[2].xyzz
> 57: MAX TEMP[3].x, TEMP[0].yyyy, IMM[5].zzzz
> 58: MOV_SAT TEMP[0].x, TEMP[0].xxxx
> 59: MAD TEMP[5].x, TEMP[3].xxxx, IMM[2].zzzz, IMM[2].wwww
> 60: MUL TEMP[4].x, TEMP[3].xxxx, TEMP[3].xxxx
> 61: MUL TEMP[4].x, TEMP[4].xxxx, TEMP[5].xxxx
> 62: MAD TEMP[3].x, TEMP[1].xxxx, -TEMP[4].xxxx, TEMP[6].wwww
> 63: MUL TEMP[4].x, TEMP[4].xxxx, TEMP[1].xxxx
> 64: MAD TEMP[1].x, TEMP[0].xxxx, IMM[2].zzzz, IMM[2].wwww
> 65: MUL TEMP[0].x, TEMP[0].xxxx, TEMP[0].xxxx
> 66: MUL TEMP[0].x, TEMP[0].xxxx, TEMP[1].xxxx
> 67: MAD TEMP[0].x, TEMP[0].xxxx, TEMP[3].xxxx, TEMP[4].xxxx
> 68: MOV TEMP[2].w, TEMP[0].xxxx
> 69: MOV OUT[0], TEMP[2]
> 70: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %24 = load <16 x i8>, <16 x i8> addrspace(2)* %23, align 16, !tbaa !0
>  %25 = call float @llvm.SI.load.const(<16 x i8> %24, i32 56)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %24, i32 60)
>  %27 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %28 = load <8 x i32>, <8 x i32> addrspace(2)* %27, align 32, !tbaa !0
>  %29 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %30 = load <4 x i32>, <4 x i32> addrspace(2)* %29, align 16, !tbaa !0
>  %31 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 1
>  %32 = load <8 x i32>, <8 x i32> addrspace(2)* %31, align 32, !tbaa !0
>  %33 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 1
>  %34 = load <4 x i32>, <4 x i32> addrspace(2)* %33, align 16, !tbaa !0
>  %35 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 2
>  %36 = load <8 x i32>, <8 x i32> addrspace(2)* %35, align 32, !tbaa !0
>  %37 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 2
>  %38 = load <4 x i32>, <4 x i32> addrspace(2)* %37, align 16, !tbaa !0
>  %39 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 3
>  %40 = load <8 x i32>, <8 x i32> addrspace(2)* %39, align 32, !tbaa !0
>  %41 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 3
>  %42 = load <4 x i32>, <4 x i32> addrspace(2)* %41, align 16, !tbaa !0
>  %43 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 4
>  %44 = load <8 x i32>, <8 x i32> addrspace(2)* %43, align 32, !tbaa !0
>  %45 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 4
>  %46 = load <4 x i32>, <4 x i32> addrspace(2)* %45, align 16, !tbaa !0
>  %47 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %8)
>  %48 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %8)
>  %49 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %6, <2 x i32> %8)
>  %50 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %6, <2 x i32> %8)
>  %51 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %6, <2 x i32> %8)
>  %52 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %6, <2 x i32> %8)
>  %53 = call float @llvm.SI.fs.interp(i32 2, i32 1, i32 %6, <2 x i32> %8)
>  %54 = call float @llvm.SI.fs.interp(i32 3, i32 1, i32 %6, <2 x i32> %8)
>  %55 = fmul float %26, 7.500000e-01
>  %56 = bitcast float %47 to i32
>  %57 = bitcast float %48 to i32
>  %58 = insertelement <2 x i32> undef, i32 %56, i32 0
>  %59 = insertelement <2 x i32> %58, i32 %57, i32 1
>  %60 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %59, <8 x i32> %28, <4 x i32> %30, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %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 = fsub float 1.000000e+00, %61
>  %66 = fmul float %65, %65
>  %67 = fmul float %66, %66
>  %68 = fmul float %67, 2.500000e-01
>  %69 = fadd float %68, %55
>  %70 = bitcast float %53 to i32
>  %71 = bitcast float %54 to i32
>  %72 = insertelement <2 x i32> undef, i32 %70, i32 0
>  %73 = insertelement <2 x i32> %72, i32 %71, i32 1
>  %74 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %73, <8 x i32> %40, <4 x i32> %42, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %75 = extractelement <4 x float> %74, i32 0
>  %76 = extractelement <4 x float> %74, i32 1
>  %77 = extractelement <4 x float> %74, i32 2
>  %78 = extractelement <4 x float> %74, i32 3
>  %79 = fadd float %75, -1.000000e+00
>  %80 = fadd float %76, -1.000000e+00
>  %81 = fadd float %77, -1.000000e+00
>  %82 = fadd float %78, -1.000000e+00
>  %83 = fmul float %69, %79
>  %84 = fadd float %83, 1.000000e+00
>  %85 = fmul float %69, %80
>  %86 = fadd float %85, 1.000000e+00
>  %87 = fmul float %69, %81
>  %88 = fadd float %87, 1.000000e+00
>  %89 = fmul float %69, %82
>  %90 = fadd float %89, 1.000000e+00
>  %91 = bitcast float %49 to i32
>  %92 = bitcast float %50 to i32
>  %93 = insertelement <2 x i32> undef, i32 %91, i32 0
>  %94 = insertelement <2 x i32> %93, i32 %92, i32 1
>  %95 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %94, <8 x i32> %44, <4 x i32> %46, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %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 %84, %96
>  %101 = fmul float %86, %97
>  %102 = fmul float %88, %98
>  %103 = fmul float %100, 0x3FD3333340000000
>  %104 = fmul float %101, 0x3FE2E147A0000000
>  %105 = fadd float %104, %103
>  %106 = fmul float %102, 0x3FBC28F5C0000000
>  %107 = fadd float %105, %106
>  %108 = fadd float %107, 0xBFB47AE140000000
>  %109 = fmul float %108, 0x402C924920000000
>  %110 = call float @llvm.AMDIL.clamp.(float %109, float 0.000000e+00, float 1.000000e+00)
>  %111 = fmul float %110, -2.000000e+00
>  %112 = fadd float %111, 3.000000e+00
>  %113 = fmul float %110, %110
>  %114 = fmul float %113, %112
>  %115 = fsub float 1.000000e+00, %114
>  %116 = fmul float %115, 0x3F9EB851E0000000
>  %117 = fadd float %116, %100
>  %118 = fmul float %115, 0x3F9EB851E0000000
>  %119 = fadd float %118, %101
>  %120 = fmul float %115, 0x3F9EB851E0000000
>  %121 = fadd float %120, %102
>  %122 = fmul float %63, %117
>  %123 = fmul float %63, %119
>  %124 = fmul float %63, %121
>  %125 = fmul float %122, 5.000000e-01
>  %126 = fadd float %125, %100
>  %127 = fmul float %123, 5.000000e-01
>  %128 = fadd float %127, %101
>  %129 = fmul float %124, 5.000000e-01
>  %130 = fadd float %129, %102
>  %131 = call float @llvm.AMDIL.clamp.(float %126, float 0.000000e+00, float 1.000000e+00)
>  %132 = call float @llvm.AMDIL.clamp.(float %128, float 0.000000e+00, float 1.000000e+00)
>  %133 = call float @llvm.AMDIL.clamp.(float %130, float 0.000000e+00, float 1.000000e+00)
>  %134 = fmul float %62, %131
>  %135 = fmul float %62, %132
>  %136 = fmul float %62, %133
>  %137 = bitcast float %47 to i32
>  %138 = bitcast float %48 to i32
>  %139 = insertelement <2 x i32> undef, i32 %137, i32 0
>  %140 = insertelement <2 x i32> %139, i32 %138, i32 1
>  %141 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %140, <8 x i32> %36, <4 x i32> %38, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %142 = extractelement <4 x float> %141, i32 0
>  %143 = extractelement <4 x float> %141, i32 1
>  %144 = extractelement <4 x float> %141, i32 2
>  %145 = extractelement <4 x float> %141, i32 3
>  %146 = fmul float %62, %131
>  %147 = fsub float %142, %146
>  %148 = fmul float %62, %132
>  %149 = fsub float %143, %148
>  %150 = fmul float %62, %133
>  %151 = fsub float %144, %150
>  %152 = fadd float %99, -5.000000e-01
>  %153 = fadd float %99, -1.000000e+00
>  %154 = fadd float %99, %99
>  %155 = call float @llvm.AMDIL.clamp.(float %154, float 0.000000e+00, float 1.000000e+00)
>  %156 = fmul float %152, 1.000000e+01
>  %157 = fmul float %153, -1.000000e+01
>  %158 = call float @llvm.AMDIL.clamp.(float %156, float 0.000000e+00, float 1.000000e+00)
>  %159 = call float @llvm.AMDIL.clamp.(float %157, float 0.000000e+00, float 1.000000e+00)
>  %160 = fmul float %158, -2.000000e+00
>  %161 = fadd float %160, 3.000000e+00
>  %162 = fmul float %159, -2.000000e+00
>  %163 = fadd float %162, 3.000000e+00
>  %164 = fmul float %158, %158
>  %165 = fmul float %159, %159
>  %166 = fmul float %164, %161
>  %167 = fmul float %165, %163
>  %168 = fmul float %167, %166
>  %169 = bitcast float %51 to i32
>  %170 = bitcast float %52 to i32
>  %171 = insertelement <2 x i32> undef, i32 %169, i32 0
>  %172 = insertelement <2 x i32> %171, i32 %170, i32 1
>  %173 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %172, <8 x i32> %32, <4 x i32> %34, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %174 = extractelement <4 x float> %173, i32 1
>  %175 = fmul float %174, %61
>  %176 = fadd float %175, %64
>  %177 = fmul float %62, %25
>  %178 = fmul float %90, %177
>  %179 = fmul float %26, 6.000000e+00
>  %180 = fadd float %179, 1.000000e+00
>  %181 = fmul float %176, %180
>  %182 = fadd float %181, %168
>  %183 = fmul float %155, -2.000000e+00
>  %184 = fadd float %183, 3.000000e+00
>  %185 = fmul float %155, %155
>  %186 = fmul float %185, %184
>  %187 = fmul float %182, %186
>  %188 = fadd float %187, 0xBFE28F5C20000000
>  %189 = fmul float %188, 1.000000e+01
>  %190 = call float @llvm.AMDIL.clamp.(float %189, float 0.000000e+00, float 1.000000e+00)
>  %191 = fmul float %190, -2.000000e+00
>  %192 = fadd float %191, 3.000000e+00
>  %193 = fmul float %190, %190
>  %194 = fmul float %193, %192
>  %195 = fmul float %192, %193
>  %196 = fadd float %195, 0xBFECCCCCC0000000
>  %197 = fmul float %192, %193
>  %198 = fadd float %197, 0xBF847AE140000000
>  %199 = fmul float %196, 1.000000e+01
>  %200 = fmul float %198, -1.000000e+02
>  %201 = fmul float %194, %147
>  %202 = fadd float %201, %134
>  %203 = fmul float %194, %149
>  %204 = fadd float %203, %135
>  %205 = fmul float %194, %151
>  %206 = fadd float %205, %136
>  %207 = call float @llvm.maxnum.f32(float %200, float 0.000000e+00)
>  %208 = call float @llvm.AMDIL.clamp.(float %199, float 0.000000e+00, float 1.000000e+00)
>  %209 = fmul float %207, -2.000000e+00
>  %210 = fadd float %209, 3.000000e+00
>  %211 = fmul float %207, %207
>  %212 = fmul float %211, %210
>  %213 = fmul float %212, %178
>  %214 = fsub float %145, %213
>  %215 = fmul float %212, %178
>  %216 = fmul float %208, -2.000000e+00
>  %217 = fadd float %216, 3.000000e+00
>  %218 = fmul float %208, %208
>  %219 = fmul float %218, %217
>  %220 = fmul float %219, %214
>  %221 = fadd float %220, %215
>  %222 = call i32 @llvm.SI.packf16(float %202, float %204)
>  %223 = bitcast i32 %222 to float
>  %224 = call i32 @llvm.SI.packf16(float %206, float %221)
>  %225 = bitcast i32 %224 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %223, float %225, float %223, float %225)
>  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.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #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 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, s10                                                            ; BEFC030A
>	v_mov_b32_e32 v2, 0x3f400000                                                 ; 7E0402FF 3F400000
>	v_mov_b32_e32 v3, 0xbda3d70a                                                 ; 7E0602FF BDA3D70A
>	v_mov_b32_e32 v4, 0x40400000                                                 ; 7E0802FF 40400000
>	v_mov_b32_e32 v5, 0x3cf5c28f                                                 ; 7E0A02FF 3CF5C28F
>	v_mov_b32_e32 v6, 0x41200000                                                 ; 7E0C02FF 41200000
>	v_mov_b32_e32 v7, 0x40c00000                                                 ; 7E0E02FF 40C00000
>	v_mov_b32_e32 v8, 0xc2c80000                                                 ; 7E1002FF C2C80000
>	v_interp_p1_f32 v9, v0, 0, 0, [m0]                                           ; C8240000
>	v_interp_p2_f32 v9, [v9], v1, 0, 0, [m0]                                     ; C8250001
>	v_interp_p1_f32 v10, v0, 1, 0, [m0]                                          ; C8280100
>	v_interp_p2_f32 v10, [v10], v1, 1, 0, [m0]                                   ; C8290101
>	v_interp_p1_f32 v11, v0, 2, 0, [m0]                                          ; C82C0200
>	v_interp_p2_f32 v11, [v11], v1, 2, 0, [m0]                                   ; C82D0201
>	v_interp_p1_f32 v12, v0, 3, 0, [m0]                                          ; C8300300
>	v_interp_p2_f32 v12, [v12], v1, 3, 0, [m0]                                   ; C8310301
>	v_interp_p1_f32 v13, v0, 0, 1, [m0]                                          ; C8340400
>	v_interp_p2_f32 v13, [v13], v1, 0, 1, [m0]                                   ; C8350401
>	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 v15, v0, 2, 1, [m0]                                          ; C83C0600
>	v_interp_p2_f32 v15, [v15], v1, 2, 1, [m0]                                   ; C83D0601
>	v_interp_p1_f32 v16, v0, 3, 1, [m0]                                          ; C8400700
>	v_interp_p2_f32 v16, [v16], v1, 3, 1, [m0]                                   ; C8410701
>	s_load_dwordx4 s[28:31], s[2:3], 0x0                                         ; C08E0300
>	s_load_dwordx4 s[32:35], s[4:5], 0x0                                         ; C0900500
>	s_load_dwordx4 s[0:3], s[4:5], 0x4                                           ; C0800504
>	s_load_dwordx4 s[16:19], s[4:5], 0x8                                         ; C0880508
>	s_load_dwordx4 s[36:39], s[4:5], 0xc                                         ; C092050C
>	s_load_dwordx4 s[40:43], s[4:5], 0x10                                        ; C0940510
>	s_load_dwordx8 s[44:51], s[6:7], 0x0                                         ; C0D60700
>	s_load_dwordx8 s[8:15], s[6:7], 0x8                                          ; C0C40708
>	s_load_dwordx8 s[20:27], s[6:7], 0x10                                        ; C0CA0710
>	s_load_dwordx8 s[52:59], s[6:7], 0x18                                        ; C0DA0718
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	s_buffer_load_dword s4, s[28:31], 0xe                                        ; C2021D0E
>	s_buffer_load_dword s5, s[28:31], 0xf                                        ; C2029D0F
>	s_load_dwordx8 s[60:67], s[6:7], 0x20                                        ; C0DE0720
>	image_sample v[17:20], 15, 0, 0, 0, 0, 0, 0, 0, v[9:10], s[44:51], s[32:35]  ; F0800F00 010B1109
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_sub_f32_e32 v0, 1.0, v17                                                   ; 080022F2
>	v_mad_f32 v0, -v17, v0, v0                                                   ; D2820000 24020111
>	image_sample v[21:24], 15, 0, 0, 0, 0, 0, 0, 0, v[15:16], s[52:59], s[36:39] ; F0800F00 012D150F
>	v_mul_f32_e32 v0, v0, v0                                                     ; 10000100
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v1, s5, v2                                                     ; 10020405
>	image_sample v[25:28], 15, 0, 0, 0, 0, 0, 0, 0, v[11:12], s[60:67], s[40:43] ; F0800F00 014F190B
>	v_madmk_f32_e32 v0, v0, v1, 0x3e800000                                       ; 40000300 3E800000
>	s_waitcnt vmcnt(1)                                                           ; BF8C0771
>	v_mad_f32 v1, v21, v0, -v0                                                   ; D2820001 84020115
>	v_mad_f32 v2, v22, v0, -v0                                                   ; D2820002 84020116
>	v_mad_f32 v11, v23, v0, -v0                                                  ; D282000B 84020117
>	v_mad_f32 v0, v24, v0, -v0                                                   ; D2820000 84020118
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v1, v25, v1, v25                                                   ; D2820001 04660319
>	v_mad_f32 v2, v26, v2, v26                                                   ; D2820002 046A051A
>	v_mad_f32 v11, v27, v11, v27                                                 ; D282000B 046E171B
>	v_mul_f32_e32 v12, 0x3e99999a, v1                                            ; 101802FF 3E99999A
>	v_madmk_f32_e32 v12, v2, v12, 0x3f170a3d                                     ; 40181902 3F170A3D
>	v_madmk_f32_e32 v12, v11, v12, 0x3de147ae                                    ; 4018190B 3DE147AE
>	v_add_f32_e32 v3, v12, v3                                                    ; 0606070C
>	v_mul_f32_e32 v3, 0x41649249, v3                                             ; 100606FF 41649249
>	v_add_f32_e64 v3, 0, v3 clamp                                                ; D2060803 00020680
>	v_mad_f32 v12, -2.0, v3, v4                                                  ; D282000C 041206F5
>	v_mul_f32_e32 v3, v3, v3                                                     ; 10060703
>	v_mad_f32 v3, -v3, v12, 1.0                                                  ; D2820003 23CA1903
>	v_mad_f32 v12, v5, v3, v1                                                    ; D282000C 04060705
>	v_mad_f32 v15, v5, v3, v2                                                    ; D282000F 040A0705
>	v_mad_f32 v3, v5, v3, v11                                                    ; D2820003 042E0705
>	v_mul_f32_e32 v5, v12, v19                                                   ; 100A270C
>	v_mul_f32_e32 v12, v15, v19                                                  ; 1018270F
>	v_mul_f32_e32 v3, v3, v19                                                    ; 10062703
>	v_mac_f32_e32 v1, 0.5, v5                                                    ; 3E020AF0
>	v_mac_f32_e32 v2, 0.5, v12                                                   ; 3E0418F0
>	v_mac_f32_e32 v11, 0.5, v3                                                   ; 3E1606F0
>	v_add_f32_e64 v1, 0, v1 clamp                                                ; D2060801 00020280
>	v_add_f32_e64 v2, 0, v2 clamp                                                ; D2060802 00020480
>	v_add_f32_e64 v3, 0, v11 clamp                                               ; D2060803 00021680
>	v_mul_f32_e32 v5, v1, v18                                                    ; 100A2501
>	v_mul_f32_e32 v11, v2, v18                                                   ; 10162502
>	v_add_f32_e32 v12, -0.5, v28                                                 ; 061838F1
>	v_add_f32_e32 v15, v28, v28                                                  ; 061E391C
>	v_madmk_f32_e32 v16, v28, v6, 0xc1200000                                     ; 40200D1C C1200000
>	v_mul_f32_e32 v19, v3, v18                                                   ; 10262503
>	v_mul_f32_e32 v12, v6, v12                                                   ; 10181906
>	v_add_f32_e64 v12, 0, v12 clamp                                              ; D206080C 00021880
>	v_add_f32_e64 v16, 0, v16 clamp                                              ; D2060810 00022080
>	v_mad_f32 v21, -2.0, v12, v4                                                 ; D2820015 041218F5
>	v_mad_f32 v22, -2.0, v16, v4                                                 ; D2820016 041220F5
>	v_mul_f32_e32 v12, v12, v12                                                  ; 1018190C
>	v_mul_f32_e32 v16, v16, v16                                                  ; 10202110
>	v_mul_f32_e32 v12, v21, v12                                                  ; 10181915
>	v_mul_f32_e32 v16, v22, v16                                                  ; 10202116
>	image_sample v[21:24], 15, 0, 0, 0, 0, 0, 0, 0, v[9:10], s[20:27], s[16:19]  ; F0800F00 00851509
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v1, -v18, v1, v21                                                  ; D2820001 24560312
>	v_mad_f32 v2, -v18, v2, v22                                                  ; D2820002 245A0512
>	v_mad_f32 v3, -v18, v3, v23                                                  ; D2820003 245E0712
>	v_add_f32_e64 v9, 0, v15 clamp                                               ; D2060809 00021E80
>	v_mul_f32_e32 v10, v12, v16                                                  ; 1014210C
>	image_sample v12, 2, 0, 0, 0, 0, 0, 0, 0, v[13:14], s[8:15], s[0:3]          ; F0800200 00020C0D
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mac_f32_e32 v20, v17, v12                                                  ; 3E281911
>	v_mul_f32_e32 v12, s4, v18                                                   ; 10182404
>	v_mac_f32_e32 v12, v12, v0                                                   ; 3E18010C
>	v_mad_f32 v0, v7, s5, 1.0                                                    ; D2820000 03C80B07
>	v_mac_f32_e32 v10, v0, v20                                                   ; 3E142900
>	v_mad_f32 v0, -2.0, v9, v4                                                   ; D2820000 041212F5
>	v_mul_f32_e32 v7, v9, v9                                                     ; 100E1309
>	v_mul_f32_e32 v0, v0, v7                                                     ; 10000F00
>	v_madak_f32_e32 v0, v10, v0, 0xbf147ae1                                      ; 4200010A BF147AE1
>	v_mul_f32_e32 v0, v6, v0                                                     ; 10000106
>	v_add_f32_e64 v0, 0, v0 clamp                                                ; D2060800 00020080
>	v_mad_f32 v7, -2.0, v0, v4                                                   ; D2820007 041200F5
>	v_mul_f32_e32 v0, v0, v0                                                     ; 10000100
>	v_mul_f32_e32 v9, v7, v0                                                     ; 10120107
>	v_madak_f32_e32 v10, v0, v7, 0xbf666666                                      ; 42140F00 BF666666
>	v_madak_f32_e32 v0, v0, v7, 0xbc23d70a                                       ; 42000F00 BC23D70A
>	v_mul_f32_e32 v6, v6, v10                                                    ; 100C1506
>	v_mul_f32_e32 v0, v0, v8                                                     ; 10001100
>	v_mac_f32_e32 v5, v1, v9                                                     ; 3E0A1301
>	v_mac_f32_e32 v11, v2, v9                                                    ; 3E161302
>	v_mac_f32_e32 v19, v3, v9                                                    ; 3E261303
>	v_max_f32_e32 v0, 0, v0                                                      ; 20000080
>	v_add_f32_e64 v1, 0, v6 clamp                                                ; D2060801 00020C80
>	v_mad_f32 v2, -2.0, v0, v4                                                   ; D2820002 041200F5
>	v_mul_f32_e32 v0, v0, v0                                                     ; 10000100
>	v_mul_f32_e32 v0, v2, v0                                                     ; 10000102
>	v_mul_f32_e32 v2, v12, v0                                                    ; 1004010C
>	v_mad_f32 v0, -v0, v12, v24                                                  ; D2820000 24621900
>	v_mac_f32_e32 v4, -2.0, v1                                                   ; 3E0802F5
>	v_mul_f32_e32 v1, v1, v1                                                     ; 10020301
>	v_mul_f32_e32 v1, v4, v1                                                     ; 10020304
>	v_mac_f32_e32 v2, v0, v1                                                     ; 3E040300
>	v_cvt_pkrtz_f16_f32_e32 v0, v5, v11                                          ; 5E001705
>	v_cvt_pkrtz_f16_f32_e32 v1, v19, v2                                          ; 5E020513
>	exp 15, 0, 1, 1, 1, v0, v1, v0, v1                                           ; F8001C0F 01000100
>	s_endpgm                                                                     ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 72
>VGPRS: 32
>Code Size: 772 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
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL OUT[1], CLIPVERTEX
>DCL OUT[2], GENERIC[0]
>DCL OUT[3], GENERIC[1]
>DCL CONST[0..54]
>DCL TEMP[0..3], LOCAL
>  0: MUL TEMP[0].xy, CONST[48].xyyy, IN[1].xyyy
>  1: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
>  2: ADD TEMP[1].x, TEMP[0].xxxx, CONST[48].wwww
>  3: MOV TEMP[1].z, TEMP[1].xxxx
>  4: MUL TEMP[0].xy, CONST[49].xyyy, IN[1].xyyy
>  5: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
>  6: ADD TEMP[2].x, TEMP[0].xxxx, CONST[49].wwww
>  7: MOV TEMP[1].w, TEMP[2].xxxx
>  8: MUL TEMP[0].xy, CONST[50].xyyy, IN[1].xyyy
>  9: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 10: ADD TEMP[2].x, TEMP[0].xxxx, CONST[50].wwww
> 11: MUL TEMP[0].xy, CONST[51].xyyy, IN[1].xyyy
> 12: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 13: ADD TEMP[3].x, TEMP[0].xxxx, CONST[51].wwww
> 14: MOV TEMP[2].y, TEMP[3].xxxx
> 15: MUL TEMP[0].xy, CONST[52].xyyy, IN[1].xyyy
> 16: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 17: ADD TEMP[3].x, TEMP[0].xxxx, CONST[52].wwww
> 18: MOV TEMP[2].z, TEMP[3].xxxx
> 19: MUL TEMP[0].xy, CONST[53].xyyy, IN[1].xyyy
> 20: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 21: ADD TEMP[0].x, TEMP[0].xxxx, CONST[53].wwww
> 22: MOV TEMP[2].w, TEMP[0].xxxx
> 23: MOV TEMP[0].xw, IN[0].xxxw
> 24: MOV TEMP[1].xy, IN[1].xyxx
> 25: MAD TEMP[3].x, IN[0].zzzz, CONST[0].zzzz, -IN[0].wwww
> 26: MOV TEMP[0].z, TEMP[3].xxxx
> 27: MOV TEMP[0].y, -IN[0].yyyy
> 28: MAD TEMP[0].xy, CONST[54].xyyy, IN[0].wwww, TEMP[0].xyyy
> 29: MOV OUT[2], TEMP[1]
> 30: MOV OUT[3], TEMP[2]
> 31: MOV OUT[0], TEMP[0]
> 32: MOV OUT[1], IN[0]
> 33: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = call float @llvm.SI.load.const(<16 x i8> %13, i32 8)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %13, i32 768)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %13, i32 772)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %13, i32 780)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %13, i32 784)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %13, i32 788)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %13, i32 796)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %13, i32 800)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %13, i32 804)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %13, i32 812)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %13, i32 816)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %13, i32 820)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %13, i32 828)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %13, i32 832)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %13, i32 836)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %13, i32 844)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %13, i32 848)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %13, i32 852)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %13, i32 860)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %13, i32 864)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %13, i32 868)
>  %35 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %36 = load <16 x i8>, <16 x i8> addrspace(2)* %35, align 16, !tbaa !0
>  %37 = add i32 %5, %8
>  %38 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %36, i32 0, i32 %37)
>  %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 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 1
>  %44 = load <16 x i8>, <16 x i8> addrspace(2)* %43, align 16, !tbaa !0
>  %45 = add i32 %5, %8
>  %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 = fmul float %15, %47
>  %50 = fmul float %16, %48
>  %51 = fadd float %50, %49
>  %52 = fadd float %51, %17
>  %53 = fmul float %18, %47
>  %54 = fmul float %19, %48
>  %55 = fadd float %54, %53
>  %56 = fadd float %55, %20
>  %57 = fmul float %21, %47
>  %58 = fmul float %22, %48
>  %59 = fadd float %58, %57
>  %60 = fadd float %59, %23
>  %61 = fmul float %24, %47
>  %62 = fmul float %25, %48
>  %63 = fadd float %62, %61
>  %64 = fadd float %63, %26
>  %65 = fmul float %27, %47
>  %66 = fmul float %28, %48
>  %67 = fadd float %66, %65
>  %68 = fadd float %67, %29
>  %69 = fmul float %30, %47
>  %70 = fmul float %31, %48
>  %71 = fadd float %70, %69
>  %72 = fadd float %71, %32
>  %73 = fmul float %41, %14
>  %74 = fsub float %73, %42
>  %75 = fmul float %33, %42
>  %76 = fadd float %75, %39
>  %77 = fmul float %34, %42
>  %78 = fsub float %77, %40
>  %79 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 16
>  %80 = load <16 x i8>, <16 x i8> addrspace(2)* %79, align 16, !tbaa !0
>  %81 = call float @llvm.SI.load.const(<16 x i8> %80, i32 0)
>  %82 = fmul float %81, %39
>  %83 = call float @llvm.SI.load.const(<16 x i8> %80, i32 4)
>  %84 = fmul float %83, %40
>  %85 = fadd float %82, %84
>  %86 = call float @llvm.SI.load.const(<16 x i8> %80, i32 8)
>  %87 = fmul float %86, %41
>  %88 = fadd float %85, %87
>  %89 = call float @llvm.SI.load.const(<16 x i8> %80, i32 12)
>  %90 = fmul float %89, %42
>  %91 = fadd float %88, %90
>  %92 = call float @llvm.SI.load.const(<16 x i8> %80, i32 16)
>  %93 = fmul float %92, %39
>  %94 = call float @llvm.SI.load.const(<16 x i8> %80, i32 20)
>  %95 = fmul float %94, %40
>  %96 = fadd float %93, %95
>  %97 = call float @llvm.SI.load.const(<16 x i8> %80, i32 24)
>  %98 = fmul float %97, %41
>  %99 = fadd float %96, %98
>  %100 = call float @llvm.SI.load.const(<16 x i8> %80, i32 28)
>  %101 = fmul float %100, %42
>  %102 = fadd float %99, %101
>  %103 = call float @llvm.SI.load.const(<16 x i8> %80, i32 32)
>  %104 = fmul float %103, %39
>  %105 = call float @llvm.SI.load.const(<16 x i8> %80, i32 36)
>  %106 = fmul float %105, %40
>  %107 = fadd float %104, %106
>  %108 = call float @llvm.SI.load.const(<16 x i8> %80, i32 40)
>  %109 = fmul float %108, %41
>  %110 = fadd float %107, %109
>  %111 = call float @llvm.SI.load.const(<16 x i8> %80, i32 44)
>  %112 = fmul float %111, %42
>  %113 = fadd float %110, %112
>  %114 = call float @llvm.SI.load.const(<16 x i8> %80, i32 48)
>  %115 = fmul float %114, %39
>  %116 = call float @llvm.SI.load.const(<16 x i8> %80, i32 52)
>  %117 = fmul float %116, %40
>  %118 = fadd float %115, %117
>  %119 = call float @llvm.SI.load.const(<16 x i8> %80, i32 56)
>  %120 = fmul float %119, %41
>  %121 = fadd float %118, %120
>  %122 = call float @llvm.SI.load.const(<16 x i8> %80, i32 60)
>  %123 = fmul float %122, %42
>  %124 = fadd float %121, %123
>  %125 = call float @llvm.SI.load.const(<16 x i8> %80, i32 64)
>  %126 = fmul float %125, %39
>  %127 = call float @llvm.SI.load.const(<16 x i8> %80, i32 68)
>  %128 = fmul float %127, %40
>  %129 = fadd float %126, %128
>  %130 = call float @llvm.SI.load.const(<16 x i8> %80, i32 72)
>  %131 = fmul float %130, %41
>  %132 = fadd float %129, %131
>  %133 = call float @llvm.SI.load.const(<16 x i8> %80, i32 76)
>  %134 = fmul float %133, %42
>  %135 = fadd float %132, %134
>  %136 = call float @llvm.SI.load.const(<16 x i8> %80, i32 80)
>  %137 = fmul float %136, %39
>  %138 = call float @llvm.SI.load.const(<16 x i8> %80, i32 84)
>  %139 = fmul float %138, %40
>  %140 = fadd float %137, %139
>  %141 = call float @llvm.SI.load.const(<16 x i8> %80, i32 88)
>  %142 = fmul float %141, %41
>  %143 = fadd float %140, %142
>  %144 = call float @llvm.SI.load.const(<16 x i8> %80, i32 92)
>  %145 = fmul float %144, %42
>  %146 = fadd float %143, %145
>  %147 = call float @llvm.SI.load.const(<16 x i8> %80, i32 96)
>  %148 = fmul float %147, %39
>  %149 = call float @llvm.SI.load.const(<16 x i8> %80, i32 100)
>  %150 = fmul float %149, %40
>  %151 = fadd float %148, %150
>  %152 = call float @llvm.SI.load.const(<16 x i8> %80, i32 104)
>  %153 = fmul float %152, %41
>  %154 = fadd float %151, %153
>  %155 = call float @llvm.SI.load.const(<16 x i8> %80, i32 108)
>  %156 = fmul float %155, %42
>  %157 = fadd float %154, %156
>  %158 = call float @llvm.SI.load.const(<16 x i8> %80, i32 112)
>  %159 = fmul float %158, %39
>  %160 = call float @llvm.SI.load.const(<16 x i8> %80, i32 116)
>  %161 = fmul float %160, %40
>  %162 = fadd float %159, %161
>  %163 = call float @llvm.SI.load.const(<16 x i8> %80, i32 120)
>  %164 = fmul float %163, %41
>  %165 = fadd float %162, %164
>  %166 = call float @llvm.SI.load.const(<16 x i8> %80, i32 124)
>  %167 = fmul float %166, %42
>  %168 = fadd float %165, %167
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %47, float %48, float %52, float %56)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %60, float %64, float %68, float %72)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 12, i32 0, float %76, float %78, float %74, float %42)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 13, i32 0, float %91, float %102, float %113, float %124)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 14, i32 0, float %135, float %146, float %157, 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
>
>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[4:7], s[8:9], 0x0                    ; C0820900
>	s_load_dwordx4 s[12:15], s[8:9], 0x4                  ; C0860904
>	v_add_i32_e32 v0, vcc, s10, v0                        ; 4A00000A
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[4:7], 0 idxen   ; E00C2000 80010100
>	buffer_load_format_xyzw v[5:8], v0, s[12:15], 0 idxen ; E00C2000 80030500
>	s_load_dwordx4 s[4:7], s[2:3], 0x0                    ; C0820300
>	s_load_dwordx4 s[0:3], s[2:3], 0x40                   ; C0800340
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	s_buffer_load_dword s8, s[4:7], 0x2                   ; C2040502
>	s_buffer_load_dword s9, s[4:7], 0xc0                  ; C20485C0
>	s_buffer_load_dword s10, s[4:7], 0xc1                 ; C20505C1
>	s_buffer_load_dword s11, s[4:7], 0xc3                 ; C20585C3
>	s_buffer_load_dword s12, s[4:7], 0xc4                 ; C20605C4
>	s_buffer_load_dword s13, s[4:7], 0xc5                 ; C20685C5
>	s_buffer_load_dword s14, s[4:7], 0xc7                 ; C20705C7
>	s_buffer_load_dword s15, s[4:7], 0xc8                 ; C20785C8
>	s_buffer_load_dword s16, s[4:7], 0xc9                 ; C20805C9
>	s_buffer_load_dword s17, s[4:7], 0xcb                 ; C20885CB
>	s_buffer_load_dword s18, s[4:7], 0xcc                 ; C20905CC
>	s_buffer_load_dword s19, s[4:7], 0xcd                 ; C20985CD
>	s_buffer_load_dword s20, s[4:7], 0xcf                 ; C20A05CF
>	s_buffer_load_dword s21, s[4:7], 0xd0                 ; C20A85D0
>	s_buffer_load_dword s22, s[4:7], 0xd1                 ; C20B05D1
>	s_buffer_load_dword s23, s[4:7], 0xd3                 ; C20B85D3
>	s_buffer_load_dword s24, s[4:7], 0xd4                 ; C20C05D4
>	s_buffer_load_dword s25, s[4:7], 0xd5                 ; C20C85D5
>	s_buffer_load_dword s26, s[4:7], 0xd7                 ; C20D05D7
>	s_buffer_load_dword s27, s[4:7], 0xd8                 ; C20D85D8
>	s_buffer_load_dword s4, s[4:7], 0xd9                  ; C20205D9
>	s_buffer_load_dword s5, s[0:3], 0x0                   ; C2028100
>	s_buffer_load_dword s6, s[0:3], 0x1                   ; C2030101
>	s_buffer_load_dword s7, s[0:3], 0x2                   ; C2038102
>	s_buffer_load_dword s28, s[0:3], 0x3                  ; C20E0103
>	s_buffer_load_dword s29, s[0:3], 0x4                  ; C20E8104
>	s_buffer_load_dword s30, s[0:3], 0x5                  ; C20F0105
>	s_buffer_load_dword s31, s[0:3], 0x6                  ; C20F8106
>	s_buffer_load_dword s32, s[0:3], 0x7                  ; C2100107
>	s_buffer_load_dword s33, s[0:3], 0x8                  ; C2108108
>	s_buffer_load_dword s34, s[0:3], 0x9                  ; C2110109
>	s_buffer_load_dword s35, s[0:3], 0xa                  ; C211810A
>	s_buffer_load_dword s36, s[0:3], 0xb                  ; C212010B
>	s_buffer_load_dword s37, s[0:3], 0xc                  ; C212810C
>	s_buffer_load_dword s38, s[0:3], 0xd                  ; C213010D
>	s_buffer_load_dword s39, s[0:3], 0xe                  ; C213810E
>	s_buffer_load_dword s40, s[0:3], 0xf                  ; C214010F
>	s_buffer_load_dword s41, s[0:3], 0x10                 ; C2148110
>	s_buffer_load_dword s42, s[0:3], 0x11                 ; C2150111
>	s_buffer_load_dword s43, s[0:3], 0x12                 ; C2158112
>	s_buffer_load_dword s44, s[0:3], 0x13                 ; C2160113
>	s_buffer_load_dword s45, s[0:3], 0x14                 ; C2168114
>	s_buffer_load_dword s46, s[0:3], 0x15                 ; C2170115
>	s_buffer_load_dword s47, s[0:3], 0x16                 ; C2178116
>	s_buffer_load_dword s48, s[0:3], 0x17                 ; C2180117
>	s_buffer_load_dword s49, s[0:3], 0x18                 ; C2188118
>	s_buffer_load_dword s50, s[0:3], 0x19                 ; C2190119
>	s_buffer_load_dword s51, s[0:3], 0x1a                 ; C219811A
>	s_buffer_load_dword s52, s[0:3], 0x1b                 ; C21A011B
>	s_buffer_load_dword s53, s[0:3], 0x1c                 ; C21A811C
>	s_buffer_load_dword s54, s[0:3], 0x1d                 ; C21B011D
>	s_buffer_load_dword s55, s[0:3], 0x1e                 ; C21B811E
>	s_buffer_load_dword s0, s[0:3], 0x1f                  ; C200011F
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mov_b32_e32 v0, s11                                 ; 7E00020B
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mov_b32_e32 v7, s14                                 ; 7E0E020E
>	v_mov_b32_e32 v8, s17                                 ; 7E100211
>	v_mov_b32_e32 v9, s20                                 ; 7E120214
>	v_mov_b32_e32 v10, s23                                ; 7E140217
>	v_mov_b32_e32 v11, s26                                ; 7E16021A
>	v_mac_f32_e32 v0, s9, v5                              ; 3E000A09
>	v_mac_f32_e32 v7, s12, v5                             ; 3E0E0A0C
>	v_mac_f32_e32 v8, s15, v5                             ; 3E100A0F
>	v_mac_f32_e32 v9, s18, v5                             ; 3E120A12
>	v_mac_f32_e32 v10, s21, v5                            ; 3E140A15
>	v_mac_f32_e32 v11, s24, v5                            ; 3E160A18
>	v_mac_f32_e32 v0, s10, v6                             ; 3E000C0A
>	v_mac_f32_e32 v7, s13, v6                             ; 3E0E0C0D
>	v_mac_f32_e32 v8, s16, v6                             ; 3E100C10
>	v_mac_f32_e32 v9, s19, v6                             ; 3E120C13
>	exp 15, 32, 0, 0, 0, v5, v6, v0, v7                   ; F800020F 07000605
>	v_mac_f32_e32 v10, s22, v6                            ; 3E140C16
>	v_mac_f32_e32 v11, s25, v6                            ; 3E160C19
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mad_f32 v0, v3, s8, -v4                             ; D2820000 84101103
>	v_mad_f32 v5, s27, v4, v1                             ; D2820005 0406081B
>	v_mad_f32 v6, s4, v4, -v2                             ; D2820006 840A0804
>	v_mul_f32_e32 v7, s6, v2                              ; 100E0406
>	exp 15, 33, 0, 0, 0, v8, v9, v10, v11                 ; F800021F 0B0A0908
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mul_f32_e32 v8, s30, v2                             ; 1010041E
>	v_mac_f32_e32 v7, s5, v1                              ; 3E0E0205
>	exp 15, 12, 0, 0, 0, v5, v6, v0, v4                   ; F80000CF 04000605
>	v_mac_f32_e32 v8, s29, v1                             ; 3E10021D
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mul_f32_e32 v0, s34, v2                             ; 10000422
>	v_mac_f32_e32 v0, s33, v1                             ; 3E000221
>	v_mul_f32_e32 v5, s38, v2                             ; 100A0426
>	v_mac_f32_e32 v5, s37, v1                             ; 3E0A0225
>	v_mul_f32_e32 v6, s42, v2                             ; 100C042A
>	v_mul_f32_e32 v9, s46, v2                             ; 1012042E
>	v_mul_f32_e32 v10, s50, v2                            ; 10140432
>	v_mul_f32_e32 v2, s54, v2                             ; 10040436
>	v_mac_f32_e32 v6, s41, v1                             ; 3E0C0229
>	v_mac_f32_e32 v9, s45, v1                             ; 3E12022D
>	v_mac_f32_e32 v10, s49, v1                            ; 3E140231
>	v_mac_f32_e32 v2, s53, v1                             ; 3E040235
>	v_mac_f32_e32 v7, s7, v3                              ; 3E0E0607
>	v_mac_f32_e32 v8, s31, v3                             ; 3E10061F
>	v_mac_f32_e32 v0, s35, v3                             ; 3E000623
>	v_mac_f32_e32 v5, s39, v3                             ; 3E0A0627
>	v_mac_f32_e32 v6, s43, v3                             ; 3E0C062B
>	v_mac_f32_e32 v9, s47, v3                             ; 3E12062F
>	v_mac_f32_e32 v10, s51, v3                            ; 3E140633
>	v_mac_f32_e32 v2, s55, v3                             ; 3E040637
>	v_mac_f32_e32 v7, s28, v4                             ; 3E0E081C
>	v_mac_f32_e32 v8, s32, v4                             ; 3E100820
>	v_mac_f32_e32 v0, s36, v4                             ; 3E000824
>	v_mac_f32_e32 v5, s40, v4                             ; 3E0A0828
>	v_mac_f32_e32 v6, s44, v4                             ; 3E0C082C
>	v_mac_f32_e32 v9, s48, v4                             ; 3E120830
>	v_mac_f32_e32 v10, s52, v4                            ; 3E140834
>	v_mac_f32_e32 v2, s0, v4                              ; 3E040800
>	exp 15, 13, 0, 0, 0, v7, v8, v0, v5                   ; F80000DF 05000807
>	exp 15, 14, 0, 1, 0, v6, v9, v10, v2                  ; F80008EF 020A0906
>	s_endpgm                                              ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 64
>VGPRS: 12
>Code Size: 544 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
>  clamp_color = 0
>FRAG
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL IN[1], GENERIC[1], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SAMP[1]
>DCL SAMP[2]
>DCL SAMP[3]
>DCL SVIEW[0], 2D, FLOAT
>DCL SVIEW[1], 2D, FLOAT
>DCL SVIEW[2], 2D, FLOAT
>DCL SVIEW[3], 2D, FLOAT
>DCL CONST[0..3]
>DCL TEMP[0..3], LOCAL
>IMM[0] FLT32 {    6.0000,     1.0000,    -0.9900,   100.0000}
>IMM[1] FLT32 {   -2.0000,     3.0000,    -1.0000,     0.0000}
>IMM[2] FLT32 {    1.0000,     0.0000,     0.0000,     0.0000}
>  0: MOV TEMP[0].xy, IN[0].xyyy
>  1: TEX TEMP[0].xw, TEMP[0], SAMP[0], 2D
>  2: MOV TEMP[1].xy, IN[1].xyyy
>  3: TEX TEMP[1].y, TEMP[1], SAMP[1], 2D
>  4: MAD TEMP[0].x, TEMP[1].yyyy, TEMP[0].xxxx, TEMP[0].wwww
>  5: MAD TEMP[1].x, CONST[3].wwww, IMM[0].xxxx, IMM[0].yyyy
>  6: MAD TEMP[0].x, TEMP[0].xxxx, TEMP[1].xxxx, IMM[0].zzzz
>  7: MUL TEMP[0].x, TEMP[0].xxxx, IMM[0].wwww
>  8: MOV_SAT TEMP[1].x, TEMP[0].xxxx
>  9: MAD TEMP[2].x, TEMP[1].xxxx, IMM[1].xxxx, IMM[1].yyyy
> 10: MUL TEMP[0].x, TEMP[1].xxxx, TEMP[1].xxxx
> 11: MAD TEMP[0].x, TEMP[2].xxxx, TEMP[0].xxxx, IMM[0].yyyy
> 12: MOV TEMP[1].xy, IN[0].xyyy
> 13: TEX TEMP[1].x, TEMP[1], SAMP[3], 2D
> 14: ADD TEMP[0].x, TEMP[0].xxxx, -TEMP[1].xxxx
> 15: MOV_SAT TEMP[0].x, TEMP[0].xxxx
> 16: MOV TEMP[1].xy, IN[0].xyyy
> 17: TEX TEMP[1].xyz, TEMP[1], SAMP[2], 2D
> 18: ADD TEMP[2].x, TEMP[1].xxxx, -CONST[3].yyyy
> 19: MAD TEMP[2].x, TEMP[0].xxxx, TEMP[2].xxxx, CONST[3].yyyy
> 20: ADD TEMP[3].x, TEMP[1].yyyy, IMM[1].zzzz
> 21: MAD TEMP[1].xy, TEMP[1].zzzz, IMM[2].xyyy, IMM[2].yxxx
> 22: MOV TEMP[2].zw, TEMP[1].yyxy
> 23: MAD TEMP[0].x, TEMP[0].xxxx, TEMP[3].xxxx, IMM[0].yyyy
> 24: MOV TEMP[2].y, TEMP[0].xxxx
> 25: MOV OUT[0], TEMP[2]
> 26: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %24 = load <16 x i8>, <16 x i8> addrspace(2)* %23, align 16, !tbaa !0
>  %25 = call float @llvm.SI.load.const(<16 x i8> %24, i32 52)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %24, i32 60)
>  %27 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %28 = load <8 x i32>, <8 x i32> addrspace(2)* %27, align 32, !tbaa !0
>  %29 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %30 = load <4 x i32>, <4 x i32> addrspace(2)* %29, align 16, !tbaa !0
>  %31 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 1
>  %32 = load <8 x i32>, <8 x i32> addrspace(2)* %31, align 32, !tbaa !0
>  %33 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 1
>  %34 = load <4 x i32>, <4 x i32> addrspace(2)* %33, align 16, !tbaa !0
>  %35 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 2
>  %36 = load <8 x i32>, <8 x i32> addrspace(2)* %35, align 32, !tbaa !0
>  %37 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 2
>  %38 = load <4 x i32>, <4 x i32> addrspace(2)* %37, align 16, !tbaa !0
>  %39 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 3
>  %40 = load <8 x i32>, <8 x i32> addrspace(2)* %39, align 32, !tbaa !0
>  %41 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 3
>  %42 = load <4 x i32>, <4 x i32> addrspace(2)* %41, align 16, !tbaa !0
>  %43 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %8)
>  %44 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %8)
>  %45 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %6, <2 x i32> %8)
>  %46 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %6, <2 x i32> %8)
>  %47 = bitcast float %43 to i32
>  %48 = bitcast float %44 to i32
>  %49 = insertelement <2 x i32> undef, i32 %47, i32 0
>  %50 = insertelement <2 x i32> %49, i32 %48, i32 1
>  %51 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %50, <8 x i32> %28, <4 x i32> %30, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %52 = extractelement <4 x float> %51, i32 0
>  %53 = extractelement <4 x float> %51, i32 3
>  %54 = bitcast float %45 to i32
>  %55 = bitcast float %46 to i32
>  %56 = insertelement <2 x i32> undef, i32 %54, i32 0
>  %57 = insertelement <2 x i32> %56, i32 %55, i32 1
>  %58 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %57, <8 x i32> %32, <4 x i32> %34, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %59 = extractelement <4 x float> %58, i32 1
>  %60 = fmul float %59, %52
>  %61 = fadd float %60, %53
>  %62 = fmul float %26, 6.000000e+00
>  %63 = fadd float %62, 1.000000e+00
>  %64 = fmul float %61, %63
>  %65 = fadd float %64, 0xBFEFAE1480000000
>  %66 = fmul float %65, 1.000000e+02
>  %67 = call float @llvm.AMDIL.clamp.(float %66, float 0.000000e+00, float 1.000000e+00)
>  %68 = fmul float %67, -2.000000e+00
>  %69 = fadd float %68, 3.000000e+00
>  %70 = fmul float %67, %67
>  %71 = fmul float %69, %70
>  %72 = fadd float %71, 1.000000e+00
>  %73 = bitcast float %43 to i32
>  %74 = bitcast float %44 to i32
>  %75 = insertelement <2 x i32> undef, i32 %73, i32 0
>  %76 = insertelement <2 x i32> %75, i32 %74, i32 1
>  %77 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %76, <8 x i32> %40, <4 x i32> %42, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %78 = extractelement <4 x float> %77, i32 0
>  %79 = fsub float %72, %78
>  %80 = call float @llvm.AMDIL.clamp.(float %79, float 0.000000e+00, float 1.000000e+00)
>  %81 = bitcast float %43 to i32
>  %82 = bitcast float %44 to i32
>  %83 = insertelement <2 x i32> undef, i32 %81, i32 0
>  %84 = insertelement <2 x i32> %83, i32 %82, i32 1
>  %85 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %84, <8 x i32> %36, <4 x i32> %38, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %86 = extractelement <4 x float> %85, i32 0
>  %87 = extractelement <4 x float> %85, i32 1
>  %88 = extractelement <4 x float> %85, i32 2
>  %89 = fsub float %86, %25
>  %90 = fmul float %80, %89
>  %91 = fadd float %90, %25
>  %92 = fadd float %87, -1.000000e+00
>  %93 = fadd float %88, 0.000000e+00
>  %94 = fmul float %88, 0.000000e+00
>  %95 = fadd float %94, 1.000000e+00
>  %96 = fmul float %80, %92
>  %97 = fadd float %96, 1.000000e+00
>  %98 = call i32 @llvm.SI.packf16(float %91, float %97)
>  %99 = bitcast i32 %98 to float
>  %100 = call i32 @llvm.SI.packf16(float %93, float %95)
>  %101 = bitcast i32 %100 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %99, float %101, float %99, float %101)
>  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.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #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_load_dwordx4 s[0:3], s[2:3], 0x0                                      ; C0800300
>	s_mov_b32 m0, s10                                                       ; BEFC030A
>	s_load_dwordx8 s[8:15], s[6:7], 0x0                                     ; C0C40700
>	s_load_dwordx8 s[16:23], s[6:7], 0x8                                    ; C0C80708
>	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
>	s_load_dwordx4 s[24:27], s[4:5], 0x0                                    ; C08C0500
>	s_load_dwordx4 s[28:31], s[4:5], 0x4                                    ; C08E0504
>	s_load_dwordx4 s[32:35], s[4:5], 0x8                                    ; C0900508
>	s_load_dwordx8 s[36:43], s[6:7], 0x18                                   ; C0D20718
>	s_load_dwordx4 s[44:47], s[4:5], 0xc                                    ; C096050C
>	s_load_dwordx8 s[48:55], s[6:7], 0x10                                   ; C0D80710
>	v_interp_p2_f32 v3, [v3], v1, 1, 0, [m0]                                ; C80D0101
>	v_interp_p1_f32 v4, v0, 0, 1, [m0]                                      ; C8100400
>	s_waitcnt lgkmcnt(0)                                                    ; BF8C007F
>	s_buffer_load_dword s4, s[0:3], 0xf                                     ; C202010F
>	v_interp_p2_f32 v4, [v4], v1, 0, 1, [m0]                                ; C8110401
>	v_interp_p1_f32 v5, v0, 1, 1, [m0]                                      ; C8140500
>	v_interp_p2_f32 v5, [v5], v1, 1, 1, [m0]                                ; C8150501
>	image_sample v[0:1], 9, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[8:15], s[24:27]  ; F0800900 00C20002
>	image_sample v4, 2, 0, 0, 0, 0, 0, 0, 0, v[4:5], s[16:23], s[28:31]     ; F0800200 00E40404
>	image_sample v5, 1, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[36:43], s[44:47]     ; F0800100 01690502
>	image_sample v[6:8], 7, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[48:55], s[32:35] ; F0800700 010C0602
>	s_buffer_load_dword s0, s[0:3], 0xd                                     ; C200010D
>	v_mov_b32_e32 v2, 0x40c00000                                            ; 7E0402FF 40C00000
>	s_waitcnt lgkmcnt(0)                                                    ; BF8C007F
>	v_mad_f32 v2, v2, s4, 1.0                                               ; D2820002 03C80902
>	s_waitcnt vmcnt(2)                                                      ; BF8C0772
>	v_mac_f32_e32 v1, v0, v4                                                ; 3E020900
>	v_madak_f32_e32 v0, v1, v2, 0xbf7d70a4                                  ; 42000501 BF7D70A4
>	v_mul_f32_e32 v0, 0x42c80000, v0                                        ; 100000FF 42C80000
>	v_add_f32_e64 v0, 0, v0 clamp                                           ; D2060800 00020080
>	v_madak_f32_e32 v1, -2.0, v0, 0x40400000                                ; 420200F5 40400000
>	v_mul_f32_e32 v0, v0, v0                                                ; 10000100
>	v_mad_f32 v0, v1, v0, 1.0                                               ; D2820000 03CA0101
>	s_waitcnt vmcnt(1)                                                      ; BF8C0771
>	v_subrev_f32_e32 v0, v5, v0                                             ; 0A000105
>	v_add_f32_e64 v0, 0, v0 clamp                                           ; D2060800 00020080
>	s_waitcnt vmcnt(0)                                                      ; BF8C0770
>	v_subrev_f32_e32 v1, s0, v6                                             ; 0A020C00
>	v_mad_f32 v1, v0, v1, s0                                                ; D2820001 00020300
>	v_add_f32_e32 v2, -1.0, v7                                              ; 06040EF3
>	v_add_f32_e32 v3, 0, v8                                                 ; 06061080
>	v_mad_f32 v4, 0, v8, 1.0                                                ; D2820004 03CA1080
>	v_mad_f32 v0, v0, v2, 1.0                                               ; D2820000 03CA0500
>	v_cvt_pkrtz_f16_f32_e32 v0, v1, v0                                      ; 5E000101
>	v_cvt_pkrtz_f16_f32_e32 v1, v3, v4                                      ; 5E020903
>	exp 15, 0, 1, 1, 1, v0, v1, v0, v1                                      ; F8001C0F 01000100
>	s_endpgm                                                                ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 56
>VGPRS: 12
>Code Size: 268 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
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL OUT[1], CLIPVERTEX
>DCL OUT[2], GENERIC[0]
>DCL OUT[3], GENERIC[1]
>DCL CONST[0..54]
>DCL TEMP[0..3], LOCAL
>  0: MUL TEMP[0].xy, CONST[48].xyyy, IN[1].xyyy
>  1: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
>  2: ADD TEMP[1].x, TEMP[0].xxxx, CONST[48].wwww
>  3: MOV TEMP[1].z, TEMP[1].xxxx
>  4: MUL TEMP[0].xy, CONST[49].xyyy, IN[1].xyyy
>  5: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
>  6: ADD TEMP[2].x, TEMP[0].xxxx, CONST[49].wwww
>  7: MOV TEMP[1].w, TEMP[2].xxxx
>  8: MUL TEMP[0].xy, CONST[50].xyyy, IN[1].xyyy
>  9: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 10: ADD TEMP[2].x, TEMP[0].xxxx, CONST[50].wwww
> 11: MUL TEMP[0].xy, CONST[51].xyyy, IN[1].xyyy
> 12: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 13: ADD TEMP[3].x, TEMP[0].xxxx, CONST[51].wwww
> 14: MOV TEMP[2].y, TEMP[3].xxxx
> 15: MUL TEMP[0].xy, CONST[52].xyyy, IN[1].xyyy
> 16: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 17: ADD TEMP[3].x, TEMP[0].xxxx, CONST[52].wwww
> 18: MOV TEMP[2].z, TEMP[3].xxxx
> 19: MUL TEMP[0].xy, CONST[53].xyyy, IN[1].xyyy
> 20: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 21: ADD TEMP[0].x, TEMP[0].xxxx, CONST[53].wwww
> 22: MOV TEMP[2].w, TEMP[0].xxxx
> 23: MOV TEMP[0].xw, IN[0].xxxw
> 24: MOV TEMP[1].xy, IN[1].xyxx
> 25: MAD TEMP[3].x, IN[0].zzzz, CONST[0].zzzz, -IN[0].wwww
> 26: MOV TEMP[0].z, TEMP[3].xxxx
> 27: MOV TEMP[0].y, -IN[0].yyyy
> 28: MAD TEMP[0].xy, CONST[54].xyyy, IN[0].wwww, TEMP[0].xyyy
> 29: MOV OUT[2], TEMP[1]
> 30: MOV OUT[3], TEMP[2]
> 31: MOV OUT[0], TEMP[0]
> 32: MOV OUT[1], IN[0]
> 33: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = call float @llvm.SI.load.const(<16 x i8> %13, i32 8)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %13, i32 768)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %13, i32 772)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %13, i32 780)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %13, i32 784)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %13, i32 788)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %13, i32 796)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %13, i32 800)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %13, i32 804)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %13, i32 812)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %13, i32 816)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %13, i32 820)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %13, i32 828)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %13, i32 832)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %13, i32 836)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %13, i32 844)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %13, i32 848)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %13, i32 852)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %13, i32 860)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %13, i32 864)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %13, i32 868)
>  %35 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %36 = load <16 x i8>, <16 x i8> addrspace(2)* %35, align 16, !tbaa !0
>  %37 = add i32 %5, %8
>  %38 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %36, i32 0, i32 %37)
>  %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 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 1
>  %44 = load <16 x i8>, <16 x i8> addrspace(2)* %43, align 16, !tbaa !0
>  %45 = add i32 %5, %8
>  %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 = fmul float %15, %47
>  %50 = fmul float %16, %48
>  %51 = fadd float %50, %49
>  %52 = fadd float %51, %17
>  %53 = fmul float %18, %47
>  %54 = fmul float %19, %48
>  %55 = fadd float %54, %53
>  %56 = fadd float %55, %20
>  %57 = fmul float %21, %47
>  %58 = fmul float %22, %48
>  %59 = fadd float %58, %57
>  %60 = fadd float %59, %23
>  %61 = fmul float %24, %47
>  %62 = fmul float %25, %48
>  %63 = fadd float %62, %61
>  %64 = fadd float %63, %26
>  %65 = fmul float %27, %47
>  %66 = fmul float %28, %48
>  %67 = fadd float %66, %65
>  %68 = fadd float %67, %29
>  %69 = fmul float %30, %47
>  %70 = fmul float %31, %48
>  %71 = fadd float %70, %69
>  %72 = fadd float %71, %32
>  %73 = fmul float %41, %14
>  %74 = fsub float %73, %42
>  %75 = fmul float %33, %42
>  %76 = fadd float %75, %39
>  %77 = fmul float %34, %42
>  %78 = fsub float %77, %40
>  %79 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 16
>  %80 = load <16 x i8>, <16 x i8> addrspace(2)* %79, align 16, !tbaa !0
>  %81 = call float @llvm.SI.load.const(<16 x i8> %80, i32 0)
>  %82 = fmul float %81, %39
>  %83 = call float @llvm.SI.load.const(<16 x i8> %80, i32 4)
>  %84 = fmul float %83, %40
>  %85 = fadd float %82, %84
>  %86 = call float @llvm.SI.load.const(<16 x i8> %80, i32 8)
>  %87 = fmul float %86, %41
>  %88 = fadd float %85, %87
>  %89 = call float @llvm.SI.load.const(<16 x i8> %80, i32 12)
>  %90 = fmul float %89, %42
>  %91 = fadd float %88, %90
>  %92 = call float @llvm.SI.load.const(<16 x i8> %80, i32 16)
>  %93 = fmul float %92, %39
>  %94 = call float @llvm.SI.load.const(<16 x i8> %80, i32 20)
>  %95 = fmul float %94, %40
>  %96 = fadd float %93, %95
>  %97 = call float @llvm.SI.load.const(<16 x i8> %80, i32 24)
>  %98 = fmul float %97, %41
>  %99 = fadd float %96, %98
>  %100 = call float @llvm.SI.load.const(<16 x i8> %80, i32 28)
>  %101 = fmul float %100, %42
>  %102 = fadd float %99, %101
>  %103 = call float @llvm.SI.load.const(<16 x i8> %80, i32 32)
>  %104 = fmul float %103, %39
>  %105 = call float @llvm.SI.load.const(<16 x i8> %80, i32 36)
>  %106 = fmul float %105, %40
>  %107 = fadd float %104, %106
>  %108 = call float @llvm.SI.load.const(<16 x i8> %80, i32 40)
>  %109 = fmul float %108, %41
>  %110 = fadd float %107, %109
>  %111 = call float @llvm.SI.load.const(<16 x i8> %80, i32 44)
>  %112 = fmul float %111, %42
>  %113 = fadd float %110, %112
>  %114 = call float @llvm.SI.load.const(<16 x i8> %80, i32 48)
>  %115 = fmul float %114, %39
>  %116 = call float @llvm.SI.load.const(<16 x i8> %80, i32 52)
>  %117 = fmul float %116, %40
>  %118 = fadd float %115, %117
>  %119 = call float @llvm.SI.load.const(<16 x i8> %80, i32 56)
>  %120 = fmul float %119, %41
>  %121 = fadd float %118, %120
>  %122 = call float @llvm.SI.load.const(<16 x i8> %80, i32 60)
>  %123 = fmul float %122, %42
>  %124 = fadd float %121, %123
>  %125 = call float @llvm.SI.load.const(<16 x i8> %80, i32 64)
>  %126 = fmul float %125, %39
>  %127 = call float @llvm.SI.load.const(<16 x i8> %80, i32 68)
>  %128 = fmul float %127, %40
>  %129 = fadd float %126, %128
>  %130 = call float @llvm.SI.load.const(<16 x i8> %80, i32 72)
>  %131 = fmul float %130, %41
>  %132 = fadd float %129, %131
>  %133 = call float @llvm.SI.load.const(<16 x i8> %80, i32 76)
>  %134 = fmul float %133, %42
>  %135 = fadd float %132, %134
>  %136 = call float @llvm.SI.load.const(<16 x i8> %80, i32 80)
>  %137 = fmul float %136, %39
>  %138 = call float @llvm.SI.load.const(<16 x i8> %80, i32 84)
>  %139 = fmul float %138, %40
>  %140 = fadd float %137, %139
>  %141 = call float @llvm.SI.load.const(<16 x i8> %80, i32 88)
>  %142 = fmul float %141, %41
>  %143 = fadd float %140, %142
>  %144 = call float @llvm.SI.load.const(<16 x i8> %80, i32 92)
>  %145 = fmul float %144, %42
>  %146 = fadd float %143, %145
>  %147 = call float @llvm.SI.load.const(<16 x i8> %80, i32 96)
>  %148 = fmul float %147, %39
>  %149 = call float @llvm.SI.load.const(<16 x i8> %80, i32 100)
>  %150 = fmul float %149, %40
>  %151 = fadd float %148, %150
>  %152 = call float @llvm.SI.load.const(<16 x i8> %80, i32 104)
>  %153 = fmul float %152, %41
>  %154 = fadd float %151, %153
>  %155 = call float @llvm.SI.load.const(<16 x i8> %80, i32 108)
>  %156 = fmul float %155, %42
>  %157 = fadd float %154, %156
>  %158 = call float @llvm.SI.load.const(<16 x i8> %80, i32 112)
>  %159 = fmul float %158, %39
>  %160 = call float @llvm.SI.load.const(<16 x i8> %80, i32 116)
>  %161 = fmul float %160, %40
>  %162 = fadd float %159, %161
>  %163 = call float @llvm.SI.load.const(<16 x i8> %80, i32 120)
>  %164 = fmul float %163, %41
>  %165 = fadd float %162, %164
>  %166 = call float @llvm.SI.load.const(<16 x i8> %80, i32 124)
>  %167 = fmul float %166, %42
>  %168 = fadd float %165, %167
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %47, float %48, float %52, float %56)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %60, float %64, float %68, float %72)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 12, i32 0, float %76, float %78, float %74, float %42)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 13, i32 0, float %91, float %102, float %113, float %124)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 14, i32 0, float %135, float %146, float %157, 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
>
>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[4:7], s[8:9], 0x0                    ; C0820900
>	s_load_dwordx4 s[12:15], s[8:9], 0x4                  ; C0860904
>	v_add_i32_e32 v0, vcc, s10, v0                        ; 4A00000A
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[4:7], 0 idxen   ; E00C2000 80010100
>	buffer_load_format_xyzw v[5:8], v0, s[12:15], 0 idxen ; E00C2000 80030500
>	s_load_dwordx4 s[4:7], s[2:3], 0x0                    ; C0820300
>	s_load_dwordx4 s[0:3], s[2:3], 0x40                   ; C0800340
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	s_buffer_load_dword s8, s[4:7], 0x2                   ; C2040502
>	s_buffer_load_dword s9, s[4:7], 0xc0                  ; C20485C0
>	s_buffer_load_dword s10, s[4:7], 0xc1                 ; C20505C1
>	s_buffer_load_dword s11, s[4:7], 0xc3                 ; C20585C3
>	s_buffer_load_dword s12, s[4:7], 0xc4                 ; C20605C4
>	s_buffer_load_dword s13, s[4:7], 0xc5                 ; C20685C5
>	s_buffer_load_dword s14, s[4:7], 0xc7                 ; C20705C7
>	s_buffer_load_dword s15, s[4:7], 0xc8                 ; C20785C8
>	s_buffer_load_dword s16, s[4:7], 0xc9                 ; C20805C9
>	s_buffer_load_dword s17, s[4:7], 0xcb                 ; C20885CB
>	s_buffer_load_dword s18, s[4:7], 0xcc                 ; C20905CC
>	s_buffer_load_dword s19, s[4:7], 0xcd                 ; C20985CD
>	s_buffer_load_dword s20, s[4:7], 0xcf                 ; C20A05CF
>	s_buffer_load_dword s21, s[4:7], 0xd0                 ; C20A85D0
>	s_buffer_load_dword s22, s[4:7], 0xd1                 ; C20B05D1
>	s_buffer_load_dword s23, s[4:7], 0xd3                 ; C20B85D3
>	s_buffer_load_dword s24, s[4:7], 0xd4                 ; C20C05D4
>	s_buffer_load_dword s25, s[4:7], 0xd5                 ; C20C85D5
>	s_buffer_load_dword s26, s[4:7], 0xd7                 ; C20D05D7
>	s_buffer_load_dword s27, s[4:7], 0xd8                 ; C20D85D8
>	s_buffer_load_dword s4, s[4:7], 0xd9                  ; C20205D9
>	s_buffer_load_dword s5, s[0:3], 0x0                   ; C2028100
>	s_buffer_load_dword s6, s[0:3], 0x1                   ; C2030101
>	s_buffer_load_dword s7, s[0:3], 0x2                   ; C2038102
>	s_buffer_load_dword s28, s[0:3], 0x3                  ; C20E0103
>	s_buffer_load_dword s29, s[0:3], 0x4                  ; C20E8104
>	s_buffer_load_dword s30, s[0:3], 0x5                  ; C20F0105
>	s_buffer_load_dword s31, s[0:3], 0x6                  ; C20F8106
>	s_buffer_load_dword s32, s[0:3], 0x7                  ; C2100107
>	s_buffer_load_dword s33, s[0:3], 0x8                  ; C2108108
>	s_buffer_load_dword s34, s[0:3], 0x9                  ; C2110109
>	s_buffer_load_dword s35, s[0:3], 0xa                  ; C211810A
>	s_buffer_load_dword s36, s[0:3], 0xb                  ; C212010B
>	s_buffer_load_dword s37, s[0:3], 0xc                  ; C212810C
>	s_buffer_load_dword s38, s[0:3], 0xd                  ; C213010D
>	s_buffer_load_dword s39, s[0:3], 0xe                  ; C213810E
>	s_buffer_load_dword s40, s[0:3], 0xf                  ; C214010F
>	s_buffer_load_dword s41, s[0:3], 0x10                 ; C2148110
>	s_buffer_load_dword s42, s[0:3], 0x11                 ; C2150111
>	s_buffer_load_dword s43, s[0:3], 0x12                 ; C2158112
>	s_buffer_load_dword s44, s[0:3], 0x13                 ; C2160113
>	s_buffer_load_dword s45, s[0:3], 0x14                 ; C2168114
>	s_buffer_load_dword s46, s[0:3], 0x15                 ; C2170115
>	s_buffer_load_dword s47, s[0:3], 0x16                 ; C2178116
>	s_buffer_load_dword s48, s[0:3], 0x17                 ; C2180117
>	s_buffer_load_dword s49, s[0:3], 0x18                 ; C2188118
>	s_buffer_load_dword s50, s[0:3], 0x19                 ; C2190119
>	s_buffer_load_dword s51, s[0:3], 0x1a                 ; C219811A
>	s_buffer_load_dword s52, s[0:3], 0x1b                 ; C21A011B
>	s_buffer_load_dword s53, s[0:3], 0x1c                 ; C21A811C
>	s_buffer_load_dword s54, s[0:3], 0x1d                 ; C21B011D
>	s_buffer_load_dword s55, s[0:3], 0x1e                 ; C21B811E
>	s_buffer_load_dword s0, s[0:3], 0x1f                  ; C200011F
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mov_b32_e32 v0, s11                                 ; 7E00020B
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mov_b32_e32 v7, s14                                 ; 7E0E020E
>	v_mov_b32_e32 v8, s17                                 ; 7E100211
>	v_mov_b32_e32 v9, s20                                 ; 7E120214
>	v_mov_b32_e32 v10, s23                                ; 7E140217
>	v_mov_b32_e32 v11, s26                                ; 7E16021A
>	v_mac_f32_e32 v0, s9, v5                              ; 3E000A09
>	v_mac_f32_e32 v7, s12, v5                             ; 3E0E0A0C
>	v_mac_f32_e32 v8, s15, v5                             ; 3E100A0F
>	v_mac_f32_e32 v9, s18, v5                             ; 3E120A12
>	v_mac_f32_e32 v10, s21, v5                            ; 3E140A15
>	v_mac_f32_e32 v11, s24, v5                            ; 3E160A18
>	v_mac_f32_e32 v0, s10, v6                             ; 3E000C0A
>	v_mac_f32_e32 v7, s13, v6                             ; 3E0E0C0D
>	v_mac_f32_e32 v8, s16, v6                             ; 3E100C10
>	v_mac_f32_e32 v9, s19, v6                             ; 3E120C13
>	exp 15, 32, 0, 0, 0, v5, v6, v0, v7                   ; F800020F 07000605
>	v_mac_f32_e32 v10, s22, v6                            ; 3E140C16
>	v_mac_f32_e32 v11, s25, v6                            ; 3E160C19
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mad_f32 v0, v3, s8, -v4                             ; D2820000 84101103
>	v_mad_f32 v5, s27, v4, v1                             ; D2820005 0406081B
>	v_mad_f32 v6, s4, v4, -v2                             ; D2820006 840A0804
>	v_mul_f32_e32 v7, s6, v2                              ; 100E0406
>	exp 15, 33, 0, 0, 0, v8, v9, v10, v11                 ; F800021F 0B0A0908
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mul_f32_e32 v8, s30, v2                             ; 1010041E
>	v_mac_f32_e32 v7, s5, v1                              ; 3E0E0205
>	exp 15, 12, 0, 0, 0, v5, v6, v0, v4                   ; F80000CF 04000605
>	v_mac_f32_e32 v8, s29, v1                             ; 3E10021D
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mul_f32_e32 v0, s34, v2                             ; 10000422
>	v_mac_f32_e32 v0, s33, v1                             ; 3E000221
>	v_mul_f32_e32 v5, s38, v2                             ; 100A0426
>	v_mac_f32_e32 v5, s37, v1                             ; 3E0A0225
>	v_mul_f32_e32 v6, s42, v2                             ; 100C042A
>	v_mul_f32_e32 v9, s46, v2                             ; 1012042E
>	v_mul_f32_e32 v10, s50, v2                            ; 10140432
>	v_mul_f32_e32 v2, s54, v2                             ; 10040436
>	v_mac_f32_e32 v6, s41, v1                             ; 3E0C0229
>	v_mac_f32_e32 v9, s45, v1                             ; 3E12022D
>	v_mac_f32_e32 v10, s49, v1                            ; 3E140231
>	v_mac_f32_e32 v2, s53, v1                             ; 3E040235
>	v_mac_f32_e32 v7, s7, v3                              ; 3E0E0607
>	v_mac_f32_e32 v8, s31, v3                             ; 3E10061F
>	v_mac_f32_e32 v0, s35, v3                             ; 3E000623
>	v_mac_f32_e32 v5, s39, v3                             ; 3E0A0627
>	v_mac_f32_e32 v6, s43, v3                             ; 3E0C062B
>	v_mac_f32_e32 v9, s47, v3                             ; 3E12062F
>	v_mac_f32_e32 v10, s51, v3                            ; 3E140633
>	v_mac_f32_e32 v2, s55, v3                             ; 3E040637
>	v_mac_f32_e32 v7, s28, v4                             ; 3E0E081C
>	v_mac_f32_e32 v8, s32, v4                             ; 3E100820
>	v_mac_f32_e32 v0, s36, v4                             ; 3E000824
>	v_mac_f32_e32 v5, s40, v4                             ; 3E0A0828
>	v_mac_f32_e32 v6, s44, v4                             ; 3E0C082C
>	v_mac_f32_e32 v9, s48, v4                             ; 3E120830
>	v_mac_f32_e32 v10, s52, v4                            ; 3E140834
>	v_mac_f32_e32 v2, s0, v4                              ; 3E040800
>	exp 15, 13, 0, 0, 0, v7, v8, v0, v5                   ; F80000DF 05000807
>	exp 15, 14, 0, 1, 0, v6, v9, v10, v2                  ; F80008EF 020A0906
>	s_endpgm                                              ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 64
>VGPRS: 12
>Code Size: 544 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
>  clamp_color = 0
>FRAG
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL IN[1], GENERIC[1], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SAMP[1]
>DCL SAMP[2]
>DCL SAMP[3]
>DCL SAMP[4]
>DCL SVIEW[0], 2D, FLOAT
>DCL SVIEW[1], 2D, FLOAT
>DCL SVIEW[2], 2D, FLOAT
>DCL SVIEW[3], 2D, FLOAT
>DCL SVIEW[4], 2D, FLOAT
>DCL CONST[0..3]
>DCL TEMP[0..6], LOCAL
>IMM[0] FLT32 {    6.0000,     1.0000,     0.7500,     0.2500}
>IMM[1] FLT32 {   -1.0000,    -0.5800,    10.0000,    -2.0000}
>IMM[2] FLT32 {    3.0000,   100.0000,     0.0500,    -0.0800}
>IMM[3] FLT32 {    0.3000,     0.5900,     0.1100,    14.2857}
>IMM[4] FLT32 {    0.0050,     2.0000,     0.0000,     0.0000}
>  0: MOV TEMP[0].xy, IN[1].zwww
>  1: TEX TEMP[0], TEMP[0], SAMP[4], 2D
>  2: MUL TEMP[1].x, IMM[0].zzzz, CONST[3].wwww
>  3: MOV TEMP[2].xy, IN[0].xyyy
>  4: TEX TEMP[2], TEMP[2], SAMP[0], 2D
>  5: ADD TEMP[3].x, -TEMP[2].xxxx, IMM[0].yyyy
>  6: MUL TEMP[3].x, TEMP[3].xxxx, TEMP[3].xxxx
>  7: MUL TEMP[3].x, TEMP[3].xxxx, TEMP[3].xxxx
>  8: MAD TEMP[1].x, TEMP[3].xxxx, IMM[0].wwww, TEMP[1].xxxx
>  9: ADD TEMP[3], TEMP[0], IMM[1].xxxx
> 10: MAD TEMP[1], TEMP[1].xxxx, TEMP[3], IMM[0].yyyy
> 11: MAD TEMP[0].x, CONST[3].wwww, IMM[0].xxxx, IMM[0].yyyy
> 12: MOV TEMP[3].xy, IN[1].xyyy
> 13: TEX TEMP[3].y, TEMP[3], SAMP[1], 2D
> 14: MAD TEMP[4].x, TEMP[3].yyyy, TEMP[2].xxxx, TEMP[2].wwww
> 15: MAD TEMP[0].x, TEMP[4].xxxx, TEMP[0].xxxx, IMM[1].yyyy
> 16: MUL TEMP[0].x, TEMP[0].xxxx, IMM[1].zzzz
> 17: MOV_SAT TEMP[4].x, TEMP[0].xxxx
> 18: MAD TEMP[5].x, TEMP[4].xxxx, IMM[1].wwww, IMM[2].xxxx
> 19: MUL TEMP[0].x, TEMP[4].xxxx, TEMP[4].xxxx
> 20: MUL TEMP[4].x, TEMP[0].xxxx, TEMP[5].xxxx
> 21: MAD TEMP[0].x, TEMP[5].xxxx, TEMP[0].xxxx, IMM[0].yyyy
> 22: MUL TEMP[4].x, TEMP[4].xxxx, IMM[2].yyyy
> 23: MOV_SAT TEMP[4].x, TEMP[4].xxxx
> 24: MAD TEMP[5].x, TEMP[4].xxxx, IMM[1].wwww, IMM[2].xxxx
> 25: MUL TEMP[4].x, TEMP[4].xxxx, TEMP[4].xxxx
> 26: MUL TEMP[4].x, TEMP[4].xxxx, TEMP[5].xxxx
> 27: LRP TEMP[5].xyz, TEMP[4].xxxx, IMM[0].yyyy, TEMP[1].xyzz
> 28: LRP TEMP[3].xyz, TEMP[4].xxxx, IMM[2].zzzz, CONST[0].xyzz
> 29: MUL TEMP[3].xyz, TEMP[5].xyzz, TEMP[3].xyzz
> 30: DP3 TEMP[5].x, TEMP[3].xyzz, IMM[3].xyzz
> 31: ADD TEMP[5].x, TEMP[5].xxxx, IMM[2].wwww
> 32: MUL TEMP[5].x, TEMP[5].xxxx, IMM[3].wwww
> 33: MOV_SAT TEMP[5].x, TEMP[5].xxxx
> 34: MAD TEMP[6].x, TEMP[5].xxxx, IMM[1].wwww, IMM[2].xxxx
> 35: MUL TEMP[5].x, TEMP[5].xxxx, TEMP[5].xxxx
> 36: MAD TEMP[5].x, TEMP[6].xxxx, -TEMP[5].xxxx, IMM[0].yyyy
> 37: MAD TEMP[1].xyz, TEMP[5].xxxx, IMM[4].xxxx, TEMP[3].xyzz
> 38: MUL TEMP[5].xyz, TEMP[2].zzzz, TEMP[1].xyzz
> 39: MAD TEMP[3].xyz, TEMP[5].xyzz, IMM[4].yyyy, TEMP[3].xyzz
> 40: MOV_SAT TEMP[3].xyz, TEMP[3].xyzz
> 41: MUL TEMP[5].xyz, TEMP[2].yyyy, TEMP[3].xyzz
> 42: MOV TEMP[6].xy, IN[0].xyyy
> 43: TEX TEMP[6], TEMP[6], SAMP[2], 2D
> 44: MAD TEMP[3].xyz, TEMP[3].xyzz, -TEMP[2].yyyy, TEMP[6].xyzz
> 45: MUL TEMP[2].x, TEMP[2].yyyy, CONST[3].zzzz
> 46: MUL TEMP[1].x, TEMP[1].wwww, TEMP[2].xxxx
> 47: MOV TEMP[2].xy, IN[0].xyyy
> 48: TEX TEMP[2].x, TEMP[2], SAMP[3], 2D
> 49: ADD TEMP[0].x, TEMP[0].xxxx, -TEMP[2].xxxx
> 50: MOV_SAT TEMP[2].x, TEMP[0].xxxx
> 51: MAD TEMP[3].xyz, TEMP[2].xxxx, TEMP[3].xyzz, TEMP[5].xyzz
> 52: LRP TEMP[0].x, TEMP[4].xxxx, IMM[0].yyyy, CONST[3].zzzz
> 53: MUL TEMP[0].x, TEMP[1].xxxx, TEMP[0].xxxx
> 54: MAD TEMP[1].x, TEMP[0].xxxx, -CONST[3].xxxx, TEMP[6].wwww
> 55: MUL TEMP[0].x, TEMP[0].xxxx, CONST[3].xxxx
> 56: MAD TEMP[0].x, TEMP[2].xxxx, TEMP[1].xxxx, TEMP[0].xxxx
> 57: MOV TEMP[3].w, TEMP[0].xxxx
> 58: MOV OUT[0], TEMP[3]
> 59: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %24 = load <16 x i8>, <16 x i8> addrspace(2)* %23, align 16, !tbaa !0
>  %25 = call float @llvm.SI.load.const(<16 x i8> %24, i32 0)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %24, i32 4)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %24, i32 8)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %24, i32 48)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %24, i32 56)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %24, i32 60)
>  %31 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %32 = load <8 x i32>, <8 x i32> addrspace(2)* %31, align 32, !tbaa !0
>  %33 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %34 = load <4 x i32>, <4 x i32> addrspace(2)* %33, align 16, !tbaa !0
>  %35 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 1
>  %36 = load <8 x i32>, <8 x i32> addrspace(2)* %35, align 32, !tbaa !0
>  %37 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 1
>  %38 = load <4 x i32>, <4 x i32> addrspace(2)* %37, align 16, !tbaa !0
>  %39 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 2
>  %40 = load <8 x i32>, <8 x i32> addrspace(2)* %39, align 32, !tbaa !0
>  %41 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 2
>  %42 = load <4 x i32>, <4 x i32> addrspace(2)* %41, align 16, !tbaa !0
>  %43 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 3
>  %44 = load <8 x i32>, <8 x i32> addrspace(2)* %43, align 32, !tbaa !0
>  %45 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 3
>  %46 = load <4 x i32>, <4 x i32> addrspace(2)* %45, align 16, !tbaa !0
>  %47 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 4
>  %48 = load <8 x i32>, <8 x i32> addrspace(2)* %47, align 32, !tbaa !0
>  %49 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 4
>  %50 = load <4 x i32>, <4 x i32> addrspace(2)* %49, align 16, !tbaa !0
>  %51 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %8)
>  %52 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %8)
>  %53 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %6, <2 x i32> %8)
>  %54 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %6, <2 x i32> %8)
>  %55 = call float @llvm.SI.fs.interp(i32 2, i32 1, i32 %6, <2 x i32> %8)
>  %56 = call float @llvm.SI.fs.interp(i32 3, i32 1, i32 %6, <2 x i32> %8)
>  %57 = bitcast float %55 to i32
>  %58 = bitcast float %56 to i32
>  %59 = insertelement <2 x i32> undef, i32 %57, i32 0
>  %60 = insertelement <2 x i32> %59, i32 %58, i32 1
>  %61 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %60, <8 x i32> %48, <4 x i32> %50, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %62 = extractelement <4 x float> %61, i32 0
>  %63 = extractelement <4 x float> %61, i32 1
>  %64 = extractelement <4 x float> %61, i32 2
>  %65 = extractelement <4 x float> %61, i32 3
>  %66 = fmul float %30, 7.500000e-01
>  %67 = bitcast float %51 to i32
>  %68 = bitcast float %52 to i32
>  %69 = insertelement <2 x i32> undef, i32 %67, i32 0
>  %70 = insertelement <2 x i32> %69, i32 %68, i32 1
>  %71 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %70, <8 x i32> %32, <4 x i32> %34, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %72 = extractelement <4 x float> %71, i32 0
>  %73 = extractelement <4 x float> %71, i32 1
>  %74 = extractelement <4 x float> %71, i32 2
>  %75 = extractelement <4 x float> %71, i32 3
>  %76 = fsub float 1.000000e+00, %72
>  %77 = fmul float %76, %76
>  %78 = fmul float %77, %77
>  %79 = fmul float %78, 2.500000e-01
>  %80 = fadd float %79, %66
>  %81 = fadd float %62, -1.000000e+00
>  %82 = fadd float %63, -1.000000e+00
>  %83 = fadd float %64, -1.000000e+00
>  %84 = fadd float %65, -1.000000e+00
>  %85 = fmul float %80, %81
>  %86 = fadd float %85, 1.000000e+00
>  %87 = fmul float %80, %82
>  %88 = fadd float %87, 1.000000e+00
>  %89 = fmul float %80, %83
>  %90 = fadd float %89, 1.000000e+00
>  %91 = fmul float %80, %84
>  %92 = fadd float %91, 1.000000e+00
>  %93 = fmul float %30, 6.000000e+00
>  %94 = fadd float %93, 1.000000e+00
>  %95 = bitcast float %53 to i32
>  %96 = bitcast float %54 to i32
>  %97 = insertelement <2 x i32> undef, i32 %95, i32 0
>  %98 = insertelement <2 x i32> %97, i32 %96, i32 1
>  %99 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %98, <8 x i32> %36, <4 x i32> %38, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %100 = extractelement <4 x float> %99, i32 1
>  %101 = fmul float %100, %72
>  %102 = fadd float %101, %75
>  %103 = fmul float %102, %94
>  %104 = fadd float %103, 0xBFE28F5C20000000
>  %105 = fmul float %104, 1.000000e+01
>  %106 = call float @llvm.AMDIL.clamp.(float %105, float 0.000000e+00, float 1.000000e+00)
>  %107 = fmul float %106, -2.000000e+00
>  %108 = fadd float %107, 3.000000e+00
>  %109 = fmul float %106, %106
>  %110 = fmul float %109, %108
>  %111 = fmul float %108, %109
>  %112 = fadd float %111, 1.000000e+00
>  %113 = fmul float %110, 1.000000e+02
>  %114 = call float @llvm.AMDIL.clamp.(float %113, float 0.000000e+00, float 1.000000e+00)
>  %115 = fmul float %114, -2.000000e+00
>  %116 = fadd float %115, 3.000000e+00
>  %117 = fmul float %114, %114
>  %118 = fmul float %117, %116
>  %119 = fsub float 1.000000e+00, %118
>  %120 = fmul float %86, %119
>  %121 = fadd float %118, %120
>  %122 = fsub float 1.000000e+00, %118
>  %123 = fmul float %88, %122
>  %124 = fadd float %118, %123
>  %125 = fsub float 1.000000e+00, %118
>  %126 = fmul float %90, %125
>  %127 = fadd float %118, %126
>  %128 = fsub float 1.000000e+00, %118
>  %129 = fmul float %118, 0x3FA99999A0000000
>  %130 = fmul float %25, %128
>  %131 = fadd float %129, %130
>  %132 = fsub float 1.000000e+00, %118
>  %133 = fmul float %118, 0x3FA99999A0000000
>  %134 = fmul float %26, %132
>  %135 = fadd float %133, %134
>  %136 = fsub float 1.000000e+00, %118
>  %137 = fmul float %118, 0x3FA99999A0000000
>  %138 = fmul float %27, %136
>  %139 = fadd float %137, %138
>  %140 = fmul float %121, %131
>  %141 = fmul float %124, %135
>  %142 = fmul float %127, %139
>  %143 = fmul float %140, 0x3FD3333340000000
>  %144 = fmul float %141, 0x3FE2E147A0000000
>  %145 = fadd float %144, %143
>  %146 = fmul float %142, 0x3FBC28F5C0000000
>  %147 = fadd float %145, %146
>  %148 = fadd float %147, 0xBFB47AE140000000
>  %149 = fmul float %148, 0x402C924920000000
>  %150 = call float @llvm.AMDIL.clamp.(float %149, float 0.000000e+00, float 1.000000e+00)
>  %151 = fmul float %150, -2.000000e+00
>  %152 = fadd float %151, 3.000000e+00
>  %153 = fmul float %150, %150
>  %154 = fmul float %153, %152
>  %155 = fsub float 1.000000e+00, %154
>  %156 = fmul float %155, 0x3F747AE140000000
>  %157 = fadd float %156, %140
>  %158 = fmul float %155, 0x3F747AE140000000
>  %159 = fadd float %158, %141
>  %160 = fmul float %155, 0x3F747AE140000000
>  %161 = fadd float %160, %142
>  %162 = fmul float %74, %157
>  %163 = fmul float %74, %159
>  %164 = fmul float %74, %161
>  %165 = fmul float %162, 2.000000e+00
>  %166 = fadd float %165, %140
>  %167 = fmul float %163, 2.000000e+00
>  %168 = fadd float %167, %141
>  %169 = fmul float %164, 2.000000e+00
>  %170 = fadd float %169, %142
>  %171 = call float @llvm.AMDIL.clamp.(float %166, float 0.000000e+00, float 1.000000e+00)
>  %172 = call float @llvm.AMDIL.clamp.(float %168, float 0.000000e+00, float 1.000000e+00)
>  %173 = call float @llvm.AMDIL.clamp.(float %170, float 0.000000e+00, float 1.000000e+00)
>  %174 = fmul float %73, %171
>  %175 = fmul float %73, %172
>  %176 = fmul float %73, %173
>  %177 = bitcast float %51 to i32
>  %178 = bitcast float %52 to i32
>  %179 = insertelement <2 x i32> undef, i32 %177, i32 0
>  %180 = insertelement <2 x i32> %179, i32 %178, i32 1
>  %181 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %180, <8 x i32> %40, <4 x i32> %42, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %182 = extractelement <4 x float> %181, i32 0
>  %183 = extractelement <4 x float> %181, i32 1
>  %184 = extractelement <4 x float> %181, i32 2
>  %185 = extractelement <4 x float> %181, i32 3
>  %186 = fmul float %73, %171
>  %187 = fsub float %182, %186
>  %188 = fmul float %73, %172
>  %189 = fsub float %183, %188
>  %190 = fmul float %73, %173
>  %191 = fsub float %184, %190
>  %192 = fmul float %73, %29
>  %193 = fmul float %92, %192
>  %194 = bitcast float %51 to i32
>  %195 = bitcast float %52 to i32
>  %196 = insertelement <2 x i32> undef, i32 %194, i32 0
>  %197 = insertelement <2 x i32> %196, i32 %195, i32 1
>  %198 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %197, <8 x i32> %44, <4 x i32> %46, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %199 = extractelement <4 x float> %198, i32 0
>  %200 = fsub float %112, %199
>  %201 = call float @llvm.AMDIL.clamp.(float %200, float 0.000000e+00, float 1.000000e+00)
>  %202 = fmul float %201, %187
>  %203 = fadd float %202, %174
>  %204 = fmul float %201, %189
>  %205 = fadd float %204, %175
>  %206 = fmul float %201, %191
>  %207 = fadd float %206, %176
>  %208 = fsub float 1.000000e+00, %118
>  %209 = fmul float %29, %208
>  %210 = fadd float %118, %209
>  %211 = fmul float %193, %210
>  %212 = fmul float %28, %211
>  %213 = fsub float %185, %212
>  %214 = fmul float %211, %28
>  %215 = fmul float %201, %213
>  %216 = fadd float %215, %214
>  %217 = call i32 @llvm.SI.packf16(float %203, float %205)
>  %218 = bitcast i32 %217 to float
>  %219 = call i32 @llvm.SI.packf16(float %207, float %216)
>  %220 = bitcast i32 %219 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %218, float %220, float %218, float %220)
>  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.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #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, s10                                                            ; BEFC030A
>	v_mov_b32_e32 v2, 0x3f400000                                                 ; 7E0402FF 3F400000
>	v_mov_b32_e32 v3, 0x40c00000                                                 ; 7E0602FF 40C00000
>	v_mov_b32_e32 v4, 0x40400000                                                 ; 7E0802FF 40400000
>	v_mov_b32_e32 v5, 0xbda3d70a                                                 ; 7E0A02FF BDA3D70A
>	v_mov_b32_e32 v6, 0x3ba3d70a                                                 ; 7E0C02FF 3BA3D70A
>	v_interp_p1_f32 v7, v0, 0, 0, [m0]                                           ; C81C0000
>	v_interp_p2_f32 v7, [v7], v1, 0, 0, [m0]                                     ; C81D0001
>	v_interp_p1_f32 v8, v0, 1, 0, [m0]                                           ; C8200100
>	v_interp_p2_f32 v8, [v8], v1, 1, 0, [m0]                                     ; C8210101
>	v_interp_p1_f32 v9, v0, 0, 1, [m0]                                           ; C8240400
>	v_interp_p2_f32 v9, [v9], v1, 0, 1, [m0]                                     ; C8250401
>	v_interp_p1_f32 v10, v0, 1, 1, [m0]                                          ; C8280500
>	v_interp_p2_f32 v10, [v10], v1, 1, 1, [m0]                                   ; C8290501
>	v_interp_p1_f32 v11, v0, 2, 1, [m0]                                          ; C82C0600
>	v_interp_p2_f32 v11, [v11], v1, 2, 1, [m0]                                   ; C82D0601
>	v_interp_p1_f32 v12, v0, 3, 1, [m0]                                          ; C8300700
>	v_interp_p2_f32 v12, [v12], v1, 3, 1, [m0]                                   ; C8310701
>	s_load_dwordx4 s[28:31], s[2:3], 0x0                                         ; C08E0300
>	s_load_dwordx4 s[32:35], s[4:5], 0x0                                         ; C0900500
>	s_load_dwordx4 s[36:39], s[4:5], 0x4                                         ; C0920504
>	s_load_dwordx4 s[16:19], s[4:5], 0x8                                         ; C0880508
>	s_load_dwordx4 s[0:3], s[4:5], 0xc                                           ; C080050C
>	s_load_dwordx4 s[40:43], s[4:5], 0x10                                        ; C0940510
>	s_load_dwordx8 s[44:51], s[6:7], 0x0                                         ; C0D60700
>	s_load_dwordx8 s[52:59], s[6:7], 0x8                                         ; C0DA0708
>	s_load_dwordx8 s[20:27], s[6:7], 0x10                                        ; C0CA0710
>	s_load_dwordx8 s[8:15], s[6:7], 0x18                                         ; C0C40718
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	s_buffer_load_dword s4, s[28:31], 0x0                                        ; C2021D00
>	s_buffer_load_dword s5, s[28:31], 0x1                                        ; C2029D01
>	s_buffer_load_dword s60, s[28:31], 0x2                                       ; C21E1D02
>	s_buffer_load_dword s61, s[28:31], 0xc                                       ; C21E9D0C
>	s_buffer_load_dword s62, s[28:31], 0xe                                       ; C21F1D0E
>	s_buffer_load_dword s28, s[28:31], 0xf                                       ; C20E1D0F
>	s_load_dwordx8 s[64:71], s[6:7], 0x20                                        ; C0E00720
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v0, s28, v2                                                    ; 1000041C
>	image_sample v[11:14], 15, 0, 0, 0, 0, 0, 0, 0, v[11:12], s[64:71], s[40:43] ; F0800F00 01500B0B
>	image_sample v[15:18], 15, 0, 0, 0, 0, 0, 0, 0, v[7:8], s[44:51], s[32:35]   ; F0800F00 010B0F07
>	image_sample v1, 2, 0, 0, 0, 0, 0, 0, 0, v[9:10], s[52:59], s[36:39]         ; F0800200 012D0109
>	s_waitcnt vmcnt(1)                                                           ; BF8C0771
>	v_sub_f32_e32 v2, 1.0, v15                                                   ; 08041EF2
>	v_mad_f32 v2, -v15, v2, v2                                                   ; D2820002 240A050F
>	v_mul_f32_e32 v2, v2, v2                                                     ; 10040502
>	v_madmk_f32_e32 v0, v2, v0, 0x3e800000                                       ; 40000102 3E800000
>	v_add_f32_e32 v2, -1.0, v11                                                  ; 060416F3
>	v_add_f32_e32 v9, -1.0, v12                                                  ; 061218F3
>	v_add_f32_e32 v10, -1.0, v13                                                 ; 06141AF3
>	v_mad_f32 v2, v0, v2, 1.0                                                    ; D2820002 03CA0500
>	v_mad_f32 v9, v0, v9, 1.0                                                    ; D2820009 03CA1300
>	v_mad_f32 v10, v0, v10, 1.0                                                  ; D282000A 03CA1500
>	v_mad_f32 v0, v14, v0, -v0                                                   ; D2820000 8402010E
>	v_mad_f32 v3, v3, s28, 1.0                                                   ; D2820003 03C83903
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mac_f32_e32 v18, v15, v1                                                   ; 3E24030F
>	v_madak_f32_e32 v1, v18, v3, 0xbf147ae1                                      ; 42020712 BF147AE1
>	v_mul_f32_e32 v1, 0x41200000, v1                                             ; 100202FF 41200000
>	v_add_f32_e64 v1, 0, v1 clamp                                                ; D2060801 00020280
>	v_mad_f32 v3, -2.0, v1, v4                                                   ; D2820003 041202F5
>	v_mul_f32_e32 v1, v1, v1                                                     ; 10020301
>	v_mul_f32_e32 v11, v3, v1                                                    ; 10160303
>	v_mad_f32 v1, v1, v3, 1.0                                                    ; D2820001 03CA0701
>	v_mul_f32_e32 v3, 0x42c80000, v11                                            ; 100616FF 42C80000
>	v_add_f32_e64 v3, 0, v3 clamp                                                ; D2060803 00020680
>	v_mad_f32 v11, -2.0, v3, v4                                                  ; D282000B 041206F5
>	v_mul_f32_e32 v3, v3, v3                                                     ; 10060703
>	v_mul_f32_e32 v12, v11, v3                                                   ; 1018070B
>	v_mad_f32 v3, -v3, v11, 1.0                                                  ; D2820003 23CA1703
>	v_mad_f32 v2, v3, v2, v12                                                    ; D2820002 04320503
>	v_mad_f32 v9, v3, v9, v12                                                    ; D2820009 04321303
>	v_mad_f32 v10, v3, v10, v12                                                  ; D282000A 04321503
>	v_mul_f32_e32 v11, 0x3d4ccccd, v12                                           ; 101618FF 3D4CCCCD
>	v_mad_f32 v13, s4, v3, v11                                                   ; D282000D 042E0604
>	v_mad_f32 v14, s5, v3, v11                                                   ; D282000E 042E0605
>	v_mad_f32 v11, s60, v3, v11                                                  ; D282000B 042E063C
>	v_mul_f32_e32 v15, v13, v2                                                   ; 101E050D
>	v_mul_f32_e32 v18, v14, v9                                                   ; 1024130E
>	v_mul_f32_e32 v19, v11, v10                                                  ; 1026150B
>	v_mul_f32_e32 v15, 0x3e99999a, v15                                           ; 101E1EFF 3E99999A
>	v_madmk_f32_e32 v15, v18, v15, 0x3f170a3d                                    ; 401E1F12 3F170A3D
>	v_madmk_f32_e32 v15, v19, v15, 0x3de147ae                                    ; 401E1F13 3DE147AE
>	v_add_f32_e32 v5, v15, v5                                                    ; 060A0B0F
>	v_mul_f32_e32 v5, 0x41649249, v5                                             ; 100A0AFF 41649249
>	v_add_f32_e64 v5, 0, v5 clamp                                                ; D2060805 00020A80
>	v_mac_f32_e32 v4, -2.0, v5                                                   ; 3E080AF5
>	v_mul_f32_e32 v5, v5, v5                                                     ; 100A0B05
>	v_mul_f32_e32 v4, v4, v5                                                     ; 10080B04
>	v_mad_f32 v4, -v4, v6, v6                                                    ; D2820004 241A0D04
>	v_mad_f32 v5, v13, v2, v4                                                    ; D2820005 0412050D
>	v_mad_f32 v6, v14, v9, v4                                                    ; D2820006 0412130E
>	v_mad_f32 v4, v11, v10, v4                                                   ; D2820004 0412150B
>	v_mul_f32_e32 v15, v5, v17                                                   ; 101E2305
>	v_mul_f32_e32 v18, v6, v17                                                   ; 10242306
>	v_mul_f32_e32 v19, v4, v17                                                   ; 10262304
>	v_mac_f32_e32 v15, v5, v17                                                   ; 3E1E2305
>	v_mac_f32_e32 v15, v13, v2                                                   ; 3E1E050D
>	v_mac_f32_e32 v18, v6, v17                                                   ; 3E242306
>	v_mac_f32_e32 v18, v14, v9                                                   ; 3E24130E
>	v_mac_f32_e32 v19, v4, v17                                                   ; 3E262304
>	v_mac_f32_e32 v19, v11, v10                                                  ; 3E26150B
>	v_add_f32_e64 v2, 0, v15 clamp                                               ; D2060802 00021E80
>	v_add_f32_e64 v4, 0, v18 clamp                                               ; D2060804 00022480
>	v_add_f32_e64 v5, 0, v19 clamp                                               ; D2060805 00022680
>	v_mul_f32_e32 v6, v2, v16                                                    ; 100C2102
>	v_mul_f32_e32 v9, v4, v16                                                    ; 10122104
>	v_mul_f32_e32 v10, v5, v16                                                   ; 10142105
>	image_sample v[17:20], 15, 0, 0, 0, 0, 0, 0, 0, v[7:8], s[20:27], s[16:19]   ; F0800F00 00851107
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v2, -v16, v2, v17                                                  ; D2820002 24460510
>	v_mad_f32 v4, -v16, v4, v18                                                  ; D2820004 244A0910
>	v_mad_f32 v5, -v16, v5, v19                                                  ; D2820005 244E0B10
>	v_mul_f32_e32 v11, s62, v16                                                  ; 1016203E
>	v_mac_f32_e32 v11, v11, v0                                                   ; 3E16010B
>	image_sample v0, 1, 0, 0, 0, 0, 0, 0, 0, v[7:8], s[8:15], s[0:3]             ; F0800100 00020007
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_subrev_f32_e32 v0, v0, v1                                                  ; 0A000300
>	v_add_f32_e64 v0, 0, v0 clamp                                                ; D2060800 00020080
>	v_mac_f32_e32 v6, v2, v0                                                     ; 3E0C0102
>	v_mac_f32_e32 v9, v4, v0                                                     ; 3E120104
>	v_mac_f32_e32 v10, v5, v0                                                    ; 3E140105
>	v_mac_f32_e32 v12, s62, v3                                                   ; 3E18063E
>	v_mul_f32_e32 v1, v12, v11                                                   ; 1002170C
>	v_mul_f32_e32 v2, s61, v1                                                    ; 1004023D
>	v_mad_f32 v1, -s61, v1, v20                                                  ; D2820001 2452023D
>	v_mac_f32_e32 v2, v1, v0                                                     ; 3E040101
>	v_cvt_pkrtz_f16_f32_e32 v0, v6, v9                                           ; 5E001306
>	v_cvt_pkrtz_f16_f32_e32 v1, v10, v2                                          ; 5E02050A
>	exp 15, 0, 1, 1, 1, v0, v1, v0, v1                                           ; F8001C0F 01000100
>	s_endpgm                                                                     ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 72
>VGPRS: 24
>Code Size: 728 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
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL OUT[1], CLIPVERTEX
>DCL OUT[2], GENERIC[0]
>DCL OUT[3], GENERIC[1]
>DCL CONST[0..54]
>DCL TEMP[0..3], LOCAL
>  0: MUL TEMP[0].xy, CONST[48].xyyy, IN[1].xyyy
>  1: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
>  2: ADD TEMP[1].x, TEMP[0].xxxx, CONST[48].wwww
>  3: MOV TEMP[1].z, TEMP[1].xxxx
>  4: MUL TEMP[0].xy, CONST[49].xyyy, IN[1].xyyy
>  5: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
>  6: ADD TEMP[2].x, TEMP[0].xxxx, CONST[49].wwww
>  7: MOV TEMP[1].w, TEMP[2].xxxx
>  8: MUL TEMP[0].xy, CONST[50].xyyy, IN[1].xyyy
>  9: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 10: ADD TEMP[2].x, TEMP[0].xxxx, CONST[50].wwww
> 11: MUL TEMP[0].xy, CONST[51].xyyy, IN[1].xyyy
> 12: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 13: ADD TEMP[3].x, TEMP[0].xxxx, CONST[51].wwww
> 14: MOV TEMP[2].y, TEMP[3].xxxx
> 15: MUL TEMP[0].xy, CONST[52].xyyy, IN[1].xyyy
> 16: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 17: ADD TEMP[3].x, TEMP[0].xxxx, CONST[52].wwww
> 18: MOV TEMP[2].z, TEMP[3].xxxx
> 19: MUL TEMP[0].xy, CONST[53].xyyy, IN[1].xyyy
> 20: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 21: ADD TEMP[0].x, TEMP[0].xxxx, CONST[53].wwww
> 22: MOV TEMP[2].w, TEMP[0].xxxx
> 23: MOV TEMP[0].xw, IN[0].xxxw
> 24: MOV TEMP[1].xy, IN[1].xyxx
> 25: MAD TEMP[3].x, IN[0].zzzz, CONST[0].zzzz, -IN[0].wwww
> 26: MOV TEMP[0].z, TEMP[3].xxxx
> 27: MOV TEMP[0].y, -IN[0].yyyy
> 28: MAD TEMP[0].xy, CONST[54].xyyy, IN[0].wwww, TEMP[0].xyyy
> 29: MOV OUT[2], TEMP[1]
> 30: MOV OUT[3], TEMP[2]
> 31: MOV OUT[0], TEMP[0]
> 32: MOV OUT[1], IN[0]
> 33: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = call float @llvm.SI.load.const(<16 x i8> %13, i32 8)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %13, i32 768)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %13, i32 772)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %13, i32 780)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %13, i32 784)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %13, i32 788)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %13, i32 796)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %13, i32 800)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %13, i32 804)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %13, i32 812)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %13, i32 816)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %13, i32 820)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %13, i32 828)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %13, i32 832)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %13, i32 836)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %13, i32 844)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %13, i32 848)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %13, i32 852)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %13, i32 860)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %13, i32 864)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %13, i32 868)
>  %35 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %36 = load <16 x i8>, <16 x i8> addrspace(2)* %35, align 16, !tbaa !0
>  %37 = add i32 %5, %8
>  %38 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %36, i32 0, i32 %37)
>  %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 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 1
>  %44 = load <16 x i8>, <16 x i8> addrspace(2)* %43, align 16, !tbaa !0
>  %45 = add i32 %5, %8
>  %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 = fmul float %15, %47
>  %50 = fmul float %16, %48
>  %51 = fadd float %50, %49
>  %52 = fadd float %51, %17
>  %53 = fmul float %18, %47
>  %54 = fmul float %19, %48
>  %55 = fadd float %54, %53
>  %56 = fadd float %55, %20
>  %57 = fmul float %21, %47
>  %58 = fmul float %22, %48
>  %59 = fadd float %58, %57
>  %60 = fadd float %59, %23
>  %61 = fmul float %24, %47
>  %62 = fmul float %25, %48
>  %63 = fadd float %62, %61
>  %64 = fadd float %63, %26
>  %65 = fmul float %27, %47
>  %66 = fmul float %28, %48
>  %67 = fadd float %66, %65
>  %68 = fadd float %67, %29
>  %69 = fmul float %30, %47
>  %70 = fmul float %31, %48
>  %71 = fadd float %70, %69
>  %72 = fadd float %71, %32
>  %73 = fmul float %41, %14
>  %74 = fsub float %73, %42
>  %75 = fmul float %33, %42
>  %76 = fadd float %75, %39
>  %77 = fmul float %34, %42
>  %78 = fsub float %77, %40
>  %79 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 16
>  %80 = load <16 x i8>, <16 x i8> addrspace(2)* %79, align 16, !tbaa !0
>  %81 = call float @llvm.SI.load.const(<16 x i8> %80, i32 0)
>  %82 = fmul float %81, %39
>  %83 = call float @llvm.SI.load.const(<16 x i8> %80, i32 4)
>  %84 = fmul float %83, %40
>  %85 = fadd float %82, %84
>  %86 = call float @llvm.SI.load.const(<16 x i8> %80, i32 8)
>  %87 = fmul float %86, %41
>  %88 = fadd float %85, %87
>  %89 = call float @llvm.SI.load.const(<16 x i8> %80, i32 12)
>  %90 = fmul float %89, %42
>  %91 = fadd float %88, %90
>  %92 = call float @llvm.SI.load.const(<16 x i8> %80, i32 16)
>  %93 = fmul float %92, %39
>  %94 = call float @llvm.SI.load.const(<16 x i8> %80, i32 20)
>  %95 = fmul float %94, %40
>  %96 = fadd float %93, %95
>  %97 = call float @llvm.SI.load.const(<16 x i8> %80, i32 24)
>  %98 = fmul float %97, %41
>  %99 = fadd float %96, %98
>  %100 = call float @llvm.SI.load.const(<16 x i8> %80, i32 28)
>  %101 = fmul float %100, %42
>  %102 = fadd float %99, %101
>  %103 = call float @llvm.SI.load.const(<16 x i8> %80, i32 32)
>  %104 = fmul float %103, %39
>  %105 = call float @llvm.SI.load.const(<16 x i8> %80, i32 36)
>  %106 = fmul float %105, %40
>  %107 = fadd float %104, %106
>  %108 = call float @llvm.SI.load.const(<16 x i8> %80, i32 40)
>  %109 = fmul float %108, %41
>  %110 = fadd float %107, %109
>  %111 = call float @llvm.SI.load.const(<16 x i8> %80, i32 44)
>  %112 = fmul float %111, %42
>  %113 = fadd float %110, %112
>  %114 = call float @llvm.SI.load.const(<16 x i8> %80, i32 48)
>  %115 = fmul float %114, %39
>  %116 = call float @llvm.SI.load.const(<16 x i8> %80, i32 52)
>  %117 = fmul float %116, %40
>  %118 = fadd float %115, %117
>  %119 = call float @llvm.SI.load.const(<16 x i8> %80, i32 56)
>  %120 = fmul float %119, %41
>  %121 = fadd float %118, %120
>  %122 = call float @llvm.SI.load.const(<16 x i8> %80, i32 60)
>  %123 = fmul float %122, %42
>  %124 = fadd float %121, %123
>  %125 = call float @llvm.SI.load.const(<16 x i8> %80, i32 64)
>  %126 = fmul float %125, %39
>  %127 = call float @llvm.SI.load.const(<16 x i8> %80, i32 68)
>  %128 = fmul float %127, %40
>  %129 = fadd float %126, %128
>  %130 = call float @llvm.SI.load.const(<16 x i8> %80, i32 72)
>  %131 = fmul float %130, %41
>  %132 = fadd float %129, %131
>  %133 = call float @llvm.SI.load.const(<16 x i8> %80, i32 76)
>  %134 = fmul float %133, %42
>  %135 = fadd float %132, %134
>  %136 = call float @llvm.SI.load.const(<16 x i8> %80, i32 80)
>  %137 = fmul float %136, %39
>  %138 = call float @llvm.SI.load.const(<16 x i8> %80, i32 84)
>  %139 = fmul float %138, %40
>  %140 = fadd float %137, %139
>  %141 = call float @llvm.SI.load.const(<16 x i8> %80, i32 88)
>  %142 = fmul float %141, %41
>  %143 = fadd float %140, %142
>  %144 = call float @llvm.SI.load.const(<16 x i8> %80, i32 92)
>  %145 = fmul float %144, %42
>  %146 = fadd float %143, %145
>  %147 = call float @llvm.SI.load.const(<16 x i8> %80, i32 96)
>  %148 = fmul float %147, %39
>  %149 = call float @llvm.SI.load.const(<16 x i8> %80, i32 100)
>  %150 = fmul float %149, %40
>  %151 = fadd float %148, %150
>  %152 = call float @llvm.SI.load.const(<16 x i8> %80, i32 104)
>  %153 = fmul float %152, %41
>  %154 = fadd float %151, %153
>  %155 = call float @llvm.SI.load.const(<16 x i8> %80, i32 108)
>  %156 = fmul float %155, %42
>  %157 = fadd float %154, %156
>  %158 = call float @llvm.SI.load.const(<16 x i8> %80, i32 112)
>  %159 = fmul float %158, %39
>  %160 = call float @llvm.SI.load.const(<16 x i8> %80, i32 116)
>  %161 = fmul float %160, %40
>  %162 = fadd float %159, %161
>  %163 = call float @llvm.SI.load.const(<16 x i8> %80, i32 120)
>  %164 = fmul float %163, %41
>  %165 = fadd float %162, %164
>  %166 = call float @llvm.SI.load.const(<16 x i8> %80, i32 124)
>  %167 = fmul float %166, %42
>  %168 = fadd float %165, %167
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %47, float %48, float %52, float %56)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %60, float %64, float %68, float %72)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 12, i32 0, float %76, float %78, float %74, float %42)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 13, i32 0, float %91, float %102, float %113, float %124)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 14, i32 0, float %135, float %146, float %157, 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
>
>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[4:7], s[8:9], 0x0                    ; C0820900
>	s_load_dwordx4 s[12:15], s[8:9], 0x4                  ; C0860904
>	v_add_i32_e32 v0, vcc, s10, v0                        ; 4A00000A
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[4:7], 0 idxen   ; E00C2000 80010100
>	buffer_load_format_xyzw v[5:8], v0, s[12:15], 0 idxen ; E00C2000 80030500
>	s_load_dwordx4 s[4:7], s[2:3], 0x0                    ; C0820300
>	s_load_dwordx4 s[0:3], s[2:3], 0x40                   ; C0800340
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	s_buffer_load_dword s8, s[4:7], 0x2                   ; C2040502
>	s_buffer_load_dword s9, s[4:7], 0xc0                  ; C20485C0
>	s_buffer_load_dword s10, s[4:7], 0xc1                 ; C20505C1
>	s_buffer_load_dword s11, s[4:7], 0xc3                 ; C20585C3
>	s_buffer_load_dword s12, s[4:7], 0xc4                 ; C20605C4
>	s_buffer_load_dword s13, s[4:7], 0xc5                 ; C20685C5
>	s_buffer_load_dword s14, s[4:7], 0xc7                 ; C20705C7
>	s_buffer_load_dword s15, s[4:7], 0xc8                 ; C20785C8
>	s_buffer_load_dword s16, s[4:7], 0xc9                 ; C20805C9
>	s_buffer_load_dword s17, s[4:7], 0xcb                 ; C20885CB
>	s_buffer_load_dword s18, s[4:7], 0xcc                 ; C20905CC
>	s_buffer_load_dword s19, s[4:7], 0xcd                 ; C20985CD
>	s_buffer_load_dword s20, s[4:7], 0xcf                 ; C20A05CF
>	s_buffer_load_dword s21, s[4:7], 0xd0                 ; C20A85D0
>	s_buffer_load_dword s22, s[4:7], 0xd1                 ; C20B05D1
>	s_buffer_load_dword s23, s[4:7], 0xd3                 ; C20B85D3
>	s_buffer_load_dword s24, s[4:7], 0xd4                 ; C20C05D4
>	s_buffer_load_dword s25, s[4:7], 0xd5                 ; C20C85D5
>	s_buffer_load_dword s26, s[4:7], 0xd7                 ; C20D05D7
>	s_buffer_load_dword s27, s[4:7], 0xd8                 ; C20D85D8
>	s_buffer_load_dword s4, s[4:7], 0xd9                  ; C20205D9
>	s_buffer_load_dword s5, s[0:3], 0x0                   ; C2028100
>	s_buffer_load_dword s6, s[0:3], 0x1                   ; C2030101
>	s_buffer_load_dword s7, s[0:3], 0x2                   ; C2038102
>	s_buffer_load_dword s28, s[0:3], 0x3                  ; C20E0103
>	s_buffer_load_dword s29, s[0:3], 0x4                  ; C20E8104
>	s_buffer_load_dword s30, s[0:3], 0x5                  ; C20F0105
>	s_buffer_load_dword s31, s[0:3], 0x6                  ; C20F8106
>	s_buffer_load_dword s32, s[0:3], 0x7                  ; C2100107
>	s_buffer_load_dword s33, s[0:3], 0x8                  ; C2108108
>	s_buffer_load_dword s34, s[0:3], 0x9                  ; C2110109
>	s_buffer_load_dword s35, s[0:3], 0xa                  ; C211810A
>	s_buffer_load_dword s36, s[0:3], 0xb                  ; C212010B
>	s_buffer_load_dword s37, s[0:3], 0xc                  ; C212810C
>	s_buffer_load_dword s38, s[0:3], 0xd                  ; C213010D
>	s_buffer_load_dword s39, s[0:3], 0xe                  ; C213810E
>	s_buffer_load_dword s40, s[0:3], 0xf                  ; C214010F
>	s_buffer_load_dword s41, s[0:3], 0x10                 ; C2148110
>	s_buffer_load_dword s42, s[0:3], 0x11                 ; C2150111
>	s_buffer_load_dword s43, s[0:3], 0x12                 ; C2158112
>	s_buffer_load_dword s44, s[0:3], 0x13                 ; C2160113
>	s_buffer_load_dword s45, s[0:3], 0x14                 ; C2168114
>	s_buffer_load_dword s46, s[0:3], 0x15                 ; C2170115
>	s_buffer_load_dword s47, s[0:3], 0x16                 ; C2178116
>	s_buffer_load_dword s48, s[0:3], 0x17                 ; C2180117
>	s_buffer_load_dword s49, s[0:3], 0x18                 ; C2188118
>	s_buffer_load_dword s50, s[0:3], 0x19                 ; C2190119
>	s_buffer_load_dword s51, s[0:3], 0x1a                 ; C219811A
>	s_buffer_load_dword s52, s[0:3], 0x1b                 ; C21A011B
>	s_buffer_load_dword s53, s[0:3], 0x1c                 ; C21A811C
>	s_buffer_load_dword s54, s[0:3], 0x1d                 ; C21B011D
>	s_buffer_load_dword s55, s[0:3], 0x1e                 ; C21B811E
>	s_buffer_load_dword s0, s[0:3], 0x1f                  ; C200011F
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mov_b32_e32 v0, s11                                 ; 7E00020B
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mov_b32_e32 v7, s14                                 ; 7E0E020E
>	v_mov_b32_e32 v8, s17                                 ; 7E100211
>	v_mov_b32_e32 v9, s20                                 ; 7E120214
>	v_mov_b32_e32 v10, s23                                ; 7E140217
>	v_mov_b32_e32 v11, s26                                ; 7E16021A
>	v_mac_f32_e32 v0, s9, v5                              ; 3E000A09
>	v_mac_f32_e32 v7, s12, v5                             ; 3E0E0A0C
>	v_mac_f32_e32 v8, s15, v5                             ; 3E100A0F
>	v_mac_f32_e32 v9, s18, v5                             ; 3E120A12
>	v_mac_f32_e32 v10, s21, v5                            ; 3E140A15
>	v_mac_f32_e32 v11, s24, v5                            ; 3E160A18
>	v_mac_f32_e32 v0, s10, v6                             ; 3E000C0A
>	v_mac_f32_e32 v7, s13, v6                             ; 3E0E0C0D
>	v_mac_f32_e32 v8, s16, v6                             ; 3E100C10
>	v_mac_f32_e32 v9, s19, v6                             ; 3E120C13
>	exp 15, 32, 0, 0, 0, v5, v6, v0, v7                   ; F800020F 07000605
>	v_mac_f32_e32 v10, s22, v6                            ; 3E140C16
>	v_mac_f32_e32 v11, s25, v6                            ; 3E160C19
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mad_f32 v0, v3, s8, -v4                             ; D2820000 84101103
>	v_mad_f32 v5, s27, v4, v1                             ; D2820005 0406081B
>	v_mad_f32 v6, s4, v4, -v2                             ; D2820006 840A0804
>	v_mul_f32_e32 v7, s6, v2                              ; 100E0406
>	exp 15, 33, 0, 0, 0, v8, v9, v10, v11                 ; F800021F 0B0A0908
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mul_f32_e32 v8, s30, v2                             ; 1010041E
>	v_mac_f32_e32 v7, s5, v1                              ; 3E0E0205
>	exp 15, 12, 0, 0, 0, v5, v6, v0, v4                   ; F80000CF 04000605
>	v_mac_f32_e32 v8, s29, v1                             ; 3E10021D
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mul_f32_e32 v0, s34, v2                             ; 10000422
>	v_mac_f32_e32 v0, s33, v1                             ; 3E000221
>	v_mul_f32_e32 v5, s38, v2                             ; 100A0426
>	v_mac_f32_e32 v5, s37, v1                             ; 3E0A0225
>	v_mul_f32_e32 v6, s42, v2                             ; 100C042A
>	v_mul_f32_e32 v9, s46, v2                             ; 1012042E
>	v_mul_f32_e32 v10, s50, v2                            ; 10140432
>	v_mul_f32_e32 v2, s54, v2                             ; 10040436
>	v_mac_f32_e32 v6, s41, v1                             ; 3E0C0229
>	v_mac_f32_e32 v9, s45, v1                             ; 3E12022D
>	v_mac_f32_e32 v10, s49, v1                            ; 3E140231
>	v_mac_f32_e32 v2, s53, v1                             ; 3E040235
>	v_mac_f32_e32 v7, s7, v3                              ; 3E0E0607
>	v_mac_f32_e32 v8, s31, v3                             ; 3E10061F
>	v_mac_f32_e32 v0, s35, v3                             ; 3E000623
>	v_mac_f32_e32 v5, s39, v3                             ; 3E0A0627
>	v_mac_f32_e32 v6, s43, v3                             ; 3E0C062B
>	v_mac_f32_e32 v9, s47, v3                             ; 3E12062F
>	v_mac_f32_e32 v10, s51, v3                            ; 3E140633
>	v_mac_f32_e32 v2, s55, v3                             ; 3E040637
>	v_mac_f32_e32 v7, s28, v4                             ; 3E0E081C
>	v_mac_f32_e32 v8, s32, v4                             ; 3E100820
>	v_mac_f32_e32 v0, s36, v4                             ; 3E000824
>	v_mac_f32_e32 v5, s40, v4                             ; 3E0A0828
>	v_mac_f32_e32 v6, s44, v4                             ; 3E0C082C
>	v_mac_f32_e32 v9, s48, v4                             ; 3E120830
>	v_mac_f32_e32 v10, s52, v4                            ; 3E140834
>	v_mac_f32_e32 v2, s0, v4                              ; 3E040800
>	exp 15, 13, 0, 0, 0, v7, v8, v0, v5                   ; F80000DF 05000807
>	exp 15, 14, 0, 1, 0, v6, v9, v10, v2                  ; F80008EF 020A0906
>	s_endpgm                                              ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 64
>VGPRS: 12
>Code Size: 544 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
>  clamp_color = 0
>FRAG
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL IN[1], GENERIC[1], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SAMP[1]
>DCL SAMP[2]
>DCL SAMP[3]
>DCL SAMP[4]
>DCL SVIEW[0], 2D, FLOAT
>DCL SVIEW[1], 2D, FLOAT
>DCL SVIEW[2], 2D, FLOAT
>DCL SVIEW[3], 2D, FLOAT
>DCL SVIEW[4], 2D, FLOAT
>DCL CONST[0..3]
>DCL TEMP[0..6], LOCAL
>IMM[0] FLT32 {    6.0000,     1.0000,     0.7500,     0.2500}
>IMM[1] FLT32 {   -1.0000,     0.3000,     0.5900,     0.1100}
>IMM[2] FLT32 {   -0.0800,    14.2857,    -2.0000,     3.0000}
>IMM[3] FLT32 {    0.0300,     0.5000,    -0.5800,    10.0000}
>IMM[4] FLT32 {   -0.9000,    -0.0100,    10.0000,  -100.0000}
>IMM[5] FLT32 {    0.0000,     0.0000,     0.0000,     0.0000}
>  0: ADD TEMP[0].xyz, -CONST[0].xyzz, CONST[1].xyzz
>  1: MOV TEMP[1].xy, IN[0].xyyy
>  2: TEX TEMP[1].xyz, TEMP[1], SAMP[3], 2D
>  3: MAD TEMP[0].xyz, TEMP[1].xxxx, TEMP[0].xyzz, CONST[0].xyzz
>  4: LRP TEMP[2].xyz, TEMP[1].yyyy, CONST[2].xyzz, TEMP[0].xyzz
>  5: ADD TEMP[0].x, -TEMP[2].xxxx, CONST[0].wwww
>  6: ADD TEMP[3].x, -TEMP[2].yyyy, CONST[1].wwww
>  7: MOV TEMP[0].y, TEMP[3].xxxx
>  8: ADD TEMP[3].x, -TEMP[2].zzzz, CONST[2].wwww
>  9: MOV TEMP[0].z, TEMP[3].xxxx
> 10: MAD TEMP[0].xyz, TEMP[1].zzzz, TEMP[0].xyzz, TEMP[2].xyzz
> 11: MOV TEMP[1].xy, IN[1].zwww
> 12: TEX TEMP[1], TEMP[1], SAMP[4], 2D
> 13: MUL TEMP[3].x, IMM[0].zzzz, CONST[3].wwww
> 14: MOV TEMP[4].xy, IN[0].xyyy
> 15: TEX TEMP[4], TEMP[4], SAMP[0], 2D
> 16: ADD TEMP[2].x, -TEMP[4].xxxx, IMM[0].yyyy
> 17: MUL TEMP[2].x, TEMP[2].xxxx, TEMP[2].xxxx
> 18: MUL TEMP[2].x, TEMP[2].xxxx, TEMP[2].xxxx
> 19: MAD TEMP[3].x, TEMP[2].xxxx, IMM[0].wwww, TEMP[3].xxxx
> 20: ADD TEMP[5], TEMP[1], IMM[1].xxxx
> 21: MAD TEMP[3], TEMP[3].xxxx, TEMP[5], IMM[0].yyyy
> 22: MUL TEMP[0].xyz, TEMP[0].xyzz, TEMP[3].xyzz
> 23: DP3 TEMP[5].x, TEMP[0].xyzz, IMM[1].yzww
> 24: ADD TEMP[5].x, TEMP[5].xxxx, IMM[2].xxxx
> 25: MUL TEMP[5].x, TEMP[5].xxxx, IMM[2].yyyy
> 26: MOV_SAT TEMP[5].x, TEMP[5].xxxx
> 27: MAD TEMP[1].x, TEMP[5].xxxx, IMM[2].zzzz, IMM[2].wwww
> 28: MUL TEMP[5].x, TEMP[5].xxxx, TEMP[5].xxxx
> 29: MAD TEMP[5].x, TEMP[1].xxxx, -TEMP[5].xxxx, IMM[0].yyyy
> 30: MAD TEMP[1].xyz, TEMP[5].xxxx, IMM[3].xxxx, TEMP[0].xyzz
> 31: MUL TEMP[1].xyz, TEMP[1].xyzz, TEMP[4].zzzz
> 32: MAD TEMP[0].xyz, TEMP[1].xyzz, IMM[3].yyyy, TEMP[0].xyzz
> 33: MOV_SAT TEMP[5].xyz, TEMP[0].xyzz
> 34: MUL TEMP[1].xyz, TEMP[4].yyyy, TEMP[5].xyzz
> 35: MOV TEMP[6].xy, IN[0].xyyy
> 36: TEX TEMP[6], TEMP[6], SAMP[2], 2D
> 37: MAD TEMP[0].xyz, TEMP[5].xyzz, -TEMP[4].yyyy, TEMP[6].xyzz
> 38: MOV TEMP[5].xy, IN[1].xyyy
> 39: TEX TEMP[5].y, TEMP[5], SAMP[1], 2D
> 40: MAD TEMP[5].x, TEMP[5].yyyy, TEMP[4].xxxx, TEMP[4].wwww
> 41: MUL TEMP[4].x, TEMP[4].yyyy, CONST[3].zzzz
> 42: MUL TEMP[3].x, TEMP[3].wwww, TEMP[4].xxxx
> 43: MAD TEMP[2].x, CONST[3].wwww, IMM[0].xxxx, IMM[0].yyyy
> 44: MAD TEMP[4].x, TEMP[5].xxxx, TEMP[2].xxxx, IMM[3].zzzz
> 45: MUL TEMP[4].x, TEMP[4].xxxx, IMM[3].wwww
> 46: MOV_SAT TEMP[4].x, TEMP[4].xxxx
> 47: MAD TEMP[2].x, TEMP[4].xxxx, IMM[2].zzzz, IMM[2].wwww
> 48: MUL TEMP[4].x, TEMP[4].xxxx, TEMP[4].xxxx
> 49: MUL TEMP[5].x, TEMP[4].xxxx, TEMP[2].xxxx
> 50: MAD TEMP[2].xy, TEMP[2].xxxx, TEMP[4].xxxx, IMM[4].xyyy
> 51: MUL TEMP[2].xy, TEMP[2].xyyy, IMM[4].zwww
> 52: MAD TEMP[1].xyz, TEMP[5].xxxx, TEMP[0].xyzz, TEMP[1].xyzz
> 53: MAX TEMP[4].x, TEMP[2].yyyy, IMM[5].xxxx
> 54: MOV_SAT TEMP[2].x, TEMP[2].xxxx
> 55: MAD TEMP[5].x, TEMP[4].xxxx, IMM[2].zzzz, IMM[2].wwww
> 56: MUL TEMP[0].x, TEMP[4].xxxx, TEMP[4].xxxx
> 57: MUL TEMP[0].x, TEMP[0].xxxx, TEMP[5].xxxx
> 58: MAD TEMP[4].x, TEMP[3].xxxx, -TEMP[0].xxxx, TEMP[6].wwww
> 59: MUL TEMP[0].x, TEMP[0].xxxx, TEMP[3].xxxx
> 60: MAD TEMP[3].x, TEMP[2].xxxx, IMM[2].zzzz, IMM[2].wwww
> 61: MUL TEMP[2].x, TEMP[2].xxxx, TEMP[2].xxxx
> 62: MUL TEMP[2].x, TEMP[2].xxxx, TEMP[3].xxxx
> 63: MAD TEMP[0].x, TEMP[2].xxxx, TEMP[4].xxxx, TEMP[0].xxxx
> 64: MOV TEMP[1].w, TEMP[0].xxxx
> 65: MOV OUT[0], TEMP[1]
> 66: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %24 = load <16 x i8>, <16 x i8> addrspace(2)* %23, align 16, !tbaa !0
>  %25 = call float @llvm.SI.load.const(<16 x i8> %24, i32 0)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %24, i32 4)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %24, i32 8)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %24, i32 12)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %24, i32 16)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %24, i32 20)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %24, i32 24)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %24, i32 28)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %24, i32 32)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %24, i32 36)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %24, i32 40)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %24, i32 44)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %24, i32 56)
>  %38 = call float @llvm.SI.load.const(<16 x i8> %24, i32 60)
>  %39 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %40 = load <8 x i32>, <8 x i32> addrspace(2)* %39, align 32, !tbaa !0
>  %41 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %42 = load <4 x i32>, <4 x i32> addrspace(2)* %41, align 16, !tbaa !0
>  %43 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 1
>  %44 = load <8 x i32>, <8 x i32> addrspace(2)* %43, align 32, !tbaa !0
>  %45 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 1
>  %46 = load <4 x i32>, <4 x i32> addrspace(2)* %45, align 16, !tbaa !0
>  %47 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 2
>  %48 = load <8 x i32>, <8 x i32> addrspace(2)* %47, align 32, !tbaa !0
>  %49 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 2
>  %50 = load <4 x i32>, <4 x i32> addrspace(2)* %49, align 16, !tbaa !0
>  %51 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 3
>  %52 = load <8 x i32>, <8 x i32> addrspace(2)* %51, align 32, !tbaa !0
>  %53 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 3
>  %54 = load <4 x i32>, <4 x i32> addrspace(2)* %53, align 16, !tbaa !0
>  %55 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 4
>  %56 = load <8 x i32>, <8 x i32> addrspace(2)* %55, align 32, !tbaa !0
>  %57 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 4
>  %58 = load <4 x i32>, <4 x i32> addrspace(2)* %57, align 16, !tbaa !0
>  %59 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %8)
>  %60 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %8)
>  %61 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %6, <2 x i32> %8)
>  %62 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %6, <2 x i32> %8)
>  %63 = call float @llvm.SI.fs.interp(i32 2, i32 1, i32 %6, <2 x i32> %8)
>  %64 = call float @llvm.SI.fs.interp(i32 3, i32 1, i32 %6, <2 x i32> %8)
>  %65 = fsub float %29, %25
>  %66 = fsub float %30, %26
>  %67 = fsub float %31, %27
>  %68 = bitcast float %59 to i32
>  %69 = bitcast float %60 to i32
>  %70 = insertelement <2 x i32> undef, i32 %68, i32 0
>  %71 = insertelement <2 x i32> %70, i32 %69, i32 1
>  %72 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %71, <8 x i32> %52, <4 x i32> %54, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %73 = extractelement <4 x float> %72, i32 0
>  %74 = extractelement <4 x float> %72, i32 1
>  %75 = extractelement <4 x float> %72, i32 2
>  %76 = fmul float %73, %65
>  %77 = fadd float %76, %25
>  %78 = fmul float %73, %66
>  %79 = fadd float %78, %26
>  %80 = fmul float %73, %67
>  %81 = fadd float %80, %27
>  %82 = fsub float 1.000000e+00, %74
>  %83 = fmul float %33, %74
>  %84 = fmul float %77, %82
>  %85 = fadd float %83, %84
>  %86 = fsub float 1.000000e+00, %74
>  %87 = fmul float %34, %74
>  %88 = fmul float %79, %86
>  %89 = fadd float %87, %88
>  %90 = fsub float 1.000000e+00, %74
>  %91 = fmul float %35, %74
>  %92 = fmul float %81, %90
>  %93 = fadd float %91, %92
>  %94 = fsub float %28, %85
>  %95 = fsub float %32, %89
>  %96 = fsub float %36, %93
>  %97 = fmul float %75, %94
>  %98 = fadd float %97, %85
>  %99 = fmul float %75, %95
>  %100 = fadd float %99, %89
>  %101 = fmul float %75, %96
>  %102 = fadd float %101, %93
>  %103 = bitcast float %63 to i32
>  %104 = bitcast float %64 to i32
>  %105 = insertelement <2 x i32> undef, i32 %103, i32 0
>  %106 = insertelement <2 x i32> %105, i32 %104, i32 1
>  %107 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %106, <8 x i32> %56, <4 x i32> %58, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %108 = extractelement <4 x float> %107, i32 0
>  %109 = extractelement <4 x float> %107, i32 1
>  %110 = extractelement <4 x float> %107, i32 2
>  %111 = extractelement <4 x float> %107, i32 3
>  %112 = fmul float %38, 7.500000e-01
>  %113 = bitcast float %59 to i32
>  %114 = bitcast float %60 to i32
>  %115 = insertelement <2 x i32> undef, i32 %113, i32 0
>  %116 = insertelement <2 x i32> %115, i32 %114, i32 1
>  %117 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %116, <8 x i32> %40, <4 x i32> %42, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %118 = extractelement <4 x float> %117, i32 0
>  %119 = extractelement <4 x float> %117, i32 1
>  %120 = extractelement <4 x float> %117, i32 2
>  %121 = extractelement <4 x float> %117, i32 3
>  %122 = fsub float 1.000000e+00, %118
>  %123 = fmul float %122, %122
>  %124 = fmul float %123, %123
>  %125 = fmul float %124, 2.500000e-01
>  %126 = fadd float %125, %112
>  %127 = fadd float %108, -1.000000e+00
>  %128 = fadd float %109, -1.000000e+00
>  %129 = fadd float %110, -1.000000e+00
>  %130 = fadd float %111, -1.000000e+00
>  %131 = fmul float %126, %127
>  %132 = fadd float %131, 1.000000e+00
>  %133 = fmul float %126, %128
>  %134 = fadd float %133, 1.000000e+00
>  %135 = fmul float %126, %129
>  %136 = fadd float %135, 1.000000e+00
>  %137 = fmul float %126, %130
>  %138 = fadd float %137, 1.000000e+00
>  %139 = fmul float %98, %132
>  %140 = fmul float %100, %134
>  %141 = fmul float %102, %136
>  %142 = fmul float %139, 0x3FD3333340000000
>  %143 = fmul float %140, 0x3FE2E147A0000000
>  %144 = fadd float %143, %142
>  %145 = fmul float %141, 0x3FBC28F5C0000000
>  %146 = fadd float %144, %145
>  %147 = fadd float %146, 0xBFB47AE140000000
>  %148 = fmul float %147, 0x402C924920000000
>  %149 = call float @llvm.AMDIL.clamp.(float %148, float 0.000000e+00, float 1.000000e+00)
>  %150 = fmul float %149, -2.000000e+00
>  %151 = fadd float %150, 3.000000e+00
>  %152 = fmul float %149, %149
>  %153 = fmul float %152, %151
>  %154 = fsub float 1.000000e+00, %153
>  %155 = fmul float %154, 0x3F9EB851E0000000
>  %156 = fadd float %155, %139
>  %157 = fmul float %154, 0x3F9EB851E0000000
>  %158 = fadd float %157, %140
>  %159 = fmul float %154, 0x3F9EB851E0000000
>  %160 = fadd float %159, %141
>  %161 = fmul float %156, %120
>  %162 = fmul float %158, %120
>  %163 = fmul float %160, %120
>  %164 = fmul float %161, 5.000000e-01
>  %165 = fadd float %164, %139
>  %166 = fmul float %162, 5.000000e-01
>  %167 = fadd float %166, %140
>  %168 = fmul float %163, 5.000000e-01
>  %169 = fadd float %168, %141
>  %170 = call float @llvm.AMDIL.clamp.(float %165, float 0.000000e+00, float 1.000000e+00)
>  %171 = call float @llvm.AMDIL.clamp.(float %167, float 0.000000e+00, float 1.000000e+00)
>  %172 = call float @llvm.AMDIL.clamp.(float %169, float 0.000000e+00, float 1.000000e+00)
>  %173 = fmul float %119, %170
>  %174 = fmul float %119, %171
>  %175 = fmul float %119, %172
>  %176 = bitcast float %59 to i32
>  %177 = bitcast float %60 to i32
>  %178 = insertelement <2 x i32> undef, i32 %176, i32 0
>  %179 = insertelement <2 x i32> %178, i32 %177, i32 1
>  %180 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %179, <8 x i32> %48, <4 x i32> %50, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %181 = extractelement <4 x float> %180, i32 0
>  %182 = extractelement <4 x float> %180, i32 1
>  %183 = extractelement <4 x float> %180, i32 2
>  %184 = extractelement <4 x float> %180, i32 3
>  %185 = fmul float %119, %170
>  %186 = fsub float %181, %185
>  %187 = fmul float %119, %171
>  %188 = fsub float %182, %187
>  %189 = fmul float %119, %172
>  %190 = fsub float %183, %189
>  %191 = bitcast float %61 to i32
>  %192 = bitcast float %62 to i32
>  %193 = insertelement <2 x i32> undef, i32 %191, i32 0
>  %194 = insertelement <2 x i32> %193, i32 %192, i32 1
>  %195 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %194, <8 x i32> %44, <4 x i32> %46, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %196 = extractelement <4 x float> %195, i32 1
>  %197 = fmul float %196, %118
>  %198 = fadd float %197, %121
>  %199 = fmul float %119, %37
>  %200 = fmul float %138, %199
>  %201 = fmul float %38, 6.000000e+00
>  %202 = fadd float %201, 1.000000e+00
>  %203 = fmul float %198, %202
>  %204 = fadd float %203, 0xBFE28F5C20000000
>  %205 = fmul float %204, 1.000000e+01
>  %206 = call float @llvm.AMDIL.clamp.(float %205, float 0.000000e+00, float 1.000000e+00)
>  %207 = fmul float %206, -2.000000e+00
>  %208 = fadd float %207, 3.000000e+00
>  %209 = fmul float %206, %206
>  %210 = fmul float %209, %208
>  %211 = fmul float %208, %209
>  %212 = fadd float %211, 0xBFECCCCCC0000000
>  %213 = fmul float %208, %209
>  %214 = fadd float %213, 0xBF847AE140000000
>  %215 = fmul float %212, 1.000000e+01
>  %216 = fmul float %214, -1.000000e+02
>  %217 = fmul float %210, %186
>  %218 = fadd float %217, %173
>  %219 = fmul float %210, %188
>  %220 = fadd float %219, %174
>  %221 = fmul float %210, %190
>  %222 = fadd float %221, %175
>  %223 = call float @llvm.maxnum.f32(float %216, float 0.000000e+00)
>  %224 = call float @llvm.AMDIL.clamp.(float %215, float 0.000000e+00, float 1.000000e+00)
>  %225 = fmul float %223, -2.000000e+00
>  %226 = fadd float %225, 3.000000e+00
>  %227 = fmul float %223, %223
>  %228 = fmul float %227, %226
>  %229 = fmul float %228, %200
>  %230 = fsub float %184, %229
>  %231 = fmul float %228, %200
>  %232 = fmul float %224, -2.000000e+00
>  %233 = fadd float %232, 3.000000e+00
>  %234 = fmul float %224, %224
>  %235 = fmul float %234, %233
>  %236 = fmul float %235, %230
>  %237 = fadd float %236, %231
>  %238 = call i32 @llvm.SI.packf16(float %218, float %220)
>  %239 = bitcast i32 %238 to float
>  %240 = call i32 @llvm.SI.packf16(float %222, float %237)
>  %241 = bitcast i32 %240 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %239, float %241, float %239, float %241)
>  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.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #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 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, s10                                                            ; BEFC030A
>	v_mov_b32_e32 v2, 0x3f400000                                                 ; 7E0402FF 3F400000
>	v_mov_b32_e32 v3, 0xbda3d70a                                                 ; 7E0602FF BDA3D70A
>	v_mov_b32_e32 v4, 0x40400000                                                 ; 7E0802FF 40400000
>	v_mov_b32_e32 v5, 0x3cf5c28f                                                 ; 7E0A02FF 3CF5C28F
>	v_mov_b32_e32 v6, 0x40c00000                                                 ; 7E0C02FF 40C00000
>	v_mov_b32_e32 v7, 0x41200000                                                 ; 7E0E02FF 41200000
>	v_mov_b32_e32 v8, 0xc2c80000                                                 ; 7E1002FF C2C80000
>	v_interp_p1_f32 v9, v0, 0, 0, [m0]                                           ; C8240000
>	v_interp_p2_f32 v9, [v9], v1, 0, 0, [m0]                                     ; C8250001
>	v_interp_p1_f32 v10, v0, 1, 0, [m0]                                          ; C8280100
>	v_interp_p2_f32 v10, [v10], v1, 1, 0, [m0]                                   ; C8290101
>	v_interp_p1_f32 v11, v0, 0, 1, [m0]                                          ; C82C0400
>	v_interp_p2_f32 v11, [v11], v1, 0, 1, [m0]                                   ; C82D0401
>	v_interp_p1_f32 v12, v0, 1, 1, [m0]                                          ; C8300500
>	v_interp_p2_f32 v12, [v12], v1, 1, 1, [m0]                                   ; C8310501
>	v_interp_p1_f32 v13, v0, 2, 1, [m0]                                          ; C8340600
>	v_interp_p2_f32 v13, [v13], v1, 2, 1, [m0]                                   ; C8350601
>	v_interp_p1_f32 v14, v0, 3, 1, [m0]                                          ; C8380700
>	v_interp_p2_f32 v14, [v14], v1, 3, 1, [m0]                                   ; C8390701
>	s_load_dwordx4 s[44:47], s[2:3], 0x0                                         ; C0960300
>	s_load_dwordx4 s[28:31], s[4:5], 0x0                                         ; C08E0500
>	s_load_dwordx4 s[0:3], s[4:5], 0x4                                           ; C0800504
>	s_load_dwordx4 s[16:19], s[4:5], 0x8                                         ; C0880508
>	s_load_dwordx4 s[52:55], s[4:5], 0xc                                         ; C09A050C
>	s_load_dwordx4 s[40:43], s[4:5], 0x10                                        ; C0940510
>	s_load_dwordx8 s[32:39], s[6:7], 0x0                                         ; C0D00700
>	s_load_dwordx8 s[8:15], s[6:7], 0x8                                          ; C0C40708
>	s_load_dwordx8 s[20:27], s[6:7], 0x10                                        ; C0CA0710
>	s_load_dwordx8 s[56:63], s[6:7], 0x18                                        ; C0DC0718
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	s_buffer_load_dword s4, s[44:47], 0x0                                        ; C2022D00
>	s_buffer_load_dword s5, s[44:47], 0x1                                        ; C202AD01
>	s_buffer_load_dword s64, s[44:47], 0x2                                       ; C2202D02
>	s_buffer_load_dword s65, s[44:47], 0x3                                       ; C220AD03
>	s_buffer_load_dword s48, s[44:47], 0x4                                       ; C2182D04
>	s_buffer_load_dword s49, s[44:47], 0x5                                       ; C218AD05
>	s_buffer_load_dword s50, s[44:47], 0x6                                       ; C2192D06
>	s_buffer_load_dword s66, s[44:47], 0x7                                       ; C2212D07
>	s_buffer_load_dword s67, s[44:47], 0x8                                       ; C221AD08
>	s_buffer_load_dword s68, s[44:47], 0x9                                       ; C2222D09
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mov_b32_e32 v0, s4                                                         ; 7E000204
>	s_buffer_load_dword s69, s[44:47], 0xa                                       ; C222AD0A
>	v_mov_b32_e32 v1, s5                                                         ; 7E020205
>	s_buffer_load_dword s70, s[44:47], 0xb                                       ; C2232D0B
>	v_mov_b32_e32 v15, s64                                                       ; 7E1E0240
>	s_buffer_load_dword s71, s[44:47], 0xe                                       ; C223AD0E
>	s_buffer_load_dword s72, s[44:47], 0xf                                       ; C2242D0F
>	v_sub_f32_e32 v0, s48, v0                                                    ; 08000030
>	v_sub_f32_e32 v1, s49, v1                                                    ; 08020231
>	v_sub_f32_e32 v15, s50, v15                                                  ; 081E1E32
>	s_load_dwordx8 s[44:51], s[6:7], 0x20                                        ; C0D60720
>	image_sample v[16:18], 7, 0, 0, 0, 0, 0, 0, 0, v[9:10], s[56:63], s[52:55]   ; F0800700 01AE1009
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v0, v16, v0, s4                                                    ; D2820000 00120110
>	v_mad_f32 v1, v16, v1, s5                                                    ; D2820001 00160310
>	v_mad_f32 v19, -v17, v0, v0                                                  ; D2820013 24020111
>	v_mac_f32_e32 v19, s67, v17                                                  ; 3E262243
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample v[20:23], 15, 0, 0, 0, 0, 0, 0, 0, v[13:14], s[44:51], s[40:43] ; F0800F00 014B140D
>	image_sample v[24:27], 15, 0, 0, 0, 0, 0, 0, 0, v[9:10], s[32:39], s[28:31]  ; F0800F00 00E81809
>	v_sub_f32_e32 v13, 1.0, v17                                                  ; 081A22F2
>	v_mad_f32 v0, -v0, v13, s65                                                  ; D2820000 21061B00
>	v_mad_f32 v0, -s67, v17, v0                                                  ; D2820000 24022243
>	v_mac_f32_e32 v19, v0, v18                                                   ; 3E262500
>	v_mad_f32 v0, -v17, v1, v1                                                   ; D2820000 24060311
>	v_mac_f32_e32 v0, s68, v17                                                   ; 3E002244
>	v_mad_f32 v1, -v1, v13, s66                                                  ; D2820001 210A1B01
>	v_mad_f32 v1, -s68, v17, v1                                                  ; D2820001 24062244
>	v_mac_f32_e32 v0, v1, v18                                                    ; 3E002501
>	v_mad_f32 v1, v16, v15, s64                                                  ; D2820001 01021F10
>	v_mad_f32 v13, -v1, v13, s70                                                 ; D282000D 211A1B01
>	v_mad_f32 v1, -v17, v1, v1                                                   ; D2820001 24060311
>	v_mac_f32_e32 v1, s69, v17                                                   ; 3E022245
>	v_mad_f32 v13, -s69, v17, v13                                                ; D282000D 24362245
>	v_mac_f32_e32 v1, v13, v18                                                   ; 3E02250D
>	v_mul_f32_e32 v2, s72, v2                                                    ; 10040448
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_sub_f32_e32 v13, 1.0, v24                                                  ; 081A30F2
>	v_mad_f32 v13, -v24, v13, v13                                                ; D282000D 24361B18
>	v_mul_f32_e32 v13, v13, v13                                                  ; 101A1B0D
>	v_madmk_f32_e32 v2, v13, v2, 0x3e800000                                      ; 4004050D 3E800000
>	v_mad_f32 v13, v20, v2, -v2                                                  ; D282000D 840A0514
>	v_mad_f32 v14, v21, v2, -v2                                                  ; D282000E 840A0515
>	v_mad_f32 v15, v22, v2, -v2                                                  ; D282000F 840A0516
>	v_mad_f32 v2, v23, v2, -v2                                                   ; D2820002 840A0517
>	v_mac_f32_e32 v19, v19, v13                                                  ; 3E261B13
>	v_mac_f32_e32 v0, v0, v14                                                    ; 3E001D00
>	v_mac_f32_e32 v1, v1, v15                                                    ; 3E021F01
>	v_mul_f32_e32 v13, 0x3e99999a, v19                                           ; 101A26FF 3E99999A
>	v_madmk_f32_e32 v13, v0, v13, 0x3f170a3d                                     ; 401A1B00 3F170A3D
>	v_madmk_f32_e32 v13, v1, v13, 0x3de147ae                                     ; 401A1B01 3DE147AE
>	v_add_f32_e32 v3, v13, v3                                                    ; 0606070D
>	v_mul_f32_e32 v3, 0x41649249, v3                                             ; 100606FF 41649249
>	v_add_f32_e64 v3, 0, v3 clamp                                                ; D2060803 00020680
>	v_mad_f32 v13, -2.0, v3, v4                                                  ; D282000D 041206F5
>	v_mul_f32_e32 v3, v3, v3                                                     ; 10060703
>	v_mad_f32 v3, -v3, v13, 1.0                                                  ; D2820003 23CA1B03
>	v_mad_f32 v13, v5, v3, v19                                                   ; D282000D 044E0705
>	v_mad_f32 v14, v5, v3, v0                                                    ; D282000E 04020705
>	v_mad_f32 v3, v5, v3, v1                                                     ; D2820003 04060705
>	v_mul_f32_e32 v5, v26, v13                                                   ; 100A1B1A
>	v_mul_f32_e32 v13, v26, v14                                                  ; 101A1D1A
>	v_mul_f32_e32 v3, v26, v3                                                    ; 1006071A
>	v_mac_f32_e32 v19, 0.5, v5                                                   ; 3E260AF0
>	v_mac_f32_e32 v0, 0.5, v13                                                   ; 3E001AF0
>	v_mac_f32_e32 v1, 0.5, v3                                                    ; 3E0206F0
>	v_add_f32_e64 v3, 0, v19 clamp                                               ; D2060803 00022680
>	v_add_f32_e64 v0, 0, v0 clamp                                                ; D2060800 00020080
>	v_add_f32_e64 v1, 0, v1 clamp                                                ; D2060801 00020280
>	v_mul_f32_e32 v5, v3, v25                                                    ; 100A3303
>	v_mul_f32_e32 v13, v0, v25                                                   ; 101A3300
>	image_sample v[14:17], 15, 0, 0, 0, 0, 0, 0, 0, v[9:10], s[20:27], s[16:19]  ; F0800F00 00850E09
>	image_sample v9, 2, 0, 0, 0, 0, 0, 0, 0, v[11:12], s[8:15], s[0:3]           ; F0800200 0002090B
>	v_mul_f32_e32 v10, v1, v25                                                   ; 10143301
>	s_waitcnt vmcnt(1)                                                           ; BF8C0771
>	v_mad_f32 v3, -v25, v3, v14                                                  ; D2820003 243A0719
>	v_mad_f32 v0, -v25, v0, v15                                                  ; D2820000 243E0119
>	v_mad_f32 v1, -v25, v1, v16                                                  ; D2820001 24420319
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mac_f32_e32 v27, v24, v9                                                   ; 3E361318
>	v_mul_f32_e32 v9, s71, v25                                                   ; 10123247
>	v_mac_f32_e32 v9, v9, v2                                                     ; 3E120509
>	v_mad_f32 v2, v6, s72, 1.0                                                   ; D2820002 03C89106
>	v_madak_f32_e32 v2, v27, v2, 0xbf147ae1                                      ; 4204051B BF147AE1
>	v_mul_f32_e32 v2, v7, v2                                                     ; 10040507
>	v_add_f32_e64 v2, 0, v2 clamp                                                ; D2060802 00020480
>	v_mad_f32 v6, -2.0, v2, v4                                                   ; D2820006 041204F5
>	v_mul_f32_e32 v2, v2, v2                                                     ; 10040502
>	v_mul_f32_e32 v11, v6, v2                                                    ; 10160506
>	v_madak_f32_e32 v12, v2, v6, 0xbf666666                                      ; 42180D02 BF666666
>	v_madak_f32_e32 v2, v2, v6, 0xbc23d70a                                       ; 42040D02 BC23D70A
>	v_mul_f32_e32 v6, v7, v12                                                    ; 100C1907
>	v_mul_f32_e32 v2, v2, v8                                                     ; 10041102
>	v_mac_f32_e32 v5, v3, v11                                                    ; 3E0A1703
>	v_mac_f32_e32 v13, v0, v11                                                   ; 3E1A1700
>	v_mac_f32_e32 v10, v1, v11                                                   ; 3E141701
>	v_max_f32_e32 v0, 0, v2                                                      ; 20000480
>	v_add_f32_e64 v1, 0, v6 clamp                                                ; D2060801 00020C80
>	v_mad_f32 v2, -2.0, v0, v4                                                   ; D2820002 041200F5
>	v_mul_f32_e32 v0, v0, v0                                                     ; 10000100
>	v_mul_f32_e32 v0, v2, v0                                                     ; 10000102
>	v_mul_f32_e32 v2, v9, v0                                                     ; 10040109
>	v_mad_f32 v0, -v0, v9, v17                                                   ; D2820000 24461300
>	v_mac_f32_e32 v4, -2.0, v1                                                   ; 3E0802F5
>	v_mul_f32_e32 v1, v1, v1                                                     ; 10020301
>	v_mul_f32_e32 v1, v4, v1                                                     ; 10020304
>	v_mac_f32_e32 v2, v0, v1                                                     ; 3E040300
>	v_cvt_pkrtz_f16_f32_e32 v0, v5, v13                                          ; 5E001B05
>	v_cvt_pkrtz_f16_f32_e32 v1, v10, v2                                          ; 5E02050A
>	exp 15, 0, 1, 1, 1, v0, v1, v0, v1                                           ; F8001C0F 01000100
>	s_endpgm                                                                     ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 80
>VGPRS: 28
>Code Size: 840 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
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL OUT[1], CLIPVERTEX
>DCL OUT[2], GENERIC[0]
>DCL OUT[3], GENERIC[1]
>DCL CONST[0..54]
>DCL TEMP[0..3], LOCAL
>  0: MUL TEMP[0].xy, CONST[48].xyyy, IN[1].xyyy
>  1: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
>  2: ADD TEMP[1].x, TEMP[0].xxxx, CONST[48].wwww
>  3: MOV TEMP[1].z, TEMP[1].xxxx
>  4: MUL TEMP[0].xy, CONST[49].xyyy, IN[1].xyyy
>  5: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
>  6: ADD TEMP[2].x, TEMP[0].xxxx, CONST[49].wwww
>  7: MOV TEMP[1].w, TEMP[2].xxxx
>  8: MUL TEMP[0].xy, CONST[50].xyyy, IN[1].xyyy
>  9: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 10: ADD TEMP[2].x, TEMP[0].xxxx, CONST[50].wwww
> 11: MUL TEMP[0].xy, CONST[51].xyyy, IN[1].xyyy
> 12: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 13: ADD TEMP[3].x, TEMP[0].xxxx, CONST[51].wwww
> 14: MOV TEMP[2].y, TEMP[3].xxxx
> 15: MUL TEMP[0].xy, CONST[52].xyyy, IN[1].xyyy
> 16: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 17: ADD TEMP[3].x, TEMP[0].xxxx, CONST[52].wwww
> 18: MOV TEMP[2].z, TEMP[3].xxxx
> 19: MUL TEMP[0].xy, CONST[53].xyyy, IN[1].xyyy
> 20: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 21: ADD TEMP[0].x, TEMP[0].xxxx, CONST[53].wwww
> 22: MOV TEMP[2].w, TEMP[0].xxxx
> 23: MOV TEMP[0].xw, IN[0].xxxw
> 24: MOV TEMP[1].xy, IN[1].xyxx
> 25: MAD TEMP[3].x, IN[0].zzzz, CONST[0].zzzz, -IN[0].wwww
> 26: MOV TEMP[0].z, TEMP[3].xxxx
> 27: MOV TEMP[0].y, -IN[0].yyyy
> 28: MAD TEMP[0].xy, CONST[54].xyyy, IN[0].wwww, TEMP[0].xyyy
> 29: MOV OUT[2], TEMP[1]
> 30: MOV OUT[3], TEMP[2]
> 31: MOV OUT[0], TEMP[0]
> 32: MOV OUT[1], IN[0]
> 33: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = call float @llvm.SI.load.const(<16 x i8> %13, i32 8)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %13, i32 768)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %13, i32 772)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %13, i32 780)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %13, i32 784)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %13, i32 788)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %13, i32 796)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %13, i32 800)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %13, i32 804)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %13, i32 812)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %13, i32 816)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %13, i32 820)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %13, i32 828)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %13, i32 832)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %13, i32 836)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %13, i32 844)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %13, i32 848)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %13, i32 852)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %13, i32 860)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %13, i32 864)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %13, i32 868)
>  %35 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %36 = load <16 x i8>, <16 x i8> addrspace(2)* %35, align 16, !tbaa !0
>  %37 = add i32 %5, %8
>  %38 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %36, i32 0, i32 %37)
>  %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 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 1
>  %44 = load <16 x i8>, <16 x i8> addrspace(2)* %43, align 16, !tbaa !0
>  %45 = add i32 %5, %8
>  %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 = fmul float %15, %47
>  %50 = fmul float %16, %48
>  %51 = fadd float %50, %49
>  %52 = fadd float %51, %17
>  %53 = fmul float %18, %47
>  %54 = fmul float %19, %48
>  %55 = fadd float %54, %53
>  %56 = fadd float %55, %20
>  %57 = fmul float %21, %47
>  %58 = fmul float %22, %48
>  %59 = fadd float %58, %57
>  %60 = fadd float %59, %23
>  %61 = fmul float %24, %47
>  %62 = fmul float %25, %48
>  %63 = fadd float %62, %61
>  %64 = fadd float %63, %26
>  %65 = fmul float %27, %47
>  %66 = fmul float %28, %48
>  %67 = fadd float %66, %65
>  %68 = fadd float %67, %29
>  %69 = fmul float %30, %47
>  %70 = fmul float %31, %48
>  %71 = fadd float %70, %69
>  %72 = fadd float %71, %32
>  %73 = fmul float %41, %14
>  %74 = fsub float %73, %42
>  %75 = fmul float %33, %42
>  %76 = fadd float %75, %39
>  %77 = fmul float %34, %42
>  %78 = fsub float %77, %40
>  %79 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 16
>  %80 = load <16 x i8>, <16 x i8> addrspace(2)* %79, align 16, !tbaa !0
>  %81 = call float @llvm.SI.load.const(<16 x i8> %80, i32 0)
>  %82 = fmul float %81, %39
>  %83 = call float @llvm.SI.load.const(<16 x i8> %80, i32 4)
>  %84 = fmul float %83, %40
>  %85 = fadd float %82, %84
>  %86 = call float @llvm.SI.load.const(<16 x i8> %80, i32 8)
>  %87 = fmul float %86, %41
>  %88 = fadd float %85, %87
>  %89 = call float @llvm.SI.load.const(<16 x i8> %80, i32 12)
>  %90 = fmul float %89, %42
>  %91 = fadd float %88, %90
>  %92 = call float @llvm.SI.load.const(<16 x i8> %80, i32 16)
>  %93 = fmul float %92, %39
>  %94 = call float @llvm.SI.load.const(<16 x i8> %80, i32 20)
>  %95 = fmul float %94, %40
>  %96 = fadd float %93, %95
>  %97 = call float @llvm.SI.load.const(<16 x i8> %80, i32 24)
>  %98 = fmul float %97, %41
>  %99 = fadd float %96, %98
>  %100 = call float @llvm.SI.load.const(<16 x i8> %80, i32 28)
>  %101 = fmul float %100, %42
>  %102 = fadd float %99, %101
>  %103 = call float @llvm.SI.load.const(<16 x i8> %80, i32 32)
>  %104 = fmul float %103, %39
>  %105 = call float @llvm.SI.load.const(<16 x i8> %80, i32 36)
>  %106 = fmul float %105, %40
>  %107 = fadd float %104, %106
>  %108 = call float @llvm.SI.load.const(<16 x i8> %80, i32 40)
>  %109 = fmul float %108, %41
>  %110 = fadd float %107, %109
>  %111 = call float @llvm.SI.load.const(<16 x i8> %80, i32 44)
>  %112 = fmul float %111, %42
>  %113 = fadd float %110, %112
>  %114 = call float @llvm.SI.load.const(<16 x i8> %80, i32 48)
>  %115 = fmul float %114, %39
>  %116 = call float @llvm.SI.load.const(<16 x i8> %80, i32 52)
>  %117 = fmul float %116, %40
>  %118 = fadd float %115, %117
>  %119 = call float @llvm.SI.load.const(<16 x i8> %80, i32 56)
>  %120 = fmul float %119, %41
>  %121 = fadd float %118, %120
>  %122 = call float @llvm.SI.load.const(<16 x i8> %80, i32 60)
>  %123 = fmul float %122, %42
>  %124 = fadd float %121, %123
>  %125 = call float @llvm.SI.load.const(<16 x i8> %80, i32 64)
>  %126 = fmul float %125, %39
>  %127 = call float @llvm.SI.load.const(<16 x i8> %80, i32 68)
>  %128 = fmul float %127, %40
>  %129 = fadd float %126, %128
>  %130 = call float @llvm.SI.load.const(<16 x i8> %80, i32 72)
>  %131 = fmul float %130, %41
>  %132 = fadd float %129, %131
>  %133 = call float @llvm.SI.load.const(<16 x i8> %80, i32 76)
>  %134 = fmul float %133, %42
>  %135 = fadd float %132, %134
>  %136 = call float @llvm.SI.load.const(<16 x i8> %80, i32 80)
>  %137 = fmul float %136, %39
>  %138 = call float @llvm.SI.load.const(<16 x i8> %80, i32 84)
>  %139 = fmul float %138, %40
>  %140 = fadd float %137, %139
>  %141 = call float @llvm.SI.load.const(<16 x i8> %80, i32 88)
>  %142 = fmul float %141, %41
>  %143 = fadd float %140, %142
>  %144 = call float @llvm.SI.load.const(<16 x i8> %80, i32 92)
>  %145 = fmul float %144, %42
>  %146 = fadd float %143, %145
>  %147 = call float @llvm.SI.load.const(<16 x i8> %80, i32 96)
>  %148 = fmul float %147, %39
>  %149 = call float @llvm.SI.load.const(<16 x i8> %80, i32 100)
>  %150 = fmul float %149, %40
>  %151 = fadd float %148, %150
>  %152 = call float @llvm.SI.load.const(<16 x i8> %80, i32 104)
>  %153 = fmul float %152, %41
>  %154 = fadd float %151, %153
>  %155 = call float @llvm.SI.load.const(<16 x i8> %80, i32 108)
>  %156 = fmul float %155, %42
>  %157 = fadd float %154, %156
>  %158 = call float @llvm.SI.load.const(<16 x i8> %80, i32 112)
>  %159 = fmul float %158, %39
>  %160 = call float @llvm.SI.load.const(<16 x i8> %80, i32 116)
>  %161 = fmul float %160, %40
>  %162 = fadd float %159, %161
>  %163 = call float @llvm.SI.load.const(<16 x i8> %80, i32 120)
>  %164 = fmul float %163, %41
>  %165 = fadd float %162, %164
>  %166 = call float @llvm.SI.load.const(<16 x i8> %80, i32 124)
>  %167 = fmul float %166, %42
>  %168 = fadd float %165, %167
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %47, float %48, float %52, float %56)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %60, float %64, float %68, float %72)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 12, i32 0, float %76, float %78, float %74, float %42)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 13, i32 0, float %91, float %102, float %113, float %124)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 14, i32 0, float %135, float %146, float %157, 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
>
>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[4:7], s[8:9], 0x0                    ; C0820900
>	s_load_dwordx4 s[12:15], s[8:9], 0x4                  ; C0860904
>	v_add_i32_e32 v0, vcc, s10, v0                        ; 4A00000A
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[4:7], 0 idxen   ; E00C2000 80010100
>	buffer_load_format_xyzw v[5:8], v0, s[12:15], 0 idxen ; E00C2000 80030500
>	s_load_dwordx4 s[4:7], s[2:3], 0x0                    ; C0820300
>	s_load_dwordx4 s[0:3], s[2:3], 0x40                   ; C0800340
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	s_buffer_load_dword s8, s[4:7], 0x2                   ; C2040502
>	s_buffer_load_dword s9, s[4:7], 0xc0                  ; C20485C0
>	s_buffer_load_dword s10, s[4:7], 0xc1                 ; C20505C1
>	s_buffer_load_dword s11, s[4:7], 0xc3                 ; C20585C3
>	s_buffer_load_dword s12, s[4:7], 0xc4                 ; C20605C4
>	s_buffer_load_dword s13, s[4:7], 0xc5                 ; C20685C5
>	s_buffer_load_dword s14, s[4:7], 0xc7                 ; C20705C7
>	s_buffer_load_dword s15, s[4:7], 0xc8                 ; C20785C8
>	s_buffer_load_dword s16, s[4:7], 0xc9                 ; C20805C9
>	s_buffer_load_dword s17, s[4:7], 0xcb                 ; C20885CB
>	s_buffer_load_dword s18, s[4:7], 0xcc                 ; C20905CC
>	s_buffer_load_dword s19, s[4:7], 0xcd                 ; C20985CD
>	s_buffer_load_dword s20, s[4:7], 0xcf                 ; C20A05CF
>	s_buffer_load_dword s21, s[4:7], 0xd0                 ; C20A85D0
>	s_buffer_load_dword s22, s[4:7], 0xd1                 ; C20B05D1
>	s_buffer_load_dword s23, s[4:7], 0xd3                 ; C20B85D3
>	s_buffer_load_dword s24, s[4:7], 0xd4                 ; C20C05D4
>	s_buffer_load_dword s25, s[4:7], 0xd5                 ; C20C85D5
>	s_buffer_load_dword s26, s[4:7], 0xd7                 ; C20D05D7
>	s_buffer_load_dword s27, s[4:7], 0xd8                 ; C20D85D8
>	s_buffer_load_dword s4, s[4:7], 0xd9                  ; C20205D9
>	s_buffer_load_dword s5, s[0:3], 0x0                   ; C2028100
>	s_buffer_load_dword s6, s[0:3], 0x1                   ; C2030101
>	s_buffer_load_dword s7, s[0:3], 0x2                   ; C2038102
>	s_buffer_load_dword s28, s[0:3], 0x3                  ; C20E0103
>	s_buffer_load_dword s29, s[0:3], 0x4                  ; C20E8104
>	s_buffer_load_dword s30, s[0:3], 0x5                  ; C20F0105
>	s_buffer_load_dword s31, s[0:3], 0x6                  ; C20F8106
>	s_buffer_load_dword s32, s[0:3], 0x7                  ; C2100107
>	s_buffer_load_dword s33, s[0:3], 0x8                  ; C2108108
>	s_buffer_load_dword s34, s[0:3], 0x9                  ; C2110109
>	s_buffer_load_dword s35, s[0:3], 0xa                  ; C211810A
>	s_buffer_load_dword s36, s[0:3], 0xb                  ; C212010B
>	s_buffer_load_dword s37, s[0:3], 0xc                  ; C212810C
>	s_buffer_load_dword s38, s[0:3], 0xd                  ; C213010D
>	s_buffer_load_dword s39, s[0:3], 0xe                  ; C213810E
>	s_buffer_load_dword s40, s[0:3], 0xf                  ; C214010F
>	s_buffer_load_dword s41, s[0:3], 0x10                 ; C2148110
>	s_buffer_load_dword s42, s[0:3], 0x11                 ; C2150111
>	s_buffer_load_dword s43, s[0:3], 0x12                 ; C2158112
>	s_buffer_load_dword s44, s[0:3], 0x13                 ; C2160113
>	s_buffer_load_dword s45, s[0:3], 0x14                 ; C2168114
>	s_buffer_load_dword s46, s[0:3], 0x15                 ; C2170115
>	s_buffer_load_dword s47, s[0:3], 0x16                 ; C2178116
>	s_buffer_load_dword s48, s[0:3], 0x17                 ; C2180117
>	s_buffer_load_dword s49, s[0:3], 0x18                 ; C2188118
>	s_buffer_load_dword s50, s[0:3], 0x19                 ; C2190119
>	s_buffer_load_dword s51, s[0:3], 0x1a                 ; C219811A
>	s_buffer_load_dword s52, s[0:3], 0x1b                 ; C21A011B
>	s_buffer_load_dword s53, s[0:3], 0x1c                 ; C21A811C
>	s_buffer_load_dword s54, s[0:3], 0x1d                 ; C21B011D
>	s_buffer_load_dword s55, s[0:3], 0x1e                 ; C21B811E
>	s_buffer_load_dword s0, s[0:3], 0x1f                  ; C200011F
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mov_b32_e32 v0, s11                                 ; 7E00020B
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mov_b32_e32 v7, s14                                 ; 7E0E020E
>	v_mov_b32_e32 v8, s17                                 ; 7E100211
>	v_mov_b32_e32 v9, s20                                 ; 7E120214
>	v_mov_b32_e32 v10, s23                                ; 7E140217
>	v_mov_b32_e32 v11, s26                                ; 7E16021A
>	v_mac_f32_e32 v0, s9, v5                              ; 3E000A09
>	v_mac_f32_e32 v7, s12, v5                             ; 3E0E0A0C
>	v_mac_f32_e32 v8, s15, v5                             ; 3E100A0F
>	v_mac_f32_e32 v9, s18, v5                             ; 3E120A12
>	v_mac_f32_e32 v10, s21, v5                            ; 3E140A15
>	v_mac_f32_e32 v11, s24, v5                            ; 3E160A18
>	v_mac_f32_e32 v0, s10, v6                             ; 3E000C0A
>	v_mac_f32_e32 v7, s13, v6                             ; 3E0E0C0D
>	v_mac_f32_e32 v8, s16, v6                             ; 3E100C10
>	v_mac_f32_e32 v9, s19, v6                             ; 3E120C13
>	exp 15, 32, 0, 0, 0, v5, v6, v0, v7                   ; F800020F 07000605
>	v_mac_f32_e32 v10, s22, v6                            ; 3E140C16
>	v_mac_f32_e32 v11, s25, v6                            ; 3E160C19
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mad_f32 v0, v3, s8, -v4                             ; D2820000 84101103
>	v_mad_f32 v5, s27, v4, v1                             ; D2820005 0406081B
>	v_mad_f32 v6, s4, v4, -v2                             ; D2820006 840A0804
>	v_mul_f32_e32 v7, s6, v2                              ; 100E0406
>	exp 15, 33, 0, 0, 0, v8, v9, v10, v11                 ; F800021F 0B0A0908
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mul_f32_e32 v8, s30, v2                             ; 1010041E
>	v_mac_f32_e32 v7, s5, v1                              ; 3E0E0205
>	exp 15, 12, 0, 0, 0, v5, v6, v0, v4                   ; F80000CF 04000605
>	v_mac_f32_e32 v8, s29, v1                             ; 3E10021D
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mul_f32_e32 v0, s34, v2                             ; 10000422
>	v_mac_f32_e32 v0, s33, v1                             ; 3E000221
>	v_mul_f32_e32 v5, s38, v2                             ; 100A0426
>	v_mac_f32_e32 v5, s37, v1                             ; 3E0A0225
>	v_mul_f32_e32 v6, s42, v2                             ; 100C042A
>	v_mul_f32_e32 v9, s46, v2                             ; 1012042E
>	v_mul_f32_e32 v10, s50, v2                            ; 10140432
>	v_mul_f32_e32 v2, s54, v2                             ; 10040436
>	v_mac_f32_e32 v6, s41, v1                             ; 3E0C0229
>	v_mac_f32_e32 v9, s45, v1                             ; 3E12022D
>	v_mac_f32_e32 v10, s49, v1                            ; 3E140231
>	v_mac_f32_e32 v2, s53, v1                             ; 3E040235
>	v_mac_f32_e32 v7, s7, v3                              ; 3E0E0607
>	v_mac_f32_e32 v8, s31, v3                             ; 3E10061F
>	v_mac_f32_e32 v0, s35, v3                             ; 3E000623
>	v_mac_f32_e32 v5, s39, v3                             ; 3E0A0627
>	v_mac_f32_e32 v6, s43, v3                             ; 3E0C062B
>	v_mac_f32_e32 v9, s47, v3                             ; 3E12062F
>	v_mac_f32_e32 v10, s51, v3                            ; 3E140633
>	v_mac_f32_e32 v2, s55, v3                             ; 3E040637
>	v_mac_f32_e32 v7, s28, v4                             ; 3E0E081C
>	v_mac_f32_e32 v8, s32, v4                             ; 3E100820
>	v_mac_f32_e32 v0, s36, v4                             ; 3E000824
>	v_mac_f32_e32 v5, s40, v4                             ; 3E0A0828
>	v_mac_f32_e32 v6, s44, v4                             ; 3E0C082C
>	v_mac_f32_e32 v9, s48, v4                             ; 3E120830
>	v_mac_f32_e32 v10, s52, v4                            ; 3E140834
>	v_mac_f32_e32 v2, s0, v4                              ; 3E040800
>	exp 15, 13, 0, 0, 0, v7, v8, v0, v5                   ; F80000DF 05000807
>	exp 15, 14, 0, 1, 0, v6, v9, v10, v2                  ; F80008EF 020A0906
>	s_endpgm                                              ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 64
>VGPRS: 12
>Code Size: 544 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
>  clamp_color = 0
>FRAG
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL IN[1], GENERIC[1], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SAMP[1]
>DCL SAMP[2]
>DCL SAMP[3]
>DCL SAMP[4]
>DCL SVIEW[0], 2D, FLOAT
>DCL SVIEW[1], 2D, FLOAT
>DCL SVIEW[2], 2D, FLOAT
>DCL SVIEW[3], 2D, FLOAT
>DCL SVIEW[4], 2D, FLOAT
>DCL CONST[0..3]
>DCL TEMP[0..4], LOCAL
>IMM[0] FLT32 {   -0.5000,    -1.0000,    10.0000,   -10.0000}
>IMM[1] FLT32 {   -2.0000,     3.0000,     6.0000,     1.0000}
>IMM[2] FLT32 {   -0.5800,    -1.0000,    -0.0000,     1.0000}
>IMM[3] FLT32 {    1.0000,     0.0000,     0.0000,     0.0000}
>  0: MOV TEMP[0].xy, IN[0].xyyy
>  1: TEX TEMP[0].xw, TEMP[0], SAMP[0], 2D
>  2: MOV TEMP[1].xy, IN[1].xyyy
>  3: TEX TEMP[1].y, TEMP[1], SAMP[1], 2D
>  4: MAD TEMP[0].x, TEMP[1].yyyy, TEMP[0].xxxx, TEMP[0].wwww
>  5: MOV TEMP[1].xy, IN[0].zwww
>  6: TEX TEMP[1].w, TEMP[1], SAMP[4], 2D
>  7: ADD TEMP[2].xy, TEMP[1].wwww, IMM[0].xyyy
>  8: ADD TEMP[3].x, TEMP[1].wwww, TEMP[1].wwww
>  9: MOV_SAT TEMP[3].x, TEMP[3].xxxx
> 10: MUL TEMP[2].xy, TEMP[2].xyyy, IMM[0].zwww
> 11: MOV_SAT TEMP[2].xy, TEMP[2].xyyy
> 12: MAD TEMP[1].xy, TEMP[2].xyyy, IMM[1].xxxx, IMM[1].yyyy
> 13: MUL TEMP[2].xy, TEMP[2].xyyy, TEMP[2].xyyy
> 14: MUL TEMP[2].xy, TEMP[2].xyyy, TEMP[1].xyyy
> 15: MUL TEMP[2].x, TEMP[2].yyyy, TEMP[2].xxxx
> 16: MAD TEMP[4].x, CONST[3].wwww, IMM[1].zzzz, IMM[1].wwww
> 17: MAD TEMP[0].x, TEMP[0].xxxx, TEMP[4].xxxx, TEMP[2].xxxx
> 18: MAD TEMP[2].x, TEMP[3].xxxx, IMM[1].xxxx, IMM[1].yyyy
> 19: MUL TEMP[3].x, TEMP[3].xxxx, TEMP[3].xxxx
> 20: MUL TEMP[2].x, TEMP[3].xxxx, TEMP[2].xxxx
> 21: MOV TEMP[3].xy, IN[0].xyyy
> 22: TEX TEMP[3].yz, TEMP[3], SAMP[3], 2D
> 23: ADD TEMP[3].x, TEMP[3].zzzz, TEMP[3].yyyy
> 24: MOV_SAT TEMP[3].x, TEMP[3].xxxx
> 25: ADD TEMP[3].x, -TEMP[3].xxxx, IMM[1].wwww
> 26: ADD TEMP[3].x, -TEMP[3].xxxx, IMM[1].wwww
> 27: MAX TEMP[2].x, TEMP[3].xxxx, TEMP[2].xxxx
> 28: MAD TEMP[0].x, TEMP[0].xxxx, TEMP[2].xxxx, IMM[2].xxxx
> 29: MUL TEMP[0].x, TEMP[0].xxxx, IMM[0].zzzz
> 30: MOV_SAT TEMP[2].x, TEMP[0].xxxx
> 31: MAD TEMP[3].x, TEMP[2].xxxx, IMM[1].xxxx, IMM[1].yyyy
> 32: MUL TEMP[0].x, TEMP[2].xxxx, TEMP[2].xxxx
> 33: MUL TEMP[0].x, TEMP[0].xxxx, TEMP[3].xxxx
> 34: MOV TEMP[2].xy, IN[0].xyyy
> 35: TEX TEMP[2].xyz, TEMP[2], SAMP[2], 2D
> 36: MAD TEMP[2].xyz, CONST[3].yyyy, IMM[2].yzzz, TEMP[2].xyzz
> 37: MUL TEMP[1].xyz, IMM[3].xyyy, CONST[3].yyyy
> 38: MAD TEMP[0].xyz, TEMP[0].xxxx, TEMP[2].xyzz, TEMP[1].xyzz
> 39: MOV TEMP[0].w, IMM[1].wwww
> 40: MOV OUT[0], TEMP[0]
> 41: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %24 = load <16 x i8>, <16 x i8> addrspace(2)* %23, align 16, !tbaa !0
>  %25 = call float @llvm.SI.load.const(<16 x i8> %24, i32 52)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %24, i32 60)
>  %27 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %28 = load <8 x i32>, <8 x i32> addrspace(2)* %27, align 32, !tbaa !0
>  %29 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %30 = load <4 x i32>, <4 x i32> addrspace(2)* %29, align 16, !tbaa !0
>  %31 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 1
>  %32 = load <8 x i32>, <8 x i32> addrspace(2)* %31, align 32, !tbaa !0
>  %33 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 1
>  %34 = load <4 x i32>, <4 x i32> addrspace(2)* %33, align 16, !tbaa !0
>  %35 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 2
>  %36 = load <8 x i32>, <8 x i32> addrspace(2)* %35, align 32, !tbaa !0
>  %37 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 2
>  %38 = load <4 x i32>, <4 x i32> addrspace(2)* %37, align 16, !tbaa !0
>  %39 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 3
>  %40 = load <8 x i32>, <8 x i32> addrspace(2)* %39, align 32, !tbaa !0
>  %41 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 3
>  %42 = load <4 x i32>, <4 x i32> addrspace(2)* %41, align 16, !tbaa !0
>  %43 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 4
>  %44 = load <8 x i32>, <8 x i32> addrspace(2)* %43, align 32, !tbaa !0
>  %45 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 4
>  %46 = load <4 x i32>, <4 x i32> addrspace(2)* %45, align 16, !tbaa !0
>  %47 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %8)
>  %48 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %8)
>  %49 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %6, <2 x i32> %8)
>  %50 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %6, <2 x i32> %8)
>  %51 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %6, <2 x i32> %8)
>  %52 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %6, <2 x i32> %8)
>  %53 = bitcast float %47 to i32
>  %54 = bitcast float %48 to i32
>  %55 = insertelement <2 x i32> undef, i32 %53, i32 0
>  %56 = insertelement <2 x i32> %55, i32 %54, i32 1
>  %57 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %56, <8 x i32> %28, <4 x i32> %30, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %58 = extractelement <4 x float> %57, i32 0
>  %59 = extractelement <4 x float> %57, i32 3
>  %60 = bitcast float %51 to i32
>  %61 = bitcast float %52 to i32
>  %62 = insertelement <2 x i32> undef, i32 %60, i32 0
>  %63 = insertelement <2 x i32> %62, i32 %61, i32 1
>  %64 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %63, <8 x i32> %32, <4 x i32> %34, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %65 = extractelement <4 x float> %64, i32 1
>  %66 = fmul float %65, %58
>  %67 = fadd float %66, %59
>  %68 = bitcast float %49 to i32
>  %69 = bitcast float %50 to i32
>  %70 = insertelement <2 x i32> undef, i32 %68, i32 0
>  %71 = insertelement <2 x i32> %70, i32 %69, i32 1
>  %72 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %71, <8 x i32> %44, <4 x i32> %46, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %73 = extractelement <4 x float> %72, i32 3
>  %74 = fadd float %73, -5.000000e-01
>  %75 = fadd float %73, -1.000000e+00
>  %76 = fadd float %73, %73
>  %77 = call float @llvm.AMDIL.clamp.(float %76, float 0.000000e+00, float 1.000000e+00)
>  %78 = fmul float %74, 1.000000e+01
>  %79 = fmul float %75, -1.000000e+01
>  %80 = call float @llvm.AMDIL.clamp.(float %78, float 0.000000e+00, float 1.000000e+00)
>  %81 = call float @llvm.AMDIL.clamp.(float %79, float 0.000000e+00, float 1.000000e+00)
>  %82 = fmul float %80, -2.000000e+00
>  %83 = fadd float %82, 3.000000e+00
>  %84 = fmul float %81, -2.000000e+00
>  %85 = fadd float %84, 3.000000e+00
>  %86 = fmul float %80, %80
>  %87 = fmul float %81, %81
>  %88 = fmul float %86, %83
>  %89 = fmul float %87, %85
>  %90 = fmul float %89, %88
>  %91 = fmul float %26, 6.000000e+00
>  %92 = fadd float %91, 1.000000e+00
>  %93 = fmul float %67, %92
>  %94 = fadd float %93, %90
>  %95 = fmul float %77, -2.000000e+00
>  %96 = fadd float %95, 3.000000e+00
>  %97 = fmul float %77, %77
>  %98 = fmul float %97, %96
>  %99 = bitcast float %47 to i32
>  %100 = bitcast float %48 to i32
>  %101 = insertelement <2 x i32> undef, i32 %99, i32 0
>  %102 = insertelement <2 x i32> %101, i32 %100, i32 1
>  %103 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %102, <8 x i32> %40, <4 x i32> %42, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %104 = extractelement <4 x float> %103, i32 1
>  %105 = extractelement <4 x float> %103, i32 2
>  %106 = fadd float %105, %104
>  %107 = call float @llvm.AMDIL.clamp.(float %106, float 0.000000e+00, float 1.000000e+00)
>  %108 = fsub float 1.000000e+00, %107
>  %109 = fsub float 1.000000e+00, %108
>  %110 = call float @llvm.maxnum.f32(float %109, float %98)
>  %111 = fmul float %94, %110
>  %112 = fadd float %111, 0xBFE28F5C20000000
>  %113 = fmul float %112, 1.000000e+01
>  %114 = call float @llvm.AMDIL.clamp.(float %113, float 0.000000e+00, float 1.000000e+00)
>  %115 = fmul float %114, -2.000000e+00
>  %116 = fadd float %115, 3.000000e+00
>  %117 = fmul float %114, %114
>  %118 = fmul float %117, %116
>  %119 = bitcast float %47 to i32
>  %120 = bitcast float %48 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.image.sample.v2i32(<2 x i32> %122, <8 x i32> %36, <4 x i32> %38, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %124 = extractelement <4 x float> %123, i32 0
>  %125 = extractelement <4 x float> %123, i32 1
>  %126 = extractelement <4 x float> %123, i32 2
>  %127 = fsub float %124, %25
>  %128 = fmul float %25, -0.000000e+00
>  %129 = fadd float %128, %125
>  %130 = fmul float %25, -0.000000e+00
>  %131 = fadd float %130, %126
>  %132 = fmul float %25, 0.000000e+00
>  %133 = fmul float %25, 0.000000e+00
>  %134 = fmul float %118, %127
>  %135 = fadd float %134, %25
>  %136 = fmul float %118, %129
>  %137 = fadd float %136, %132
>  %138 = fmul float %118, %131
>  %139 = fadd float %138, %133
>  %140 = call i32 @llvm.SI.packf16(float %135, float %137)
>  %141 = bitcast i32 %140 to float
>  %142 = call i32 @llvm.SI.packf16(float %139, float 1.000000e+00)
>  %143 = bitcast i32 %142 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %141, float %143, float %141, float %143)
>  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.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #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 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, s10                                                         ; BEFC030A
>	v_mov_b32_e32 v2, 0x41200000                                              ; 7E0402FF 41200000
>	v_mov_b32_e32 v3, 0x40400000                                              ; 7E0602FF 40400000
>	v_mov_b32_e32 v4, 0x40c00000                                              ; 7E0802FF 40C00000
>	v_mov_b32_e32 v5, 0x80000000                                              ; 7E0A02FF 80000000
>	v_interp_p1_f32 v6, v0, 0, 0, [m0]                                        ; C8180000
>	v_interp_p2_f32 v6, [v6], v1, 0, 0, [m0]                                  ; C8190001
>	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 v8, v0, 2, 0, [m0]                                        ; C8200200
>	v_interp_p2_f32 v8, [v8], v1, 2, 0, [m0]                                  ; C8210201
>	v_interp_p1_f32 v9, v0, 3, 0, [m0]                                        ; C8240300
>	v_interp_p2_f32 v9, [v9], v1, 3, 0, [m0]                                  ; C8250301
>	v_interp_p1_f32 v10, v0, 0, 1, [m0]                                       ; C8280400
>	v_interp_p2_f32 v10, [v10], v1, 0, 1, [m0]                                ; C8290401
>	v_interp_p1_f32 v11, v0, 1, 1, [m0]                                       ; C82C0500
>	v_interp_p2_f32 v11, [v11], v1, 1, 1, [m0]                                ; C82D0501
>	s_load_dwordx4 s[28:31], s[2:3], 0x0                                      ; C08E0300
>	s_load_dwordx4 s[32:35], s[4:5], 0x0                                      ; C0900500
>	s_load_dwordx4 s[36:39], s[4:5], 0x4                                      ; C0920504
>	s_load_dwordx4 s[0:3], s[4:5], 0x8                                        ; C0800508
>	s_load_dwordx4 s[16:19], s[4:5], 0xc                                      ; C088050C
>	s_load_dwordx4 s[40:43], s[4:5], 0x10                                     ; C0940510
>	s_load_dwordx8 s[44:51], s[6:7], 0x0                                      ; C0D60700
>	s_load_dwordx8 s[52:59], s[6:7], 0x8                                      ; C0DA0708
>	s_load_dwordx8 s[8:15], s[6:7], 0x10                                      ; C0C40710
>	s_load_dwordx8 s[20:27], s[6:7], 0x18                                     ; C0CA0718
>	s_waitcnt lgkmcnt(0)                                                      ; BF8C007F
>	s_buffer_load_dword s4, s[28:31], 0xd                                     ; C2021D0D
>	s_buffer_load_dword s5, s[28:31], 0xf                                     ; C2029D0F
>	s_load_dwordx8 s[60:67], s[6:7], 0x20                                     ; C0DE0720
>	image_sample v[0:1], 9, 0, 0, 0, 0, 0, 0, 0, v[6:7], s[44:51], s[32:35]   ; F0800900 010B0006
>	image_sample v10, 2, 0, 0, 0, 0, 0, 0, 0, v[10:11], s[52:59], s[36:39]    ; F0800200 012D0A0A
>	s_waitcnt vmcnt(0)                                                        ; BF8C0770
>	v_mac_f32_e32 v1, v0, v10                                                 ; 3E021500
>	s_waitcnt lgkmcnt(0)                                                      ; BF8C007F
>	v_mul_f32_e64 v0, 0, s4                                                   ; D2100000 00000880
>	v_mad_f32 v4, v4, s5, 1.0                                                 ; D2820004 03C80B04
>	image_sample v8, 8, 0, 0, 0, 0, 0, 0, 0, v[8:9], s[60:67], s[40:43]       ; F0800800 014F0808
>	s_waitcnt vmcnt(0)                                                        ; BF8C0770
>	v_add_f32_e32 v9, -0.5, v8                                                ; 061210F1
>	v_mul_f32_e32 v9, v2, v9                                                  ; 10121302
>	v_madmk_f32_e32 v10, v8, v2, 0xc1200000                                   ; 40140508 C1200000
>	v_add_f32_e64 v9, 0, v9 clamp                                             ; D2060809 00021280
>	v_add_f32_e64 v10, 0, v10 clamp                                           ; D206080A 00021480
>	v_mad_f32 v11, -2.0, v9, v3                                               ; D282000B 040E12F5
>	v_mul_f32_e32 v9, v9, v9                                                  ; 10121309
>	v_mul_f32_e32 v9, v11, v9                                                 ; 1012130B
>	v_mad_f32 v11, -2.0, v10, v3                                              ; D282000B 040E14F5
>	v_mul_f32_e32 v10, v10, v10                                               ; 1014150A
>	v_mul_f32_e32 v10, v11, v10                                               ; 1014150B
>	v_mul_f32_e32 v9, v9, v10                                                 ; 10121509
>	v_add_f32_e32 v8, v8, v8                                                  ; 06101108
>	v_add_f32_e64 v8, 0, v8 clamp                                             ; D2060808 00021080
>	v_mac_f32_e32 v9, v4, v1                                                  ; 3E120304
>	v_mad_f32 v1, -2.0, v8, v3                                                ; D2820001 040E10F5
>	v_mul_f32_e32 v4, v8, v8                                                  ; 10081108
>	v_mul_f32_e32 v1, v1, v4                                                  ; 10020901
>	image_sample v[10:11], 6, 0, 0, 0, 0, 0, 0, 0, v[6:7], s[20:27], s[16:19] ; F0800600 00850A06
>	s_waitcnt vmcnt(0)                                                        ; BF8C0770
>	v_add_f32_e32 v4, v10, v11                                                ; 0608170A
>	v_add_f32_e64 v4, 0, v4 clamp                                             ; D2060804 00020880
>	v_sub_f32_e32 v4, 1.0, v4                                                 ; 080808F2
>	v_sub_f32_e32 v4, 1.0, v4                                                 ; 080808F2
>	v_max_f32_e32 v1, v1, v4                                                  ; 20020901
>	v_madak_f32_e32 v1, v9, v1, 0xbf147ae1                                    ; 42020309 BF147AE1
>	v_mul_f32_e32 v1, v2, v1                                                  ; 10020302
>	v_add_f32_e64 v1, 0, v1 clamp                                             ; D2060801 00020280
>	v_mac_f32_e32 v3, -2.0, v1                                                ; 3E0602F5
>	v_mul_f32_e32 v1, v1, v1                                                  ; 10020301
>	v_mul_f32_e32 v1, v3, v1                                                  ; 10020303
>	image_sample v[2:4], 7, 0, 0, 0, 0, 0, 0, 0, v[6:7], s[8:15], s[0:3]      ; F0800700 00020206
>	s_waitcnt vmcnt(0)                                                        ; BF8C0770
>	v_subrev_f32_e32 v2, s4, v2                                               ; 0A040404
>	v_mad_f32 v3, s4, v5, v3                                                  ; D2820003 040E0A04
>	v_mac_f32_e32 v4, s4, v5                                                  ; 3E080A04
>	v_mad_f32 v2, v1, v2, s4                                                  ; D2820002 00120501
>	v_mad_f32 v3, v3, v1, v0                                                  ; D2820003 04020303
>	v_mac_f32_e32 v0, v4, v1                                                  ; 3E000304
>	v_cvt_pkrtz_f16_f32_e32 v1, v2, v3                                        ; 5E020702
>	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
>
>*** SHADER STATS ***
>SGPRS: 72
>VGPRS: 12
>Code Size: 440 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
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL OUT[1], CLIPVERTEX
>DCL OUT[2], GENERIC[0]
>DCL OUT[3], GENERIC[1]
>DCL CONST[0..54]
>DCL TEMP[0..3], LOCAL
>  0: MUL TEMP[0].xy, CONST[48].xyyy, IN[1].xyyy
>  1: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
>  2: ADD TEMP[1].x, TEMP[0].xxxx, CONST[48].wwww
>  3: MOV TEMP[1].z, TEMP[1].xxxx
>  4: MUL TEMP[0].xy, CONST[49].xyyy, IN[1].xyyy
>  5: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
>  6: ADD TEMP[2].x, TEMP[0].xxxx, CONST[49].wwww
>  7: MOV TEMP[1].w, TEMP[2].xxxx
>  8: MUL TEMP[0].xy, CONST[50].xyyy, IN[1].xyyy
>  9: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 10: ADD TEMP[2].x, TEMP[0].xxxx, CONST[50].wwww
> 11: MUL TEMP[0].xy, CONST[51].xyyy, IN[1].xyyy
> 12: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 13: ADD TEMP[3].x, TEMP[0].xxxx, CONST[51].wwww
> 14: MOV TEMP[2].y, TEMP[3].xxxx
> 15: MUL TEMP[0].xy, CONST[52].xyyy, IN[1].xyyy
> 16: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 17: ADD TEMP[3].x, TEMP[0].xxxx, CONST[52].wwww
> 18: MOV TEMP[2].z, TEMP[3].xxxx
> 19: MUL TEMP[0].xy, CONST[53].xyyy, IN[1].xyyy
> 20: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 21: ADD TEMP[0].x, TEMP[0].xxxx, CONST[53].wwww
> 22: MOV TEMP[2].w, TEMP[0].xxxx
> 23: MOV TEMP[0].xw, IN[0].xxxw
> 24: MOV TEMP[1].xy, IN[1].xyxx
> 25: MAD TEMP[3].x, IN[0].zzzz, CONST[0].zzzz, -IN[0].wwww
> 26: MOV TEMP[0].z, TEMP[3].xxxx
> 27: MOV TEMP[0].y, -IN[0].yyyy
> 28: MAD TEMP[0].xy, CONST[54].xyyy, IN[0].wwww, TEMP[0].xyyy
> 29: MOV OUT[2], TEMP[1]
> 30: MOV OUT[3], TEMP[2]
> 31: MOV OUT[0], TEMP[0]
> 32: MOV OUT[1], IN[0]
> 33: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = call float @llvm.SI.load.const(<16 x i8> %13, i32 8)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %13, i32 768)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %13, i32 772)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %13, i32 780)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %13, i32 784)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %13, i32 788)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %13, i32 796)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %13, i32 800)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %13, i32 804)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %13, i32 812)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %13, i32 816)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %13, i32 820)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %13, i32 828)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %13, i32 832)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %13, i32 836)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %13, i32 844)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %13, i32 848)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %13, i32 852)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %13, i32 860)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %13, i32 864)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %13, i32 868)
>  %35 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %36 = load <16 x i8>, <16 x i8> addrspace(2)* %35, align 16, !tbaa !0
>  %37 = add i32 %5, %8
>  %38 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %36, i32 0, i32 %37)
>  %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 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 1
>  %44 = load <16 x i8>, <16 x i8> addrspace(2)* %43, align 16, !tbaa !0
>  %45 = add i32 %5, %8
>  %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 = fmul float %15, %47
>  %50 = fmul float %16, %48
>  %51 = fadd float %50, %49
>  %52 = fadd float %51, %17
>  %53 = fmul float %18, %47
>  %54 = fmul float %19, %48
>  %55 = fadd float %54, %53
>  %56 = fadd float %55, %20
>  %57 = fmul float %21, %47
>  %58 = fmul float %22, %48
>  %59 = fadd float %58, %57
>  %60 = fadd float %59, %23
>  %61 = fmul float %24, %47
>  %62 = fmul float %25, %48
>  %63 = fadd float %62, %61
>  %64 = fadd float %63, %26
>  %65 = fmul float %27, %47
>  %66 = fmul float %28, %48
>  %67 = fadd float %66, %65
>  %68 = fadd float %67, %29
>  %69 = fmul float %30, %47
>  %70 = fmul float %31, %48
>  %71 = fadd float %70, %69
>  %72 = fadd float %71, %32
>  %73 = fmul float %41, %14
>  %74 = fsub float %73, %42
>  %75 = fmul float %33, %42
>  %76 = fadd float %75, %39
>  %77 = fmul float %34, %42
>  %78 = fsub float %77, %40
>  %79 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 16
>  %80 = load <16 x i8>, <16 x i8> addrspace(2)* %79, align 16, !tbaa !0
>  %81 = call float @llvm.SI.load.const(<16 x i8> %80, i32 0)
>  %82 = fmul float %81, %39
>  %83 = call float @llvm.SI.load.const(<16 x i8> %80, i32 4)
>  %84 = fmul float %83, %40
>  %85 = fadd float %82, %84
>  %86 = call float @llvm.SI.load.const(<16 x i8> %80, i32 8)
>  %87 = fmul float %86, %41
>  %88 = fadd float %85, %87
>  %89 = call float @llvm.SI.load.const(<16 x i8> %80, i32 12)
>  %90 = fmul float %89, %42
>  %91 = fadd float %88, %90
>  %92 = call float @llvm.SI.load.const(<16 x i8> %80, i32 16)
>  %93 = fmul float %92, %39
>  %94 = call float @llvm.SI.load.const(<16 x i8> %80, i32 20)
>  %95 = fmul float %94, %40
>  %96 = fadd float %93, %95
>  %97 = call float @llvm.SI.load.const(<16 x i8> %80, i32 24)
>  %98 = fmul float %97, %41
>  %99 = fadd float %96, %98
>  %100 = call float @llvm.SI.load.const(<16 x i8> %80, i32 28)
>  %101 = fmul float %100, %42
>  %102 = fadd float %99, %101
>  %103 = call float @llvm.SI.load.const(<16 x i8> %80, i32 32)
>  %104 = fmul float %103, %39
>  %105 = call float @llvm.SI.load.const(<16 x i8> %80, i32 36)
>  %106 = fmul float %105, %40
>  %107 = fadd float %104, %106
>  %108 = call float @llvm.SI.load.const(<16 x i8> %80, i32 40)
>  %109 = fmul float %108, %41
>  %110 = fadd float %107, %109
>  %111 = call float @llvm.SI.load.const(<16 x i8> %80, i32 44)
>  %112 = fmul float %111, %42
>  %113 = fadd float %110, %112
>  %114 = call float @llvm.SI.load.const(<16 x i8> %80, i32 48)
>  %115 = fmul float %114, %39
>  %116 = call float @llvm.SI.load.const(<16 x i8> %80, i32 52)
>  %117 = fmul float %116, %40
>  %118 = fadd float %115, %117
>  %119 = call float @llvm.SI.load.const(<16 x i8> %80, i32 56)
>  %120 = fmul float %119, %41
>  %121 = fadd float %118, %120
>  %122 = call float @llvm.SI.load.const(<16 x i8> %80, i32 60)
>  %123 = fmul float %122, %42
>  %124 = fadd float %121, %123
>  %125 = call float @llvm.SI.load.const(<16 x i8> %80, i32 64)
>  %126 = fmul float %125, %39
>  %127 = call float @llvm.SI.load.const(<16 x i8> %80, i32 68)
>  %128 = fmul float %127, %40
>  %129 = fadd float %126, %128
>  %130 = call float @llvm.SI.load.const(<16 x i8> %80, i32 72)
>  %131 = fmul float %130, %41
>  %132 = fadd float %129, %131
>  %133 = call float @llvm.SI.load.const(<16 x i8> %80, i32 76)
>  %134 = fmul float %133, %42
>  %135 = fadd float %132, %134
>  %136 = call float @llvm.SI.load.const(<16 x i8> %80, i32 80)
>  %137 = fmul float %136, %39
>  %138 = call float @llvm.SI.load.const(<16 x i8> %80, i32 84)
>  %139 = fmul float %138, %40
>  %140 = fadd float %137, %139
>  %141 = call float @llvm.SI.load.const(<16 x i8> %80, i32 88)
>  %142 = fmul float %141, %41
>  %143 = fadd float %140, %142
>  %144 = call float @llvm.SI.load.const(<16 x i8> %80, i32 92)
>  %145 = fmul float %144, %42
>  %146 = fadd float %143, %145
>  %147 = call float @llvm.SI.load.const(<16 x i8> %80, i32 96)
>  %148 = fmul float %147, %39
>  %149 = call float @llvm.SI.load.const(<16 x i8> %80, i32 100)
>  %150 = fmul float %149, %40
>  %151 = fadd float %148, %150
>  %152 = call float @llvm.SI.load.const(<16 x i8> %80, i32 104)
>  %153 = fmul float %152, %41
>  %154 = fadd float %151, %153
>  %155 = call float @llvm.SI.load.const(<16 x i8> %80, i32 108)
>  %156 = fmul float %155, %42
>  %157 = fadd float %154, %156
>  %158 = call float @llvm.SI.load.const(<16 x i8> %80, i32 112)
>  %159 = fmul float %158, %39
>  %160 = call float @llvm.SI.load.const(<16 x i8> %80, i32 116)
>  %161 = fmul float %160, %40
>  %162 = fadd float %159, %161
>  %163 = call float @llvm.SI.load.const(<16 x i8> %80, i32 120)
>  %164 = fmul float %163, %41
>  %165 = fadd float %162, %164
>  %166 = call float @llvm.SI.load.const(<16 x i8> %80, i32 124)
>  %167 = fmul float %166, %42
>  %168 = fadd float %165, %167
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %47, float %48, float %52, float %56)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %60, float %64, float %68, float %72)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 12, i32 0, float %76, float %78, float %74, float %42)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 13, i32 0, float %91, float %102, float %113, float %124)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 14, i32 0, float %135, float %146, float %157, 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
>
>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[4:7], s[8:9], 0x0                    ; C0820900
>	s_load_dwordx4 s[12:15], s[8:9], 0x4                  ; C0860904
>	v_add_i32_e32 v0, vcc, s10, v0                        ; 4A00000A
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[4:7], 0 idxen   ; E00C2000 80010100
>	buffer_load_format_xyzw v[5:8], v0, s[12:15], 0 idxen ; E00C2000 80030500
>	s_load_dwordx4 s[4:7], s[2:3], 0x0                    ; C0820300
>	s_load_dwordx4 s[0:3], s[2:3], 0x40                   ; C0800340
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	s_buffer_load_dword s8, s[4:7], 0x2                   ; C2040502
>	s_buffer_load_dword s9, s[4:7], 0xc0                  ; C20485C0
>	s_buffer_load_dword s10, s[4:7], 0xc1                 ; C20505C1
>	s_buffer_load_dword s11, s[4:7], 0xc3                 ; C20585C3
>	s_buffer_load_dword s12, s[4:7], 0xc4                 ; C20605C4
>	s_buffer_load_dword s13, s[4:7], 0xc5                 ; C20685C5
>	s_buffer_load_dword s14, s[4:7], 0xc7                 ; C20705C7
>	s_buffer_load_dword s15, s[4:7], 0xc8                 ; C20785C8
>	s_buffer_load_dword s16, s[4:7], 0xc9                 ; C20805C9
>	s_buffer_load_dword s17, s[4:7], 0xcb                 ; C20885CB
>	s_buffer_load_dword s18, s[4:7], 0xcc                 ; C20905CC
>	s_buffer_load_dword s19, s[4:7], 0xcd                 ; C20985CD
>	s_buffer_load_dword s20, s[4:7], 0xcf                 ; C20A05CF
>	s_buffer_load_dword s21, s[4:7], 0xd0                 ; C20A85D0
>	s_buffer_load_dword s22, s[4:7], 0xd1                 ; C20B05D1
>	s_buffer_load_dword s23, s[4:7], 0xd3                 ; C20B85D3
>	s_buffer_load_dword s24, s[4:7], 0xd4                 ; C20C05D4
>	s_buffer_load_dword s25, s[4:7], 0xd5                 ; C20C85D5
>	s_buffer_load_dword s26, s[4:7], 0xd7                 ; C20D05D7
>	s_buffer_load_dword s27, s[4:7], 0xd8                 ; C20D85D8
>	s_buffer_load_dword s4, s[4:7], 0xd9                  ; C20205D9
>	s_buffer_load_dword s5, s[0:3], 0x0                   ; C2028100
>	s_buffer_load_dword s6, s[0:3], 0x1                   ; C2030101
>	s_buffer_load_dword s7, s[0:3], 0x2                   ; C2038102
>	s_buffer_load_dword s28, s[0:3], 0x3                  ; C20E0103
>	s_buffer_load_dword s29, s[0:3], 0x4                  ; C20E8104
>	s_buffer_load_dword s30, s[0:3], 0x5                  ; C20F0105
>	s_buffer_load_dword s31, s[0:3], 0x6                  ; C20F8106
>	s_buffer_load_dword s32, s[0:3], 0x7                  ; C2100107
>	s_buffer_load_dword s33, s[0:3], 0x8                  ; C2108108
>	s_buffer_load_dword s34, s[0:3], 0x9                  ; C2110109
>	s_buffer_load_dword s35, s[0:3], 0xa                  ; C211810A
>	s_buffer_load_dword s36, s[0:3], 0xb                  ; C212010B
>	s_buffer_load_dword s37, s[0:3], 0xc                  ; C212810C
>	s_buffer_load_dword s38, s[0:3], 0xd                  ; C213010D
>	s_buffer_load_dword s39, s[0:3], 0xe                  ; C213810E
>	s_buffer_load_dword s40, s[0:3], 0xf                  ; C214010F
>	s_buffer_load_dword s41, s[0:3], 0x10                 ; C2148110
>	s_buffer_load_dword s42, s[0:3], 0x11                 ; C2150111
>	s_buffer_load_dword s43, s[0:3], 0x12                 ; C2158112
>	s_buffer_load_dword s44, s[0:3], 0x13                 ; C2160113
>	s_buffer_load_dword s45, s[0:3], 0x14                 ; C2168114
>	s_buffer_load_dword s46, s[0:3], 0x15                 ; C2170115
>	s_buffer_load_dword s47, s[0:3], 0x16                 ; C2178116
>	s_buffer_load_dword s48, s[0:3], 0x17                 ; C2180117
>	s_buffer_load_dword s49, s[0:3], 0x18                 ; C2188118
>	s_buffer_load_dword s50, s[0:3], 0x19                 ; C2190119
>	s_buffer_load_dword s51, s[0:3], 0x1a                 ; C219811A
>	s_buffer_load_dword s52, s[0:3], 0x1b                 ; C21A011B
>	s_buffer_load_dword s53, s[0:3], 0x1c                 ; C21A811C
>	s_buffer_load_dword s54, s[0:3], 0x1d                 ; C21B011D
>	s_buffer_load_dword s55, s[0:3], 0x1e                 ; C21B811E
>	s_buffer_load_dword s0, s[0:3], 0x1f                  ; C200011F
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mov_b32_e32 v0, s11                                 ; 7E00020B
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mov_b32_e32 v7, s14                                 ; 7E0E020E
>	v_mov_b32_e32 v8, s17                                 ; 7E100211
>	v_mov_b32_e32 v9, s20                                 ; 7E120214
>	v_mov_b32_e32 v10, s23                                ; 7E140217
>	v_mov_b32_e32 v11, s26                                ; 7E16021A
>	v_mac_f32_e32 v0, s9, v5                              ; 3E000A09
>	v_mac_f32_e32 v7, s12, v5                             ; 3E0E0A0C
>	v_mac_f32_e32 v8, s15, v5                             ; 3E100A0F
>	v_mac_f32_e32 v9, s18, v5                             ; 3E120A12
>	v_mac_f32_e32 v10, s21, v5                            ; 3E140A15
>	v_mac_f32_e32 v11, s24, v5                            ; 3E160A18
>	v_mac_f32_e32 v0, s10, v6                             ; 3E000C0A
>	v_mac_f32_e32 v7, s13, v6                             ; 3E0E0C0D
>	v_mac_f32_e32 v8, s16, v6                             ; 3E100C10
>	v_mac_f32_e32 v9, s19, v6                             ; 3E120C13
>	exp 15, 32, 0, 0, 0, v5, v6, v0, v7                   ; F800020F 07000605
>	v_mac_f32_e32 v10, s22, v6                            ; 3E140C16
>	v_mac_f32_e32 v11, s25, v6                            ; 3E160C19
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mad_f32 v0, v3, s8, -v4                             ; D2820000 84101103
>	v_mad_f32 v5, s27, v4, v1                             ; D2820005 0406081B
>	v_mad_f32 v6, s4, v4, -v2                             ; D2820006 840A0804
>	v_mul_f32_e32 v7, s6, v2                              ; 100E0406
>	exp 15, 33, 0, 0, 0, v8, v9, v10, v11                 ; F800021F 0B0A0908
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mul_f32_e32 v8, s30, v2                             ; 1010041E
>	v_mac_f32_e32 v7, s5, v1                              ; 3E0E0205
>	exp 15, 12, 0, 0, 0, v5, v6, v0, v4                   ; F80000CF 04000605
>	v_mac_f32_e32 v8, s29, v1                             ; 3E10021D
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mul_f32_e32 v0, s34, v2                             ; 10000422
>	v_mac_f32_e32 v0, s33, v1                             ; 3E000221
>	v_mul_f32_e32 v5, s38, v2                             ; 100A0426
>	v_mac_f32_e32 v5, s37, v1                             ; 3E0A0225
>	v_mul_f32_e32 v6, s42, v2                             ; 100C042A
>	v_mul_f32_e32 v9, s46, v2                             ; 1012042E
>	v_mul_f32_e32 v10, s50, v2                            ; 10140432
>	v_mul_f32_e32 v2, s54, v2                             ; 10040436
>	v_mac_f32_e32 v6, s41, v1                             ; 3E0C0229
>	v_mac_f32_e32 v9, s45, v1                             ; 3E12022D
>	v_mac_f32_e32 v10, s49, v1                            ; 3E140231
>	v_mac_f32_e32 v2, s53, v1                             ; 3E040235
>	v_mac_f32_e32 v7, s7, v3                              ; 3E0E0607
>	v_mac_f32_e32 v8, s31, v3                             ; 3E10061F
>	v_mac_f32_e32 v0, s35, v3                             ; 3E000623
>	v_mac_f32_e32 v5, s39, v3                             ; 3E0A0627
>	v_mac_f32_e32 v6, s43, v3                             ; 3E0C062B
>	v_mac_f32_e32 v9, s47, v3                             ; 3E12062F
>	v_mac_f32_e32 v10, s51, v3                            ; 3E140633
>	v_mac_f32_e32 v2, s55, v3                             ; 3E040637
>	v_mac_f32_e32 v7, s28, v4                             ; 3E0E081C
>	v_mac_f32_e32 v8, s32, v4                             ; 3E100820
>	v_mac_f32_e32 v0, s36, v4                             ; 3E000824
>	v_mac_f32_e32 v5, s40, v4                             ; 3E0A0828
>	v_mac_f32_e32 v6, s44, v4                             ; 3E0C082C
>	v_mac_f32_e32 v9, s48, v4                             ; 3E120830
>	v_mac_f32_e32 v10, s52, v4                            ; 3E140834
>	v_mac_f32_e32 v2, s0, v4                              ; 3E040800
>	exp 15, 13, 0, 0, 0, v7, v8, v0, v5                   ; F80000DF 05000807
>	exp 15, 14, 0, 1, 0, v6, v9, v10, v2                  ; F80008EF 020A0906
>	s_endpgm                                              ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 64
>VGPRS: 12
>Code Size: 544 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
>  clamp_color = 0
>FRAG
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL IN[1], GENERIC[1], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SAMP[1]
>DCL SAMP[2]
>DCL SAMP[3]
>DCL SAMP[4]
>DCL SAMP[5]
>DCL SVIEW[0], 2D, FLOAT
>DCL SVIEW[1], 2D, FLOAT
>DCL SVIEW[2], 2D, FLOAT
>DCL SVIEW[3], 2D, FLOAT
>DCL SVIEW[4], 2D, FLOAT
>DCL SVIEW[5], 2D, FLOAT
>DCL CONST[0..3]
>DCL TEMP[0..10], LOCAL
>IMM[0] FLT32 {    1.0000,     6.0000,     0.7500,     0.2500}
>IMM[1] FLT32 {   -1.0000,     0.3000,     0.5900,     0.1100}
>IMM[2] FLT32 {   -0.0800,    14.2857,    -2.0000,     3.0000}
>IMM[3] FLT32 {    0.0300,     0.5000,    -0.5000,    -1.0000}
>IMM[4] FLT32 {   10.0000,   -10.0000,    -0.5800,  -100.0000}
>IMM[5] FLT32 {   -0.9000,    -0.0100,     0.0000,     0.0000}
>  0: ADD TEMP[0].xyz, -CONST[0].xyzz, CONST[1].xyzz
>  1: MOV TEMP[1].xy, IN[0].zwww
>  2: TEX TEMP[1], TEMP[1], SAMP[5], 2D
>  3: MAD TEMP[0].xyz, TEMP[1].xxxx, TEMP[0].xyzz, CONST[0].xyzz
>  4: LRP TEMP[2].xyz, TEMP[1].yyyy, CONST[2].xyzz, TEMP[0].xyzz
>  5: MOV TEMP[0].x, CONST[0].wwww
>  6: MOV TEMP[0].y, CONST[1].wwww
>  7: MOV TEMP[0].z, CONST[2].wwww
>  8: LRP TEMP[3].xyz, TEMP[1].zzzz, TEMP[0].xyzz, TEMP[2].xyzz
>  9: MOV TEMP[4].xy, IN[0].xyyy
> 10: TEX TEMP[5].yz, TEMP[4], SAMP[3], 2D
> 11: LRP TEMP[4].xyz, TEMP[5].yyyy, CONST[2].xyzz, TEMP[3].xyzz
> 12: LRP TEMP[3].xyz, TEMP[5].zzzz, TEMP[0].xyzz, TEMP[4].xyzz
> 13: ADD TEMP[0].x, TEMP[5].zzzz, TEMP[5].yyyy
> 14: MOV_SAT TEMP[5].x, TEMP[0].xxxx
> 15: ADD TEMP[0].x, -TEMP[5].xxxx, IMM[0].xxxx
> 16: ADD TEMP[0].x, -TEMP[0].xxxx, IMM[0].xxxx
> 17: MOV TEMP[5].xy, IN[1].zwww
> 18: TEX TEMP[5], TEMP[5], SAMP[4], 2D
> 19: MUL TEMP[6].x, IMM[0].zzzz, CONST[3].wwww
> 20: MOV TEMP[7].xy, IN[0].xyyy
> 21: TEX TEMP[7], TEMP[7], SAMP[0], 2D
> 22: ADD TEMP[8].x, -TEMP[7].xxxx, IMM[0].xxxx
> 23: MUL TEMP[8].x, TEMP[8].xxxx, TEMP[8].xxxx
> 24: MUL TEMP[8].x, TEMP[8].xxxx, TEMP[8].xxxx
> 25: MAD TEMP[6].x, TEMP[8].xxxx, IMM[0].wwww, TEMP[6].xxxx
> 26: ADD TEMP[5], TEMP[5], IMM[1].xxxx
> 27: MAD TEMP[5], TEMP[6].xxxx, TEMP[5], IMM[0].xxxx
> 28: MUL TEMP[6].xyz, TEMP[3].xyzz, TEMP[5].xyzz
> 29: DP3 TEMP[8].x, TEMP[6].xyzz, IMM[1].yzww
> 30: ADD TEMP[4].x, TEMP[8].xxxx, IMM[2].xxxx
> 31: MUL TEMP[4].x, TEMP[4].xxxx, IMM[2].yyyy
> 32: MOV_SAT TEMP[8].x, TEMP[4].xxxx
> 33: MAD TEMP[2].x, TEMP[8].xxxx, IMM[2].zzzz, IMM[2].wwww
> 34: MUL TEMP[4].x, TEMP[8].xxxx, TEMP[8].xxxx
> 35: MAD TEMP[4].x, TEMP[2].xxxx, -TEMP[4].xxxx, IMM[0].xxxx
> 36: MAD TEMP[2].xyz, TEMP[4].xxxx, IMM[3].xxxx, TEMP[6].xyzz
> 37: MUL TEMP[2].xyz, TEMP[2].xyzz, TEMP[7].zzzz
> 38: MAD TEMP[6].xyz, TEMP[2].xyzz, IMM[3].yyyy, TEMP[6].xyzz
> 39: MOV_SAT TEMP[6].xyz, TEMP[6].xyzz
> 40: MUL TEMP[2].xyz, TEMP[7].yyyy, TEMP[6].xyzz
> 41: MOV TEMP[8].xy, IN[0].xyyy
> 42: TEX TEMP[8], TEMP[8], SAMP[2], 2D
> 43: MAD TEMP[6].xyz, TEMP[6].xyzz, -TEMP[7].yyyy, TEMP[8].xyzz
> 44: ADD TEMP[4].x, TEMP[1].wwww, TEMP[1].wwww
> 45: MOV_SAT TEMP[9].x, TEMP[4].xxxx
> 46: ADD TEMP[3].xy, TEMP[1].wwww, IMM[3].zwww
> 47: MUL TEMP[3].xy, TEMP[3].xyyy, IMM[4].xyyy
> 48: MOV_SAT TEMP[1].xy, TEMP[3].xyyy
> 49: MAD TEMP[10].x, TEMP[9].xxxx, IMM[2].zzzz, IMM[2].wwww
> 50: MUL TEMP[4].x, TEMP[9].xxxx, TEMP[9].xxxx
> 51: MUL TEMP[4].x, TEMP[4].xxxx, TEMP[10].xxxx
> 52: MAX TEMP[9].x, TEMP[0].xxxx, TEMP[4].xxxx
> 53: MAD TEMP[10].xy, TEMP[1].xyyy, IMM[2].zzzz, IMM[2].wwww
> 54: MUL TEMP[3].xy, TEMP[1].xyyy, TEMP[1].xyyy
> 55: MUL TEMP[1].xy, TEMP[10].xyyy, TEMP[3].xyyy
> 56: MUL TEMP[0].x, TEMP[1].yyyy, TEMP[1].xxxx
> 57: MOV TEMP[1].xy, IN[1].xyyy
> 58: TEX TEMP[1].y, TEMP[1], SAMP[1], 2D
> 59: MAD TEMP[4].x, TEMP[1].yyyy, TEMP[7].xxxx, TEMP[7].wwww
> 60: MUL TEMP[1].x, TEMP[7].yyyy, CONST[3].zzzz
> 61: MUL TEMP[1].x, TEMP[5].wwww, TEMP[1].xxxx
> 62: MAD TEMP[3].x, CONST[3].wwww, IMM[0].yyyy, IMM[0].xxxx
> 63: MAD TEMP[0].x, TEMP[4].xxxx, TEMP[3].xxxx, TEMP[0].xxxx
> 64: MAD TEMP[0].x, TEMP[0].xxxx, TEMP[9].xxxx, IMM[4].zzzz
> 65: MUL TEMP[0].x, TEMP[0].xxxx, IMM[4].xxxx
> 66: MOV_SAT TEMP[3].x, TEMP[0].xxxx
> 67: MAD TEMP[4].x, TEMP[3].xxxx, IMM[2].zzzz, IMM[2].wwww
> 68: MUL TEMP[0].x, TEMP[3].xxxx, TEMP[3].xxxx
> 69: MUL TEMP[3].x, TEMP[0].xxxx, TEMP[4].xxxx
> 70: MAD TEMP[4].xy, TEMP[4].xxxx, TEMP[0].xxxx, IMM[5].xyyy
> 71: MUL TEMP[4].xy, TEMP[4].xyyy, IMM[4].xwww
> 72: MAD TEMP[2].xyz, TEMP[3].xxxx, TEMP[6].xyzz, TEMP[2].xyzz
> 73: MAX TEMP[3].x, TEMP[4].yyyy, IMM[5].zzzz
> 74: MOV_SAT TEMP[4].x, TEMP[4].xxxx
> 75: MAD TEMP[5].x, TEMP[3].xxxx, IMM[2].zzzz, IMM[2].wwww
> 76: MUL TEMP[0].x, TEMP[3].xxxx, TEMP[3].xxxx
> 77: MUL TEMP[0].x, TEMP[0].xxxx, TEMP[5].xxxx
> 78: MAD TEMP[3].x, TEMP[1].xxxx, -TEMP[0].xxxx, TEMP[8].wwww
> 79: MUL TEMP[0].x, TEMP[0].xxxx, TEMP[1].xxxx
> 80: MAD TEMP[1].x, TEMP[4].xxxx, IMM[2].zzzz, IMM[2].wwww
> 81: MUL TEMP[4].x, TEMP[4].xxxx, TEMP[4].xxxx
> 82: MUL TEMP[1].x, TEMP[4].xxxx, TEMP[1].xxxx
> 83: MAD TEMP[0].x, TEMP[1].xxxx, TEMP[3].xxxx, TEMP[0].xxxx
> 84: MOV TEMP[2].w, TEMP[0].xxxx
> 85: MOV OUT[0], TEMP[2]
> 86: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %24 = load <16 x i8>, <16 x i8> addrspace(2)* %23, align 16, !tbaa !0
>  %25 = call float @llvm.SI.load.const(<16 x i8> %24, i32 0)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %24, i32 4)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %24, i32 8)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %24, i32 12)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %24, i32 16)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %24, i32 20)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %24, i32 24)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %24, i32 28)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %24, i32 32)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %24, i32 36)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %24, i32 40)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %24, i32 44)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %24, i32 56)
>  %38 = call float @llvm.SI.load.const(<16 x i8> %24, i32 60)
>  %39 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %40 = load <8 x i32>, <8 x i32> addrspace(2)* %39, align 32, !tbaa !0
>  %41 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %42 = load <4 x i32>, <4 x i32> addrspace(2)* %41, align 16, !tbaa !0
>  %43 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 1
>  %44 = load <8 x i32>, <8 x i32> addrspace(2)* %43, align 32, !tbaa !0
>  %45 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 1
>  %46 = load <4 x i32>, <4 x i32> addrspace(2)* %45, align 16, !tbaa !0
>  %47 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 2
>  %48 = load <8 x i32>, <8 x i32> addrspace(2)* %47, align 32, !tbaa !0
>  %49 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 2
>  %50 = load <4 x i32>, <4 x i32> addrspace(2)* %49, align 16, !tbaa !0
>  %51 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 3
>  %52 = load <8 x i32>, <8 x i32> addrspace(2)* %51, align 32, !tbaa !0
>  %53 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 3
>  %54 = load <4 x i32>, <4 x i32> addrspace(2)* %53, align 16, !tbaa !0
>  %55 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 4
>  %56 = load <8 x i32>, <8 x i32> addrspace(2)* %55, align 32, !tbaa !0
>  %57 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 4
>  %58 = load <4 x i32>, <4 x i32> addrspace(2)* %57, align 16, !tbaa !0
>  %59 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 5
>  %60 = load <8 x i32>, <8 x i32> addrspace(2)* %59, align 32, !tbaa !0
>  %61 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 5
>  %62 = load <4 x i32>, <4 x i32> addrspace(2)* %61, align 16, !tbaa !0
>  %63 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %8)
>  %64 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %8)
>  %65 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %6, <2 x i32> %8)
>  %66 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %6, <2 x i32> %8)
>  %67 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %6, <2 x i32> %8)
>  %68 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %6, <2 x i32> %8)
>  %69 = call float @llvm.SI.fs.interp(i32 2, i32 1, i32 %6, <2 x i32> %8)
>  %70 = call float @llvm.SI.fs.interp(i32 3, i32 1, i32 %6, <2 x i32> %8)
>  %71 = fsub float %29, %25
>  %72 = fsub float %30, %26
>  %73 = fsub float %31, %27
>  %74 = bitcast float %65 to i32
>  %75 = bitcast float %66 to i32
>  %76 = insertelement <2 x i32> undef, i32 %74, i32 0
>  %77 = insertelement <2 x i32> %76, i32 %75, i32 1
>  %78 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %77, <8 x i32> %60, <4 x i32> %62, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %79 = extractelement <4 x float> %78, i32 0
>  %80 = extractelement <4 x float> %78, i32 1
>  %81 = extractelement <4 x float> %78, i32 2
>  %82 = extractelement <4 x float> %78, i32 3
>  %83 = fmul float %79, %71
>  %84 = fadd float %83, %25
>  %85 = fmul float %79, %72
>  %86 = fadd float %85, %26
>  %87 = fmul float %79, %73
>  %88 = fadd float %87, %27
>  %89 = fsub float 1.000000e+00, %80
>  %90 = fmul float %33, %80
>  %91 = fmul float %84, %89
>  %92 = fadd float %90, %91
>  %93 = fsub float 1.000000e+00, %80
>  %94 = fmul float %34, %80
>  %95 = fmul float %86, %93
>  %96 = fadd float %94, %95
>  %97 = fsub float 1.000000e+00, %80
>  %98 = fmul float %35, %80
>  %99 = fmul float %88, %97
>  %100 = fadd float %98, %99
>  %101 = fsub float 1.000000e+00, %81
>  %102 = fmul float %28, %81
>  %103 = fmul float %92, %101
>  %104 = fadd float %102, %103
>  %105 = fsub float 1.000000e+00, %81
>  %106 = fmul float %32, %81
>  %107 = fmul float %96, %105
>  %108 = fadd float %106, %107
>  %109 = fsub float 1.000000e+00, %81
>  %110 = fmul float %36, %81
>  %111 = fmul float %100, %109
>  %112 = fadd float %110, %111
>  %113 = bitcast float %63 to i32
>  %114 = bitcast float %64 to i32
>  %115 = insertelement <2 x i32> undef, i32 %113, i32 0
>  %116 = insertelement <2 x i32> %115, i32 %114, i32 1
>  %117 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %116, <8 x i32> %52, <4 x i32> %54, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %118 = extractelement <4 x float> %117, i32 1
>  %119 = extractelement <4 x float> %117, i32 2
>  %120 = fsub float 1.000000e+00, %118
>  %121 = fmul float %33, %118
>  %122 = fmul float %104, %120
>  %123 = fadd float %121, %122
>  %124 = fsub float 1.000000e+00, %118
>  %125 = fmul float %34, %118
>  %126 = fmul float %108, %124
>  %127 = fadd float %125, %126
>  %128 = fsub float 1.000000e+00, %118
>  %129 = fmul float %35, %118
>  %130 = fmul float %112, %128
>  %131 = fadd float %129, %130
>  %132 = fsub float 1.000000e+00, %119
>  %133 = fmul float %28, %119
>  %134 = fmul float %123, %132
>  %135 = fadd float %133, %134
>  %136 = fsub float 1.000000e+00, %119
>  %137 = fmul float %32, %119
>  %138 = fmul float %127, %136
>  %139 = fadd float %137, %138
>  %140 = fsub float 1.000000e+00, %119
>  %141 = fmul float %36, %119
>  %142 = fmul float %131, %140
>  %143 = fadd float %141, %142
>  %144 = fadd float %119, %118
>  %145 = call float @llvm.AMDIL.clamp.(float %144, float 0.000000e+00, float 1.000000e+00)
>  %146 = fsub float 1.000000e+00, %145
>  %147 = fsub float 1.000000e+00, %146
>  %148 = bitcast float %69 to i32
>  %149 = bitcast float %70 to i32
>  %150 = insertelement <2 x i32> undef, i32 %148, i32 0
>  %151 = insertelement <2 x i32> %150, i32 %149, i32 1
>  %152 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %151, <8 x i32> %56, <4 x i32> %58, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %153 = extractelement <4 x float> %152, i32 0
>  %154 = extractelement <4 x float> %152, i32 1
>  %155 = extractelement <4 x float> %152, i32 2
>  %156 = extractelement <4 x float> %152, i32 3
>  %157 = fmul float %38, 7.500000e-01
>  %158 = bitcast float %63 to i32
>  %159 = bitcast float %64 to i32
>  %160 = insertelement <2 x i32> undef, i32 %158, i32 0
>  %161 = insertelement <2 x i32> %160, i32 %159, i32 1
>  %162 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %161, <8 x i32> %40, <4 x i32> %42, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %163 = extractelement <4 x float> %162, i32 0
>  %164 = extractelement <4 x float> %162, i32 1
>  %165 = extractelement <4 x float> %162, i32 2
>  %166 = extractelement <4 x float> %162, i32 3
>  %167 = fsub float 1.000000e+00, %163
>  %168 = fmul float %167, %167
>  %169 = fmul float %168, %168
>  %170 = fmul float %169, 2.500000e-01
>  %171 = fadd float %170, %157
>  %172 = fadd float %153, -1.000000e+00
>  %173 = fadd float %154, -1.000000e+00
>  %174 = fadd float %155, -1.000000e+00
>  %175 = fadd float %156, -1.000000e+00
>  %176 = fmul float %171, %172
>  %177 = fadd float %176, 1.000000e+00
>  %178 = fmul float %171, %173
>  %179 = fadd float %178, 1.000000e+00
>  %180 = fmul float %171, %174
>  %181 = fadd float %180, 1.000000e+00
>  %182 = fmul float %171, %175
>  %183 = fadd float %182, 1.000000e+00
>  %184 = fmul float %135, %177
>  %185 = fmul float %139, %179
>  %186 = fmul float %143, %181
>  %187 = fmul float %184, 0x3FD3333340000000
>  %188 = fmul float %185, 0x3FE2E147A0000000
>  %189 = fadd float %188, %187
>  %190 = fmul float %186, 0x3FBC28F5C0000000
>  %191 = fadd float %189, %190
>  %192 = fadd float %191, 0xBFB47AE140000000
>  %193 = fmul float %192, 0x402C924920000000
>  %194 = call float @llvm.AMDIL.clamp.(float %193, float 0.000000e+00, float 1.000000e+00)
>  %195 = fmul float %194, -2.000000e+00
>  %196 = fadd float %195, 3.000000e+00
>  %197 = fmul float %194, %194
>  %198 = fmul float %197, %196
>  %199 = fsub float 1.000000e+00, %198
>  %200 = fmul float %199, 0x3F9EB851E0000000
>  %201 = fadd float %200, %184
>  %202 = fmul float %199, 0x3F9EB851E0000000
>  %203 = fadd float %202, %185
>  %204 = fmul float %199, 0x3F9EB851E0000000
>  %205 = fadd float %204, %186
>  %206 = fmul float %201, %165
>  %207 = fmul float %203, %165
>  %208 = fmul float %205, %165
>  %209 = fmul float %206, 5.000000e-01
>  %210 = fadd float %209, %184
>  %211 = fmul float %207, 5.000000e-01
>  %212 = fadd float %211, %185
>  %213 = fmul float %208, 5.000000e-01
>  %214 = fadd float %213, %186
>  %215 = call float @llvm.AMDIL.clamp.(float %210, float 0.000000e+00, float 1.000000e+00)
>  %216 = call float @llvm.AMDIL.clamp.(float %212, float 0.000000e+00, float 1.000000e+00)
>  %217 = call float @llvm.AMDIL.clamp.(float %214, float 0.000000e+00, float 1.000000e+00)
>  %218 = fmul float %164, %215
>  %219 = fmul float %164, %216
>  %220 = fmul float %164, %217
>  %221 = bitcast float %63 to i32
>  %222 = bitcast float %64 to i32
>  %223 = insertelement <2 x i32> undef, i32 %221, i32 0
>  %224 = insertelement <2 x i32> %223, i32 %222, i32 1
>  %225 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %224, <8 x i32> %48, <4 x i32> %50, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %226 = extractelement <4 x float> %225, i32 0
>  %227 = extractelement <4 x float> %225, i32 1
>  %228 = extractelement <4 x float> %225, i32 2
>  %229 = extractelement <4 x float> %225, i32 3
>  %230 = fmul float %164, %215
>  %231 = fsub float %226, %230
>  %232 = fmul float %164, %216
>  %233 = fsub float %227, %232
>  %234 = fmul float %164, %217
>  %235 = fsub float %228, %234
>  %236 = fadd float %82, %82
>  %237 = call float @llvm.AMDIL.clamp.(float %236, float 0.000000e+00, float 1.000000e+00)
>  %238 = fadd float %82, -5.000000e-01
>  %239 = fadd float %82, -1.000000e+00
>  %240 = fmul float %238, 1.000000e+01
>  %241 = fmul float %239, -1.000000e+01
>  %242 = call float @llvm.AMDIL.clamp.(float %240, float 0.000000e+00, float 1.000000e+00)
>  %243 = call float @llvm.AMDIL.clamp.(float %241, float 0.000000e+00, float 1.000000e+00)
>  %244 = fmul float %237, -2.000000e+00
>  %245 = fadd float %244, 3.000000e+00
>  %246 = fmul float %237, %237
>  %247 = fmul float %246, %245
>  %248 = call float @llvm.maxnum.f32(float %147, float %247)
>  %249 = fmul float %242, -2.000000e+00
>  %250 = fadd float %249, 3.000000e+00
>  %251 = fmul float %243, -2.000000e+00
>  %252 = fadd float %251, 3.000000e+00
>  %253 = fmul float %242, %242
>  %254 = fmul float %243, %243
>  %255 = fmul float %250, %253
>  %256 = fmul float %252, %254
>  %257 = fmul float %256, %255
>  %258 = bitcast float %67 to i32
>  %259 = bitcast float %68 to i32
>  %260 = insertelement <2 x i32> undef, i32 %258, i32 0
>  %261 = insertelement <2 x i32> %260, i32 %259, i32 1
>  %262 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %261, <8 x i32> %44, <4 x i32> %46, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %263 = extractelement <4 x float> %262, i32 1
>  %264 = fmul float %263, %163
>  %265 = fadd float %264, %166
>  %266 = fmul float %164, %37
>  %267 = fmul float %183, %266
>  %268 = fmul float %38, 6.000000e+00
>  %269 = fadd float %268, 1.000000e+00
>  %270 = fmul float %265, %269
>  %271 = fadd float %270, %257
>  %272 = fmul float %271, %248
>  %273 = fadd float %272, 0xBFE28F5C20000000
>  %274 = fmul float %273, 1.000000e+01
>  %275 = call float @llvm.AMDIL.clamp.(float %274, float 0.000000e+00, float 1.000000e+00)
>  %276 = fmul float %275, -2.000000e+00
>  %277 = fadd float %276, 3.000000e+00
>  %278 = fmul float %275, %275
>  %279 = fmul float %278, %277
>  %280 = fmul float %277, %278
>  %281 = fadd float %280, 0xBFECCCCCC0000000
>  %282 = fmul float %277, %278
>  %283 = fadd float %282, 0xBF847AE140000000
>  %284 = fmul float %281, 1.000000e+01
>  %285 = fmul float %283, -1.000000e+02
>  %286 = fmul float %279, %231
>  %287 = fadd float %286, %218
>  %288 = fmul float %279, %233
>  %289 = fadd float %288, %219
>  %290 = fmul float %279, %235
>  %291 = fadd float %290, %220
>  %292 = call float @llvm.maxnum.f32(float %285, float 0.000000e+00)
>  %293 = call float @llvm.AMDIL.clamp.(float %284, float 0.000000e+00, float 1.000000e+00)
>  %294 = fmul float %292, -2.000000e+00
>  %295 = fadd float %294, 3.000000e+00
>  %296 = fmul float %292, %292
>  %297 = fmul float %296, %295
>  %298 = fmul float %297, %267
>  %299 = fsub float %229, %298
>  %300 = fmul float %297, %267
>  %301 = fmul float %293, -2.000000e+00
>  %302 = fadd float %301, 3.000000e+00
>  %303 = fmul float %293, %293
>  %304 = fmul float %303, %302
>  %305 = fmul float %304, %299
>  %306 = fadd float %305, %300
>  %307 = call i32 @llvm.SI.packf16(float %287, float %289)
>  %308 = bitcast i32 %307 to float
>  %309 = call i32 @llvm.SI.packf16(float %291, float %306)
>  %310 = bitcast i32 %309 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %308, float %310, float %308, float %310)
>  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.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #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 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, s10                                                            ; BEFC030A
>	v_mov_b32_e32 v2, 0x3f400000                                                 ; 7E0402FF 3F400000
>	v_mov_b32_e32 v3, 0xbda3d70a                                                 ; 7E0602FF BDA3D70A
>	v_mov_b32_e32 v4, 0x40400000                                                 ; 7E0802FF 40400000
>	v_mov_b32_e32 v5, 0x3cf5c28f                                                 ; 7E0A02FF 3CF5C28F
>	v_mov_b32_e32 v6, 0x41200000                                                 ; 7E0C02FF 41200000
>	v_mov_b32_e32 v7, 0x40c00000                                                 ; 7E0E02FF 40C00000
>	v_mov_b32_e32 v8, 0xc2c80000                                                 ; 7E1002FF C2C80000
>	v_interp_p1_f32 v9, v0, 0, 0, [m0]                                           ; C8240000
>	v_interp_p2_f32 v9, [v9], v1, 0, 0, [m0]                                     ; C8250001
>	v_interp_p1_f32 v10, v0, 1, 0, [m0]                                          ; C8280100
>	v_interp_p2_f32 v10, [v10], v1, 1, 0, [m0]                                   ; C8290101
>	v_interp_p1_f32 v11, v0, 2, 0, [m0]                                          ; C82C0200
>	v_interp_p2_f32 v11, [v11], v1, 2, 0, [m0]                                   ; C82D0201
>	v_interp_p1_f32 v12, v0, 3, 0, [m0]                                          ; C8300300
>	v_interp_p2_f32 v12, [v12], v1, 3, 0, [m0]                                   ; C8310301
>	v_interp_p1_f32 v13, v0, 0, 1, [m0]                                          ; C8340400
>	v_interp_p2_f32 v13, [v13], v1, 0, 1, [m0]                                   ; C8350401
>	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 v15, v0, 2, 1, [m0]                                          ; C83C0600
>	v_interp_p2_f32 v15, [v15], v1, 2, 1, [m0]                                   ; C83D0601
>	v_interp_p1_f32 v16, v0, 3, 1, [m0]                                          ; C8400700
>	v_interp_p2_f32 v16, [v16], v1, 3, 1, [m0]                                   ; C8410701
>	s_load_dwordx4 s[52:55], s[2:3], 0x0                                         ; C09A0300
>	s_load_dwordx4 s[28:31], s[4:5], 0x0                                         ; C08E0500
>	s_load_dwordx4 s[0:3], s[4:5], 0x4                                           ; C0800504
>	s_load_dwordx4 s[16:19], s[4:5], 0x8                                         ; C0880508
>	s_load_dwordx4 s[32:35], s[4:5], 0xc                                         ; C090050C
>	s_load_dwordx4 s[36:39], s[4:5], 0x10                                        ; C0920510
>	s_load_dwordx4 s[48:51], s[4:5], 0x14                                        ; C0980514
>	s_load_dwordx8 s[40:47], s[6:7], 0x0                                         ; C0D40700
>	s_load_dwordx8 s[8:15], s[6:7], 0x8                                          ; C0C40708
>	s_load_dwordx8 s[20:27], s[6:7], 0x10                                        ; C0CA0710
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	s_buffer_load_dword s4, s[52:55], 0x0                                        ; C2023500
>	s_buffer_load_dword s5, s[52:55], 0x1                                        ; C202B501
>	s_buffer_load_dword s56, s[52:55], 0x2                                       ; C21C3502
>	s_buffer_load_dword s57, s[52:55], 0x3                                       ; C21CB503
>	s_buffer_load_dword s58, s[52:55], 0x4                                       ; C21D3504
>	s_buffer_load_dword s59, s[52:55], 0x5                                       ; C21DB505
>	s_buffer_load_dword s60, s[52:55], 0x6                                       ; C21E3506
>	s_buffer_load_dword s61, s[52:55], 0x7                                       ; C21EB507
>	s_buffer_load_dword s62, s[52:55], 0x8                                       ; C21F3508
>	s_buffer_load_dword s63, s[52:55], 0x9                                       ; C21FB509
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mov_b32_e32 v0, s4                                                         ; 7E000204
>	s_buffer_load_dword s64, s[52:55], 0xa                                       ; C220350A
>	v_mov_b32_e32 v1, s5                                                         ; 7E020205
>	s_buffer_load_dword s65, s[52:55], 0xb                                       ; C220B50B
>	v_mov_b32_e32 v17, s56                                                       ; 7E220238
>	s_buffer_load_dword s66, s[52:55], 0xe                                       ; C221350E
>	s_buffer_load_dword s52, s[52:55], 0xf                                       ; C21A350F
>	v_sub_f32_e32 v0, s58, v0                                                    ; 0800003A
>	v_sub_f32_e32 v1, s59, v1                                                    ; 0802023B
>	v_sub_f32_e32 v17, s60, v17                                                  ; 0822223C
>	s_load_dwordx8 s[68:75], s[6:7], 0x18                                        ; C0E20718
>	s_load_dwordx8 s[76:83], s[6:7], 0x20                                        ; C0E60720
>	s_load_dwordx8 s[84:91], s[6:7], 0x28                                        ; C0EA0728
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mul_f32_e32 v2, s52, v2                                                    ; 10040434
>	image_sample v[18:21], 15, 0, 0, 0, 0, 0, 0, 0, v[11:12], s[84:91], s[48:51] ; F0800F00 0195120B
>	image_sample v[11:12], 6, 0, 0, 0, 0, 0, 0, 0, v[9:10], s[68:75], s[32:35]   ; F0800600 01110B09
>	image_sample v[22:25], 15, 0, 0, 0, 0, 0, 0, 0, v[15:16], s[76:83], s[36:39] ; F0800F00 0133160F
>	image_sample v[26:29], 15, 0, 0, 0, 0, 0, 0, 0, v[9:10], s[40:47], s[28:31]  ; F0800F00 00EA1A09
>	s_waitcnt vmcnt(3)                                                           ; BF8C0773
>	v_mad_f32 v0, v18, v0, s4                                                    ; D2820000 00120112
>	v_mad_f32 v1, v18, v1, s5                                                    ; D2820001 00160312
>	v_mad_f32 v15, v18, v17, s56                                                 ; D282000F 00E22312
>	v_mad_f32 v0, -v19, v0, v0                                                   ; D2820000 24020113
>	v_mac_f32_e32 v0, s62, v19                                                   ; 3E00263E
>	v_mad_f32 v1, -v19, v1, v1                                                   ; D2820001 24060313
>	v_mac_f32_e32 v1, s63, v19                                                   ; 3E02263F
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_sub_f32_e32 v16, 1.0, v26                                                  ; 082034F2
>	v_mad_f32 v16, -v26, v16, v16                                                ; D2820010 2442211A
>	v_mul_f32_e32 v16, v16, v16                                                  ; 10202110
>	v_madmk_f32_e32 v2, v16, v2, 0x3e800000                                      ; 40040510 3E800000
>	v_mad_f32 v16, v22, v2, -v2                                                  ; D2820010 840A0516
>	v_mad_f32 v17, v23, v2, -v2                                                  ; D2820011 840A0517
>	v_mad_f32 v18, v24, v2, -v2                                                  ; D2820012 840A0518
>	v_mad_f32 v2, v25, v2, -v2                                                   ; D2820002 840A0519
>	v_mad_f32 v0, -v20, v0, v0                                                   ; D2820000 24020114
>	v_mac_f32_e32 v0, s57, v20                                                   ; 3E002839
>	v_mad_f32 v0, -v11, v0, v0                                                   ; D2820000 2402010B
>	v_mac_f32_e32 v0, s62, v11                                                   ; 3E00163E
>	v_mad_f32 v0, -v12, v0, v0                                                   ; D2820000 2402010C
>	v_mac_f32_e32 v0, s57, v12                                                   ; 3E001839
>	v_mac_f32_e32 v0, v0, v16                                                    ; 3E002100
>	v_mad_f32 v1, -v20, v1, v1                                                   ; D2820001 24060314
>	v_mac_f32_e32 v1, s61, v20                                                   ; 3E02283D
>	v_mad_f32 v1, -v11, v1, v1                                                   ; D2820001 2406030B
>	v_mac_f32_e32 v1, s63, v11                                                   ; 3E02163F
>	v_mad_f32 v1, -v12, v1, v1                                                   ; D2820001 2406030C
>	v_mac_f32_e32 v1, s61, v12                                                   ; 3E02183D
>	v_mac_f32_e32 v1, v1, v17                                                    ; 3E022301
>	v_mad_f32 v15, -v19, v15, v15                                                ; D282000F 243E1F13
>	v_mac_f32_e32 v15, s64, v19                                                  ; 3E1E2640
>	v_mad_f32 v15, -v20, v15, v15                                                ; D282000F 243E1F14
>	v_mac_f32_e32 v15, s65, v20                                                  ; 3E1E2841
>	v_mad_f32 v15, -v11, v15, v15                                                ; D282000F 243E1F0B
>	v_mac_f32_e32 v15, s64, v11                                                  ; 3E1E1640
>	v_mad_f32 v15, -v12, v15, v15                                                ; D282000F 243E1F0C
>	v_mac_f32_e32 v15, s65, v12                                                  ; 3E1E1841
>	v_mac_f32_e32 v15, v15, v18                                                  ; 3E1E250F
>	v_mul_f32_e32 v16, 0x3e99999a, v0                                            ; 102000FF 3E99999A
>	v_madmk_f32_e32 v16, v1, v16, 0x3f170a3d                                     ; 40202101 3F170A3D
>	v_madmk_f32_e32 v16, v15, v16, 0x3de147ae                                    ; 4020210F 3DE147AE
>	v_add_f32_e32 v3, v16, v3                                                    ; 06060710
>	v_mul_f32_e32 v3, 0x41649249, v3                                             ; 100606FF 41649249
>	v_add_f32_e64 v3, 0, v3 clamp                                                ; D2060803 00020680
>	v_mad_f32 v16, -2.0, v3, v4                                                  ; D2820010 041206F5
>	v_mul_f32_e32 v3, v3, v3                                                     ; 10060703
>	v_mad_f32 v3, -v3, v16, 1.0                                                  ; D2820003 23CA2103
>	v_mad_f32 v16, v5, v3, v0                                                    ; D2820010 04020705
>	v_mad_f32 v17, v5, v3, v1                                                    ; D2820011 04060705
>	v_mad_f32 v3, v5, v3, v15                                                    ; D2820003 043E0705
>	v_mul_f32_e32 v5, v28, v16                                                   ; 100A211C
>	v_mac_f32_e32 v0, 0.5, v5                                                    ; 3E000AF0
>	v_mul_f32_e32 v5, v28, v17                                                   ; 100A231C
>	v_mul_f32_e32 v3, v28, v3                                                    ; 1006071C
>	v_mac_f32_e32 v1, 0.5, v5                                                    ; 3E020AF0
>	v_mac_f32_e32 v15, 0.5, v3                                                   ; 3E1E06F0
>	v_add_f32_e64 v0, 0, v0 clamp                                                ; D2060800 00020080
>	v_add_f32_e64 v1, 0, v1 clamp                                                ; D2060801 00020280
>	v_add_f32_e64 v3, 0, v15 clamp                                               ; D2060803 00021E80
>	v_add_f32_e32 v5, v21, v21                                                   ; 060A2B15
>	v_add_f32_e32 v15, -0.5, v21                                                 ; 061E2AF1
>	v_madmk_f32_e32 v16, v21, v6, 0xc1200000                                     ; 40200D15 C1200000
>	v_mul_f32_e32 v17, v0, v27                                                   ; 10223700
>	v_mul_f32_e32 v18, v1, v27                                                   ; 10243701
>	v_mul_f32_e32 v15, v6, v15                                                   ; 101E1F06
>	v_add_f32_e64 v15, 0, v15 clamp                                              ; D206080F 00021E80
>	v_add_f32_e64 v16, 0, v16 clamp                                              ; D2060810 00022080
>	v_mad_f32 v19, -2.0, v15, v4                                                 ; D2820013 04121EF5
>	v_mul_f32_e32 v15, v15, v15                                                  ; 101E1F0F
>	v_mul_f32_e32 v15, v15, v19                                                  ; 101E270F
>	v_mad_f32 v19, -2.0, v16, v4                                                 ; D2820013 041220F5
>	v_mul_f32_e32 v16, v16, v16                                                  ; 10202110
>	v_mul_f32_e32 v16, v16, v19                                                  ; 10202710
>	v_mul_f32_e32 v15, v15, v16                                                  ; 101E210F
>	image_sample v[19:22], 15, 0, 0, 0, 0, 0, 0, 0, v[9:10], s[20:27], s[16:19]  ; F0800F00 00851309
>	image_sample v9, 2, 0, 0, 0, 0, 0, 0, 0, v[13:14], s[8:15], s[0:3]           ; F0800200 0002090D
>	v_mul_f32_e32 v10, v3, v27                                                   ; 10143703
>	s_waitcnt vmcnt(1)                                                           ; BF8C0771
>	v_mad_f32 v0, -v27, v0, v19                                                  ; D2820000 244E011B
>	v_mad_f32 v1, -v27, v1, v20                                                  ; D2820001 2452031B
>	v_mad_f32 v3, -v27, v3, v21                                                  ; D2820003 2456071B
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mac_f32_e32 v29, v26, v9                                                   ; 3E3A131A
>	v_mad_f32 v7, v7, s52, 1.0                                                   ; D2820007 03C86907
>	v_mac_f32_e32 v15, v7, v29                                                   ; 3E1E3B07
>	v_mul_f32_e32 v7, s66, v27                                                   ; 100E3642
>	v_add_f32_e32 v9, v11, v12                                                   ; 0612190B
>	v_add_f32_e64 v5, 0, v5 clamp                                                ; D2060805 00020A80
>	v_mad_f32 v11, -2.0, v5, v4                                                  ; D282000B 04120AF5
>	v_mul_f32_e32 v5, v5, v5                                                     ; 100A0B05
>	v_mul_f32_e32 v5, v11, v5                                                    ; 100A0B0B
>	v_add_f32_e64 v9, 0, v9 clamp                                                ; D2060809 00021280
>	v_sub_f32_e32 v9, 1.0, v9                                                    ; 081212F2
>	v_sub_f32_e32 v9, 1.0, v9                                                    ; 081212F2
>	v_max_f32_e32 v5, v5, v9                                                     ; 200A1305
>	v_mac_f32_e32 v7, v7, v2                                                     ; 3E0E0507
>	v_madak_f32_e32 v2, v15, v5, 0xbf147ae1                                      ; 42040B0F BF147AE1
>	v_mul_f32_e32 v2, v6, v2                                                     ; 10040506
>	v_add_f32_e64 v2, 0, v2 clamp                                                ; D2060802 00020480
>	v_mad_f32 v5, -2.0, v2, v4                                                   ; D2820005 041204F5
>	v_mul_f32_e32 v2, v2, v2                                                     ; 10040502
>	v_mul_f32_e32 v9, v5, v2                                                     ; 10120505
>	v_madak_f32_e32 v11, v2, v5, 0xbf666666                                      ; 42160B02 BF666666
>	v_madak_f32_e32 v2, v2, v5, 0xbc23d70a                                       ; 42040B02 BC23D70A
>	v_mul_f32_e32 v5, v6, v11                                                    ; 100A1706
>	v_mul_f32_e32 v2, v2, v8                                                     ; 10041102
>	v_mac_f32_e32 v17, v0, v9                                                    ; 3E221300
>	v_mac_f32_e32 v18, v1, v9                                                    ; 3E241301
>	v_mac_f32_e32 v10, v3, v9                                                    ; 3E141303
>	v_max_f32_e32 v0, 0, v2                                                      ; 20000480
>	v_add_f32_e64 v1, 0, v5 clamp                                                ; D2060801 00020A80
>	v_mad_f32 v2, -2.0, v0, v4                                                   ; D2820002 041200F5
>	v_mul_f32_e32 v0, v0, v0                                                     ; 10000100
>	v_mul_f32_e32 v0, v2, v0                                                     ; 10000102
>	v_mul_f32_e32 v2, v7, v0                                                     ; 10040107
>	v_mad_f32 v0, -v0, v7, v22                                                   ; D2820000 245A0F00
>	v_mac_f32_e32 v4, -2.0, v1                                                   ; 3E0802F5
>	v_mul_f32_e32 v1, v1, v1                                                     ; 10020301
>	v_mul_f32_e32 v1, v4, v1                                                     ; 10020304
>	v_mac_f32_e32 v2, v0, v1                                                     ; 3E040300
>	v_cvt_pkrtz_f16_f32_e32 v0, v17, v18                                         ; 5E002511
>	v_cvt_pkrtz_f16_f32_e32 v1, v10, v2                                          ; 5E02050A
>	exp 15, 0, 1, 1, 1, v0, v1, v0, v1                                           ; F8001C0F 01000100
>	s_endpgm                                                                     ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 96
>VGPRS: 32
>Code Size: 1040 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
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL OUT[1], CLIPVERTEX
>DCL OUT[2], GENERIC[0]
>DCL OUT[3], GENERIC[1]
>DCL CONST[0..54]
>DCL TEMP[0..3], LOCAL
>  0: MUL TEMP[0].xy, CONST[48].xyyy, IN[1].xyyy
>  1: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
>  2: ADD TEMP[1].x, TEMP[0].xxxx, CONST[48].wwww
>  3: MOV TEMP[1].z, TEMP[1].xxxx
>  4: MUL TEMP[0].xy, CONST[49].xyyy, IN[1].xyyy
>  5: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
>  6: ADD TEMP[2].x, TEMP[0].xxxx, CONST[49].wwww
>  7: MOV TEMP[1].w, TEMP[2].xxxx
>  8: MUL TEMP[0].xy, CONST[50].xyyy, IN[1].xyyy
>  9: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 10: ADD TEMP[2].x, TEMP[0].xxxx, CONST[50].wwww
> 11: MUL TEMP[0].xy, CONST[51].xyyy, IN[1].xyyy
> 12: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 13: ADD TEMP[3].x, TEMP[0].xxxx, CONST[51].wwww
> 14: MOV TEMP[2].y, TEMP[3].xxxx
> 15: MUL TEMP[0].xy, CONST[52].xyyy, IN[1].xyyy
> 16: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 17: ADD TEMP[3].x, TEMP[0].xxxx, CONST[52].wwww
> 18: MOV TEMP[2].z, TEMP[3].xxxx
> 19: MUL TEMP[0].xy, CONST[53].xyyy, IN[1].xyyy
> 20: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 21: ADD TEMP[0].x, TEMP[0].xxxx, CONST[53].wwww
> 22: MOV TEMP[2].w, TEMP[0].xxxx
> 23: MOV TEMP[0].xw, IN[0].xxxw
> 24: MOV TEMP[1].xy, IN[1].xyxx
> 25: MAD TEMP[3].x, IN[0].zzzz, CONST[0].zzzz, -IN[0].wwww
> 26: MOV TEMP[0].z, TEMP[3].xxxx
> 27: MOV TEMP[0].y, -IN[0].yyyy
> 28: MAD TEMP[0].xy, CONST[54].xyyy, IN[0].wwww, TEMP[0].xyyy
> 29: MOV OUT[2], TEMP[1]
> 30: MOV OUT[3], TEMP[2]
> 31: MOV OUT[0], TEMP[0]
> 32: MOV OUT[1], IN[0]
> 33: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = call float @llvm.SI.load.const(<16 x i8> %13, i32 8)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %13, i32 768)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %13, i32 772)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %13, i32 780)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %13, i32 784)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %13, i32 788)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %13, i32 796)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %13, i32 800)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %13, i32 804)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %13, i32 812)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %13, i32 816)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %13, i32 820)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %13, i32 828)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %13, i32 832)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %13, i32 836)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %13, i32 844)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %13, i32 848)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %13, i32 852)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %13, i32 860)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %13, i32 864)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %13, i32 868)
>  %35 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %36 = load <16 x i8>, <16 x i8> addrspace(2)* %35, align 16, !tbaa !0
>  %37 = add i32 %5, %8
>  %38 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %36, i32 0, i32 %37)
>  %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 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 1
>  %44 = load <16 x i8>, <16 x i8> addrspace(2)* %43, align 16, !tbaa !0
>  %45 = add i32 %5, %8
>  %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 = fmul float %15, %47
>  %50 = fmul float %16, %48
>  %51 = fadd float %50, %49
>  %52 = fadd float %51, %17
>  %53 = fmul float %18, %47
>  %54 = fmul float %19, %48
>  %55 = fadd float %54, %53
>  %56 = fadd float %55, %20
>  %57 = fmul float %21, %47
>  %58 = fmul float %22, %48
>  %59 = fadd float %58, %57
>  %60 = fadd float %59, %23
>  %61 = fmul float %24, %47
>  %62 = fmul float %25, %48
>  %63 = fadd float %62, %61
>  %64 = fadd float %63, %26
>  %65 = fmul float %27, %47
>  %66 = fmul float %28, %48
>  %67 = fadd float %66, %65
>  %68 = fadd float %67, %29
>  %69 = fmul float %30, %47
>  %70 = fmul float %31, %48
>  %71 = fadd float %70, %69
>  %72 = fadd float %71, %32
>  %73 = fmul float %41, %14
>  %74 = fsub float %73, %42
>  %75 = fmul float %33, %42
>  %76 = fadd float %75, %39
>  %77 = fmul float %34, %42
>  %78 = fsub float %77, %40
>  %79 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 16
>  %80 = load <16 x i8>, <16 x i8> addrspace(2)* %79, align 16, !tbaa !0
>  %81 = call float @llvm.SI.load.const(<16 x i8> %80, i32 0)
>  %82 = fmul float %81, %39
>  %83 = call float @llvm.SI.load.const(<16 x i8> %80, i32 4)
>  %84 = fmul float %83, %40
>  %85 = fadd float %82, %84
>  %86 = call float @llvm.SI.load.const(<16 x i8> %80, i32 8)
>  %87 = fmul float %86, %41
>  %88 = fadd float %85, %87
>  %89 = call float @llvm.SI.load.const(<16 x i8> %80, i32 12)
>  %90 = fmul float %89, %42
>  %91 = fadd float %88, %90
>  %92 = call float @llvm.SI.load.const(<16 x i8> %80, i32 16)
>  %93 = fmul float %92, %39
>  %94 = call float @llvm.SI.load.const(<16 x i8> %80, i32 20)
>  %95 = fmul float %94, %40
>  %96 = fadd float %93, %95
>  %97 = call float @llvm.SI.load.const(<16 x i8> %80, i32 24)
>  %98 = fmul float %97, %41
>  %99 = fadd float %96, %98
>  %100 = call float @llvm.SI.load.const(<16 x i8> %80, i32 28)
>  %101 = fmul float %100, %42
>  %102 = fadd float %99, %101
>  %103 = call float @llvm.SI.load.const(<16 x i8> %80, i32 32)
>  %104 = fmul float %103, %39
>  %105 = call float @llvm.SI.load.const(<16 x i8> %80, i32 36)
>  %106 = fmul float %105, %40
>  %107 = fadd float %104, %106
>  %108 = call float @llvm.SI.load.const(<16 x i8> %80, i32 40)
>  %109 = fmul float %108, %41
>  %110 = fadd float %107, %109
>  %111 = call float @llvm.SI.load.const(<16 x i8> %80, i32 44)
>  %112 = fmul float %111, %42
>  %113 = fadd float %110, %112
>  %114 = call float @llvm.SI.load.const(<16 x i8> %80, i32 48)
>  %115 = fmul float %114, %39
>  %116 = call float @llvm.SI.load.const(<16 x i8> %80, i32 52)
>  %117 = fmul float %116, %40
>  %118 = fadd float %115, %117
>  %119 = call float @llvm.SI.load.const(<16 x i8> %80, i32 56)
>  %120 = fmul float %119, %41
>  %121 = fadd float %118, %120
>  %122 = call float @llvm.SI.load.const(<16 x i8> %80, i32 60)
>  %123 = fmul float %122, %42
>  %124 = fadd float %121, %123
>  %125 = call float @llvm.SI.load.const(<16 x i8> %80, i32 64)
>  %126 = fmul float %125, %39
>  %127 = call float @llvm.SI.load.const(<16 x i8> %80, i32 68)
>  %128 = fmul float %127, %40
>  %129 = fadd float %126, %128
>  %130 = call float @llvm.SI.load.const(<16 x i8> %80, i32 72)
>  %131 = fmul float %130, %41
>  %132 = fadd float %129, %131
>  %133 = call float @llvm.SI.load.const(<16 x i8> %80, i32 76)
>  %134 = fmul float %133, %42
>  %135 = fadd float %132, %134
>  %136 = call float @llvm.SI.load.const(<16 x i8> %80, i32 80)
>  %137 = fmul float %136, %39
>  %138 = call float @llvm.SI.load.const(<16 x i8> %80, i32 84)
>  %139 = fmul float %138, %40
>  %140 = fadd float %137, %139
>  %141 = call float @llvm.SI.load.const(<16 x i8> %80, i32 88)
>  %142 = fmul float %141, %41
>  %143 = fadd float %140, %142
>  %144 = call float @llvm.SI.load.const(<16 x i8> %80, i32 92)
>  %145 = fmul float %144, %42
>  %146 = fadd float %143, %145
>  %147 = call float @llvm.SI.load.const(<16 x i8> %80, i32 96)
>  %148 = fmul float %147, %39
>  %149 = call float @llvm.SI.load.const(<16 x i8> %80, i32 100)
>  %150 = fmul float %149, %40
>  %151 = fadd float %148, %150
>  %152 = call float @llvm.SI.load.const(<16 x i8> %80, i32 104)
>  %153 = fmul float %152, %41
>  %154 = fadd float %151, %153
>  %155 = call float @llvm.SI.load.const(<16 x i8> %80, i32 108)
>  %156 = fmul float %155, %42
>  %157 = fadd float %154, %156
>  %158 = call float @llvm.SI.load.const(<16 x i8> %80, i32 112)
>  %159 = fmul float %158, %39
>  %160 = call float @llvm.SI.load.const(<16 x i8> %80, i32 116)
>  %161 = fmul float %160, %40
>  %162 = fadd float %159, %161
>  %163 = call float @llvm.SI.load.const(<16 x i8> %80, i32 120)
>  %164 = fmul float %163, %41
>  %165 = fadd float %162, %164
>  %166 = call float @llvm.SI.load.const(<16 x i8> %80, i32 124)
>  %167 = fmul float %166, %42
>  %168 = fadd float %165, %167
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %47, float %48, float %52, float %56)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %60, float %64, float %68, float %72)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 12, i32 0, float %76, float %78, float %74, float %42)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 13, i32 0, float %91, float %102, float %113, float %124)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 14, i32 0, float %135, float %146, float %157, 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
>
>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[4:7], s[8:9], 0x0                    ; C0820900
>	s_load_dwordx4 s[12:15], s[8:9], 0x4                  ; C0860904
>	v_add_i32_e32 v0, vcc, s10, v0                        ; 4A00000A
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[4:7], 0 idxen   ; E00C2000 80010100
>	buffer_load_format_xyzw v[5:8], v0, s[12:15], 0 idxen ; E00C2000 80030500
>	s_load_dwordx4 s[4:7], s[2:3], 0x0                    ; C0820300
>	s_load_dwordx4 s[0:3], s[2:3], 0x40                   ; C0800340
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	s_buffer_load_dword s8, s[4:7], 0x2                   ; C2040502
>	s_buffer_load_dword s9, s[4:7], 0xc0                  ; C20485C0
>	s_buffer_load_dword s10, s[4:7], 0xc1                 ; C20505C1
>	s_buffer_load_dword s11, s[4:7], 0xc3                 ; C20585C3
>	s_buffer_load_dword s12, s[4:7], 0xc4                 ; C20605C4
>	s_buffer_load_dword s13, s[4:7], 0xc5                 ; C20685C5
>	s_buffer_load_dword s14, s[4:7], 0xc7                 ; C20705C7
>	s_buffer_load_dword s15, s[4:7], 0xc8                 ; C20785C8
>	s_buffer_load_dword s16, s[4:7], 0xc9                 ; C20805C9
>	s_buffer_load_dword s17, s[4:7], 0xcb                 ; C20885CB
>	s_buffer_load_dword s18, s[4:7], 0xcc                 ; C20905CC
>	s_buffer_load_dword s19, s[4:7], 0xcd                 ; C20985CD
>	s_buffer_load_dword s20, s[4:7], 0xcf                 ; C20A05CF
>	s_buffer_load_dword s21, s[4:7], 0xd0                 ; C20A85D0
>	s_buffer_load_dword s22, s[4:7], 0xd1                 ; C20B05D1
>	s_buffer_load_dword s23, s[4:7], 0xd3                 ; C20B85D3
>	s_buffer_load_dword s24, s[4:7], 0xd4                 ; C20C05D4
>	s_buffer_load_dword s25, s[4:7], 0xd5                 ; C20C85D5
>	s_buffer_load_dword s26, s[4:7], 0xd7                 ; C20D05D7
>	s_buffer_load_dword s27, s[4:7], 0xd8                 ; C20D85D8
>	s_buffer_load_dword s4, s[4:7], 0xd9                  ; C20205D9
>	s_buffer_load_dword s5, s[0:3], 0x0                   ; C2028100
>	s_buffer_load_dword s6, s[0:3], 0x1                   ; C2030101
>	s_buffer_load_dword s7, s[0:3], 0x2                   ; C2038102
>	s_buffer_load_dword s28, s[0:3], 0x3                  ; C20E0103
>	s_buffer_load_dword s29, s[0:3], 0x4                  ; C20E8104
>	s_buffer_load_dword s30, s[0:3], 0x5                  ; C20F0105
>	s_buffer_load_dword s31, s[0:3], 0x6                  ; C20F8106
>	s_buffer_load_dword s32, s[0:3], 0x7                  ; C2100107
>	s_buffer_load_dword s33, s[0:3], 0x8                  ; C2108108
>	s_buffer_load_dword s34, s[0:3], 0x9                  ; C2110109
>	s_buffer_load_dword s35, s[0:3], 0xa                  ; C211810A
>	s_buffer_load_dword s36, s[0:3], 0xb                  ; C212010B
>	s_buffer_load_dword s37, s[0:3], 0xc                  ; C212810C
>	s_buffer_load_dword s38, s[0:3], 0xd                  ; C213010D
>	s_buffer_load_dword s39, s[0:3], 0xe                  ; C213810E
>	s_buffer_load_dword s40, s[0:3], 0xf                  ; C214010F
>	s_buffer_load_dword s41, s[0:3], 0x10                 ; C2148110
>	s_buffer_load_dword s42, s[0:3], 0x11                 ; C2150111
>	s_buffer_load_dword s43, s[0:3], 0x12                 ; C2158112
>	s_buffer_load_dword s44, s[0:3], 0x13                 ; C2160113
>	s_buffer_load_dword s45, s[0:3], 0x14                 ; C2168114
>	s_buffer_load_dword s46, s[0:3], 0x15                 ; C2170115
>	s_buffer_load_dword s47, s[0:3], 0x16                 ; C2178116
>	s_buffer_load_dword s48, s[0:3], 0x17                 ; C2180117
>	s_buffer_load_dword s49, s[0:3], 0x18                 ; C2188118
>	s_buffer_load_dword s50, s[0:3], 0x19                 ; C2190119
>	s_buffer_load_dword s51, s[0:3], 0x1a                 ; C219811A
>	s_buffer_load_dword s52, s[0:3], 0x1b                 ; C21A011B
>	s_buffer_load_dword s53, s[0:3], 0x1c                 ; C21A811C
>	s_buffer_load_dword s54, s[0:3], 0x1d                 ; C21B011D
>	s_buffer_load_dword s55, s[0:3], 0x1e                 ; C21B811E
>	s_buffer_load_dword s0, s[0:3], 0x1f                  ; C200011F
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mov_b32_e32 v0, s11                                 ; 7E00020B
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mov_b32_e32 v7, s14                                 ; 7E0E020E
>	v_mov_b32_e32 v8, s17                                 ; 7E100211
>	v_mov_b32_e32 v9, s20                                 ; 7E120214
>	v_mov_b32_e32 v10, s23                                ; 7E140217
>	v_mov_b32_e32 v11, s26                                ; 7E16021A
>	v_mac_f32_e32 v0, s9, v5                              ; 3E000A09
>	v_mac_f32_e32 v7, s12, v5                             ; 3E0E0A0C
>	v_mac_f32_e32 v8, s15, v5                             ; 3E100A0F
>	v_mac_f32_e32 v9, s18, v5                             ; 3E120A12
>	v_mac_f32_e32 v10, s21, v5                            ; 3E140A15
>	v_mac_f32_e32 v11, s24, v5                            ; 3E160A18
>	v_mac_f32_e32 v0, s10, v6                             ; 3E000C0A
>	v_mac_f32_e32 v7, s13, v6                             ; 3E0E0C0D
>	v_mac_f32_e32 v8, s16, v6                             ; 3E100C10
>	v_mac_f32_e32 v9, s19, v6                             ; 3E120C13
>	exp 15, 32, 0, 0, 0, v5, v6, v0, v7                   ; F800020F 07000605
>	v_mac_f32_e32 v10, s22, v6                            ; 3E140C16
>	v_mac_f32_e32 v11, s25, v6                            ; 3E160C19
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mad_f32 v0, v3, s8, -v4                             ; D2820000 84101103
>	v_mad_f32 v5, s27, v4, v1                             ; D2820005 0406081B
>	v_mad_f32 v6, s4, v4, -v2                             ; D2820006 840A0804
>	v_mul_f32_e32 v7, s6, v2                              ; 100E0406
>	exp 15, 33, 0, 0, 0, v8, v9, v10, v11                 ; F800021F 0B0A0908
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mul_f32_e32 v8, s30, v2                             ; 1010041E
>	v_mac_f32_e32 v7, s5, v1                              ; 3E0E0205
>	exp 15, 12, 0, 0, 0, v5, v6, v0, v4                   ; F80000CF 04000605
>	v_mac_f32_e32 v8, s29, v1                             ; 3E10021D
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mul_f32_e32 v0, s34, v2                             ; 10000422
>	v_mac_f32_e32 v0, s33, v1                             ; 3E000221
>	v_mul_f32_e32 v5, s38, v2                             ; 100A0426
>	v_mac_f32_e32 v5, s37, v1                             ; 3E0A0225
>	v_mul_f32_e32 v6, s42, v2                             ; 100C042A
>	v_mul_f32_e32 v9, s46, v2                             ; 1012042E
>	v_mul_f32_e32 v10, s50, v2                            ; 10140432
>	v_mul_f32_e32 v2, s54, v2                             ; 10040436
>	v_mac_f32_e32 v6, s41, v1                             ; 3E0C0229
>	v_mac_f32_e32 v9, s45, v1                             ; 3E12022D
>	v_mac_f32_e32 v10, s49, v1                            ; 3E140231
>	v_mac_f32_e32 v2, s53, v1                             ; 3E040235
>	v_mac_f32_e32 v7, s7, v3                              ; 3E0E0607
>	v_mac_f32_e32 v8, s31, v3                             ; 3E10061F
>	v_mac_f32_e32 v0, s35, v3                             ; 3E000623
>	v_mac_f32_e32 v5, s39, v3                             ; 3E0A0627
>	v_mac_f32_e32 v6, s43, v3                             ; 3E0C062B
>	v_mac_f32_e32 v9, s47, v3                             ; 3E12062F
>	v_mac_f32_e32 v10, s51, v3                            ; 3E140633
>	v_mac_f32_e32 v2, s55, v3                             ; 3E040637
>	v_mac_f32_e32 v7, s28, v4                             ; 3E0E081C
>	v_mac_f32_e32 v8, s32, v4                             ; 3E100820
>	v_mac_f32_e32 v0, s36, v4                             ; 3E000824
>	v_mac_f32_e32 v5, s40, v4                             ; 3E0A0828
>	v_mac_f32_e32 v6, s44, v4                             ; 3E0C082C
>	v_mac_f32_e32 v9, s48, v4                             ; 3E120830
>	v_mac_f32_e32 v10, s52, v4                            ; 3E140834
>	v_mac_f32_e32 v2, s0, v4                              ; 3E040800
>	exp 15, 13, 0, 0, 0, v7, v8, v0, v5                   ; F80000DF 05000807
>	exp 15, 14, 0, 1, 0, v6, v9, v10, v2                  ; F80008EF 020A0906
>	s_endpgm                                              ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 64
>VGPRS: 12
>Code Size: 544 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
>  clamp_color = 0
>FRAG
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL IN[1], GENERIC[1], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SAMP[1]
>DCL SAMP[2]
>DCL SAMP[3]
>DCL SAMP[4]
>DCL SVIEW[0], 2D, FLOAT
>DCL SVIEW[1], 2D, FLOAT
>DCL SVIEW[2], 2D, FLOAT
>DCL SVIEW[3], 2D, FLOAT
>DCL SVIEW[4], 2D, FLOAT
>DCL CONST[0..3]
>DCL TEMP[0..4], LOCAL
>IMM[0] FLT32 {   -0.5000,    -1.0000,    10.0000,   -10.0000}
>IMM[1] FLT32 {   -2.0000,     3.0000,     6.0000,     1.0000}
>IMM[2] FLT32 {   -0.9900,   100.0000,     1.0000,     0.0000}
>  0: MOV TEMP[0].xy, IN[0].xyyy
>  1: TEX TEMP[0].xw, TEMP[0], SAMP[0], 2D
>  2: MOV TEMP[1].xy, IN[1].xyyy
>  3: TEX TEMP[1].y, TEMP[1], SAMP[1], 2D
>  4: MAD TEMP[0].x, TEMP[1].yyyy, TEMP[0].xxxx, TEMP[0].wwww
>  5: MOV TEMP[1].xy, IN[0].zwww
>  6: TEX TEMP[1].w, TEMP[1], SAMP[4], 2D
>  7: ADD TEMP[2].xy, TEMP[1].wwww, IMM[0].xyyy
>  8: ADD TEMP[3].x, TEMP[1].wwww, TEMP[1].wwww
>  9: MOV_SAT TEMP[3].x, TEMP[3].xxxx
> 10: MUL TEMP[2].xy, TEMP[2].xyyy, IMM[0].zwww
> 11: MOV_SAT TEMP[2].xy, TEMP[2].xyyy
> 12: MAD TEMP[1].xy, TEMP[2].xyyy, IMM[1].xxxx, IMM[1].yyyy
> 13: MUL TEMP[2].xy, TEMP[2].xyyy, TEMP[2].xyyy
> 14: MUL TEMP[1].xy, TEMP[2].xyyy, TEMP[1].xyyy
> 15: MUL TEMP[1].x, TEMP[1].yyyy, TEMP[1].xxxx
> 16: MAD TEMP[2].x, CONST[3].wwww, IMM[1].zzzz, IMM[1].wwww
> 17: MAD TEMP[0].x, TEMP[0].xxxx, TEMP[2].xxxx, TEMP[1].xxxx
> 18: ADD TEMP[0].x, TEMP[0].xxxx, IMM[2].xxxx
> 19: MUL TEMP[0].x, TEMP[0].xxxx, IMM[2].yyyy
> 20: MOV_SAT TEMP[1].x, TEMP[0].xxxx
> 21: MAD TEMP[2].x, TEMP[1].xxxx, IMM[1].xxxx, IMM[1].yyyy
> 22: MUL TEMP[0].x, TEMP[1].xxxx, TEMP[1].xxxx
> 23: MAD TEMP[0].x, TEMP[2].xxxx, TEMP[0].xxxx, IMM[1].wwww
> 24: MOV TEMP[1].xy, IN[0].xyyy
> 25: TEX TEMP[1].xyz, TEMP[1], SAMP[3], 2D
> 26: ADD TEMP[0].x, TEMP[0].xxxx, -TEMP[1].xxxx
> 27: MOV_SAT TEMP[0].x, TEMP[0].xxxx
> 28: MAX TEMP[1].x, TEMP[1].yyyy, TEMP[1].zzzz
> 29: MAX TEMP[1].x, TEMP[1].xxxx, TEMP[0].xxxx
> 30: MOV TEMP[2].xy, IN[0].xyyy
> 31: TEX TEMP[2].xyz, TEMP[2], SAMP[2], 2D
> 32: ADD TEMP[4].x, -TEMP[3].xxxx, TEMP[2].xxxx
> 33: MAD TEMP[1].x, TEMP[1].xxxx, TEMP[4].xxxx, TEMP[3].xxxx
> 34: ADD TEMP[3].x, TEMP[2].yyyy, IMM[0].yyyy
> 35: MAD TEMP[2].xy, TEMP[2].zzzz, IMM[2].zwww, IMM[2].wzzz
> 36: MOV TEMP[1].zw, TEMP[2].yyxy
> 37: MAD TEMP[0].x, TEMP[0].xxxx, TEMP[3].xxxx, IMM[1].wwww
> 38: MOV TEMP[1].y, TEMP[0].xxxx
> 39: MOV OUT[0], TEMP[1]
> 40: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %24 = load <16 x i8>, <16 x i8> addrspace(2)* %23, align 16, !tbaa !0
>  %25 = call float @llvm.SI.load.const(<16 x i8> %24, i32 60)
>  %26 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %27 = load <8 x i32>, <8 x i32> addrspace(2)* %26, align 32, !tbaa !0
>  %28 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %29 = load <4 x i32>, <4 x i32> addrspace(2)* %28, align 16, !tbaa !0
>  %30 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 1
>  %31 = load <8 x i32>, <8 x i32> addrspace(2)* %30, align 32, !tbaa !0
>  %32 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 1
>  %33 = load <4 x i32>, <4 x i32> addrspace(2)* %32, align 16, !tbaa !0
>  %34 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 2
>  %35 = load <8 x i32>, <8 x i32> addrspace(2)* %34, align 32, !tbaa !0
>  %36 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 2
>  %37 = load <4 x i32>, <4 x i32> addrspace(2)* %36, align 16, !tbaa !0
>  %38 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 3
>  %39 = load <8 x i32>, <8 x i32> addrspace(2)* %38, align 32, !tbaa !0
>  %40 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 3
>  %41 = load <4 x i32>, <4 x i32> addrspace(2)* %40, align 16, !tbaa !0
>  %42 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 4
>  %43 = load <8 x i32>, <8 x i32> addrspace(2)* %42, align 32, !tbaa !0
>  %44 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 4
>  %45 = load <4 x i32>, <4 x i32> addrspace(2)* %44, align 16, !tbaa !0
>  %46 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %8)
>  %47 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %8)
>  %48 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %6, <2 x i32> %8)
>  %49 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %6, <2 x i32> %8)
>  %50 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %6, <2 x i32> %8)
>  %51 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %6, <2 x i32> %8)
>  %52 = bitcast float %46 to i32
>  %53 = bitcast float %47 to i32
>  %54 = insertelement <2 x i32> undef, i32 %52, i32 0
>  %55 = insertelement <2 x i32> %54, i32 %53, i32 1
>  %56 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %55, <8 x i32> %27, <4 x i32> %29, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %57 = extractelement <4 x float> %56, i32 0
>  %58 = extractelement <4 x float> %56, i32 3
>  %59 = bitcast float %50 to i32
>  %60 = bitcast float %51 to i32
>  %61 = insertelement <2 x i32> undef, i32 %59, i32 0
>  %62 = insertelement <2 x i32> %61, i32 %60, i32 1
>  %63 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %62, <8 x i32> %31, <4 x i32> %33, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %64 = extractelement <4 x float> %63, i32 1
>  %65 = fmul float %64, %57
>  %66 = fadd float %65, %58
>  %67 = bitcast float %48 to i32
>  %68 = bitcast float %49 to i32
>  %69 = insertelement <2 x i32> undef, i32 %67, i32 0
>  %70 = insertelement <2 x i32> %69, i32 %68, i32 1
>  %71 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %70, <8 x i32> %43, <4 x i32> %45, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %72 = extractelement <4 x float> %71, i32 3
>  %73 = fadd float %72, -5.000000e-01
>  %74 = fadd float %72, -1.000000e+00
>  %75 = fadd float %72, %72
>  %76 = call float @llvm.AMDIL.clamp.(float %75, float 0.000000e+00, float 1.000000e+00)
>  %77 = fmul float %73, 1.000000e+01
>  %78 = fmul float %74, -1.000000e+01
>  %79 = call float @llvm.AMDIL.clamp.(float %77, float 0.000000e+00, float 1.000000e+00)
>  %80 = call float @llvm.AMDIL.clamp.(float %78, float 0.000000e+00, float 1.000000e+00)
>  %81 = fmul float %79, -2.000000e+00
>  %82 = fadd float %81, 3.000000e+00
>  %83 = fmul float %80, -2.000000e+00
>  %84 = fadd float %83, 3.000000e+00
>  %85 = fmul float %79, %79
>  %86 = fmul float %80, %80
>  %87 = fmul float %85, %82
>  %88 = fmul float %86, %84
>  %89 = fmul float %88, %87
>  %90 = fmul float %25, 6.000000e+00
>  %91 = fadd float %90, 1.000000e+00
>  %92 = fmul float %66, %91
>  %93 = fadd float %92, %89
>  %94 = fadd float %93, 0xBFEFAE1480000000
>  %95 = fmul float %94, 1.000000e+02
>  %96 = call float @llvm.AMDIL.clamp.(float %95, float 0.000000e+00, float 1.000000e+00)
>  %97 = fmul float %96, -2.000000e+00
>  %98 = fadd float %97, 3.000000e+00
>  %99 = fmul float %96, %96
>  %100 = fmul float %98, %99
>  %101 = fadd float %100, 1.000000e+00
>  %102 = bitcast float %46 to i32
>  %103 = bitcast float %47 to i32
>  %104 = insertelement <2 x i32> undef, i32 %102, i32 0
>  %105 = insertelement <2 x i32> %104, i32 %103, i32 1
>  %106 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %105, <8 x i32> %39, <4 x i32> %41, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %107 = extractelement <4 x float> %106, i32 0
>  %108 = extractelement <4 x float> %106, i32 1
>  %109 = extractelement <4 x float> %106, i32 2
>  %110 = fsub float %101, %107
>  %111 = call float @llvm.AMDIL.clamp.(float %110, float 0.000000e+00, float 1.000000e+00)
>  %112 = call float @llvm.maxnum.f32(float %108, float %109)
>  %113 = call float @llvm.maxnum.f32(float %112, float %111)
>  %114 = bitcast float %46 to i32
>  %115 = bitcast float %47 to i32
>  %116 = insertelement <2 x i32> undef, i32 %114, i32 0
>  %117 = insertelement <2 x i32> %116, i32 %115, i32 1
>  %118 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %117, <8 x i32> %35, <4 x i32> %37, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %119 = extractelement <4 x float> %118, i32 0
>  %120 = extractelement <4 x float> %118, i32 1
>  %121 = extractelement <4 x float> %118, i32 2
>  %122 = fsub float %119, %76
>  %123 = fmul float %113, %122
>  %124 = fadd float %123, %76
>  %125 = fadd float %120, -1.000000e+00
>  %126 = fadd float %121, 0.000000e+00
>  %127 = fmul float %121, 0.000000e+00
>  %128 = fadd float %127, 1.000000e+00
>  %129 = fmul float %111, %125
>  %130 = fadd float %129, 1.000000e+00
>  %131 = call i32 @llvm.SI.packf16(float %124, float %130)
>  %132 = bitcast i32 %131 to float
>  %133 = call i32 @llvm.SI.packf16(float %126, float %128)
>  %134 = bitcast i32 %133 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %132, float %134, float %132, float %134)
>  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.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #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 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, s10                                                       ; BEFC030A
>	v_mov_b32_e32 v2, 0x41200000                                            ; 7E0402FF 41200000
>	v_mov_b32_e32 v3, 0x40400000                                            ; 7E0602FF 40400000
>	v_mov_b32_e32 v4, 0x40c00000                                            ; 7E0802FF 40C00000
>	v_interp_p1_f32 v5, v0, 0, 0, [m0]                                      ; C8140000
>	v_interp_p2_f32 v5, [v5], v1, 0, 0, [m0]                                ; C8150001
>	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, 2, 0, [m0]                                      ; C81C0200
>	v_interp_p2_f32 v7, [v7], v1, 2, 0, [m0]                                ; C81D0201
>	v_interp_p1_f32 v8, v0, 3, 0, [m0]                                      ; C8200300
>	v_interp_p2_f32 v8, [v8], v1, 3, 0, [m0]                                ; C8210301
>	v_interp_p1_f32 v9, v0, 0, 1, [m0]                                      ; C8240400
>	v_interp_p2_f32 v9, [v9], v1, 0, 1, [m0]                                ; C8250401
>	v_interp_p1_f32 v10, v0, 1, 1, [m0]                                     ; C8280500
>	v_interp_p2_f32 v10, [v10], v1, 1, 1, [m0]                              ; C8290501
>	s_load_dwordx4 s[28:31], s[2:3], 0x0                                    ; C08E0300
>	s_load_dwordx4 s[32:35], s[4:5], 0x0                                    ; C0900500
>	s_load_dwordx4 s[36:39], s[4:5], 0x4                                    ; C0920504
>	s_load_dwordx4 s[0:3], s[4:5], 0x8                                      ; C0800508
>	s_load_dwordx4 s[16:19], s[4:5], 0xc                                    ; C088050C
>	s_load_dwordx4 s[40:43], s[4:5], 0x10                                   ; C0940510
>	s_load_dwordx8 s[44:51], s[6:7], 0x0                                    ; C0D60700
>	s_load_dwordx8 s[52:59], s[6:7], 0x8                                    ; C0DA0708
>	s_load_dwordx8 s[8:15], s[6:7], 0x10                                    ; C0C40710
>	s_load_dwordx8 s[20:27], s[6:7], 0x18                                   ; C0CA0718
>	s_waitcnt lgkmcnt(0)                                                    ; BF8C007F
>	s_buffer_load_dword s4, s[28:31], 0xf                                   ; C2021D0F
>	s_load_dwordx8 s[60:67], s[6:7], 0x20                                   ; C0DE0720
>	image_sample v[0:1], 9, 0, 0, 0, 0, 0, 0, 0, v[5:6], s[44:51], s[32:35] ; F0800900 010B0005
>	image_sample v9, 2, 0, 0, 0, 0, 0, 0, 0, v[9:10], s[52:59], s[36:39]    ; F0800200 012D0909
>	s_waitcnt vmcnt(0)                                                      ; BF8C0770
>	v_mac_f32_e32 v1, v0, v9                                                ; 3E021300
>	s_waitcnt lgkmcnt(0)                                                    ; BF8C007F
>	v_mad_f32 v0, v4, s4, 1.0                                               ; D2820000 03C80904
>	image_sample v4, 8, 0, 0, 0, 0, 0, 0, 0, v[7:8], s[60:67], s[40:43]     ; F0800800 014F0407
>	s_waitcnt vmcnt(0)                                                      ; BF8C0770
>	v_add_f32_e32 v7, -0.5, v4                                              ; 060E08F1
>	v_add_f32_e32 v8, v4, v4                                                ; 06100904
>	v_add_f32_e64 v8, 0, v8 clamp                                           ; D2060808 00021080
>	v_mul_f32_e32 v7, v2, v7                                                ; 100E0F02
>	v_madmk_f32_e32 v2, v4, v2, 0xc1200000                                  ; 40040504 C1200000
>	v_add_f32_e64 v4, 0, v7 clamp                                           ; D2060804 00020E80
>	v_add_f32_e64 v2, 0, v2 clamp                                           ; D2060802 00020480
>	v_mad_f32 v7, -2.0, v4, v3                                              ; D2820007 040E08F5
>	v_mul_f32_e32 v4, v4, v4                                                ; 10080904
>	v_mul_f32_e32 v4, v7, v4                                                ; 10080907
>	v_mad_f32 v7, -2.0, v2, v3                                              ; D2820007 040E04F5
>	v_mul_f32_e32 v2, v2, v2                                                ; 10040502
>	v_mul_f32_e32 v2, v7, v2                                                ; 10040507
>	v_madak_f32_e32 v2, v2, v4, 0xbf7d70a4                                  ; 42040902 BF7D70A4
>	v_mac_f32_e32 v2, v0, v1                                                ; 3E040300
>	v_mul_f32_e32 v0, 0x42c80000, v2                                        ; 100004FF 42C80000
>	v_add_f32_e64 v0, 0, v0 clamp                                           ; D2060800 00020080
>	v_mac_f32_e32 v3, -2.0, v0                                              ; 3E0600F5
>	v_mul_f32_e32 v0, v0, v0                                                ; 10000100
>	v_mad_f32 v0, v3, v0, 1.0                                               ; D2820000 03CA0103
>	image_sample v[1:3], 7, 0, 0, 0, 0, 0, 0, 0, v[5:6], s[20:27], s[16:19] ; F0800700 00850105
>	s_waitcnt vmcnt(0)                                                      ; BF8C0770
>	v_subrev_f32_e32 v0, v1, v0                                             ; 0A000101
>	v_add_f32_e64 v0, 0, v0 clamp                                           ; D2060800 00020080
>	v_max3_f32 v1, v2, v3, v0                                               ; D2A80001 04020702
>	image_sample v[2:4], 7, 0, 0, 0, 0, 0, 0, 0, v[5:6], s[8:15], s[0:3]    ; F0800700 00020205
>	s_waitcnt vmcnt(0)                                                      ; BF8C0770
>	v_subrev_f32_e32 v2, v8, v2                                             ; 0A040508
>	v_mac_f32_e32 v8, v2, v1                                                ; 3E100302
>	v_add_f32_e32 v1, -1.0, v3                                              ; 060206F3
>	v_add_f32_e32 v2, 0, v4                                                 ; 06040880
>	v_mad_f32 v3, 0, v4, 1.0                                                ; D2820003 03CA0880
>	v_mad_f32 v0, v0, v1, 1.0                                               ; D2820000 03CA0300
>	v_cvt_pkrtz_f16_f32_e32 v0, v8, v0                                      ; 5E000108
>	v_cvt_pkrtz_f16_f32_e32 v1, v2, v3                                      ; 5E020702
>	exp 15, 0, 1, 1, 1, v0, v1, v0, v1                                      ; F8001C0F 01000100
>	s_endpgm                                                                ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 72
>VGPRS: 12
>Code Size: 396 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
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL OUT[1], CLIPVERTEX
>DCL OUT[2], GENERIC[0]
>DCL OUT[3], GENERIC[1]
>DCL CONST[0..54]
>DCL TEMP[0..3], LOCAL
>  0: MUL TEMP[0].xy, CONST[48].xyyy, IN[1].xyyy
>  1: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
>  2: ADD TEMP[1].x, TEMP[0].xxxx, CONST[48].wwww
>  3: MOV TEMP[1].z, TEMP[1].xxxx
>  4: MUL TEMP[0].xy, CONST[49].xyyy, IN[1].xyyy
>  5: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
>  6: ADD TEMP[2].x, TEMP[0].xxxx, CONST[49].wwww
>  7: MOV TEMP[1].w, TEMP[2].xxxx
>  8: MUL TEMP[0].xy, CONST[50].xyyy, IN[1].xyyy
>  9: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 10: ADD TEMP[2].x, TEMP[0].xxxx, CONST[50].wwww
> 11: MUL TEMP[0].xy, CONST[51].xyyy, IN[1].xyyy
> 12: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 13: ADD TEMP[3].x, TEMP[0].xxxx, CONST[51].wwww
> 14: MOV TEMP[2].y, TEMP[3].xxxx
> 15: MUL TEMP[0].xy, CONST[52].xyyy, IN[1].xyyy
> 16: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 17: ADD TEMP[3].x, TEMP[0].xxxx, CONST[52].wwww
> 18: MOV TEMP[2].z, TEMP[3].xxxx
> 19: MUL TEMP[0].xy, CONST[53].xyyy, IN[1].xyyy
> 20: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 21: ADD TEMP[0].x, TEMP[0].xxxx, CONST[53].wwww
> 22: MOV TEMP[2].w, TEMP[0].xxxx
> 23: MOV TEMP[0].xw, IN[0].xxxw
> 24: MOV TEMP[1].xy, IN[1].xyxx
> 25: MAD TEMP[3].x, IN[0].zzzz, CONST[0].zzzz, -IN[0].wwww
> 26: MOV TEMP[0].z, TEMP[3].xxxx
> 27: MOV TEMP[0].y, -IN[0].yyyy
> 28: MAD TEMP[0].xy, CONST[54].xyyy, IN[0].wwww, TEMP[0].xyyy
> 29: MOV OUT[2], TEMP[1]
> 30: MOV OUT[3], TEMP[2]
> 31: MOV OUT[0], TEMP[0]
> 32: MOV OUT[1], IN[0]
> 33: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = call float @llvm.SI.load.const(<16 x i8> %13, i32 8)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %13, i32 768)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %13, i32 772)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %13, i32 780)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %13, i32 784)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %13, i32 788)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %13, i32 796)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %13, i32 800)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %13, i32 804)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %13, i32 812)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %13, i32 816)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %13, i32 820)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %13, i32 828)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %13, i32 832)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %13, i32 836)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %13, i32 844)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %13, i32 848)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %13, i32 852)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %13, i32 860)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %13, i32 864)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %13, i32 868)
>  %35 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %36 = load <16 x i8>, <16 x i8> addrspace(2)* %35, align 16, !tbaa !0
>  %37 = add i32 %5, %8
>  %38 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %36, i32 0, i32 %37)
>  %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 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 1
>  %44 = load <16 x i8>, <16 x i8> addrspace(2)* %43, align 16, !tbaa !0
>  %45 = add i32 %5, %8
>  %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 = fmul float %15, %47
>  %50 = fmul float %16, %48
>  %51 = fadd float %50, %49
>  %52 = fadd float %51, %17
>  %53 = fmul float %18, %47
>  %54 = fmul float %19, %48
>  %55 = fadd float %54, %53
>  %56 = fadd float %55, %20
>  %57 = fmul float %21, %47
>  %58 = fmul float %22, %48
>  %59 = fadd float %58, %57
>  %60 = fadd float %59, %23
>  %61 = fmul float %24, %47
>  %62 = fmul float %25, %48
>  %63 = fadd float %62, %61
>  %64 = fadd float %63, %26
>  %65 = fmul float %27, %47
>  %66 = fmul float %28, %48
>  %67 = fadd float %66, %65
>  %68 = fadd float %67, %29
>  %69 = fmul float %30, %47
>  %70 = fmul float %31, %48
>  %71 = fadd float %70, %69
>  %72 = fadd float %71, %32
>  %73 = fmul float %41, %14
>  %74 = fsub float %73, %42
>  %75 = fmul float %33, %42
>  %76 = fadd float %75, %39
>  %77 = fmul float %34, %42
>  %78 = fsub float %77, %40
>  %79 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 16
>  %80 = load <16 x i8>, <16 x i8> addrspace(2)* %79, align 16, !tbaa !0
>  %81 = call float @llvm.SI.load.const(<16 x i8> %80, i32 0)
>  %82 = fmul float %81, %39
>  %83 = call float @llvm.SI.load.const(<16 x i8> %80, i32 4)
>  %84 = fmul float %83, %40
>  %85 = fadd float %82, %84
>  %86 = call float @llvm.SI.load.const(<16 x i8> %80, i32 8)
>  %87 = fmul float %86, %41
>  %88 = fadd float %85, %87
>  %89 = call float @llvm.SI.load.const(<16 x i8> %80, i32 12)
>  %90 = fmul float %89, %42
>  %91 = fadd float %88, %90
>  %92 = call float @llvm.SI.load.const(<16 x i8> %80, i32 16)
>  %93 = fmul float %92, %39
>  %94 = call float @llvm.SI.load.const(<16 x i8> %80, i32 20)
>  %95 = fmul float %94, %40
>  %96 = fadd float %93, %95
>  %97 = call float @llvm.SI.load.const(<16 x i8> %80, i32 24)
>  %98 = fmul float %97, %41
>  %99 = fadd float %96, %98
>  %100 = call float @llvm.SI.load.const(<16 x i8> %80, i32 28)
>  %101 = fmul float %100, %42
>  %102 = fadd float %99, %101
>  %103 = call float @llvm.SI.load.const(<16 x i8> %80, i32 32)
>  %104 = fmul float %103, %39
>  %105 = call float @llvm.SI.load.const(<16 x i8> %80, i32 36)
>  %106 = fmul float %105, %40
>  %107 = fadd float %104, %106
>  %108 = call float @llvm.SI.load.const(<16 x i8> %80, i32 40)
>  %109 = fmul float %108, %41
>  %110 = fadd float %107, %109
>  %111 = call float @llvm.SI.load.const(<16 x i8> %80, i32 44)
>  %112 = fmul float %111, %42
>  %113 = fadd float %110, %112
>  %114 = call float @llvm.SI.load.const(<16 x i8> %80, i32 48)
>  %115 = fmul float %114, %39
>  %116 = call float @llvm.SI.load.const(<16 x i8> %80, i32 52)
>  %117 = fmul float %116, %40
>  %118 = fadd float %115, %117
>  %119 = call float @llvm.SI.load.const(<16 x i8> %80, i32 56)
>  %120 = fmul float %119, %41
>  %121 = fadd float %118, %120
>  %122 = call float @llvm.SI.load.const(<16 x i8> %80, i32 60)
>  %123 = fmul float %122, %42
>  %124 = fadd float %121, %123
>  %125 = call float @llvm.SI.load.const(<16 x i8> %80, i32 64)
>  %126 = fmul float %125, %39
>  %127 = call float @llvm.SI.load.const(<16 x i8> %80, i32 68)
>  %128 = fmul float %127, %40
>  %129 = fadd float %126, %128
>  %130 = call float @llvm.SI.load.const(<16 x i8> %80, i32 72)
>  %131 = fmul float %130, %41
>  %132 = fadd float %129, %131
>  %133 = call float @llvm.SI.load.const(<16 x i8> %80, i32 76)
>  %134 = fmul float %133, %42
>  %135 = fadd float %132, %134
>  %136 = call float @llvm.SI.load.const(<16 x i8> %80, i32 80)
>  %137 = fmul float %136, %39
>  %138 = call float @llvm.SI.load.const(<16 x i8> %80, i32 84)
>  %139 = fmul float %138, %40
>  %140 = fadd float %137, %139
>  %141 = call float @llvm.SI.load.const(<16 x i8> %80, i32 88)
>  %142 = fmul float %141, %41
>  %143 = fadd float %140, %142
>  %144 = call float @llvm.SI.load.const(<16 x i8> %80, i32 92)
>  %145 = fmul float %144, %42
>  %146 = fadd float %143, %145
>  %147 = call float @llvm.SI.load.const(<16 x i8> %80, i32 96)
>  %148 = fmul float %147, %39
>  %149 = call float @llvm.SI.load.const(<16 x i8> %80, i32 100)
>  %150 = fmul float %149, %40
>  %151 = fadd float %148, %150
>  %152 = call float @llvm.SI.load.const(<16 x i8> %80, i32 104)
>  %153 = fmul float %152, %41
>  %154 = fadd float %151, %153
>  %155 = call float @llvm.SI.load.const(<16 x i8> %80, i32 108)
>  %156 = fmul float %155, %42
>  %157 = fadd float %154, %156
>  %158 = call float @llvm.SI.load.const(<16 x i8> %80, i32 112)
>  %159 = fmul float %158, %39
>  %160 = call float @llvm.SI.load.const(<16 x i8> %80, i32 116)
>  %161 = fmul float %160, %40
>  %162 = fadd float %159, %161
>  %163 = call float @llvm.SI.load.const(<16 x i8> %80, i32 120)
>  %164 = fmul float %163, %41
>  %165 = fadd float %162, %164
>  %166 = call float @llvm.SI.load.const(<16 x i8> %80, i32 124)
>  %167 = fmul float %166, %42
>  %168 = fadd float %165, %167
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %47, float %48, float %52, float %56)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %60, float %64, float %68, float %72)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 12, i32 0, float %76, float %78, float %74, float %42)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 13, i32 0, float %91, float %102, float %113, float %124)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 14, i32 0, float %135, float %146, float %157, 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
>
>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[4:7], s[8:9], 0x0                    ; C0820900
>	s_load_dwordx4 s[12:15], s[8:9], 0x4                  ; C0860904
>	v_add_i32_e32 v0, vcc, s10, v0                        ; 4A00000A
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[4:7], 0 idxen   ; E00C2000 80010100
>	buffer_load_format_xyzw v[5:8], v0, s[12:15], 0 idxen ; E00C2000 80030500
>	s_load_dwordx4 s[4:7], s[2:3], 0x0                    ; C0820300
>	s_load_dwordx4 s[0:3], s[2:3], 0x40                   ; C0800340
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	s_buffer_load_dword s8, s[4:7], 0x2                   ; C2040502
>	s_buffer_load_dword s9, s[4:7], 0xc0                  ; C20485C0
>	s_buffer_load_dword s10, s[4:7], 0xc1                 ; C20505C1
>	s_buffer_load_dword s11, s[4:7], 0xc3                 ; C20585C3
>	s_buffer_load_dword s12, s[4:7], 0xc4                 ; C20605C4
>	s_buffer_load_dword s13, s[4:7], 0xc5                 ; C20685C5
>	s_buffer_load_dword s14, s[4:7], 0xc7                 ; C20705C7
>	s_buffer_load_dword s15, s[4:7], 0xc8                 ; C20785C8
>	s_buffer_load_dword s16, s[4:7], 0xc9                 ; C20805C9
>	s_buffer_load_dword s17, s[4:7], 0xcb                 ; C20885CB
>	s_buffer_load_dword s18, s[4:7], 0xcc                 ; C20905CC
>	s_buffer_load_dword s19, s[4:7], 0xcd                 ; C20985CD
>	s_buffer_load_dword s20, s[4:7], 0xcf                 ; C20A05CF
>	s_buffer_load_dword s21, s[4:7], 0xd0                 ; C20A85D0
>	s_buffer_load_dword s22, s[4:7], 0xd1                 ; C20B05D1
>	s_buffer_load_dword s23, s[4:7], 0xd3                 ; C20B85D3
>	s_buffer_load_dword s24, s[4:7], 0xd4                 ; C20C05D4
>	s_buffer_load_dword s25, s[4:7], 0xd5                 ; C20C85D5
>	s_buffer_load_dword s26, s[4:7], 0xd7                 ; C20D05D7
>	s_buffer_load_dword s27, s[4:7], 0xd8                 ; C20D85D8
>	s_buffer_load_dword s4, s[4:7], 0xd9                  ; C20205D9
>	s_buffer_load_dword s5, s[0:3], 0x0                   ; C2028100
>	s_buffer_load_dword s6, s[0:3], 0x1                   ; C2030101
>	s_buffer_load_dword s7, s[0:3], 0x2                   ; C2038102
>	s_buffer_load_dword s28, s[0:3], 0x3                  ; C20E0103
>	s_buffer_load_dword s29, s[0:3], 0x4                  ; C20E8104
>	s_buffer_load_dword s30, s[0:3], 0x5                  ; C20F0105
>	s_buffer_load_dword s31, s[0:3], 0x6                  ; C20F8106
>	s_buffer_load_dword s32, s[0:3], 0x7                  ; C2100107
>	s_buffer_load_dword s33, s[0:3], 0x8                  ; C2108108
>	s_buffer_load_dword s34, s[0:3], 0x9                  ; C2110109
>	s_buffer_load_dword s35, s[0:3], 0xa                  ; C211810A
>	s_buffer_load_dword s36, s[0:3], 0xb                  ; C212010B
>	s_buffer_load_dword s37, s[0:3], 0xc                  ; C212810C
>	s_buffer_load_dword s38, s[0:3], 0xd                  ; C213010D
>	s_buffer_load_dword s39, s[0:3], 0xe                  ; C213810E
>	s_buffer_load_dword s40, s[0:3], 0xf                  ; C214010F
>	s_buffer_load_dword s41, s[0:3], 0x10                 ; C2148110
>	s_buffer_load_dword s42, s[0:3], 0x11                 ; C2150111
>	s_buffer_load_dword s43, s[0:3], 0x12                 ; C2158112
>	s_buffer_load_dword s44, s[0:3], 0x13                 ; C2160113
>	s_buffer_load_dword s45, s[0:3], 0x14                 ; C2168114
>	s_buffer_load_dword s46, s[0:3], 0x15                 ; C2170115
>	s_buffer_load_dword s47, s[0:3], 0x16                 ; C2178116
>	s_buffer_load_dword s48, s[0:3], 0x17                 ; C2180117
>	s_buffer_load_dword s49, s[0:3], 0x18                 ; C2188118
>	s_buffer_load_dword s50, s[0:3], 0x19                 ; C2190119
>	s_buffer_load_dword s51, s[0:3], 0x1a                 ; C219811A
>	s_buffer_load_dword s52, s[0:3], 0x1b                 ; C21A011B
>	s_buffer_load_dword s53, s[0:3], 0x1c                 ; C21A811C
>	s_buffer_load_dword s54, s[0:3], 0x1d                 ; C21B011D
>	s_buffer_load_dword s55, s[0:3], 0x1e                 ; C21B811E
>	s_buffer_load_dword s0, s[0:3], 0x1f                  ; C200011F
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mov_b32_e32 v0, s11                                 ; 7E00020B
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mov_b32_e32 v7, s14                                 ; 7E0E020E
>	v_mov_b32_e32 v8, s17                                 ; 7E100211
>	v_mov_b32_e32 v9, s20                                 ; 7E120214
>	v_mov_b32_e32 v10, s23                                ; 7E140217
>	v_mov_b32_e32 v11, s26                                ; 7E16021A
>	v_mac_f32_e32 v0, s9, v5                              ; 3E000A09
>	v_mac_f32_e32 v7, s12, v5                             ; 3E0E0A0C
>	v_mac_f32_e32 v8, s15, v5                             ; 3E100A0F
>	v_mac_f32_e32 v9, s18, v5                             ; 3E120A12
>	v_mac_f32_e32 v10, s21, v5                            ; 3E140A15
>	v_mac_f32_e32 v11, s24, v5                            ; 3E160A18
>	v_mac_f32_e32 v0, s10, v6                             ; 3E000C0A
>	v_mac_f32_e32 v7, s13, v6                             ; 3E0E0C0D
>	v_mac_f32_e32 v8, s16, v6                             ; 3E100C10
>	v_mac_f32_e32 v9, s19, v6                             ; 3E120C13
>	exp 15, 32, 0, 0, 0, v5, v6, v0, v7                   ; F800020F 07000605
>	v_mac_f32_e32 v10, s22, v6                            ; 3E140C16
>	v_mac_f32_e32 v11, s25, v6                            ; 3E160C19
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mad_f32 v0, v3, s8, -v4                             ; D2820000 84101103
>	v_mad_f32 v5, s27, v4, v1                             ; D2820005 0406081B
>	v_mad_f32 v6, s4, v4, -v2                             ; D2820006 840A0804
>	v_mul_f32_e32 v7, s6, v2                              ; 100E0406
>	exp 15, 33, 0, 0, 0, v8, v9, v10, v11                 ; F800021F 0B0A0908
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mul_f32_e32 v8, s30, v2                             ; 1010041E
>	v_mac_f32_e32 v7, s5, v1                              ; 3E0E0205
>	exp 15, 12, 0, 0, 0, v5, v6, v0, v4                   ; F80000CF 04000605
>	v_mac_f32_e32 v8, s29, v1                             ; 3E10021D
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mul_f32_e32 v0, s34, v2                             ; 10000422
>	v_mac_f32_e32 v0, s33, v1                             ; 3E000221
>	v_mul_f32_e32 v5, s38, v2                             ; 100A0426
>	v_mac_f32_e32 v5, s37, v1                             ; 3E0A0225
>	v_mul_f32_e32 v6, s42, v2                             ; 100C042A
>	v_mul_f32_e32 v9, s46, v2                             ; 1012042E
>	v_mul_f32_e32 v10, s50, v2                            ; 10140432
>	v_mul_f32_e32 v2, s54, v2                             ; 10040436
>	v_mac_f32_e32 v6, s41, v1                             ; 3E0C0229
>	v_mac_f32_e32 v9, s45, v1                             ; 3E12022D
>	v_mac_f32_e32 v10, s49, v1                            ; 3E140231
>	v_mac_f32_e32 v2, s53, v1                             ; 3E040235
>	v_mac_f32_e32 v7, s7, v3                              ; 3E0E0607
>	v_mac_f32_e32 v8, s31, v3                             ; 3E10061F
>	v_mac_f32_e32 v0, s35, v3                             ; 3E000623
>	v_mac_f32_e32 v5, s39, v3                             ; 3E0A0627
>	v_mac_f32_e32 v6, s43, v3                             ; 3E0C062B
>	v_mac_f32_e32 v9, s47, v3                             ; 3E12062F
>	v_mac_f32_e32 v10, s51, v3                            ; 3E140633
>	v_mac_f32_e32 v2, s55, v3                             ; 3E040637
>	v_mac_f32_e32 v7, s28, v4                             ; 3E0E081C
>	v_mac_f32_e32 v8, s32, v4                             ; 3E100820
>	v_mac_f32_e32 v0, s36, v4                             ; 3E000824
>	v_mac_f32_e32 v5, s40, v4                             ; 3E0A0828
>	v_mac_f32_e32 v6, s44, v4                             ; 3E0C082C
>	v_mac_f32_e32 v9, s48, v4                             ; 3E120830
>	v_mac_f32_e32 v10, s52, v4                            ; 3E140834
>	v_mac_f32_e32 v2, s0, v4                              ; 3E040800
>	exp 15, 13, 0, 0, 0, v7, v8, v0, v5                   ; F80000DF 05000807
>	exp 15, 14, 0, 1, 0, v6, v9, v10, v2                  ; F80008EF 020A0906
>	s_endpgm                                              ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 64
>VGPRS: 12
>Code Size: 544 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
>  clamp_color = 0
>FRAG
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL IN[1], GENERIC[1], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SAMP[1]
>DCL SAMP[2]
>DCL SAMP[3]
>DCL SAMP[4]
>DCL SAMP[5]
>DCL SVIEW[0], 2D, FLOAT
>DCL SVIEW[1], 2D, FLOAT
>DCL SVIEW[2], 2D, FLOAT
>DCL SVIEW[3], 2D, FLOAT
>DCL SVIEW[4], 2D, FLOAT
>DCL SVIEW[5], 2D, FLOAT
>DCL CONST[0..3]
>DCL TEMP[0..7], LOCAL
>IMM[0] FLT32 {   -0.5000,    -1.0000,    10.0000,   -10.0000}
>IMM[1] FLT32 {   -2.0000,     3.0000,     1.0000,     6.0000}
>IMM[2] FLT32 {   -0.5800,   100.0000,     0.0500,     0.7500}
>IMM[3] FLT32 {    0.2500,     0.3000,     0.5900,     0.1100}
>IMM[4] FLT32 {   -0.0800,    14.2857,     0.0050,     2.0000}
>  0: ADD TEMP[0].xyz, -CONST[0].xyzz, CONST[1].xyzz
>  1: MOV TEMP[1].xy, IN[0].zwww
>  2: TEX TEMP[1], TEMP[1], SAMP[5], 2D
>  3: MAD TEMP[0].xyz, TEMP[1].xxxx, TEMP[0].xyzz, CONST[0].xyzz
>  4: LRP TEMP[2].xyz, TEMP[1].yyyy, CONST[2].xyzz, TEMP[0].xyzz
>  5: MOV TEMP[0].x, CONST[0].wwww
>  6: MOV TEMP[0].y, CONST[1].wwww
>  7: MOV TEMP[0].z, CONST[2].wwww
>  8: LRP TEMP[3].xyz, TEMP[1].zzzz, TEMP[0].xyzz, TEMP[2].xyzz
>  9: ADD TEMP[1].xy, TEMP[1].wwww, IMM[0].xyyy
> 10: MUL TEMP[1].xy, TEMP[1].xyyy, IMM[0].zwww
> 11: MOV_SAT TEMP[1].xy, TEMP[1].xyyy
> 12: MOV TEMP[4].xy, IN[0].xyyy
> 13: TEX TEMP[4].xyz, TEMP[4], SAMP[3], 2D
> 14: LRP TEMP[5].xyz, TEMP[4].yyyy, CONST[2].xyzz, TEMP[3].xyzz
> 15: LRP TEMP[3].xyz, TEMP[4].zzzz, TEMP[0].xyzz, TEMP[5].xyzz
> 16: MAD TEMP[0].xy, TEMP[1].xyyy, IMM[1].xxxx, IMM[1].yyyy
> 17: MUL TEMP[1].xy, TEMP[1].xyyy, TEMP[1].xyyy
> 18: MUL TEMP[0].xy, TEMP[1].xyyy, TEMP[0].xyyy
> 19: MUL TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 20: MAD TEMP[1].x, CONST[3].wwww, IMM[1].wwww, IMM[1].zzzz
> 21: MOV TEMP[6].xy, IN[0].xyyy
> 22: TEX TEMP[6], TEMP[6], SAMP[0], 2D
> 23: MOV TEMP[7].xy, IN[1].xyyy
> 24: TEX TEMP[7].y, TEMP[7], SAMP[1], 2D
> 25: MAD TEMP[7].x, TEMP[7].yyyy, TEMP[6].xxxx, TEMP[6].wwww
> 26: MAD TEMP[0].x, TEMP[7].xxxx, TEMP[1].xxxx, TEMP[0].xxxx
> 27: ADD TEMP[0].x, TEMP[0].xxxx, IMM[2].xxxx
> 28: MUL TEMP[0].x, TEMP[0].xxxx, IMM[0].zzzz
> 29: MOV_SAT TEMP[1].x, TEMP[0].xxxx
> 30: MAD TEMP[7].x, TEMP[1].xxxx, IMM[1].xxxx, IMM[1].yyyy
> 31: MUL TEMP[0].x, TEMP[1].xxxx, TEMP[1].xxxx
> 32: MUL TEMP[1].x, TEMP[0].xxxx, TEMP[7].xxxx
> 33: MAD TEMP[0].x, TEMP[7].xxxx, TEMP[0].xxxx, IMM[1].zzzz
> 34: ADD TEMP[0].x, -TEMP[4].xxxx, TEMP[0].xxxx
> 35: MOV_SAT TEMP[0].x, TEMP[0].xxxx
> 36: MUL TEMP[1].x, TEMP[1].xxxx, IMM[2].yyyy
> 37: MOV_SAT TEMP[1].x, TEMP[1].xxxx
> 38: MAD TEMP[4].x, TEMP[1].xxxx, IMM[1].xxxx, IMM[1].yyyy
> 39: MUL TEMP[1].x, TEMP[1].xxxx, TEMP[1].xxxx
> 40: MUL TEMP[1].x, TEMP[1].xxxx, TEMP[4].xxxx
> 41: LRP TEMP[2].xyz, TEMP[1].xxxx, IMM[2].zzzz, TEMP[3].xyzz
> 42: MUL TEMP[3].x, IMM[2].wwww, CONST[3].wwww
> 43: ADD TEMP[4].x, -TEMP[6].xxxx, IMM[1].zzzz
> 44: MUL TEMP[4].x, TEMP[4].xxxx, TEMP[4].xxxx
> 45: MUL TEMP[4].x, TEMP[4].xxxx, TEMP[4].xxxx
> 46: MAD TEMP[3].x, TEMP[4].xxxx, IMM[3].xxxx, TEMP[3].xxxx
> 47: MOV TEMP[4].xy, IN[1].zwww
> 48: TEX TEMP[4], TEMP[4], SAMP[4], 2D
> 49: ADD TEMP[4], TEMP[4], IMM[0].yyyy
> 50: MAD TEMP[5], TEMP[3].xxxx, TEMP[4], IMM[1].zzzz
> 51: LRP TEMP[3].xyz, TEMP[1].xxxx, IMM[1].zzzz, TEMP[5].xyzz
> 52: MUL TEMP[3].xyz, TEMP[3].xyzz, TEMP[2].xyzz
> 53: DP3 TEMP[4].x, TEMP[3].xyzz, IMM[3].yzww
> 54: ADD TEMP[4].x, TEMP[4].xxxx, IMM[4].xxxx
> 55: MUL TEMP[4].x, TEMP[4].xxxx, IMM[4].yyyy
> 56: MOV_SAT TEMP[4].x, TEMP[4].xxxx
> 57: MAD TEMP[7].x, TEMP[4].xxxx, IMM[1].xxxx, IMM[1].yyyy
> 58: MUL TEMP[4].x, TEMP[4].xxxx, TEMP[4].xxxx
> 59: MAD TEMP[4].x, TEMP[7].xxxx, -TEMP[4].xxxx, IMM[1].zzzz
> 60: MAD TEMP[2].xyz, TEMP[4].xxxx, IMM[4].zzzz, TEMP[3].xyzz
> 61: MUL TEMP[2].xyz, TEMP[2].xyzz, TEMP[6].zzzz
> 62: MAD TEMP[3].xyz, TEMP[2].xyzz, IMM[4].wwww, TEMP[3].xyzz
> 63: MOV_SAT TEMP[3].xyz, TEMP[3].xyzz
> 64: MUL TEMP[2].xyz, TEMP[6].yyyy, TEMP[3].xyzz
> 65: MOV TEMP[4].xy, IN[0].xyyy
> 66: TEX TEMP[4], TEMP[4], SAMP[2], 2D
> 67: MAD TEMP[3].xyz, TEMP[3].xyzz, -TEMP[6].yyyy, TEMP[4].xyzz
> 68: MUL TEMP[6].x, TEMP[6].yyyy, CONST[3].zzzz
> 69: MUL TEMP[5].x, TEMP[5].wwww, TEMP[6].xxxx
> 70: MAD TEMP[3].xyz, TEMP[0].xxxx, TEMP[3].xyzz, TEMP[2].xyzz
> 71: LRP TEMP[2].x, TEMP[1].xxxx, IMM[1].zzzz, CONST[3].zzzz
> 72: MUL TEMP[1].x, TEMP[5].xxxx, TEMP[2].xxxx
> 73: MAD TEMP[2].x, TEMP[1].xxxx, -CONST[3].xxxx, TEMP[4].wwww
> 74: MUL TEMP[1].x, TEMP[1].xxxx, CONST[3].xxxx
> 75: MAD TEMP[0].x, TEMP[0].xxxx, TEMP[2].xxxx, TEMP[1].xxxx
> 76: MOV TEMP[3].w, TEMP[0].xxxx
> 77: MOV OUT[0], TEMP[3]
> 78: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %24 = load <16 x i8>, <16 x i8> addrspace(2)* %23, align 16, !tbaa !0
>  %25 = call float @llvm.SI.load.const(<16 x i8> %24, i32 0)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %24, i32 4)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %24, i32 8)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %24, i32 12)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %24, i32 16)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %24, i32 20)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %24, i32 24)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %24, i32 28)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %24, i32 32)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %24, i32 36)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %24, i32 40)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %24, i32 44)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %24, i32 48)
>  %38 = call float @llvm.SI.load.const(<16 x i8> %24, i32 56)
>  %39 = call float @llvm.SI.load.const(<16 x i8> %24, i32 60)
>  %40 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %41 = load <8 x i32>, <8 x i32> addrspace(2)* %40, align 32, !tbaa !0
>  %42 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %43 = load <4 x i32>, <4 x i32> addrspace(2)* %42, align 16, !tbaa !0
>  %44 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 1
>  %45 = load <8 x i32>, <8 x i32> addrspace(2)* %44, align 32, !tbaa !0
>  %46 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 1
>  %47 = load <4 x i32>, <4 x i32> addrspace(2)* %46, align 16, !tbaa !0
>  %48 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 2
>  %49 = load <8 x i32>, <8 x i32> addrspace(2)* %48, align 32, !tbaa !0
>  %50 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 2
>  %51 = load <4 x i32>, <4 x i32> addrspace(2)* %50, align 16, !tbaa !0
>  %52 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 3
>  %53 = load <8 x i32>, <8 x i32> addrspace(2)* %52, align 32, !tbaa !0
>  %54 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 3
>  %55 = load <4 x i32>, <4 x i32> addrspace(2)* %54, align 16, !tbaa !0
>  %56 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 4
>  %57 = load <8 x i32>, <8 x i32> addrspace(2)* %56, align 32, !tbaa !0
>  %58 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 4
>  %59 = load <4 x i32>, <4 x i32> addrspace(2)* %58, align 16, !tbaa !0
>  %60 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 5
>  %61 = load <8 x i32>, <8 x i32> addrspace(2)* %60, align 32, !tbaa !0
>  %62 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 5
>  %63 = load <4 x i32>, <4 x i32> addrspace(2)* %62, align 16, !tbaa !0
>  %64 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %8)
>  %65 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %8)
>  %66 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %6, <2 x i32> %8)
>  %67 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %6, <2 x i32> %8)
>  %68 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %6, <2 x i32> %8)
>  %69 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %6, <2 x i32> %8)
>  %70 = call float @llvm.SI.fs.interp(i32 2, i32 1, i32 %6, <2 x i32> %8)
>  %71 = call float @llvm.SI.fs.interp(i32 3, i32 1, i32 %6, <2 x i32> %8)
>  %72 = fsub float %29, %25
>  %73 = fsub float %30, %26
>  %74 = fsub float %31, %27
>  %75 = bitcast float %66 to i32
>  %76 = bitcast float %67 to i32
>  %77 = insertelement <2 x i32> undef, i32 %75, i32 0
>  %78 = insertelement <2 x i32> %77, i32 %76, i32 1
>  %79 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %78, <8 x i32> %61, <4 x i32> %63, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %80 = extractelement <4 x float> %79, i32 0
>  %81 = extractelement <4 x float> %79, i32 1
>  %82 = extractelement <4 x float> %79, i32 2
>  %83 = extractelement <4 x float> %79, i32 3
>  %84 = fmul float %80, %72
>  %85 = fadd float %84, %25
>  %86 = fmul float %80, %73
>  %87 = fadd float %86, %26
>  %88 = fmul float %80, %74
>  %89 = fadd float %88, %27
>  %90 = fsub float 1.000000e+00, %81
>  %91 = fmul float %33, %81
>  %92 = fmul float %85, %90
>  %93 = fadd float %91, %92
>  %94 = fsub float 1.000000e+00, %81
>  %95 = fmul float %34, %81
>  %96 = fmul float %87, %94
>  %97 = fadd float %95, %96
>  %98 = fsub float 1.000000e+00, %81
>  %99 = fmul float %35, %81
>  %100 = fmul float %89, %98
>  %101 = fadd float %99, %100
>  %102 = fsub float 1.000000e+00, %82
>  %103 = fmul float %28, %82
>  %104 = fmul float %93, %102
>  %105 = fadd float %103, %104
>  %106 = fsub float 1.000000e+00, %82
>  %107 = fmul float %32, %82
>  %108 = fmul float %97, %106
>  %109 = fadd float %107, %108
>  %110 = fsub float 1.000000e+00, %82
>  %111 = fmul float %36, %82
>  %112 = fmul float %101, %110
>  %113 = fadd float %111, %112
>  %114 = fadd float %83, -5.000000e-01
>  %115 = fadd float %83, -1.000000e+00
>  %116 = fmul float %114, 1.000000e+01
>  %117 = fmul float %115, -1.000000e+01
>  %118 = call float @llvm.AMDIL.clamp.(float %116, float 0.000000e+00, float 1.000000e+00)
>  %119 = call float @llvm.AMDIL.clamp.(float %117, float 0.000000e+00, float 1.000000e+00)
>  %120 = bitcast float %64 to i32
>  %121 = bitcast float %65 to i32
>  %122 = insertelement <2 x i32> undef, i32 %120, i32 0
>  %123 = insertelement <2 x i32> %122, i32 %121, i32 1
>  %124 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %123, <8 x i32> %53, <4 x i32> %55, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %125 = extractelement <4 x float> %124, i32 0
>  %126 = extractelement <4 x float> %124, i32 1
>  %127 = extractelement <4 x float> %124, i32 2
>  %128 = fsub float 1.000000e+00, %126
>  %129 = fmul float %33, %126
>  %130 = fmul float %105, %128
>  %131 = fadd float %129, %130
>  %132 = fsub float 1.000000e+00, %126
>  %133 = fmul float %34, %126
>  %134 = fmul float %109, %132
>  %135 = fadd float %133, %134
>  %136 = fsub float 1.000000e+00, %126
>  %137 = fmul float %35, %126
>  %138 = fmul float %113, %136
>  %139 = fadd float %137, %138
>  %140 = fsub float 1.000000e+00, %127
>  %141 = fmul float %28, %127
>  %142 = fmul float %131, %140
>  %143 = fadd float %141, %142
>  %144 = fsub float 1.000000e+00, %127
>  %145 = fmul float %32, %127
>  %146 = fmul float %135, %144
>  %147 = fadd float %145, %146
>  %148 = fsub float 1.000000e+00, %127
>  %149 = fmul float %36, %127
>  %150 = fmul float %139, %148
>  %151 = fadd float %149, %150
>  %152 = fmul float %118, -2.000000e+00
>  %153 = fadd float %152, 3.000000e+00
>  %154 = fmul float %119, -2.000000e+00
>  %155 = fadd float %154, 3.000000e+00
>  %156 = fmul float %118, %118
>  %157 = fmul float %119, %119
>  %158 = fmul float %156, %153
>  %159 = fmul float %157, %155
>  %160 = fmul float %159, %158
>  %161 = fmul float %39, 6.000000e+00
>  %162 = fadd float %161, 1.000000e+00
>  %163 = bitcast float %64 to i32
>  %164 = bitcast float %65 to i32
>  %165 = insertelement <2 x i32> undef, i32 %163, i32 0
>  %166 = insertelement <2 x i32> %165, i32 %164, i32 1
>  %167 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %166, <8 x i32> %41, <4 x i32> %43, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %168 = extractelement <4 x float> %167, i32 0
>  %169 = extractelement <4 x float> %167, i32 1
>  %170 = extractelement <4 x float> %167, i32 2
>  %171 = extractelement <4 x float> %167, i32 3
>  %172 = bitcast float %68 to i32
>  %173 = bitcast float %69 to i32
>  %174 = insertelement <2 x i32> undef, i32 %172, i32 0
>  %175 = insertelement <2 x i32> %174, i32 %173, i32 1
>  %176 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %175, <8 x i32> %45, <4 x i32> %47, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %177 = extractelement <4 x float> %176, i32 1
>  %178 = fmul float %177, %168
>  %179 = fadd float %178, %171
>  %180 = fmul float %179, %162
>  %181 = fadd float %180, %160
>  %182 = fadd float %181, 0xBFE28F5C20000000
>  %183 = fmul float %182, 1.000000e+01
>  %184 = call float @llvm.AMDIL.clamp.(float %183, float 0.000000e+00, float 1.000000e+00)
>  %185 = fmul float %184, -2.000000e+00
>  %186 = fadd float %185, 3.000000e+00
>  %187 = fmul float %184, %184
>  %188 = fmul float %187, %186
>  %189 = fmul float %186, %187
>  %190 = fadd float %189, 1.000000e+00
>  %191 = fsub float %190, %125
>  %192 = call float @llvm.AMDIL.clamp.(float %191, float 0.000000e+00, float 1.000000e+00)
>  %193 = fmul float %188, 1.000000e+02
>  %194 = call float @llvm.AMDIL.clamp.(float %193, float 0.000000e+00, float 1.000000e+00)
>  %195 = fmul float %194, -2.000000e+00
>  %196 = fadd float %195, 3.000000e+00
>  %197 = fmul float %194, %194
>  %198 = fmul float %197, %196
>  %199 = fsub float 1.000000e+00, %198
>  %200 = fmul float %198, 0x3FA99999A0000000
>  %201 = fmul float %143, %199
>  %202 = fadd float %200, %201
>  %203 = fsub float 1.000000e+00, %198
>  %204 = fmul float %198, 0x3FA99999A0000000
>  %205 = fmul float %147, %203
>  %206 = fadd float %204, %205
>  %207 = fsub float 1.000000e+00, %198
>  %208 = fmul float %198, 0x3FA99999A0000000
>  %209 = fmul float %151, %207
>  %210 = fadd float %208, %209
>  %211 = fmul float %39, 7.500000e-01
>  %212 = fsub float 1.000000e+00, %168
>  %213 = fmul float %212, %212
>  %214 = fmul float %213, %213
>  %215 = fmul float %214, 2.500000e-01
>  %216 = fadd float %215, %211
>  %217 = bitcast float %70 to i32
>  %218 = bitcast float %71 to i32
>  %219 = insertelement <2 x i32> undef, i32 %217, i32 0
>  %220 = insertelement <2 x i32> %219, i32 %218, i32 1
>  %221 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %220, <8 x i32> %57, <4 x i32> %59, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %222 = extractelement <4 x float> %221, i32 0
>  %223 = extractelement <4 x float> %221, i32 1
>  %224 = extractelement <4 x float> %221, i32 2
>  %225 = extractelement <4 x float> %221, i32 3
>  %226 = fadd float %222, -1.000000e+00
>  %227 = fadd float %223, -1.000000e+00
>  %228 = fadd float %224, -1.000000e+00
>  %229 = fadd float %225, -1.000000e+00
>  %230 = fmul float %216, %226
>  %231 = fadd float %230, 1.000000e+00
>  %232 = fmul float %216, %227
>  %233 = fadd float %232, 1.000000e+00
>  %234 = fmul float %216, %228
>  %235 = fadd float %234, 1.000000e+00
>  %236 = fmul float %216, %229
>  %237 = fadd float %236, 1.000000e+00
>  %238 = fsub float 1.000000e+00, %198
>  %239 = fmul float %231, %238
>  %240 = fadd float %198, %239
>  %241 = fsub float 1.000000e+00, %198
>  %242 = fmul float %233, %241
>  %243 = fadd float %198, %242
>  %244 = fsub float 1.000000e+00, %198
>  %245 = fmul float %235, %244
>  %246 = fadd float %198, %245
>  %247 = fmul float %240, %202
>  %248 = fmul float %243, %206
>  %249 = fmul float %246, %210
>  %250 = fmul float %247, 0x3FD3333340000000
>  %251 = fmul float %248, 0x3FE2E147A0000000
>  %252 = fadd float %251, %250
>  %253 = fmul float %249, 0x3FBC28F5C0000000
>  %254 = fadd float %252, %253
>  %255 = fadd float %254, 0xBFB47AE140000000
>  %256 = fmul float %255, 0x402C924920000000
>  %257 = call float @llvm.AMDIL.clamp.(float %256, float 0.000000e+00, float 1.000000e+00)
>  %258 = fmul float %257, -2.000000e+00
>  %259 = fadd float %258, 3.000000e+00
>  %260 = fmul float %257, %257
>  %261 = fmul float %260, %259
>  %262 = fsub float 1.000000e+00, %261
>  %263 = fmul float %262, 0x3F747AE140000000
>  %264 = fadd float %263, %247
>  %265 = fmul float %262, 0x3F747AE140000000
>  %266 = fadd float %265, %248
>  %267 = fmul float %262, 0x3F747AE140000000
>  %268 = fadd float %267, %249
>  %269 = fmul float %264, %170
>  %270 = fmul float %266, %170
>  %271 = fmul float %268, %170
>  %272 = fmul float %269, 2.000000e+00
>  %273 = fadd float %272, %247
>  %274 = fmul float %270, 2.000000e+00
>  %275 = fadd float %274, %248
>  %276 = fmul float %271, 2.000000e+00
>  %277 = fadd float %276, %249
>  %278 = call float @llvm.AMDIL.clamp.(float %273, float 0.000000e+00, float 1.000000e+00)
>  %279 = call float @llvm.AMDIL.clamp.(float %275, float 0.000000e+00, float 1.000000e+00)
>  %280 = call float @llvm.AMDIL.clamp.(float %277, float 0.000000e+00, float 1.000000e+00)
>  %281 = fmul float %169, %278
>  %282 = fmul float %169, %279
>  %283 = fmul float %169, %280
>  %284 = bitcast float %64 to i32
>  %285 = bitcast float %65 to i32
>  %286 = insertelement <2 x i32> undef, i32 %284, i32 0
>  %287 = insertelement <2 x i32> %286, i32 %285, i32 1
>  %288 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %287, <8 x i32> %49, <4 x i32> %51, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %289 = extractelement <4 x float> %288, i32 0
>  %290 = extractelement <4 x float> %288, i32 1
>  %291 = extractelement <4 x float> %288, i32 2
>  %292 = extractelement <4 x float> %288, i32 3
>  %293 = fmul float %169, %278
>  %294 = fsub float %289, %293
>  %295 = fmul float %169, %279
>  %296 = fsub float %290, %295
>  %297 = fmul float %169, %280
>  %298 = fsub float %291, %297
>  %299 = fmul float %169, %38
>  %300 = fmul float %237, %299
>  %301 = fmul float %192, %294
>  %302 = fadd float %301, %281
>  %303 = fmul float %192, %296
>  %304 = fadd float %303, %282
>  %305 = fmul float %192, %298
>  %306 = fadd float %305, %283
>  %307 = fsub float 1.000000e+00, %198
>  %308 = fmul float %38, %307
>  %309 = fadd float %198, %308
>  %310 = fmul float %300, %309
>  %311 = fmul float %37, %310
>  %312 = fsub float %292, %311
>  %313 = fmul float %310, %37
>  %314 = fmul float %192, %312
>  %315 = fadd float %314, %313
>  %316 = call i32 @llvm.SI.packf16(float %302, float %304)
>  %317 = bitcast i32 %316 to float
>  %318 = call i32 @llvm.SI.packf16(float %306, float %315)
>  %319 = bitcast i32 %318 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %317, float %319, float %317, float %319)
>  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.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #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, s10                                                            ; BEFC030A
>	v_mov_b32_e32 v2, 0x41200000                                                 ; 7E0402FF 41200000
>	v_mov_b32_e32 v3, 0x40400000                                                 ; 7E0602FF 40400000
>	v_mov_b32_e32 v4, 0x40c00000                                                 ; 7E0802FF 40C00000
>	v_mov_b32_e32 v5, 0x3f400000                                                 ; 7E0A02FF 3F400000
>	v_mov_b32_e32 v6, 0xbda3d70a                                                 ; 7E0C02FF BDA3D70A
>	v_mov_b32_e32 v7, 0x3ba3d70a                                                 ; 7E0E02FF 3BA3D70A
>	v_interp_p1_f32 v8, v0, 0, 0, [m0]                                           ; C8200000
>	v_interp_p2_f32 v8, [v8], v1, 0, 0, [m0]                                     ; C8210001
>	v_interp_p1_f32 v9, v0, 1, 0, [m0]                                           ; C8240100
>	v_interp_p2_f32 v9, [v9], v1, 1, 0, [m0]                                     ; C8250101
>	v_interp_p1_f32 v10, v0, 2, 0, [m0]                                          ; C8280200
>	v_interp_p2_f32 v10, [v10], v1, 2, 0, [m0]                                   ; C8290201
>	v_interp_p1_f32 v11, v0, 3, 0, [m0]                                          ; C82C0300
>	v_interp_p2_f32 v11, [v11], v1, 3, 0, [m0]                                   ; C82D0301
>	v_interp_p1_f32 v12, v0, 0, 1, [m0]                                          ; C8300400
>	v_interp_p2_f32 v12, [v12], v1, 0, 1, [m0]                                   ; C8310401
>	v_interp_p1_f32 v13, v0, 1, 1, [m0]                                          ; C8340500
>	v_interp_p2_f32 v13, [v13], v1, 1, 1, [m0]                                   ; C8350501
>	v_interp_p1_f32 v14, v0, 2, 1, [m0]                                          ; C8380600
>	v_interp_p2_f32 v14, [v14], v1, 2, 1, [m0]                                   ; C8390601
>	v_interp_p1_f32 v15, v0, 3, 1, [m0]                                          ; C83C0700
>	v_interp_p2_f32 v15, [v15], v1, 3, 1, [m0]                                   ; C83D0701
>	s_load_dwordx4 s[24:27], s[2:3], 0x0                                         ; C08C0300
>	s_load_dwordx4 s[32:35], s[4:5], 0x0                                         ; C0900500
>	s_load_dwordx4 s[36:39], s[4:5], 0x4                                         ; C0920504
>	s_load_dwordx4 s[8:11], s[4:5], 0x8                                          ; C0840508
>	s_load_dwordx4 s[40:43], s[4:5], 0xc                                         ; C094050C
>	s_load_dwordx4 s[20:23], s[4:5], 0x10                                        ; C08A0510
>	s_load_dwordx4 s[60:63], s[4:5], 0x14                                        ; C09E0514
>	s_load_dwordx8 s[44:51], s[6:7], 0x0                                         ; C0D60700
>	s_load_dwordx8 s[52:59], s[6:7], 0x8                                         ; C0DA0708
>	s_load_dwordx8 s[12:19], s[6:7], 0x10                                        ; C0C60710
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	s_buffer_load_dword s2, s[24:27], 0x0                                        ; C2011900
>	s_buffer_load_dword s3, s[24:27], 0x1                                        ; C2019901
>	s_buffer_load_dword s4, s[24:27], 0x2                                        ; C2021902
>	s_buffer_load_dword s5, s[24:27], 0x3                                        ; C2029903
>	s_buffer_load_dword s28, s[24:27], 0x4                                       ; C20E1904
>	s_buffer_load_dword s29, s[24:27], 0x5                                       ; C20E9905
>	s_buffer_load_dword s30, s[24:27], 0x6                                       ; C20F1906
>	s_buffer_load_dword s64, s[24:27], 0x7                                       ; C2201907
>	s_buffer_load_dword s65, s[24:27], 0x8                                       ; C2209908
>	s_buffer_load_dword s66, s[24:27], 0x9                                       ; C2211909
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mov_b32_e32 v0, s2                                                         ; 7E000202
>	s_buffer_load_dword s67, s[24:27], 0xa                                       ; C221990A
>	v_mov_b32_e32 v1, s3                                                         ; 7E020203
>	s_buffer_load_dword s68, s[24:27], 0xb                                       ; C222190B
>	v_mov_b32_e32 v16, s4                                                        ; 7E200204
>	s_buffer_load_dword s0, s[24:27], 0xc                                        ; C200190C
>	s_buffer_load_dword s1, s[24:27], 0xe                                        ; C200990E
>	s_buffer_load_dword s69, s[24:27], 0xf                                       ; C222990F
>	v_sub_f32_e32 v0, s28, v0                                                    ; 0800001C
>	v_sub_f32_e32 v1, s29, v1                                                    ; 0802021D
>	v_sub_f32_e32 v16, s30, v16                                                  ; 0820201E
>	s_load_dwordx8 s[72:79], s[6:7], 0x18                                        ; C0E40718
>	s_load_dwordx8 s[24:31], s[6:7], 0x20                                        ; C0CC0720
>	s_load_dwordx8 s[80:87], s[6:7], 0x28                                        ; C0E80728
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v4, v4, s69, 1.0                                                   ; D2820004 03C88B04
>	image_sample v[17:20], 15, 0, 0, 0, 0, 0, 0, 0, v[10:11], s[80:87], s[60:63] ; F0800F00 01F4110A
>	image_sample v[21:23], 7, 0, 0, 0, 0, 0, 0, 0, v[8:9], s[72:79], s[40:43]    ; F0800700 01521508
>	image_sample v[24:27], 15, 0, 0, 0, 0, 0, 0, 0, v[8:9], s[44:51], s[32:35]   ; F0800F00 010B1808
>	image_sample v10, 2, 0, 0, 0, 0, 0, 0, 0, v[12:13], s[52:59], s[36:39]       ; F0800200 012D0A0C
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mac_f32_e32 v27, v24, v10                                                  ; 3E361518
>	v_mad_f32 v0, v17, v0, s2                                                    ; D2820000 000A0111
>	v_mad_f32 v1, v17, v1, s3                                                    ; D2820001 000E0311
>	v_mad_f32 v10, v17, v16, s4                                                  ; D282000A 00122111
>	v_mad_f32 v0, -v18, v0, v0                                                   ; D2820000 24020112
>	v_mac_f32_e32 v0, s65, v18                                                   ; 3E002441
>	v_mad_f32 v1, -v18, v1, v1                                                   ; D2820001 24060312
>	v_mac_f32_e32 v1, s66, v18                                                   ; 3E022442
>	v_mad_f32 v10, -v18, v10, v10                                                ; D282000A 242A1512
>	v_mac_f32_e32 v10, s67, v18                                                  ; 3E142443
>	v_mad_f32 v0, -v19, v0, v0                                                   ; D2820000 24020113
>	v_mac_f32_e32 v0, s5, v19                                                    ; 3E002605
>	v_mad_f32 v1, -v19, v1, v1                                                   ; D2820001 24060313
>	v_mac_f32_e32 v1, s64, v19                                                   ; 3E022640
>	v_mad_f32 v10, -v19, v10, v10                                                ; D282000A 242A1513
>	v_mac_f32_e32 v10, s68, v19                                                  ; 3E142644
>	v_add_f32_e32 v11, -0.5, v20                                                 ; 061628F1
>	v_madmk_f32_e32 v12, v20, v2, 0xc1200000                                     ; 40180514 C1200000
>	v_mul_f32_e32 v11, v2, v11                                                   ; 10161702
>	v_add_f32_e64 v11, 0, v11 clamp                                              ; D206080B 00021680
>	v_add_f32_e64 v12, 0, v12 clamp                                              ; D206080C 00021880
>	v_mad_f32 v13, -2.0, v11, v3                                                 ; D282000D 040E16F5
>	v_mul_f32_e32 v11, v11, v11                                                  ; 1016170B
>	v_mul_f32_e32 v11, v13, v11                                                  ; 1016170D
>	v_mad_f32 v13, -2.0, v12, v3                                                 ; D282000D 040E18F5
>	v_mul_f32_e32 v12, v12, v12                                                  ; 1018190C
>	v_mul_f32_e32 v12, v13, v12                                                  ; 1018190D
>	v_mad_f32 v0, -v22, v0, v0                                                   ; D2820000 24020116
>	v_mac_f32_e32 v0, s65, v22                                                   ; 3E002C41
>	v_mad_f32 v1, -v22, v1, v1                                                   ; D2820001 24060316
>	v_mac_f32_e32 v1, s66, v22                                                   ; 3E022C42
>	v_mad_f32 v10, -v22, v10, v10                                                ; D282000A 242A1516
>	v_mac_f32_e32 v10, s67, v22                                                  ; 3E142C43
>	v_mad_f32 v0, -v23, v0, v0                                                   ; D2820000 24020117
>	v_mac_f32_e32 v0, s5, v23                                                    ; 3E002E05
>	v_mad_f32 v1, -v23, v1, v1                                                   ; D2820001 24060317
>	v_mac_f32_e32 v1, s64, v23                                                   ; 3E022E40
>	v_mad_f32 v10, -v23, v10, v10                                                ; D282000A 242A1517
>	v_mac_f32_e32 v10, s68, v23                                                  ; 3E142E44
>	v_madak_f32_e32 v11, v12, v11, 0xbf147ae1                                    ; 4216170C BF147AE1
>	v_mac_f32_e32 v11, v4, v27                                                   ; 3E163704
>	v_mul_f32_e32 v2, v2, v11                                                    ; 10041702
>	v_add_f32_e64 v2, 0, v2 clamp                                                ; D2060802 00020480
>	v_mad_f32 v4, -2.0, v2, v3                                                   ; D2820004 040E04F5
>	v_mul_f32_e32 v2, v2, v2                                                     ; 10040502
>	v_mul_f32_e32 v11, v4, v2                                                    ; 10160504
>	v_mad_f32 v2, v2, v4, 1.0                                                    ; D2820002 03CA0902
>	v_subrev_f32_e32 v2, v21, v2                                                 ; 0A040515
>	v_add_f32_e64 v2, 0, v2 clamp                                                ; D2060802 00020480
>	v_mul_f32_e32 v4, 0x42c80000, v11                                            ; 100816FF 42C80000
>	v_add_f32_e64 v4, 0, v4 clamp                                                ; D2060804 00020880
>	v_mul_f32_e32 v5, s69, v5                                                    ; 100A0A45
>	v_sub_f32_e32 v11, 1.0, v24                                                  ; 081630F2
>	v_mad_f32 v11, -v24, v11, v11                                                ; D282000B 242E1718
>	v_mul_f32_e32 v11, v11, v11                                                  ; 1016170B
>	v_madmk_f32_e32 v5, v11, v5, 0x3e800000                                      ; 400A0B0B 3E800000
>	image_sample v[11:14], 15, 0, 0, 0, 0, 0, 0, 0, v[14:15], s[24:31], s[20:23] ; F0800F00 00A60B0E
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_add_f32_e32 v11, -1.0, v11                                                 ; 061616F3
>	v_add_f32_e32 v12, -1.0, v12                                                 ; 061818F3
>	v_add_f32_e32 v13, -1.0, v13                                                 ; 061A1AF3
>	v_mad_f32 v14, v14, v5, -v5                                                  ; D282000E 84160B0E
>	v_mad_f32 v15, -2.0, v4, v3                                                  ; D282000F 040E08F5
>	v_mul_f32_e32 v4, v4, v4                                                     ; 10080904
>	v_mul_f32_e32 v16, v15, v4                                                   ; 1020090F
>	v_mad_f32 v4, -v4, v15, 1.0                                                  ; D2820004 23CA1F04
>	v_mul_f32_e32 v15, 0x3d4ccccd, v16                                           ; 101E20FF 3D4CCCCD
>	v_mad_f32 v0, v4, v0, v15                                                    ; D2820000 043E0104
>	v_mad_f32 v1, v4, v1, v15                                                    ; D2820001 043E0304
>	v_mad_f32 v10, v4, v10, v15                                                  ; D282000A 043E1504
>	v_mad_f32 v11, v5, v11, 1.0                                                  ; D282000B 03CA1705
>	v_mad_f32 v12, v5, v12, 1.0                                                  ; D282000C 03CA1905
>	v_mad_f32 v5, v5, v13, 1.0                                                   ; D2820005 03CA1B05
>	v_mad_f32 v11, v4, v11, v16                                                  ; D282000B 04421704
>	v_mad_f32 v12, v4, v12, v16                                                  ; D282000C 04421904
>	v_mad_f32 v5, v4, v5, v16                                                    ; D2820005 04420B04
>	v_mul_f32_e32 v13, v0, v11                                                   ; 101A1700
>	v_mul_f32_e32 v15, v1, v12                                                   ; 101E1901
>	v_mul_f32_e32 v17, v10, v5                                                   ; 10220B0A
>	v_mul_f32_e32 v13, 0x3e99999a, v13                                           ; 101A1AFF 3E99999A
>	v_madmk_f32_e32 v13, v15, v13, 0x3f170a3d                                    ; 401A1B0F 3F170A3D
>	v_madmk_f32_e32 v13, v17, v13, 0x3de147ae                                    ; 401A1B11 3DE147AE
>	v_add_f32_e32 v6, v13, v6                                                    ; 060C0D0D
>	v_mul_f32_e32 v6, 0x41649249, v6                                             ; 100C0CFF 41649249
>	v_add_f32_e64 v6, 0, v6 clamp                                                ; D2060806 00020C80
>	v_mac_f32_e32 v3, -2.0, v6                                                   ; 3E060CF5
>	v_mul_f32_e32 v6, v6, v6                                                     ; 100C0D06
>	v_mul_f32_e32 v3, v3, v6                                                     ; 10060D03
>	v_mad_f32 v3, -v3, v7, v7                                                    ; D2820003 241E0F03
>	v_mad_f32 v6, v0, v11, v3                                                    ; D2820006 040E1700
>	v_mad_f32 v7, v1, v12, v3                                                    ; D2820007 040E1901
>	v_mad_f32 v3, v10, v5, v3                                                    ; D2820003 040E0B0A
>	v_mul_f32_e32 v13, v26, v6                                                   ; 101A0D1A
>	v_mul_f32_e32 v15, v26, v7                                                   ; 101E0F1A
>	v_mul_f32_e32 v17, v26, v3                                                   ; 1022071A
>	v_mac_f32_e32 v13, v26, v6                                                   ; 3E1A0D1A
>	v_mac_f32_e32 v13, v0, v11                                                   ; 3E1A1700
>	v_mac_f32_e32 v15, v26, v7                                                   ; 3E1E0F1A
>	v_mac_f32_e32 v15, v1, v12                                                   ; 3E1E1901
>	v_mac_f32_e32 v17, v26, v3                                                   ; 3E22071A
>	v_mac_f32_e32 v17, v10, v5                                                   ; 3E220B0A
>	v_add_f32_e64 v0, 0, v13 clamp                                               ; D2060800 00021A80
>	v_add_f32_e64 v1, 0, v15 clamp                                               ; D2060801 00021E80
>	v_add_f32_e64 v3, 0, v17 clamp                                               ; D2060803 00022280
>	v_mul_f32_e32 v5, v0, v25                                                    ; 100A3300
>	v_mul_f32_e32 v6, v1, v25                                                    ; 100C3301
>	v_mul_f32_e32 v7, v3, v25                                                    ; 100E3303
>	image_sample v[8:11], 15, 0, 0, 0, 0, 0, 0, 0, v[8:9], s[12:19], s[8:11]     ; F0800F00 00430808
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v0, -v25, v0, v8                                                   ; D2820000 24220119
>	v_mad_f32 v1, -v25, v1, v9                                                   ; D2820001 24260319
>	v_mad_f32 v3, -v25, v3, v10                                                  ; D2820003 242A0719
>	v_mul_f32_e32 v8, s1, v25                                                    ; 10103201
>	v_mac_f32_e32 v8, v8, v14                                                    ; 3E101D08
>	v_mac_f32_e32 v5, v0, v2                                                     ; 3E0A0500
>	v_mac_f32_e32 v6, v1, v2                                                     ; 3E0C0501
>	v_mac_f32_e32 v7, v3, v2                                                     ; 3E0E0503
>	v_mac_f32_e32 v16, s1, v4                                                    ; 3E200801
>	v_mul_f32_e32 v0, v16, v8                                                    ; 10001110
>	v_mul_f32_e32 v1, s0, v0                                                     ; 10020000
>	v_mad_f32 v0, -s0, v0, v11                                                   ; D2820000 242E0000
>	v_mac_f32_e32 v1, v0, v2                                                     ; 3E020500
>	v_cvt_pkrtz_f16_f32_e32 v0, v5, v6                                           ; 5E000D05
>	v_cvt_pkrtz_f16_f32_e32 v1, v7, v1                                           ; 5E020307
>	exp 15, 0, 1, 1, 1, v0, v1, v0, v1                                           ; F8001C0F 01000100
>	s_endpgm                                                                     ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 88
>VGPRS: 28
>Code Size: 1060 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
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL OUT[1], CLIPVERTEX
>DCL OUT[2], GENERIC[0]
>DCL OUT[3], GENERIC[1]
>DCL CONST[0..54]
>DCL TEMP[0..3], LOCAL
>  0: MUL TEMP[0].xy, CONST[48].xyyy, IN[1].xyyy
>  1: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
>  2: ADD TEMP[1].x, TEMP[0].xxxx, CONST[48].wwww
>  3: MOV TEMP[1].z, TEMP[1].xxxx
>  4: MUL TEMP[0].xy, CONST[49].xyyy, IN[1].xyyy
>  5: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
>  6: ADD TEMP[2].x, TEMP[0].xxxx, CONST[49].wwww
>  7: MOV TEMP[1].w, TEMP[2].xxxx
>  8: MUL TEMP[0].xy, CONST[50].xyyy, IN[1].xyyy
>  9: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 10: ADD TEMP[2].x, TEMP[0].xxxx, CONST[50].wwww
> 11: MUL TEMP[0].xy, CONST[51].xyyy, IN[1].xyyy
> 12: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 13: ADD TEMP[3].x, TEMP[0].xxxx, CONST[51].wwww
> 14: MOV TEMP[2].y, TEMP[3].xxxx
> 15: MUL TEMP[0].xy, CONST[52].xyyy, IN[1].xyyy
> 16: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 17: ADD TEMP[3].x, TEMP[0].xxxx, CONST[52].wwww
> 18: MOV TEMP[2].z, TEMP[3].xxxx
> 19: MUL TEMP[0].xy, CONST[53].xyyy, IN[1].xyyy
> 20: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 21: ADD TEMP[0].x, TEMP[0].xxxx, CONST[53].wwww
> 22: MOV TEMP[2].w, TEMP[0].xxxx
> 23: MOV TEMP[0].xw, IN[0].xxxw
> 24: MOV TEMP[1].xy, IN[1].xyxx
> 25: MAD TEMP[3].x, IN[0].zzzz, CONST[0].zzzz, -IN[0].wwww
> 26: MOV TEMP[0].z, TEMP[3].xxxx
> 27: MOV TEMP[0].y, -IN[0].yyyy
> 28: MAD TEMP[0].xy, CONST[54].xyyy, IN[0].wwww, TEMP[0].xyyy
> 29: MOV OUT[2], TEMP[1]
> 30: MOV OUT[3], TEMP[2]
> 31: MOV OUT[0], TEMP[0]
> 32: MOV OUT[1], IN[0]
> 33: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = call float @llvm.SI.load.const(<16 x i8> %13, i32 8)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %13, i32 768)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %13, i32 772)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %13, i32 780)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %13, i32 784)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %13, i32 788)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %13, i32 796)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %13, i32 800)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %13, i32 804)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %13, i32 812)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %13, i32 816)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %13, i32 820)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %13, i32 828)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %13, i32 832)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %13, i32 836)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %13, i32 844)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %13, i32 848)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %13, i32 852)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %13, i32 860)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %13, i32 864)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %13, i32 868)
>  %35 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %36 = load <16 x i8>, <16 x i8> addrspace(2)* %35, align 16, !tbaa !0
>  %37 = add i32 %5, %8
>  %38 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %36, i32 0, i32 %37)
>  %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 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 1
>  %44 = load <16 x i8>, <16 x i8> addrspace(2)* %43, align 16, !tbaa !0
>  %45 = add i32 %5, %8
>  %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 = fmul float %15, %47
>  %50 = fmul float %16, %48
>  %51 = fadd float %50, %49
>  %52 = fadd float %51, %17
>  %53 = fmul float %18, %47
>  %54 = fmul float %19, %48
>  %55 = fadd float %54, %53
>  %56 = fadd float %55, %20
>  %57 = fmul float %21, %47
>  %58 = fmul float %22, %48
>  %59 = fadd float %58, %57
>  %60 = fadd float %59, %23
>  %61 = fmul float %24, %47
>  %62 = fmul float %25, %48
>  %63 = fadd float %62, %61
>  %64 = fadd float %63, %26
>  %65 = fmul float %27, %47
>  %66 = fmul float %28, %48
>  %67 = fadd float %66, %65
>  %68 = fadd float %67, %29
>  %69 = fmul float %30, %47
>  %70 = fmul float %31, %48
>  %71 = fadd float %70, %69
>  %72 = fadd float %71, %32
>  %73 = fmul float %41, %14
>  %74 = fsub float %73, %42
>  %75 = fmul float %33, %42
>  %76 = fadd float %75, %39
>  %77 = fmul float %34, %42
>  %78 = fsub float %77, %40
>  %79 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 16
>  %80 = load <16 x i8>, <16 x i8> addrspace(2)* %79, align 16, !tbaa !0
>  %81 = call float @llvm.SI.load.const(<16 x i8> %80, i32 0)
>  %82 = fmul float %81, %39
>  %83 = call float @llvm.SI.load.const(<16 x i8> %80, i32 4)
>  %84 = fmul float %83, %40
>  %85 = fadd float %82, %84
>  %86 = call float @llvm.SI.load.const(<16 x i8> %80, i32 8)
>  %87 = fmul float %86, %41
>  %88 = fadd float %85, %87
>  %89 = call float @llvm.SI.load.const(<16 x i8> %80, i32 12)
>  %90 = fmul float %89, %42
>  %91 = fadd float %88, %90
>  %92 = call float @llvm.SI.load.const(<16 x i8> %80, i32 16)
>  %93 = fmul float %92, %39
>  %94 = call float @llvm.SI.load.const(<16 x i8> %80, i32 20)
>  %95 = fmul float %94, %40
>  %96 = fadd float %93, %95
>  %97 = call float @llvm.SI.load.const(<16 x i8> %80, i32 24)
>  %98 = fmul float %97, %41
>  %99 = fadd float %96, %98
>  %100 = call float @llvm.SI.load.const(<16 x i8> %80, i32 28)
>  %101 = fmul float %100, %42
>  %102 = fadd float %99, %101
>  %103 = call float @llvm.SI.load.const(<16 x i8> %80, i32 32)
>  %104 = fmul float %103, %39
>  %105 = call float @llvm.SI.load.const(<16 x i8> %80, i32 36)
>  %106 = fmul float %105, %40
>  %107 = fadd float %104, %106
>  %108 = call float @llvm.SI.load.const(<16 x i8> %80, i32 40)
>  %109 = fmul float %108, %41
>  %110 = fadd float %107, %109
>  %111 = call float @llvm.SI.load.const(<16 x i8> %80, i32 44)
>  %112 = fmul float %111, %42
>  %113 = fadd float %110, %112
>  %114 = call float @llvm.SI.load.const(<16 x i8> %80, i32 48)
>  %115 = fmul float %114, %39
>  %116 = call float @llvm.SI.load.const(<16 x i8> %80, i32 52)
>  %117 = fmul float %116, %40
>  %118 = fadd float %115, %117
>  %119 = call float @llvm.SI.load.const(<16 x i8> %80, i32 56)
>  %120 = fmul float %119, %41
>  %121 = fadd float %118, %120
>  %122 = call float @llvm.SI.load.const(<16 x i8> %80, i32 60)
>  %123 = fmul float %122, %42
>  %124 = fadd float %121, %123
>  %125 = call float @llvm.SI.load.const(<16 x i8> %80, i32 64)
>  %126 = fmul float %125, %39
>  %127 = call float @llvm.SI.load.const(<16 x i8> %80, i32 68)
>  %128 = fmul float %127, %40
>  %129 = fadd float %126, %128
>  %130 = call float @llvm.SI.load.const(<16 x i8> %80, i32 72)
>  %131 = fmul float %130, %41
>  %132 = fadd float %129, %131
>  %133 = call float @llvm.SI.load.const(<16 x i8> %80, i32 76)
>  %134 = fmul float %133, %42
>  %135 = fadd float %132, %134
>  %136 = call float @llvm.SI.load.const(<16 x i8> %80, i32 80)
>  %137 = fmul float %136, %39
>  %138 = call float @llvm.SI.load.const(<16 x i8> %80, i32 84)
>  %139 = fmul float %138, %40
>  %140 = fadd float %137, %139
>  %141 = call float @llvm.SI.load.const(<16 x i8> %80, i32 88)
>  %142 = fmul float %141, %41
>  %143 = fadd float %140, %142
>  %144 = call float @llvm.SI.load.const(<16 x i8> %80, i32 92)
>  %145 = fmul float %144, %42
>  %146 = fadd float %143, %145
>  %147 = call float @llvm.SI.load.const(<16 x i8> %80, i32 96)
>  %148 = fmul float %147, %39
>  %149 = call float @llvm.SI.load.const(<16 x i8> %80, i32 100)
>  %150 = fmul float %149, %40
>  %151 = fadd float %148, %150
>  %152 = call float @llvm.SI.load.const(<16 x i8> %80, i32 104)
>  %153 = fmul float %152, %41
>  %154 = fadd float %151, %153
>  %155 = call float @llvm.SI.load.const(<16 x i8> %80, i32 108)
>  %156 = fmul float %155, %42
>  %157 = fadd float %154, %156
>  %158 = call float @llvm.SI.load.const(<16 x i8> %80, i32 112)
>  %159 = fmul float %158, %39
>  %160 = call float @llvm.SI.load.const(<16 x i8> %80, i32 116)
>  %161 = fmul float %160, %40
>  %162 = fadd float %159, %161
>  %163 = call float @llvm.SI.load.const(<16 x i8> %80, i32 120)
>  %164 = fmul float %163, %41
>  %165 = fadd float %162, %164
>  %166 = call float @llvm.SI.load.const(<16 x i8> %80, i32 124)
>  %167 = fmul float %166, %42
>  %168 = fadd float %165, %167
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %47, float %48, float %52, float %56)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %60, float %64, float %68, float %72)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 12, i32 0, float %76, float %78, float %74, float %42)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 13, i32 0, float %91, float %102, float %113, float %124)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 14, i32 0, float %135, float %146, float %157, 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
>
>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[4:7], s[8:9], 0x0                    ; C0820900
>	s_load_dwordx4 s[12:15], s[8:9], 0x4                  ; C0860904
>	v_add_i32_e32 v0, vcc, s10, v0                        ; 4A00000A
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[4:7], 0 idxen   ; E00C2000 80010100
>	buffer_load_format_xyzw v[5:8], v0, s[12:15], 0 idxen ; E00C2000 80030500
>	s_load_dwordx4 s[4:7], s[2:3], 0x0                    ; C0820300
>	s_load_dwordx4 s[0:3], s[2:3], 0x40                   ; C0800340
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	s_buffer_load_dword s8, s[4:7], 0x2                   ; C2040502
>	s_buffer_load_dword s9, s[4:7], 0xc0                  ; C20485C0
>	s_buffer_load_dword s10, s[4:7], 0xc1                 ; C20505C1
>	s_buffer_load_dword s11, s[4:7], 0xc3                 ; C20585C3
>	s_buffer_load_dword s12, s[4:7], 0xc4                 ; C20605C4
>	s_buffer_load_dword s13, s[4:7], 0xc5                 ; C20685C5
>	s_buffer_load_dword s14, s[4:7], 0xc7                 ; C20705C7
>	s_buffer_load_dword s15, s[4:7], 0xc8                 ; C20785C8
>	s_buffer_load_dword s16, s[4:7], 0xc9                 ; C20805C9
>	s_buffer_load_dword s17, s[4:7], 0xcb                 ; C20885CB
>	s_buffer_load_dword s18, s[4:7], 0xcc                 ; C20905CC
>	s_buffer_load_dword s19, s[4:7], 0xcd                 ; C20985CD
>	s_buffer_load_dword s20, s[4:7], 0xcf                 ; C20A05CF
>	s_buffer_load_dword s21, s[4:7], 0xd0                 ; C20A85D0
>	s_buffer_load_dword s22, s[4:7], 0xd1                 ; C20B05D1
>	s_buffer_load_dword s23, s[4:7], 0xd3                 ; C20B85D3
>	s_buffer_load_dword s24, s[4:7], 0xd4                 ; C20C05D4
>	s_buffer_load_dword s25, s[4:7], 0xd5                 ; C20C85D5
>	s_buffer_load_dword s26, s[4:7], 0xd7                 ; C20D05D7
>	s_buffer_load_dword s27, s[4:7], 0xd8                 ; C20D85D8
>	s_buffer_load_dword s4, s[4:7], 0xd9                  ; C20205D9
>	s_buffer_load_dword s5, s[0:3], 0x0                   ; C2028100
>	s_buffer_load_dword s6, s[0:3], 0x1                   ; C2030101
>	s_buffer_load_dword s7, s[0:3], 0x2                   ; C2038102
>	s_buffer_load_dword s28, s[0:3], 0x3                  ; C20E0103
>	s_buffer_load_dword s29, s[0:3], 0x4                  ; C20E8104
>	s_buffer_load_dword s30, s[0:3], 0x5                  ; C20F0105
>	s_buffer_load_dword s31, s[0:3], 0x6                  ; C20F8106
>	s_buffer_load_dword s32, s[0:3], 0x7                  ; C2100107
>	s_buffer_load_dword s33, s[0:3], 0x8                  ; C2108108
>	s_buffer_load_dword s34, s[0:3], 0x9                  ; C2110109
>	s_buffer_load_dword s35, s[0:3], 0xa                  ; C211810A
>	s_buffer_load_dword s36, s[0:3], 0xb                  ; C212010B
>	s_buffer_load_dword s37, s[0:3], 0xc                  ; C212810C
>	s_buffer_load_dword s38, s[0:3], 0xd                  ; C213010D
>	s_buffer_load_dword s39, s[0:3], 0xe                  ; C213810E
>	s_buffer_load_dword s40, s[0:3], 0xf                  ; C214010F
>	s_buffer_load_dword s41, s[0:3], 0x10                 ; C2148110
>	s_buffer_load_dword s42, s[0:3], 0x11                 ; C2150111
>	s_buffer_load_dword s43, s[0:3], 0x12                 ; C2158112
>	s_buffer_load_dword s44, s[0:3], 0x13                 ; C2160113
>	s_buffer_load_dword s45, s[0:3], 0x14                 ; C2168114
>	s_buffer_load_dword s46, s[0:3], 0x15                 ; C2170115
>	s_buffer_load_dword s47, s[0:3], 0x16                 ; C2178116
>	s_buffer_load_dword s48, s[0:3], 0x17                 ; C2180117
>	s_buffer_load_dword s49, s[0:3], 0x18                 ; C2188118
>	s_buffer_load_dword s50, s[0:3], 0x19                 ; C2190119
>	s_buffer_load_dword s51, s[0:3], 0x1a                 ; C219811A
>	s_buffer_load_dword s52, s[0:3], 0x1b                 ; C21A011B
>	s_buffer_load_dword s53, s[0:3], 0x1c                 ; C21A811C
>	s_buffer_load_dword s54, s[0:3], 0x1d                 ; C21B011D
>	s_buffer_load_dword s55, s[0:3], 0x1e                 ; C21B811E
>	s_buffer_load_dword s0, s[0:3], 0x1f                  ; C200011F
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mov_b32_e32 v0, s11                                 ; 7E00020B
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mov_b32_e32 v7, s14                                 ; 7E0E020E
>	v_mov_b32_e32 v8, s17                                 ; 7E100211
>	v_mov_b32_e32 v9, s20                                 ; 7E120214
>	v_mov_b32_e32 v10, s23                                ; 7E140217
>	v_mov_b32_e32 v11, s26                                ; 7E16021A
>	v_mac_f32_e32 v0, s9, v5                              ; 3E000A09
>	v_mac_f32_e32 v7, s12, v5                             ; 3E0E0A0C
>	v_mac_f32_e32 v8, s15, v5                             ; 3E100A0F
>	v_mac_f32_e32 v9, s18, v5                             ; 3E120A12
>	v_mac_f32_e32 v10, s21, v5                            ; 3E140A15
>	v_mac_f32_e32 v11, s24, v5                            ; 3E160A18
>	v_mac_f32_e32 v0, s10, v6                             ; 3E000C0A
>	v_mac_f32_e32 v7, s13, v6                             ; 3E0E0C0D
>	v_mac_f32_e32 v8, s16, v6                             ; 3E100C10
>	v_mac_f32_e32 v9, s19, v6                             ; 3E120C13
>	exp 15, 32, 0, 0, 0, v5, v6, v0, v7                   ; F800020F 07000605
>	v_mac_f32_e32 v10, s22, v6                            ; 3E140C16
>	v_mac_f32_e32 v11, s25, v6                            ; 3E160C19
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mad_f32 v0, v3, s8, -v4                             ; D2820000 84101103
>	v_mad_f32 v5, s27, v4, v1                             ; D2820005 0406081B
>	v_mad_f32 v6, s4, v4, -v2                             ; D2820006 840A0804
>	v_mul_f32_e32 v7, s6, v2                              ; 100E0406
>	exp 15, 33, 0, 0, 0, v8, v9, v10, v11                 ; F800021F 0B0A0908
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mul_f32_e32 v8, s30, v2                             ; 1010041E
>	v_mac_f32_e32 v7, s5, v1                              ; 3E0E0205
>	exp 15, 12, 0, 0, 0, v5, v6, v0, v4                   ; F80000CF 04000605
>	v_mac_f32_e32 v8, s29, v1                             ; 3E10021D
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mul_f32_e32 v0, s34, v2                             ; 10000422
>	v_mac_f32_e32 v0, s33, v1                             ; 3E000221
>	v_mul_f32_e32 v5, s38, v2                             ; 100A0426
>	v_mac_f32_e32 v5, s37, v1                             ; 3E0A0225
>	v_mul_f32_e32 v6, s42, v2                             ; 100C042A
>	v_mul_f32_e32 v9, s46, v2                             ; 1012042E
>	v_mul_f32_e32 v10, s50, v2                            ; 10140432
>	v_mul_f32_e32 v2, s54, v2                             ; 10040436
>	v_mac_f32_e32 v6, s41, v1                             ; 3E0C0229
>	v_mac_f32_e32 v9, s45, v1                             ; 3E12022D
>	v_mac_f32_e32 v10, s49, v1                            ; 3E140231
>	v_mac_f32_e32 v2, s53, v1                             ; 3E040235
>	v_mac_f32_e32 v7, s7, v3                              ; 3E0E0607
>	v_mac_f32_e32 v8, s31, v3                             ; 3E10061F
>	v_mac_f32_e32 v0, s35, v3                             ; 3E000623
>	v_mac_f32_e32 v5, s39, v3                             ; 3E0A0627
>	v_mac_f32_e32 v6, s43, v3                             ; 3E0C062B
>	v_mac_f32_e32 v9, s47, v3                             ; 3E12062F
>	v_mac_f32_e32 v10, s51, v3                            ; 3E140633
>	v_mac_f32_e32 v2, s55, v3                             ; 3E040637
>	v_mac_f32_e32 v7, s28, v4                             ; 3E0E081C
>	v_mac_f32_e32 v8, s32, v4                             ; 3E100820
>	v_mac_f32_e32 v0, s36, v4                             ; 3E000824
>	v_mac_f32_e32 v5, s40, v4                             ; 3E0A0828
>	v_mac_f32_e32 v6, s44, v4                             ; 3E0C082C
>	v_mac_f32_e32 v9, s48, v4                             ; 3E120830
>	v_mac_f32_e32 v10, s52, v4                            ; 3E140834
>	v_mac_f32_e32 v2, s0, v4                              ; 3E040800
>	exp 15, 13, 0, 0, 0, v7, v8, v0, v5                   ; F80000DF 05000807
>	exp 15, 14, 0, 1, 0, v6, v9, v10, v2                  ; F80008EF 020A0906
>	s_endpgm                                              ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 64
>VGPRS: 12
>Code Size: 544 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
>  clamp_color = 0
>FRAG
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL IN[1], GENERIC[1], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SAMP[1]
>DCL SAMP[2]
>DCL SVIEW[0], 2D, FLOAT
>DCL SVIEW[1], 2D, FLOAT
>DCL SVIEW[2], 2D, FLOAT
>DCL CONST[0..3]
>DCL TEMP[0..3], LOCAL
>IMM[0] FLT32 {    6.0000,     1.0000,    -0.5800,    10.0000}
>IMM[1] FLT32 {   -2.0000,     3.0000,    -1.0000,    -0.0000}
>IMM[2] FLT32 {    1.0000,     0.0000,     0.0000,     0.0000}
>  0: MOV TEMP[0].xy, IN[0].xyyy
>  1: TEX TEMP[0].xw, TEMP[0], SAMP[0], 2D
>  2: MOV TEMP[1].xy, IN[1].xyyy
>  3: TEX TEMP[1].y, TEMP[1], SAMP[1], 2D
>  4: MAD TEMP[0].x, TEMP[1].yyyy, TEMP[0].xxxx, TEMP[0].wwww
>  5: MAD TEMP[2].x, CONST[3].wwww, IMM[0].xxxx, IMM[0].yyyy
>  6: MAD TEMP[0].x, TEMP[0].xxxx, TEMP[2].xxxx, IMM[0].zzzz
>  7: MUL TEMP[0].x, TEMP[0].xxxx, IMM[0].wwww
>  8: MOV_SAT TEMP[2].x, TEMP[0].xxxx
>  9: MAD TEMP[3].x, TEMP[2].xxxx, IMM[1].xxxx, IMM[1].yyyy
> 10: MUL TEMP[0].x, TEMP[2].xxxx, TEMP[2].xxxx
> 11: MUL TEMP[0].x, TEMP[0].xxxx, TEMP[3].xxxx
> 12: MOV TEMP[2].xy, IN[0].xyyy
> 13: TEX TEMP[2].xyz, TEMP[2], SAMP[2], 2D
> 14: MAD TEMP[2].xyz, CONST[3].yyyy, IMM[1].zwww, TEMP[2].xyzz
> 15: MUL TEMP[1].xyz, IMM[2].xyyy, CONST[3].yyyy
> 16: MAD TEMP[0].xyz, TEMP[0].xxxx, TEMP[2].xyzz, TEMP[1].xyzz
> 17: MOV TEMP[0].w, IMM[0].yyyy
> 18: MOV OUT[0], TEMP[0]
> 19: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %24 = load <16 x i8>, <16 x i8> addrspace(2)* %23, align 16, !tbaa !0
>  %25 = call float @llvm.SI.load.const(<16 x i8> %24, i32 52)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %24, i32 60)
>  %27 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %28 = load <8 x i32>, <8 x i32> addrspace(2)* %27, align 32, !tbaa !0
>  %29 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %30 = load <4 x i32>, <4 x i32> addrspace(2)* %29, align 16, !tbaa !0
>  %31 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 1
>  %32 = load <8 x i32>, <8 x i32> addrspace(2)* %31, align 32, !tbaa !0
>  %33 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 1
>  %34 = load <4 x i32>, <4 x i32> addrspace(2)* %33, align 16, !tbaa !0
>  %35 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 2
>  %36 = load <8 x i32>, <8 x i32> addrspace(2)* %35, align 32, !tbaa !0
>  %37 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 2
>  %38 = load <4 x i32>, <4 x i32> addrspace(2)* %37, align 16, !tbaa !0
>  %39 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %8)
>  %40 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %8)
>  %41 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %6, <2 x i32> %8)
>  %42 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %6, <2 x i32> %8)
>  %43 = bitcast float %39 to i32
>  %44 = bitcast float %40 to i32
>  %45 = insertelement <2 x i32> undef, i32 %43, i32 0
>  %46 = insertelement <2 x i32> %45, i32 %44, i32 1
>  %47 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %46, <8 x i32> %28, <4 x i32> %30, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %48 = extractelement <4 x float> %47, i32 0
>  %49 = extractelement <4 x float> %47, i32 3
>  %50 = bitcast float %41 to i32
>  %51 = bitcast float %42 to i32
>  %52 = insertelement <2 x i32> undef, i32 %50, i32 0
>  %53 = insertelement <2 x i32> %52, i32 %51, i32 1
>  %54 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %53, <8 x i32> %32, <4 x i32> %34, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %55 = extractelement <4 x float> %54, i32 1
>  %56 = fmul float %55, %48
>  %57 = fadd float %56, %49
>  %58 = fmul float %26, 6.000000e+00
>  %59 = fadd float %58, 1.000000e+00
>  %60 = fmul float %57, %59
>  %61 = fadd float %60, 0xBFE28F5C20000000
>  %62 = fmul float %61, 1.000000e+01
>  %63 = call float @llvm.AMDIL.clamp.(float %62, float 0.000000e+00, float 1.000000e+00)
>  %64 = fmul float %63, -2.000000e+00
>  %65 = fadd float %64, 3.000000e+00
>  %66 = fmul float %63, %63
>  %67 = fmul float %66, %65
>  %68 = bitcast float %39 to i32
>  %69 = bitcast float %40 to i32
>  %70 = insertelement <2 x i32> undef, i32 %68, i32 0
>  %71 = insertelement <2 x i32> %70, i32 %69, i32 1
>  %72 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %71, <8 x i32> %36, <4 x i32> %38, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %73 = extractelement <4 x float> %72, i32 0
>  %74 = extractelement <4 x float> %72, i32 1
>  %75 = extractelement <4 x float> %72, i32 2
>  %76 = fsub float %73, %25
>  %77 = fmul float %25, -0.000000e+00
>  %78 = fadd float %77, %74
>  %79 = fmul float %25, -0.000000e+00
>  %80 = fadd float %79, %75
>  %81 = fmul float %25, 0.000000e+00
>  %82 = fmul float %25, 0.000000e+00
>  %83 = fmul float %67, %76
>  %84 = fadd float %83, %25
>  %85 = fmul float %67, %78
>  %86 = fadd float %85, %81
>  %87 = fmul float %67, %80
>  %88 = fadd float %87, %82
>  %89 = call i32 @llvm.SI.packf16(float %84, float %86)
>  %90 = bitcast i32 %89 to float
>  %91 = call i32 @llvm.SI.packf16(float %88, float 1.000000e+00)
>  %92 = bitcast i32 %91 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %90, float %92, float %90, float %92)
>  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.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #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, s10                                                       ; BEFC030A
>	v_interp_p1_f32 v2, v0, 0, 0, [m0]                                      ; C8080000
>	s_load_dwordx4 s[0:3], s[2:3], 0x0                                      ; C0800300
>	s_load_dwordx4 s[8:11], s[4:5], 0x0                                     ; C0840500
>	s_load_dwordx4 s[12:15], s[4:5], 0x4                                    ; C0860504
>	s_load_dwordx4 s[16:19], s[4:5], 0x8                                    ; C0880508
>	v_interp_p2_f32 v2, [v2], v1, 0, 0, [m0]                                ; C8090001
>	s_load_dwordx8 s[20:27], s[6:7], 0x0                                    ; C0CA0700
>	s_load_dwordx8 s[28:35], s[6:7], 0x8                                    ; C0CE0708
>	s_load_dwordx8 s[36:43], s[6:7], 0x10                                   ; C0D20710
>	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, 1, [m0]                                      ; C8100400
>	s_waitcnt lgkmcnt(0)                                                    ; BF8C007F
>	s_buffer_load_dword s4, s[0:3], 0xf                                     ; C202010F
>	v_interp_p2_f32 v4, [v4], v1, 0, 1, [m0]                                ; C8110401
>	v_interp_p1_f32 v5, v0, 1, 1, [m0]                                      ; C8140500
>	v_interp_p2_f32 v5, [v5], v1, 1, 1, [m0]                                ; C8150501
>	image_sample v[0:1], 9, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[20:27], s[8:11]  ; F0800900 00450002
>	image_sample v4, 2, 0, 0, 0, 0, 0, 0, 0, v[4:5], s[28:35], s[12:15]     ; F0800200 00670404
>	image_sample v[5:7], 7, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[36:43], s[16:19] ; F0800700 00890502
>	s_buffer_load_dword s0, s[0:3], 0xd                                     ; C200010D
>	s_waitcnt vmcnt(1)                                                      ; BF8C0771
>	v_mac_f32_e32 v1, v0, v4                                                ; 3E020900
>	v_mov_b32_e32 v0, 0x40c00000                                            ; 7E0002FF 40C00000
>	s_waitcnt lgkmcnt(0)                                                    ; BF8C007F
>	v_mad_f32 v0, v0, s4, 1.0                                               ; D2820000 03C80900
>	v_madak_f32_e32 v0, v1, v0, 0xbf147ae1                                  ; 42000101 BF147AE1
>	v_mul_f32_e32 v0, 0x41200000, v0                                        ; 100000FF 41200000
>	v_add_f32_e64 v0, 0, v0 clamp                                           ; D2060800 00020080
>	v_madak_f32_e32 v1, -2.0, v0, 0x40400000                                ; 420200F5 40400000
>	v_mul_f32_e32 v0, v0, v0                                                ; 10000100
>	v_mul_f32_e32 v0, v1, v0                                                ; 10000101
>	s_waitcnt vmcnt(0)                                                      ; BF8C0770
>	v_subrev_f32_e32 v1, s0, v5                                             ; 0A020A00
>	v_mov_b32_e32 v2, 0x80000000                                            ; 7E0402FF 80000000
>	v_mad_f32 v3, s0, v2, v6                                                ; D2820003 041A0400
>	v_mac_f32_e32 v7, s0, v2                                                ; 3E0E0400
>	v_mul_f32_e64 v2, 0, s0                                                 ; D2100002 00000080
>	v_mad_f32 v1, v0, v1, s0                                                ; D2820001 00020300
>	v_mad_f32 v3, v3, v0, v2                                                ; D2820003 040A0103
>	v_mac_f32_e32 v2, v7, v0                                                ; 3E040107
>	v_cvt_pkrtz_f16_f32_e32 v0, v1, v3                                      ; 5E000701
>	v_cvt_pkrtz_f16_f32_e64 v1, v2, 1.0                                     ; D25E0001 0001E502
>	exp 15, 0, 1, 1, 1, v0, v1, v0, v1                                      ; F8001C0F 01000100
>	s_endpgm                                                                ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 48
>VGPRS: 8
>Code Size: 252 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
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL OUT[1], CLIPVERTEX
>DCL OUT[2], GENERIC[0]
>DCL OUT[3], GENERIC[1]
>DCL CONST[0..54]
>DCL TEMP[0..3], LOCAL
>  0: MUL TEMP[0].xy, CONST[48].xyyy, IN[1].xyyy
>  1: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
>  2: ADD TEMP[1].x, TEMP[0].xxxx, CONST[48].wwww
>  3: MOV TEMP[1].z, TEMP[1].xxxx
>  4: MUL TEMP[0].xy, CONST[49].xyyy, IN[1].xyyy
>  5: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
>  6: ADD TEMP[2].x, TEMP[0].xxxx, CONST[49].wwww
>  7: MOV TEMP[1].w, TEMP[2].xxxx
>  8: MUL TEMP[0].xy, CONST[50].xyyy, IN[1].xyyy
>  9: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 10: ADD TEMP[2].x, TEMP[0].xxxx, CONST[50].wwww
> 11: MUL TEMP[0].xy, CONST[51].xyyy, IN[1].xyyy
> 12: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 13: ADD TEMP[3].x, TEMP[0].xxxx, CONST[51].wwww
> 14: MOV TEMP[2].y, TEMP[3].xxxx
> 15: MUL TEMP[0].xy, CONST[52].xyyy, IN[1].xyyy
> 16: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 17: ADD TEMP[3].x, TEMP[0].xxxx, CONST[52].wwww
> 18: MOV TEMP[2].z, TEMP[3].xxxx
> 19: MUL TEMP[0].xy, CONST[53].xyyy, IN[1].xyyy
> 20: ADD TEMP[0].x, TEMP[0].yyyy, TEMP[0].xxxx
> 21: ADD TEMP[0].x, TEMP[0].xxxx, CONST[53].wwww
> 22: MOV TEMP[2].w, TEMP[0].xxxx
> 23: MOV TEMP[0].xw, IN[0].xxxw
> 24: MOV TEMP[1].xy, IN[1].xyxx
> 25: MAD TEMP[3].x, IN[0].zzzz, CONST[0].zzzz, -IN[0].wwww
> 26: MOV TEMP[0].z, TEMP[3].xxxx
> 27: MOV TEMP[0].y, -IN[0].yyyy
> 28: MAD TEMP[0].xy, CONST[54].xyyy, IN[0].wwww, TEMP[0].xyyy
> 29: MOV OUT[2], TEMP[1]
> 30: MOV OUT[3], TEMP[2]
> 31: MOV OUT[0], TEMP[0]
> 32: MOV OUT[1], IN[0]
> 33: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = call float @llvm.SI.load.const(<16 x i8> %13, i32 8)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %13, i32 768)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %13, i32 772)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %13, i32 780)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %13, i32 784)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %13, i32 788)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %13, i32 796)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %13, i32 800)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %13, i32 804)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %13, i32 812)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %13, i32 816)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %13, i32 820)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %13, i32 828)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %13, i32 832)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %13, i32 836)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %13, i32 844)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %13, i32 848)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %13, i32 852)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %13, i32 860)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %13, i32 864)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %13, i32 868)
>  %35 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %36 = load <16 x i8>, <16 x i8> addrspace(2)* %35, align 16, !tbaa !0
>  %37 = add i32 %5, %8
>  %38 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %36, i32 0, i32 %37)
>  %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 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 1
>  %44 = load <16 x i8>, <16 x i8> addrspace(2)* %43, align 16, !tbaa !0
>  %45 = add i32 %5, %8
>  %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 = fmul float %15, %47
>  %50 = fmul float %16, %48
>  %51 = fadd float %50, %49
>  %52 = fadd float %51, %17
>  %53 = fmul float %18, %47
>  %54 = fmul float %19, %48
>  %55 = fadd float %54, %53
>  %56 = fadd float %55, %20
>  %57 = fmul float %21, %47
>  %58 = fmul float %22, %48
>  %59 = fadd float %58, %57
>  %60 = fadd float %59, %23
>  %61 = fmul float %24, %47
>  %62 = fmul float %25, %48
>  %63 = fadd float %62, %61
>  %64 = fadd float %63, %26
>  %65 = fmul float %27, %47
>  %66 = fmul float %28, %48
>  %67 = fadd float %66, %65
>  %68 = fadd float %67, %29
>  %69 = fmul float %30, %47
>  %70 = fmul float %31, %48
>  %71 = fadd float %70, %69
>  %72 = fadd float %71, %32
>  %73 = fmul float %41, %14
>  %74 = fsub float %73, %42
>  %75 = fmul float %33, %42
>  %76 = fadd float %75, %39
>  %77 = fmul float %34, %42
>  %78 = fsub float %77, %40
>  %79 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 16
>  %80 = load <16 x i8>, <16 x i8> addrspace(2)* %79, align 16, !tbaa !0
>  %81 = call float @llvm.SI.load.const(<16 x i8> %80, i32 0)
>  %82 = fmul float %81, %39
>  %83 = call float @llvm.SI.load.const(<16 x i8> %80, i32 4)
>  %84 = fmul float %83, %40
>  %85 = fadd float %82, %84
>  %86 = call float @llvm.SI.load.const(<16 x i8> %80, i32 8)
>  %87 = fmul float %86, %41
>  %88 = fadd float %85, %87
>  %89 = call float @llvm.SI.load.const(<16 x i8> %80, i32 12)
>  %90 = fmul float %89, %42
>  %91 = fadd float %88, %90
>  %92 = call float @llvm.SI.load.const(<16 x i8> %80, i32 16)
>  %93 = fmul float %92, %39
>  %94 = call float @llvm.SI.load.const(<16 x i8> %80, i32 20)
>  %95 = fmul float %94, %40
>  %96 = fadd float %93, %95
>  %97 = call float @llvm.SI.load.const(<16 x i8> %80, i32 24)
>  %98 = fmul float %97, %41
>  %99 = fadd float %96, %98
>  %100 = call float @llvm.SI.load.const(<16 x i8> %80, i32 28)
>  %101 = fmul float %100, %42
>  %102 = fadd float %99, %101
>  %103 = call float @llvm.SI.load.const(<16 x i8> %80, i32 32)
>  %104 = fmul float %103, %39
>  %105 = call float @llvm.SI.load.const(<16 x i8> %80, i32 36)
>  %106 = fmul float %105, %40
>  %107 = fadd float %104, %106
>  %108 = call float @llvm.SI.load.const(<16 x i8> %80, i32 40)
>  %109 = fmul float %108, %41
>  %110 = fadd float %107, %109
>  %111 = call float @llvm.SI.load.const(<16 x i8> %80, i32 44)
>  %112 = fmul float %111, %42
>  %113 = fadd float %110, %112
>  %114 = call float @llvm.SI.load.const(<16 x i8> %80, i32 48)
>  %115 = fmul float %114, %39
>  %116 = call float @llvm.SI.load.const(<16 x i8> %80, i32 52)
>  %117 = fmul float %116, %40
>  %118 = fadd float %115, %117
>  %119 = call float @llvm.SI.load.const(<16 x i8> %80, i32 56)
>  %120 = fmul float %119, %41
>  %121 = fadd float %118, %120
>  %122 = call float @llvm.SI.load.const(<16 x i8> %80, i32 60)
>  %123 = fmul float %122, %42
>  %124 = fadd float %121, %123
>  %125 = call float @llvm.SI.load.const(<16 x i8> %80, i32 64)
>  %126 = fmul float %125, %39
>  %127 = call float @llvm.SI.load.const(<16 x i8> %80, i32 68)
>  %128 = fmul float %127, %40
>  %129 = fadd float %126, %128
>  %130 = call float @llvm.SI.load.const(<16 x i8> %80, i32 72)
>  %131 = fmul float %130, %41
>  %132 = fadd float %129, %131
>  %133 = call float @llvm.SI.load.const(<16 x i8> %80, i32 76)
>  %134 = fmul float %133, %42
>  %135 = fadd float %132, %134
>  %136 = call float @llvm.SI.load.const(<16 x i8> %80, i32 80)
>  %137 = fmul float %136, %39
>  %138 = call float @llvm.SI.load.const(<16 x i8> %80, i32 84)
>  %139 = fmul float %138, %40
>  %140 = fadd float %137, %139
>  %141 = call float @llvm.SI.load.const(<16 x i8> %80, i32 88)
>  %142 = fmul float %141, %41
>  %143 = fadd float %140, %142
>  %144 = call float @llvm.SI.load.const(<16 x i8> %80, i32 92)
>  %145 = fmul float %144, %42
>  %146 = fadd float %143, %145
>  %147 = call float @llvm.SI.load.const(<16 x i8> %80, i32 96)
>  %148 = fmul float %147, %39
>  %149 = call float @llvm.SI.load.const(<16 x i8> %80, i32 100)
>  %150 = fmul float %149, %40
>  %151 = fadd float %148, %150
>  %152 = call float @llvm.SI.load.const(<16 x i8> %80, i32 104)
>  %153 = fmul float %152, %41
>  %154 = fadd float %151, %153
>  %155 = call float @llvm.SI.load.const(<16 x i8> %80, i32 108)
>  %156 = fmul float %155, %42
>  %157 = fadd float %154, %156
>  %158 = call float @llvm.SI.load.const(<16 x i8> %80, i32 112)
>  %159 = fmul float %158, %39
>  %160 = call float @llvm.SI.load.const(<16 x i8> %80, i32 116)
>  %161 = fmul float %160, %40
>  %162 = fadd float %159, %161
>  %163 = call float @llvm.SI.load.const(<16 x i8> %80, i32 120)
>  %164 = fmul float %163, %41
>  %165 = fadd float %162, %164
>  %166 = call float @llvm.SI.load.const(<16 x i8> %80, i32 124)
>  %167 = fmul float %166, %42
>  %168 = fadd float %165, %167
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %47, float %48, float %52, float %56)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %60, float %64, float %68, float %72)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 12, i32 0, float %76, float %78, float %74, float %42)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 13, i32 0, float %91, float %102, float %113, float %124)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 14, i32 0, float %135, float %146, float %157, 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
>
>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[4:7], s[8:9], 0x0                    ; C0820900
>	s_load_dwordx4 s[12:15], s[8:9], 0x4                  ; C0860904
>	v_add_i32_e32 v0, vcc, s10, v0                        ; 4A00000A
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	buffer_load_format_xyzw v[1:4], v0, s[4:7], 0 idxen   ; E00C2000 80010100
>	buffer_load_format_xyzw v[5:8], v0, s[12:15], 0 idxen ; E00C2000 80030500
>	s_load_dwordx4 s[4:7], s[2:3], 0x0                    ; C0820300
>	s_load_dwordx4 s[0:3], s[2:3], 0x40                   ; C0800340
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	s_buffer_load_dword s8, s[4:7], 0x2                   ; C2040502
>	s_buffer_load_dword s9, s[4:7], 0xc0                  ; C20485C0
>	s_buffer_load_dword s10, s[4:7], 0xc1                 ; C20505C1
>	s_buffer_load_dword s11, s[4:7], 0xc3                 ; C20585C3
>	s_buffer_load_dword s12, s[4:7], 0xc4                 ; C20605C4
>	s_buffer_load_dword s13, s[4:7], 0xc5                 ; C20685C5
>	s_buffer_load_dword s14, s[4:7], 0xc7                 ; C20705C7
>	s_buffer_load_dword s15, s[4:7], 0xc8                 ; C20785C8
>	s_buffer_load_dword s16, s[4:7], 0xc9                 ; C20805C9
>	s_buffer_load_dword s17, s[4:7], 0xcb                 ; C20885CB
>	s_buffer_load_dword s18, s[4:7], 0xcc                 ; C20905CC
>	s_buffer_load_dword s19, s[4:7], 0xcd                 ; C20985CD
>	s_buffer_load_dword s20, s[4:7], 0xcf                 ; C20A05CF
>	s_buffer_load_dword s21, s[4:7], 0xd0                 ; C20A85D0
>	s_buffer_load_dword s22, s[4:7], 0xd1                 ; C20B05D1
>	s_buffer_load_dword s23, s[4:7], 0xd3                 ; C20B85D3
>	s_buffer_load_dword s24, s[4:7], 0xd4                 ; C20C05D4
>	s_buffer_load_dword s25, s[4:7], 0xd5                 ; C20C85D5
>	s_buffer_load_dword s26, s[4:7], 0xd7                 ; C20D05D7
>	s_buffer_load_dword s27, s[4:7], 0xd8                 ; C20D85D8
>	s_buffer_load_dword s4, s[4:7], 0xd9                  ; C20205D9
>	s_buffer_load_dword s5, s[0:3], 0x0                   ; C2028100
>	s_buffer_load_dword s6, s[0:3], 0x1                   ; C2030101
>	s_buffer_load_dword s7, s[0:3], 0x2                   ; C2038102
>	s_buffer_load_dword s28, s[0:3], 0x3                  ; C20E0103
>	s_buffer_load_dword s29, s[0:3], 0x4                  ; C20E8104
>	s_buffer_load_dword s30, s[0:3], 0x5                  ; C20F0105
>	s_buffer_load_dword s31, s[0:3], 0x6                  ; C20F8106
>	s_buffer_load_dword s32, s[0:3], 0x7                  ; C2100107
>	s_buffer_load_dword s33, s[0:3], 0x8                  ; C2108108
>	s_buffer_load_dword s34, s[0:3], 0x9                  ; C2110109
>	s_buffer_load_dword s35, s[0:3], 0xa                  ; C211810A
>	s_buffer_load_dword s36, s[0:3], 0xb                  ; C212010B
>	s_buffer_load_dword s37, s[0:3], 0xc                  ; C212810C
>	s_buffer_load_dword s38, s[0:3], 0xd                  ; C213010D
>	s_buffer_load_dword s39, s[0:3], 0xe                  ; C213810E
>	s_buffer_load_dword s40, s[0:3], 0xf                  ; C214010F
>	s_buffer_load_dword s41, s[0:3], 0x10                 ; C2148110
>	s_buffer_load_dword s42, s[0:3], 0x11                 ; C2150111
>	s_buffer_load_dword s43, s[0:3], 0x12                 ; C2158112
>	s_buffer_load_dword s44, s[0:3], 0x13                 ; C2160113
>	s_buffer_load_dword s45, s[0:3], 0x14                 ; C2168114
>	s_buffer_load_dword s46, s[0:3], 0x15                 ; C2170115
>	s_buffer_load_dword s47, s[0:3], 0x16                 ; C2178116
>	s_buffer_load_dword s48, s[0:3], 0x17                 ; C2180117
>	s_buffer_load_dword s49, s[0:3], 0x18                 ; C2188118
>	s_buffer_load_dword s50, s[0:3], 0x19                 ; C2190119
>	s_buffer_load_dword s51, s[0:3], 0x1a                 ; C219811A
>	s_buffer_load_dword s52, s[0:3], 0x1b                 ; C21A011B
>	s_buffer_load_dword s53, s[0:3], 0x1c                 ; C21A811C
>	s_buffer_load_dword s54, s[0:3], 0x1d                 ; C21B011D
>	s_buffer_load_dword s55, s[0:3], 0x1e                 ; C21B811E
>	s_buffer_load_dword s0, s[0:3], 0x1f                  ; C200011F
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mov_b32_e32 v0, s11                                 ; 7E00020B
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mov_b32_e32 v7, s14                                 ; 7E0E020E
>	v_mov_b32_e32 v8, s17                                 ; 7E100211
>	v_mov_b32_e32 v9, s20                                 ; 7E120214
>	v_mov_b32_e32 v10, s23                                ; 7E140217
>	v_mov_b32_e32 v11, s26                                ; 7E16021A
>	v_mac_f32_e32 v0, s9, v5                              ; 3E000A09
>	v_mac_f32_e32 v7, s12, v5                             ; 3E0E0A0C
>	v_mac_f32_e32 v8, s15, v5                             ; 3E100A0F
>	v_mac_f32_e32 v9, s18, v5                             ; 3E120A12
>	v_mac_f32_e32 v10, s21, v5                            ; 3E140A15
>	v_mac_f32_e32 v11, s24, v5                            ; 3E160A18
>	v_mac_f32_e32 v0, s10, v6                             ; 3E000C0A
>	v_mac_f32_e32 v7, s13, v6                             ; 3E0E0C0D
>	v_mac_f32_e32 v8, s16, v6                             ; 3E100C10
>	v_mac_f32_e32 v9, s19, v6                             ; 3E120C13
>	exp 15, 32, 0, 0, 0, v5, v6, v0, v7                   ; F800020F 07000605
>	v_mac_f32_e32 v10, s22, v6                            ; 3E140C16
>	v_mac_f32_e32 v11, s25, v6                            ; 3E160C19
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mad_f32 v0, v3, s8, -v4                             ; D2820000 84101103
>	v_mad_f32 v5, s27, v4, v1                             ; D2820005 0406081B
>	v_mad_f32 v6, s4, v4, -v2                             ; D2820006 840A0804
>	v_mul_f32_e32 v7, s6, v2                              ; 100E0406
>	exp 15, 33, 0, 0, 0, v8, v9, v10, v11                 ; F800021F 0B0A0908
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mul_f32_e32 v8, s30, v2                             ; 1010041E
>	v_mac_f32_e32 v7, s5, v1                              ; 3E0E0205
>	exp 15, 12, 0, 0, 0, v5, v6, v0, v4                   ; F80000CF 04000605
>	v_mac_f32_e32 v8, s29, v1                             ; 3E10021D
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mul_f32_e32 v0, s34, v2                             ; 10000422
>	v_mac_f32_e32 v0, s33, v1                             ; 3E000221
>	v_mul_f32_e32 v5, s38, v2                             ; 100A0426
>	v_mac_f32_e32 v5, s37, v1                             ; 3E0A0225
>	v_mul_f32_e32 v6, s42, v2                             ; 100C042A
>	v_mul_f32_e32 v9, s46, v2                             ; 1012042E
>	v_mul_f32_e32 v10, s50, v2                            ; 10140432
>	v_mul_f32_e32 v2, s54, v2                             ; 10040436
>	v_mac_f32_e32 v6, s41, v1                             ; 3E0C0229
>	v_mac_f32_e32 v9, s45, v1                             ; 3E12022D
>	v_mac_f32_e32 v10, s49, v1                            ; 3E140231
>	v_mac_f32_e32 v2, s53, v1                             ; 3E040235
>	v_mac_f32_e32 v7, s7, v3                              ; 3E0E0607
>	v_mac_f32_e32 v8, s31, v3                             ; 3E10061F
>	v_mac_f32_e32 v0, s35, v3                             ; 3E000623
>	v_mac_f32_e32 v5, s39, v3                             ; 3E0A0627
>	v_mac_f32_e32 v6, s43, v3                             ; 3E0C062B
>	v_mac_f32_e32 v9, s47, v3                             ; 3E12062F
>	v_mac_f32_e32 v10, s51, v3                            ; 3E140633
>	v_mac_f32_e32 v2, s55, v3                             ; 3E040637
>	v_mac_f32_e32 v7, s28, v4                             ; 3E0E081C
>	v_mac_f32_e32 v8, s32, v4                             ; 3E100820
>	v_mac_f32_e32 v0, s36, v4                             ; 3E000824
>	v_mac_f32_e32 v5, s40, v4                             ; 3E0A0828
>	v_mac_f32_e32 v6, s44, v4                             ; 3E0C082C
>	v_mac_f32_e32 v9, s48, v4                             ; 3E120830
>	v_mac_f32_e32 v10, s52, v4                            ; 3E140834
>	v_mac_f32_e32 v2, s0, v4                              ; 3E040800
>	exp 15, 13, 0, 0, 0, v7, v8, v0, v5                   ; F80000DF 05000807
>	exp 15, 14, 0, 1, 0, v6, v9, v10, v2                  ; F80008EF 020A0906
>	s_endpgm                                              ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 64
>VGPRS: 12
>Code Size: 544 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
>  clamp_color = 0
>FRAG
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL IN[1], GENERIC[1], 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 SVIEW[0], 2D, FLOAT
>DCL SVIEW[1], 2D, FLOAT
>DCL SVIEW[2], 2D, FLOAT
>DCL SVIEW[3], 2D, FLOAT
>DCL SVIEW[4], 2D, FLOAT
>DCL SVIEW[5], 2D, FLOAT
>DCL SVIEW[6], 2D, FLOAT
>DCL CONST[0..11]
>DCL TEMP[0..8], LOCAL
>IMM[0] FLT32 {    1.0000,    -1.0000,     0.0588,    16.0000}
>IMM[1] FLT32 {   -0.0011,     0.9960,     0.0020,     0.9966}
>IMM[2] FLT32 {    0.0040,     0.9989,    -0.0007,     0.9973}
>IMM[3] FLT32 {    0.0010,     0.9983,     0.0027,     0.9993}
>IMM[4] FLT32 {   -0.0034,     0.9980,    -0.0017,     0.9990}
>IMM[5] FLT32 {    0.0000,     1.0000,     0.0017,     1.0010}
>IMM[6] FLT32 {    0.0034,     1.0020,    -0.0027,     1.0007}
>IMM[7] FLT32 {   -0.0010,     1.0017,     0.0007,     1.0027}
>IMM[8] FLT32 {   -0.0040,     1.0011,    -0.0020,     1.0034}
>IMM[9] FLT32 {    0.0011,     1.0040,     2.0000,     7.0000}
>IMM[10] FLT32 {    0.0600,     0.1200,     0.1800,     6.0000}
>IMM[11] FLT32 {   -0.0600,    -0.1200,    -0.1800,    -0.5800}
>IMM[12] FLT32 {    0.5600,    -0.0200,     0.5400,     0.5200}
>IMM[13] FLT32 {   10.0000,    -2.0000,     3.0000,     0.7500}
>IMM[14] FLT32 {    0.2500,     0.3000,     0.5900,     0.1100}
>IMM[15] FLT32 {   -0.0800,    14.2857,     0.0300,     0.5000}
>IMM[16] FLT32 {   -0.9000,    -0.0100,    10.0000,  -100.0000}
>  0: MAD TEMP[0], IN[0].xyxy, IMM[0].xyxy, IMM[1]
>  1: MOV TEMP[1].xy, TEMP[0].zwww
>  2: TEX TEMP[1], TEMP[1], SAMP[5], 2D
>  3: MUL TEMP[1], TEMP[1], IMM[0].zzzz
>  4: MOV TEMP[2].xy, TEMP[0].xyyy
>  5: TEX TEMP[2], TEMP[2], SAMP[5], 2D
>  6: MAD TEMP[0], TEMP[2], IMM[0].zzzz, TEMP[1]
>  7: MAD TEMP[1], IN[0].xyxy, IMM[0].xyxy, IMM[2]
>  8: MOV TEMP[2].xy, TEMP[1].xyyy
>  9: TEX TEMP[2], TEMP[2], SAMP[5], 2D
> 10: MAD TEMP[0], TEMP[2], IMM[0].zzzz, TEMP[0]
> 11: MOV TEMP[2].xy, TEMP[1].zwww
> 12: TEX TEMP[2], TEMP[2], SAMP[5], 2D
> 13: MAD TEMP[0], TEMP[2], IMM[0].zzzz, TEMP[0]
> 14: MAD TEMP[1], IN[0].xyxy, IMM[0].xyxy, IMM[3]
> 15: MOV TEMP[2].xy, TEMP[1].xyyy
> 16: TEX TEMP[2], TEMP[2], SAMP[5], 2D
> 17: MAD TEMP[0], TEMP[2], IMM[0].zzzz, TEMP[0]
> 18: MOV TEMP[2].xy, TEMP[1].zwww
> 19: TEX TEMP[2], TEMP[2], SAMP[5], 2D
> 20: MAD TEMP[0], TEMP[2], IMM[0].zzzz, TEMP[0]
> 21: MAD TEMP[1], IN[0].xyxy, IMM[0].xyxy, IMM[4]
> 22: MOV TEMP[2].xy, TEMP[1].xyyy
> 23: TEX TEMP[2], TEMP[2], SAMP[5], 2D
> 24: MAD TEMP[0], TEMP[2], IMM[0].zzzz, TEMP[0]
> 25: MOV TEMP[2].xy, TEMP[1].zwww
> 26: TEX TEMP[2], TEMP[2], SAMP[5], 2D
> 27: MAD TEMP[0], TEMP[2], IMM[0].zzzz, TEMP[0]
> 28: MAD TEMP[1], IN[0].xyxy, IMM[0].xyxy, IMM[5]
> 29: MOV TEMP[2].xy, TEMP[1].xyyy
> 30: TEX TEMP[2], TEMP[2], SAMP[5], 2D
> 31: MAD TEMP[0], TEMP[2], IMM[0].zzzz, TEMP[0]
> 32: MOV TEMP[2].xy, TEMP[1].zwww
> 33: TEX TEMP[2], TEMP[2], SAMP[5], 2D
> 34: MAD TEMP[0], TEMP[2], IMM[0].zzzz, TEMP[0]
> 35: MAD TEMP[1], IN[0].xyxy, IMM[0].xyxy, IMM[6]
> 36: MOV TEMP[2].xy, TEMP[1].xyyy
> 37: TEX TEMP[2], TEMP[2], SAMP[5], 2D
> 38: MAD TEMP[0], TEMP[2], IMM[0].zzzz, TEMP[0]
> 39: MOV TEMP[2].xy, TEMP[1].zwww
> 40: TEX TEMP[2], TEMP[2], SAMP[5], 2D
> 41: MAD TEMP[0], TEMP[2], IMM[0].zzzz, TEMP[0]
> 42: MAD TEMP[1], IN[0].xyxy, IMM[0].xyxy, IMM[7]
> 43: MOV TEMP[2].xy, TEMP[1].xyyy
> 44: TEX TEMP[2], TEMP[2], SAMP[5], 2D
> 45: MAD TEMP[0], TEMP[2], IMM[0].zzzz, TEMP[0]
> 46: MOV TEMP[2].xy, TEMP[1].zwww
> 47: TEX TEMP[2], TEMP[2], SAMP[5], 2D
> 48: MAD TEMP[0], TEMP[2], IMM[0].zzzz, TEMP[0]
> 49: MAD TEMP[2], IN[0].xyxy, IMM[0].xyxy, IMM[8]
> 50: MOV TEMP[3].xy, TEMP[2].xyyy
> 51: TEX TEMP[3], TEMP[3], SAMP[5], 2D
> 52: MAD TEMP[0], TEMP[3], IMM[0].zzzz, TEMP[0]
> 53: MOV TEMP[3].xy, TEMP[2].zwww
> 54: TEX TEMP[3], TEMP[3], SAMP[5], 2D
> 55: MAD TEMP[2], TEMP[3], IMM[0].zzzz, TEMP[0]
> 56: MAD TEMP[0].xy, IN[0].xyyy, IMM[0].xyyy, IMM[9].xyyy
> 57: MOV TEMP[3].xy, TEMP[0].xyyy
> 58: TEX TEMP[3], TEMP[3], SAMP[5], 2D
> 59: MAD TEMP[2], TEMP[3], IMM[0].zzzz, TEMP[2]
> 60: MUL TEMP[3].x, TEMP[2].wwww, IMM[0].wwww
> 61: MUL TEMP[2].xyz, TEMP[3].xxxx, TEMP[2].xyzz
> 62: DP2 TEMP[3].x, TEMP[2].yzzz, CONST[10].xyyy
> 63: ADD TEMP[0].x, TEMP[3].xxxx, CONST[10].wwww
> 64: DP2 TEMP[3].x, TEMP[2].yzzz, CONST[11].xyyy
> 65: ADD TEMP[3].x, TEMP[3].xxxx, CONST[11].wwww
> 66: MOV TEMP[0].y, TEMP[3].xxxx
> 67: MOV TEMP[3].xy, TEMP[0].xyyy
> 68: TEX TEMP[3].xyz, TEMP[3], SAMP[6], 2D
> 69: DP2 TEMP[4].x, TEMP[2].xzzz, CONST[10].xyyy
> 70: ADD TEMP[1].x, TEMP[4].xxxx, CONST[10].wwww
> 71: DP2 TEMP[4].x, TEMP[2].xzzz, CONST[11].xyyy
> 72: ADD TEMP[4].x, TEMP[4].xxxx, CONST[11].wwww
> 73: MOV TEMP[1].y, TEMP[4].xxxx
> 74: MOV TEMP[4].xy, IN[0].xyyy
> 75: TEX TEMP[4].xyz, TEMP[4], SAMP[4], 2D
> 76: MAD TEMP[4].xyz, TEMP[4].xyzz, IMM[9].zzzz, IMM[0].yyyy
> 77: DP3 TEMP[5].x, TEMP[4].xyzz, TEMP[4].xyzz
> 78: RSQ TEMP[5].x, TEMP[5].xxxx
> 79: MUL TEMP[5].xy, TEMP[5].xxxx, TEMP[4].yzzz
> 80: ABS TEMP[6].x, TEMP[5].xxxx
> 81: POW TEMP[6].x, TEMP[6].xxxx, IMM[9].wwww
> 82: MOV TEMP[7].xy, TEMP[1].xyyy
> 83: TEX TEMP[7].xyz, TEMP[7], SAMP[6], 2D
> 84: LRP TEMP[4].xyz, TEMP[6].xxxx, TEMP[7].xyzz, TEMP[3].xyzz
> 85: DP2 TEMP[3].x, TEMP[2].yxxx, CONST[10].xyyy
> 86: ADD TEMP[0].x, TEMP[3].xxxx, CONST[10].wwww
> 87: DP2 TEMP[3].x, TEMP[2].yxxx, CONST[11].xyyy
> 88: ADD TEMP[3].x, TEMP[3].xxxx, CONST[11].wwww
> 89: MOV TEMP[0].y, TEMP[3].xxxx
> 90: MOV TEMP[3].xy, TEMP[0].xyyy
> 91: TEX TEMP[3].xyz, TEMP[3], SAMP[6], 2D
> 92: ABS TEMP[5].x, TEMP[5].yyyy
> 93: POW TEMP[5].x, TEMP[5].xxxx, IMM[9].wwww
> 94: LRP TEMP[0].xyz, TEMP[5].xxxx, TEMP[3].xyzz, TEMP[4].xyzz
> 95: MAD TEMP[1], CONST[3].wwww, IMM[11].xxyz, IMM[12]
> 96: ADD TEMP[2].xy, -TEMP[1].xzzz, TEMP[1].zwww
> 97: RCP TEMP[2].x, TEMP[2].xxxx
> 98: RCP TEMP[3].x, TEMP[2].yyyy
> 99: MOV TEMP[2].y, TEMP[3].xxxx
>100: MAD TEMP[3].x, CONST[3].wwww, IMM[10].wwww, IMM[0].xxxx
>101: MOV TEMP[5].xy, IN[1].xyyy
>102: TEX TEMP[5].y, TEMP[5], SAMP[1], 2D
>103: MOV TEMP[6].xy, IN[0].xyyy
>104: TEX TEMP[6], TEMP[6], SAMP[0], 2D
>105: MAD TEMP[5].x, TEMP[5].yyyy, TEMP[6].xxxx, TEMP[6].wwww
>106: MAD TEMP[7].xy, TEMP[5].xxxx, TEMP[3].xxxx, -TEMP[1].xzzz
>107: MAD TEMP[3].x, TEMP[5].xxxx, TEMP[3].xxxx, IMM[11].wwww
>108: RCP TEMP[5].x, TEMP[1].yyyy
>109: MUL TEMP[5].x, TEMP[3].xxxx, TEMP[5].xxxx
>110: MOV_SAT TEMP[5].x, TEMP[5].xxxx
>111: MUL TEMP[3].x, TEMP[3].xxxx, IMM[13].xxxx
>112: MOV_SAT TEMP[3].x, TEMP[3].xxxx
>113: MUL TEMP[2].xy, TEMP[2].xyyy, TEMP[7].xyyy
>114: MOV_SAT TEMP[7].xy, TEMP[2].xyyy
>115: MAD TEMP[1].x, TEMP[7].xxxx, IMM[13].yyyy, IMM[13].zzzz
>116: MUL TEMP[2].x, TEMP[7].xxxx, TEMP[7].xxxx
>117: MUL TEMP[8].x, TEMP[2].xxxx, TEMP[1].xxxx
>118: MOV TEMP[1].y, TEMP[8].xxxx
>119: MAD TEMP[2].x, TEMP[7].yyyy, IMM[13].yyyy, IMM[13].zzzz
>120: MUL TEMP[7].x, TEMP[7].yyyy, TEMP[7].yyyy
>121: MUL TEMP[7].x, TEMP[7].xxxx, TEMP[2].xxxx
>122: MOV TEMP[1].z, TEMP[7].xxxx
>123: MAD TEMP[2].x, TEMP[5].xxxx, IMM[13].yyyy, IMM[13].zzzz
>124: MUL TEMP[5].x, TEMP[5].xxxx, TEMP[5].xxxx
>125: MUL TEMP[1].x, TEMP[5].xxxx, TEMP[2].xxxx
>126: MUL TEMP[0].xyz, TEMP[0].xyzz, TEMP[1].xyzz
>127: ADD TEMP[1].xyz, -CONST[0].xyzz, CONST[1].xyzz
>128: MAD TEMP[1].xyz, TEMP[0].xxxx, TEMP[1].xyzz, CONST[0].xyzz
>129: LRP TEMP[4].xyz, TEMP[0].yyyy, CONST[2].xyzz, TEMP[1].xyzz
>130: ADD TEMP[1].x, -TEMP[4].xxxx, CONST[0].wwww
>131: ADD TEMP[5].x, -TEMP[4].yyyy, CONST[1].wwww
>132: MOV TEMP[1].y, TEMP[5].xxxx
>133: ADD TEMP[5].x, -TEMP[4].zzzz, CONST[2].wwww
>134: MOV TEMP[1].z, TEMP[5].xxxx
>135: MAD TEMP[0].xyz, TEMP[0].zzzz, TEMP[1].xyzz, TEMP[4].xyzz
>136: MUL TEMP[2].x, CONST[3].wwww, IMM[13].wwww
>137: ADD TEMP[5].x, -TEMP[6].xxxx, IMM[0].xxxx
>138: MUL TEMP[5].x, TEMP[5].xxxx, TEMP[5].xxxx
>139: MUL TEMP[5].x, TEMP[5].xxxx, TEMP[5].xxxx
>140: MAD TEMP[2].x, TEMP[5].xxxx, IMM[14].xxxx, TEMP[2].xxxx
>141: MOV TEMP[5].xy, IN[1].zwww
>142: TEX TEMP[5], TEMP[5], SAMP[3], 2D
>143: ADD TEMP[5], TEMP[5], IMM[0].yyyy
>144: MAD TEMP[4], TEMP[2].xxxx, TEMP[5], IMM[0].xxxx
>145: MUL TEMP[5].xyz, TEMP[0].xyzz, TEMP[4].xyzz
>146: DP3 TEMP[7].x, TEMP[5].xyzz, IMM[14].yzww
>147: ADD TEMP[0].x, TEMP[7].xxxx, IMM[15].xxxx
>148: MUL TEMP[0].x, TEMP[0].xxxx, IMM[15].yyyy
>149: MOV_SAT TEMP[7].x, TEMP[0].xxxx
>150: MAD TEMP[8].x, TEMP[7].xxxx, IMM[13].yyyy, IMM[13].zzzz
>151: MUL TEMP[0].x, TEMP[7].xxxx, TEMP[7].xxxx
>152: MAD TEMP[0].x, TEMP[8].xxxx, -TEMP[0].xxxx, IMM[0].xxxx
>153: MAD TEMP[0].xyz, TEMP[0].xxxx, IMM[15].zzzz, TEMP[5].xyzz
>154: MUL TEMP[0].xyz, TEMP[0].xyzz, TEMP[6].zzzz
>155: MAD TEMP[5].xyz, TEMP[0].xyzz, IMM[15].wwww, TEMP[5].xyzz
>156: MOV_SAT TEMP[5].xyz, TEMP[5].xyzz
>157: MUL TEMP[0].xyz, TEMP[6].yyyy, TEMP[5].xyzz
>158: MOV TEMP[7].xy, IN[0].xyyy
>159: TEX TEMP[7], TEMP[7], SAMP[2], 2D
>160: MAD TEMP[5].xyz, TEMP[5].xyzz, -TEMP[6].yyyy, TEMP[7].xyzz
>161: MUL TEMP[6].x, TEMP[6].yyyy, CONST[3].zzzz
>162: MUL TEMP[4].x, TEMP[4].wwww, TEMP[6].xxxx
>163: MAD TEMP[1].x, TEMP[3].xxxx, IMM[13].yyyy, IMM[13].zzzz
>164: MUL TEMP[3].x, TEMP[3].xxxx, TEMP[3].xxxx
>165: MUL TEMP[6].x, TEMP[3].xxxx, TEMP[1].xxxx
>166: MAD TEMP[1].xy, TEMP[1].xxxx, TEMP[3].xxxx, IMM[16].xyyy
>167: MUL TEMP[1].xy, TEMP[1].xyyy, IMM[16].zwww
>168: MAD TEMP[0].xyz, TEMP[6].xxxx, TEMP[5].xyzz, TEMP[0].xyzz
>169: MAX TEMP[3].x, TEMP[1].yyyy, IMM[5].xxxx
>170: MOV_SAT TEMP[1].x, TEMP[1].xxxx
>171: MAD TEMP[5].x, TEMP[3].xxxx, IMM[13].yyyy, IMM[13].zzzz
>172: MUL TEMP[2].x, TEMP[3].xxxx, TEMP[3].xxxx
>173: MUL TEMP[2].x, TEMP[2].xxxx, TEMP[5].xxxx
>174: MAD TEMP[3].x, TEMP[4].xxxx, -TEMP[2].xxxx, TEMP[7].wwww
>175: MUL TEMP[2].x, TEMP[2].xxxx, TEMP[4].xxxx
>176: MAD TEMP[4].x, TEMP[1].xxxx, IMM[13].yyyy, IMM[13].zzzz
>177: MUL TEMP[1].x, TEMP[1].xxxx, TEMP[1].xxxx
>178: MUL TEMP[1].x, TEMP[1].xxxx, TEMP[4].xxxx
>179: MAD TEMP[1].x, TEMP[1].xxxx, TEMP[3].xxxx, TEMP[2].xxxx
>180: MOV TEMP[0].w, TEMP[1].xxxx
>181: MOV OUT[0], TEMP[0]
>182: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %24 = load <16 x i8>, <16 x i8> addrspace(2)* %23, align 16, !tbaa !0
>  %25 = call float @llvm.SI.load.const(<16 x i8> %24, i32 0)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %24, i32 4)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %24, i32 8)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %24, i32 12)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %24, i32 16)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %24, i32 20)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %24, i32 24)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %24, i32 28)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %24, i32 32)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %24, i32 36)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %24, i32 40)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %24, i32 44)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %24, i32 56)
>  %38 = call float @llvm.SI.load.const(<16 x i8> %24, i32 60)
>  %39 = call float @llvm.SI.load.const(<16 x i8> %24, i32 160)
>  %40 = call float @llvm.SI.load.const(<16 x i8> %24, i32 164)
>  %41 = call float @llvm.SI.load.const(<16 x i8> %24, i32 172)
>  %42 = call float @llvm.SI.load.const(<16 x i8> %24, i32 176)
>  %43 = call float @llvm.SI.load.const(<16 x i8> %24, i32 180)
>  %44 = call float @llvm.SI.load.const(<16 x i8> %24, i32 188)
>  %45 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %46 = load <8 x i32>, <8 x i32> addrspace(2)* %45, align 32, !tbaa !0
>  %47 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %48 = load <4 x i32>, <4 x i32> addrspace(2)* %47, align 16, !tbaa !0
>  %49 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 1
>  %50 = load <8 x i32>, <8 x i32> addrspace(2)* %49, align 32, !tbaa !0
>  %51 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 1
>  %52 = load <4 x i32>, <4 x i32> addrspace(2)* %51, align 16, !tbaa !0
>  %53 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 2
>  %54 = load <8 x i32>, <8 x i32> addrspace(2)* %53, align 32, !tbaa !0
>  %55 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 2
>  %56 = load <4 x i32>, <4 x i32> addrspace(2)* %55, align 16, !tbaa !0
>  %57 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 3
>  %58 = load <8 x i32>, <8 x i32> addrspace(2)* %57, align 32, !tbaa !0
>  %59 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 3
>  %60 = load <4 x i32>, <4 x i32> addrspace(2)* %59, align 16, !tbaa !0
>  %61 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 4
>  %62 = load <8 x i32>, <8 x i32> addrspace(2)* %61, align 32, !tbaa !0
>  %63 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 4
>  %64 = load <4 x i32>, <4 x i32> addrspace(2)* %63, align 16, !tbaa !0
>  %65 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 5
>  %66 = load <8 x i32>, <8 x i32> addrspace(2)* %65, align 32, !tbaa !0
>  %67 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 5
>  %68 = load <4 x i32>, <4 x i32> addrspace(2)* %67, align 16, !tbaa !0
>  %69 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 6
>  %70 = load <8 x i32>, <8 x i32> addrspace(2)* %69, align 32, !tbaa !0
>  %71 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 6
>  %72 = load <4 x i32>, <4 x i32> addrspace(2)* %71, align 16, !tbaa !0
>  %73 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %8)
>  %74 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %8)
>  %75 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %6, <2 x i32> %8)
>  %76 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %6, <2 x i32> %8)
>  %77 = call float @llvm.SI.fs.interp(i32 2, i32 1, i32 %6, <2 x i32> %8)
>  %78 = call float @llvm.SI.fs.interp(i32 3, i32 1, i32 %6, <2 x i32> %8)
>  %79 = fadd float %73, 0xBF5191B840000000
>  %80 = fsub float 0x3FEFDF3720000000, %74
>  %81 = fadd float %73, 0x3F5FFFFAA0000000
>  %82 = fsub float 0x3FEFE44980000000, %74
>  %83 = bitcast float %81 to i32
>  %84 = bitcast float %82 to i32
>  %85 = insertelement <2 x i32> undef, i32 %83, i32 0
>  %86 = insertelement <2 x i32> %85, i32 %84, i32 1
>  %87 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %86, <8 x i32> %66, <4 x i32> %68, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %88 = extractelement <4 x float> %87, i32 0
>  %89 = extractelement <4 x float> %87, i32 1
>  %90 = extractelement <4 x float> %87, i32 2
>  %91 = extractelement <4 x float> %87, i32 3
>  %92 = fmul float %88, 0x3FAE1E1E20000000
>  %93 = fmul float %89, 0x3FAE1E1E20000000
>  %94 = fmul float %90, 0x3FAE1E1E20000000
>  %95 = fmul float %91, 0x3FAE1E1E20000000
>  %96 = bitcast float %79 to i32
>  %97 = bitcast float %80 to i32
>  %98 = insertelement <2 x i32> undef, i32 %96, i32 0
>  %99 = insertelement <2 x i32> %98, i32 %97, i32 1
>  %100 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %99, <8 x i32> %66, <4 x i32> %68, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %101 = extractelement <4 x float> %100, i32 0
>  %102 = extractelement <4 x float> %100, i32 1
>  %103 = extractelement <4 x float> %100, i32 2
>  %104 = extractelement <4 x float> %100, i32 3
>  %105 = fmul float %101, 0x3FAE1E1E20000000
>  %106 = fadd float %105, %92
>  %107 = fmul float %102, 0x3FAE1E1E20000000
>  %108 = fadd float %107, %93
>  %109 = fmul float %103, 0x3FAE1E1E20000000
>  %110 = fadd float %109, %94
>  %111 = fmul float %104, 0x3FAE1E1E20000000
>  %112 = fadd float %111, %95
>  %113 = fadd float %73, 0x3F70646EC0000000
>  %114 = fsub float 0x3FEFF73720000000, %74
>  %115 = fadd float %73, 0xBF476CFB80000000
>  %116 = fsub float 0x3FEFEA24C0000000, %74
>  %117 = bitcast float %113 to i32
>  %118 = bitcast float %114 to i32
>  %119 = insertelement <2 x i32> undef, i32 %117, i32 0
>  %120 = insertelement <2 x i32> %119, i32 %118, i32 1
>  %121 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %120, <8 x i32> %66, <4 x i32> %68, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %122 = extractelement <4 x float> %121, i32 0
>  %123 = extractelement <4 x float> %121, i32 1
>  %124 = extractelement <4 x float> %121, i32 2
>  %125 = extractelement <4 x float> %121, i32 3
>  %126 = fmul float %122, 0x3FAE1E1E20000000
>  %127 = fadd float %126, %106
>  %128 = fmul float %123, 0x3FAE1E1E20000000
>  %129 = fadd float %128, %108
>  %130 = fmul float %124, 0x3FAE1E1E20000000
>  %131 = fadd float %130, %110
>  %132 = fmul float %125, 0x3FAE1E1E20000000
>  %133 = fadd float %132, %112
>  %134 = bitcast float %115 to i32
>  %135 = bitcast float %116 to i32
>  %136 = insertelement <2 x i32> undef, i32 %134, i32 0
>  %137 = insertelement <2 x i32> %136, i32 %135, i32 1
>  %138 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %137, <8 x i32> %66, <4 x i32> %68, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %139 = extractelement <4 x float> %138, i32 0
>  %140 = extractelement <4 x float> %138, i32 1
>  %141 = extractelement <4 x float> %138, i32 2
>  %142 = extractelement <4 x float> %138, i32 3
>  %143 = fmul float %139, 0x3FAE1E1E20000000
>  %144 = fadd float %143, %127
>  %145 = fmul float %140, 0x3FAE1E1E20000000
>  %146 = fadd float %145, %129
>  %147 = fmul float %141, 0x3FAE1E1E20000000
>  %148 = fadd float %147, %131
>  %149 = fmul float %142, 0x3FAE1E1E20000000
>  %150 = fadd float %149, %133
>  %151 = fadd float %73, 0x3F500002A0000000
>  %152 = fsub float 0x3FEFF224C0000000, %74
>  %153 = fadd float %73, 0x3F65DB3E60000000
>  %154 = fsub float 0x3FEFFA24C0000000, %74
>  %155 = bitcast float %151 to i32
>  %156 = bitcast float %152 to i32
>  %157 = insertelement <2 x i32> undef, i32 %155, i32 0
>  %158 = insertelement <2 x i32> %157, i32 %156, i32 1
>  %159 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %158, <8 x i32> %66, <4 x i32> %68, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %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 = fmul float %160, 0x3FAE1E1E20000000
>  %165 = fadd float %164, %144
>  %166 = fmul float %161, 0x3FAE1E1E20000000
>  %167 = fadd float %166, %146
>  %168 = fmul float %162, 0x3FAE1E1E20000000
>  %169 = fadd float %168, %148
>  %170 = fmul float %163, 0x3FAE1E1E20000000
>  %171 = fadd float %170, %150
>  %172 = bitcast float %153 to i32
>  %173 = bitcast float %154 to i32
>  %174 = insertelement <2 x i32> undef, i32 %172, i32 0
>  %175 = insertelement <2 x i32> %174, i32 %173, i32 1
>  %176 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %175, <8 x i32> %66, <4 x i32> %68, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %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 = extractelement <4 x float> %176, i32 3
>  %181 = fmul float %177, 0x3FAE1E1E20000000
>  %182 = fadd float %181, %165
>  %183 = fmul float %178, 0x3FAE1E1E20000000
>  %184 = fadd float %183, %167
>  %185 = fmul float %179, 0x3FAE1E1E20000000
>  %186 = fadd float %185, %169
>  %187 = fmul float %180, 0x3FAE1E1E20000000
>  %188 = fadd float %187, %171
>  %189 = fadd float %73, 0xBF6BB67A00000000
>  %190 = fsub float 0x3FEFF00000000000, %74
>  %191 = fadd float %73, 0xBF5BB67F60000000
>  %192 = fsub float 0x3FEFF80000000000, %74
>  %193 = bitcast float %189 to i32
>  %194 = bitcast float %190 to i32
>  %195 = insertelement <2 x i32> undef, i32 %193, i32 0
>  %196 = insertelement <2 x i32> %195, i32 %194, i32 1
>  %197 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %196, <8 x i32> %66, <4 x i32> %68, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %198 = extractelement <4 x float> %197, i32 0
>  %199 = extractelement <4 x float> %197, i32 1
>  %200 = extractelement <4 x float> %197, i32 2
>  %201 = extractelement <4 x float> %197, i32 3
>  %202 = fmul float %198, 0x3FAE1E1E20000000
>  %203 = fadd float %202, %182
>  %204 = fmul float %199, 0x3FAE1E1E20000000
>  %205 = fadd float %204, %184
>  %206 = fmul float %200, 0x3FAE1E1E20000000
>  %207 = fadd float %206, %186
>  %208 = fmul float %201, 0x3FAE1E1E20000000
>  %209 = fadd float %208, %188
>  %210 = bitcast float %191 to i32
>  %211 = bitcast float %192 to i32
>  %212 = insertelement <2 x i32> undef, i32 %210, i32 0
>  %213 = insertelement <2 x i32> %212, i32 %211, i32 1
>  %214 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %213, <8 x i32> %66, <4 x i32> %68, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %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 %215, 0x3FAE1E1E20000000
>  %220 = fadd float %219, %203
>  %221 = fmul float %216, 0x3FAE1E1E20000000
>  %222 = fadd float %221, %205
>  %223 = fmul float %217, 0x3FAE1E1E20000000
>  %224 = fadd float %223, %207
>  %225 = fmul float %218, 0x3FAE1E1E20000000
>  %226 = fadd float %225, %209
>  %227 = fadd float %73, 0.000000e+00
>  %228 = fsub float 1.000000e+00, %74
>  %229 = fadd float %73, 0x3F5BB67F60000000
>  %230 = fsub float 0x3FF0040000000000, %74
>  %231 = bitcast float %227 to i32
>  %232 = bitcast float %228 to i32
>  %233 = insertelement <2 x i32> undef, i32 %231, i32 0
>  %234 = insertelement <2 x i32> %233, i32 %232, i32 1
>  %235 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %234, <8 x i32> %66, <4 x i32> %68, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %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 = extractelement <4 x float> %235, i32 3
>  %240 = fmul float %236, 0x3FAE1E1E20000000
>  %241 = fadd float %240, %220
>  %242 = fmul float %237, 0x3FAE1E1E20000000
>  %243 = fadd float %242, %222
>  %244 = fmul float %238, 0x3FAE1E1E20000000
>  %245 = fadd float %244, %224
>  %246 = fmul float %239, 0x3FAE1E1E20000000
>  %247 = fadd float %246, %226
>  %248 = bitcast float %229 to i32
>  %249 = bitcast float %230 to i32
>  %250 = insertelement <2 x i32> undef, i32 %248, i32 0
>  %251 = insertelement <2 x i32> %250, i32 %249, i32 1
>  %252 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %251, <8 x i32> %66, <4 x i32> %68, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %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 = fmul float %253, 0x3FAE1E1E20000000
>  %258 = fadd float %257, %241
>  %259 = fmul float %254, 0x3FAE1E1E20000000
>  %260 = fadd float %259, %243
>  %261 = fmul float %255, 0x3FAE1E1E20000000
>  %262 = fadd float %261, %245
>  %263 = fmul float %256, 0x3FAE1E1E20000000
>  %264 = fadd float %263, %247
>  %265 = fadd float %73, 0x3F6BB67A00000000
>  %266 = fsub float 0x3FF0080000000000, %74
>  %267 = fadd float %73, 0xBF65DB3E60000000
>  %268 = fsub float 0x3FF002EDA0000000, %74
>  %269 = bitcast float %265 to i32
>  %270 = bitcast float %266 to i32
>  %271 = insertelement <2 x i32> undef, i32 %269, i32 0
>  %272 = insertelement <2 x i32> %271, i32 %270, i32 1
>  %273 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %272, <8 x i32> %66, <4 x i32> %68, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %274 = extractelement <4 x float> %273, i32 0
>  %275 = extractelement <4 x float> %273, i32 1
>  %276 = extractelement <4 x float> %273, i32 2
>  %277 = extractelement <4 x float> %273, i32 3
>  %278 = fmul float %274, 0x3FAE1E1E20000000
>  %279 = fadd float %278, %258
>  %280 = fmul float %275, 0x3FAE1E1E20000000
>  %281 = fadd float %280, %260
>  %282 = fmul float %276, 0x3FAE1E1E20000000
>  %283 = fadd float %282, %262
>  %284 = fmul float %277, 0x3FAE1E1E20000000
>  %285 = fadd float %284, %264
>  %286 = bitcast float %267 to i32
>  %287 = bitcast float %268 to i32
>  %288 = insertelement <2 x i32> undef, i32 %286, i32 0
>  %289 = insertelement <2 x i32> %288, i32 %287, i32 1
>  %290 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %289, <8 x i32> %66, <4 x i32> %68, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %291 = extractelement <4 x float> %290, i32 0
>  %292 = extractelement <4 x float> %290, i32 1
>  %293 = extractelement <4 x float> %290, i32 2
>  %294 = extractelement <4 x float> %290, i32 3
>  %295 = fmul float %291, 0x3FAE1E1E20000000
>  %296 = fadd float %295, %279
>  %297 = fmul float %292, 0x3FAE1E1E20000000
>  %298 = fadd float %297, %281
>  %299 = fmul float %293, 0x3FAE1E1E20000000
>  %300 = fadd float %299, %283
>  %301 = fmul float %294, 0x3FAE1E1E20000000
>  %302 = fadd float %301, %285
>  %303 = fadd float %73, 0xBF500002A0000000
>  %304 = fsub float 0x3FF006EDA0000000, %74
>  %305 = fadd float %73, 0x3F476CFB80000000
>  %306 = fsub float 0x3FF00AEDA0000000, %74
>  %307 = bitcast float %303 to i32
>  %308 = bitcast float %304 to i32
>  %309 = insertelement <2 x i32> undef, i32 %307, i32 0
>  %310 = insertelement <2 x i32> %309, i32 %308, i32 1
>  %311 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %310, <8 x i32> %66, <4 x i32> %68, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %312 = extractelement <4 x float> %311, i32 0
>  %313 = extractelement <4 x float> %311, i32 1
>  %314 = extractelement <4 x float> %311, i32 2
>  %315 = extractelement <4 x float> %311, i32 3
>  %316 = fmul float %312, 0x3FAE1E1E20000000
>  %317 = fadd float %316, %296
>  %318 = fmul float %313, 0x3FAE1E1E20000000
>  %319 = fadd float %318, %298
>  %320 = fmul float %314, 0x3FAE1E1E20000000
>  %321 = fadd float %320, %300
>  %322 = fmul float %315, 0x3FAE1E1E20000000
>  %323 = fadd float %322, %302
>  %324 = bitcast float %305 to i32
>  %325 = bitcast float %306 to i32
>  %326 = insertelement <2 x i32> undef, i32 %324, i32 0
>  %327 = insertelement <2 x i32> %326, i32 %325, i32 1
>  %328 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %327, <8 x i32> %66, <4 x i32> %68, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %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 = extractelement <4 x float> %328, i32 3
>  %333 = fmul float %329, 0x3FAE1E1E20000000
>  %334 = fadd float %333, %317
>  %335 = fmul float %330, 0x3FAE1E1E20000000
>  %336 = fadd float %335, %319
>  %337 = fmul float %331, 0x3FAE1E1E20000000
>  %338 = fadd float %337, %321
>  %339 = fmul float %332, 0x3FAE1E1E20000000
>  %340 = fadd float %339, %323
>  %341 = fadd float %73, 0xBF70646EC0000000
>  %342 = fsub float 0x3FF0046460000000, %74
>  %343 = fadd float %73, 0xBF5FFFFAA0000000
>  %344 = fsub float 0x3FF00DDB40000000, %74
>  %345 = bitcast float %341 to i32
>  %346 = bitcast float %342 to i32
>  %347 = insertelement <2 x i32> undef, i32 %345, i32 0
>  %348 = insertelement <2 x i32> %347, i32 %346, i32 1
>  %349 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %348, <8 x i32> %66, <4 x i32> %68, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %350 = extractelement <4 x float> %349, i32 0
>  %351 = extractelement <4 x float> %349, i32 1
>  %352 = extractelement <4 x float> %349, i32 2
>  %353 = extractelement <4 x float> %349, i32 3
>  %354 = fmul float %350, 0x3FAE1E1E20000000
>  %355 = fadd float %354, %334
>  %356 = fmul float %351, 0x3FAE1E1E20000000
>  %357 = fadd float %356, %336
>  %358 = fmul float %352, 0x3FAE1E1E20000000
>  %359 = fadd float %358, %338
>  %360 = fmul float %353, 0x3FAE1E1E20000000
>  %361 = fadd float %360, %340
>  %362 = bitcast float %343 to i32
>  %363 = bitcast float %344 to i32
>  %364 = insertelement <2 x i32> undef, i32 %362, i32 0
>  %365 = insertelement <2 x i32> %364, i32 %363, i32 1
>  %366 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %365, <8 x i32> %66, <4 x i32> %68, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %367 = extractelement <4 x float> %366, i32 0
>  %368 = extractelement <4 x float> %366, i32 1
>  %369 = extractelement <4 x float> %366, i32 2
>  %370 = extractelement <4 x float> %366, i32 3
>  %371 = fmul float %367, 0x3FAE1E1E20000000
>  %372 = fadd float %371, %355
>  %373 = fmul float %368, 0x3FAE1E1E20000000
>  %374 = fadd float %373, %357
>  %375 = fmul float %369, 0x3FAE1E1E20000000
>  %376 = fadd float %375, %359
>  %377 = fmul float %370, 0x3FAE1E1E20000000
>  %378 = fadd float %377, %361
>  %379 = fadd float %73, 0x3F5191B840000000
>  %380 = fsub float 0x3FF0106460000000, %74
>  %381 = bitcast float %379 to i32
>  %382 = bitcast float %380 to i32
>  %383 = insertelement <2 x i32> undef, i32 %381, i32 0
>  %384 = insertelement <2 x i32> %383, i32 %382, i32 1
>  %385 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %384, <8 x i32> %66, <4 x i32> %68, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %386 = extractelement <4 x float> %385, i32 0
>  %387 = extractelement <4 x float> %385, i32 1
>  %388 = extractelement <4 x float> %385, i32 2
>  %389 = extractelement <4 x float> %385, i32 3
>  %390 = fmul float %386, 0x3FAE1E1E20000000
>  %391 = fadd float %390, %372
>  %392 = fmul float %387, 0x3FAE1E1E20000000
>  %393 = fadd float %392, %374
>  %394 = fmul float %388, 0x3FAE1E1E20000000
>  %395 = fadd float %394, %376
>  %396 = fmul float %389, 0x3FAE1E1E20000000
>  %397 = fadd float %396, %378
>  %398 = fmul float %397, 1.600000e+01
>  %399 = fmul float %398, %391
>  %400 = fmul float %398, %393
>  %401 = fmul float %398, %395
>  %402 = fmul float %400, %39
>  %403 = fmul float %401, %40
>  %404 = fadd float %402, %403
>  %405 = fadd float %404, %41
>  %406 = fmul float %400, %42
>  %407 = fmul float %401, %43
>  %408 = fadd float %406, %407
>  %409 = fadd float %408, %44
>  %410 = bitcast float %405 to i32
>  %411 = bitcast float %409 to i32
>  %412 = insertelement <2 x i32> undef, i32 %410, i32 0
>  %413 = insertelement <2 x i32> %412, i32 %411, i32 1
>  %414 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %413, <8 x i32> %70, <4 x i32> %72, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %415 = extractelement <4 x float> %414, i32 0
>  %416 = extractelement <4 x float> %414, i32 1
>  %417 = extractelement <4 x float> %414, i32 2
>  %418 = fmul float %399, %39
>  %419 = fmul float %401, %40
>  %420 = fadd float %418, %419
>  %421 = fadd float %420, %41
>  %422 = fmul float %399, %42
>  %423 = fmul float %401, %43
>  %424 = fadd float %422, %423
>  %425 = fadd float %424, %44
>  %426 = bitcast float %73 to i32
>  %427 = bitcast float %74 to i32
>  %428 = insertelement <2 x i32> undef, i32 %426, i32 0
>  %429 = insertelement <2 x i32> %428, i32 %427, i32 1
>  %430 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %429, <8 x i32> %62, <4 x i32> %64, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %431 = extractelement <4 x float> %430, i32 0
>  %432 = extractelement <4 x float> %430, i32 1
>  %433 = extractelement <4 x float> %430, i32 2
>  %434 = fmul float %431, 2.000000e+00
>  %435 = fadd float %434, -1.000000e+00
>  %436 = fmul float %432, 2.000000e+00
>  %437 = fadd float %436, -1.000000e+00
>  %438 = fmul float %433, 2.000000e+00
>  %439 = fadd float %438, -1.000000e+00
>  %440 = fmul float %435, %435
>  %441 = fmul float %437, %437
>  %442 = fadd float %441, %440
>  %443 = fmul float %439, %439
>  %444 = fadd float %442, %443
>  %445 = call float @llvm.AMDGPU.rsq.clamped.f32(float %444)
>  %446 = fmul float %445, %437
>  %447 = fmul float %445, %439
>  %448 = call float @llvm.fabs.f32(float %446)
>  %449 = call float @llvm.pow.f32(float %448, float 7.000000e+00)
>  %450 = bitcast float %421 to i32
>  %451 = bitcast float %425 to i32
>  %452 = insertelement <2 x i32> undef, i32 %450, i32 0
>  %453 = insertelement <2 x i32> %452, i32 %451, i32 1
>  %454 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %453, <8 x i32> %70, <4 x i32> %72, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %455 = extractelement <4 x float> %454, i32 0
>  %456 = extractelement <4 x float> %454, i32 1
>  %457 = extractelement <4 x float> %454, i32 2
>  %458 = fsub float 1.000000e+00, %449
>  %459 = fmul float %455, %449
>  %460 = fmul float %415, %458
>  %461 = fadd float %459, %460
>  %462 = fsub float 1.000000e+00, %449
>  %463 = fmul float %456, %449
>  %464 = fmul float %416, %462
>  %465 = fadd float %463, %464
>  %466 = fsub float 1.000000e+00, %449
>  %467 = fmul float %457, %449
>  %468 = fmul float %417, %466
>  %469 = fadd float %467, %468
>  %470 = fmul float %400, %39
>  %471 = fmul float %399, %40
>  %472 = fadd float %470, %471
>  %473 = fadd float %472, %41
>  %474 = fmul float %400, %42
>  %475 = fmul float %399, %43
>  %476 = fadd float %474, %475
>  %477 = fadd float %476, %44
>  %478 = bitcast float %473 to i32
>  %479 = bitcast float %477 to i32
>  %480 = insertelement <2 x i32> undef, i32 %478, i32 0
>  %481 = insertelement <2 x i32> %480, i32 %479, i32 1
>  %482 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %481, <8 x i32> %70, <4 x i32> %72, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %483 = extractelement <4 x float> %482, i32 0
>  %484 = extractelement <4 x float> %482, i32 1
>  %485 = extractelement <4 x float> %482, i32 2
>  %486 = call float @llvm.fabs.f32(float %447)
>  %487 = call float @llvm.pow.f32(float %486, float 7.000000e+00)
>  %488 = fsub float 1.000000e+00, %487
>  %489 = fmul float %483, %487
>  %490 = fmul float %461, %488
>  %491 = fadd float %489, %490
>  %492 = fsub float 1.000000e+00, %487
>  %493 = fmul float %484, %487
>  %494 = fmul float %465, %492
>  %495 = fadd float %493, %494
>  %496 = fsub float 1.000000e+00, %487
>  %497 = fmul float %485, %487
>  %498 = fmul float %469, %496
>  %499 = fadd float %497, %498
>  %500 = fmul float %38, 0xBFAEB851E0000000
>  %501 = fadd float %500, 0x3FE1EB8520000000
>  %502 = fmul float %38, 0xBFAEB851E0000000
>  %503 = fadd float %502, 0xBF947AE140000000
>  %504 = fmul float %38, 0xBFBEB851E0000000
>  %505 = fadd float %504, 0x3FE147AE20000000
>  %506 = fmul float %38, 0xBFC70A3D80000000
>  %507 = fadd float %506, 0x3FE0A3D700000000
>  %508 = fsub float %505, %501
>  %509 = fsub float %507, %505
>  %510 = fdiv float 1.000000e+00, %508
>  %511 = fdiv float 1.000000e+00, %509
>  %512 = fmul float %38, 6.000000e+00
>  %513 = fadd float %512, 1.000000e+00
>  %514 = bitcast float %75 to i32
>  %515 = bitcast float %76 to i32
>  %516 = insertelement <2 x i32> undef, i32 %514, i32 0
>  %517 = insertelement <2 x i32> %516, i32 %515, i32 1
>  %518 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %517, <8 x i32> %50, <4 x i32> %52, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %519 = extractelement <4 x float> %518, i32 1
>  %520 = bitcast float %73 to i32
>  %521 = bitcast float %74 to i32
>  %522 = insertelement <2 x i32> undef, i32 %520, i32 0
>  %523 = insertelement <2 x i32> %522, i32 %521, i32 1
>  %524 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %523, <8 x i32> %46, <4 x i32> %48, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %525 = extractelement <4 x float> %524, i32 0
>  %526 = extractelement <4 x float> %524, i32 1
>  %527 = extractelement <4 x float> %524, i32 2
>  %528 = extractelement <4 x float> %524, i32 3
>  %529 = fmul float %519, %525
>  %530 = fadd float %529, %528
>  %531 = fmul float %530, %513
>  %532 = fsub float %531, %501
>  %533 = fmul float %530, %513
>  %534 = fsub float %533, %505
>  %535 = fmul float %530, %513
>  %536 = fadd float %535, 0xBFE28F5C20000000
>  %537 = fdiv float 1.000000e+00, %503
>  %538 = fmul float %536, %537
>  %539 = call float @llvm.AMDIL.clamp.(float %538, float 0.000000e+00, float 1.000000e+00)
>  %540 = fmul float %536, 1.000000e+01
>  %541 = call float @llvm.AMDIL.clamp.(float %540, float 0.000000e+00, float 1.000000e+00)
>  %542 = fmul float %510, %532
>  %543 = fmul float %511, %534
>  %544 = call float @llvm.AMDIL.clamp.(float %542, float 0.000000e+00, float 1.000000e+00)
>  %545 = call float @llvm.AMDIL.clamp.(float %543, float 0.000000e+00, float 1.000000e+00)
>  %546 = fmul float %544, -2.000000e+00
>  %547 = fadd float %546, 3.000000e+00
>  %548 = fmul float %544, %544
>  %549 = fmul float %548, %547
>  %550 = fmul float %545, -2.000000e+00
>  %551 = fadd float %550, 3.000000e+00
>  %552 = fmul float %545, %545
>  %553 = fmul float %552, %551
>  %554 = fmul float %539, -2.000000e+00
>  %555 = fadd float %554, 3.000000e+00
>  %556 = fmul float %539, %539
>  %557 = fmul float %556, %555
>  %558 = fmul float %491, %557
>  %559 = fmul float %495, %549
>  %560 = fmul float %499, %553
>  %561 = fsub float %29, %25
>  %562 = fsub float %30, %26
>  %563 = fsub float %31, %27
>  %564 = fmul float %558, %561
>  %565 = fadd float %564, %25
>  %566 = fmul float %558, %562
>  %567 = fadd float %566, %26
>  %568 = fmul float %558, %563
>  %569 = fadd float %568, %27
>  %570 = fsub float 1.000000e+00, %559
>  %571 = fmul float %33, %559
>  %572 = fmul float %565, %570
>  %573 = fadd float %571, %572
>  %574 = fsub float 1.000000e+00, %559
>  %575 = fmul float %34, %559
>  %576 = fmul float %567, %574
>  %577 = fadd float %575, %576
>  %578 = fsub float 1.000000e+00, %559
>  %579 = fmul float %35, %559
>  %580 = fmul float %569, %578
>  %581 = fadd float %579, %580
>  %582 = fsub float %28, %573
>  %583 = fsub float %32, %577
>  %584 = fsub float %36, %581
>  %585 = fmul float %560, %582
>  %586 = fadd float %585, %573
>  %587 = fmul float %560, %583
>  %588 = fadd float %587, %577
>  %589 = fmul float %560, %584
>  %590 = fadd float %589, %581
>  %591 = fmul float %38, 7.500000e-01
>  %592 = fsub float 1.000000e+00, %525
>  %593 = fmul float %592, %592
>  %594 = fmul float %593, %593
>  %595 = fmul float %594, 2.500000e-01
>  %596 = fadd float %595, %591
>  %597 = bitcast float %77 to i32
>  %598 = bitcast float %78 to i32
>  %599 = insertelement <2 x i32> undef, i32 %597, i32 0
>  %600 = insertelement <2 x i32> %599, i32 %598, i32 1
>  %601 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %600, <8 x i32> %58, <4 x i32> %60, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %602 = extractelement <4 x float> %601, i32 0
>  %603 = extractelement <4 x float> %601, i32 1
>  %604 = extractelement <4 x float> %601, i32 2
>  %605 = extractelement <4 x float> %601, i32 3
>  %606 = fadd float %602, -1.000000e+00
>  %607 = fadd float %603, -1.000000e+00
>  %608 = fadd float %604, -1.000000e+00
>  %609 = fadd float %605, -1.000000e+00
>  %610 = fmul float %596, %606
>  %611 = fadd float %610, 1.000000e+00
>  %612 = fmul float %596, %607
>  %613 = fadd float %612, 1.000000e+00
>  %614 = fmul float %596, %608
>  %615 = fadd float %614, 1.000000e+00
>  %616 = fmul float %596, %609
>  %617 = fadd float %616, 1.000000e+00
>  %618 = fmul float %586, %611
>  %619 = fmul float %588, %613
>  %620 = fmul float %590, %615
>  %621 = fmul float %618, 0x3FD3333340000000
>  %622 = fmul float %619, 0x3FE2E147A0000000
>  %623 = fadd float %622, %621
>  %624 = fmul float %620, 0x3FBC28F5C0000000
>  %625 = fadd float %623, %624
>  %626 = fadd float %625, 0xBFB47AE140000000
>  %627 = fmul float %626, 0x402C924920000000
>  %628 = call float @llvm.AMDIL.clamp.(float %627, float 0.000000e+00, float 1.000000e+00)
>  %629 = fmul float %628, -2.000000e+00
>  %630 = fadd float %629, 3.000000e+00
>  %631 = fmul float %628, %628
>  %632 = fmul float %631, %630
>  %633 = fsub float 1.000000e+00, %632
>  %634 = fmul float %633, 0x3F9EB851E0000000
>  %635 = fadd float %634, %618
>  %636 = fmul float %633, 0x3F9EB851E0000000
>  %637 = fadd float %636, %619
>  %638 = fmul float %633, 0x3F9EB851E0000000
>  %639 = fadd float %638, %620
>  %640 = fmul float %635, %527
>  %641 = fmul float %637, %527
>  %642 = fmul float %639, %527
>  %643 = fmul float %640, 5.000000e-01
>  %644 = fadd float %643, %618
>  %645 = fmul float %641, 5.000000e-01
>  %646 = fadd float %645, %619
>  %647 = fmul float %642, 5.000000e-01
>  %648 = fadd float %647, %620
>  %649 = call float @llvm.AMDIL.clamp.(float %644, float 0.000000e+00, float 1.000000e+00)
>  %650 = call float @llvm.AMDIL.clamp.(float %646, float 0.000000e+00, float 1.000000e+00)
>  %651 = call float @llvm.AMDIL.clamp.(float %648, float 0.000000e+00, float 1.000000e+00)
>  %652 = fmul float %526, %649
>  %653 = fmul float %526, %650
>  %654 = fmul float %526, %651
>  %655 = bitcast float %73 to i32
>  %656 = bitcast float %74 to i32
>  %657 = insertelement <2 x i32> undef, i32 %655, i32 0
>  %658 = insertelement <2 x i32> %657, i32 %656, i32 1
>  %659 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %658, <8 x i32> %54, <4 x i32> %56, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %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 %526, %649
>  %665 = fsub float %660, %664
>  %666 = fmul float %526, %650
>  %667 = fsub float %661, %666
>  %668 = fmul float %526, %651
>  %669 = fsub float %662, %668
>  %670 = fmul float %526, %37
>  %671 = fmul float %617, %670
>  %672 = fmul float %541, -2.000000e+00
>  %673 = fadd float %672, 3.000000e+00
>  %674 = fmul float %541, %541
>  %675 = fmul float %674, %673
>  %676 = fmul float %673, %674
>  %677 = fadd float %676, 0xBFECCCCCC0000000
>  %678 = fmul float %673, %674
>  %679 = fadd float %678, 0xBF847AE140000000
>  %680 = fmul float %677, 1.000000e+01
>  %681 = fmul float %679, -1.000000e+02
>  %682 = fmul float %675, %665
>  %683 = fadd float %682, %652
>  %684 = fmul float %675, %667
>  %685 = fadd float %684, %653
>  %686 = fmul float %675, %669
>  %687 = fadd float %686, %654
>  %688 = call float @llvm.maxnum.f32(float %681, float 0.000000e+00)
>  %689 = call float @llvm.AMDIL.clamp.(float %680, float 0.000000e+00, float 1.000000e+00)
>  %690 = fmul float %688, -2.000000e+00
>  %691 = fadd float %690, 3.000000e+00
>  %692 = fmul float %688, %688
>  %693 = fmul float %692, %691
>  %694 = fmul float %693, %671
>  %695 = fsub float %663, %694
>  %696 = fmul float %693, %671
>  %697 = fmul float %689, -2.000000e+00
>  %698 = fadd float %697, 3.000000e+00
>  %699 = fmul float %689, %689
>  %700 = fmul float %699, %698
>  %701 = fmul float %700, %695
>  %702 = fadd float %701, %696
>  %703 = call i32 @llvm.SI.packf16(float %683, float %685)
>  %704 = bitcast i32 %703 to float
>  %705 = call i32 @llvm.SI.packf16(float %687, float %702)
>  %706 = bitcast i32 %705 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %704, float %706, float %704, float %706)
>  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.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.AMDGPU.rsq.clamped.f32(float) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.fabs.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: nounwind readnone
>declare float @llvm.maxnum.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_mov_b32 m0, s10                                                            ; BEFC030A
>	v_mov_b32_e32 v10, 0xba8c8dc2                                                ; 7E1402FF BA8C8DC2
>	v_mov_b32_e32 v11, 0x3d70f0f1                                                ; 7E1602FF 3D70F0F1
>	v_mov_b32_e32 v12, 0xba3b67dc                                                ; 7E1802FF BA3B67DC
>	v_mov_b32_e32 v13, 0xbb5db3d0                                                ; 7E1A02FF BB5DB3D0
>	v_mov_b32_e32 v14, 0xbaddb3fb                                                ; 7E1C02FF BADDB3FB
>	v_mov_b32_e32 v15, 0xbb2ed9f3                                                ; 7E1E02FF BB2ED9F3
>	v_mov_b32_e32 v16, 0xba800015                                                ; 7E2002FF BA800015
>	v_mov_b32_e32 v17, 0xbb832376                                                ; 7E2202FF BB832376
>	v_mov_b32_e32 v18, 0xbaffffd5                                                ; 7E2402FF BAFFFFD5
>	v_mov_b32_e32 v7, 0x7fffffff                                                 ; 7E0E02FF 7FFFFFFF
>	v_mov_b32_e32 v6, 0x40e00000                                                 ; 7E0C02FF 40E00000
>	v_mov_b32_e32 v4, 0x3f0f5c29                                                 ; 7E0802FF 3F0F5C29
>	v_mov_b32_e32 v5, 0xbd75c28f                                                 ; 7E0A02FF BD75C28F
>	s_load_dwordx4 s[12:15], s[2:3], 0x0                                         ; C0860300
>	s_load_dwordx4 s[16:19], s[4:5], 0x0                                         ; C0880500
>	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[20:23], s[4:5], 0x4                                         ; C08A0504
>	v_interp_p1_f32 v3, v0, 1, 0, [m0]                                           ; C80C0100
>	s_load_dwordx4 s[0:3], s[4:5], 0x8                                           ; C0800508
>	v_interp_p2_f32 v3, [v3], v1, 1, 0, [m0]                                     ; C80D0101
>	v_interp_p1_f32 v19, v0, 0, 1, [m0]                                          ; C84C0400
>	v_interp_p2_f32 v19, [v19], v1, 0, 1, [m0]                                   ; C84D0401
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	s_buffer_load_dword s8, s[12:15], 0x0                                        ; C2040D00
>	v_interp_p1_f32 v20, v0, 1, 1, [m0]                                          ; C8500500
>	v_interp_p2_f32 v20, [v20], v1, 1, 1, [m0]                                   ; C8510501
>	v_interp_p1_f32 v8, v0, 2, 1, [m0]                                           ; C8200600
>	s_buffer_load_dword s11, s[12:15], 0x29                                      ; C2058D29
>	s_buffer_load_dword s33, s[12:15], 0x2b                                      ; C2108D2B
>	s_buffer_load_dword s10, s[12:15], 0x2c                                      ; C2050D2C
>	s_buffer_load_dword s9, s[12:15], 0x2d                                       ; C2048D2D
>	s_buffer_load_dword s32, s[12:15], 0x2f                                      ; C2100D2F
>	s_load_dwordx8 s[24:31], s[6:7], 0x0                                         ; C0CC0700
>	v_interp_p2_f32 v8, [v8], v1, 2, 1, [m0]                                     ; C8210601
>	v_interp_p1_f32 v9, v0, 3, 1, [m0]                                           ; C8240700
>	v_interp_p2_f32 v9, [v9], v1, 3, 1, [m0]                                     ; C8250701
>	v_add_f32_e32 v0, v2, v10                                                    ; 06001502
>	v_sub_f32_e32 v1, 0x3f7ef9b9, v3                                             ; 080206FF 3F7EF9B9
>	v_add_f32_e32 v21, 0x3affffd5, v2                                            ; 062A04FF 3AFFFFD5
>	v_sub_f32_e32 v22, 0x3f7f224c, v3                                            ; 082C06FF 3F7F224C
>	v_add_f32_e32 v23, 0x3b832376, v2                                            ; 062E04FF 3B832376
>	v_sub_f32_e32 v24, 0x3f7fb9b9, v3                                            ; 083006FF 3F7FB9B9
>	v_add_f32_e32 v25, v2, v12                                                   ; 06321902
>	v_sub_f32_e32 v26, 0x3f7f5126, v3                                            ; 083406FF 3F7F5126
>	v_add_f32_e32 v27, 0x3a800015, v2                                            ; 063604FF 3A800015
>	v_sub_f32_e32 v28, 0x3f7f9126, v3                                            ; 083806FF 3F7F9126
>	v_add_f32_e32 v29, 0x3b2ed9f3, v2                                            ; 063A04FF 3B2ED9F3
>	v_sub_f32_e32 v30, 0x3f7fd126, v3                                            ; 083C06FF 3F7FD126
>	v_add_f32_e32 v12, v2, v13                                                   ; 06181B02
>	v_sub_f32_e32 v13, 0x3f7f8000, v3                                            ; 081A06FF 3F7F8000
>	v_add_f32_e32 v31, v2, v14                                                   ; 063E1D02
>	v_sub_f32_e32 v32, 0x3f7fc000, v3                                            ; 084006FF 3F7FC000
>	v_add_f32_e32 v33, 0, v2                                                     ; 06420480
>	v_sub_f32_e32 v34, 1.0, v3                                                   ; 084406F2
>	v_add_f32_e32 v35, 0x3addb3fb, v2                                            ; 064604FF 3ADDB3FB
>	v_sub_f32_e32 v36, 0x3f802000, v3                                            ; 084806FF 3F802000
>	v_add_f32_e32 v37, 0x3b5db3d0, v2                                            ; 064A04FF 3B5DB3D0
>	v_sub_f32_e32 v38, 0x3f804000, v3                                            ; 084C06FF 3F804000
>	v_add_f32_e32 v14, v2, v15                                                   ; 061C1F02
>	v_sub_f32_e32 v15, 0x3f80176d, v3                                            ; 081E06FF 3F80176D
>	v_add_f32_e32 v39, v2, v16                                                   ; 064E2102
>	v_sub_f32_e32 v40, 0x3f80376d, v3                                            ; 085006FF 3F80376D
>	v_add_f32_e32 v41, 0x3a3b67dc, v2                                            ; 065204FF 3A3B67DC
>	s_load_dwordx8 s[48:55], s[6:7], 0x28                                        ; C0D80728
>	s_load_dwordx4 s[56:59], s[4:5], 0x14                                        ; C09C0514
>	v_sub_f32_e32 v42, 0x3f80576d, v3                                            ; 085406FF 3F80576D
>	v_add_f32_e32 v16, v2, v17                                                   ; 06202302
>	v_sub_f32_e32 v17, 0x3f802323, v3                                            ; 082206FF 3F802323
>	v_add_f32_e32 v43, v2, v18                                                   ; 06562502
>	v_sub_f32_e32 v44, 0x3f806eda, v3                                            ; 085806FF 3F806EDA
>	v_add_f32_e32 v45, 0x3a8c8dc2, v2                                            ; 065A04FF 3A8C8DC2
>	v_sub_f32_e32 v46, 0x3f808323, v3                                            ; 085C06FF 3F808323
>	s_load_dwordx4 s[36:39], s[4:5], 0x18                                        ; C0920518
>	s_load_dwordx8 s[40:47], s[6:7], 0x30                                        ; C0D40730
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample v[47:50], 15, 0, 0, 0, 0, 0, 0, 0, v[21:22], s[48:55], s[56:59] ; F0800F00 01CC2F15
>	image_sample v[51:54], 15, 0, 0, 0, 0, 0, 0, 0, v[0:1], s[48:55], s[56:59]   ; F0800F00 01CC3300
>	image_sample v[21:24], 15, 0, 0, 0, 0, 0, 0, 0, v[23:24], s[48:55], s[56:59] ; F0800F00 01CC1517
>	image_sample v[55:58], 15, 0, 0, 0, 0, 0, 0, 0, v[25:26], s[48:55], s[56:59] ; F0800F00 01CC3719
>	image_sample v[25:28], 15, 0, 0, 0, 0, 0, 0, 0, v[27:28], s[48:55], s[56:59] ; F0800F00 01CC191B
>	image_sample v[59:62], 15, 0, 0, 0, 0, 0, 0, 0, v[29:30], s[48:55], s[56:59] ; F0800F00 01CC3B1D
>	image_sample v[63:66], 15, 0, 0, 0, 0, 0, 0, 0, v[12:13], s[48:55], s[56:59] ; F0800F00 01CC3F0C
>	image_sample v[29:32], 15, 0, 0, 0, 0, 0, 0, 0, v[31:32], s[48:55], s[56:59] ; F0800F00 01CC1D1F
>	image_sample v[67:70], 15, 0, 0, 0, 0, 0, 0, 0, v[33:34], s[48:55], s[56:59] ; F0800F00 01CC4321
>	image_sample v[33:36], 15, 0, 0, 0, 0, 0, 0, 0, v[35:36], s[48:55], s[56:59] ; F0800F00 01CC2123
>	image_sample v[71:74], 15, 0, 0, 0, 0, 0, 0, 0, v[37:38], s[48:55], s[56:59] ; F0800F00 01CC4725
>	image_sample v[12:15], 15, 0, 0, 0, 0, 0, 0, 0, v[14:15], s[48:55], s[56:59] ; F0800F00 01CC0C0E
>	image_sample v[37:40], 15, 0, 0, 0, 0, 0, 0, 0, v[39:40], s[48:55], s[56:59] ; F0800F00 01CC2527
>	image_sample v[75:78], 15, 0, 0, 0, 0, 0, 0, 0, v[41:42], s[48:55], s[56:59] ; F0800F00 01CC4B29
>	image_sample v[79:82], 15, 0, 0, 0, 0, 0, 0, 0, v[16:17], s[48:55], s[56:59] ; F0800F00 01CC4F10
>	image_sample v[41:44], 15, 0, 0, 0, 0, 0, 0, 0, v[43:44], s[48:55], s[56:59] ; F0800F00 01CC292B
>	image_sample v[83:86], 15, 0, 0, 0, 0, 0, 0, 0, v[45:46], s[48:55], s[56:59] ; F0800F00 01CC532D
>	v_mov_b32_e32 v0, s33                                                        ; 7E000221
>	s_waitcnt                                                                    ; BF8C077F
>	v_mul_f32_e32 v1, v11, v49                                                   ; 1002630B
>	v_mac_f32_e32 v1, v11, v53                                                   ; 3E026B0B
>	v_mul_f32_e32 v10, v11, v50                                                  ; 1014650B
>	v_mac_f32_e32 v10, v11, v54                                                  ; 3E146D0B
>	s_waitcnt vmcnt(14)                                                          ; BF8C077E
>	v_mac_f32_e32 v1, v11, v23                                                   ; 3E022F0B
>	v_mac_f32_e32 v10, v11, v24                                                  ; 3E14310B
>	s_waitcnt vmcnt(13)                                                          ; BF8C077D
>	v_mac_f32_e32 v1, v11, v57                                                   ; 3E02730B
>	v_mac_f32_e32 v10, v11, v58                                                  ; 3E14750B
>	s_waitcnt vmcnt(12)                                                          ; BF8C077C
>	v_mac_f32_e32 v1, v11, v27                                                   ; 3E02370B
>	v_mac_f32_e32 v10, v11, v28                                                  ; 3E14390B
>	s_waitcnt vmcnt(11)                                                          ; BF8C077B
>	v_mac_f32_e32 v1, v11, v61                                                   ; 3E027B0B
>	v_mac_f32_e32 v10, v11, v62                                                  ; 3E147D0B
>	s_waitcnt vmcnt(10)                                                          ; BF8C077A
>	v_mac_f32_e32 v1, v11, v65                                                   ; 3E02830B
>	v_mac_f32_e32 v10, v11, v66                                                  ; 3E14850B
>	s_waitcnt vmcnt(9)                                                           ; BF8C0779
>	v_mac_f32_e32 v1, v11, v31                                                   ; 3E023F0B
>	v_mac_f32_e32 v10, v11, v32                                                  ; 3E14410B
>	s_waitcnt vmcnt(8)                                                           ; BF8C0778
>	v_mac_f32_e32 v1, v11, v69                                                   ; 3E028B0B
>	v_mac_f32_e32 v10, v11, v70                                                  ; 3E148D0B
>	s_waitcnt vmcnt(7)                                                           ; BF8C0777
>	v_mac_f32_e32 v1, v11, v35                                                   ; 3E02470B
>	v_mac_f32_e32 v10, v11, v36                                                  ; 3E14490B
>	s_waitcnt vmcnt(6)                                                           ; BF8C0776
>	v_mac_f32_e32 v1, v11, v73                                                   ; 3E02930B
>	v_mac_f32_e32 v10, v11, v74                                                  ; 3E14950B
>	s_waitcnt vmcnt(5)                                                           ; BF8C0775
>	v_mac_f32_e32 v1, v11, v14                                                   ; 3E021D0B
>	v_mac_f32_e32 v10, v11, v15                                                  ; 3E141F0B
>	s_waitcnt vmcnt(4)                                                           ; BF8C0774
>	v_mac_f32_e32 v1, v11, v39                                                   ; 3E024F0B
>	v_mac_f32_e32 v10, v11, v40                                                  ; 3E14510B
>	s_waitcnt vmcnt(3)                                                           ; BF8C0773
>	v_mac_f32_e32 v1, v11, v77                                                   ; 3E029B0B
>	v_mac_f32_e32 v10, v11, v78                                                  ; 3E149D0B
>	s_waitcnt vmcnt(2)                                                           ; BF8C0772
>	v_mac_f32_e32 v1, v11, v81                                                   ; 3E02A30B
>	v_mac_f32_e32 v10, v11, v82                                                  ; 3E14A50B
>	s_waitcnt vmcnt(1)                                                           ; BF8C0771
>	v_mac_f32_e32 v1, v11, v43                                                   ; 3E02570B
>	v_mac_f32_e32 v10, v11, v44                                                  ; 3E14590B
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mac_f32_e32 v1, v11, v85                                                   ; 3E02AB0B
>	v_mac_f32_e32 v10, v11, v86                                                  ; 3E14AD0B
>	v_mul_f32_e32 v10, 0x41800000, v10                                           ; 101414FF 41800000
>	v_mul_f32_e32 v14, v1, v10                                                   ; 101C1501
>	v_mad_f32 v15, s11, v14, v0                                                  ; D282000F 04021C0B
>	v_mov_b32_e32 v1, s32                                                        ; 7E020220
>	v_mul_f32_e32 v16, v11, v48                                                  ; 1020610B
>	v_mac_f32_e32 v16, v11, v52                                                  ; 3E20690B
>	v_mac_f32_e32 v16, v11, v22                                                  ; 3E202D0B
>	v_mac_f32_e32 v16, v11, v56                                                  ; 3E20710B
>	v_mac_f32_e32 v16, v11, v26                                                  ; 3E20350B
>	v_mac_f32_e32 v16, v11, v60                                                  ; 3E20790B
>	v_mac_f32_e32 v16, v11, v64                                                  ; 3E20810B
>	v_mac_f32_e32 v16, v11, v30                                                  ; 3E203D0B
>	v_mac_f32_e32 v16, v11, v68                                                  ; 3E20890B
>	v_mac_f32_e32 v16, v11, v34                                                  ; 3E20450B
>	v_mac_f32_e32 v16, v11, v72                                                  ; 3E20910B
>	s_buffer_load_dword s32, s[12:15], 0x28                                      ; C2100D28
>	v_mac_f32_e32 v16, v11, v13                                                  ; 3E201B0B
>	v_mac_f32_e32 v16, v11, v38                                                  ; 3E204D0B
>	s_load_dwordx4 s[48:51], s[4:5], 0x10                                        ; C0980510
>	s_load_dwordx8 s[52:59], s[6:7], 0x20                                        ; C0DA0720
>	v_mac_f32_e32 v16, v11, v76                                                  ; 3E20990B
>	v_mac_f32_e32 v16, v11, v80                                                  ; 3E20A10B
>	v_mac_f32_e32 v16, v11, v42                                                  ; 3E20550B
>	v_mac_f32_e32 v16, v11, v84                                                  ; 3E20A90B
>	v_mul_f32_e32 v13, v16, v10                                                  ; 101A1510
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_mad_f32 v17, s32, v13, v15                                                 ; D2820011 043E1A20
>	v_mad_f32 v16, s9, v14, v1                                                   ; D2820010 04061C09
>	v_mad_f32 v18, s10, v13, v16                                                 ; D2820012 04421A0A
>	image_sample v[22:24], 7, 0, 0, 0, 0, 0, 0, 0, v[17:18], s[40:47], s[36:39]  ; F0800700 012A1611
>	image_sample v[26:28], 7, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[52:59], s[48:51]    ; F0800700 018D1A02
>	v_mul_f32_e32 v14, v11, v47                                                  ; 101C5F0B
>	v_mac_f32_e32 v14, v11, v51                                                  ; 3E1C670B
>	v_mac_f32_e32 v14, v11, v21                                                  ; 3E1C2B0B
>	v_mac_f32_e32 v14, v11, v55                                                  ; 3E1C6F0B
>	v_mac_f32_e32 v14, v11, v25                                                  ; 3E1C330B
>	v_mac_f32_e32 v14, v11, v59                                                  ; 3E1C770B
>	v_mac_f32_e32 v14, v11, v63                                                  ; 3E1C7F0B
>	v_mac_f32_e32 v14, v11, v29                                                  ; 3E1C3B0B
>	v_mac_f32_e32 v14, v11, v67                                                  ; 3E1C870B
>	v_mac_f32_e32 v14, v11, v33                                                  ; 3E1C430B
>	v_mac_f32_e32 v14, v11, v71                                                  ; 3E1C8F0B
>	v_mac_f32_e32 v14, v11, v12                                                  ; 3E1C190B
>	v_mac_f32_e32 v14, v11, v37                                                  ; 3E1C4B0B
>	v_mac_f32_e32 v14, v11, v75                                                  ; 3E1C970B
>	v_mac_f32_e32 v14, v11, v79                                                  ; 3E1C9F0B
>	v_mac_f32_e32 v14, v11, v41                                                  ; 3E1C530B
>	v_mac_f32_e32 v14, v11, v83                                                  ; 3E1CA70B
>	v_mul_f32_e32 v10, v14, v10                                                  ; 1014150E
>	v_mac_f32_e32 v15, s32, v10                                                  ; 3E1E1420
>	v_mac_f32_e32 v0, s32, v13                                                   ; 3E001A20
>	v_mac_f32_e32 v0, s11, v10                                                   ; 3E00140B
>	v_mac_f32_e32 v16, s10, v10                                                  ; 3E20140A
>	v_mac_f32_e32 v1, s10, v13                                                   ; 3E021A0A
>	v_mac_f32_e32 v1, s9, v10                                                    ; 3E021409
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_add_f32_e32 v10, v26, v26                                                  ; 0614351A
>	v_mad_f32 v11, 2.0, v26, -1.0                                                ; D282000B 03CE34F4
>	v_mad_f32 v10, v10, v11, -v11                                                ; D282000A 842E170A
>	v_mad_f32 v11, 2.0, v27, -1.0                                                ; D282000B 03CE36F4
>	v_mac_f32_e32 v10, v11, v11                                                  ; 3E14170B
>	v_mad_f32 v11, 2.0, v28, -1.0                                                ; D282000B 03CE38F4
>	v_mac_f32_e32 v10, v11, v11                                                  ; 3E14170B
>	v_rsq_clamp_f32_e32 v10, v10                                                 ; 7E14590A
>	s_load_dwordx8 s[52:59], s[6:7], 0x8                                         ; C0DA0708
>	image_sample v[11:13], 7, 0, 0, 0, 0, 0, 0, 0, v[15:16], s[40:47], s[36:39]  ; F0800700 012A0B0F
>	image_sample v[14:16], 7, 0, 0, 0, 0, 0, 0, 0, v[0:1], s[40:47], s[36:39]    ; F0800700 012A0E00
>	v_add_f32_e32 v0, v27, v27                                                   ; 0600371B
>	v_mad_f32 v0, v0, v10, -v10                                                  ; D2820000 842A1500
>	v_and_b32_e32 v0, v0, v7                                                     ; 36000F00
>	v_log_f32_e32 v0, v0                                                         ; 7E004F00
>	s_load_dwordx4 s[40:43], s[4:5], 0xc                                         ; C094050C
>	s_load_dwordx8 s[32:39], s[6:7], 0x10                                        ; C0D00710
>	s_load_dwordx8 s[44:51], s[6:7], 0x18                                        ; C0D60718
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	image_sample v1, 2, 0, 0, 0, 0, 0, 0, 0, v[19:20], s[52:59], s[20:23]        ; F0800200 00AD0113
>	image_sample v[17:20], 15, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[24:31], s[16:19]   ; F0800F00 00861102
>	v_mul_legacy_f32_e32 v0, v6, v0                                              ; 0E000106
>	v_exp_f32_e32 v0, v0                                                         ; 7E004B00
>	v_mad_f32 v21, -v0, v22, v22                                                 ; D2820015 245A2D00
>	v_mad_f32 v22, -v0, v23, v23                                                 ; D2820016 245E2F00
>	v_mad_f32 v23, -v0, v24, v24                                                 ; D2820017 24623100
>	s_waitcnt vmcnt(3)                                                           ; BF8C0773
>	v_mac_f32_e32 v21, v0, v11                                                   ; 3E2A1700
>	v_mac_f32_e32 v22, v0, v12                                                   ; 3E2C1900
>	v_mac_f32_e32 v23, v0, v13                                                   ; 3E2E1B00
>	v_add_f32_e32 v0, v28, v28                                                   ; 0600391C
>	image_sample v[24:27], 15, 0, 0, 0, 0, 0, 0, 0, v[8:9], s[44:51], s[40:43]   ; F0800F00 014B1808
>	s_buffer_load_dword s4, s[12:15], 0x1                                        ; C2020D01
>	s_buffer_load_dword s5, s[12:15], 0x2                                        ; C2028D02
>	s_buffer_load_dword s6, s[12:15], 0x3                                        ; C2030D03
>	s_buffer_load_dword s7, s[12:15], 0x4                                        ; C2038D04
>	s_buffer_load_dword s9, s[12:15], 0x5                                        ; C2048D05
>	s_buffer_load_dword s10, s[12:15], 0x6                                       ; C2050D06
>	s_buffer_load_dword s11, s[12:15], 0xa                                       ; C2058D0A
>	s_buffer_load_dword s16, s[12:15], 0xb                                       ; C2080D0B
>	s_buffer_load_dword s17, s[12:15], 0xe                                       ; C2088D0E
>	s_buffer_load_dword s18, s[12:15], 0xf                                       ; C2090D0F
>	s_buffer_load_dword s19, s[12:15], 0x7                                       ; C2098D07
>	s_buffer_load_dword s20, s[12:15], 0x8                                       ; C20A0D08
>	s_buffer_load_dword s12, s[12:15], 0x9                                       ; C2060D09
>	v_mad_f32 v0, v0, v10, -v10                                                  ; D2820000 842A1500
>	v_and_b32_e32 v0, v0, v7                                                     ; 36000F00
>	v_log_f32_e32 v0, v0                                                         ; 7E004F00
>	v_mov_b32_e32 v7, s8                                                         ; 7E0E0208
>	s_waitcnt lgkmcnt(0)                                                         ; BF8C007F
>	v_sub_f32_e32 v7, s7, v7                                                     ; 080E0E07
>	v_mov_b32_e32 v8, s4                                                         ; 7E100204
>	v_sub_f32_e32 v8, s9, v8                                                     ; 08101009
>	v_mov_b32_e32 v9, s5                                                         ; 7E120205
>	v_sub_f32_e32 v9, s10, v9                                                    ; 0812120A
>	v_mul_legacy_f32_e32 v0, v6, v0                                              ; 0E000106
>	v_exp_f32_e32 v0, v0                                                         ; 7E004B00
>	v_mad_f32 v6, -v0, v21, v21                                                  ; D2820006 24562B00
>	s_waitcnt vmcnt(3)                                                           ; BF8C0773
>	v_mac_f32_e32 v6, v0, v14                                                    ; 3E0C1D00
>	v_mad_f32 v10, -v0, v22, v22                                                 ; D282000A 245A2D00
>	v_mac_f32_e32 v10, v0, v15                                                   ; 3E141F00
>	v_mad_f32 v11, -v0, v23, v23                                                 ; D282000B 245E2F00
>	v_mac_f32_e32 v11, v0, v16                                                   ; 3E162100
>	v_mov_b32_e32 v0, 0xbca3d70a                                                 ; 7E0002FF BCA3D70A
>	v_mac_f32_e32 v0, s18, v5                                                    ; 3E000A12
>	v_mov_b32_e32 v12, 0x40c00000                                                ; 7E1802FF 40C00000
>	v_rcp_f32_e32 v0, v0                                                         ; 7E005500
>	v_mad_f32 v12, v12, s18, 1.0                                                 ; D282000C 03C8250C
>	s_waitcnt vmcnt(1)                                                           ; BF8C0771
>	v_mac_f32_e32 v20, v17, v1                                                   ; 3E280311
>	v_madak_f32_e32 v1, v20, v12, 0xbf147ae1                                     ; 42021914 BF147AE1
>	v_mul_f32_e32 v0, v0, v1                                                     ; 10000300
>	v_add_f32_e64 v0, 0, v0 clamp                                                ; D2060800 00020080
>	v_mov_b32_e32 v13, 0x40400000                                                ; 7E1A02FF 40400000
>	v_mad_f32 v14, -2.0, v0, v13                                                 ; D282000E 043600F5
>	v_mul_f32_e32 v0, v0, v0                                                     ; 10000100
>	v_mul_f32_e32 v0, v14, v0                                                    ; 1000010E
>	v_mul_f32_e32 v0, v0, v6                                                     ; 10000D00
>	v_mad_f32 v6, v0, v7, s8                                                     ; D2820006 00220F00
>	v_mac_f32_e32 v4, s18, v5                                                    ; 3E080A12
>	v_mov_b32_e32 v5, 0x3f0a3d71                                                 ; 7E0A02FF 3F0A3D71
>	v_mov_b32_e32 v7, 0xbdf5c28f                                                 ; 7E0E02FF BDF5C28F
>	v_mac_f32_e32 v5, s18, v7                                                    ; 3E0A0E12
>	v_subrev_f32_e32 v7, v4, v5                                                  ; 0A0E0B04
>	v_rcp_f32_e32 v7, v7                                                         ; 7E0E5507
>	v_mad_f32 v8, v0, v8, s4                                                     ; D2820008 00121100
>	v_mad_f32 v0, v0, v9, s5                                                     ; D2820000 00161300
>	v_mad_f32 v4, v20, v12, -v4                                                  ; D2820004 84121914
>	v_mul_f32_e32 v4, v4, v7                                                     ; 10080F04
>	v_add_f32_e64 v4, 0, v4 clamp                                                ; D2060804 00020880
>	v_mad_f32 v7, -2.0, v4, v13                                                  ; D2820007 043608F5
>	v_mul_f32_e32 v4, v4, v4                                                     ; 10080904
>	v_mul_f32_e32 v4, v7, v4                                                     ; 10080907
>	v_mad_f32 v7, -v10, v4, 1.0                                                  ; D2820007 23CA090A
>	v_mad_f32 v9, -v6, v7, s6                                                    ; D2820009 201A0F06
>	v_mul_f32_e32 v4, v4, v10                                                    ; 10081504
>	v_mad_f32 v6, -v4, v6, v6                                                    ; D2820006 241A0D04
>	v_mac_f32_e32 v6, s20, v4                                                    ; 3E0C0814
>	v_mad_f32 v9, -s20, v4, v9                                                   ; D2820009 24260814
>	v_mad_f32 v10, -v8, v7, s19                                                  ; D282000A 204E0F08
>	v_mad_f32 v8, -v4, v8, v8                                                    ; D2820008 24221104
>	v_mac_f32_e32 v8, s12, v4                                                    ; 3E10080C
>	v_mad_f32 v10, -s12, v4, v10                                                 ; D282000A 242A080C
>	v_mad_f32 v7, -v0, v7, s16                                                   ; D2820007 20420F00
>	v_mad_f32 v0, -v4, v0, v0                                                    ; D2820000 24020104
>	v_mac_f32_e32 v0, s11, v4                                                    ; 3E00080B
>	v_mad_f32 v4, -s11, v4, v7                                                   ; D2820004 241E080B
>	v_mov_b32_e32 v7, 0x3f051eb8                                                 ; 7E0E02FF 3F051EB8
>	v_mov_b32_e32 v14, 0xbe3851ec                                                ; 7E1C02FF BE3851EC
>	v_mac_f32_e32 v7, s18, v14                                                   ; 3E0E1C12
>	image_sample v[28:31], 15, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[32:39], s[0:3]     ; F0800F00 00081C02
>	v_subrev_f32_e32 v2, v5, v7                                                  ; 0A040F05
>	v_rcp_f32_e32 v2, v2                                                         ; 7E045502
>	v_mad_f32 v3, v20, v12, -v5                                                  ; D2820003 84161914
>	v_mov_b32_e32 v5, 0x3f400000                                                 ; 7E0A02FF 3F400000
>	v_mul_f32_e32 v5, s18, v5                                                    ; 100A0A12
>	v_mul_f32_e32 v2, v3, v2                                                     ; 10040503
>	v_add_f32_e64 v2, 0, v2 clamp                                                ; D2060802 00020480
>	v_mad_f32 v3, -2.0, v2, v13                                                  ; D2820003 043604F5
>	v_mul_f32_e32 v2, v2, v2                                                     ; 10040502
>	v_mul_f32_e32 v2, v3, v2                                                     ; 10040503
>	v_mul_f32_e32 v3, s17, v18                                                   ; 10062411
>	v_mul_f32_e32 v2, v2, v11                                                    ; 10041702
>	v_mac_f32_e32 v6, v9, v2                                                     ; 3E0C0509
>	v_mov_b32_e32 v7, 0x41200000                                                 ; 7E0E02FF 41200000
>	v_mul_f32_e32 v1, v7, v1                                                     ; 10020307
>	v_add_f32_e64 v1, 0, v1 clamp                                                ; D2060801 00020280
>	v_mac_f32_e32 v8, v10, v2                                                    ; 3E10050A
>	v_mac_f32_e32 v0, v4, v2                                                     ; 3E000504
>	v_sub_f32_e32 v2, 1.0, v17                                                   ; 080422F2
>	v_mad_f32 v2, -v17, v2, v2                                                   ; D2820002 240A0511
>	v_mul_f32_e32 v2, v2, v2                                                     ; 10040502
>	v_madmk_f32_e32 v2, v2, v5, 0x3e800000                                       ; 40040B02 3E800000
>	s_waitcnt vmcnt(1)                                                           ; BF8C0771
>	v_mad_f32 v4, v24, v2, -v2                                                   ; D2820004 840A0518
>	v_mad_f32 v5, v25, v2, -v2                                                   ; D2820005 840A0519
>	v_mad_f32 v9, v26, v2, -v2                                                   ; D2820009 840A051A
>	v_mad_f32 v2, v27, v2, -v2                                                   ; D2820002 840A051B
>	v_mac_f32_e32 v6, v6, v4                                                     ; 3E0C0906
>	v_mac_f32_e32 v8, v8, v5                                                     ; 3E100B08
>	v_mac_f32_e32 v0, v0, v9                                                     ; 3E001300
>	v_mul_f32_e32 v4, 0x3e99999a, v6                                             ; 10080CFF 3E99999A
>	v_madmk_f32_e32 v4, v8, v4, 0x3f170a3d                                       ; 40080908 3F170A3D
>	v_madmk_f32_e32 v4, v0, v4, 0x3de147ae                                       ; 40080900 3DE147AE
>	v_mov_b32_e32 v5, 0xbda3d70a                                                 ; 7E0A02FF BDA3D70A
>	v_add_f32_e32 v4, v4, v5                                                     ; 06080B04
>	v_mul_f32_e32 v4, 0x41649249, v4                                             ; 100808FF 41649249
>	v_add_f32_e64 v4, 0, v4 clamp                                                ; D2060804 00020880
>	v_mad_f32 v5, -2.0, v4, v13                                                  ; D2820005 043608F5
>	v_mul_f32_e32 v4, v4, v4                                                     ; 10080904
>	v_mad_f32 v4, -v4, v5, 1.0                                                   ; D2820004 23CA0B04
>	v_mov_b32_e32 v5, 0x3cf5c28f                                                 ; 7E0A02FF 3CF5C28F
>	v_mad_f32 v9, v5, v4, v6                                                     ; D2820009 041A0905
>	v_mad_f32 v10, v5, v4, v8                                                    ; D282000A 04220905
>	v_mad_f32 v4, v5, v4, v0                                                     ; D2820004 04020905
>	v_mul_f32_e32 v5, v19, v9                                                    ; 100A1313
>	v_mul_f32_e32 v9, v19, v10                                                   ; 10121513
>	v_mul_f32_e32 v4, v19, v4                                                    ; 10080913
>	v_mac_f32_e32 v6, 0.5, v5                                                    ; 3E0C0AF0
>	v_mac_f32_e32 v8, 0.5, v9                                                    ; 3E1012F0
>	v_mac_f32_e32 v0, 0.5, v4                                                    ; 3E0008F0
>	v_add_f32_e64 v4, 0, v6 clamp                                                ; D2060804 00020C80
>	v_add_f32_e64 v5, 0, v8 clamp                                                ; D2060805 00021080
>	v_add_f32_e64 v0, 0, v0 clamp                                                ; D2060800 00020080
>	v_mul_f32_e32 v6, v4, v18                                                    ; 100C2504
>	v_mul_f32_e32 v8, v5, v18                                                    ; 10102505
>	v_mul_f32_e32 v9, v0, v18                                                    ; 10122500
>	s_waitcnt vmcnt(0)                                                           ; BF8C0770
>	v_mad_f32 v4, -v18, v4, v28                                                  ; D2820004 24720912
>	v_mad_f32 v5, -v18, v5, v29                                                  ; D2820005 24760B12
>	v_mad_f32 v0, -v18, v0, v30                                                  ; D2820000 247A0112
>	v_mac_f32_e32 v3, v3, v2                                                     ; 3E060503
>	v_mad_f32 v2, -2.0, v1, v13                                                  ; D2820002 043602F5
>	v_mul_f32_e32 v1, v1, v1                                                     ; 10020301
>	v_mul_f32_e32 v10, v2, v1                                                    ; 10140302
>	v_madak_f32_e32 v11, v1, v2, 0xbf666666                                      ; 42160501 BF666666
>	v_madak_f32_e32 v1, v1, v2, 0xbc23d70a                                       ; 42020501 BC23D70A
>	v_mul_f32_e32 v2, v7, v11                                                    ; 10041707
>	v_mov_b32_e32 v7, 0xc2c80000                                                 ; 7E0E02FF C2C80000
>	v_mul_f32_e32 v1, v1, v7                                                     ; 10020F01
>	v_mac_f32_e32 v6, v4, v10                                                    ; 3E0C1504
>	v_mac_f32_e32 v8, v5, v10                                                    ; 3E101505
>	v_mac_f32_e32 v9, v0, v10                                                    ; 3E121500
>	v_max_f32_e32 v0, 0, v1                                                      ; 20000280
>	v_add_f32_e64 v1, 0, v2 clamp                                                ; D2060801 00020480
>	v_mad_f32 v2, -2.0, v0, v13                                                  ; D2820002 043600F5
>	v_mul_f32_e32 v0, v0, v0                                                     ; 10000100
>	v_mul_f32_e32 v0, v2, v0                                                     ; 10000102
>	v_mul_f32_e32 v2, v3, v0                                                     ; 10040103
>	v_mad_f32 v0, -v0, v3, v31                                                   ; D2820000 247E0700
>	v_mac_f32_e32 v13, -2.0, v1                                                  ; 3E1A02F5
>	v_mul_f32_e32 v1, v1, v1                                                     ; 10020301
>	v_mul_f32_e32 v1, v13, v1                                                    ; 1002030D
>	v_mac_f32_e32 v2, v0, v1                                                     ; 3E040300
>	v_cvt_pkrtz_f16_f32_e32 v0, v6, v8                                           ; 5E001106
>	v_cvt_pkrtz_f16_f32_e32 v1, v9, v2                                           ; 5E020509
>	exp 15, 0, 1, 1, 1, v0, v1, v0, v1                                           ; F8001C0F 01000100
>	s_endpgm                                                                     ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 64
>VGPRS: 88
>Code Size: 2180 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
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[0]
>DCL OUT[2], GENERIC[1]
>DCL OUT[3], GENERIC[2]
>DCL CONST[0..5]
>DCL TEMP[0..2], LOCAL
>IMM[0] FLT32 {    0.0000,     1.0000,     0.0000,     0.0000}
>  0: MOV TEMP[0].zw, IMM[0].yyxy
>  1: DP4 TEMP[0].x, IN[0], CONST[2]
>  2: DP4 TEMP[1].x, IN[0], CONST[3]
>  3: MOV TEMP[0].y, TEMP[1].xxxx
>  4: DP4 TEMP[1].x, IN[0], CONST[4]
>  5: DP4 TEMP[2].x, IN[0], CONST[5]
>  6: MOV TEMP[1].y, TEMP[2].xxxx
>  7: MOV TEMP[1].xy, TEMP[1].xyxx
>  8: MOV OUT[1], CONST[0]
>  9: MOV OUT[2], CONST[1]
> 10: MOV OUT[0], TEMP[0]
> 11: MOV OUT[3], TEMP[1]
> 12: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = call float @llvm.SI.load.const(<16 x i8> %13, i32 0)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %13, i32 4)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %13, i32 8)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %13, i32 12)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %13, i32 16)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %13, i32 20)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %13, i32 24)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %13, i32 28)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %13, i32 32)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %13, i32 36)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %13, i32 40)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %13, i32 44)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %13, i32 48)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %13, i32 52)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %13, i32 56)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %13, i32 60)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %13, i32 64)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %13, i32 68)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %13, i32 72)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %13, i32 76)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %13, i32 80)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %13, i32 84)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %13, i32 88)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %13, i32 92)
>  %38 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %39 = load <16 x i8>, <16 x i8> addrspace(2)* %38, align 16, !tbaa !0
>  %40 = add i32 %5, %8
>  %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 = extractelement <4 x float> %41, i32 2
>  %45 = extractelement <4 x float> %41, i32 3
>  %46 = fmul float %42, %22
>  %47 = fmul float %43, %23
>  %48 = fadd float %46, %47
>  %49 = fmul float %44, %24
>  %50 = fadd float %48, %49
>  %51 = fmul float %45, %25
>  %52 = fadd float %50, %51
>  %53 = fmul float %42, %26
>  %54 = fmul float %43, %27
>  %55 = fadd float %53, %54
>  %56 = fmul float %44, %28
>  %57 = fadd float %55, %56
>  %58 = fmul float %45, %29
>  %59 = fadd float %57, %58
>  %60 = fmul float %42, %30
>  %61 = fmul float %43, %31
>  %62 = fadd float %60, %61
>  %63 = fmul float %44, %32
>  %64 = fadd float %62, %63
>  %65 = fmul float %45, %33
>  %66 = fadd float %64, %65
>  %67 = fmul float %42, %34
>  %68 = fmul float %43, %35
>  %69 = fadd float %67, %68
>  %70 = fmul float %44, %36
>  %71 = fadd float %69, %70
>  %72 = fmul float %45, %37
>  %73 = fadd float %71, %72
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %14, float %15, float %16, float %17)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %18, float %19, float %20, float %21)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 34, i32 0, float %66, float %73, float undef, float undef)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %52, float %59, 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[4:7], s[8:9], 0x0                  ; C0820900
>	v_add_i32_e32 v0, vcc, s10, v0                      ; 4A00000A
>	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 lgkmcnt(0)                                ; BF8C007F
>	s_buffer_load_dword s4, s[0:3], 0x0                 ; C2020100
>	s_buffer_load_dword s5, s[0:3], 0x1                 ; C2028101
>	s_buffer_load_dword s6, s[0:3], 0x2                 ; C2030102
>	s_buffer_load_dword s7, s[0:3], 0x3                 ; C2038103
>	s_buffer_load_dword s8, s[0:3], 0x4                 ; C2040104
>	s_buffer_load_dword s9, s[0:3], 0x5                 ; C2048105
>	s_buffer_load_dword s10, s[0:3], 0x6                ; C2050106
>	s_buffer_load_dword s11, s[0:3], 0x7                ; C2058107
>	s_buffer_load_dword s12, s[0:3], 0x8                ; C2060108
>	s_buffer_load_dword s13, s[0:3], 0x9                ; C2068109
>	s_buffer_load_dword s14, s[0:3], 0xa                ; C207010A
>	s_buffer_load_dword s15, s[0:3], 0xb                ; C207810B
>	s_buffer_load_dword s16, s[0:3], 0xc                ; C208010C
>	s_buffer_load_dword s17, s[0:3], 0xd                ; C208810D
>	s_buffer_load_dword s18, s[0:3], 0xe                ; C209010E
>	s_buffer_load_dword s19, s[0:3], 0xf                ; C209810F
>	s_buffer_load_dword s20, s[0:3], 0x10               ; C20A0110
>	s_buffer_load_dword s21, s[0:3], 0x11               ; C20A8111
>	s_buffer_load_dword s22, s[0:3], 0x14               ; C20B0114
>	s_buffer_load_dword s23, s[0:3], 0x15               ; C20B8115
>	s_buffer_load_dword s24, s[0:3], 0x12               ; C20C0112
>	s_buffer_load_dword s25, s[0:3], 0x13               ; C20C8113
>	s_buffer_load_dword s26, s[0:3], 0x16               ; C20D0116
>	s_buffer_load_dword s0, s[0:3], 0x17                ; C2000117
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	v_mov_b32_e32 v4, s4                                ; 7E080204
>	v_mov_b32_e32 v5, s5                                ; 7E0A0205
>	v_mov_b32_e32 v6, s6                                ; 7E0C0206
>	v_mov_b32_e32 v7, s7                                ; 7E0E0207
>	exp 15, 32, 0, 0, 0, v4, v5, v6, v7                 ; F800020F 07060504
>	s_waitcnt expcnt(0)                                 ; BF8C070F
>	v_mov_b32_e32 v4, s8                                ; 7E080208
>	v_mov_b32_e32 v5, s9                                ; 7E0A0209
>	v_mov_b32_e32 v6, s10                               ; 7E0C020A
>	v_mov_b32_e32 v7, s11                               ; 7E0E020B
>	exp 15, 33, 0, 0, 0, v4, v5, v6, v7                 ; F800021F 07060504
>	s_waitcnt vmcnt(0) expcnt(0)                        ; BF8C0700
>	v_mul_f32_e32 v4, s13, v1                           ; 1008020D
>	v_mul_f32_e32 v5, s17, v1                           ; 100A0211
>	v_mul_f32_e32 v6, s21, v1                           ; 100C0215
>	v_mul_f32_e32 v1, s23, v1                           ; 10020217
>	v_mac_f32_e32 v4, s12, v0                           ; 3E08000C
>	v_mac_f32_e32 v5, s16, v0                           ; 3E0A0010
>	v_mac_f32_e32 v6, s20, v0                           ; 3E0C0014
>	v_mac_f32_e32 v1, s22, v0                           ; 3E020016
>	v_mac_f32_e32 v4, s14, v2                           ; 3E08040E
>	v_mac_f32_e32 v5, s18, v2                           ; 3E0A0412
>	v_mac_f32_e32 v6, s24, v2                           ; 3E0C0418
>	v_mac_f32_e32 v1, s26, v2                           ; 3E02041A
>	v_mac_f32_e32 v4, s15, v3                           ; 3E08060F
>	v_mac_f32_e32 v5, s19, v3                           ; 3E0A0613
>	v_mac_f32_e32 v6, s25, v3                           ; 3E0C0619
>	v_mac_f32_e32 v1, s0, v3                            ; 3E020600
>	v_mov_b32_e32 v0, 1.0                               ; 7E0002F2
>	exp 15, 34, 0, 0, 0, v6, v1, v0, v0                 ; F800022F 00000106
>	s_waitcnt expcnt(0)                                 ; BF8C070F
>	v_mov_b32_e32 v1, 0                                 ; 7E020280
>	exp 15, 12, 0, 1, 0, v4, v5, v1, v0                 ; F80008CF 00010504
>	s_endpgm                                            ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 32
>VGPRS: 8
>Code Size: 280 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
>  clamp_color = 0
>FRAG
>PROPERTY FS_COLOR0_WRITES_ALL_CBUFS 1
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL IN[1], GENERIC[1], PERSPECTIVE
>DCL IN[2], GENERIC[2], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SVIEW[0], 2D, FLOAT
>DCL CONST[0]
>DCL CONST[2]
>DCL TEMP[0..7], LOCAL
>IMM[0] FLT32 {    0.0000,     1.0000,     0.0000,     0.0000}
>  0: MOV TEMP[0].xy, IN[2].xyxx
>  1: MOV TEMP[1], IMM[0].xxxx
>  2: MOV TEMP[2], IMM[0].xxxx
>  3: MOV TEMP[3].x, -CONST[0].xxxx
>  4: BGNLOOP :0
>  5:   FSLT TEMP[4].x, CONST[0].xxxx, TEMP[3].xxxx
>  6:   UIF TEMP[4].xxxx :0
>  7:     BRK
>  8:   ENDIF
>  9:   MAD TEMP[5].xy, TEMP[3].xxxx, CONST[2].xyyy, TEMP[0].xyyy
> 10:   MOV TEMP[6].xy, TEMP[5].xyyy
> 11:   MOV TEMP[6].w, IMM[0].xxxx
> 12:   TXB TEMP[7], TEMP[6], SAMP[0], 2D
> 13:   ADD TEMP[2], TEMP[2], TEMP[7]
> 14:   ADD TEMP[3].x, TEMP[3].xxxx, IMM[0].yyyy
> 15: ENDLOOP :0
> 16: MUL TEMP[1], TEMP[2], CONST[0].wwww
> 17: MOV TEMP[0].w, IMM[0].yyyy
> 18: MOV TEMP[0].xyz, IN[1].xyzx
> 19: MUL TEMP[0], TEMP[1], TEMP[0]
> 20: MUL TEMP[1], TEMP[0], IN[1].wwww
> 21: MAD TEMP[1], IN[0], TEMP[1].wwww, TEMP[1]
> 22: MOV OUT[0], TEMP[1]
> 23: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %24 = load <16 x i8>, <16 x i8> addrspace(2)* %23, align 16, !tbaa !0
>  %25 = call float @llvm.SI.load.const(<16 x i8> %24, i32 0)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %24, i32 12)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %24, i32 32)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %24, i32 36)
>  %29 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %30 = load <8 x i32>, <8 x i32> addrspace(2)* %29, align 32, !tbaa !0
>  %31 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %32 = load <4 x i32>, <4 x i32> addrspace(2)* %31, align 16, !tbaa !0
>  %33 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %8)
>  %34 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %8)
>  %35 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %6, <2 x i32> %8)
>  %36 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %6, <2 x i32> %8)
>  %37 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %6, <2 x i32> %8)
>  %38 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %6, <2 x i32> %8)
>  %39 = call float @llvm.SI.fs.interp(i32 2, i32 1, i32 %6, <2 x i32> %8)
>  %40 = call float @llvm.SI.fs.interp(i32 3, i32 1, i32 %6, <2 x i32> %8)
>  %41 = call float @llvm.SI.fs.interp(i32 0, i32 2, i32 %6, <2 x i32> %8)
>  %42 = call float @llvm.SI.fs.interp(i32 1, i32 2, i32 %6, <2 x i32> %8)
>  %43 = fsub float -0.000000e+00, %25
>  br label %LOOP
>
>LOOP:                                             ; preds = %ENDIF, %main_body
>  %temp8.0 = phi float [ 0.000000e+00, %main_body ], [ %81, %ENDIF ]
>  %temp9.0 = phi float [ 0.000000e+00, %main_body ], [ %82, %ENDIF ]
>  %temp10.0 = phi float [ 0.000000e+00, %main_body ], [ %83, %ENDIF ]
>  %temp11.0 = phi float [ 0.000000e+00, %main_body ], [ %84, %ENDIF ]
>  %temp12.0 = phi float [ %43, %main_body ], [ %85, %ENDIF ]
>  %44 = fcmp olt float %25, %temp12.0
>  br i1 %44, label %IF, label %ENDIF
>
>IF:                                               ; preds = %LOOP
>  %45 = fmul float %temp8.0, %26
>  %46 = fmul float %temp9.0, %26
>  %47 = fmul float %temp10.0, %26
>  %48 = fmul float %temp11.0, %26
>  %49 = fmul float %45, %37
>  %50 = fmul float %46, %38
>  %51 = fmul float %47, %39
>  %52 = fmul float %49, %40
>  %53 = fmul float %50, %40
>  %54 = fmul float %51, %40
>  %55 = fmul float %48, %40
>  %56 = fmul float %33, %55
>  %57 = fadd float %56, %52
>  %58 = fmul float %34, %55
>  %59 = fadd float %58, %53
>  %60 = fmul float %35, %55
>  %61 = fadd float %60, %54
>  %62 = fmul float %36, %55
>  %63 = fadd float %62, %55
>  %64 = call i32 @llvm.SI.packf16(float %57, float %59)
>  %65 = bitcast i32 %64 to float
>  %66 = call i32 @llvm.SI.packf16(float %61, float %63)
>  %67 = bitcast i32 %66 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %65, float %67, float %65, float %67)
>  ret void
>
>ENDIF:                                            ; preds = %LOOP
>  %68 = fmul float %temp12.0, %27
>  %69 = fadd float %68, %41
>  %70 = fmul float %temp12.0, %28
>  %71 = fadd float %70, %42
>  %72 = bitcast float %69 to i32
>  %73 = bitcast float %71 to i32
>  %74 = insertelement <4 x i32> <i32 0, i32 undef, i32 undef, i32 undef>, i32 %72, i32 1
>  %75 = insertelement <4 x i32> %74, i32 %73, i32 2
>  %76 = call <4 x float> @llvm.SI.image.sample.b.v4i32(<4 x i32> %75, <8 x i32> %30, <4 x i32> %32, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %77 = extractelement <4 x float> %76, i32 0
>  %78 = extractelement <4 x float> %76, i32 1
>  %79 = extractelement <4 x float> %76, i32 2
>  %80 = extractelement <4 x float> %76, i32 3
>  %81 = fadd float %temp8.0, %77
>  %82 = fadd float %temp9.0, %78
>  %83 = fadd float %temp10.0, %79
>  %84 = fadd float %temp11.0, %80
>  %85 = fadd float %temp12.0, 1.000000e+00
>  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: nounwind readnone
>declare <4 x float> @llvm.SI.image.sample.b.v4i32(<4 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, 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 }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                        ; BEFE0A7E
>	s_load_dwordx4 s[16:19], s[2:3], 0x0                                        ; C0880300
>	s_mov_b32 m0, s10                                                           ; BEFC030A
>	s_load_dwordx8 s[8:15], s[6:7], 0x0                                         ; C0C40700
>	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
>	s_waitcnt lgkmcnt(0)                                                        ; BF8C007F
>	s_buffer_load_dword s1, s[16:19], 0x0                                       ; C2009100
>	s_buffer_load_dword s0, s[16:19], 0x3                                       ; C2001103
>	s_buffer_load_dword s2, s[16:19], 0x8                                       ; C2011108
>	s_buffer_load_dword s3, s[16:19], 0x9                                       ; C2019109
>	v_interp_p1_f32 v4, v0, 2, 0, [m0]                                          ; C8100200
>	v_interp_p2_f32 v4, [v4], v1, 2, 0, [m0]                                    ; C8110201
>	v_interp_p1_f32 v5, v0, 3, 0, [m0]                                          ; C8140300
>	v_interp_p2_f32 v5, [v5], v1, 3, 0, [m0]                                    ; C8150301
>	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 v7, v0, 1, 1, [m0]                                          ; C81C0500
>	v_interp_p2_f32 v7, [v7], v1, 1, 1, [m0]                                    ; C81D0501
>	v_interp_p1_f32 v8, v0, 2, 1, [m0]                                          ; C8200600
>	v_interp_p2_f32 v8, [v8], v1, 2, 1, [m0]                                    ; C8210601
>	v_interp_p1_f32 v9, v0, 3, 1, [m0]                                          ; C8240700
>	v_interp_p2_f32 v9, [v9], v1, 3, 1, [m0]                                    ; C8250701
>	v_interp_p1_f32 v10, v0, 0, 2, [m0]                                         ; C8280800
>	s_load_dwordx4 s[4:7], s[4:5], 0x0                                          ; C0820500
>	v_interp_p2_f32 v10, [v10], v1, 0, 2, [m0]                                  ; C8290801
>	v_interp_p1_f32 v0, v0, 1, 2, [m0]                                          ; C8000900
>	v_interp_p2_f32 v0, [v0], v1, 1, 2, [m0]                                    ; C8010901
>	v_mov_b32_e32 v1, 0x80000000                                                ; 7E0202FF 80000000
>	s_waitcnt lgkmcnt(0)                                                        ; BF8C007F
>	v_xor_b32_e32 v11, s1, v1                                                   ; 3A160201
>	v_mov_b32_e32 v15, 0                                                        ; 7E1E0280
>	s_mov_b64 s[16:17], 0                                                       ; BE900480
>	v_mov_b32_e32 v16, 0                                                        ; 7E200280
>	v_mov_b32_e32 v17, 0                                                        ; 7E220280
>	v_mov_b32_e32 v18, 0                                                        ; 7E240280
>	v_mov_b32_e32 v1, v18                                                       ; 7E020312
>	v_mov_b32_e32 v12, v17                                                      ; 7E180311
>	v_mov_b32_e32 v13, v16                                                      ; 7E1A0310
>	v_mov_b32_e32 v14, v15                                                      ; 7E1C030F
>	v_cmp_nlt_f32_e32 vcc, s1, v11                                              ; 7C1C1601
>	s_and_saveexec_b64 s[18:19], vcc                                            ; BE92246A
>	s_xor_b64 s[18:19], exec, s[18:19]                                          ; 8992127E
>	v_mad_f32 v16, s2, v11, v10                                                 ; D2820010 042A1602
>	v_mad_f32 v17, s3, v11, v0                                                  ; D2820011 04021603
>	v_mov_b32_e32 v15, 0                                                        ; 7E1E0280
>	image_sample_b v[18:21], 15, 0, 0, 0, 0, 0, 0, 0, v[15:18], s[8:15], s[4:7] ; F0940F00 0022120F
>	s_waitcnt vmcnt(0)                                                          ; BF8C0770
>	v_add_f32_e32 v15, v18, v14                                                 ; 061E1D12
>	v_add_f32_e32 v16, v19, v13                                                 ; 06201B13
>	v_add_f32_e32 v17, v20, v12                                                 ; 06221914
>	v_add_f32_e32 v18, v21, v1                                                  ; 06240315
>	v_add_f32_e32 v11, 1.0, v11                                                 ; 061616F2
>	s_or_b64 exec, exec, s[18:19]                                               ; 88FE127E
>	s_or_b64 s[16:17], s[18:19], s[16:17]                                       ; 88901012
>	s_andn2_b64 exec, exec, s[16:17]                                            ; 8AFE107E
>	s_cbranch_execnz BB0_1                                                      ; BF890000
>	s_or_b64 exec, exec, s[16:17]                                               ; 88FE107E
>	v_mul_f32_e32 v0, s0, v14                                                   ; 10001C00
>	v_mul_f32_e32 v10, s0, v13                                                  ; 10141A00
>	v_mul_f32_e32 v11, s0, v12                                                  ; 10161800
>	v_mul_f32_e32 v1, s0, v1                                                    ; 10020200
>	v_mul_f32_e32 v0, v6, v0                                                    ; 10000106
>	v_mul_f32_e32 v6, v7, v10                                                   ; 100C1507
>	v_mul_f32_e32 v7, v8, v11                                                   ; 100E1708
>	v_mul_f32_e32 v0, v9, v0                                                    ; 10000109
>	v_mul_f32_e32 v6, v9, v6                                                    ; 100C0D09
>	v_mul_f32_e32 v7, v9, v7                                                    ; 100E0F09
>	v_mul_f32_e32 v1, v9, v1                                                    ; 10020309
>	v_mac_f32_e32 v0, v1, v2                                                    ; 3E000501
>	v_mac_f32_e32 v6, v1, v3                                                    ; 3E0C0701
>	v_mac_f32_e32 v7, v1, v4                                                    ; 3E0E0901
>	v_mac_f32_e32 v1, v1, v5                                                    ; 3E020B01
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v6                                          ; 5E000D00
>	v_cvt_pkrtz_f16_f32_e32 v1, v7, v1                                          ; 5E020307
>	exp 15, 0, 1, 1, 1, v0, v1, v0, v1                                          ; F8001C0F 01000100
>	s_endpgm                                                                    ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 24
>VGPRS: 24
>Code Size: 336 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
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[0]
>DCL OUT[2], GENERIC[1]
>DCL OUT[3], GENERIC[2]
>DCL CONST[0..5]
>DCL TEMP[0..2], LOCAL
>IMM[0] FLT32 {    0.0000,     1.0000,     0.0000,     0.0000}
>  0: MOV TEMP[0].zw, IMM[0].yyxy
>  1: DP4 TEMP[0].x, IN[0], CONST[2]
>  2: DP4 TEMP[1].x, IN[0], CONST[3]
>  3: MOV TEMP[0].y, TEMP[1].xxxx
>  4: DP4 TEMP[1].x, IN[0], CONST[4]
>  5: DP4 TEMP[2].x, IN[0], CONST[5]
>  6: MOV TEMP[1].y, TEMP[2].xxxx
>  7: MOV TEMP[1].xy, TEMP[1].xyxx
>  8: MOV OUT[1], CONST[0]
>  9: MOV OUT[2], CONST[1]
> 10: MOV OUT[0], TEMP[0]
> 11: MOV OUT[3], TEMP[1]
> 12: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = call float @llvm.SI.load.const(<16 x i8> %13, i32 0)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %13, i32 4)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %13, i32 8)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %13, i32 12)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %13, i32 16)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %13, i32 20)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %13, i32 24)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %13, i32 28)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %13, i32 32)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %13, i32 36)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %13, i32 40)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %13, i32 44)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %13, i32 48)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %13, i32 52)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %13, i32 56)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %13, i32 60)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %13, i32 64)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %13, i32 68)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %13, i32 72)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %13, i32 76)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %13, i32 80)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %13, i32 84)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %13, i32 88)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %13, i32 92)
>  %38 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %39 = load <16 x i8>, <16 x i8> addrspace(2)* %38, align 16, !tbaa !0
>  %40 = add i32 %5, %8
>  %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 = extractelement <4 x float> %41, i32 2
>  %45 = extractelement <4 x float> %41, i32 3
>  %46 = fmul float %42, %22
>  %47 = fmul float %43, %23
>  %48 = fadd float %46, %47
>  %49 = fmul float %44, %24
>  %50 = fadd float %48, %49
>  %51 = fmul float %45, %25
>  %52 = fadd float %50, %51
>  %53 = fmul float %42, %26
>  %54 = fmul float %43, %27
>  %55 = fadd float %53, %54
>  %56 = fmul float %44, %28
>  %57 = fadd float %55, %56
>  %58 = fmul float %45, %29
>  %59 = fadd float %57, %58
>  %60 = fmul float %42, %30
>  %61 = fmul float %43, %31
>  %62 = fadd float %60, %61
>  %63 = fmul float %44, %32
>  %64 = fadd float %62, %63
>  %65 = fmul float %45, %33
>  %66 = fadd float %64, %65
>  %67 = fmul float %42, %34
>  %68 = fmul float %43, %35
>  %69 = fadd float %67, %68
>  %70 = fmul float %44, %36
>  %71 = fadd float %69, %70
>  %72 = fmul float %45, %37
>  %73 = fadd float %71, %72
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %14, float %15, float %16, float %17)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %18, float %19, float %20, float %21)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 34, i32 0, float %66, float %73, float undef, float undef)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %52, float %59, 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[4:7], s[8:9], 0x0                  ; C0820900
>	v_add_i32_e32 v0, vcc, s10, v0                      ; 4A00000A
>	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 lgkmcnt(0)                                ; BF8C007F
>	s_buffer_load_dword s4, s[0:3], 0x0                 ; C2020100
>	s_buffer_load_dword s5, s[0:3], 0x1                 ; C2028101
>	s_buffer_load_dword s6, s[0:3], 0x2                 ; C2030102
>	s_buffer_load_dword s7, s[0:3], 0x3                 ; C2038103
>	s_buffer_load_dword s8, s[0:3], 0x4                 ; C2040104
>	s_buffer_load_dword s9, s[0:3], 0x5                 ; C2048105
>	s_buffer_load_dword s10, s[0:3], 0x6                ; C2050106
>	s_buffer_load_dword s11, s[0:3], 0x7                ; C2058107
>	s_buffer_load_dword s12, s[0:3], 0x8                ; C2060108
>	s_buffer_load_dword s13, s[0:3], 0x9                ; C2068109
>	s_buffer_load_dword s14, s[0:3], 0xa                ; C207010A
>	s_buffer_load_dword s15, s[0:3], 0xb                ; C207810B
>	s_buffer_load_dword s16, s[0:3], 0xc                ; C208010C
>	s_buffer_load_dword s17, s[0:3], 0xd                ; C208810D
>	s_buffer_load_dword s18, s[0:3], 0xe                ; C209010E
>	s_buffer_load_dword s19, s[0:3], 0xf                ; C209810F
>	s_buffer_load_dword s20, s[0:3], 0x10               ; C20A0110
>	s_buffer_load_dword s21, s[0:3], 0x11               ; C20A8111
>	s_buffer_load_dword s22, s[0:3], 0x14               ; C20B0114
>	s_buffer_load_dword s23, s[0:3], 0x15               ; C20B8115
>	s_buffer_load_dword s24, s[0:3], 0x12               ; C20C0112
>	s_buffer_load_dword s25, s[0:3], 0x13               ; C20C8113
>	s_buffer_load_dword s26, s[0:3], 0x16               ; C20D0116
>	s_buffer_load_dword s0, s[0:3], 0x17                ; C2000117
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	v_mov_b32_e32 v4, s4                                ; 7E080204
>	v_mov_b32_e32 v5, s5                                ; 7E0A0205
>	v_mov_b32_e32 v6, s6                                ; 7E0C0206
>	v_mov_b32_e32 v7, s7                                ; 7E0E0207
>	exp 15, 32, 0, 0, 0, v4, v5, v6, v7                 ; F800020F 07060504
>	s_waitcnt expcnt(0)                                 ; BF8C070F
>	v_mov_b32_e32 v4, s8                                ; 7E080208
>	v_mov_b32_e32 v5, s9                                ; 7E0A0209
>	v_mov_b32_e32 v6, s10                               ; 7E0C020A
>	v_mov_b32_e32 v7, s11                               ; 7E0E020B
>	exp 15, 33, 0, 0, 0, v4, v5, v6, v7                 ; F800021F 07060504
>	s_waitcnt vmcnt(0) expcnt(0)                        ; BF8C0700
>	v_mul_f32_e32 v4, s13, v1                           ; 1008020D
>	v_mul_f32_e32 v5, s17, v1                           ; 100A0211
>	v_mul_f32_e32 v6, s21, v1                           ; 100C0215
>	v_mul_f32_e32 v1, s23, v1                           ; 10020217
>	v_mac_f32_e32 v4, s12, v0                           ; 3E08000C
>	v_mac_f32_e32 v5, s16, v0                           ; 3E0A0010
>	v_mac_f32_e32 v6, s20, v0                           ; 3E0C0014
>	v_mac_f32_e32 v1, s22, v0                           ; 3E020016
>	v_mac_f32_e32 v4, s14, v2                           ; 3E08040E
>	v_mac_f32_e32 v5, s18, v2                           ; 3E0A0412
>	v_mac_f32_e32 v6, s24, v2                           ; 3E0C0418
>	v_mac_f32_e32 v1, s26, v2                           ; 3E02041A
>	v_mac_f32_e32 v4, s15, v3                           ; 3E08060F
>	v_mac_f32_e32 v5, s19, v3                           ; 3E0A0613
>	v_mac_f32_e32 v6, s25, v3                           ; 3E0C0619
>	v_mac_f32_e32 v1, s0, v3                            ; 3E020600
>	v_mov_b32_e32 v0, 1.0                               ; 7E0002F2
>	exp 15, 34, 0, 0, 0, v6, v1, v0, v0                 ; F800022F 00000106
>	s_waitcnt expcnt(0)                                 ; BF8C070F
>	v_mov_b32_e32 v1, 0                                 ; 7E020280
>	exp 15, 12, 0, 1, 0, v4, v5, v1, v0                 ; F80008CF 00010504
>	s_endpgm                                            ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 32
>VGPRS: 8
>Code Size: 280 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
>  clamp_color = 0
>FRAG
>PROPERTY FS_COLOR0_WRITES_ALL_CBUFS 1
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL IN[1], GENERIC[1], PERSPECTIVE
>DCL IN[2], GENERIC[2], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SAMP[1]
>DCL SVIEW[0], 2D, FLOAT
>DCL SVIEW[1], 2D, FLOAT
>DCL CONST[0..2]
>DCL CONST[4]
>DCL CONST[6]
>DCL TEMP[0..10], LOCAL
>IMM[0] FLT32 {    0.0000,     1.0000,     0.0000,     0.0000}
>  0: MOV TEMP[0].xy, IN[2].xyxx
>  1: MOV TEMP[1], IMM[0].xxxx
>  2: MOV TEMP[2], IMM[0].xxxx
>  3: MOV TEMP[3].y, IMM[0].xxxx
>  4: MOV TEMP[3].x, -CONST[0].xxxx
>  5: BGNLOOP :0
>  6:   FSLT TEMP[4].x, CONST[0].xxxx, TEMP[3].xxxx
>  7:   UIF TEMP[4].xxxx :0
>  8:     BRK
>  9:   ENDIF
> 10:   MOV TEMP[3].y, -CONST[0].yyyy
> 11:   BGNLOOP :0
> 12:     FSLT TEMP[5].x, CONST[0].yyyy, TEMP[3].yyyy
> 13:     UIF TEMP[5].xxxx :0
> 14:       BRK
> 15:     ENDIF
> 16:     ADD TEMP[6].xy, CONST[1].xyyy, TEMP[3].xyyy
> 17:     MAD TEMP[7].xy, TEMP[6].xyyy, CONST[6].xyyy, TEMP[0].xyyy
> 18:     MOV TEMP[8].xy, TEMP[7].xyyy
> 19:     MOV TEMP[8].w, IMM[0].xxxx
> 20:     TXB TEMP[9], TEMP[8], SAMP[1], 2D
> 21:     ADD TEMP[2], TEMP[2], TEMP[9]
> 22:     ADD TEMP[10].x, TEMP[3].yyyy, IMM[0].yyyy
> 23:     MOV TEMP[3].y, TEMP[10].xxxx
> 24:   ENDLOOP :0
> 25:   ADD TEMP[3].x, TEMP[3].xxxx, IMM[0].yyyy
> 26: ENDLOOP :0
> 27: MUL TEMP[1].w, TEMP[2], CONST[0].wwww
> 28: MUL TEMP[0].xy, IN[2].xyyy, CONST[4].xyyy
> 29: MOV TEMP[0].xy, TEMP[0].xyyy
> 30: MOV TEMP[0].w, IMM[0].xxxx
> 31: TXB TEMP[0], TEMP[0], SAMP[0], 2D
> 32: ADD TEMP[2].x, IMM[0].yyyy, -TEMP[0].wwww
> 33: MUL TEMP[2].x, TEMP[1].wwww, TEMP[2].xxxx
> 34: MUL TEMP[2].x, TEMP[2].xxxx, CONST[0].zzzz
> 35: MOV_SAT TEMP[2].x, TEMP[2].xxxx
> 36: MAD TEMP[1], CONST[2], TEMP[2].xxxx, TEMP[0]
> 37: MOV TEMP[0].w, IMM[0].yyyy
> 38: MOV TEMP[0].xyz, IN[1].xyzx
> 39: MUL TEMP[0], TEMP[1], TEMP[0]
> 40: MUL TEMP[1], TEMP[0], IN[1].wwww
> 41: MAD TEMP[1], IN[0], TEMP[1].wwww, TEMP[1]
> 42: MOV OUT[0], TEMP[1]
> 43: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %24 = load <16 x i8>, <16 x i8> addrspace(2)* %23, align 16, !tbaa !0
>  %25 = call float @llvm.SI.load.const(<16 x i8> %24, i32 0)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %24, i32 4)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %24, i32 8)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %24, i32 12)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %24, i32 16)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %24, i32 20)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %24, i32 32)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %24, i32 36)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %24, i32 40)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %24, i32 44)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %24, i32 64)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %24, i32 68)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %24, i32 96)
>  %38 = call float @llvm.SI.load.const(<16 x i8> %24, i32 100)
>  %39 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %40 = load <8 x i32>, <8 x i32> addrspace(2)* %39, align 32, !tbaa !0
>  %41 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %42 = load <4 x i32>, <4 x i32> addrspace(2)* %41, align 16, !tbaa !0
>  %43 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 1
>  %44 = load <8 x i32>, <8 x i32> addrspace(2)* %43, align 32, !tbaa !0
>  %45 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 1
>  %46 = load <4 x i32>, <4 x i32> addrspace(2)* %45, align 16, !tbaa !0
>  %47 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %8)
>  %48 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %8)
>  %49 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %6, <2 x i32> %8)
>  %50 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %6, <2 x i32> %8)
>  %51 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %6, <2 x i32> %8)
>  %52 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %6, <2 x i32> %8)
>  %53 = call float @llvm.SI.fs.interp(i32 2, i32 1, i32 %6, <2 x i32> %8)
>  %54 = call float @llvm.SI.fs.interp(i32 3, i32 1, i32 %6, <2 x i32> %8)
>  %55 = call float @llvm.SI.fs.interp(i32 0, i32 2, i32 %6, <2 x i32> %8)
>  %56 = call float @llvm.SI.fs.interp(i32 1, i32 2, i32 %6, <2 x i32> %8)
>  %57 = fsub float -0.000000e+00, %25
>  %58 = fsub float -0.000000e+00, %26
>  br label %LOOP
>
>LOOP:                                             ; preds = %IF47, %main_body
>  %temp11.0 = phi float [ 0.000000e+00, %main_body ], [ %temp11.1, %IF47 ]
>  %temp12.0 = phi float [ %57, %main_body ], [ %109, %IF47 ]
>  %59 = fcmp olt float %25, %temp12.0
>  br i1 %59, label %IF, label %ENDIF
>
>IF:                                               ; preds = %LOOP
>  %60 = fmul float %temp11.0, %28
>  %61 = fmul float %55, %35
>  %62 = fmul float %56, %36
>  %63 = bitcast float %61 to i32
>  %64 = bitcast float %62 to i32
>  %65 = insertelement <4 x i32> <i32 0, i32 undef, i32 undef, i32 undef>, i32 %63, i32 1
>  %66 = insertelement <4 x i32> %65, i32 %64, i32 2
>  %67 = call <4 x float> @llvm.SI.image.sample.b.v4i32(<4 x i32> %66, <8 x i32> %40, <4 x i32> %42, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %68 = extractelement <4 x float> %67, i32 0
>  %69 = extractelement <4 x float> %67, i32 1
>  %70 = extractelement <4 x float> %67, i32 2
>  %71 = extractelement <4 x float> %67, i32 3
>  %72 = fsub float 1.000000e+00, %71
>  %73 = fmul float %60, %72
>  %74 = fmul float %73, %27
>  %75 = call float @llvm.AMDIL.clamp.(float %74, float 0.000000e+00, float 1.000000e+00)
>  %76 = fmul float %31, %75
>  %77 = fadd float %76, %68
>  %78 = fmul float %32, %75
>  %79 = fadd float %78, %69
>  %80 = fmul float %33, %75
>  %81 = fadd float %80, %70
>  %82 = fmul float %34, %75
>  %83 = fadd float %82, %71
>  %84 = fmul float %77, %51
>  %85 = fmul float %79, %52
>  %86 = fmul float %81, %53
>  %87 = fmul float %84, %54
>  %88 = fmul float %85, %54
>  %89 = fmul float %86, %54
>  %90 = fmul float %83, %54
>  %91 = fmul float %47, %90
>  %92 = fadd float %91, %87
>  %93 = fmul float %48, %90
>  %94 = fadd float %93, %88
>  %95 = fmul float %49, %90
>  %96 = fadd float %95, %89
>  %97 = fmul float %50, %90
>  %98 = fadd float %97, %90
>  %99 = call i32 @llvm.SI.packf16(float %92, float %94)
>  %100 = bitcast i32 %99 to float
>  %101 = call i32 @llvm.SI.packf16(float %96, float %98)
>  %102 = bitcast i32 %101 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %100, float %102, float %100, float %102)
>  ret void
>
>ENDIF:                                            ; preds = %LOOP
>  %103 = fadd float %29, %temp12.0
>  %104 = fmul float %103, %37
>  %105 = fadd float %104, %55
>  %106 = bitcast float %105 to i32
>  %107 = insertelement <4 x i32> <i32 0, i32 undef, i32 undef, i32 undef>, i32 %106, i32 1
>  br label %LOOP45
>
>LOOP45:                                           ; preds = %ENDIF46, %ENDIF
>  %temp11.1 = phi float [ %temp11.0, %ENDIF ], [ %117, %ENDIF46 ]
>  %temp13.0 = phi float [ %58, %ENDIF ], [ %118, %ENDIF46 ]
>  %108 = fcmp olt float %26, %temp13.0
>  br i1 %108, label %IF47, label %ENDIF46
>
>IF47:                                             ; preds = %LOOP45
>  %109 = fadd float %temp12.0, 1.000000e+00
>  br label %LOOP
>
>ENDIF46:                                          ; preds = %LOOP45
>  %110 = fadd float %30, %temp13.0
>  %111 = fmul float %110, %38
>  %112 = fadd float %111, %56
>  %113 = bitcast float %112 to i32
>  %114 = insertelement <4 x i32> %107, i32 %113, i32 2
>  %115 = call <4 x float> @llvm.SI.image.sample.b.v4i32(<4 x i32> %114, <8 x i32> %44, <4 x i32> %46, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %116 = extractelement <4 x float> %115, i32 3
>  %117 = fadd float %temp11.1, %116
>  %118 = fadd float %temp13.0, 1.000000e+00
>  br label %LOOP45
>}
>
>; 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.image.sample.b.v4i32(<4 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #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_load_dwordx4 s[32:35], s[2:3], 0x0                                           ; C0900300
>	s_load_dwordx8 s[12:19], s[6:7], 0x0                                           ; C0C60700
>	s_load_dwordx8 s[20:27], s[6:7], 0x8                                           ; C0CA0708
>	s_mov_b32 m0, s10                                                              ; BEFC030A
>	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 v3, v0, 1, 0, [m0]                                             ; C80C0100
>	v_interp_p2_f32 v3, [v3], v1, 1, 0, [m0]                                       ; C80D0101
>	v_interp_p1_f32 v2, v0, 2, 0, [m0]                                             ; C8080200
>	v_interp_p2_f32 v2, [v2], v1, 2, 0, [m0]                                       ; C8090201
>	s_waitcnt lgkmcnt(0)                                                           ; BF8C007F
>	s_buffer_load_dword s8, s[32:35], 0x0                                          ; C2042100
>	s_buffer_load_dword s9, s[32:35], 0x1                                          ; C204A101
>	s_buffer_load_dword s3, s[32:35], 0x2                                          ; C201A102
>	s_buffer_load_dword s7, s[32:35], 0x3                                          ; C203A103
>	s_buffer_load_dword s28, s[32:35], 0x4                                         ; C20E2104
>	s_buffer_load_dword s29, s[32:35], 0x5                                         ; C20EA105
>	s_buffer_load_dword s6, s[32:35], 0x8                                          ; C2032108
>	s_buffer_load_dword s2, s[32:35], 0x9                                          ; C2012109
>	s_buffer_load_dword s1, s[32:35], 0xa                                          ; C200A10A
>	s_buffer_load_dword s0, s[32:35], 0xb                                          ; C200210B
>	s_buffer_load_dword s10, s[32:35], 0x10                                        ; C2052110
>	s_buffer_load_dword s11, s[32:35], 0x11                                        ; C205A111
>	s_buffer_load_dword s30, s[32:35], 0x18                                        ; C20F2118
>	s_buffer_load_dword s31, s[32:35], 0x19                                        ; C20FA119
>	v_interp_p1_f32 v5, v0, 3, 0, [m0]                                             ; C8140300
>	v_interp_p2_f32 v5, [v5], v1, 3, 0, [m0]                                       ; C8150301
>	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 v7, v0, 1, 1, [m0]                                             ; C81C0500
>	v_interp_p2_f32 v7, [v7], v1, 1, 1, [m0]                                       ; C81D0501
>	v_interp_p1_f32 v8, v0, 2, 1, [m0]                                             ; C8200600
>	v_interp_p2_f32 v8, [v8], v1, 2, 1, [m0]                                       ; C8210601
>	v_interp_p1_f32 v9, v0, 3, 1, [m0]                                             ; C8240700
>	v_interp_p2_f32 v9, [v9], v1, 3, 1, [m0]                                       ; C8250701
>	s_load_dwordx4 s[32:35], s[4:5], 0x0                                           ; C0900500
>	s_load_dwordx4 s[36:39], s[4:5], 0x4                                           ; C0920504
>	v_interp_p1_f32 v10, v0, 0, 2, [m0]                                            ; C8280800
>	v_interp_p2_f32 v10, [v10], v1, 0, 2, [m0]                                     ; C8290801
>	v_interp_p1_f32 v0, v0, 1, 2, [m0]                                             ; C8000900
>	v_interp_p2_f32 v0, [v0], v1, 1, 2, [m0]                                       ; C8010901
>	v_mov_b32_e32 v1, 0x80000000                                                   ; 7E0202FF 80000000
>	s_waitcnt lgkmcnt(0)                                                           ; BF8C007F
>	v_xor_b32_e32 v11, s8, v1                                                      ; 3A160208
>	v_xor_b32_e32 v12, s9, v1                                                      ; 3A180209
>	v_mov_b32_e32 v13, 0                                                           ; 7E1A0280
>	s_mov_b64 s[4:5], 0                                                            ; BE840480
>	v_mov_b32_e32 v1, v13                                                          ; 7E02030D
>	v_cmp_nlt_f32_e32 vcc, s8, v11                                                 ; 7C1C1608
>	s_and_saveexec_b64 s[40:41], vcc                                               ; BEA8246A
>	s_xor_b64 s[40:41], exec, s[40:41]                                             ; 89A8287E
>	s_cbranch_execz BB0_4                                                          ; BF880000
>	v_add_f32_e32 v13, s28, v11                                                    ; 061A161C
>	v_mad_f32 v15, s30, v13, v10                                                   ; D282000F 042A1A1E
>	v_mov_b32_e32 v14, 0                                                           ; 7E1C0280
>	s_mov_b64 s[42:43], 0                                                          ; BEAA0480
>	v_mov_b32_e32 v16, v1                                                          ; 7E200301
>	v_mov_b32_e32 v17, v12                                                         ; 7E22030C
>	v_mov_b32_e32 v13, v16                                                         ; 7E1A0310
>	v_cmp_nlt_f32_e32 vcc, s9, v17                                                 ; 7C1C2209
>	s_and_saveexec_b64 s[44:45], vcc                                               ; BEAC246A
>	s_xor_b64 s[44:45], exec, s[44:45]                                             ; 89AC2C7E
>	v_add_f32_e32 v16, s29, v17                                                    ; 0620221D
>	v_mad_f32 v16, s31, v16, v0                                                    ; D2820010 0402201F
>	image_sample_b v16, 8, 0, 0, 0, 0, 0, 0, 0, v[14:17], s[20:27], s[36:39]       ; F0940800 0125100E
>	s_waitcnt vmcnt(0)                                                             ; BF8C0770
>	v_add_f32_e32 v16, v16, v13                                                    ; 06201B10
>	v_add_f32_e32 v17, 1.0, v17                                                    ; 062222F2
>	s_or_b64 exec, exec, s[44:45]                                                  ; 88FE2C7E
>	s_or_b64 s[42:43], s[44:45], s[42:43]                                          ; 88AA2A2C
>	s_andn2_b64 exec, exec, s[42:43]                                               ; 8AFE2A7E
>	s_cbranch_execnz BB0_5                                                         ; BF890000
>	s_or_b64 exec, exec, s[42:43]                                                  ; 88FE2A7E
>	v_add_f32_e32 v11, 1.0, v11                                                    ; 061616F2
>	s_or_b64 exec, exec, s[40:41]                                                  ; 88FE287E
>	s_or_b64 s[4:5], s[40:41], s[4:5]                                              ; 88840428
>	s_andn2_b64 exec, exec, s[4:5]                                                 ; 8AFE047E
>	s_cbranch_execnz BB0_1                                                         ; BF890000
>	s_or_b64 exec, exec, s[4:5]                                                    ; 88FE047E
>	v_mul_f32_e32 v1, s7, v1                                                       ; 10020207
>	v_mul_f32_e32 v11, s10, v10                                                    ; 1016140A
>	v_mul_f32_e32 v12, s11, v0                                                     ; 1018000B
>	v_mov_b32_e32 v10, 0                                                           ; 7E140280
>	image_sample_b v[10:13], 15, 0, 0, 0, 0, 0, 0, 0, v[10:13], s[12:19], s[32:35] ; F0940F00 01030A0A
>	s_waitcnt vmcnt(0)                                                             ; BF8C0770
>	v_mad_f32 v0, -v13, v1, v1                                                     ; D2820000 2406030D
>	v_mul_f32_e32 v0, s3, v0                                                       ; 10000003
>	v_add_f32_e64 v0, 0, v0 clamp                                                  ; D2060800 00020080
>	v_mad_f32 v1, s6, v0, v10                                                      ; D2820001 042A0006
>	v_mad_f32 v10, s2, v0, v11                                                     ; D282000A 042E0002
>	v_mad_f32 v11, s1, v0, v12                                                     ; D282000B 04320001
>	v_mac_f32_e32 v13, s0, v0                                                      ; 3E1A0000
>	v_mul_f32_e32 v0, v6, v1                                                       ; 10000306
>	v_mul_f32_e32 v1, v7, v10                                                      ; 10021507
>	v_mul_f32_e32 v6, v8, v11                                                      ; 100C1708
>	v_mul_f32_e32 v0, v9, v0                                                       ; 10000109
>	v_mul_f32_e32 v1, v9, v1                                                       ; 10020309
>	v_mul_f32_e32 v6, v9, v6                                                       ; 100C0D09
>	v_mul_f32_e32 v7, v9, v13                                                      ; 100E1B09
>	v_mac_f32_e32 v0, v7, v4                                                       ; 3E000907
>	v_mac_f32_e32 v1, v7, v3                                                       ; 3E020707
>	v_mac_f32_e32 v6, v7, v2                                                       ; 3E0C0507
>	v_mac_f32_e32 v7, v7, v5                                                       ; 3E0E0B07
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v1                                             ; 5E000300
>	v_cvt_pkrtz_f16_f32_e32 v1, v6, v7                                             ; 5E020F06
>	exp 15, 0, 1, 1, 1, v0, v1, v0, v1                                             ; F8001C0F 01000100
>	s_endpgm                                                                       ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 48
>VGPRS: 20
>Code Size: 476 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
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL CONST[0..47]
>DCL TEMP[0..1], LOCAL
>DCL ADDR[0]
>IMM[0] FLT32 {    0.0000,     1.0000,     2.0000,     0.1000}
>IMM[1] FLT32 {    1.1000,     0.0000,     0.0000,     0.0000}
>  0: MOV TEMP[0].zw, IMM[0].yyxy
>  1: MAD TEMP[1].x, IN[1].xxxx, IMM[0].zzzz, IMM[0].wwww
>  2: F2I TEMP[1].x, TEMP[1].xxxx
>  3: UARL ADDR[0].x, TEMP[1].xxxx
>  4: UARL ADDR[0].x, TEMP[1].xxxx
>  5: DP4 TEMP[0].x, IN[0], CONST[ADDR[0].x]
>  6: MAD TEMP[1].x, IN[1].xxxx, IMM[0].zzzz, IMM[1].xxxx
>  7: F2I TEMP[1].x, TEMP[1].xxxx
>  8: UARL ADDR[0].x, TEMP[1].xxxx
>  9: DP4 TEMP[1].x, IN[0], CONST[ADDR[0].x]
> 10: MOV TEMP[0].y, TEMP[1].xxxx
> 11: MOV OUT[0], TEMP[0]
> 12: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %15 = load <16 x i8>, <16 x i8> addrspace(2)* %14, align 16, !tbaa !0
>  %16 = add i32 %5, %8
>  %17 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %15, i32 0, i32 %16)
>  %18 = extractelement <4 x float> %17, i32 0
>  %19 = extractelement <4 x float> %17, i32 1
>  %20 = extractelement <4 x float> %17, i32 2
>  %21 = extractelement <4 x float> %17, i32 3
>  %22 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 1
>  %23 = load <16 x i8>, <16 x i8> addrspace(2)* %22, align 16, !tbaa !0
>  %24 = add i32 %5, %8
>  %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 = fmul float %26, 2.000000e+00
>  %28 = fadd float %27, 0x3FB99999A0000000
>  %29 = fptosi float %28 to i32
>  %30 = shl i32 %29, 4
>  %31 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %30)
>  %32 = shl i32 %29, 4
>  %33 = or i32 %32, 4
>  %34 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %33)
>  %35 = shl i32 %29, 4
>  %36 = or i32 %35, 8
>  %37 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %36)
>  %38 = shl i32 %29, 4
>  %39 = or i32 %38, 12
>  %40 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %39)
>  %41 = fmul float %18, %31
>  %42 = fmul float %19, %34
>  %43 = fadd float %41, %42
>  %44 = fmul float %20, %37
>  %45 = fadd float %43, %44
>  %46 = fmul float %21, %40
>  %47 = fadd float %45, %46
>  %48 = fmul float %26, 2.000000e+00
>  %49 = fadd float %48, 0x3FF19999A0000000
>  %50 = fptosi float %49 to i32
>  %51 = shl i32 %50, 4
>  %52 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %51)
>  %53 = shl i32 %50, 4
>  %54 = or i32 %53, 4
>  %55 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %54)
>  %56 = shl i32 %50, 4
>  %57 = or i32 %56, 8
>  %58 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %57)
>  %59 = shl i32 %50, 4
>  %60 = or i32 %59, 12
>  %61 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %60)
>  %62 = fmul float %18, %52
>  %63 = fmul float %19, %55
>  %64 = fadd float %62, %63
>  %65 = fmul float %20, %58
>  %66 = fadd float %64, %65
>  %67 = fmul float %21, %61
>  %68 = fadd float %66, %67
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %47, float %68, 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_load_dwordx4 s[4:7], s[8:9], 0x0                    ; C0820900
>	s_load_dwordx4 s[12:15], s[8:9], 0x4                  ; C0860904
>	v_mov_b32_e32 v1, 1.0                                 ; 7E0202F2
>	v_add_i32_e32 v0, vcc, s10, v0                        ; 4A00000A
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	buffer_load_format_xyzw v[2:5], v0, s[4:7], 0 idxen   ; E00C2000 80010200
>	buffer_load_format_xyzw v[6:9], v0, s[12:15], 0 idxen ; E00C2000 80030600
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_madak_f32_e32 v0, 2.0, v6, 0x3dcccccd               ; 42000CF4 3DCCCCCD
>	v_madak_f32_e32 v6, 2.0, v6, 0x3f8ccccd               ; 420C0CF4 3F8CCCCD
>	v_cvt_i32_f32_e32 v0, v0                              ; 7E001100
>	v_cvt_i32_f32_e32 v6, v6                              ; 7E0C1106
>	v_lshlrev_b32_e32 v0, 4, v0                           ; 34000084
>	v_lshlrev_b32_e32 v6, 4, v6                           ; 340C0C84
>	buffer_load_dword v7, v0, s[0:3], 0 offen             ; E0301000 80000700
>	v_or_b32_e32 v8, 4, v0                                ; 38100084
>	v_or_b32_e32 v9, 8, v0                                ; 38120088
>	v_or_b32_e32 v0, 12, v0                               ; 3800008C
>	buffer_load_dword v10, v6, s[0:3], 0 offen            ; E0301000 80000A06
>	v_or_b32_e32 v11, 4, v6                               ; 38160C84
>	buffer_load_dword v8, v8, s[0:3], 0 offen             ; E0301000 80000808
>	v_or_b32_e32 v12, 8, v6                               ; 38180C88
>	buffer_load_dword v11, v11, s[0:3], 0 offen           ; E0301000 80000B0B
>	v_or_b32_e32 v6, 12, v6                               ; 380C0C8C
>	buffer_load_dword v9, v9, s[0:3], 0 offen             ; E0301000 80000909
>	buffer_load_dword v12, v12, s[0:3], 0 offen           ; E0301000 80000C0C
>	buffer_load_dword v0, v0, s[0:3], 0 offen             ; E0301000 80000000
>	buffer_load_dword v6, v6, s[0:3], 0 offen             ; E0301000 80000606
>	s_waitcnt vmcnt(5)                                    ; BF8C0775
>	v_mul_f32_e32 v8, v8, v3                              ; 10100708
>	v_mac_f32_e32 v8, v7, v2                              ; 3E100507
>	s_waitcnt vmcnt(4)                                    ; BF8C0774
>	v_mul_f32_e32 v3, v11, v3                             ; 1006070B
>	v_mac_f32_e32 v3, v10, v2                             ; 3E06050A
>	s_waitcnt vmcnt(3)                                    ; BF8C0773
>	v_mac_f32_e32 v8, v9, v4                              ; 3E100909
>	s_waitcnt vmcnt(2)                                    ; BF8C0772
>	v_mac_f32_e32 v3, v12, v4                             ; 3E06090C
>	s_waitcnt vmcnt(1)                                    ; BF8C0771
>	v_mac_f32_e32 v8, v0, v5                              ; 3E100B00
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mac_f32_e32 v3, v6, v5                              ; 3E060B06
>	v_mov_b32_e32 v0, 0                                   ; 7E000280
>	exp 15, 12, 0, 1, 0, v8, v3, v0, v1                   ; F80008CF 01000308
>	s_endpgm                                              ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 24
>VGPRS: 16
>Code Size: 236 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
>  clamp_color = 0
>FRAG
>PROPERTY FS_COLOR0_WRITES_ALL_CBUFS 1
>DCL OUT[0], COLOR
>DCL CONST[0]
>  0: MOV OUT[0], CONST[0]
>  1: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %24 = load <16 x i8>, <16 x i8> addrspace(2)* %23, align 16, !tbaa !0
>  %25 = call float @llvm.SI.load.const(<16 x i8> %24, i32 0)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %24, i32 4)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %24, i32 8)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %24, i32 12)
>  %29 = call i32 @llvm.SI.packf16(float %25, float %26)
>  %30 = bitcast i32 %29 to float
>  %31 = call i32 @llvm.SI.packf16(float %27, float %28)
>  %32 = bitcast i32 %31 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %30, float %32, float %30, float %32)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<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_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_buffer_load_dword s5, s[0:3], 0x1 ; C2028101
>	s_buffer_load_dword s6, s[0:3], 0x2 ; C2030102
>	s_buffer_load_dword s0, s[0:3], 0x3 ; C2000103
>	s_waitcnt lgkmcnt(0)                ; BF8C007F
>	v_mov_b32_e32 v0, s5                ; 7E000205
>	v_cvt_pkrtz_f16_f32_e32 v0, s4, v0  ; 5E000004
>	v_mov_b32_e32 v1, s0                ; 7E020200
>	v_cvt_pkrtz_f16_f32_e32 v1, s6, v1  ; 5E020206
>	exp 15, 0, 1, 1, 1, v0, v1, v0, v1  ; F8001C0F 01000100
>	s_endpgm                            ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 8
>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
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL OUT[0], POSITION
>DCL CONST[0..1]
>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: DP4 TEMP[0].x, IN[0], CONST[0]
>  2: DP4 TEMP[1].x, IN[0], CONST[1]
>  3: MOV TEMP[0].y, TEMP[1].xxxx
>  4: MOV OUT[0], TEMP[0]
>  5: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = call float @llvm.SI.load.const(<16 x i8> %13, i32 0)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %13, i32 4)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %13, i32 8)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %13, i32 12)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %13, i32 16)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %13, i32 20)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %13, i32 24)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %13, i32 28)
>  %22 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %23 = load <16 x i8>, <16 x i8> addrspace(2)* %22, align 16, !tbaa !0
>  %24 = add i32 %5, %8
>  %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 = extractelement <4 x float> %25, i32 3
>  %30 = fmul float %26, %14
>  %31 = fmul float %27, %15
>  %32 = fadd float %30, %31
>  %33 = fmul float %28, %16
>  %34 = fadd float %32, %33
>  %35 = fmul float %29, %17
>  %36 = fadd float %34, %35
>  %37 = fmul float %26, %18
>  %38 = fmul float %27, %19
>  %39 = fadd float %37, %38
>  %40 = fmul float %28, %20
>  %41 = fadd float %39, %40
>  %42 = fmul float %29, %21
>  %43 = fadd float %41, %42
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %36, float %43, 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[4:7], s[8:9], 0x0                  ; C0820900
>	v_add_i32_e32 v0, vcc, s10, v0                      ; 4A00000A
>	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 lgkmcnt(0)                                ; BF8C007F
>	s_buffer_load_dword s4, s[0:3], 0x1                 ; C2020101
>	s_buffer_load_dword s5, s[0:3], 0x5                 ; C2028105
>	s_buffer_load_dword s6, s[0:3], 0x0                 ; C2030100
>	s_buffer_load_dword s7, s[0:3], 0x4                 ; C2038104
>	s_buffer_load_dword s8, s[0:3], 0x2                 ; C2040102
>	s_buffer_load_dword s9, s[0:3], 0x6                 ; C2048106
>	s_buffer_load_dword s10, s[0:3], 0x3                ; C2050103
>	s_buffer_load_dword s0, s[0:3], 0x7                 ; C2000107
>	s_waitcnt vmcnt(0) lgkmcnt(0)                       ; BF8C0070
>	v_mul_f32_e32 v4, s4, v1                            ; 10080204
>	v_mul_f32_e32 v1, s5, v1                            ; 10020205
>	v_mac_f32_e32 v4, s6, v0                            ; 3E080006
>	v_mac_f32_e32 v1, s7, v0                            ; 3E020007
>	v_mac_f32_e32 v4, s8, v2                            ; 3E080408
>	v_mac_f32_e32 v1, s9, v2                            ; 3E020409
>	v_mac_f32_e32 v4, s10, v3                           ; 3E08060A
>	v_mac_f32_e32 v1, s0, v3                            ; 3E020600
>	v_mov_b32_e32 v0, 1.0                               ; 7E0002F2
>	v_mov_b32_e32 v2, 0                                 ; 7E040280
>	exp 15, 12, 0, 1, 0, v4, v1, v2, v0                 ; F80008CF 00020104
>	s_endpgm                                            ; BF810000
>
>*** SHADER STATS ***
>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
>  clamp_color = 0
>FRAG
>PROPERTY FS_COLOR0_WRITES_ALL_CBUFS 1
>DCL OUT[0], COLOR
>DCL CONST[0]
>  0: MOV OUT[0], CONST[0]
>  1: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %24 = load <16 x i8>, <16 x i8> addrspace(2)* %23, align 16, !tbaa !0
>  %25 = call float @llvm.SI.load.const(<16 x i8> %24, i32 0)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %24, i32 4)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %24, i32 8)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %24, i32 12)
>  %29 = call i32 @llvm.SI.packf16(float %25, float %26)
>  %30 = bitcast i32 %29 to float
>  %31 = call i32 @llvm.SI.packf16(float %27, float %28)
>  %32 = bitcast i32 %31 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %30, float %32, float %30, float %32)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.load.const(<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_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_buffer_load_dword s5, s[0:3], 0x1 ; C2028101
>	s_buffer_load_dword s6, s[0:3], 0x2 ; C2030102
>	s_buffer_load_dword s0, s[0:3], 0x3 ; C2000103
>	s_waitcnt lgkmcnt(0)                ; BF8C007F
>	v_mov_b32_e32 v0, s5                ; 7E000205
>	v_cvt_pkrtz_f16_f32_e32 v0, s4, v0  ; 5E000004
>	v_mov_b32_e32 v1, s0                ; 7E020200
>	v_cvt_pkrtz_f16_f32_e32 v1, s6, v1  ; 5E020206
>	exp 15, 0, 1, 1, 1, v0, v1, v0, v1  ; F8001C0F 01000100
>	s_endpgm                            ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 8
>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
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL IN[2]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[0]
>DCL OUT[2], GENERIC[1]
>DCL CONST[0..1]
>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: DP4 TEMP[0].x, IN[2], CONST[0]
>  2: DP4 TEMP[1].x, IN[2], CONST[1]
>  3: MOV TEMP[0].y, TEMP[1].xxxx
>  4: MOV OUT[2], IN[1]
>  5: MOV OUT[1], IN[0]
>  6: MOV OUT[0], TEMP[0]
>  7: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = call float @llvm.SI.load.const(<16 x i8> %13, i32 0)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %13, i32 4)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %13, i32 8)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %13, i32 12)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %13, i32 16)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %13, i32 20)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %13, i32 24)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %13, i32 28)
>  %22 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %23 = load <16 x i8>, <16 x i8> addrspace(2)* %22, align 16, !tbaa !0
>  %24 = add i32 %5, %8
>  %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 = extractelement <4 x float> %25, i32 3
>  %30 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 1
>  %31 = load <16 x i8>, <16 x i8> addrspace(2)* %30, align 16, !tbaa !0
>  %32 = add i32 %5, %8
>  %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 = extractelement <4 x float> %33, i32 2
>  %37 = extractelement <4 x float> %33, i32 3
>  %38 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 2
>  %39 = load <16 x i8>, <16 x i8> addrspace(2)* %38, align 16, !tbaa !0
>  %40 = add i32 %5, %8
>  %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 = extractelement <4 x float> %41, i32 2
>  %45 = extractelement <4 x float> %41, i32 3
>  %46 = fmul float %42, %14
>  %47 = fmul float %43, %15
>  %48 = fadd float %46, %47
>  %49 = fmul float %44, %16
>  %50 = fadd float %48, %49
>  %51 = fmul float %45, %17
>  %52 = fadd float %50, %51
>  %53 = fmul float %42, %18
>  %54 = fmul float %43, %19
>  %55 = fadd float %53, %54
>  %56 = fmul float %44, %20
>  %57 = fadd float %55, %56
>  %58 = fmul float %45, %21
>  %59 = fadd float %57, %58
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %26, float %27, float %28, float %29)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %34, float %35, float %36, float %37)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %52, float %59, 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_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
>	v_add_i32_e32 v0, vcc, s10, v0                          ; 4A00000A
>	v_mov_b32_e32 v1, 1.0                                   ; 7E0202F2
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	s_buffer_load_dword s8, s[0:3], 0x0                     ; C2040100
>	s_buffer_load_dword s9, s[0:3], 0x1                     ; C2048101
>	s_buffer_load_dword s10, s[0:3], 0x2                    ; C2050102
>	s_buffer_load_dword s11, s[0:3], 0x3                    ; C2058103
>	s_buffer_load_dword s20, s[0:3], 0x4                    ; C20A0104
>	buffer_load_format_xyzw v[2:5], v0, s[4:7], 0 idxen     ; E00C2000 80010200
>	buffer_load_format_xyzw v[6:9], v0, s[12:15], 0 idxen   ; E00C2000 80030600
>	buffer_load_format_xyzw v[10:13], v0, s[16:19], 0 idxen ; E00C2000 80040A00
>	s_buffer_load_dword s4, s[0:3], 0x5                     ; C2020105
>	s_buffer_load_dword s5, s[0:3], 0x6                     ; C2028106
>	s_buffer_load_dword s0, s[0:3], 0x7                     ; C2000107
>	s_waitcnt vmcnt(2)                                      ; BF8C0772
>	exp 15, 32, 0, 0, 0, v2, v3, v4, v5                     ; F800020F 05040302
>	s_waitcnt vmcnt(1)                                      ; BF8C0771
>	exp 15, 33, 0, 0, 0, v6, v7, v8, v9                     ; F800021F 09080706
>	s_waitcnt vmcnt(0) lgkmcnt(0)                           ; BF8C0070
>	v_mul_f32_e32 v0, s9, v11                               ; 10001609
>	s_waitcnt expcnt(0)                                     ; BF8C070F
>	v_mul_f32_e32 v2, s4, v11                               ; 10041604
>	v_mac_f32_e32 v0, s8, v10                               ; 3E001408
>	v_mac_f32_e32 v2, s20, v10                              ; 3E041414
>	v_mac_f32_e32 v0, s10, v12                              ; 3E00180A
>	v_mac_f32_e32 v2, s5, v12                               ; 3E041805
>	v_mac_f32_e32 v0, s11, v13                              ; 3E001A0B
>	v_mac_f32_e32 v2, s0, v13                               ; 3E041A00
>	v_mov_b32_e32 v3, 0                                     ; 7E060280
>	exp 15, 12, 0, 1, 0, v0, v2, v3, v1                     ; F80008CF 01030200
>	s_endpgm                                                ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 24
>VGPRS: 16
>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
>  clamp_color = 0
>FRAG
>PROPERTY FS_COLOR0_WRITES_ALL_CBUFS 1
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL IN[1], GENERIC[1], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL TEMP[0..1], LOCAL
>  0: MOV TEMP[0].xyz, IN[0].xyzx
>  1: MUL TEMP[1].x, IN[0].wwww, IN[1].wwww
>  2: MOV TEMP[0].w, TEMP[1].xxxx
>  3: MOV OUT[0], TEMP[0]
>  4: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %8)
>  %24 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %8)
>  %25 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %6, <2 x i32> %8)
>  %26 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %6, <2 x i32> %8)
>  %27 = call float @llvm.SI.fs.interp(i32 3, i32 1, i32 %6, <2 x i32> %8)
>  %28 = fmul float %26, %27
>  %29 = call i32 @llvm.SI.packf16(float %23, float %24)
>  %30 = bitcast i32 %29 to float
>  %31 = call i32 @llvm.SI.packf16(float %25, float %28)
>  %32 = bitcast i32 %31 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %30, float %32, float %30, float %32)
>  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_mov_b32 m0, s10                        ; BEFC030A
>	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_interp_p1_f32 v4, v0, 2, 0, [m0]       ; C8100200
>	v_interp_p2_f32 v4, [v4], v1, 2, 0, [m0] ; C8110201
>	v_interp_p1_f32 v5, v0, 3, 0, [m0]       ; C8140300
>	v_interp_p2_f32 v5, [v5], v1, 3, 0, [m0] ; C8150301
>	v_interp_p1_f32 v0, v0, 3, 1, [m0]       ; C8000700
>	v_interp_p2_f32 v0, [v0], v1, 3, 1, [m0] ; C8010701
>	v_mul_f32_e32 v0, v0, v5                 ; 10000B00
>	v_cvt_pkrtz_f16_f32_e32 v1, v2, v3       ; 5E020702
>	v_cvt_pkrtz_f16_f32_e32 v0, v4, v0       ; 5E000104
>	exp 15, 0, 1, 1, 1, v1, v0, v1, v0       ; F8001C0F 00010001
>	s_endpgm                                 ; BF810000
>
>*** SHADER STATS ***
>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
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL IN[2]
>DCL IN[3]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[0]
>DCL OUT[2], GENERIC[1]
>DCL CONST[0..95]
>DCL TEMP[0..2], LOCAL
>DCL ADDR[0]
>IMM[0] FLT32 {    0.0000,     1.0000,     4.0000,     2.1000}
>IMM[1] FLT32 {    3.1000,     1.1000,     0.1000,     0.0000}
>  0: MOV TEMP[0].zw, IMM[0].yyxy
>  1: MAD TEMP[1].x, IN[3].xxxx, IMM[0].zzzz, IMM[0].wwww
>  2: F2I TEMP[1].x, TEMP[1].xxxx
>  3: UARL ADDR[0].x, TEMP[1].xxxx
>  4: UARL ADDR[0].x, TEMP[1].xxxx
>  5: DP4 TEMP[0].x, IN[2], CONST[ADDR[0].x]
>  6: MAD TEMP[1].x, IN[3].xxxx, IMM[0].zzzz, IMM[1].xxxx
>  7: F2I TEMP[1].x, TEMP[1].xxxx
>  8: UARL ADDR[0].x, TEMP[1].xxxx
>  9: DP4 TEMP[1].x, IN[2], CONST[ADDR[0].x]
> 10: MOV TEMP[0].y, TEMP[1].xxxx
> 11: MAD TEMP[1].x, IN[3].xxxx, IMM[0].zzzz, IMM[1].yyyy
> 12: F2I TEMP[1].x, TEMP[1].xxxx
> 13: MAD TEMP[2].x, IN[3].xxxx, IMM[0].zzzz, IMM[1].zzzz
> 14: F2I TEMP[2].x, TEMP[2].xxxx
> 15: UARL ADDR[0].x, TEMP[2].xxxx
> 16: UARL ADDR[0].x, TEMP[2].xxxx
> 17: MOV TEMP[2], CONST[ADDR[0].x]
> 18: UARL ADDR[0].x, TEMP[1].xxxx
> 19: UARL ADDR[0].x, TEMP[1].xxxx
> 20: MAD TEMP[1], IN[0], CONST[ADDR[0].x], TEMP[2]
> 21: MOV TEMP[2].xy, IN[1].xyxx
> 22: MOV OUT[1], TEMP[1]
> 23: MOV OUT[0], TEMP[0]
> 24: MOV OUT[2], TEMP[2]
> 25: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %15 = load <16 x i8>, <16 x i8> addrspace(2)* %14, align 16, !tbaa !0
>  %16 = add i32 %5, %8
>  %17 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %15, i32 0, i32 %16)
>  %18 = extractelement <4 x float> %17, i32 0
>  %19 = extractelement <4 x float> %17, i32 1
>  %20 = extractelement <4 x float> %17, i32 2
>  %21 = extractelement <4 x float> %17, i32 3
>  %22 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 1
>  %23 = load <16 x i8>, <16 x i8> addrspace(2)* %22, align 16, !tbaa !0
>  %24 = add i32 %5, %8
>  %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 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 2
>  %29 = load <16 x i8>, <16 x i8> addrspace(2)* %28, align 16, !tbaa !0
>  %30 = add i32 %5, %8
>  %31 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %29, i32 0, i32 %30)
>  %32 = extractelement <4 x float> %31, i32 0
>  %33 = extractelement <4 x float> %31, i32 1
>  %34 = extractelement <4 x float> %31, i32 2
>  %35 = extractelement <4 x float> %31, i32 3
>  %36 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 3
>  %37 = load <16 x i8>, <16 x i8> addrspace(2)* %36, align 16, !tbaa !0
>  %38 = add i32 %5, %8
>  %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 = fmul float %40, 4.000000e+00
>  %42 = fadd float %41, 0x4000CCCCC0000000
>  %43 = fptosi float %42 to i32
>  %44 = shl i32 %43, 4
>  %45 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %44)
>  %46 = shl i32 %43, 4
>  %47 = or i32 %46, 4
>  %48 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %47)
>  %49 = shl i32 %43, 4
>  %50 = or i32 %49, 8
>  %51 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %50)
>  %52 = shl i32 %43, 4
>  %53 = or i32 %52, 12
>  %54 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %53)
>  %55 = fmul float %32, %45
>  %56 = fmul float %33, %48
>  %57 = fadd float %55, %56
>  %58 = fmul float %34, %51
>  %59 = fadd float %57, %58
>  %60 = fmul float %35, %54
>  %61 = fadd float %59, %60
>  %62 = fmul float %40, 4.000000e+00
>  %63 = fadd float %62, 0x4008CCCCC0000000
>  %64 = fptosi float %63 to i32
>  %65 = shl i32 %64, 4
>  %66 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %65)
>  %67 = shl i32 %64, 4
>  %68 = or i32 %67, 4
>  %69 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %68)
>  %70 = shl i32 %64, 4
>  %71 = or i32 %70, 8
>  %72 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %71)
>  %73 = shl i32 %64, 4
>  %74 = or i32 %73, 12
>  %75 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %74)
>  %76 = fmul float %32, %66
>  %77 = fmul float %33, %69
>  %78 = fadd float %76, %77
>  %79 = fmul float %34, %72
>  %80 = fadd float %78, %79
>  %81 = fmul float %35, %75
>  %82 = fadd float %80, %81
>  %83 = fmul float %40, 4.000000e+00
>  %84 = fadd float %83, 0x3FF19999A0000000
>  %85 = fptosi float %84 to i32
>  %86 = fmul float %40, 4.000000e+00
>  %87 = fadd float %86, 0x3FB99999A0000000
>  %88 = fptosi float %87 to i32
>  %89 = shl i32 %88, 4
>  %90 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %89)
>  %91 = shl i32 %88, 4
>  %92 = or i32 %91, 4
>  %93 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %92)
>  %94 = shl i32 %88, 4
>  %95 = or i32 %94, 8
>  %96 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %95)
>  %97 = shl i32 %88, 4
>  %98 = or i32 %97, 12
>  %99 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %98)
>  %100 = shl i32 %85, 4
>  %101 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %100)
>  %102 = fmul float %18, %101
>  %103 = fadd float %102, %90
>  %104 = shl i32 %85, 4
>  %105 = or i32 %104, 4
>  %106 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %105)
>  %107 = fmul float %19, %106
>  %108 = fadd float %107, %93
>  %109 = shl i32 %85, 4
>  %110 = or i32 %109, 8
>  %111 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %110)
>  %112 = fmul float %20, %111
>  %113 = fadd float %112, %96
>  %114 = shl i32 %85, 4
>  %115 = or i32 %114, 12
>  %116 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %115)
>  %117 = fmul float %21, %116
>  %118 = fadd float %117, %99
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %103, float %108, float %113, float %118)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %26, float %27, float %96, float %99)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %61, float %82, 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_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
>	v_mov_b32_e32 v1, 1.0                                   ; 7E0202F2
>	v_add_i32_e32 v0, vcc, s10, v0                          ; 4A00000A
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	buffer_load_format_xyzw v[2:5], v0, s[4:7], 0 idxen     ; E00C2000 80010200
>	buffer_load_format_xyzw v[6:9], v0, s[12:15], 0 idxen   ; E00C2000 80030600
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	buffer_load_format_xyzw v[8:11], v0, s[16:19], 0 idxen  ; E00C2000 80040800
>	buffer_load_format_xyzw v[12:15], v0, s[20:23], 0 idxen ; E00C2000 80050C00
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	v_madak_f32_e32 v0, 4.0, v12, 0x40066666                ; 420018F6 40066666
>	v_madak_f32_e32 v13, 4.0, v12, 0x40466666               ; 421A18F6 40466666
>	v_madak_f32_e32 v14, 4.0, v12, 0x3f8ccccd               ; 421C18F6 3F8CCCCD
>	v_madak_f32_e32 v12, 4.0, v12, 0x3dcccccd               ; 421818F6 3DCCCCCD
>	v_cvt_i32_f32_e32 v0, v0                                ; 7E001100
>	v_cvt_i32_f32_e32 v13, v13                              ; 7E1A110D
>	v_cvt_i32_f32_e32 v12, v12                              ; 7E18110C
>	v_cvt_i32_f32_e32 v14, v14                              ; 7E1C110E
>	v_lshlrev_b32_e32 v0, 4, v0                             ; 34000084
>	v_lshlrev_b32_e32 v13, 4, v13                           ; 341A1A84
>	v_lshlrev_b32_e32 v12, 4, v12                           ; 34181884
>	v_lshlrev_b32_e32 v14, 4, v14                           ; 341C1C84
>	buffer_load_dword v15, v0, s[0:3], 0 offen              ; E0301000 80000F00
>	v_or_b32_e32 v16, 4, v0                                 ; 38200084
>	v_or_b32_e32 v17, 8, v0                                 ; 38220088
>	v_or_b32_e32 v0, 12, v0                                 ; 3800008C
>	buffer_load_dword v18, v12, s[0:3], 0 offen             ; E0301000 8000120C
>	v_or_b32_e32 v19, 4, v12                                ; 38261884
>	v_or_b32_e32 v20, 8, v12                                ; 38281888
>	v_or_b32_e32 v12, 12, v12                               ; 3818188C
>	buffer_load_dword v21, v14, s[0:3], 0 offen             ; E0301000 8000150E
>	v_or_b32_e32 v22, 4, v14                                ; 382C1C84
>	v_or_b32_e32 v23, 8, v14                                ; 382E1C88
>	v_or_b32_e32 v14, 12, v14                               ; 381C1C8C
>	buffer_load_dword v19, v19, s[0:3], 0 offen             ; E0301000 80001313
>	buffer_load_dword v20, v20, s[0:3], 0 offen             ; E0301000 80001414
>	buffer_load_dword v12, v12, s[0:3], 0 offen             ; E0301000 80000C0C
>	buffer_load_dword v22, v22, s[0:3], 0 offen             ; E0301000 80001616
>	buffer_load_dword v23, v23, s[0:3], 0 offen             ; E0301000 80001717
>	buffer_load_dword v14, v14, s[0:3], 0 offen             ; E0301000 80000E0E
>	buffer_load_dword v16, v16, s[0:3], 0 offen             ; E0301000 80001010
>	v_or_b32_e32 v24, 4, v13                                ; 38301A84
>	buffer_load_dword v24, v24, s[0:3], 0 offen             ; E0301000 80001818
>	buffer_load_dword v25, v13, s[0:3], 0 offen             ; E0301000 8000190D
>	buffer_load_dword v17, v17, s[0:3], 0 offen             ; E0301000 80001111
>	v_or_b32_e32 v26, 8, v13                                ; 38341A88
>	buffer_load_dword v26, v26, s[0:3], 0 offen             ; E0301000 80001A1A
>	v_or_b32_e32 v13, 12, v13                               ; 381A1A8C
>	buffer_load_dword v0, v0, s[0:3], 0 offen               ; E0301000 80000000
>	buffer_load_dword v13, v13, s[0:3], 0 offen             ; E0301000 80000D0D
>	s_waitcnt vmcnt(13)                                     ; BF8C077D
>	v_mac_f32_e32 v18, v21, v2                              ; 3E240515
>	s_waitcnt vmcnt(9)                                      ; BF8C0779
>	v_mac_f32_e32 v19, v22, v3                              ; 3E260716
>	s_waitcnt vmcnt(8)                                      ; BF8C0778
>	v_mad_f32 v2, v23, v4, v20                              ; D2820002 04520917
>	s_waitcnt vmcnt(7)                                      ; BF8C0777
>	v_mad_f32 v3, v14, v5, v12                              ; D2820003 04320B0E
>	s_waitcnt vmcnt(6)                                      ; BF8C0776
>	v_mul_f32_e32 v4, v16, v9                               ; 10081310
>	v_mac_f32_e32 v4, v15, v8                               ; 3E08110F
>	s_waitcnt vmcnt(5)                                      ; BF8C0775
>	v_mul_f32_e32 v5, v24, v9                               ; 100A1318
>	s_waitcnt vmcnt(4)                                      ; BF8C0774
>	v_mac_f32_e32 v5, v25, v8                               ; 3E0A1119
>	exp 15, 32, 0, 0, 0, v18, v19, v2, v3                   ; F800020F 03021312
>	s_waitcnt vmcnt(3)                                      ; BF8C0773
>	v_mac_f32_e32 v4, v17, v10                              ; 3E081511
>	s_waitcnt vmcnt(2)                                      ; BF8C0772
>	v_mac_f32_e32 v5, v26, v10                              ; 3E0A151A
>	exp 15, 33, 0, 0, 0, v6, v7, v20, v12                   ; F800021F 0C140706
>	s_waitcnt vmcnt(1)                                      ; BF8C0771
>	v_mac_f32_e32 v4, v0, v11                               ; 3E081700
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	v_mac_f32_e32 v5, v13, v11                              ; 3E0A170D
>	v_mov_b32_e32 v0, 0                                     ; 7E000280
>	exp 15, 12, 0, 1, 0, v4, v5, v0, v1                     ; F80008CF 01000504
>	s_endpgm                                                ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 32
>VGPRS: 28
>Code Size: 444 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
>  clamp_color = 0
>FRAG
>PROPERTY FS_COLOR0_WRITES_ALL_CBUFS 1
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL IN[1], GENERIC[1], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SVIEW[0], 2D, FLOAT
>DCL TEMP[0..1], LOCAL
>  0: MOV TEMP[0].xyz, IN[0].xyzx
>  1: MOV TEMP[1].xy, IN[1].xyyy
>  2: TEX TEMP[1].w, TEMP[1], SAMP[0], 2D
>  3: MUL TEMP[1].x, IN[0].wwww, TEMP[1].wwww
>  4: MOV TEMP[0].w, TEMP[1].xxxx
>  5: MOV OUT[0], TEMP[0]
>  6: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %24 = load <8 x i32>, <8 x i32> addrspace(2)* %23, align 32, !tbaa !0
>  %25 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %26 = load <4 x i32>, <4 x i32> addrspace(2)* %25, align 16, !tbaa !0
>  %27 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %8)
>  %28 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %8)
>  %29 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %6, <2 x i32> %8)
>  %30 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %6, <2 x i32> %8)
>  %31 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %6, <2 x i32> %8)
>  %32 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %6, <2 x i32> %8)
>  %33 = bitcast float %31 to i32
>  %34 = bitcast float %32 to i32
>  %35 = insertelement <2 x i32> undef, i32 %33, i32 0
>  %36 = insertelement <2 x i32> %35, i32 %34, i32 1
>  %37 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %36, <8 x i32> %24, <4 x i32> %26, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %38 = extractelement <4 x float> %37, i32 3
>  %39 = fmul float %30, %38
>  %40 = call i32 @llvm.SI.packf16(float %27, float %28)
>  %41 = bitcast i32 %40 to float
>  %42 = call i32 @llvm.SI.packf16(float %29, float %39)
>  %43 = bitcast i32 %42 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %41, float %43, float %41, float %43)
>  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.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, 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 }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                             ; BEFE0A7E
>	s_mov_b32 m0, s10                                                ; BEFC030A
>	s_load_dwordx8 s[8:15], s[6:7], 0x0                              ; C0C40700
>	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
>	s_load_dwordx4 s[0:3], s[4:5], 0x0                               ; C0800500
>	v_interp_p2_f32 v3, [v3], v1, 1, 0, [m0]                         ; C80D0101
>	v_interp_p1_f32 v4, v0, 2, 0, [m0]                               ; C8100200
>	v_interp_p2_f32 v4, [v4], v1, 2, 0, [m0]                         ; C8110201
>	v_interp_p1_f32 v5, v0, 3, 0, [m0]                               ; C8140300
>	v_interp_p2_f32 v5, [v5], v1, 3, 0, [m0]                         ; C8150301
>	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 v7, v0, 1, 1, [m0]                               ; C81C0500
>	v_interp_p2_f32 v7, [v7], v1, 1, 1, [m0]                         ; C81D0501
>	s_waitcnt lgkmcnt(0)                                             ; BF8C007F
>	image_sample v0, 8, 0, 0, 0, 0, 0, 0, 0, v[6:7], s[8:15], s[0:3] ; F0800800 00020006
>	s_waitcnt vmcnt(0)                                               ; BF8C0770
>	v_mul_f32_e32 v0, v0, v5                                         ; 10000B00
>	v_cvt_pkrtz_f16_f32_e32 v1, v2, v3                               ; 5E020702
>	v_cvt_pkrtz_f16_f32_e32 v0, v4, v0                               ; 5E000104
>	exp 15, 0, 1, 1, 1, v1, v0, v1, v0                               ; F8001C0F 00010001
>	s_endpgm                                                         ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 16
>VGPRS: 8
>Code Size: 104 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
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[0]
>DCL OUT[2], GENERIC[1]
>DCL OUT[3], GENERIC[2]
>DCL CONST[0..5]
>DCL TEMP[0..2], LOCAL
>IMM[0] FLT32 {    0.0000,     1.0000,     0.0000,     0.0000}
>  0: MOV TEMP[0].zw, IMM[0].yyxy
>  1: DP4 TEMP[0].x, IN[0], CONST[2]
>  2: DP4 TEMP[1].x, IN[0], CONST[3]
>  3: MOV TEMP[0].y, TEMP[1].xxxx
>  4: DP4 TEMP[1].x, IN[0], CONST[4]
>  5: DP4 TEMP[2].x, IN[0], CONST[5]
>  6: MOV TEMP[1].y, TEMP[2].xxxx
>  7: MOV TEMP[1].xy, TEMP[1].xyxx
>  8: MOV OUT[1], CONST[0]
>  9: MOV OUT[2], CONST[1]
> 10: MOV OUT[0], TEMP[0]
> 11: MOV OUT[3], TEMP[1]
> 12: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = call float @llvm.SI.load.const(<16 x i8> %13, i32 0)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %13, i32 4)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %13, i32 8)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %13, i32 12)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %13, i32 16)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %13, i32 20)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %13, i32 24)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %13, i32 28)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %13, i32 32)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %13, i32 36)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %13, i32 40)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %13, i32 44)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %13, i32 48)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %13, i32 52)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %13, i32 56)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %13, i32 60)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %13, i32 64)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %13, i32 68)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %13, i32 72)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %13, i32 76)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %13, i32 80)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %13, i32 84)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %13, i32 88)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %13, i32 92)
>  %38 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %39 = load <16 x i8>, <16 x i8> addrspace(2)* %38, align 16, !tbaa !0
>  %40 = add i32 %5, %8
>  %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 = extractelement <4 x float> %41, i32 2
>  %45 = extractelement <4 x float> %41, i32 3
>  %46 = fmul float %42, %22
>  %47 = fmul float %43, %23
>  %48 = fadd float %46, %47
>  %49 = fmul float %44, %24
>  %50 = fadd float %48, %49
>  %51 = fmul float %45, %25
>  %52 = fadd float %50, %51
>  %53 = fmul float %42, %26
>  %54 = fmul float %43, %27
>  %55 = fadd float %53, %54
>  %56 = fmul float %44, %28
>  %57 = fadd float %55, %56
>  %58 = fmul float %45, %29
>  %59 = fadd float %57, %58
>  %60 = fmul float %42, %30
>  %61 = fmul float %43, %31
>  %62 = fadd float %60, %61
>  %63 = fmul float %44, %32
>  %64 = fadd float %62, %63
>  %65 = fmul float %45, %33
>  %66 = fadd float %64, %65
>  %67 = fmul float %42, %34
>  %68 = fmul float %43, %35
>  %69 = fadd float %67, %68
>  %70 = fmul float %44, %36
>  %71 = fadd float %69, %70
>  %72 = fmul float %45, %37
>  %73 = fadd float %71, %72
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %14, float %15, float %16, float %17)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %18, float %19, float %20, float %21)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 34, i32 0, float %66, float %73, float undef, float undef)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %52, float %59, 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[4:7], s[8:9], 0x0                  ; C0820900
>	v_add_i32_e32 v0, vcc, s10, v0                      ; 4A00000A
>	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 lgkmcnt(0)                                ; BF8C007F
>	s_buffer_load_dword s4, s[0:3], 0x0                 ; C2020100
>	s_buffer_load_dword s5, s[0:3], 0x1                 ; C2028101
>	s_buffer_load_dword s6, s[0:3], 0x2                 ; C2030102
>	s_buffer_load_dword s7, s[0:3], 0x3                 ; C2038103
>	s_buffer_load_dword s8, s[0:3], 0x4                 ; C2040104
>	s_buffer_load_dword s9, s[0:3], 0x5                 ; C2048105
>	s_buffer_load_dword s10, s[0:3], 0x6                ; C2050106
>	s_buffer_load_dword s11, s[0:3], 0x7                ; C2058107
>	s_buffer_load_dword s12, s[0:3], 0x8                ; C2060108
>	s_buffer_load_dword s13, s[0:3], 0x9                ; C2068109
>	s_buffer_load_dword s14, s[0:3], 0xa                ; C207010A
>	s_buffer_load_dword s15, s[0:3], 0xb                ; C207810B
>	s_buffer_load_dword s16, s[0:3], 0xc                ; C208010C
>	s_buffer_load_dword s17, s[0:3], 0xd                ; C208810D
>	s_buffer_load_dword s18, s[0:3], 0xe                ; C209010E
>	s_buffer_load_dword s19, s[0:3], 0xf                ; C209810F
>	s_buffer_load_dword s20, s[0:3], 0x10               ; C20A0110
>	s_buffer_load_dword s21, s[0:3], 0x11               ; C20A8111
>	s_buffer_load_dword s22, s[0:3], 0x14               ; C20B0114
>	s_buffer_load_dword s23, s[0:3], 0x15               ; C20B8115
>	s_buffer_load_dword s24, s[0:3], 0x12               ; C20C0112
>	s_buffer_load_dword s25, s[0:3], 0x13               ; C20C8113
>	s_buffer_load_dword s26, s[0:3], 0x16               ; C20D0116
>	s_buffer_load_dword s0, s[0:3], 0x17                ; C2000117
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	v_mov_b32_e32 v4, s4                                ; 7E080204
>	v_mov_b32_e32 v5, s5                                ; 7E0A0205
>	v_mov_b32_e32 v6, s6                                ; 7E0C0206
>	v_mov_b32_e32 v7, s7                                ; 7E0E0207
>	exp 15, 32, 0, 0, 0, v4, v5, v6, v7                 ; F800020F 07060504
>	s_waitcnt expcnt(0)                                 ; BF8C070F
>	v_mov_b32_e32 v4, s8                                ; 7E080208
>	v_mov_b32_e32 v5, s9                                ; 7E0A0209
>	v_mov_b32_e32 v6, s10                               ; 7E0C020A
>	v_mov_b32_e32 v7, s11                               ; 7E0E020B
>	exp 15, 33, 0, 0, 0, v4, v5, v6, v7                 ; F800021F 07060504
>	s_waitcnt vmcnt(0) expcnt(0)                        ; BF8C0700
>	v_mul_f32_e32 v4, s13, v1                           ; 1008020D
>	v_mul_f32_e32 v5, s17, v1                           ; 100A0211
>	v_mul_f32_e32 v6, s21, v1                           ; 100C0215
>	v_mul_f32_e32 v1, s23, v1                           ; 10020217
>	v_mac_f32_e32 v4, s12, v0                           ; 3E08000C
>	v_mac_f32_e32 v5, s16, v0                           ; 3E0A0010
>	v_mac_f32_e32 v6, s20, v0                           ; 3E0C0014
>	v_mac_f32_e32 v1, s22, v0                           ; 3E020016
>	v_mac_f32_e32 v4, s14, v2                           ; 3E08040E
>	v_mac_f32_e32 v5, s18, v2                           ; 3E0A0412
>	v_mac_f32_e32 v6, s24, v2                           ; 3E0C0418
>	v_mac_f32_e32 v1, s26, v2                           ; 3E02041A
>	v_mac_f32_e32 v4, s15, v3                           ; 3E08060F
>	v_mac_f32_e32 v5, s19, v3                           ; 3E0A0613
>	v_mac_f32_e32 v6, s25, v3                           ; 3E0C0619
>	v_mac_f32_e32 v1, s0, v3                            ; 3E020600
>	v_mov_b32_e32 v0, 1.0                               ; 7E0002F2
>	exp 15, 34, 0, 0, 0, v6, v1, v0, v0                 ; F800022F 00000106
>	s_waitcnt expcnt(0)                                 ; BF8C070F
>	v_mov_b32_e32 v1, 0                                 ; 7E020280
>	exp 15, 12, 0, 1, 0, v4, v5, v1, v0                 ; F80008CF 00010504
>	s_endpgm                                            ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 32
>VGPRS: 8
>Code Size: 280 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
>  clamp_color = 0
>FRAG
>PROPERTY FS_COLOR0_WRITES_ALL_CBUFS 1
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL IN[1], GENERIC[1], PERSPECTIVE
>DCL IN[2], GENERIC[2], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SVIEW[0], 2D, FLOAT
>DCL TEMP[0..1], LOCAL
>IMM[0] FLT32 {    1.0000,     0.0000,     0.0000,     0.0000}
>  0: MOV TEMP[0].w, IMM[0].xxxx
>  1: MOV TEMP[0].xyz, IN[1].xyzx
>  2: MOV TEMP[1].xy, IN[2].xyyy
>  3: TEX TEMP[1], TEMP[1], SAMP[0], 2D
>  4: MUL TEMP[0], TEMP[1], TEMP[0]
>  5: MUL TEMP[0], TEMP[0], IN[1].wwww
>  6: MAD TEMP[0], IN[0], TEMP[0].wwww, TEMP[0]
>  7: MOV OUT[0], TEMP[0]
>  8: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %24 = load <8 x i32>, <8 x i32> addrspace(2)* %23, align 32, !tbaa !0
>  %25 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %26 = load <4 x i32>, <4 x i32> addrspace(2)* %25, align 16, !tbaa !0
>  %27 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %8)
>  %28 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %8)
>  %29 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %6, <2 x i32> %8)
>  %30 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %6, <2 x i32> %8)
>  %31 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %6, <2 x i32> %8)
>  %32 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %6, <2 x i32> %8)
>  %33 = call float @llvm.SI.fs.interp(i32 2, i32 1, i32 %6, <2 x i32> %8)
>  %34 = call float @llvm.SI.fs.interp(i32 3, i32 1, i32 %6, <2 x i32> %8)
>  %35 = call float @llvm.SI.fs.interp(i32 0, i32 2, i32 %6, <2 x i32> %8)
>  %36 = call float @llvm.SI.fs.interp(i32 1, i32 2, i32 %6, <2 x i32> %8)
>  %37 = bitcast float %35 to i32
>  %38 = bitcast float %36 to i32
>  %39 = insertelement <2 x i32> undef, i32 %37, i32 0
>  %40 = insertelement <2 x i32> %39, i32 %38, i32 1
>  %41 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %40, <8 x i32> %24, <4 x i32> %26, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %42 = extractelement <4 x float> %41, i32 0
>  %43 = extractelement <4 x float> %41, i32 1
>  %44 = extractelement <4 x float> %41, i32 2
>  %45 = extractelement <4 x float> %41, i32 3
>  %46 = fmul float %42, %31
>  %47 = fmul float %43, %32
>  %48 = fmul float %44, %33
>  %49 = fmul float %46, %34
>  %50 = fmul float %47, %34
>  %51 = fmul float %48, %34
>  %52 = fmul float %45, %34
>  %53 = fmul float %27, %52
>  %54 = fadd float %53, %49
>  %55 = fmul float %28, %52
>  %56 = fadd float %55, %50
>  %57 = fmul float %29, %52
>  %58 = fadd float %57, %51
>  %59 = fmul float %30, %52
>  %60 = fadd float %59, %52
>  %61 = call i32 @llvm.SI.packf16(float %54, float %56)
>  %62 = bitcast i32 %61 to float
>  %63 = call i32 @llvm.SI.packf16(float %58, float %60)
>  %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.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, 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 }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                      ; BEFE0A7E
>	s_mov_b32 m0, s10                                                         ; BEFC030A
>	s_load_dwordx8 s[8:15], s[6:7], 0x0                                       ; C0C40700
>	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_interp_p1_f32 v4, v0, 2, 0, [m0]                                        ; C8100200
>	v_interp_p2_f32 v4, [v4], v1, 2, 0, [m0]                                  ; C8110201
>	v_interp_p1_f32 v5, v0, 3, 0, [m0]                                        ; C8140300
>	v_interp_p2_f32 v5, [v5], v1, 3, 0, [m0]                                  ; C8150301
>	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 v7, v0, 1, 1, [m0]                                        ; C81C0500
>	s_load_dwordx4 s[0:3], s[4:5], 0x0                                        ; C0800500
>	v_interp_p2_f32 v7, [v7], v1, 1, 1, [m0]                                  ; C81D0501
>	v_interp_p1_f32 v8, v0, 2, 1, [m0]                                        ; C8200600
>	v_interp_p2_f32 v8, [v8], v1, 2, 1, [m0]                                  ; C8210601
>	v_interp_p1_f32 v9, v0, 3, 1, [m0]                                        ; C8240700
>	v_interp_p2_f32 v9, [v9], v1, 3, 1, [m0]                                  ; C8250701
>	v_interp_p1_f32 v10, v0, 0, 2, [m0]                                       ; C8280800
>	v_interp_p2_f32 v10, [v10], v1, 0, 2, [m0]                                ; C8290801
>	v_interp_p1_f32 v11, v0, 1, 2, [m0]                                       ; C82C0900
>	v_interp_p2_f32 v11, [v11], v1, 1, 2, [m0]                                ; C82D0901
>	s_waitcnt lgkmcnt(0)                                                      ; BF8C007F
>	image_sample v[10:13], 15, 0, 0, 0, 0, 0, 0, 0, v[10:11], s[8:15], s[0:3] ; F0800F00 00020A0A
>	s_waitcnt vmcnt(0)                                                        ; BF8C0770
>	v_mul_f32_e32 v0, v6, v10                                                 ; 10001506
>	v_mul_f32_e32 v1, v7, v11                                                 ; 10021707
>	v_mul_f32_e32 v6, v8, v12                                                 ; 100C1908
>	v_mul_f32_e32 v0, v9, v0                                                  ; 10000109
>	v_mul_f32_e32 v1, v9, v1                                                  ; 10020309
>	v_mul_f32_e32 v6, v9, v6                                                  ; 100C0D09
>	v_mul_f32_e32 v7, v9, v13                                                 ; 100E1B09
>	v_mac_f32_e32 v0, v7, v2                                                  ; 3E000507
>	v_mac_f32_e32 v1, v7, v3                                                  ; 3E020707
>	v_mac_f32_e32 v6, v7, v4                                                  ; 3E0C0907
>	v_mac_f32_e32 v7, v7, v5                                                  ; 3E0E0B07
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v1                                        ; 5E000300
>	v_cvt_pkrtz_f16_f32_e32 v1, v6, v7                                        ; 5E020F06
>	exp 15, 0, 1, 1, 1, v0, v1, v0, v1                                        ; F8001C0F 01000100
>	s_endpgm                                                                  ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 16
>VGPRS: 16
>Code Size: 176 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
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[0]
>DCL OUT[2], GENERIC[1]
>DCL OUT[3], GENERIC[2]
>DCL OUT[4], GENERIC[3]
>DCL CONST[0..143]
>DCL TEMP[0..3], LOCAL
>DCL ADDR[0]
>IMM[0] FLT32 {    0.0000,     1.0000,  1530.0599,     2.1000}
>IMM[1] FLT32 {    3.1000,     4.1000,     5.1000,     0.1000}
>IMM[2] FLT32 {    1.1000,     0.0000,     0.0000,     0.0000}
>  0: MOV TEMP[0].zw, IMM[0].yyxy
>  1: MAD TEMP[1].x, IMM[0].zzzz, IN[0].zzzz, IMM[0].wwww
>  2: F2I TEMP[1].x, TEMP[1].xxxx
>  3: UARL ADDR[0].x, TEMP[1].xxxx
>  4: UARL ADDR[0].x, TEMP[1].xxxx
>  5: DP4 TEMP[0].x, IN[1], CONST[ADDR[0].x]
>  6: MAD TEMP[1].x, IMM[0].zzzz, IN[0].zzzz, IMM[1].xxxx
>  7: F2I TEMP[1].x, TEMP[1].xxxx
>  8: UARL ADDR[0].x, TEMP[1].xxxx
>  9: DP4 TEMP[1].x, IN[1], CONST[ADDR[0].x]
> 10: MOV TEMP[0].y, TEMP[1].xxxx
> 11: MAD TEMP[1].x, IMM[0].zzzz, IN[0].zzzz, IMM[1].yyyy
> 12: F2I TEMP[1].x, TEMP[1].xxxx
> 13: UARL ADDR[0].x, TEMP[1].xxxx
> 14: UARL ADDR[0].x, TEMP[1].xxxx
> 15: DP4 TEMP[1].x, IN[1], CONST[ADDR[0].x]
> 16: MAD TEMP[2].x, IMM[0].zzzz, IN[0].zzzz, IMM[1].zzzz
> 17: F2I TEMP[2].x, TEMP[2].xxxx
> 18: UARL ADDR[0].x, TEMP[2].xxxx
> 19: DP4 TEMP[2].x, IN[1], CONST[ADDR[0].x]
> 20: MOV TEMP[1].y, TEMP[2].xxxx
> 21: MAD TEMP[2].x, IMM[0].zzzz, IN[0].zzzz, IMM[1].wwww
> 22: F2I TEMP[2].x, TEMP[2].xxxx
> 23: UARL ADDR[0].x, TEMP[2].xxxx
> 24: MOV TEMP[2], CONST[ADDR[0].x]
> 25: MAD TEMP[3].x, IMM[0].zzzz, IN[0].zzzz, IMM[2].xxxx
> 26: F2I TEMP[3].x, TEMP[3].xxxx
> 27: UARL ADDR[0].x, TEMP[3].xxxx
> 28: MOV TEMP[3], CONST[ADDR[0].x]
> 29: MOV TEMP[1].xy, TEMP[1].xyxx
> 30: MOV OUT[1], IN[0]
> 31: MOV OUT[2], TEMP[2]
> 32: MOV OUT[3], TEMP[3]
> 33: MOV OUT[0], TEMP[0]
> 34: MOV OUT[4], TEMP[1]
> 35: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %15 = load <16 x i8>, <16 x i8> addrspace(2)* %14, align 16, !tbaa !0
>  %16 = add i32 %5, %8
>  %17 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %15, i32 0, i32 %16)
>  %18 = extractelement <4 x float> %17, i32 0
>  %19 = extractelement <4 x float> %17, i32 1
>  %20 = extractelement <4 x float> %17, i32 2
>  %21 = extractelement <4 x float> %17, i32 3
>  %22 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 1
>  %23 = load <16 x i8>, <16 x i8> addrspace(2)* %22, align 16, !tbaa !0
>  %24 = add i32 %5, %8
>  %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 = extractelement <4 x float> %25, i32 3
>  %30 = fmul float %20, 0x4097E83D60000000
>  %31 = fadd float %30, 0x4000CCCCC0000000
>  %32 = fptosi float %31 to i32
>  %33 = shl i32 %32, 4
>  %34 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %33)
>  %35 = shl i32 %32, 4
>  %36 = or i32 %35, 4
>  %37 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %36)
>  %38 = shl i32 %32, 4
>  %39 = or i32 %38, 8
>  %40 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %39)
>  %41 = shl i32 %32, 4
>  %42 = or i32 %41, 12
>  %43 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %42)
>  %44 = fmul float %26, %34
>  %45 = fmul float %27, %37
>  %46 = fadd float %44, %45
>  %47 = fmul float %28, %40
>  %48 = fadd float %46, %47
>  %49 = fmul float %29, %43
>  %50 = fadd float %48, %49
>  %51 = fmul float %20, 0x4097E83D60000000
>  %52 = fadd float %51, 0x4008CCCCC0000000
>  %53 = fptosi float %52 to i32
>  %54 = shl i32 %53, 4
>  %55 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %54)
>  %56 = shl i32 %53, 4
>  %57 = or i32 %56, 4
>  %58 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %57)
>  %59 = shl i32 %53, 4
>  %60 = or i32 %59, 8
>  %61 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %60)
>  %62 = shl i32 %53, 4
>  %63 = or i32 %62, 12
>  %64 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %63)
>  %65 = fmul float %26, %55
>  %66 = fmul float %27, %58
>  %67 = fadd float %65, %66
>  %68 = fmul float %28, %61
>  %69 = fadd float %67, %68
>  %70 = fmul float %29, %64
>  %71 = fadd float %69, %70
>  %72 = fmul float %20, 0x4097E83D60000000
>  %73 = fadd float %72, 0x4010666660000000
>  %74 = fptosi float %73 to i32
>  %75 = shl i32 %74, 4
>  %76 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %75)
>  %77 = shl i32 %74, 4
>  %78 = or i32 %77, 4
>  %79 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %78)
>  %80 = shl i32 %74, 4
>  %81 = or i32 %80, 8
>  %82 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %81)
>  %83 = shl i32 %74, 4
>  %84 = or i32 %83, 12
>  %85 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %84)
>  %86 = fmul float %26, %76
>  %87 = fmul float %27, %79
>  %88 = fadd float %86, %87
>  %89 = fmul float %28, %82
>  %90 = fadd float %88, %89
>  %91 = fmul float %29, %85
>  %92 = fadd float %90, %91
>  %93 = fmul float %20, 0x4097E83D60000000
>  %94 = fadd float %93, 0x4014666660000000
>  %95 = fptosi float %94 to i32
>  %96 = shl i32 %95, 4
>  %97 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %96)
>  %98 = shl i32 %95, 4
>  %99 = or i32 %98, 4
>  %100 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %99)
>  %101 = shl i32 %95, 4
>  %102 = or i32 %101, 8
>  %103 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %102)
>  %104 = shl i32 %95, 4
>  %105 = or i32 %104, 12
>  %106 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %105)
>  %107 = fmul float %26, %97
>  %108 = fmul float %27, %100
>  %109 = fadd float %107, %108
>  %110 = fmul float %28, %103
>  %111 = fadd float %109, %110
>  %112 = fmul float %29, %106
>  %113 = fadd float %111, %112
>  %114 = fmul float %20, 0x4097E83D60000000
>  %115 = fadd float %114, 0x3FB99999A0000000
>  %116 = fptosi float %115 to i32
>  %117 = shl i32 %116, 4
>  %118 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %117)
>  %119 = shl i32 %116, 4
>  %120 = or i32 %119, 4
>  %121 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %120)
>  %122 = shl i32 %116, 4
>  %123 = or i32 %122, 8
>  %124 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %123)
>  %125 = shl i32 %116, 4
>  %126 = or i32 %125, 12
>  %127 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %126)
>  %128 = fmul float %20, 0x4097E83D60000000
>  %129 = fadd float %128, 0x3FF19999A0000000
>  %130 = fptosi float %129 to i32
>  %131 = shl i32 %130, 4
>  %132 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %131)
>  %133 = shl i32 %130, 4
>  %134 = or i32 %133, 4
>  %135 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %134)
>  %136 = shl i32 %130, 4
>  %137 = or i32 %136, 8
>  %138 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %137)
>  %139 = shl i32 %130, 4
>  %140 = or i32 %139, 12
>  %141 = call float @llvm.SI.load.const(<16 x i8> %13, i32 %140)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %18, float %19, float %20, float %21)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %118, float %121, float %124, float %127)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 34, i32 0, float %132, float %135, float %138, float %141)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 35, i32 0, float %92, float %113, float undef, float undef)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %50, float %71, 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_load_dwordx4 s[4:7], s[8:9], 0x0                    ; C0820900
>	s_load_dwordx4 s[12:15], s[8:9], 0x4                  ; C0860904
>	v_mov_b32_e32 v1, 0x44bf41eb                          ; 7E0202FF 44BF41EB
>	v_add_i32_e32 v0, vcc, s10, v0                        ; 4A00000A
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	buffer_load_format_xyzw v[2:5], v0, s[4:7], 0 idxen   ; E00C2000 80010200
>	buffer_load_format_xyzw v[6:9], v0, s[12:15], 0 idxen ; E00C2000 80030600
>	s_waitcnt vmcnt(1)                                    ; BF8C0771
>	v_madak_f32_e32 v0, v4, v1, 0x40833333                ; 42000304 40833333
>	v_madak_f32_e32 v10, v4, v1, 0x40a33333               ; 42140304 40A33333
>	v_madak_f32_e32 v11, v4, v1, 0x40066666               ; 42160304 40066666
>	v_madak_f32_e32 v12, v4, v1, 0x40466666               ; 42180304 40466666
>	v_cvt_i32_f32_e32 v0, v0                              ; 7E001100
>	v_cvt_i32_f32_e32 v10, v10                            ; 7E14110A
>	v_madak_f32_e32 v13, v4, v1, 0x3dcccccd               ; 421A0304 3DCCCCCD
>	v_madak_f32_e32 v1, v4, v1, 0x3f8ccccd                ; 42020304 3F8CCCCD
>	v_cvt_i32_f32_e32 v11, v11                            ; 7E16110B
>	v_cvt_i32_f32_e32 v12, v12                            ; 7E18110C
>	v_cvt_i32_f32_e32 v13, v13                            ; 7E1A110D
>	v_cvt_i32_f32_e32 v1, v1                              ; 7E021101
>	v_lshlrev_b32_e32 v11, 4, v11                         ; 34161684
>	v_lshlrev_b32_e32 v12, 4, v12                         ; 34181884
>	v_lshlrev_b32_e32 v0, 4, v0                           ; 34000084
>	v_lshlrev_b32_e32 v10, 4, v10                         ; 34141484
>	v_lshlrev_b32_e32 v13, 4, v13                         ; 341A1A84
>	v_lshlrev_b32_e32 v1, 4, v1                           ; 34020284
>	buffer_load_dword v14, v11, s[0:3], 0 offen           ; E0301000 80000E0B
>	v_or_b32_e32 v15, 4, v11                              ; 381E1684
>	v_or_b32_e32 v16, 8, v11                              ; 38201688
>	v_or_b32_e32 v11, 12, v11                             ; 3816168C
>	buffer_load_dword v17, v12, s[0:3], 0 offen           ; E0301000 8000110C
>	v_or_b32_e32 v18, 4, v12                              ; 38241884
>	v_or_b32_e32 v19, 8, v12                              ; 38261888
>	v_or_b32_e32 v12, 12, v12                             ; 3818188C
>	buffer_load_dword v20, v13, s[0:3], 0 offen           ; E0301000 8000140D
>	v_or_b32_e32 v21, 4, v13                              ; 382A1A84
>	v_or_b32_e32 v22, 8, v13                              ; 382C1A88
>	v_or_b32_e32 v13, 12, v13                             ; 381A1A8C
>	buffer_load_dword v23, v1, s[0:3], 0 offen            ; E0301000 80001701
>	v_or_b32_e32 v24, 4, v1                               ; 38300284
>	v_or_b32_e32 v25, 8, v1                               ; 38320288
>	v_or_b32_e32 v1, 12, v1                               ; 3802028C
>	buffer_load_dword v15, v15, s[0:3], 0 offen           ; E0301000 80000F0F
>	buffer_load_dword v21, v21, s[0:3], 0 offen           ; E0301000 80001515
>	buffer_load_dword v22, v22, s[0:3], 0 offen           ; E0301000 80001616
>	buffer_load_dword v13, v13, s[0:3], 0 offen           ; E0301000 80000D0D
>	buffer_load_dword v24, v24, s[0:3], 0 offen           ; E0301000 80001818
>	buffer_load_dword v25, v25, s[0:3], 0 offen           ; E0301000 80001919
>	buffer_load_dword v1, v1, s[0:3], 0 offen             ; E0301000 80000101
>	buffer_load_dword v26, v0, s[0:3], 0 offen            ; E0301000 80001A00
>	v_or_b32_e32 v27, 4, v0                               ; 38360084
>	v_or_b32_e32 v28, 8, v0                               ; 38380088
>	v_or_b32_e32 v0, 12, v0                               ; 3800008C
>	buffer_load_dword v18, v18, s[0:3], 0 offen           ; E0301000 80001212
>	buffer_load_dword v16, v16, s[0:3], 0 offen           ; E0301000 80001010
>	buffer_load_dword v27, v27, s[0:3], 0 offen           ; E0301000 80001B1B
>	v_or_b32_e32 v29, 4, v10                              ; 383A1484
>	buffer_load_dword v29, v29, s[0:3], 0 offen           ; E0301000 80001D1D
>	buffer_load_dword v30, v10, s[0:3], 0 offen           ; E0301000 80001E0A
>	v_or_b32_e32 v31, 8, v10                              ; 383E1488
>	v_or_b32_e32 v10, 12, v10                             ; 3814148C
>	buffer_load_dword v19, v19, s[0:3], 0 offen           ; E0301000 80001313
>	buffer_load_dword v28, v28, s[0:3], 0 offen           ; E0301000 80001C1C
>	buffer_load_dword v31, v31, s[0:3], 0 offen           ; E0301000 80001F1F
>	buffer_load_dword v11, v11, s[0:3], 0 offen           ; E0301000 80000B0B
>	buffer_load_dword v12, v12, s[0:3], 0 offen           ; E0301000 80000C0C
>	buffer_load_dword v0, v0, s[0:3], 0 offen             ; E0301000 80000000
>	buffer_load_dword v10, v10, s[0:3], 0 offen           ; E0301000 80000A0A
>	exp 15, 32, 0, 0, 0, v2, v3, v4, v5                   ; F800020F 05040302
>	s_waitcnt                                             ; BF8C077F
>	exp 15, 33, 0, 0, 0, v20, v21, v22, v13               ; F800021F 0D161514
>	s_waitcnt vmcnt(13)                                   ; BF8C077D
>	exp 15, 34, 0, 0, 0, v23, v24, v25, v1                ; F800022F 01191817
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mul_f32_e32 v1, v15, v7                             ; 10020F0F
>	v_mac_f32_e32 v1, v14, v6                             ; 3E020D0E
>	s_waitcnt vmcnt(11)                                   ; BF8C077B
>	v_mul_f32_e32 v2, v18, v7                             ; 10040F12
>	v_mac_f32_e32 v2, v17, v6                             ; 3E040D11
>	s_waitcnt vmcnt(9)                                    ; BF8C0779
>	v_mul_f32_e32 v3, v27, v7                             ; 10060F1B
>	v_mac_f32_e32 v3, v26, v6                             ; 3E060D1A
>	s_waitcnt vmcnt(8)                                    ; BF8C0778
>	v_mul_f32_e32 v4, v29, v7                             ; 10080F1D
>	s_waitcnt vmcnt(7)                                    ; BF8C0777
>	v_mac_f32_e32 v4, v30, v6                             ; 3E080D1E
>	v_mac_f32_e32 v1, v16, v8                             ; 3E021110
>	v_mov_b32_e32 v5, 1.0                                 ; 7E0A02F2
>	s_waitcnt vmcnt(6)                                    ; BF8C0776
>	v_mac_f32_e32 v2, v19, v8                             ; 3E041113
>	s_waitcnt vmcnt(5)                                    ; BF8C0775
>	v_mac_f32_e32 v3, v28, v8                             ; 3E06111C
>	s_waitcnt vmcnt(4)                                    ; BF8C0774
>	v_mac_f32_e32 v4, v31, v8                             ; 3E08111F
>	s_waitcnt vmcnt(3)                                    ; BF8C0773
>	v_mac_f32_e32 v1, v11, v9                             ; 3E02130B
>	s_waitcnt vmcnt(2)                                    ; BF8C0772
>	v_mac_f32_e32 v2, v12, v9                             ; 3E04130C
>	s_waitcnt vmcnt(1)                                    ; BF8C0771
>	v_mac_f32_e32 v3, v0, v9                              ; 3E061300
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	v_mac_f32_e32 v4, v10, v9                             ; 3E08130A
>	exp 15, 35, 0, 0, 0, v3, v4, v0, v0                   ; F800023F 00000403
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mov_b32_e32 v0, 0                                   ; 7E000280
>	exp 15, 12, 0, 1, 0, v1, v2, v0, v5                   ; F80008CF 05000201
>	s_endpgm                                              ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 24
>VGPRS: 32
>Code Size: 584 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
>  clamp_color = 0
>FRAG
>PROPERTY FS_COLOR0_WRITES_ALL_CBUFS 1
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL IN[1], GENERIC[1], PERSPECTIVE
>DCL IN[2], GENERIC[2], PERSPECTIVE
>DCL IN[3], GENERIC[3], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SVIEW[0], 2D, FLOAT
>DCL TEMP[0..1], LOCAL
>  0: MOV TEMP[0].xy, IN[3].xyyy
>  1: TEX TEMP[0], TEMP[0], SAMP[0], 2D
>  2: MAD TEMP[0], TEMP[0], IN[2], IN[1]
>  3: MUL TEMP[1].x, TEMP[0].wwww, IN[0].wwww
>  4: MOV TEMP[0].w, TEMP[1].xxxx
>  5: MOV OUT[0], TEMP[0]
>  6: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %24 = load <8 x i32>, <8 x i32> addrspace(2)* %23, align 32, !tbaa !0
>  %25 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %26 = load <4 x i32>, <4 x i32> addrspace(2)* %25, align 16, !tbaa !0
>  %27 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %6, <2 x i32> %8)
>  %28 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %6, <2 x i32> %8)
>  %29 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %6, <2 x i32> %8)
>  %30 = call float @llvm.SI.fs.interp(i32 2, i32 1, i32 %6, <2 x i32> %8)
>  %31 = call float @llvm.SI.fs.interp(i32 3, i32 1, i32 %6, <2 x i32> %8)
>  %32 = call float @llvm.SI.fs.interp(i32 0, i32 2, i32 %6, <2 x i32> %8)
>  %33 = call float @llvm.SI.fs.interp(i32 1, i32 2, i32 %6, <2 x i32> %8)
>  %34 = call float @llvm.SI.fs.interp(i32 2, i32 2, i32 %6, <2 x i32> %8)
>  %35 = call float @llvm.SI.fs.interp(i32 3, i32 2, i32 %6, <2 x i32> %8)
>  %36 = call float @llvm.SI.fs.interp(i32 0, i32 3, i32 %6, <2 x i32> %8)
>  %37 = call float @llvm.SI.fs.interp(i32 1, i32 3, i32 %6, <2 x i32> %8)
>  %38 = bitcast float %36 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 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %41, <8 x i32> %24, <4 x i32> %26, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %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 = fmul float %43, %32
>  %48 = fadd float %47, %28
>  %49 = fmul float %44, %33
>  %50 = fadd float %49, %29
>  %51 = fmul float %45, %34
>  %52 = fadd float %51, %30
>  %53 = fmul float %46, %35
>  %54 = fadd float %53, %31
>  %55 = fmul float %54, %27
>  %56 = call i32 @llvm.SI.packf16(float %48, float %50)
>  %57 = bitcast i32 %56 to float
>  %58 = call i32 @llvm.SI.packf16(float %52, float %55)
>  %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.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, 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 }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                      ; BEFE0A7E
>	s_mov_b32 m0, s10                                                         ; BEFC030A
>	s_load_dwordx8 s[8:15], s[6:7], 0x0                                       ; C0C40700
>	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, 0, 1, [m0]                                        ; C80C0400
>	v_interp_p2_f32 v3, [v3], v1, 0, 1, [m0]                                  ; C80D0401
>	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 v5, v0, 2, 1, [m0]                                        ; C8140600
>	v_interp_p2_f32 v5, [v5], v1, 2, 1, [m0]                                  ; C8150601
>	v_interp_p1_f32 v6, v0, 3, 1, [m0]                                        ; C8180700
>	v_interp_p2_f32 v6, [v6], v1, 3, 1, [m0]                                  ; C8190701
>	v_interp_p1_f32 v7, v0, 0, 2, [m0]                                        ; C81C0800
>	v_interp_p2_f32 v7, [v7], v1, 0, 2, [m0]                                  ; C81D0801
>	v_interp_p1_f32 v8, v0, 1, 2, [m0]                                        ; C8200900
>	s_load_dwordx4 s[0:3], s[4:5], 0x0                                        ; C0800500
>	v_interp_p2_f32 v8, [v8], v1, 1, 2, [m0]                                  ; C8210901
>	v_interp_p1_f32 v9, v0, 2, 2, [m0]                                        ; C8240A00
>	v_interp_p2_f32 v9, [v9], v1, 2, 2, [m0]                                  ; C8250A01
>	v_interp_p1_f32 v10, v0, 3, 2, [m0]                                       ; C8280B00
>	v_interp_p2_f32 v10, [v10], v1, 3, 2, [m0]                                ; C8290B01
>	v_interp_p1_f32 v11, v0, 0, 3, [m0]                                       ; C82C0C00
>	v_interp_p2_f32 v11, [v11], v1, 0, 3, [m0]                                ; C82D0C01
>	v_interp_p1_f32 v12, v0, 1, 3, [m0]                                       ; C8300D00
>	v_interp_p2_f32 v12, [v12], v1, 1, 3, [m0]                                ; C8310D01
>	s_waitcnt lgkmcnt(0)                                                      ; BF8C007F
>	image_sample v[11:14], 15, 0, 0, 0, 0, 0, 0, 0, v[11:12], s[8:15], s[0:3] ; F0800F00 00020B0B
>	s_waitcnt vmcnt(0)                                                        ; BF8C0770
>	v_mac_f32_e32 v3, v7, v11                                                 ; 3E061707
>	v_mac_f32_e32 v4, v8, v12                                                 ; 3E081908
>	v_mac_f32_e32 v5, v9, v13                                                 ; 3E0A1B09
>	v_mac_f32_e32 v6, v10, v14                                                ; 3E0C1D0A
>	v_mul_f32_e32 v0, v2, v6                                                  ; 10000D02
>	v_cvt_pkrtz_f16_f32_e32 v1, v3, v4                                        ; 5E020903
>	v_cvt_pkrtz_f16_f32_e32 v0, v5, v0                                        ; 5E000105
>	exp 15, 0, 1, 1, 1, v1, v0, v1, v0                                        ; F8001C0F 00010001
>	s_endpgm                                                                  ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 16
>VGPRS: 16
>Code Size: 160 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
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL IN[2]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[0]
>DCL OUT[2], GENERIC[1]
>DCL OUT[3], GENERIC[2]
>DCL CONST[0..3]
>DCL TEMP[0..2], LOCAL
>IMM[0] FLT32 {    0.0000,     1.0000,     0.0000,     0.0000}
>  0: MOV TEMP[0].zw, IMM[0].yyxy
>  1: DP4 TEMP[0].x, IN[2], CONST[0]
>  2: DP4 TEMP[1].x, IN[2], CONST[1]
>  3: MOV TEMP[0].y, TEMP[1].xxxx
>  4: DP4 TEMP[1].x, IN[2], CONST[2]
>  5: DP4 TEMP[2].x, IN[2], CONST[3]
>  6: MOV TEMP[1].y, TEMP[2].xxxx
>  7: MOV TEMP[1].xy, TEMP[1].xyxx
>  8: MOV OUT[2], IN[1]
>  9: MOV OUT[1], IN[0]
> 10: MOV OUT[0], TEMP[0]
> 11: MOV OUT[3], TEMP[1]
> 12: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = call float @llvm.SI.load.const(<16 x i8> %13, i32 0)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %13, i32 4)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %13, i32 8)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %13, i32 12)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %13, i32 16)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %13, i32 20)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %13, i32 24)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %13, i32 28)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %13, i32 32)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %13, i32 36)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %13, i32 40)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %13, i32 44)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %13, i32 48)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %13, i32 52)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %13, i32 56)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %13, i32 60)
>  %30 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %31 = load <16 x i8>, <16 x i8> addrspace(2)* %30, align 16, !tbaa !0
>  %32 = add i32 %5, %8
>  %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 = extractelement <4 x float> %33, i32 2
>  %37 = extractelement <4 x float> %33, i32 3
>  %38 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 1
>  %39 = load <16 x i8>, <16 x i8> addrspace(2)* %38, align 16, !tbaa !0
>  %40 = add i32 %5, %8
>  %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 = extractelement <4 x float> %41, i32 2
>  %45 = extractelement <4 x float> %41, i32 3
>  %46 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 2
>  %47 = load <16 x i8>, <16 x i8> addrspace(2)* %46, align 16, !tbaa !0
>  %48 = add i32 %5, %8
>  %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 = extractelement <4 x float> %49, i32 3
>  %54 = fmul float %50, %14
>  %55 = fmul float %51, %15
>  %56 = fadd float %54, %55
>  %57 = fmul float %52, %16
>  %58 = fadd float %56, %57
>  %59 = fmul float %53, %17
>  %60 = fadd float %58, %59
>  %61 = fmul float %50, %18
>  %62 = fmul float %51, %19
>  %63 = fadd float %61, %62
>  %64 = fmul float %52, %20
>  %65 = fadd float %63, %64
>  %66 = fmul float %53, %21
>  %67 = fadd float %65, %66
>  %68 = fmul float %50, %22
>  %69 = fmul float %51, %23
>  %70 = fadd float %68, %69
>  %71 = fmul float %52, %24
>  %72 = fadd float %70, %71
>  %73 = fmul float %53, %25
>  %74 = fadd float %72, %73
>  %75 = fmul float %50, %26
>  %76 = fmul float %51, %27
>  %77 = fadd float %75, %76
>  %78 = fmul float %52, %28
>  %79 = fadd float %77, %78
>  %80 = fmul float %53, %29
>  %81 = fadd float %79, %80
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %34, float %35, float %36, float %37)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %42, float %43, float %44, float %45)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 34, i32 0, float %74, float %81, float undef, float undef)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %60, float %67, 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_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
>	v_add_i32_e32 v0, vcc, s10, v0                          ; 4A00000A
>	v_mov_b32_e32 v1, 1.0                                   ; 7E0202F2
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	s_buffer_load_dword s8, s[0:3], 0x5                     ; C2040105
>	s_buffer_load_dword s9, s[0:3], 0x6                     ; C2048106
>	s_buffer_load_dword s10, s[0:3], 0x7                    ; C2050107
>	s_buffer_load_dword s11, s[0:3], 0x8                    ; C2058108
>	s_buffer_load_dword s20, s[0:3], 0x9                    ; C20A0109
>	buffer_load_format_xyzw v[2:5], v0, s[4:7], 0 idxen     ; E00C2000 80010200
>	buffer_load_format_xyzw v[6:9], v0, s[12:15], 0 idxen   ; E00C2000 80030600
>	buffer_load_format_xyzw v[10:13], v0, s[16:19], 0 idxen ; E00C2000 80040A00
>	s_buffer_load_dword s4, s[0:3], 0xa                     ; C202010A
>	s_buffer_load_dword s5, s[0:3], 0xb                     ; C202810B
>	s_buffer_load_dword s6, s[0:3], 0xc                     ; C203010C
>	s_buffer_load_dword s7, s[0:3], 0xd                     ; C203810D
>	s_buffer_load_dword s12, s[0:3], 0xe                    ; C206010E
>	s_buffer_load_dword s13, s[0:3], 0x0                    ; C2068100
>	s_buffer_load_dword s14, s[0:3], 0x1                    ; C2070101
>	s_buffer_load_dword s15, s[0:3], 0x2                    ; C2078102
>	s_buffer_load_dword s16, s[0:3], 0x3                    ; C2080103
>	s_buffer_load_dword s17, s[0:3], 0x4                    ; C2088104
>	s_buffer_load_dword s0, s[0:3], 0xf                     ; C200010F
>	s_waitcnt vmcnt(2)                                      ; BF8C0772
>	exp 15, 32, 0, 0, 0, v2, v3, v4, v5                     ; F800020F 05040302
>	s_waitcnt vmcnt(1)                                      ; BF8C0771
>	exp 15, 33, 0, 0, 0, v6, v7, v8, v9                     ; F800021F 09080706
>	s_waitcnt vmcnt(0) lgkmcnt(0)                           ; BF8C0070
>	v_mul_f32_e32 v0, s14, v11                              ; 1000160E
>	s_waitcnt expcnt(0)                                     ; BF8C070F
>	v_mul_f32_e32 v2, s8, v11                               ; 10041608
>	v_mul_f32_e32 v3, s20, v11                              ; 10061614
>	v_mul_f32_e32 v4, s7, v11                               ; 10081607
>	v_mac_f32_e32 v0, s13, v10                              ; 3E00140D
>	v_mac_f32_e32 v2, s17, v10                              ; 3E041411
>	v_mac_f32_e32 v3, s11, v10                              ; 3E06140B
>	v_mac_f32_e32 v4, s6, v10                               ; 3E081406
>	v_mac_f32_e32 v0, s15, v12                              ; 3E00180F
>	v_mac_f32_e32 v2, s9, v12                               ; 3E041809
>	v_mac_f32_e32 v3, s4, v12                               ; 3E061804
>	v_mac_f32_e32 v4, s12, v12                              ; 3E08180C
>	v_mac_f32_e32 v0, s16, v13                              ; 3E001A10
>	v_mac_f32_e32 v2, s10, v13                              ; 3E041A0A
>	v_mac_f32_e32 v3, s5, v13                               ; 3E061A05
>	v_mac_f32_e32 v4, s0, v13                               ; 3E081A00
>	exp 15, 34, 0, 0, 0, v3, v4, v0, v0                     ; F800022F 00000403
>	s_waitcnt expcnt(0)                                     ; BF8C070F
>	v_mov_b32_e32 v3, 0                                     ; 7E060280
>	exp 15, 12, 0, 1, 0, v0, v2, v3, v1                     ; F80008CF 01030200
>	s_endpgm                                                ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 24
>VGPRS: 16
>Code Size: 240 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
>  clamp_color = 0
>FRAG
>PROPERTY FS_COLOR0_WRITES_ALL_CBUFS 1
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL IN[1], GENERIC[1], PERSPECTIVE
>DCL IN[2], GENERIC[2], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SVIEW[0], 2D, FLOAT
>DCL TEMP[0..1], LOCAL
>  0: MOV TEMP[0].xy, IN[2].xyyy
>  1: TEX TEMP[0], TEMP[0], SAMP[0], 2D
>  2: LRP TEMP[0], IN[1].xxxx, TEMP[0], IN[0]
>  3: MOV TEMP[1].xyz, TEMP[0].xyzx
>  4: MUL TEMP[0].x, TEMP[0].wwww, IN[1].wwww
>  5: MOV TEMP[1].w, TEMP[0].xxxx
>  6: MOV OUT[0], TEMP[1]
>  7: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %24 = load <8 x i32>, <8 x i32> addrspace(2)* %23, align 32, !tbaa !0
>  %25 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %26 = load <4 x i32>, <4 x i32> addrspace(2)* %25, align 16, !tbaa !0
>  %27 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %8)
>  %28 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %8)
>  %29 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %6, <2 x i32> %8)
>  %30 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %6, <2 x i32> %8)
>  %31 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %6, <2 x i32> %8)
>  %32 = call float @llvm.SI.fs.interp(i32 3, i32 1, i32 %6, <2 x i32> %8)
>  %33 = call float @llvm.SI.fs.interp(i32 0, i32 2, i32 %6, <2 x i32> %8)
>  %34 = call float @llvm.SI.fs.interp(i32 1, i32 2, i32 %6, <2 x i32> %8)
>  %35 = bitcast float %33 to i32
>  %36 = bitcast float %34 to i32
>  %37 = insertelement <2 x i32> undef, i32 %35, i32 0
>  %38 = insertelement <2 x i32> %37, i32 %36, i32 1
>  %39 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %38, <8 x i32> %24, <4 x i32> %26, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %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 = extractelement <4 x float> %39, i32 3
>  %44 = fsub float 1.000000e+00, %31
>  %45 = fmul float %40, %31
>  %46 = fmul float %27, %44
>  %47 = fadd float %45, %46
>  %48 = fsub float 1.000000e+00, %31
>  %49 = fmul float %41, %31
>  %50 = fmul float %28, %48
>  %51 = fadd float %49, %50
>  %52 = fsub float 1.000000e+00, %31
>  %53 = fmul float %42, %31
>  %54 = fmul float %29, %52
>  %55 = fadd float %53, %54
>  %56 = fsub float 1.000000e+00, %31
>  %57 = fmul float %43, %31
>  %58 = fmul float %30, %56
>  %59 = fadd float %57, %58
>  %60 = fmul float %59, %32
>  %61 = call i32 @llvm.SI.packf16(float %47, float %51)
>  %62 = bitcast i32 %61 to float
>  %63 = call i32 @llvm.SI.packf16(float %55, float %60)
>  %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.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, 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 }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                   ; BEFE0A7E
>	s_mov_b32 m0, s10                                                      ; BEFC030A
>	s_load_dwordx8 s[8:15], s[6:7], 0x0                                    ; C0C40700
>	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_interp_p1_f32 v4, v0, 2, 0, [m0]                                     ; C8100200
>	v_interp_p2_f32 v4, [v4], v1, 2, 0, [m0]                               ; C8110201
>	v_interp_p1_f32 v5, v0, 3, 0, [m0]                                     ; C8140300
>	s_load_dwordx4 s[0:3], s[4:5], 0x0                                     ; C0800500
>	v_interp_p2_f32 v5, [v5], v1, 3, 0, [m0]                               ; C8150301
>	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 v7, v0, 3, 1, [m0]                                     ; C81C0700
>	v_interp_p2_f32 v7, [v7], v1, 3, 1, [m0]                               ; C81D0701
>	v_interp_p1_f32 v8, v0, 0, 2, [m0]                                     ; C8200800
>	v_interp_p2_f32 v8, [v8], v1, 0, 2, [m0]                               ; C8210801
>	v_interp_p1_f32 v9, v0, 1, 2, [m0]                                     ; C8240900
>	v_interp_p2_f32 v9, [v9], v1, 1, 2, [m0]                               ; C8250901
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	image_sample v[8:11], 15, 0, 0, 0, 0, 0, 0, 0, v[8:9], s[8:15], s[0:3] ; F0800F00 00020808
>	v_mad_f32 v0, -v6, v2, v2                                              ; D2820000 240A0506
>	s_waitcnt vmcnt(0)                                                     ; BF8C0770
>	v_mac_f32_e32 v0, v6, v8                                               ; 3E001106
>	v_mad_f32 v1, -v6, v3, v3                                              ; D2820001 240E0706
>	v_mac_f32_e32 v1, v6, v9                                               ; 3E021306
>	v_mad_f32 v2, -v6, v4, v4                                              ; D2820002 24120906
>	v_mac_f32_e32 v2, v6, v10                                              ; 3E041506
>	v_mad_f32 v3, -v6, v5, v5                                              ; D2820003 24160B06
>	v_mac_f32_e32 v3, v6, v11                                              ; 3E061706
>	v_mul_f32_e32 v3, v7, v3                                               ; 10060707
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v1                                     ; 5E000300
>	v_cvt_pkrtz_f16_f32_e32 v1, v2, v3                                     ; 5E020702
>	exp 15, 0, 1, 1, 1, v0, v1, v0, v1                                     ; F8001C0F 01000100
>	s_endpgm                                                               ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 16
>VGPRS: 12
>Code Size: 168 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
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL IN[2]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[0]
>DCL OUT[2], GENERIC[1]
>DCL OUT[3], GENERIC[2]
>DCL OUT[4], GENERIC[3]
>DCL OUT[5], GENERIC[4]
>DCL CONST[0..5]
>DCL TEMP[0..2], LOCAL
>IMM[0] FLT32 {    0.0000,     1.0000,     0.0000,     0.0000}
>  0: MOV TEMP[0].zw, IMM[0].yyxy
>  1: DP4 TEMP[0].x, IN[2], CONST[2]
>  2: DP4 TEMP[1].x, IN[2], CONST[3]
>  3: MOV TEMP[0].y, TEMP[1].xxxx
>  4: DP4 TEMP[1].x, IN[2], CONST[4]
>  5: DP4 TEMP[2].x, IN[2], CONST[5]
>  6: MOV TEMP[1].y, TEMP[2].xxxx
>  7: MOV TEMP[1].xy, TEMP[1].xyxx
>  8: MOV OUT[2], IN[1]
>  9: MOV OUT[3], CONST[0]
> 10: MOV OUT[1], IN[0]
> 11: MOV OUT[4], CONST[1]
> 12: MOV OUT[0], TEMP[0]
> 13: MOV OUT[5], TEMP[1]
> 14: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = call float @llvm.SI.load.const(<16 x i8> %13, i32 0)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %13, i32 4)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %13, i32 8)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %13, i32 12)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %13, i32 16)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %13, i32 20)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %13, i32 24)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %13, i32 28)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %13, i32 32)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %13, i32 36)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %13, i32 40)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %13, i32 44)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %13, i32 48)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %13, i32 52)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %13, i32 56)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %13, i32 60)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %13, i32 64)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %13, i32 68)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %13, i32 72)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %13, i32 76)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %13, i32 80)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %13, i32 84)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %13, i32 88)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %13, i32 92)
>  %38 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %39 = load <16 x i8>, <16 x i8> addrspace(2)* %38, align 16, !tbaa !0
>  %40 = add i32 %5, %8
>  %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 = extractelement <4 x float> %41, i32 2
>  %45 = extractelement <4 x float> %41, i32 3
>  %46 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 1
>  %47 = load <16 x i8>, <16 x i8> addrspace(2)* %46, align 16, !tbaa !0
>  %48 = add i32 %5, %8
>  %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 = extractelement <4 x float> %49, i32 3
>  %54 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 2
>  %55 = load <16 x i8>, <16 x i8> addrspace(2)* %54, align 16, !tbaa !0
>  %56 = add i32 %5, %8
>  %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 = extractelement <4 x float> %57, i32 1
>  %60 = extractelement <4 x float> %57, i32 2
>  %61 = extractelement <4 x float> %57, i32 3
>  %62 = fmul float %58, %22
>  %63 = fmul float %59, %23
>  %64 = fadd float %62, %63
>  %65 = fmul float %60, %24
>  %66 = fadd float %64, %65
>  %67 = fmul float %61, %25
>  %68 = fadd float %66, %67
>  %69 = fmul float %58, %26
>  %70 = fmul float %59, %27
>  %71 = fadd float %69, %70
>  %72 = fmul float %60, %28
>  %73 = fadd float %71, %72
>  %74 = fmul float %61, %29
>  %75 = fadd float %73, %74
>  %76 = fmul float %58, %30
>  %77 = fmul float %59, %31
>  %78 = fadd float %76, %77
>  %79 = fmul float %60, %32
>  %80 = fadd float %78, %79
>  %81 = fmul float %61, %33
>  %82 = fadd float %80, %81
>  %83 = fmul float %58, %34
>  %84 = fmul float %59, %35
>  %85 = fadd float %83, %84
>  %86 = fmul float %60, %36
>  %87 = fadd float %85, %86
>  %88 = fmul float %61, %37
>  %89 = fadd float %87, %88
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %42, float %43, float %44, float %45)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %50, float %51, float %52, float %53)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 34, i32 0, float %14, float %15, float %16, float %17)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 35, i32 0, float %18, float %19, float %20, float %21)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 36, i32 0, float %82, float %89, float undef, float undef)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %68, float %75, 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_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
>	v_add_i32_e32 v0, vcc, s10, v0                          ; 4A00000A
>	v_mov_b32_e32 v1, 1.0                                   ; 7E0202F2
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	s_buffer_load_dword s8, s[0:3], 0xf                     ; C204010F
>	s_buffer_load_dword s9, s[0:3], 0x10                    ; C2048110
>	s_buffer_load_dword s10, s[0:3], 0x11                   ; C2050111
>	s_buffer_load_dword s11, s[0:3], 0x12                   ; C2058112
>	s_buffer_load_dword s20, s[0:3], 0x13                   ; C20A0113
>	buffer_load_format_xyzw v[2:5], v0, s[4:7], 0 idxen     ; E00C2000 80010200
>	buffer_load_format_xyzw v[6:9], v0, s[12:15], 0 idxen   ; E00C2000 80030600
>	buffer_load_format_xyzw v[10:13], v0, s[16:19], 0 idxen ; E00C2000 80040A00
>	s_buffer_load_dword s4, s[0:3], 0x14                    ; C2020114
>	s_buffer_load_dword s5, s[0:3], 0x15                    ; C2028115
>	s_buffer_load_dword s6, s[0:3], 0x16                    ; C2030116
>	s_buffer_load_dword s7, s[0:3], 0x17                    ; C2038117
>	s_buffer_load_dword s12, s[0:3], 0x5                    ; C2060105
>	s_buffer_load_dword s13, s[0:3], 0x6                    ; C2068106
>	s_buffer_load_dword s14, s[0:3], 0x7                    ; C2070107
>	s_buffer_load_dword s15, s[0:3], 0x8                    ; C2078108
>	s_buffer_load_dword s16, s[0:3], 0x9                    ; C2080109
>	s_buffer_load_dword s17, s[0:3], 0xa                    ; C208810A
>	s_buffer_load_dword s18, s[0:3], 0xb                    ; C209010B
>	s_buffer_load_dword s19, s[0:3], 0xc                    ; C209810C
>	s_buffer_load_dword s21, s[0:3], 0xd                    ; C20A810D
>	s_buffer_load_dword s22, s[0:3], 0xe                    ; C20B010E
>	s_buffer_load_dword s23, s[0:3], 0x0                    ; C20B8100
>	s_buffer_load_dword s24, s[0:3], 0x1                    ; C20C0101
>	s_buffer_load_dword s25, s[0:3], 0x2                    ; C20C8102
>	s_buffer_load_dword s26, s[0:3], 0x3                    ; C20D0103
>	s_buffer_load_dword s0, s[0:3], 0x4                     ; C2000104
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	v_mov_b32_e32 v0, s12                                   ; 7E00020C
>	v_mov_b32_e32 v14, s23                                  ; 7E1C0217
>	v_mov_b32_e32 v15, s24                                  ; 7E1E0218
>	v_mov_b32_e32 v16, s25                                  ; 7E200219
>	v_mov_b32_e32 v17, s26                                  ; 7E22021A
>	v_mov_b32_e32 v18, s0                                   ; 7E240200
>	s_waitcnt vmcnt(2)                                      ; BF8C0772
>	exp 15, 32, 0, 0, 0, v2, v3, v4, v5                     ; F800020F 05040302
>	s_waitcnt expcnt(0)                                     ; BF8C070F
>	v_mov_b32_e32 v2, s13                                   ; 7E04020D
>	s_waitcnt vmcnt(1)                                      ; BF8C0771
>	exp 15, 33, 0, 0, 0, v6, v7, v8, v9                     ; F800021F 09080706
>	v_mov_b32_e32 v3, s14                                   ; 7E06020E
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	v_mul_f32_e32 v4, s16, v11                              ; 10081610
>	v_mul_f32_e32 v5, s21, v11                              ; 100A1615
>	s_waitcnt expcnt(0)                                     ; BF8C070F
>	v_mul_f32_e32 v6, s10, v11                              ; 100C160A
>	exp 15, 34, 0, 0, 0, v14, v15, v16, v17                 ; F800022F 11100F0E
>	v_mul_f32_e32 v7, s5, v11                               ; 100E1605
>	v_mac_f32_e32 v4, s15, v10                              ; 3E08140F
>	v_mac_f32_e32 v5, s19, v10                              ; 3E0A1413
>	v_mac_f32_e32 v6, s9, v10                               ; 3E0C1409
>	v_mac_f32_e32 v7, s4, v10                               ; 3E0E1404
>	v_mac_f32_e32 v4, s17, v12                              ; 3E081811
>	v_mac_f32_e32 v5, s22, v12                              ; 3E0A1816
>	v_mac_f32_e32 v6, s11, v12                              ; 3E0C180B
>	v_mac_f32_e32 v7, s6, v12                               ; 3E0E1806
>	v_mac_f32_e32 v4, s18, v13                              ; 3E081A12
>	v_mac_f32_e32 v5, s8, v13                               ; 3E0A1A08
>	v_mac_f32_e32 v6, s20, v13                              ; 3E0C1A14
>	v_mac_f32_e32 v7, s7, v13                               ; 3E0E1A07
>	exp 15, 35, 0, 0, 0, v18, v0, v2, v3                    ; F800023F 03020012
>	exp 15, 36, 0, 0, 0, v6, v7, v0, v0                     ; F800024F 00000706
>	s_waitcnt expcnt(0)                                     ; BF8C070F
>	v_mov_b32_e32 v0, 0                                     ; 7E000280
>	exp 15, 12, 0, 1, 0, v4, v5, v0, v1                     ; F80008CF 01000504
>	s_endpgm                                                ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 32
>VGPRS: 20
>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
>  clamp_color = 0
>FRAG
>PROPERTY FS_COLOR0_WRITES_ALL_CBUFS 1
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL IN[1], GENERIC[1], PERSPECTIVE
>DCL IN[2], GENERIC[2], PERSPECTIVE
>DCL IN[3], GENERIC[3], PERSPECTIVE
>DCL IN[4], GENERIC[4], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SVIEW[0], 2D, FLOAT
>DCL TEMP[0..1], LOCAL
>  0: MOV TEMP[0].xy, IN[4].xyyy
>  1: TEX TEMP[0], TEMP[0], SAMP[0], 2D
>  2: LRP TEMP[0], IN[1].xxxx, TEMP[0], IN[0]
>  3: MAD TEMP[0], TEMP[0], IN[3], IN[2]
>  4: MUL TEMP[1].x, TEMP[0].wwww, IN[1].wwww
>  5: MOV TEMP[0].w, TEMP[1].xxxx
>  6: MOV OUT[0], TEMP[0]
>  7: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %24 = load <8 x i32>, <8 x i32> addrspace(2)* %23, align 32, !tbaa !0
>  %25 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %26 = load <4 x i32>, <4 x i32> addrspace(2)* %25, align 16, !tbaa !0
>  %27 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %8)
>  %28 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %8)
>  %29 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %6, <2 x i32> %8)
>  %30 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %6, <2 x i32> %8)
>  %31 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %6, <2 x i32> %8)
>  %32 = call float @llvm.SI.fs.interp(i32 3, i32 1, i32 %6, <2 x i32> %8)
>  %33 = call float @llvm.SI.fs.interp(i32 0, i32 2, i32 %6, <2 x i32> %8)
>  %34 = call float @llvm.SI.fs.interp(i32 1, i32 2, i32 %6, <2 x i32> %8)
>  %35 = call float @llvm.SI.fs.interp(i32 2, i32 2, i32 %6, <2 x i32> %8)
>  %36 = call float @llvm.SI.fs.interp(i32 3, i32 2, i32 %6, <2 x i32> %8)
>  %37 = call float @llvm.SI.fs.interp(i32 0, i32 3, i32 %6, <2 x i32> %8)
>  %38 = call float @llvm.SI.fs.interp(i32 1, i32 3, i32 %6, <2 x i32> %8)
>  %39 = call float @llvm.SI.fs.interp(i32 2, i32 3, i32 %6, <2 x i32> %8)
>  %40 = call float @llvm.SI.fs.interp(i32 3, i32 3, i32 %6, <2 x i32> %8)
>  %41 = call float @llvm.SI.fs.interp(i32 0, i32 4, i32 %6, <2 x i32> %8)
>  %42 = call float @llvm.SI.fs.interp(i32 1, i32 4, i32 %6, <2 x i32> %8)
>  %43 = bitcast float %41 to i32
>  %44 = bitcast float %42 to i32
>  %45 = insertelement <2 x i32> undef, i32 %43, i32 0
>  %46 = insertelement <2 x i32> %45, i32 %44, i32 1
>  %47 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %46, <8 x i32> %24, <4 x i32> %26, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %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 = fsub float 1.000000e+00, %31
>  %53 = fmul float %48, %31
>  %54 = fmul float %27, %52
>  %55 = fadd float %53, %54
>  %56 = fsub float 1.000000e+00, %31
>  %57 = fmul float %49, %31
>  %58 = fmul float %28, %56
>  %59 = fadd float %57, %58
>  %60 = fsub float 1.000000e+00, %31
>  %61 = fmul float %50, %31
>  %62 = fmul float %29, %60
>  %63 = fadd float %61, %62
>  %64 = fsub float 1.000000e+00, %31
>  %65 = fmul float %51, %31
>  %66 = fmul float %30, %64
>  %67 = fadd float %65, %66
>  %68 = fmul float %55, %37
>  %69 = fadd float %68, %33
>  %70 = fmul float %59, %38
>  %71 = fadd float %70, %34
>  %72 = fmul float %63, %39
>  %73 = fadd float %72, %35
>  %74 = fmul float %67, %40
>  %75 = fadd float %74, %36
>  %76 = fmul float %75, %32
>  %77 = call i32 @llvm.SI.packf16(float %69, float %71)
>  %78 = bitcast i32 %77 to float
>  %79 = call i32 @llvm.SI.packf16(float %73, float %76)
>  %80 = bitcast i32 %79 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %78, float %80, float %78, float %80)
>  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.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, 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 }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                      ; BEFE0A7E
>	s_mov_b32 m0, s10                                                         ; BEFC030A
>	s_load_dwordx8 s[8:15], s[6:7], 0x0                                       ; C0C40700
>	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_interp_p1_f32 v4, v0, 2, 0, [m0]                                        ; C8100200
>	v_interp_p2_f32 v4, [v4], v1, 2, 0, [m0]                                  ; C8110201
>	v_interp_p1_f32 v5, v0, 3, 0, [m0]                                        ; C8140300
>	v_interp_p2_f32 v5, [v5], v1, 3, 0, [m0]                                  ; C8150301
>	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 v7, v0, 3, 1, [m0]                                        ; C81C0700
>	v_interp_p2_f32 v7, [v7], v1, 3, 1, [m0]                                  ; C81D0701
>	v_interp_p1_f32 v8, v0, 0, 2, [m0]                                        ; C8200800
>	v_interp_p2_f32 v8, [v8], v1, 0, 2, [m0]                                  ; C8210801
>	v_interp_p1_f32 v9, v0, 1, 2, [m0]                                        ; C8240900
>	v_interp_p2_f32 v9, [v9], v1, 1, 2, [m0]                                  ; C8250901
>	v_interp_p1_f32 v10, v0, 2, 2, [m0]                                       ; C8280A00
>	v_interp_p2_f32 v10, [v10], v1, 2, 2, [m0]                                ; C8290A01
>	v_interp_p1_f32 v11, v0, 3, 2, [m0]                                       ; C82C0B00
>	v_interp_p2_f32 v11, [v11], v1, 3, 2, [m0]                                ; C82D0B01
>	v_interp_p1_f32 v12, v0, 0, 3, [m0]                                       ; C8300C00
>	v_interp_p2_f32 v12, [v12], v1, 0, 3, [m0]                                ; C8310C01
>	v_interp_p1_f32 v13, v0, 1, 3, [m0]                                       ; C8340D00
>	s_load_dwordx4 s[0:3], s[4:5], 0x0                                        ; C0800500
>	v_interp_p2_f32 v13, [v13], v1, 1, 3, [m0]                                ; C8350D01
>	v_interp_p1_f32 v14, v0, 2, 3, [m0]                                       ; C8380E00
>	v_interp_p2_f32 v14, [v14], v1, 2, 3, [m0]                                ; C8390E01
>	v_interp_p1_f32 v15, v0, 3, 3, [m0]                                       ; C83C0F00
>	v_interp_p2_f32 v15, [v15], v1, 3, 3, [m0]                                ; C83D0F01
>	v_interp_p1_f32 v16, v0, 0, 4, [m0]                                       ; C8401000
>	v_interp_p2_f32 v16, [v16], v1, 0, 4, [m0]                                ; C8411001
>	v_interp_p1_f32 v17, v0, 1, 4, [m0]                                       ; C8441100
>	v_interp_p2_f32 v17, [v17], v1, 1, 4, [m0]                                ; C8451101
>	s_waitcnt lgkmcnt(0)                                                      ; BF8C007F
>	image_sample v[16:19], 15, 0, 0, 0, 0, 0, 0, 0, v[16:17], s[8:15], s[0:3] ; F0800F00 00021010
>	v_mad_f32 v0, -v6, v2, v2                                                 ; D2820000 240A0506
>	s_waitcnt vmcnt(0)                                                        ; BF8C0770
>	v_mac_f32_e32 v0, v6, v16                                                 ; 3E002106
>	v_mad_f32 v1, -v6, v3, v3                                                 ; D2820001 240E0706
>	v_mac_f32_e32 v1, v6, v17                                                 ; 3E022306
>	v_mad_f32 v2, -v6, v4, v4                                                 ; D2820002 24120906
>	v_mac_f32_e32 v2, v6, v18                                                 ; 3E042506
>	v_mad_f32 v3, -v6, v5, v5                                                 ; D2820003 24160B06
>	v_mac_f32_e32 v3, v6, v19                                                 ; 3E062706
>	v_mac_f32_e32 v8, v12, v0                                                 ; 3E10010C
>	v_mac_f32_e32 v9, v13, v1                                                 ; 3E12030D
>	v_mac_f32_e32 v10, v14, v2                                                ; 3E14050E
>	v_mac_f32_e32 v11, v15, v3                                                ; 3E16070F
>	v_mul_f32_e32 v0, v7, v11                                                 ; 10001707
>	v_cvt_pkrtz_f16_f32_e32 v1, v8, v9                                        ; 5E021308
>	v_cvt_pkrtz_f16_f32_e32 v0, v10, v0                                       ; 5E00010A
>	exp 15, 0, 1, 1, 1, v1, v0, v1, v0                                        ; F8001C0F 00010001
>	s_endpgm                                                                  ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 16
>VGPRS: 20
>Code Size: 248 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
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL IN[1]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[0]
>DCL OUT[2], GENERIC[1]
>DCL OUT[3], GENERIC[2]
>DCL CONST[0..3]
>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: DP4 TEMP[0].x, IN[1], CONST[2]
>  2: DP4 TEMP[1].x, IN[1], CONST[3]
>  3: MOV TEMP[0].y, TEMP[1].xxxx
>  4: MOV TEMP[1].xy, IN[0].xyxx
>  5: MOV OUT[1], CONST[0]
>  6: MOV OUT[2], CONST[1]
>  7: MOV OUT[0], TEMP[0]
>  8: MOV OUT[3], TEMP[1]
>  9: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = call float @llvm.SI.load.const(<16 x i8> %13, i32 0)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %13, i32 4)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %13, i32 8)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %13, i32 12)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %13, i32 16)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %13, i32 20)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %13, i32 24)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %13, i32 28)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %13, i32 32)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %13, i32 36)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %13, i32 40)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %13, i32 44)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %13, i32 48)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %13, i32 52)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %13, i32 56)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %13, i32 60)
>  %30 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %31 = load <16 x i8>, <16 x i8> addrspace(2)* %30, align 16, !tbaa !0
>  %32 = add i32 %5, %8
>  %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>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 1
>  %37 = load <16 x i8>, <16 x i8> addrspace(2)* %36, align 16, !tbaa !0
>  %38 = add i32 %5, %8
>  %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 = extractelement <4 x float> %39, i32 3
>  %44 = fmul float %40, %22
>  %45 = fmul float %41, %23
>  %46 = fadd float %44, %45
>  %47 = fmul float %42, %24
>  %48 = fadd float %46, %47
>  %49 = fmul float %43, %25
>  %50 = fadd float %48, %49
>  %51 = fmul float %40, %26
>  %52 = fmul float %41, %27
>  %53 = fadd float %51, %52
>  %54 = fmul float %42, %28
>  %55 = fadd float %53, %54
>  %56 = fmul float %43, %29
>  %57 = fadd float %55, %56
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %14, float %15, float %16, float %17)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %18, float %19, float %20, float %21)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 34, i32 0, float %34, float %35, float undef, float undef)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %50, float %57, 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_load_dwordx4 s[4:7], s[8:9], 0x0                    ; C0820900
>	s_load_dwordx4 s[12:15], s[8:9], 0x4                  ; C0860904
>	v_add_i32_e32 v0, vcc, s10, v0                        ; 4A00000A
>	v_mov_b32_e32 v1, 1.0                                 ; 7E0202F2
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	s_buffer_load_dword s8, s[0:3], 0x5                   ; C2040105
>	s_buffer_load_dword s9, s[0:3], 0x6                   ; C2048106
>	s_buffer_load_dword s10, s[0:3], 0x7                  ; C2050107
>	s_buffer_load_dword s11, s[0:3], 0x8                  ; C2058108
>	s_buffer_load_dword s16, s[0:3], 0x9                  ; C2080109
>	buffer_load_format_xyzw v[2:5], v0, s[4:7], 0 idxen   ; E00C2000 80010200
>	s_waitcnt vmcnt(0)                                    ; BF8C0770
>	buffer_load_format_xyzw v[4:7], v0, s[12:15], 0 idxen ; E00C2000 80030400
>	s_buffer_load_dword s4, s[0:3], 0xa                   ; C202010A
>	s_buffer_load_dword s5, s[0:3], 0xb                   ; C202810B
>	s_buffer_load_dword s6, s[0:3], 0xc                   ; C203010C
>	s_buffer_load_dword s7, s[0:3], 0xd                   ; C203810D
>	s_buffer_load_dword s12, s[0:3], 0xe                  ; C206010E
>	s_buffer_load_dword s13, s[0:3], 0x0                  ; C2068100
>	s_buffer_load_dword s14, s[0:3], 0x1                  ; C2070101
>	s_buffer_load_dword s15, s[0:3], 0x2                  ; C2078102
>	s_buffer_load_dword s17, s[0:3], 0x3                  ; C2088103
>	s_buffer_load_dword s18, s[0:3], 0x4                  ; C2090104
>	s_buffer_load_dword s0, s[0:3], 0xf                   ; C200010F
>	s_waitcnt lgkmcnt(0)                                  ; BF8C007F
>	v_mov_b32_e32 v0, s8                                  ; 7E000208
>	v_mov_b32_e32 v8, s13                                 ; 7E10020D
>	v_mov_b32_e32 v9, s14                                 ; 7E12020E
>	v_mov_b32_e32 v10, s15                                ; 7E14020F
>	v_mov_b32_e32 v11, s17                                ; 7E160211
>	exp 15, 32, 0, 0, 0, v8, v9, v10, v11                 ; F800020F 0B0A0908
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mov_b32_e32 v8, s18                                 ; 7E100212
>	v_mov_b32_e32 v9, s9                                  ; 7E120209
>	v_mov_b32_e32 v10, s10                                ; 7E14020A
>	exp 15, 33, 0, 0, 0, v8, v0, v9, v10                  ; F800021F 0A090008
>	s_waitcnt vmcnt(0) expcnt(0)                          ; BF8C0700
>	v_mul_f32_e32 v0, s16, v5                             ; 10000A10
>	exp 15, 34, 0, 0, 0, v2, v3, v0, v0                   ; F800022F 00000302
>	s_waitcnt expcnt(0)                                   ; BF8C070F
>	v_mul_f32_e32 v2, s7, v5                              ; 10040A07
>	v_mac_f32_e32 v0, s11, v4                             ; 3E00080B
>	v_mac_f32_e32 v2, s6, v4                              ; 3E040806
>	v_mac_f32_e32 v0, s4, v6                              ; 3E000C04
>	v_mac_f32_e32 v2, s12, v6                             ; 3E040C0C
>	v_mac_f32_e32 v0, s5, v7                              ; 3E000E05
>	v_mac_f32_e32 v2, s0, v7                              ; 3E040E00
>	v_mov_b32_e32 v3, 0                                   ; 7E060280
>	exp 15, 12, 0, 1, 0, v0, v2, v3, v1                   ; F80008CF 01030200
>	s_endpgm                                              ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 24
>VGPRS: 12
>Code Size: 228 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
>  clamp_color = 0
>FRAG
>PROPERTY FS_COLOR0_WRITES_ALL_CBUFS 1
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL IN[1], GENERIC[1], PERSPECTIVE
>DCL IN[2], GENERIC[2], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SVIEW[0], 2D, FLOAT
>DCL TEMP[0], LOCAL
>  0: MOV TEMP[0].xy, IN[2].xyyy
>  1: TEX TEMP[0], TEMP[0], SAMP[0], 2D
>  2: MAD TEMP[0], TEMP[0], IN[1], IN[0]
>  3: MOV OUT[0], TEMP[0]
>  4: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %24 = load <8 x i32>, <8 x i32> addrspace(2)* %23, align 32, !tbaa !0
>  %25 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %26 = load <4 x i32>, <4 x i32> addrspace(2)* %25, align 16, !tbaa !0
>  %27 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %8)
>  %28 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %8)
>  %29 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %6, <2 x i32> %8)
>  %30 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %6, <2 x i32> %8)
>  %31 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %6, <2 x i32> %8)
>  %32 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %6, <2 x i32> %8)
>  %33 = call float @llvm.SI.fs.interp(i32 2, i32 1, i32 %6, <2 x i32> %8)
>  %34 = call float @llvm.SI.fs.interp(i32 3, i32 1, i32 %6, <2 x i32> %8)
>  %35 = call float @llvm.SI.fs.interp(i32 0, i32 2, i32 %6, <2 x i32> %8)
>  %36 = call float @llvm.SI.fs.interp(i32 1, i32 2, i32 %6, <2 x i32> %8)
>  %37 = bitcast float %35 to i32
>  %38 = bitcast float %36 to i32
>  %39 = insertelement <2 x i32> undef, i32 %37, i32 0
>  %40 = insertelement <2 x i32> %39, i32 %38, i32 1
>  %41 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %40, <8 x i32> %24, <4 x i32> %26, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %42 = extractelement <4 x float> %41, i32 0
>  %43 = extractelement <4 x float> %41, i32 1
>  %44 = extractelement <4 x float> %41, i32 2
>  %45 = extractelement <4 x float> %41, i32 3
>  %46 = fmul float %42, %31
>  %47 = fadd float %46, %27
>  %48 = fmul float %43, %32
>  %49 = fadd float %48, %28
>  %50 = fmul float %44, %33
>  %51 = fadd float %50, %29
>  %52 = fmul float %45, %34
>  %53 = fadd float %52, %30
>  %54 = call i32 @llvm.SI.packf16(float %47, float %49)
>  %55 = bitcast i32 %54 to float
>  %56 = call i32 @llvm.SI.packf16(float %51, float %53)
>  %57 = bitcast i32 %56 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %55, float %57, float %55, float %57)
>  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.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, 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 }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                      ; BEFE0A7E
>	s_mov_b32 m0, s10                                                         ; BEFC030A
>	s_load_dwordx8 s[8:15], s[6:7], 0x0                                       ; C0C40700
>	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_interp_p1_f32 v4, v0, 2, 0, [m0]                                        ; C8100200
>	v_interp_p2_f32 v4, [v4], v1, 2, 0, [m0]                                  ; C8110201
>	v_interp_p1_f32 v5, v0, 3, 0, [m0]                                        ; C8140300
>	v_interp_p2_f32 v5, [v5], v1, 3, 0, [m0]                                  ; C8150301
>	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 v7, v0, 1, 1, [m0]                                        ; C81C0500
>	s_load_dwordx4 s[0:3], s[4:5], 0x0                                        ; C0800500
>	v_interp_p2_f32 v7, [v7], v1, 1, 1, [m0]                                  ; C81D0501
>	v_interp_p1_f32 v8, v0, 2, 1, [m0]                                        ; C8200600
>	v_interp_p2_f32 v8, [v8], v1, 2, 1, [m0]                                  ; C8210601
>	v_interp_p1_f32 v9, v0, 3, 1, [m0]                                        ; C8240700
>	v_interp_p2_f32 v9, [v9], v1, 3, 1, [m0]                                  ; C8250701
>	v_interp_p1_f32 v10, v0, 0, 2, [m0]                                       ; C8280800
>	v_interp_p2_f32 v10, [v10], v1, 0, 2, [m0]                                ; C8290801
>	v_interp_p1_f32 v11, v0, 1, 2, [m0]                                       ; C82C0900
>	v_interp_p2_f32 v11, [v11], v1, 1, 2, [m0]                                ; C82D0901
>	s_waitcnt lgkmcnt(0)                                                      ; BF8C007F
>	image_sample v[10:13], 15, 0, 0, 0, 0, 0, 0, 0, v[10:11], s[8:15], s[0:3] ; F0800F00 00020A0A
>	s_waitcnt vmcnt(0)                                                        ; BF8C0770
>	v_mac_f32_e32 v2, v6, v10                                                 ; 3E041506
>	v_mac_f32_e32 v3, v7, v11                                                 ; 3E061707
>	v_mac_f32_e32 v4, v8, v12                                                 ; 3E081908
>	v_mac_f32_e32 v5, v9, v13                                                 ; 3E0A1B09
>	v_cvt_pkrtz_f16_f32_e32 v0, v2, v3                                        ; 5E000702
>	v_cvt_pkrtz_f16_f32_e32 v1, v4, v5                                        ; 5E020B04
>	exp 15, 0, 1, 1, 1, v0, v1, v0, v1                                        ; F8001C0F 01000100
>	s_endpgm                                                                  ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 16
>VGPRS: 16
>Code Size: 148 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
>  clamp_color = 0
>FRAG
>  0: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  call void @llvm.SI.export(i32 0, i32 1, i32 1, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  ret void
>}
>
>declare void @llvm.SI.export(i32, i32, i32, i32, i32, i32, i32, i32, i32)
>
>attributes #0 = { "ShaderType"="0" "enable-no-nans-fp-math"="true" }
>
>Shader Disassembly:
>
>	v_mov_b32_e32 v0, 0               ; 7E000280
>	exp 0, 0, 0, 1, 1, v0, v0, v0, v0 ; F8001800 00000000
>	s_endpgm                          ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 8
>VGPRS: 4
>Code Size: 16 bytes
>LDS: 0 blocks
>Scratch: 0 bytes per wave
>********************
>Looking up breakpad interfaces from steamclient
>Calling BreakpadMiniDumpSystemInit
>Steam_SetMinidumpSteamID:  Caching Steam ID:  76561197960306978 [API loaded yes]
>Steam_SetMinidumpSteamID:  Setting Steam ID:  76561197960306978
>SHADER KEY
>  instance_divisors = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
>  as_es = 0
>  as_ls = 0
>  export_prim_id = 0
>VERT
>DCL IN[0]
>DCL OUT[0], POSITION
>DCL OUT[1], GENERIC[0]
>DCL OUT[2], GENERIC[1]
>DCL OUT[3], GENERIC[2]
>DCL CONST[0..5]
>DCL TEMP[0..2], LOCAL
>IMM[0] FLT32 {    0.0000,     1.0000,     0.0000,     0.0000}
>  0: MOV TEMP[0].zw, IMM[0].yyxy
>  1: DP4 TEMP[0].x, IN[0], CONST[2]
>  2: DP4 TEMP[1].x, IN[0], CONST[3]
>  3: MOV TEMP[0].y, TEMP[1].xxxx
>  4: DP4 TEMP[1].x, IN[0], CONST[4]
>  5: DP4 TEMP[2].x, IN[0], CONST[5]
>  6: MOV TEMP[1].y, TEMP[2].xxxx
>  7: MOV TEMP[1].xy, TEMP[1].xyxx
>  8: MOV OUT[1], CONST[0]
>  9: MOV OUT[2], CONST[1]
> 10: MOV OUT[0], TEMP[0]
> 11: MOV OUT[3], TEMP[1]
> 12: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = call float @llvm.SI.load.const(<16 x i8> %13, i32 0)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %13, i32 4)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %13, i32 8)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %13, i32 12)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %13, i32 16)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %13, i32 20)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %13, i32 24)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %13, i32 28)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %13, i32 32)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %13, i32 36)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %13, i32 40)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %13, i32 44)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %13, i32 48)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %13, i32 52)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %13, i32 56)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %13, i32 60)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %13, i32 64)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %13, i32 68)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %13, i32 72)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %13, i32 76)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %13, i32 80)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %13, i32 84)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %13, i32 88)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %13, i32 92)
>  %38 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %39 = load <16 x i8>, <16 x i8> addrspace(2)* %38, align 16, !tbaa !0
>  %40 = add i32 %5, %8
>  %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 = extractelement <4 x float> %41, i32 2
>  %45 = extractelement <4 x float> %41, i32 3
>  %46 = fmul float %42, %22
>  %47 = fmul float %43, %23
>  %48 = fadd float %46, %47
>  %49 = fmul float %44, %24
>  %50 = fadd float %48, %49
>  %51 = fmul float %45, %25
>  %52 = fadd float %50, %51
>  %53 = fmul float %42, %26
>  %54 = fmul float %43, %27
>  %55 = fadd float %53, %54
>  %56 = fmul float %44, %28
>  %57 = fadd float %55, %56
>  %58 = fmul float %45, %29
>  %59 = fadd float %57, %58
>  %60 = fmul float %42, %30
>  %61 = fmul float %43, %31
>  %62 = fadd float %60, %61
>  %63 = fmul float %44, %32
>  %64 = fadd float %62, %63
>  %65 = fmul float %45, %33
>  %66 = fadd float %64, %65
>  %67 = fmul float %42, %34
>  %68 = fmul float %43, %35
>  %69 = fadd float %67, %68
>  %70 = fmul float %44, %36
>  %71 = fadd float %69, %70
>  %72 = fmul float %45, %37
>  %73 = fadd float %71, %72
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %14, float %15, float %16, float %17)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %18, float %19, float %20, float %21)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 34, i32 0, float %66, float %73, float undef, float undef)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %52, float %59, 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[4:7], s[8:9], 0x0                  ; C0820900
>	v_add_i32_e32 v0, vcc, s10, v0                      ; 4A00000A
>	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 lgkmcnt(0)                                ; BF8C007F
>	s_buffer_load_dword s4, s[0:3], 0x0                 ; C2020100
>	s_buffer_load_dword s5, s[0:3], 0x1                 ; C2028101
>	s_buffer_load_dword s6, s[0:3], 0x2                 ; C2030102
>	s_buffer_load_dword s7, s[0:3], 0x3                 ; C2038103
>	s_buffer_load_dword s8, s[0:3], 0x4                 ; C2040104
>	s_buffer_load_dword s9, s[0:3], 0x5                 ; C2048105
>	s_buffer_load_dword s10, s[0:3], 0x6                ; C2050106
>	s_buffer_load_dword s11, s[0:3], 0x7                ; C2058107
>	s_buffer_load_dword s12, s[0:3], 0x8                ; C2060108
>	s_buffer_load_dword s13, s[0:3], 0x9                ; C2068109
>	s_buffer_load_dword s14, s[0:3], 0xa                ; C207010A
>	s_buffer_load_dword s15, s[0:3], 0xb                ; C207810B
>	s_buffer_load_dword s16, s[0:3], 0xc                ; C208010C
>	s_buffer_load_dword s17, s[0:3], 0xd                ; C208810D
>	s_buffer_load_dword s18, s[0:3], 0xe                ; C209010E
>	s_buffer_load_dword s19, s[0:3], 0xf                ; C209810F
>	s_buffer_load_dword s20, s[0:3], 0x10               ; C20A0110
>	s_buffer_load_dword s21, s[0:3], 0x11               ; C20A8111
>	s_buffer_load_dword s22, s[0:3], 0x14               ; C20B0114
>	s_buffer_load_dword s23, s[0:3], 0x15               ; C20B8115
>	s_buffer_load_dword s24, s[0:3], 0x12               ; C20C0112
>	s_buffer_load_dword s25, s[0:3], 0x13               ; C20C8113
>	s_buffer_load_dword s26, s[0:3], 0x16               ; C20D0116
>	s_buffer_load_dword s0, s[0:3], 0x17                ; C2000117
>	s_waitcnt lgkmcnt(0)                                ; BF8C007F
>	v_mov_b32_e32 v4, s4                                ; 7E080204
>	v_mov_b32_e32 v5, s5                                ; 7E0A0205
>	v_mov_b32_e32 v6, s6                                ; 7E0C0206
>	v_mov_b32_e32 v7, s7                                ; 7E0E0207
>	exp 15, 32, 0, 0, 0, v4, v5, v6, v7                 ; F800020F 07060504
>	s_waitcnt expcnt(0)                                 ; BF8C070F
>	v_mov_b32_e32 v4, s8                                ; 7E080208
>	v_mov_b32_e32 v5, s9                                ; 7E0A0209
>	v_mov_b32_e32 v6, s10                               ; 7E0C020A
>	v_mov_b32_e32 v7, s11                               ; 7E0E020B
>	exp 15, 33, 0, 0, 0, v4, v5, v6, v7                 ; F800021F 07060504
>	s_waitcnt vmcnt(0) expcnt(0)                        ; BF8C0700
>	v_mul_f32_e32 v4, s13, v1                           ; 1008020D
>	v_mul_f32_e32 v5, s17, v1                           ; 100A0211
>	v_mul_f32_e32 v6, s21, v1                           ; 100C0215
>	v_mul_f32_e32 v1, s23, v1                           ; 10020217
>	v_mac_f32_e32 v4, s12, v0                           ; 3E08000C
>	v_mac_f32_e32 v5, s16, v0                           ; 3E0A0010
>	v_mac_f32_e32 v6, s20, v0                           ; 3E0C0014
>	v_mac_f32_e32 v1, s22, v0                           ; 3E020016
>	v_mac_f32_e32 v4, s14, v2                           ; 3E08040E
>	v_mac_f32_e32 v5, s18, v2                           ; 3E0A0412
>	v_mac_f32_e32 v6, s24, v2                           ; 3E0C0418
>	v_mac_f32_e32 v1, s26, v2                           ; 3E02041A
>	v_mac_f32_e32 v4, s15, v3                           ; 3E08060F
>	v_mac_f32_e32 v5, s19, v3                           ; 3E0A0613
>	v_mac_f32_e32 v6, s25, v3                           ; 3E0C0619
>	v_mac_f32_e32 v1, s0, v3                            ; 3E020600
>	v_mov_b32_e32 v0, 1.0                               ; 7E0002F2
>	exp 15, 34, 0, 0, 0, v6, v1, v0, v0                 ; F800022F 00000106
>	s_waitcnt expcnt(0)                                 ; BF8C070F
>	v_mov_b32_e32 v1, 0                                 ; 7E020280
>	exp 15, 12, 0, 1, 0, v4, v5, v1, v0                 ; F80008CF 00010504
>	s_endpgm                                            ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 32
>VGPRS: 8
>Code Size: 280 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
>  clamp_color = 0
>FRAG
>PROPERTY FS_COLOR0_WRITES_ALL_CBUFS 1
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL IN[1], GENERIC[1], PERSPECTIVE
>DCL IN[2], GENERIC[2], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SVIEW[0], 2D, FLOAT
>DCL CONST[0]
>DCL CONST[2]
>DCL TEMP[0..9], LOCAL
>IMM[0] FLT32 {    0.0000,     1.0000,     0.0000,     0.0000}
>  0: MOV TEMP[0].xy, IN[2].xyxx
>  1: MOV TEMP[1], IMM[0].xxxx
>  2: MOV TEMP[2], IMM[0].xxxx
>  3: MOV TEMP[3].y, IMM[0].xxxx
>  4: MOV TEMP[3].x, -CONST[0].xxxx
>  5: BGNLOOP :0
>  6:   FSLT TEMP[4].x, CONST[0].xxxx, TEMP[3].xxxx
>  7:   UIF TEMP[4].xxxx :0
>  8:     BRK
>  9:   ENDIF
> 10:   MOV TEMP[3].y, -CONST[0].yyyy
> 11:   BGNLOOP :0
> 12:     FSLT TEMP[5].x, CONST[0].yyyy, TEMP[3].yyyy
> 13:     UIF TEMP[5].xxxx :0
> 14:       BRK
> 15:     ENDIF
> 16:     MAD TEMP[6].xy, TEMP[3].xyyy, CONST[2].xyyy, TEMP[0].xyyy
> 17:     MOV TEMP[7].xy, TEMP[6].xyyy
> 18:     MOV TEMP[7].w, IMM[0].xxxx
> 19:     TXB TEMP[8], TEMP[7], SAMP[0], 2D
> 20:     ADD TEMP[2], TEMP[2], TEMP[8]
> 21:     ADD TEMP[9].x, TEMP[3].yyyy, IMM[0].yyyy
> 22:     MOV TEMP[3].y, TEMP[9].xxxx
> 23:   ENDLOOP :0
> 24:   ADD TEMP[3].x, TEMP[3].xxxx, IMM[0].yyyy
> 25: ENDLOOP :0
> 26: MUL TEMP[1], TEMP[2], CONST[0].wwww
> 27: MOV TEMP[0].w, IMM[0].yyyy
> 28: MOV TEMP[0].xyz, IN[1].xyzx
> 29: MUL TEMP[0], TEMP[1], TEMP[0]
> 30: MUL TEMP[1], TEMP[0], IN[1].wwww
> 31: MAD TEMP[1], IN[0], TEMP[1].wwww, TEMP[1]
> 32: MOV OUT[0], TEMP[1]
> 33: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %24 = load <16 x i8>, <16 x i8> addrspace(2)* %23, align 16, !tbaa !0
>  %25 = call float @llvm.SI.load.const(<16 x i8> %24, i32 0)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %24, i32 4)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %24, i32 12)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %24, i32 32)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %24, i32 36)
>  %30 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %31 = load <8 x i32>, <8 x i32> addrspace(2)* %30, align 32, !tbaa !0
>  %32 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %33 = load <4 x i32>, <4 x i32> addrspace(2)* %32, align 16, !tbaa !0
>  %34 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %8)
>  %35 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %8)
>  %36 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %6, <2 x i32> %8)
>  %37 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %6, <2 x i32> %8)
>  %38 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %6, <2 x i32> %8)
>  %39 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %6, <2 x i32> %8)
>  %40 = call float @llvm.SI.fs.interp(i32 2, i32 1, i32 %6, <2 x i32> %8)
>  %41 = call float @llvm.SI.fs.interp(i32 3, i32 1, i32 %6, <2 x i32> %8)
>  %42 = call float @llvm.SI.fs.interp(i32 0, i32 2, i32 %6, <2 x i32> %8)
>  %43 = call float @llvm.SI.fs.interp(i32 1, i32 2, i32 %6, <2 x i32> %8)
>  %44 = fsub float -0.000000e+00, %25
>  %45 = fsub float -0.000000e+00, %26
>  br label %LOOP
>
>LOOP:                                             ; preds = %IF43, %main_body
>  %temp9.0 = phi float [ 0.000000e+00, %main_body ], [ %temp9.1, %IF43 ]
>  %temp10.0 = phi float [ 0.000000e+00, %main_body ], [ %temp10.1, %IF43 ]
>  %temp11.0 = phi float [ 0.000000e+00, %main_body ], [ %temp11.1, %IF43 ]
>  %temp12.0 = phi float [ %44, %main_body ], [ %75, %IF43 ]
>  %temp8.0 = phi float [ 0.000000e+00, %main_body ], [ %temp8.1, %IF43 ]
>  %46 = fcmp olt float %25, %temp12.0
>  br i1 %46, label %IF, label %ENDIF
>
>IF:                                               ; preds = %LOOP
>  %47 = fmul float %temp8.0, %27
>  %48 = fmul float %temp9.0, %27
>  %49 = fmul float %temp10.0, %27
>  %50 = fmul float %temp11.0, %27
>  %51 = fmul float %47, %38
>  %52 = fmul float %48, %39
>  %53 = fmul float %49, %40
>  %54 = fmul float %51, %41
>  %55 = fmul float %52, %41
>  %56 = fmul float %53, %41
>  %57 = fmul float %50, %41
>  %58 = fmul float %34, %57
>  %59 = fadd float %58, %54
>  %60 = fmul float %35, %57
>  %61 = fadd float %60, %55
>  %62 = fmul float %36, %57
>  %63 = fadd float %62, %56
>  %64 = fmul float %37, %57
>  %65 = fadd float %64, %57
>  %66 = call i32 @llvm.SI.packf16(float %59, float %61)
>  %67 = bitcast i32 %66 to float
>  %68 = call i32 @llvm.SI.packf16(float %63, float %65)
>  %69 = bitcast i32 %68 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %67, float %69, float %67, float %69)
>  ret void
>
>ENDIF:                                            ; preds = %LOOP
>  %70 = fmul float %temp12.0, %28
>  %71 = fadd float %70, %42
>  %72 = bitcast float %71 to i32
>  %73 = insertelement <4 x i32> <i32 0, i32 undef, i32 undef, i32 undef>, i32 %72, i32 1
>  br label %LOOP41
>
>LOOP41:                                           ; preds = %ENDIF42, %ENDIF
>  %temp9.1 = phi float [ %temp9.0, %ENDIF ], [ %86, %ENDIF42 ]
>  %temp10.1 = phi float [ %temp10.0, %ENDIF ], [ %87, %ENDIF42 ]
>  %temp11.1 = phi float [ %temp11.0, %ENDIF ], [ %88, %ENDIF42 ]
>  %temp13.0 = phi float [ %45, %ENDIF ], [ %89, %ENDIF42 ]
>  %temp8.1 = phi float [ %temp8.0, %ENDIF ], [ %85, %ENDIF42 ]
>  %74 = fcmp olt float %26, %temp13.0
>  br i1 %74, label %IF43, label %ENDIF42
>
>IF43:                                             ; preds = %LOOP41
>  %75 = fadd float %temp12.0, 1.000000e+00
>  br label %LOOP
>
>ENDIF42:                                          ; preds = %LOOP41
>  %76 = fmul float %temp13.0, %29
>  %77 = fadd float %76, %43
>  %78 = bitcast float %77 to i32
>  %79 = insertelement <4 x i32> %73, i32 %78, i32 2
>  %80 = call <4 x float> @llvm.SI.image.sample.b.v4i32(<4 x i32> %79, <8 x i32> %31, <4 x i32> %33, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %81 = extractelement <4 x float> %80, i32 0
>  %82 = extractelement <4 x float> %80, i32 1
>  %83 = extractelement <4 x float> %80, i32 2
>  %84 = extractelement <4 x float> %80, i32 3
>  %85 = fadd float %temp8.1, %81
>  %86 = fadd float %temp9.1, %82
>  %87 = fadd float %temp10.1, %83
>  %88 = fadd float %temp11.1, %84
>  %89 = fadd float %temp13.0, 1.000000e+00
>  br label %LOOP41
>}
>
>; 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.image.sample.b.v4i32(<4 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, 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 }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                          ; BEFE0A7E
>	s_load_dwordx4 s[16:19], s[2:3], 0x0                                          ; C0880300
>	s_mov_b32 m0, s10                                                             ; BEFC030A
>	s_load_dwordx8 s[8:15], s[6:7], 0x0                                           ; C0C40700
>	v_interp_p1_f32 v2, v0, 0, 0, [m0]                                            ; C8080000
>	v_interp_p2_f32 v2, [v2], v1, 0, 0, [m0]                                      ; C8090001
>	s_waitcnt lgkmcnt(0)                                                          ; BF8C007F
>	s_buffer_load_dword s1, s[16:19], 0x0                                         ; C2009100
>	s_buffer_load_dword s2, s[16:19], 0x1                                         ; C2011101
>	s_buffer_load_dword s0, s[16:19], 0x3                                         ; C2001103
>	s_buffer_load_dword s3, s[16:19], 0x8                                         ; C2019108
>	s_buffer_load_dword s6, s[16:19], 0x9                                         ; C2031109
>	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, 2, 0, [m0]                                            ; C8100200
>	v_interp_p2_f32 v4, [v4], v1, 2, 0, [m0]                                      ; C8110201
>	v_interp_p1_f32 v5, v0, 3, 0, [m0]                                            ; C8140300
>	v_interp_p2_f32 v5, [v5], v1, 3, 0, [m0]                                      ; C8150301
>	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 v7, v0, 1, 1, [m0]                                            ; C81C0500
>	v_interp_p2_f32 v7, [v7], v1, 1, 1, [m0]                                      ; C81D0501
>	v_interp_p1_f32 v8, v0, 2, 1, [m0]                                            ; C8200600
>	v_interp_p2_f32 v8, [v8], v1, 2, 1, [m0]                                      ; C8210601
>	v_interp_p1_f32 v9, v0, 3, 1, [m0]                                            ; C8240700
>	v_interp_p2_f32 v9, [v9], v1, 3, 1, [m0]                                      ; C8250701
>	v_interp_p1_f32 v10, v0, 0, 2, [m0]                                           ; C8280800
>	s_load_dwordx4 s[16:19], s[4:5], 0x0                                          ; C0880500
>	v_interp_p2_f32 v10, [v10], v1, 0, 2, [m0]                                    ; C8290801
>	v_interp_p1_f32 v0, v0, 1, 2, [m0]                                            ; C8000900
>	v_interp_p2_f32 v0, [v0], v1, 1, 2, [m0]                                      ; C8010901
>	v_mov_b32_e32 v11, 0x80000000                                                 ; 7E1602FF 80000000
>	s_waitcnt lgkmcnt(0)                                                          ; BF8C007F
>	v_xor_b32_e32 v1, s1, v11                                                     ; 3A021601
>	v_xor_b32_e32 v12, s2, v11                                                    ; 3A181602
>	v_mov_b32_e32 v16, 0                                                          ; 7E200280
>	s_mov_b64 s[4:5], 0                                                           ; BE840480
>	v_mov_b32_e32 v17, 0                                                          ; 7E220280
>	v_mov_b32_e32 v18, 0                                                          ; 7E240280
>	v_mov_b32_e32 v19, 0                                                          ; 7E260280
>	v_mov_b32_e32 v13, v19                                                        ; 7E1A0313
>	v_mov_b32_e32 v11, v18                                                        ; 7E160312
>	v_mov_b32_e32 v14, v17                                                        ; 7E1C0311
>	v_mov_b32_e32 v15, v16                                                        ; 7E1E0310
>	v_cmp_nlt_f32_e32 vcc, s1, v1                                                 ; 7C1C0201
>	s_and_saveexec_b64 s[20:21], vcc                                              ; BE94246A
>	s_xor_b64 s[20:21], exec, s[20:21]                                            ; 8994147E
>	s_cbranch_execz BB0_4                                                         ; BF880000
>	v_mad_f32 v21, s3, v1, v10                                                    ; D2820015 042A0203
>	v_mov_b32_e32 v20, 0                                                          ; 7E280280
>	s_mov_b64 s[22:23], 0                                                         ; BE960480
>	v_mov_b32_e32 v22, v15                                                        ; 7E2C030F
>	v_mov_b32_e32 v24, v14                                                        ; 7E30030E
>	v_mov_b32_e32 v26, v11                                                        ; 7E34030B
>	v_mov_b32_e32 v23, v12                                                        ; 7E2E030C
>	v_mov_b32_e32 v25, v13                                                        ; 7E32030D
>	v_mov_b32_e32 v19, v25                                                        ; 7E260319
>	v_mov_b32_e32 v18, v26                                                        ; 7E24031A
>	v_mov_b32_e32 v17, v24                                                        ; 7E220318
>	v_mov_b32_e32 v16, v22                                                        ; 7E200316
>	v_cmp_nlt_f32_e32 vcc, s2, v23                                                ; 7C1C2E02
>	s_and_saveexec_b64 s[24:25], vcc                                              ; BE98246A
>	s_xor_b64 s[24:25], exec, s[24:25]                                            ; 8998187E
>	v_mad_f32 v22, s6, v23, v0                                                    ; D2820016 04022E06
>	image_sample_b v[26:29], 15, 0, 0, 0, 0, 0, 0, 0, v[20:23], s[8:15], s[16:19] ; F0940F00 00821A14
>	s_waitcnt vmcnt(0)                                                            ; BF8C0770
>	v_add_f32_e32 v25, v26, v19                                                   ; 0632271A
>	v_add_f32_e32 v22, v27, v16                                                   ; 062C211B
>	v_add_f32_e32 v24, v28, v17                                                   ; 0630231C
>	v_add_f32_e32 v26, v29, v18                                                   ; 0634251D
>	v_add_f32_e32 v23, 1.0, v23                                                   ; 062E2EF2
>	s_or_b64 exec, exec, s[24:25]                                                 ; 88FE187E
>	s_or_b64 s[22:23], s[24:25], s[22:23]                                         ; 88961618
>	s_andn2_b64 exec, exec, s[22:23]                                              ; 8AFE167E
>	s_cbranch_execnz BB0_5                                                        ; BF890000
>	s_or_b64 exec, exec, s[22:23]                                                 ; 88FE167E
>	v_add_f32_e32 v1, 1.0, v1                                                     ; 060202F2
>	s_or_b64 exec, exec, s[20:21]                                                 ; 88FE147E
>	s_or_b64 s[4:5], s[20:21], s[4:5]                                             ; 88840414
>	s_andn2_b64 exec, exec, s[4:5]                                                ; 8AFE047E
>	s_cbranch_execnz BB0_1                                                        ; BF890000
>	s_or_b64 exec, exec, s[4:5]                                                   ; 88FE047E
>	v_mul_f32_e32 v0, s0, v13                                                     ; 10001A00
>	v_mul_f32_e32 v1, s0, v15                                                     ; 10021E00
>	v_mul_f32_e32 v10, s0, v14                                                    ; 10141C00
>	v_mul_f32_e32 v11, s0, v11                                                    ; 10161600
>	v_mul_f32_e32 v0, v6, v0                                                      ; 10000106
>	v_mul_f32_e32 v1, v7, v1                                                      ; 10020307
>	v_mul_f32_e32 v6, v8, v10                                                     ; 100C1508
>	v_mul_f32_e32 v0, v9, v0                                                      ; 10000109
>	v_mul_f32_e32 v1, v9, v1                                                      ; 10020309
>	v_mul_f32_e32 v6, v9, v6                                                      ; 100C0D09
>	v_mul_f32_e32 v7, v9, v11                                                     ; 100E1709
>	v_mac_f32_e32 v0, v7, v2                                                      ; 3E000507
>	v_mac_f32_e32 v1, v7, v3                                                      ; 3E020707
>	v_mac_f32_e32 v6, v7, v4                                                      ; 3E0C0907
>	v_mac_f32_e32 v7, v7, v5                                                      ; 3E0E0B07
>	v_cvt_pkrtz_f16_f32_e32 v0, v0, v1                                            ; 5E000300
>	v_cvt_pkrtz_f16_f32_e32 v1, v6, v7                                            ; 5E020F06
>	exp 15, 0, 1, 1, 1, v0, v1, v0, v1                                            ; F8001C0F 01000100
>	s_endpgm                                                                      ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 32
>VGPRS: 32
>Code Size: 424 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
>  clamp_color = 0
>FRAG
>DCL IN[0], GENERIC[0], LINEAR
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SVIEW[0], 2D, FLOAT
>  0: TEX OUT[0], IN[0], SAMP[0], 2D
>  1: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %24 = load <8 x i32>, <8 x i32> addrspace(2)* %23, align 32, !tbaa !0
>  %25 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %26 = load <4 x i32>, <4 x i32> addrspace(2)* %25, align 16, !tbaa !0
>  %27 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %12)
>  %28 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %12)
>  %29 = bitcast float %27 to i32
>  %30 = bitcast float %28 to i32
>  %31 = insertelement <2 x i32> undef, i32 %29, i32 0
>  %32 = insertelement <2 x i32> %31, i32 %30, i32 1
>  %33 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %32, <8 x i32> %24, <4 x i32> %26, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %34 = extractelement <4 x float> %33, i32 0
>  %35 = extractelement <4 x float> %33, i32 1
>  %36 = extractelement <4 x float> %33, i32 2
>  %37 = extractelement <4 x float> %33, i32 3
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 0, float %34, float %35, float %36, float %37)
>  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.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, 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_load_dwordx8 s[12:19], s[6:7], 0x0                                   ; C0C60700
>	s_load_dwordx4 s[0:3], s[4:5], 0x0                                     ; C0800500
>	s_mov_b32 m0, s10                                                      ; BEFC030A
>	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
>	s_waitcnt lgkmcnt(0)                                                   ; BF8C007F
>	image_sample v[0:3], 15, 0, 0, 0, 0, 0, 0, 0, v[2:3], s[12:19], s[0:3] ; F0800F00 00030002
>	s_waitcnt vmcnt(0)                                                     ; BF8C0770
>	exp 15, 0, 0, 1, 1, v0, v1, v2, v3                                     ; F800180F 03020100
>	s_endpgm                                                               ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 24
>VGPRS: 4
>Code Size: 60 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
>  clamp_color = 0
>FRAG
>DCL IN[0], GENERIC[0], LINEAR
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SVIEW[0], CUBE, UINT
>  0: TEX OUT[0], IN[0], SAMP[0], CUBE
>  1: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %24 = load <8 x i32>, <8 x i32> addrspace(2)* %23, align 32, !tbaa !0
>  %25 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %26 = load <4 x i32>, <4 x i32> addrspace(2)* %25, align 16, !tbaa !0
>  %27 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %12)
>  %28 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %12)
>  %29 = call float @llvm.SI.fs.interp(i32 2, i32 0, i32 %6, <2 x i32> %12)
>  %30 = call float @llvm.SI.fs.interp(i32 3, i32 0, i32 %6, <2 x i32> %12)
>  %31 = insertelement <4 x float> undef, float %27, i32 0
>  %32 = insertelement <4 x float> %31, float %28, i32 1
>  %33 = insertelement <4 x float> %32, float %29, i32 2
>  %34 = insertelement <4 x float> %33, float %30, i32 3
>  %35 = call <4 x float> @llvm.AMDGPU.cube(<4 x float> %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 = extractelement <4 x float> %35, i32 3
>  %40 = call float @llvm.fabs.f32(float %38)
>  %41 = fdiv float 1.000000e+00, %40
>  %42 = fmul float %36, %41
>  %43 = fadd float %42, 1.500000e+00
>  %44 = fmul float %37, %41
>  %45 = fadd float %44, 1.500000e+00
>  %46 = bitcast float %45 to i32
>  %47 = bitcast float %43 to i32
>  %48 = bitcast float %39 to i32
>  %49 = insertelement <4 x i32> undef, i32 %46, i32 0
>  %50 = insertelement <4 x i32> %49, i32 %47, i32 1
>  %51 = insertelement <4 x i32> %50, i32 %48, i32 2
>  %52 = call <4 x float> @llvm.SI.image.sample.v4i32(<4 x i32> %51, <8 x i32> %24, <4 x i32> %26, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %53 = extractelement <4 x float> %52, i32 0
>  %54 = extractelement <4 x float> %52, i32 1
>  %55 = extractelement <4 x float> %52, i32 2
>  %56 = extractelement <4 x float> %52, i32 3
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 0, float %53, float %54, float %55, float %56)
>  ret void
>}
>
>; Function Attrs: nounwind readnone
>declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
>
>; Function Attrs: readnone
>declare <4 x float> @llvm.AMDGPU.cube(<4 x float>) #2
>
>; Function Attrs: nounwind readnone
>declare float @llvm.fabs.f32(float) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.image.sample.v4i32(<4 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, 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 }
>attributes #2 = { readnone }
>
>!0 = !{!"const", null, i32 1}
>
>Shader Disassembly:
>
>	s_wqm_b64 exec, exec                                                  ; BEFE0A7E
>	s_mov_b32 m0, s10                                                     ; BEFC030A
>	s_load_dwordx8 s[8:15], s[6:7], 0x0                                   ; C0C40700
>	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_interp_p1_f32 v4, v0, 2, 0, [m0]                                    ; C8100200
>	s_load_dwordx4 s[0:3], s[4:5], 0x0                                    ; C0800500
>	v_interp_p2_f32 v4, [v4], v1, 2, 0, [m0]                              ; C8110201
>	v_interp_p1_f32 v0, v0, 3, 0, [m0]                                    ; C8000300
>	v_interp_p2_f32 v0, [v0], v1, 3, 0, [m0]                              ; C8010301
>	v_cubeid_f32 v7, v2, v3, v4                                           ; D2880007 04120702
>	v_cubema_f32 v0, v2, v3, v4                                           ; D28E0000 04120702
>	v_rcp_f32_e64 v0, |v0|                                                ; D3540100 00000100
>	v_cubesc_f32 v1, v2, v3, v4                                           ; D28A0001 04120702
>	v_cubetc_f32 v2, v2, v3, v4                                           ; D28C0002 04120702
>	v_mov_b32_e32 v5, 0x3fc00000                                          ; 7E0A02FF 3FC00000
>	v_mad_f32 v6, v0, v2, v5                                              ; D2820006 04160500
>	v_mac_f32_e32 v5, v0, v1                                              ; 3E0A0300
>	s_waitcnt lgkmcnt(0)                                                  ; BF8C007F
>	image_sample v[0:3], 15, 0, 0, 0, 0, 0, 0, 0, v[5:8], s[8:15], s[0:3] ; F0800F00 00020005
>	s_waitcnt vmcnt(0)                                                    ; BF8C0770
>	exp 15, 0, 0, 1, 1, v0, v1, v2, v3                                    ; F800180F 03020100
>	s_endpgm                                                              ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 16
>VGPRS: 12
>Code Size: 136 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
>  as_ls = 0
>  export_prim_id = 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], CLIPVERTEX
>DCL OUT[2], GENERIC[0]
>DCL OUT[3], GENERIC[1]
>DCL OUT[4], GENERIC[2]
>DCL OUT[5], GENERIC[3]
>DCL OUT[6], GENERIC[4]
>DCL OUT[7], GENERIC[5]
>DCL CONST[0..57]
>DCL TEMP[0..10], LOCAL
>IMM[0] FLT32 { -128.0000,     1.0000,    -2.0000,   -64.0000}
>IMM[1] FLT32 {   -0.0159,     0.0159,     2.2000,     0.2125}
>IMM[2] FLT32 {    0.2125,     0.7154,     0.0721,     0.0000}
>  0: ADD TEMP[0].xy, IMM[0].xxxx, IN[1].xyyy
>  1: FSLT TEMP[1].xy, TEMP[0].xyyy, CONST[0].xxxx
>  2: AND TEMP[1].xy, TEMP[1].xyyy, IMM[0].yyyy
>  3: ABS TEMP[2].xy, TEMP[0].xyyy
>  4: ADD TEMP[0].xy, -TEMP[1].xyyy, TEMP[2].xyyy
>  5: MAD TEMP[1].x, TEMP[1].xxxx, IMM[0].zzzz, IMM[0].yyyy
>  6: ADD TEMP[0].xy, TEMP[0].xyyy, IMM[0].wwww
>  7: FSLT TEMP[2].xy, TEMP[0].xyyy, CONST[0].xxxx
>  8: AND TEMP[2].xy, TEMP[2].xyyy, IMM[0].yyyy
>  9: ABS TEMP[3].xy, TEMP[0].xyyy
> 10: ADD TEMP[0].xy, TEMP[3].xyyy, -TEMP[2].xyyy
> 11: MAD TEMP[2].xy, TEMP[2].xyyy, IMM[0].zzzz, IMM[0].yyyy
> 12: MAD TEMP[3].x, TEMP[0].xxxx, IMM[1].xxxx, IMM[0].yyyy
> 13: MAD TEMP[3].x, TEMP[0].yyyy, IMM[1].xxxx, TEMP[3].xxxx
> 14: MOV TEMP[3].z, TEMP[3].xxxx
> 15: MUL TEMP[3].xy, TEMP[0].xyyy, IMM[1].yyyy
> 16: DP3 TEMP[4].x, TEMP[3].xyzz, TEMP[3].xyzz
> 17: RSQ TEMP[4].x, TEMP[4].xxxx
> 18: MUL TEMP[3].xyz, TEMP[3].xyzz, TEMP[4].xxxx
> 19: MUL TEMP[2].xy, TEMP[2].xyyy, TEMP[3].xyyy
> 20: MUL TEMP[1].x, TEMP[1].xxxx, TEMP[3].zzzz
> 21: MOV TEMP[2].z, TEMP[1].xxxx
> 22: MAD TEMP[0].xyz, IN[5].xyzz, CONST[13].xxxx, TEMP[2].xyzz
> 23: DP3 TEMP[2].x, TEMP[0].xyzz, CONST[54].xyzz
> 24: DP3 TEMP[1].x, TEMP[0].xyzz, CONST[55].xyzz
> 25: MOV TEMP[2].y, TEMP[1].xxxx
> 26: DP3 TEMP[1].x, TEMP[0].xyzz, CONST[56].xyzz
> 27: MOV TEMP[2].z, TEMP[1].xxxx
> 28: DP3 TEMP[0].x, TEMP[2].xyzz, TEMP[2].xyzz
> 29: RSQ TEMP[1].x, TEMP[0].xxxx
> 30: MUL TEMP[1].xyz, TEMP[1].xxxx, TEMP[2].xyzz
> 31: ADD TEMP[0].xyz, IN[2].xyzz, IN[2].xyzz
> 32: LG2 TEMP[2].x, TEMP[0].xxxx
> 33: LG2 TEMP[3].x, TEMP[0].yyyy
> 34: MOV TEMP[2].y, TEMP[3].xxxx
> 35: LG2 TEMP[3].x, TEMP[0].zzzz
> 36: MOV TEMP[2].z, TEMP[3].xxxx
> 37: MUL TEMP[0].xyz, TEMP[2].xyzz, IMM[1].zzzz
> 38: EX2 TEMP[2].x, TEMP[0].xxxx
> 39: EX2 TEMP[3].x, TEMP[0].yyyy
> 40: MOV TEMP[2].y, TEMP[3].xxxx
> 41: EX2 TEMP[3].x, TEMP[0].zzzz
> 42: MOV TEMP[2].z, TEMP[3].xxxx
> 43: ABS TEMP[3].x, CONST[50].xxxx
> 44: FSLT TEMP[3].x, -TEMP[3].xxxx, TEMP[3].xxxx
> 45: AND TEMP[0].x, TEMP[3].xxxx, IMM[0].yyyy
> 46: MUL TEMP[3].xyz, TEMP[2].xyzz, TEMP[0].xxxx
> 47: DP3 TEMP[4].x, TEMP[2].xyzz, IMM[2].xyzz
> 48: ADD TEMP[5].x, CONST[0].yyyy, -IN[2].wwww
> 49: MUL TEMP[4].x, TEMP[4].xxxx, TEMP[5].xxxx
> 50: MUL TEMP[4].x, TEMP[4].xxxx, TEMP[0].xxxx
> 51: MOV TEMP[4].w, TEMP[4].xxxx
> 52: DP4 TEMP[5].x, IN[3], CONST[48]
> 53: DP4 TEMP[6].x, IN[3], CONST[49]
> 54: MOV TEMP[5].y, TEMP[6].xxxx
> 55: MOV TEMP[0].w, IN[0].wwww
> 56: MAD TEMP[0].xyz, IN[4].xyzz, CONST[13].xxxx, IN[0].xyzz
> 57: DP4 TEMP[2].x, TEMP[0], CONST[54]
> 58: DP4 TEMP[6].x, TEMP[0], CONST[55]
> 59: MOV TEMP[2].y, TEMP[6].xxxx
> 60: DP4 TEMP[0].x, TEMP[0], CONST[56]
> 61: MOV TEMP[2].z, TEMP[0].xxxx
> 62: ADD TEMP[0].xyz, -TEMP[2].xyzz, CONST[2].xyzz
> 63: MOV TEMP[2].w, CONST[0].yyyy
> 64: DP4 TEMP[6].x, TEMP[2], CONST[8]
> 65: DP4 TEMP[7].x, TEMP[2], CONST[9]
> 66: MOV TEMP[6].y, TEMP[7].xxxx
> 67: DP4 TEMP[8].x, TEMP[2], CONST[11]
> 68: MOV TEMP[6].w, TEMP[8].xxxx
> 69: DP4 TEMP[9].x, TEMP[2], CONST[10]
> 70: MOV TEMP[2].xyz, TEMP[2].xyzx
> 71: MOV TEMP[6].z, TEMP[9].xxxx
> 72: MOV TEMP[2].w, TEMP[9].xxxx
> 73: MOV TEMP[5].zw, CONST[0].xxxx
> 74: MOV TEMP[0].w, CONST[0].xxxx
> 75: MOV TEMP[1].w, CONST[0].xxxx
> 76: MOV TEMP[3].w, CONST[0].xxxx
> 77: MOV TEMP[4].xyz, CONST[0].xxxx
> 78: MOV TEMP[10], TEMP[6]
> 79: MAD TEMP[9].x, TEMP[9].xxxx, CONST[0].zzzz, -TEMP[8].xxxx
> 80: MOV TEMP[6].z, TEMP[9].xxxx
> 81: MOV TEMP[6].y, -TEMP[7].xxxx
> 82: MAD TEMP[6].xy, CONST[57].xyyy, TEMP[8].xxxx, TEMP[6].xyyy
> 83: MOV OUT[2], TEMP[5]
> 84: MOV OUT[3], TEMP[4]
> 85: MOV OUT[4], TEMP[0]
> 86: MOV OUT[5], TEMP[1]
> 87: MOV OUT[0], TEMP[6]
> 88: MOV OUT[1], TEMP[10]
> 89: MOV OUT[6], TEMP[3]
> 90: MOV OUT[7], TEMP[2]
> 91: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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 inreg, i32, i32, i32, i32) #0 {
>main_body:
>  %12 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %13 = load <16 x i8>, <16 x i8> addrspace(2)* %12, align 16, !tbaa !0
>  %14 = call float @llvm.SI.load.const(<16 x i8> %13, i32 0)
>  %15 = call float @llvm.SI.load.const(<16 x i8> %13, i32 4)
>  %16 = call float @llvm.SI.load.const(<16 x i8> %13, i32 8)
>  %17 = call float @llvm.SI.load.const(<16 x i8> %13, i32 32)
>  %18 = call float @llvm.SI.load.const(<16 x i8> %13, i32 36)
>  %19 = call float @llvm.SI.load.const(<16 x i8> %13, i32 40)
>  %20 = call float @llvm.SI.load.const(<16 x i8> %13, i32 128)
>  %21 = call float @llvm.SI.load.const(<16 x i8> %13, i32 132)
>  %22 = call float @llvm.SI.load.const(<16 x i8> %13, i32 136)
>  %23 = call float @llvm.SI.load.const(<16 x i8> %13, i32 140)
>  %24 = call float @llvm.SI.load.const(<16 x i8> %13, i32 144)
>  %25 = call float @llvm.SI.load.const(<16 x i8> %13, i32 148)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %13, i32 152)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %13, i32 156)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %13, i32 160)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %13, i32 164)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %13, i32 168)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %13, i32 172)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %13, i32 176)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %13, i32 180)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %13, i32 184)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %13, i32 188)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %13, i32 208)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %13, i32 768)
>  %38 = call float @llvm.SI.load.const(<16 x i8> %13, i32 772)
>  %39 = call float @llvm.SI.load.const(<16 x i8> %13, i32 776)
>  %40 = call float @llvm.SI.load.const(<16 x i8> %13, i32 780)
>  %41 = call float @llvm.SI.load.const(<16 x i8> %13, i32 784)
>  %42 = call float @llvm.SI.load.const(<16 x i8> %13, i32 788)
>  %43 = call float @llvm.SI.load.const(<16 x i8> %13, i32 792)
>  %44 = call float @llvm.SI.load.const(<16 x i8> %13, i32 796)
>  %45 = call float @llvm.SI.load.const(<16 x i8> %13, i32 800)
>  %46 = call float @llvm.SI.load.const(<16 x i8> %13, i32 864)
>  %47 = call float @llvm.SI.load.const(<16 x i8> %13, i32 868)
>  %48 = call float @llvm.SI.load.const(<16 x i8> %13, i32 872)
>  %49 = call float @llvm.SI.load.const(<16 x i8> %13, i32 876)
>  %50 = call float @llvm.SI.load.const(<16 x i8> %13, i32 880)
>  %51 = call float @llvm.SI.load.const(<16 x i8> %13, i32 884)
>  %52 = call float @llvm.SI.load.const(<16 x i8> %13, i32 888)
>  %53 = call float @llvm.SI.load.const(<16 x i8> %13, i32 892)
>  %54 = call float @llvm.SI.load.const(<16 x i8> %13, i32 896)
>  %55 = call float @llvm.SI.load.const(<16 x i8> %13, i32 900)
>  %56 = call float @llvm.SI.load.const(<16 x i8> %13, i32 904)
>  %57 = call float @llvm.SI.load.const(<16 x i8> %13, i32 908)
>  %58 = call float @llvm.SI.load.const(<16 x i8> %13, i32 912)
>  %59 = call float @llvm.SI.load.const(<16 x i8> %13, i32 916)
>  %60 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
>  %61 = load <16 x i8>, <16 x i8> addrspace(2)* %60, align 16, !tbaa !0
>  %62 = add i32 %5, %8
>  %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 = extractelement <4 x float> %63, i32 2
>  %67 = extractelement <4 x float> %63, i32 3
>  %68 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 1
>  %69 = load <16 x i8>, <16 x i8> addrspace(2)* %68, align 16, !tbaa !0
>  %70 = add i32 %5, %8
>  %71 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %69, i32 0, i32 %70)
>  %72 = extractelement <4 x float> %71, i32 0
>  %73 = extractelement <4 x float> %71, i32 1
>  %74 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 2
>  %75 = load <16 x i8>, <16 x i8> addrspace(2)* %74, align 16, !tbaa !0
>  %76 = add i32 %5, %8
>  %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 = extractelement <4 x float> %77, i32 2
>  %81 = extractelement <4 x float> %77, i32 3
>  %82 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 3
>  %83 = load <16 x i8>, <16 x i8> addrspace(2)* %82, align 16, !tbaa !0
>  %84 = add i32 %5, %8
>  %85 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %83, i32 0, i32 %84)
>  %86 = extractelement <4 x float> %85, i32 0
>  %87 = extractelement <4 x float> %85, i32 1
>  %88 = extractelement <4 x float> %85, i32 2
>  %89 = extractelement <4 x float> %85, i32 3
>  %90 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 4
>  %91 = load <16 x i8>, <16 x i8> addrspace(2)* %90, align 16, !tbaa !0
>  %92 = add i32 %5, %8
>  %93 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %91, i32 0, i32 %92)
>  %94 = extractelement <4 x float> %93, i32 0
>  %95 = extractelement <4 x float> %93, i32 1
>  %96 = extractelement <4 x float> %93, i32 2
>  %97 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 5
>  %98 = load <16 x i8>, <16 x i8> addrspace(2)* %97, align 16, !tbaa !0
>  %99 = add i32 %5, %8
>  %100 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %98, i32 0, i32 %99)
>  %101 = extractelement <4 x float> %100, i32 0
>  %102 = extractelement <4 x float> %100, i32 1
>  %103 = extractelement <4 x float> %100, i32 2
>  %104 = fadd float %72, -1.280000e+02
>  %105 = fadd float %73, -1.280000e+02
>  %106 = fcmp olt float %104, %14
>  %107 = fcmp olt float %105, %14
>  %108 = select i1 %106, float 1.000000e+00, float 0.000000e+00
>  %109 = call float @llvm.fabs.f32(float %104)
>  %110 = call float @llvm.fabs.f32(float %105)
>  %111 = fsub float %109, %108
>  %112 = select i1 %107, float -1.000000e+00, float -0.000000e+00
>  %113 = fadd float %112, %110
>  %114 = fmul float %108, -2.000000e+00
>  %115 = fadd float %114, 1.000000e+00
>  %116 = fadd float %111, -6.400000e+01
>  %117 = fadd float %113, -6.400000e+01
>  %118 = fcmp olt float %116, %14
>  %119 = fcmp olt float %117, %14
>  %120 = select i1 %118, float 1.000000e+00, float 0.000000e+00
>  %121 = select i1 %119, float 1.000000e+00, float 0.000000e+00
>  %122 = call float @llvm.fabs.f32(float %116)
>  %123 = call float @llvm.fabs.f32(float %117)
>  %124 = fsub float %122, %120
>  %125 = fsub float %123, %121
>  %126 = fmul float %120, -2.000000e+00
>  %127 = fadd float %126, 1.000000e+00
>  %128 = fmul float %121, -2.000000e+00
>  %129 = fadd float %128, 1.000000e+00
>  %130 = fmul float %124, 0xBF90410420000000
>  %131 = fadd float %130, 1.000000e+00
>  %132 = fmul float %125, 0xBF90410420000000
>  %133 = fadd float %132, %131
>  %134 = fmul float %124, 0x3F90410420000000
>  %135 = fmul float %125, 0x3F90410420000000
>  %136 = fmul float %134, %134
>  %137 = fmul float %135, %135
>  %138 = fadd float %137, %136
>  %139 = fmul float %133, %133
>  %140 = fadd float %138, %139
>  %141 = call float @llvm.AMDGPU.rsq.clamped.f32(float %140)
>  %142 = fmul float %134, %141
>  %143 = fmul float %135, %141
>  %144 = fmul float %133, %141
>  %145 = fmul float %127, %142
>  %146 = fmul float %129, %143
>  %147 = fmul float %115, %144
>  %148 = fmul float %101, %36
>  %149 = fadd float %148, %145
>  %150 = fmul float %102, %36
>  %151 = fadd float %150, %146
>  %152 = fmul float %103, %36
>  %153 = fadd float %152, %147
>  %154 = fmul float %149, %46
>  %155 = fmul float %151, %47
>  %156 = fadd float %155, %154
>  %157 = fmul float %153, %48
>  %158 = fadd float %156, %157
>  %159 = fmul float %149, %50
>  %160 = fmul float %151, %51
>  %161 = fadd float %160, %159
>  %162 = fmul float %153, %52
>  %163 = fadd float %161, %162
>  %164 = fmul float %149, %54
>  %165 = fmul float %151, %55
>  %166 = fadd float %165, %164
>  %167 = fmul float %153, %56
>  %168 = fadd float %166, %167
>  %169 = fmul float %158, %158
>  %170 = fmul float %163, %163
>  %171 = fadd float %170, %169
>  %172 = fmul float %168, %168
>  %173 = fadd float %171, %172
>  %174 = call float @llvm.AMDGPU.rsq.clamped.f32(float %173)
>  %175 = fmul float %174, %158
>  %176 = fmul float %174, %163
>  %177 = fmul float %174, %168
>  %178 = fadd float %78, %78
>  %179 = fadd float %79, %79
>  %180 = fadd float %80, %80
>  %181 = call float @llvm.log2.f32(float %178)
>  %182 = call float @llvm.log2.f32(float %179)
>  %183 = call float @llvm.log2.f32(float %180)
>  %184 = fmul float %181, 0x40019999A0000000
>  %185 = fmul float %182, 0x40019999A0000000
>  %186 = fmul float %183, 0x40019999A0000000
>  %187 = call float @llvm.AMDIL.exp.(float %184)
>  %188 = call float @llvm.AMDIL.exp.(float %185)
>  %189 = call float @llvm.AMDIL.exp.(float %186)
>  %190 = call float @llvm.fabs.f32(float %45)
>  %191 = fsub float -0.000000e+00, %190
>  %192 = fcmp ogt float %190, %191
>  %193 = select i1 %192, float 1.000000e+00, float 0.000000e+00
>  %194 = fmul float %187, %193
>  %195 = fmul float %188, %193
>  %196 = fmul float %189, %193
>  %197 = fmul float %187, 0x3FCB333340000000
>  %198 = fmul float %188, 0x3FE6E48E80000000
>  %199 = fadd float %198, %197
>  %200 = fmul float %189, 0x3FB2752540000000
>  %201 = fadd float %199, %200
>  %202 = fsub float %15, %81
>  %203 = fmul float %201, %202
>  %204 = fmul float %203, %193
>  %205 = fmul float %86, %37
>  %206 = fmul float %87, %38
>  %207 = fadd float %205, %206
>  %208 = fmul float %88, %39
>  %209 = fadd float %207, %208
>  %210 = fmul float %89, %40
>  %211 = fadd float %209, %210
>  %212 = fmul float %86, %41
>  %213 = fmul float %87, %42
>  %214 = fadd float %212, %213
>  %215 = fmul float %88, %43
>  %216 = fadd float %214, %215
>  %217 = fmul float %89, %44
>  %218 = fadd float %216, %217
>  %219 = fmul float %94, %36
>  %220 = fadd float %219, %64
>  %221 = fmul float %95, %36
>  %222 = fadd float %221, %65
>  %223 = fmul float %96, %36
>  %224 = fadd float %223, %66
>  %225 = fmul float %220, %46
>  %226 = fmul float %222, %47
>  %227 = fadd float %225, %226
>  %228 = fmul float %224, %48
>  %229 = fadd float %227, %228
>  %230 = fmul float %67, %49
>  %231 = fadd float %229, %230
>  %232 = fmul float %220, %50
>  %233 = fmul float %222, %51
>  %234 = fadd float %232, %233
>  %235 = fmul float %224, %52
>  %236 = fadd float %234, %235
>  %237 = fmul float %67, %53
>  %238 = fadd float %236, %237
>  %239 = fmul float %220, %54
>  %240 = fmul float %222, %55
>  %241 = fadd float %239, %240
>  %242 = fmul float %224, %56
>  %243 = fadd float %241, %242
>  %244 = fmul float %67, %57
>  %245 = fadd float %243, %244
>  %246 = fsub float %17, %231
>  %247 = fsub float %18, %238
>  %248 = fsub float %19, %245
>  %249 = fmul float %231, %20
>  %250 = fmul float %238, %21
>  %251 = fadd float %249, %250
>  %252 = fmul float %245, %22
>  %253 = fadd float %251, %252
>  %254 = fmul float %15, %23
>  %255 = fadd float %253, %254
>  %256 = fmul float %231, %24
>  %257 = fmul float %238, %25
>  %258 = fadd float %256, %257
>  %259 = fmul float %245, %26
>  %260 = fadd float %258, %259
>  %261 = fmul float %15, %27
>  %262 = fadd float %260, %261
>  %263 = fmul float %231, %32
>  %264 = fmul float %238, %33
>  %265 = fadd float %263, %264
>  %266 = fmul float %245, %34
>  %267 = fadd float %265, %266
>  %268 = fmul float %15, %35
>  %269 = fadd float %267, %268
>  %270 = fmul float %231, %28
>  %271 = fmul float %238, %29
>  %272 = fadd float %270, %271
>  %273 = fmul float %245, %30
>  %274 = fadd float %272, %273
>  %275 = fmul float %15, %31
>  %276 = fadd float %274, %275
>  %277 = fmul float %276, %16
>  %278 = fsub float %277, %269
>  %279 = fmul float %58, %269
>  %280 = fadd float %279, %255
>  %281 = fmul float %59, %269
>  %282 = fsub float %281, %262
>  %283 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 16
>  %284 = load <16 x i8>, <16 x i8> addrspace(2)* %283, align 16, !tbaa !0
>  %285 = call float @llvm.SI.load.const(<16 x i8> %284, i32 0)
>  %286 = fmul float %285, %255
>  %287 = call float @llvm.SI.load.const(<16 x i8> %284, i32 4)
>  %288 = fmul float %287, %262
>  %289 = fadd float %286, %288
>  %290 = call float @llvm.SI.load.const(<16 x i8> %284, i32 8)
>  %291 = fmul float %290, %276
>  %292 = fadd float %289, %291
>  %293 = call float @llvm.SI.load.const(<16 x i8> %284, i32 12)
>  %294 = fmul float %293, %269
>  %295 = fadd float %292, %294
>  %296 = call float @llvm.SI.load.const(<16 x i8> %284, i32 16)
>  %297 = fmul float %296, %255
>  %298 = call float @llvm.SI.load.const(<16 x i8> %284, i32 20)
>  %299 = fmul float %298, %262
>  %300 = fadd float %297, %299
>  %301 = call float @llvm.SI.load.const(<16 x i8> %284, i32 24)
>  %302 = fmul float %301, %276
>  %303 = fadd float %300, %302
>  %304 = call float @llvm.SI.load.const(<16 x i8> %284, i32 28)
>  %305 = fmul float %304, %269
>  %306 = fadd float %303, %305
>  %307 = call float @llvm.SI.load.const(<16 x i8> %284, i32 32)
>  %308 = fmul float %307, %255
>  %309 = call float @llvm.SI.load.const(<16 x i8> %284, i32 36)
>  %310 = fmul float %309, %262
>  %311 = fadd float %308, %310
>  %312 = call float @llvm.SI.load.const(<16 x i8> %284, i32 40)
>  %313 = fmul float %312, %276
>  %314 = fadd float %311, %313
>  %315 = call float @llvm.SI.load.const(<16 x i8> %284, i32 44)
>  %316 = fmul float %315, %269
>  %317 = fadd float %314, %316
>  %318 = call float @llvm.SI.load.const(<16 x i8> %284, i32 48)
>  %319 = fmul float %318, %255
>  %320 = call float @llvm.SI.load.const(<16 x i8> %284, i32 52)
>  %321 = fmul float %320, %262
>  %322 = fadd float %319, %321
>  %323 = call float @llvm.SI.load.const(<16 x i8> %284, i32 56)
>  %324 = fmul float %323, %276
>  %325 = fadd float %322, %324
>  %326 = call float @llvm.SI.load.const(<16 x i8> %284, i32 60)
>  %327 = fmul float %326, %269
>  %328 = fadd float %325, %327
>  %329 = call float @llvm.SI.load.const(<16 x i8> %284, i32 64)
>  %330 = fmul float %329, %255
>  %331 = call float @llvm.SI.load.const(<16 x i8> %284, i32 68)
>  %332 = fmul float %331, %262
>  %333 = fadd float %330, %332
>  %334 = call float @llvm.SI.load.const(<16 x i8> %284, i32 72)
>  %335 = fmul float %334, %276
>  %336 = fadd float %333, %335
>  %337 = call float @llvm.SI.load.const(<16 x i8> %284, i32 76)
>  %338 = fmul float %337, %269
>  %339 = fadd float %336, %338
>  %340 = call float @llvm.SI.load.const(<16 x i8> %284, i32 80)
>  %341 = fmul float %340, %255
>  %342 = call float @llvm.SI.load.const(<16 x i8> %284, i32 84)
>  %343 = fmul float %342, %262
>  %344 = fadd float %341, %343
>  %345 = call float @llvm.SI.load.const(<16 x i8> %284, i32 88)
>  %346 = fmul float %345, %276
>  %347 = fadd float %344, %346
>  %348 = call float @llvm.SI.load.const(<16 x i8> %284, i32 92)
>  %349 = fmul float %348, %269
>  %350 = fadd float %347, %349
>  %351 = call float @llvm.SI.load.const(<16 x i8> %284, i32 96)
>  %352 = fmul float %351, %255
>  %353 = call float @llvm.SI.load.const(<16 x i8> %284, i32 100)
>  %354 = fmul float %353, %262
>  %355 = fadd float %352, %354
>  %356 = call float @llvm.SI.load.const(<16 x i8> %284, i32 104)
>  %357 = fmul float %356, %276
>  %358 = fadd float %355, %357
>  %359 = call float @llvm.SI.load.const(<16 x i8> %284, i32 108)
>  %360 = fmul float %359, %269
>  %361 = fadd float %358, %360
>  %362 = call float @llvm.SI.load.const(<16 x i8> %284, i32 112)
>  %363 = fmul float %362, %255
>  %364 = call float @llvm.SI.load.const(<16 x i8> %284, i32 116)
>  %365 = fmul float %364, %262
>  %366 = fadd float %363, %365
>  %367 = call float @llvm.SI.load.const(<16 x i8> %284, i32 120)
>  %368 = fmul float %367, %276
>  %369 = fadd float %366, %368
>  %370 = call float @llvm.SI.load.const(<16 x i8> %284, i32 124)
>  %371 = fmul float %370, %269
>  %372 = fadd float %369, %371
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %211, float %218, float %14, float %14)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 33, i32 0, float %14, float %14, float %14, float %204)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 34, i32 0, float %246, float %247, float %248, float %14)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 35, i32 0, float %175, float %176, float %177, float %14)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 36, i32 0, float %194, float %195, float %196, float %14)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 37, i32 0, float %231, float %238, float %245, float %276)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 12, i32 0, float %280, float %282, float %278, float %269)
>  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 13, i32 0, float %295, float %306, float %317, float %328)
>  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 14, i32 0, float %339, float %350, float %361, float %372)
>  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.fabs.f32(float) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.AMDGPU.rsq.clamped.f32(float) #1
>
>; 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_mov_b32_e32 v1, 0xc3000000                            ; 7E0202FF C3000000
>	v_mov_b32_e32 v2, 0x80000000                            ; 7E0402FF 80000000
>	v_mov_b32_e32 v3, 0xc2800000                            ; 7E0602FF C2800000
>	v_mov_b32_e32 v4, 0x3c820821                            ; 7E0802FF 3C820821
>	v_mov_b32_e32 v5, 0xbc820821                            ; 7E0A02FF BC820821
>	v_mov_b32_e32 v6, 0x400ccccd                            ; 7E0C02FF 400CCCCD
>	v_add_i32_e32 v0, vcc, 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_load_dwordx4 s[28:31], s[2:3], 0x0                    ; C08E0300
>	s_load_dwordx4 s[60:63], s[2:3], 0x40                   ; C09E0340
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	buffer_load_format_xyzw v[7:10], v0, s[4:7], 0 idxen    ; E00C2000 80010700
>	buffer_load_format_xyzw v[11:14], v0, s[12:15], 0 idxen ; E00C2000 80030B00
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	buffer_load_format_xyzw v[13:16], v0, s[16:19], 0 idxen ; E00C2000 80040D00
>	buffer_load_format_xyzw v[17:20], v0, s[20:23], 0 idxen ; E00C2000 80051100
>	buffer_load_format_xyzw v[21:24], v0, s[24:27], 0 idxen ; E00C2000 80061500
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	buffer_load_format_xyzw v[24:27], v0, s[8:11], 0 idxen  ; E00C2000 80021800
>	s_buffer_load_dword s0, s[28:31], 0x0                   ; C2001D00
>	s_buffer_load_dword s43, s[28:31], 0x1                  ; C2159D01
>	s_buffer_load_dword s2, s[28:31], 0x2                   ; C2011D02
>	s_buffer_load_dword s44, s[28:31], 0x8                  ; C2161D08
>	s_buffer_load_dword s45, s[28:31], 0x9                  ; C2169D09
>	s_buffer_load_dword s47, s[28:31], 0x2e                 ; C2179D2E
>	s_buffer_load_dword s1, s[28:31], 0x2f                  ; C2009D2F
>	s_buffer_load_dword s38, s[28:31], 0x34                 ; C2131D34
>	s_buffer_load_dword s58, s[28:31], 0xc0                 ; C21D1DC0
>	s_buffer_load_dword s59, s[28:31], 0xc1                 ; C21D9DC1
>	s_buffer_load_dword s64, s[28:31], 0xc7                 ; C2201DC7
>	s_buffer_load_dword s6, s[28:31], 0xc8                  ; C2031DC8
>	s_buffer_load_dword s9, s[28:31], 0xd8                  ; C2049DD8
>	s_buffer_load_dword s7, s[28:31], 0xd9                  ; C2039DD9
>	s_buffer_load_dword s3, s[28:31], 0xda                  ; C2019DDA
>	s_buffer_load_dword s65, s[28:31], 0xdb                 ; C2209DDB
>	s_buffer_load_dword s11, s[28:31], 0xdc                 ; C2059DDC
>	s_buffer_load_dword s8, s[28:31], 0xdd                  ; C2041DDD
>	s_buffer_load_dword s4, s[28:31], 0xde                  ; C2021DDE
>	s_buffer_load_dword s66, s[28:31], 0xdf                 ; C2211DDF
>	s_buffer_load_dword s12, s[28:31], 0xe0                 ; C2061DE0
>	s_buffer_load_dword s10, s[28:31], 0xe1                 ; C2051DE1
>	s_buffer_load_dword s5, s[28:31], 0xe2                  ; C2029DE2
>	s_buffer_load_dword s67, s[28:31], 0xe3                 ; C2219DE3
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	v_cmp_gt_f32_e64 s[14:15], |s6|, -|s6|                  ; D008030E 40000C06
>	s_buffer_load_dword s6, s[28:31], 0xe4                  ; C2031DE4
>	s_buffer_load_dword s68, s[28:31], 0xc2                 ; C2221DC2
>	s_buffer_load_dword s69, s[28:31], 0xc3                 ; C2229DC3
>	s_buffer_load_dword s70, s[28:31], 0xc4                 ; C2231DC4
>	v_cndmask_b32_e64 v0, 0, 1.0, s[14:15]                  ; D2000000 0039E480
>	s_buffer_load_dword s71, s[28:31], 0xc5                 ; C2239DC5
>	s_buffer_load_dword s72, s[28:31], 0xc6                 ; C2241DC6
>	s_waitcnt vmcnt(0)                                      ; BF8C0770
>	v_mov_b32_e32 v27, s1                                   ; 7E360201
>	s_buffer_load_dword s57, s[28:31], 0x24                 ; C21C9D24
>	v_mov_b32_e32 v28, s0                                   ; 7E380200
>	s_buffer_load_dword s73, s[28:31], 0x25                 ; C2249D25
>	s_buffer_load_dword s50, s[28:31], 0x26                 ; C2191D26
>	s_buffer_load_dword s1, s[28:31], 0x27                  ; C2009D27
>	s_buffer_load_dword s51, s[28:31], 0x28                 ; C2199D28
>	s_buffer_load_dword s48, s[28:31], 0xa                  ; C2181D0A
>	s_buffer_load_dword s74, s[28:31], 0x20                 ; C2251D20
>	s_buffer_load_dword s75, s[28:31], 0x21                 ; C2259D21
>	s_buffer_load_dword s52, s[28:31], 0x22                 ; C21A1D22
>	s_buffer_load_dword s13, s[28:31], 0x23                 ; C2069D23
>	s_buffer_load_dword s55, s[28:31], 0x29                 ; C21B9D29
>	s_buffer_load_dword s53, s[28:31], 0x2a                 ; C21A9D2A
>	s_buffer_load_dword s14, s[28:31], 0x2b                 ; C2071D2B
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	v_mov_b32_e32 v29, s1                                   ; 7E3A0201
>	s_buffer_load_dword s76, s[28:31], 0x2c                 ; C2261D2C
>	s_buffer_load_dword s77, s[28:31], 0x2d                 ; C2269D2D
>	s_buffer_load_dword s21, s[28:31], 0xe5                 ; C20A9DE5
>	s_buffer_load_dword s26, s[60:63], 0x0                  ; C20D3D00
>	s_buffer_load_dword s56, s[60:63], 0x1                  ; C21C3D01
>	s_buffer_load_dword s22, s[60:63], 0x2                  ; C20B3D02
>	v_mov_b32_e32 v30, s13                                  ; 7E3C020D
>	s_buffer_load_dword s18, s[60:63], 0x3                  ; C2093D03
>	s_buffer_load_dword s28, s[60:63], 0x4                  ; C20E3D04
>	s_buffer_load_dword s54, s[60:63], 0x5                  ; C21B3D05
>	v_mov_b32_e32 v31, s14                                  ; 7E3E020E
>	s_buffer_load_dword s23, s[60:63], 0x6                  ; C20BBD06
>	s_buffer_load_dword s13, s[60:63], 0x7                  ; C206BD07
>	s_buffer_load_dword s31, s[60:63], 0x8                  ; C20FBD08
>	s_buffer_load_dword s49, s[60:63], 0x9                  ; C218BD09
>	s_buffer_load_dword s24, s[60:63], 0xa                  ; C20C3D0A
>	s_buffer_load_dword s14, s[60:63], 0xb                  ; C2073D0B
>	s_buffer_load_dword s33, s[60:63], 0xc                  ; C210BD0C
>	s_buffer_load_dword s46, s[60:63], 0xd                  ; C2173D0D
>	s_buffer_load_dword s25, s[60:63], 0xe                  ; C20CBD0E
>	s_buffer_load_dword s16, s[60:63], 0xf                  ; C2083D0F
>	s_buffer_load_dword s34, s[60:63], 0x10                 ; C2113D10
>	s_buffer_load_dword s42, s[60:63], 0x11                 ; C2153D11
>	s_buffer_load_dword s27, s[60:63], 0x12                 ; C20DBD12
>	s_buffer_load_dword s15, s[60:63], 0x13                 ; C207BD13
>	s_buffer_load_dword s35, s[60:63], 0x14                 ; C211BD14
>	s_buffer_load_dword s41, s[60:63], 0x15                 ; C214BD15
>	s_buffer_load_dword s29, s[60:63], 0x16                 ; C20EBD16
>	s_buffer_load_dword s17, s[60:63], 0x17                 ; C208BD17
>	s_buffer_load_dword s36, s[60:63], 0x18                 ; C2123D18
>	s_buffer_load_dword s40, s[60:63], 0x19                 ; C2143D19
>	s_buffer_load_dword s30, s[60:63], 0x1a                 ; C20F3D1A
>	s_buffer_load_dword s19, s[60:63], 0x1b                 ; C209BD1B
>	s_buffer_load_dword s37, s[60:63], 0x1c                 ; C212BD1C
>	s_buffer_load_dword s39, s[60:63], 0x1d                 ; C213BD1D
>	s_buffer_load_dword s32, s[60:63], 0x1e                 ; C2103D1E
>	s_buffer_load_dword s20, s[60:63], 0x1f                 ; C20A3D1F
>	v_add_f32_e32 v12, v1, v12                              ; 06181901
>	v_cmp_gt_f32_e32 vcc, s0, v12                           ; 7C081800
>	v_cndmask_b32_e64 v2, v2, -1.0, vcc                     ; D2000002 01A9E702
>	v_add_f32_e32 v1, v1, v11                               ; 06021701
>	v_cmp_gt_f32_e32 vcc, s0, v1                            ; 7C080200
>	v_cndmask_b32_e64 v11, 0, 1.0, vcc                      ; D200000B 01A9E480
>	v_add_f32_e64 v2, v2, |v12|                             ; D2060202 00021902
>	v_sub_f32_e64 v1, |v1|, v11                             ; D2080101 00021701
>	v_add_f32_e32 v1, v3, v1                                ; 06020303
>	v_add_f32_e32 v2, v3, v2                                ; 06040503
>	v_cmp_gt_f32_e32 vcc, s0, v1                            ; 7C080200
>	v_cmp_gt_f32_e64 s[0:1], s0, v2                         ; D0080000 00020400
>	v_mul_f32_e32 v3, s59, v18                              ; 1006243B
>	v_mul_f32_e32 v12, s71, v18                             ; 10182447
>	v_mac_f32_e32 v3, s58, v17                              ; 3E06223A
>	v_mac_f32_e32 v12, s70, v17                             ; 3E182246
>	v_mac_f32_e32 v3, s68, v19                              ; 3E062644
>	v_mac_f32_e32 v12, s72, v19                             ; 3E182648
>	v_mac_f32_e32 v3, s69, v20                              ; 3E062845
>	v_mac_f32_e32 v12, s64, v20                             ; 3E182840
>	v_mad_f32 v7, s38, v21, v7                              ; D2820007 041E2A26
>	v_mad_f32 v8, s38, v22, v8                              ; D2820008 04222C26
>	v_mad_f32 v9, s38, v23, v9                              ; D2820009 04262E26
>	v_mul_f32_e32 v17, s7, v8                               ; 10221007
>	v_mac_f32_e32 v17, s9, v7                               ; 3E220E09
>	v_mac_f32_e32 v17, s3, v9                               ; 3E221203
>	v_mac_f32_e32 v17, s65, v10                             ; 3E221441
>	v_mul_f32_e32 v18, s8, v8                               ; 10241008
>	v_mac_f32_e32 v18, s11, v7                              ; 3E240E0B
>	v_mac_f32_e32 v18, s4, v9                               ; 3E241204
>	v_mac_f32_e32 v18, s66, v10                             ; 3E241442
>	v_mul_f32_e32 v8, s10, v8                               ; 1010100A
>	v_mac_f32_e32 v8, s12, v7                               ; 3E100E0C
>	v_mac_f32_e32 v8, s5, v9                                ; 3E101205
>	v_mac_f32_e32 v8, s67, v10                              ; 3E101443
>	v_mul_f32_e32 v7, s75, v18                              ; 100E244B
>	v_mac_f32_e32 v7, s74, v17                              ; 3E0E224A
>	v_mul_f32_e32 v9, s73, v18                              ; 10122449
>	v_mac_f32_e32 v9, s57, v17                              ; 3E122239
>	s_waitcnt lgkmcnt(0)                                    ; BF8C007F
>	v_mul_f32_e32 v10, s77, v18                             ; 1014244D
>	v_mac_f32_e32 v10, s76, v17                             ; 3E14224C
>	v_mul_f32_e32 v19, s55, v18                             ; 10262437
>	v_mac_f32_e32 v19, s51, v17                             ; 3E262233
>	v_mac_f32_e32 v7, s52, v8                               ; 3E0E1034
>	v_mac_f32_e32 v9, s50, v8                               ; 3E121032
>	v_mac_f32_e32 v10, s47, v8                              ; 3E14102F
>	v_mac_f32_e32 v19, s53, v8                              ; 3E261035
>	v_mac_f32_e32 v7, s43, v30                              ; 3E0E3C2B
>	v_mac_f32_e32 v9, s43, v29                              ; 3E123A2B
>	v_mac_f32_e32 v10, s43, v27                             ; 3E14362B
>	v_mac_f32_e32 v19, s43, v31                             ; 3E263E2B
>	v_sub_f32_e32 v16, s43, v16                             ; 0820202B
>	v_sub_f32_e32 v20, s44, v17                             ; 0828222C
>	v_sub_f32_e32 v21, s45, v18                             ; 082A242D
>	v_sub_f32_e32 v22, s48, v8                              ; 082C1030
>	v_mul_f32_e32 v23, s56, v9                              ; 102E1238
>	v_mul_f32_e32 v27, s54, v9                              ; 10361236
>	v_cndmask_b32_e64 v29, 0, 1.0, vcc                      ; D200001D 01A9E480
>	v_cndmask_b32_e64 v30, 0, 1.0, s[0:1]                   ; D200001E 0001E480
>	v_sub_f32_e64 v1, |v1|, v29                             ; D2080101 00023B01
>	v_sub_f32_e64 v2, |v2|, v30                             ; D2080102 00023D02
>	v_mad_f32 v31, v1, v5, 1.0                              ; D282001F 03CA0B01
>	v_mac_f32_e32 v31, v5, v2                               ; 3E3E0505
>	v_mul_f32_e32 v1, v4, v1                                ; 10020304
>	v_mul_f32_e32 v2, v4, v2                                ; 10040504
>	v_mul_f32_e32 v4, v1, v1                                ; 10080301
>	v_mac_f32_e32 v4, v2, v2                                ; 3E080502
>	v_mac_f32_e32 v4, v31, v31                              ; 3E083F1F
>	v_rsq_clamp_f32_e32 v4, v4                              ; 7E085904
>	v_mul_f32_e32 v5, s49, v9                               ; 100A1231
>	v_mul_f32_e32 v32, s46, v9                              ; 1040122E
>	v_mul_f32_e32 v33, s42, v9                              ; 1042122A
>	v_mul_f32_e32 v1, v4, v1                                ; 10020304
>	v_mul_f32_e32 v2, v4, v2                                ; 10040504
>	v_mul_f32_e32 v4, v4, v31                               ; 10083F04
>	v_mul_f32_e32 v29, -2.0, v29                            ; 103A3AF5
>	v_mac_f32_e32 v1, v1, v29                               ; 3E023B01
>	v_mul_f32_e32 v29, -2.0, v30                            ; 103A3CF5
>	v_mac_f32_e32 v2, v2, v29                               ; 3E043B02
>	v_mul_f32_e32 v11, -2.0, v11                            ; 101616F5
>	v_mac_f32_e32 v4, v4, v11                               ; 3E081704
>	v_mac_f32_e32 v1, s38, v24                              ; 3E023026
>	v_mac_f32_e32 v2, s38, v25                              ; 3E043226
>	v_mac_f32_e32 v4, s38, v26                              ; 3E083426
>	v_mul_f32_e32 v11, s41, v9                              ; 10161229
>	v_mul_f32_e32 v24, s40, v9                              ; 10301228
>	v_add_f32_e32 v13, v13, v13                             ; 061A1B0D
>	v_add_f32_e32 v14, v14, v14                             ; 061C1D0E
>	v_add_f32_e32 v15, v15, v15                             ; 061E1F0F
>	v_log_f32_e32 v13, v13                                  ; 7E1A4F0D
>	v_log_f32_e32 v14, v14                                  ; 7E1C4F0E
>	v_log_f32_e32 v15, v15                                  ; 7E1E4F0F
>	exp 15, 32, 0, 0, 0, v3, v12, v28, v28                  ; F800020F 1C1C0C03
>	s_waitcnt expcnt(0)                                     ; BF8C070F
>	v_mul_f32_e32 v3, v6, v13                               ; 10061B06
>	v_mul_f32_e32 v12, v6, v14                              ; 10181D06
>	v_mul_f32_e32 v6, v6, v15                               ; 100C1F06
>	v_exp_f32_e32 v3, v3                                    ; 7E064B03
>	v_exp_f32_e32 v12, v12                                  ; 7E184B0C
>	v_exp_f32_e32 v6, v6                                    ; 7E0C4B06
>	v_mul_f32_e32 v13, 0x3e59999a, v3                       ; 101A06FF 3E59999A
>	v_madmk_f32_e32 v13, v12, v13, 0x3f372474               ; 401A1B0C 3F372474
>	v_madmk_f32_e32 v13, v6, v13, 0x3d93a92a                ; 401A1B06 3D93A92A
>	v_mul_f32_e32 v13, v16, v13                             ; 101A1B10
>	v_mul_f32_e32 v13, v0, v13                              ; 101A1B00
>	exp 15, 33, 0, 0, 0, v28, v28, v28, v13                 ; F800021F 0D1C1C1C
>	exp 15, 34, 0, 0, 0, v20, v21, v22, v28                 ; F800022F 1C161514
>	s_waitcnt expcnt(0)                                     ; BF8C070F
>	v_mul_f32_e32 v13, s9, v1                               ; 101A0209
>	v_mul_f32_e32 v14, s11, v1                              ; 101C020B
>	v_mul_f32_e32 v1, s12, v1                               ; 1002020C
>	v_mac_f32_e32 v13, s7, v2                               ; 3E1A0407
>	v_mac_f32_e32 v14, s8, v2                               ; 3E1C0408
>	v_mac_f32_e32 v1, s10, v2                               ; 3E02040A
>	v_mul_f32_e32 v2, s39, v9                               ; 10041227
>	v_mac_f32_e32 v23, s26, v7                              ; 3E2E0E1A
>	v_mac_f32_e32 v27, s28, v7                              ; 3E360E1C
>	v_mac_f32_e32 v5, s31, v7                               ; 3E0A0E1F
>	v_mac_f32_e32 v32, s33, v7                              ; 3E400E21
>	v_mac_f32_e32 v33, s34, v7                              ; 3E420E22
>	v_mac_f32_e32 v11, s35, v7                              ; 3E160E23
>	v_mac_f32_e32 v24, s36, v7                              ; 3E300E24
>	v_mac_f32_e32 v2, s37, v7                               ; 3E040E25
>	v_mad_f32 v7, s6, v10, v7                               ; D2820007 041E1406
>	v_mad_f32 v9, s21, v10, -v9                             ; D2820009 84261415
>	v_mac_f32_e32 v23, s22, v19                             ; 3E2E2616
>	v_mac_f32_e32 v27, s23, v19                             ; 3E362617
>	v_mac_f32_e32 v5, s24, v19                              ; 3E0A2618
>	v_mac_f32_e32 v32, s25, v19                             ; 3E402619
>	v_mac_f32_e32 v33, s27, v19                             ; 3E42261B
>	v_mac_f32_e32 v11, s29, v19                             ; 3E16261D
>	v_mac_f32_e32 v24, s30, v19                             ; 3E30261E
>	v_mac_f32_e32 v2, s32, v19                              ; 3E042620
>	v_mac_f32_e32 v23, s18, v10                             ; 3E2E1412
>	v_mac_f32_e32 v13, s3, v4                               ; 3E1A0803
>	v_mac_f32_e32 v14, s4, v4                               ; 3E1C0804
>	v_mac_f32_e32 v1, s5, v4                                ; 3E020805
>	v_mul_f32_e32 v4, v13, v13                              ; 10081B0D
>	v_mac_f32_e32 v4, v14, v14                              ; 3E081D0E
>	v_mac_f32_e32 v4, v1, v1                                ; 3E080301
>	v_rsq_clamp_f32_e32 v4, v4                              ; 7E085904
>	v_mac_f32_e32 v27, s13, v10                             ; 3E36140D
>	v_mac_f32_e32 v5, s14, v10                              ; 3E0A140E
>	v_mac_f32_e32 v32, s16, v10                             ; 3E401410
>	v_mul_f32_e32 v13, v13, v4                              ; 101A090D
>	v_mul_f32_e32 v14, v14, v4                              ; 101C090E
>	v_mul_f32_e32 v1, v1, v4                                ; 10020901
>	exp 15, 35, 0, 0, 0, v13, v14, v1, v28                  ; F800023F 1C010E0D
>	v_mac_f32_e32 v33, s15, v10                             ; 3E42140F
>	v_mac_f32_e32 v11, s17, v10                             ; 3E161411
>	v_mac_f32_e32 v24, s19, v10                             ; 3E301413
>	v_mac_f32_e32 v2, s20, v10                              ; 3E041414
>	s_waitcnt expcnt(0)                                     ; BF8C070F
>	v_mul_f32_e32 v1, v0, v3                                ; 10020700
>	v_mul_f32_e32 v3, v0, v12                               ; 10061900
>	v_mad_f32 v4, v19, s2, -v10                             ; D2820004 84280513
>	v_mul_f32_e32 v0, v0, v6                                ; 10000D00
>	exp 15, 36, 0, 0, 0, v1, v3, v0, v28                    ; F800024F 1C000301
>	exp 15, 37, 0, 0, 0, v17, v18, v8, v19                  ; F800025F 13081211
>	exp 15, 12, 0, 0, 0, v7, v9, v4, v10                    ; F80000CF 0A040907
>	exp 15, 13, 0, 0, 0, v23, v27, v5, v32                  ; F80000DF 20051B17
>	exp 15, 14, 0, 1, 0, v33, v11, v24, v2                  ; F80008EF 02180B21
>	s_endpgm                                                ; BF810000
>
>*** SHADER STATS ***
>SGPRS: 80
>VGPRS: 36
>Code Size: 1272 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
>  clamp_color = 0
>FRAG
>DCL IN[0], GENERIC[0], PERSPECTIVE
>DCL IN[1], GENERIC[1], PERSPECTIVE
>DCL IN[2], GENERIC[2], PERSPECTIVE
>DCL IN[3], GENERIC[3], PERSPECTIVE
>DCL IN[4], GENERIC[4], PERSPECTIVE
>DCL IN[5], GENERIC[5], PERSPECTIVE
>DCL OUT[0], COLOR
>DCL SAMP[0]
>DCL SAMP[1]
>DCL SAMP[2]
>DCL SAMP[3]
>DCL SVIEW[0], 2D, FLOAT
>DCL SVIEW[1], CUBE, FLOAT
>DCL SVIEW[2], 2D, FLOAT
>DCL SVIEW[3], SHADOW2D, FLOAT
>DCL CONST[0..90]
>DCL TEMP[0..13], LOCAL
>IMM[0] FLT32 {    1.0000,     0.0000,     2.0000,    -0.5000}
>IMM[1] FLT32 {   -0.0000,    -1.0000,    -2.0000,     0.0625}
>IMM[2] FLT32 {    0.0005,     0.0000,    -0.0005,     0.1250}
>IMM[3] FLT32 {    0.2500,     0.0000,    -1.0000,    -2.0000}
>IMM[4] FLT32 {    0.2125,     0.7154,     0.0721,     0.5000}
>IMM[5] FLT32 {    0.2990,     0.5870,     0.1140,     0.0000}
>  0: MOV TEMP[0].xy, IN[0].xyyy
>  1: TEX TEMP[0], TEMP[0], SAMP[0], 2D
>  2: MOV TEMP[1].xy, IN[0].xyyy
>  3: TEX TEMP[1], TEMP[1], SAMP[2], 2D
>  4: MOV TEMP[1], TEMP[1]
>  5: UIF CONST[90].xxxx :0
>  6:   MAD TEMP[2], IN[5].xyzx, IMM[0].xxxy, IMM[0].yyyx
>  7:   DP4 TEMP[3].x, TEMP[2], CONST[69]
>  8:   DP4 TEMP[4].x, TEMP[2], CONST[70]
>  9:   MOV TEMP[3].y, TEMP[4].xxxx
> 10:   MOV_SAT TEMP[5].xy, TEMP[3].xyyy
> 11:   ADD TEMP[5].xy, -TEMP[3].xyyy, TEMP[5].xyyy
> 12:   MOV TEMP[1].yz, TEMP[5].yxyy
> 13:   DP2 TEMP[5].x, TEMP[5].xyyy, IMM[0].xxxx
> 14:   MOV TEMP[1].y, TEMP[5].xxxx
> 15:   DP4 TEMP[6].x, TEMP[2], CONST[73]
> 16:   DP4 TEMP[7].x, TEMP[2], CONST[74]
> 17:   MOV TEMP[6].y, TEMP[7].xxxx
> 18:   MOV_SAT TEMP[8].xy, TEMP[6].xyyy
> 19:   ADD TEMP[8].xy, -TEMP[6].xyyy, TEMP[8].xyyy
> 20:   MOV TEMP[1].zw, TEMP[8].yyxy
> 21:   DP2 TEMP[8].x, TEMP[8].xyyy, IMM[0].xxxx
> 22:   MOV TEMP[1].z, TEMP[8].xxxx
> 23:   DP4 TEMP[9].x, TEMP[2], CONST[77]
> 24:   DP4 TEMP[10].x, TEMP[2], CONST[78]
> 25:   MOV TEMP[6].z, IMM[0].xxxx
> 26:   MOV TEMP[9].z, IMM[0].zzzz
> 27:   MOV TEMP[11].w, TEMP[6]
> 28:   ABS TEMP[12].x, TEMP[8].xxxx
> 29:   FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
> 30:   UIF TEMP[12].xxxx :0
> 31:     MOV TEMP[12].x, TEMP[6].xxxx
> 32:   ELSE :0
> 33:     MOV TEMP[12].x, TEMP[9].xxxx
> 34:   ENDIF
> 35:   MOV TEMP[11].x, TEMP[12].xxxx
> 36:   ABS TEMP[12].x, TEMP[8].xxxx
> 37:   FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
> 38:   UIF TEMP[12].xxxx :0
> 39:     MOV TEMP[7].x, TEMP[7].xxxx
> 40:   ELSE :0
> 41:     MOV TEMP[7].x, TEMP[10].xxxx
> 42:   ENDIF
> 43:   MOV TEMP[11].y, TEMP[7].xxxx
> 44:   ABS TEMP[7].x, TEMP[8].xxxx
> 45:   FSGE TEMP[7].x, -TEMP[7].xxxx, IMM[0].yyyy
> 46:   UIF TEMP[7].xxxx :0
> 47:     MOV TEMP[7].x, IMM[0].xxxx
> 48:   ELSE :0
> 49:     MOV TEMP[7].x, IMM[0].zzzz
> 50:   ENDIF
> 51:   MOV TEMP[11].z, TEMP[7].xxxx
> 52:   MOV TEMP[6].xyz, TEMP[11]
> 53:   MOV TEMP[3].w, IMM[0].yyyy
> 54:   MOV TEMP[11].x, TEMP[1]
> 55:   ABS TEMP[7].x, TEMP[5].xxxx
> 56:   FSGE TEMP[7].x, -TEMP[7].xxxx, IMM[0].yyyy
> 57:   UIF TEMP[7].xxxx :0
> 58:     MOV TEMP[7].x, TEMP[3].xxxx
> 59:   ELSE :0
> 60:     MOV TEMP[7].x, TEMP[6].xxxx
> 61:   ENDIF
> 62:   MOV TEMP[11].y, TEMP[7].xxxx
> 63:   ABS TEMP[7].x, TEMP[5].xxxx
> 64:   FSGE TEMP[7].x, -TEMP[7].xxxx, IMM[0].yyyy
> 65:   UIF TEMP[7].xxxx :0
> 66:     MOV TEMP[4].x, TEMP[4].xxxx
> 67:   ELSE :0
> 68:     MOV TEMP[4].x, TEMP[6].yyyy
> 69:   ENDIF
> 70:   MOV TEMP[11].z, TEMP[4].xxxx
> 71:   ABS TEMP[4].x, TEMP[5].xxxx
> 72:   FSGE TEMP[4].x, -TEMP[4].xxxx, IMM[0].yyyy
> 73:   UIF TEMP[4].xxxx :0
> 74:     MOV TEMP[4].x, IMM[0].yyyy
> 75:   ELSE :0
> 76:     MOV TEMP[4].x, TEMP[6].zzzz
> 77:   ENDIF
> 78:   MOV TEMP[11].w, TEMP[4].xxxx
> 79:   MOV TEMP[1].xw, TEMP[11].xxxw
> 80:   DP4 TEMP[5].x, TEMP[2], CONST[71]
> 81:   MOV TEMP[3].z, TEMP[5].xxxx
> 82:   ADD TEMP[6].xy, TEMP[11].yzzz, IMM[0].wwww
> 83:   ABS TEMP[7].xy, TEMP[6].xyyy
> 84:   ADD TEMP[6].xy, TEMP[7].xyyy, -CONST[67].zzzz
> 85:   MUL TEMP[6].xy, TEMP[6].xyyy, CONST[67].wwww
> 86:   MOV_SAT TEMP[7].xy, TEMP[6].xyyy
> 87:   ADD TEMP[6].xy, -TEMP[7].xyyy, IMM[0].xxxx
> 88:   MUL TEMP[6].x, TEMP[6].yyyy, TEMP[6].xxxx
> 89:   MOV_SAT TEMP[7].xy, TEMP[11].yzzz
> 90:   ADD TEMP[4].xyz, TEMP[4].xxxx, IMM[1].xyzz
> 91:   MOV TEMP[9].y, IMM[0].yyyy
> 92:   ABS TEMP[8].x, TEMP[4].xxxx
> 93:   FSGE TEMP[8].x, -TEMP[8].xxxx, IMM[0].yyyy
> 94:   UIF TEMP[8].xxxx :0
> 95:     MOV TEMP[8].x, CONST[85].zzzz
> 96:   ELSE :0
> 97:     MOV TEMP[8].x, IMM[0].yyyy
> 98:   ENDIF
> 99:   ABS TEMP[10].x, TEMP[4].xxxx
>100:   FSGE TEMP[10].x, -TEMP[10].xxxx, IMM[0].yyyy
>101:   UIF TEMP[10].xxxx :0
>102:     MOV TEMP[10].x, CONST[85].wwww
>103:   ELSE :0
>104:     MOV TEMP[10].x, IMM[0].yyyy
>105:   ENDIF
>106:   MOV TEMP[11].y, TEMP[10].xxxx
>107:   ABS TEMP[10].x, TEMP[4].xxxx
>108:   FSGE TEMP[10].x, -TEMP[10].xxxx, IMM[0].yyyy
>109:   UIF TEMP[10].xxxx :0
>110:     MOV TEMP[10].x, CONST[85].xxxx
>111:   ELSE :0
>112:     MOV TEMP[10].x, IMM[0].yyyy
>113:   ENDIF
>114:   MOV TEMP[11].z, TEMP[10].xxxx
>115:   ABS TEMP[10].x, TEMP[4].xxxx
>116:   FSGE TEMP[10].x, -TEMP[10].xxxx, IMM[0].yyyy
>117:   UIF TEMP[10].xxxx :0
>118:     MOV TEMP[10].x, CONST[85].yyyy
>119:   ELSE :0
>120:     MOV TEMP[10].x, IMM[0].yyyy
>121:   ENDIF
>122:   MOV TEMP[11].w, TEMP[10].xxxx
>123:   ABS TEMP[10].x, TEMP[4].yyyy
>124:   FSGE TEMP[10].x, -TEMP[10].xxxx, IMM[0].yyyy
>125:   UIF TEMP[10].xxxx :0
>126:     MOV TEMP[10].x, CONST[86].zzzz
>127:   ELSE :0
>128:     MOV TEMP[10].x, TEMP[8].xxxx
>129:   ENDIF
>130:   ABS TEMP[8].x, TEMP[4].yyyy
>131:   FSGE TEMP[8].x, -TEMP[8].xxxx, IMM[0].yyyy
>132:   UIF TEMP[8].xxxx :0
>133:     MOV TEMP[8].x, CONST[86].wwww
>134:   ELSE :0
>135:     MOV TEMP[8].x, TEMP[11].yyyy
>136:   ENDIF
>137:   MOV TEMP[11].y, TEMP[8].xxxx
>138:   ABS TEMP[8].x, TEMP[4].yyyy
>139:   FSGE TEMP[8].x, -TEMP[8].xxxx, IMM[0].yyyy
>140:   UIF TEMP[8].xxxx :0
>141:     MOV TEMP[8].x, CONST[86].xxxx
>142:   ELSE :0
>143:     MOV TEMP[8].x, TEMP[11].zzzz
>144:   ENDIF
>145:   MOV TEMP[11].z, TEMP[8].xxxx
>146:   ABS TEMP[8].x, TEMP[4].yyyy
>147:   FSGE TEMP[8].x, -TEMP[8].xxxx, IMM[0].yyyy
>148:   UIF TEMP[8].xxxx :0
>149:     MOV TEMP[8].x, CONST[86].yyyy
>150:   ELSE :0
>151:     MOV TEMP[8].x, TEMP[11].wwww
>152:   ENDIF
>153:   MOV TEMP[11].w, TEMP[8].xxxx
>154:   ABS TEMP[8].x, TEMP[4].zzzz
>155:   FSGE TEMP[8].x, -TEMP[8].xxxx, IMM[0].yyyy
>156:   UIF TEMP[8].xxxx :0
>157:     MOV TEMP[8].x, CONST[87].zzzz
>158:   ELSE :0
>159:     MOV TEMP[8].x, TEMP[10].xxxx
>160:   ENDIF
>161:   MOV TEMP[11].x, TEMP[8].xxxx
>162:   ABS TEMP[8].x, TEMP[4].zzzz
>163:   FSGE TEMP[8].x, -TEMP[8].xxxx, IMM[0].yyyy
>164:   UIF TEMP[8].xxxx :0
>165:     MOV TEMP[8].x, CONST[87].wwww
>166:   ELSE :0
>167:     MOV TEMP[8].x, TEMP[11].yyyy
>168:   ENDIF
>169:   MOV TEMP[11].y, TEMP[8].xxxx
>170:   ABS TEMP[8].x, TEMP[4].zzzz
>171:   FSGE TEMP[8].x, -TEMP[8].xxxx, IMM[0].yyyy
>172:   UIF TEMP[8].xxxx :0
>173:     MOV TEMP[8].x, CONST[87].xxxx
>174:   ELSE :0
>175:     MOV TEMP[8].x, TEMP[11].zzzz
>176:   ENDIF
>177:   MOV TEMP[11].z, TEMP[8].xxxx
>178:   ABS TEMP[8].x, TEMP[4].zzzz
>179:   FSGE TEMP[8].x, -TEMP[8].xxxx, IMM[0].yyyy
>180:   UIF TEMP[8].xxxx :0
>181:     MOV TEMP[8].x, CONST[87].yyyy
>182:   ELSE :0
>183:     MOV TEMP[8].x, TEMP[11].wwww
>184:   ENDIF
>185:   MOV TEMP[11].w, TEMP[8].xxxx
>186:   MAD TEMP[3].xy, TEMP[7].xyyy, TEMP[11].xyyy, TEMP[11].zwww
>187:   ADD TEMP[4], TEMP[3], IMM[2].xxyy
>188:   TXL TEMP[7].x, TEMP[4], SAMP[3], SHADOW2D
>189:   MOV TEMP[4].x, TEMP[7].xxxx
>190:   ADD TEMP[7], TEMP[3], IMM[2].zxyy
>191:   ADD TEMP[8], TEMP[3], IMM[2].xzyy
>192:   ADD TEMP[10], TEMP[3], IMM[2].zzyy
>193:   TXL TEMP[12].x, TEMP[7], SAMP[3], SHADOW2D
>194:   MOV TEMP[4].y, TEMP[12].xxxx
>195:   TXL TEMP[12].x, TEMP[8], SAMP[3], SHADOW2D
>196:   MOV TEMP[4].z, TEMP[12].xxxx
>197:   TXL TEMP[12].x, TEMP[10], SAMP[3], SHADOW2D
>198:   MOV TEMP[4].w, TEMP[12].xxxx
>199:   DP4 TEMP[12].x, TEMP[4], IMM[1].wwww
>200:   ADD TEMP[4], TEMP[3], IMM[2].xyyy
>201:   TXL TEMP[13].x, TEMP[4], SAMP[3], SHADOW2D
>202:   MOV TEMP[4].x, TEMP[13].xxxx
>203:   ADD TEMP[7], TEMP[3], IMM[2].zyyy
>204:   TXL TEMP[13], TEMP[7], SAMP[3], SHADOW2D
>205:   MOV TEMP[7], TEMP[13]
>206:   ADD TEMP[8], TEMP[3], IMM[2].yzyy
>207:   TXL TEMP[8].x, TEMP[8], SAMP[3], SHADOW2D
>208:   ADD TEMP[10], TEMP[3], IMM[2].yxyy
>209:   TXL TEMP[10].x, TEMP[10], SAMP[3], SHADOW2D
>210:   MOV TEMP[4].y, TEMP[13].xxxx
>211:   MOV TEMP[4].z, TEMP[8].xxxx
>212:   MOV TEMP[4].w, TEMP[10].xxxx
>213:   DP4 TEMP[8].x, TEMP[4], IMM[2].wwww
>214:   MOV TEMP[1].z, TEMP[8].xxxx
>215:   MOV TEMP[10].xy, TEMP[3].xyyy
>216:   MOV TEMP[10].z, TEMP[5].xxxx
>217:   MOV TEMP[10].w, IMM[0].yyyy
>218:   TXL TEMP[10], TEMP[10], SAMP[3], SHADOW2D
>219:   MOV TEMP[4], TEMP[10]
>220:   ADD TEMP[8].x, TEMP[8].xxxx, TEMP[12].xxxx
>221:   MAD TEMP[8].x, TEMP[10].xxxx, IMM[3].xxxx, TEMP[8].xxxx
>222:   MOV TEMP[1].y, TEMP[8].xxxx
>223:   FSLT TEMP[8].x, TEMP[6].xxxx, IMM[0].xxxx
>224:   UIF TEMP[8].xxxx :0
>225:     ADD TEMP[8].xyz, TEMP[1].wwww, IMM[3].yzww
>226:     ABS TEMP[10].x, TEMP[8].xxxx
>227:     FSGE TEMP[10].x, -TEMP[10].xxxx, IMM[0].yyyy
>228:     UIF TEMP[10].xxxx :0
>229:       MOV TEMP[10].x, CONST[73].xxxx
>230:     ELSE :0
>231:       MOV TEMP[10].x, IMM[0].yyyy
>232:     ENDIF
>233:     MOV TEMP[11].x, TEMP[10].xxxx
>234:     ABS TEMP[10].x, TEMP[8].xxxx
>235:     FSGE TEMP[10].x, -TEMP[10].xxxx, IMM[0].yyyy
>236:     UIF TEMP[10].xxxx :0
>237:       MOV TEMP[10].x, CONST[73].yyyy
>238:     ELSE :0
>239:       MOV TEMP[10].x, IMM[0].yyyy
>240:     ENDIF
>241:     MOV TEMP[11].y, TEMP[10].xxxx
>242:     ABS TEMP[10].x, TEMP[8].xxxx
>243:     FSGE TEMP[10].x, -TEMP[10].xxxx, IMM[0].yyyy
>244:     UIF TEMP[10].xxxx :0
>245:       MOV TEMP[10].x, CONST[73].zzzz
>246:     ELSE :0
>247:       MOV TEMP[10].x, IMM[0].yyyy
>248:     ENDIF
>249:     MOV TEMP[11].z, TEMP[10].xxxx
>250:     ABS TEMP[10].x, TEMP[8].xxxx
>251:     FSGE TEMP[10].x, -TEMP[10].xxxx, IMM[0].yyyy
>252:     UIF TEMP[10].xxxx :0
>253:       MOV TEMP[10].x, CONST[73].wwww
>254:     ELSE :0
>255:       MOV TEMP[10].x, IMM[0].yyyy
>256:     ENDIF
>257:     MOV TEMP[11].w, TEMP[10].xxxx
>258:     MOV TEMP[4], TEMP[11]
>259:     ABS TEMP[10].x, TEMP[8].xxxx
>260:     FSGE TEMP[10].x, -TEMP[10].xxxx, IMM[0].yyyy
>261:     UIF TEMP[10].xxxx :0
>262:       MOV TEMP[10].x, CONST[74].xxxx
>263:     ELSE :0
>264:       MOV TEMP[10].x, IMM[0].yyyy
>265:     ENDIF
>266:     MOV TEMP[11].x, TEMP[10].xxxx
>267:     ABS TEMP[10].x, TEMP[8].xxxx
>268:     FSGE TEMP[10].x, -TEMP[10].xxxx, IMM[0].yyyy
>269:     UIF TEMP[10].xxxx :0
>270:       MOV TEMP[10].x, CONST[74].yyyy
>271:     ELSE :0
>272:       MOV TEMP[10].x, IMM[0].yyyy
>273:     ENDIF
>274:     MOV TEMP[11].y, TEMP[10].xxxx
>275:     ABS TEMP[10].x, TEMP[8].xxxx
>276:     FSGE TEMP[10].x, -TEMP[10].xxxx, IMM[0].yyyy
>277:     UIF TEMP[10].xxxx :0
>278:       MOV TEMP[10].x, CONST[74].zzzz
>279:     ELSE :0
>280:       MOV TEMP[10].x, IMM[0].yyyy
>281:     ENDIF
>282:     MOV TEMP[11].z, TEMP[10].xxxx
>283:     ABS TEMP[10].x, TEMP[8].xxxx
>284:     FSGE TEMP[10].x, -TEMP[10].xxxx, IMM[0].yyyy
>285:     UIF TEMP[10].xxxx :0
>286:       MOV TEMP[10].x, CONST[74].wwww
>287:     ELSE :0
>288:       MOV TEMP[10].x, IMM[0].yyyy
>289:     ENDIF
>290:     MOV TEMP[11].w, TEMP[10].xxxx
>291:     MOV TEMP[7], TEMP[11]
>292:     ABS TEMP[10].x, TEMP[8].yyyy
>293:     FSGE TEMP[10].x, -TEMP[10].xxxx, IMM[0].yyyy
>294:     UIF TEMP[10].xxxx :0
>295:       MOV TEMP[10].x, CONST[77].xxxx
>296:     ELSE :0
>297:       MOV TEMP[10].x, TEMP[4].xxxx
>298:     ENDIF
>299:     MOV TEMP[11].x, TEMP[10].xxxx
>300:     ABS TEMP[10].x, TEMP[8].yyyy
>301:     FSGE TEMP[10].x, -TEMP[10].xxxx, IMM[0].yyyy
>302:     UIF TEMP[10].xxxx :0
>303:       MOV TEMP[10].x, CONST[77].yyyy
>304:     ELSE :0
>305:       MOV TEMP[10].x, TEMP[4].yyyy
>306:     ENDIF
>307:     MOV TEMP[11].y, TEMP[10].xxxx
>308:     ABS TEMP[10].x, TEMP[8].yyyy
>309:     FSGE TEMP[10].x, -TEMP[10].xxxx, IMM[0].yyyy
>310:     UIF TEMP[10].xxxx :0
>311:       MOV TEMP[10].x, CONST[77].zzzz
>312:     ELSE :0
>313:       MOV TEMP[10].x, TEMP[4].zzzz
>314:     ENDIF
>315:     MOV TEMP[11].z, TEMP[10].xxxx
>316:     ABS TEMP[10].x, TEMP[8].yyyy
>317:     FSGE TEMP[10].x, -TEMP[10].xxxx, IMM[0].yyyy
>318:     UIF TEMP[10].xxxx :0
>319:       MOV TEMP[10].x, CONST[77].wwww
>320:     ELSE :0
>321:       MOV TEMP[10].x, TEMP[4].wwww
>322:     ENDIF
>323:     MOV TEMP[11].w, TEMP[10].xxxx
>324:     MOV TEMP[4], TEMP[11]
>325:     ABS TEMP[10].x, TEMP[8].yyyy
>326:     FSGE TEMP[10].x, -TEMP[10].xxxx, IMM[0].yyyy
>327:     UIF TEMP[10].xxxx :0
>328:       MOV TEMP[10].x, CONST[78].xxxx
>329:     ELSE :0
>330:       MOV TEMP[10].x, TEMP[7].xxxx
>331:     ENDIF
>332:     MOV TEMP[11].x, TEMP[10].xxxx
>333:     ABS TEMP[10].x, TEMP[8].yyyy
>334:     FSGE TEMP[10].x, -TEMP[10].xxxx, IMM[0].yyyy
>335:     UIF TEMP[10].xxxx :0
>336:       MOV TEMP[10].x, CONST[78].yyyy
>337:     ELSE :0
>338:       MOV TEMP[10].x, TEMP[7].yyyy
>339:     ENDIF
>340:     MOV TEMP[11].y, TEMP[10].xxxx
>341:     ABS TEMP[10].x, TEMP[8].yyyy
>342:     FSGE TEMP[10].x, -TEMP[10].xxxx, IMM[0].yyyy
>343:     UIF TEMP[10].xxxx :0
>344:       MOV TEMP[10].x, CONST[78].zzzz
>345:     ELSE :0
>346:       MOV TEMP[10].x, TEMP[7].zzzz
>347:     ENDIF
>348:     MOV TEMP[11].z, TEMP[10].xxxx
>349:     ABS TEMP[10].x, TEMP[8].yyyy
>350:     FSGE TEMP[10].x, -TEMP[10].xxxx, IMM[0].yyyy
>351:     UIF TEMP[10].xxxx :0
>352:       MOV TEMP[10].x, CONST[78].wwww
>353:     ELSE :0
>354:       MOV TEMP[10].x, TEMP[7].wwww
>355:     ENDIF
>356:     MOV TEMP[11].w, TEMP[10].xxxx
>357:     MOV TEMP[7], TEMP[11]
>358:     ABS TEMP[10].x, TEMP[8].zzzz
>359:     FSGE TEMP[10].x, -TEMP[10].xxxx, IMM[0].yyyy
>360:     UIF TEMP[10].xxxx :0
>361:       MOV TEMP[10].x, CONST[81].xxxx
>362:     ELSE :0
>363:       MOV TEMP[10].x, TEMP[4].xxxx
>364:     ENDIF
>365:     MOV TEMP[11].x, TEMP[10].xxxx
>366:     ABS TEMP[10].x, TEMP[8].zzzz
>367:     FSGE TEMP[10].x, -TEMP[10].xxxx, IMM[0].yyyy
>368:     UIF TEMP[10].xxxx :0
>369:       MOV TEMP[10].x, CONST[81].yyyy
>370:     ELSE :0
>371:       MOV TEMP[10].x, TEMP[4].yyyy
>372:     ENDIF
>373:     MOV TEMP[11].y, TEMP[10].xxxx
>374:     ABS TEMP[10].x, TEMP[8].zzzz
>375:     FSGE TEMP[10].x, -TEMP[10].xxxx, IMM[0].yyyy
>376:     UIF TEMP[10].xxxx :0
>377:       MOV TEMP[10].x, CONST[81].zzzz
>378:     ELSE :0
>379:       MOV TEMP[10].x, TEMP[4].zzzz
>380:     ENDIF
>381:     MOV TEMP[11].z, TEMP[10].xxxx
>382:     ABS TEMP[10].x, TEMP[8].zzzz
>383:     FSGE TEMP[10].x, -TEMP[10].xxxx, IMM[0].yyyy
>384:     UIF TEMP[10].xxxx :0
>385:       MOV TEMP[10].x, CONST[81].wwww
>386:     ELSE :0
>387:       MOV TEMP[10].x, TEMP[4].wwww
>388:     ENDIF
>389:     MOV TEMP[11].w, TEMP[10].xxxx
>390:     MOV TEMP[4], TEMP[11]
>391:     ABS TEMP[10].x, TEMP[8].zzzz
>392:     FSGE TEMP[10].x, -TEMP[10].xxxx, IMM[0].yyyy
>393:     UIF TEMP[10].xxxx :0
>394:       MOV TEMP[10].x, CONST[82].xxxx
>395:     ELSE :0
>396:       MOV TEMP[10].x, TEMP[7].xxxx
>397:     ENDIF
>398:     MOV TEMP[11].x, TEMP[10].xxxx
>399:     ABS TEMP[10].x, TEMP[8].zzzz
>400:     FSGE TEMP[10].x, -TEMP[10].xxxx, IMM[0].yyyy
>401:     UIF TEMP[10].xxxx :0
>402:       MOV TEMP[10].x, CONST[82].yyyy
>403:     ELSE :0
>404:       MOV TEMP[10].x, TEMP[7].yyyy
>405:     ENDIF
>406:     MOV TEMP[11].y, TEMP[10].xxxx
>407:     ABS TEMP[10].x, TEMP[8].zzzz
>408:     FSGE TEMP[10].x, -TEMP[10].xxxx, IMM[0].yyyy
>409:     UIF TEMP[10].xxxx :0
>410:       MOV TEMP[10].x, CONST[82].zzzz
>411:     ELSE :0
>412:       MOV TEMP[10].x, TEMP[7].zzzz
>413:     ENDIF
>414:     MOV TEMP[11].z, TEMP[10].xxxx
>415:     ABS TEMP[10].x, TEMP[8].zzzz
>416:     FSGE TEMP[10].x, -TEMP[10].xxxx, IMM[0].yyyy
>417:     UIF TEMP[10].xxxx :0
>418:       MOV TEMP[10].x, CONST[82].wwww
>419:     ELSE :0
>420:       MOV TEMP[10].x, TEMP[7].wwww
>421:     ENDIF
>422:     MOV TEMP[11].w, TEMP[10].xxxx
>423:     DP4 TEMP[4].x, TEMP[2], TEMP[4]
>424:     MOV_SAT TEMP[4].x, TEMP[4].xxxx
>425:     DP4 TEMP[10].x, TEMP[2], TEMP[11]
>426:     MOV_SAT TEMP[10].x, TEMP[10].xxxx
>427:     MOV TEMP[4].y, TEMP[10].xxxx
>428:     ABS TEMP[10].x, TEMP[8].xxxx
>429:     FSGE TEMP[10].x, -TEMP[10].xxxx, IMM[0].yyyy
>430:     UIF TEMP[10].xxxx :0
>431:       MOV TEMP[10].x, CONST[86].zzzz
>432:     ELSE :0
>433:       MOV TEMP[10].x, IMM[0].yyyy
>434:     ENDIF
>435:     ABS TEMP[12].x, TEMP[8].xxxx
>436:     FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>437:     UIF TEMP[12].xxxx :0
>438:       MOV TEMP[12].x, CONST[86].wwww
>439:     ELSE :0
>440:       MOV TEMP[12].x, IMM[0].yyyy
>441:     ENDIF
>442:     MOV TEMP[11].y, TEMP[12].xxxx
>443:     ABS TEMP[12].x, TEMP[8].xxxx
>444:     FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>445:     UIF TEMP[12].xxxx :0
>446:       MOV TEMP[12].x, CONST[86].xxxx
>447:     ELSE :0
>448:       MOV TEMP[12].x, IMM[0].yyyy
>449:     ENDIF
>450:     MOV TEMP[11].z, TEMP[12].xxxx
>451:     ABS TEMP[12].x, TEMP[8].xxxx
>452:     FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>453:     UIF TEMP[12].xxxx :0
>454:       MOV TEMP[12].x, CONST[86].yyyy
>455:     ELSE :0
>456:       MOV TEMP[12].x, IMM[0].yyyy
>457:     ENDIF
>458:     MOV TEMP[11].w, TEMP[12].xxxx
>459:     ABS TEMP[12].x, TEMP[8].yyyy
>460:     FSGE TEMP[12].x, -TEMP[12].xxxx, IMM[0].yyyy
>461:     UIF TEMP[12].xxxx :0
>462:       MOV TEMP[12].x, CONST[87].zzzz
>463:     ELSE :0
>464:       MOV TEMP[12].x, TEMP[10].xxxx
>465:     ENDIF
>466:     ABS TEMP[10].x, TEMP[8].yyyy
>467:     FSGE TEMP[10].x, -TEMP[10].xxxx, IMM[0].yyyy
>468:     UIF TEMP[10].xxxx :0
>469:       MOV TEMP[10].x, CONST[87].wwww
>470:     ELSE :0
>471:       MOV TEMP[10].x, TEMP[11].yyyy
>472:     ENDIF
>473:     MOV TEMP[11].y, TEMP[10].xxxx
>474:     ABS TEMP[10].x, TEMP[8].yyyy
>475:     FSGE TEMP[10].x, -TEMP[10].xxxx, IMM[0].yyyy
>476:     UIF TEMP[10].xxxx :0
>477:       MOV TEMP[10].x, CONST[87].xxxx
>478:     ELSE :0
>479:       MOV TEMP[10].x, TEMP[11].zzzz
>480:     ENDIF
>481:     MOV TEMP[11].z, TEMP[10].xxxx
>482:     ABS TEMP[10].x, TEMP[8].yyyy
>483:     FSGE TEMP[10].x, -TEMP[10].xxxx, IMM[0].yyyy
>484:     UIF TEMP[10].xxxx :0
>485:       MOV TEMP[10].x, CONST[87].yyyy
>486:     ELSE :0
>487:       MOV TEMP[10].x, TEMP[11].wwww
>488:     ENDIF
>489:     MOV TEMP[11].w, TEMP[10].xxxx
>490:     ABS TEMP[10].x, TEMP[8].zzzz
>491:     FSGE TEMP[10].x, -TEMP[10].xxxx, IMM[0].yyyy
>492:     UIF TEMP[10].xxxx :0
>493:       MOV TEMP[10].x, CONST[88].zzzz
>494:     ELSE :0
>495:       MOV TEMP[10].x, TEMP[12].xxxx
>496:     ENDIF
>497:     MOV TEMP[11].x, TEMP[10].xxxx
>498:     ABS TEMP[10].x, TEMP[8].zzzz
>499:     FSGE TEMP[10].x, -TEMP[10].xxxx, IMM[0].yyyy
>500:     UIF TEMP[10].xxxx :0
>501:       MOV TEMP[10].x, CONST[88].wwww
>502:     ELSE :0
>503:       MOV TEMP[10].x, TEMP[11].yyyy
>504:     ENDIF
>505:     MOV TEMP[11].y, TEMP[10].xxxx
>506:     ABS TEMP[10].x, TEMP[8].zzzz
>507:     FSGE TEMP[10].x, -TEMP[10].xxxx, IMM[0].yyyy
>508:     UIF TEMP[10].xxxx :0
>509:       MOV TEMP[10].x, CONST[88].xxxx
>510:     ELSE :0
>511:       MOV TEMP[10].x, TEMP[11].zzzz
>512:     ENDIF
>513:     MOV TEMP[11].z, TEMP[10].xxxx
>514:     ABS TEMP[10].x, TEMP[8].zzzz
>515:     FSGE TEMP[10].x, -TEMP[10].xxxx, IMM[0].yyyy
>516:     UIF TEMP[10].xxxx :0
>517:       MOV TEMP[10].x, CONST[88].yyyy
>518:     ELSE :0
>519:       MOV TEMP[10].x, TEMP[11].wwww
>520:     ENDIF
>521:     MOV TEMP[11].w, TEMP[10].xxxx
>522:     MAD TEMP[3].xy, TEMP[4].xyyy, TEMP[11].xyyy, TEMP[11].zwww
>523:     ADD TEMP[2], TEMP[3], IMM[2].xxyy
>524:     TXL TEMP[10].x, TEMP[2], SAMP[3], SHADOW2D
>525:     MOV TEMP[2].x, TEMP[10].xxxx
>526:     ADD TEMP[9], TEMP[3], IMM[2].zxyy
>527:     ADD TEMP[4], TEMP[3], IMM[2].xzyy
>528:     ADD TEMP[7], TEMP[3], IMM[2].zzyy
>529:     TXL TEMP[10].x, TEMP[9], SAMP[3], SHADOW2D
>530:     MOV TEMP[2].y, TEMP[10].xxxx
>531:     TXL TEMP[10].x, TEMP[4], SAMP[3], SHADOW2D
>532:     MOV TEMP[2].z, TEMP[10].xxxx
>533:     TXL TEMP[10].x, TEMP[7], SAMP[3], SHADOW2D
>534:     MOV TEMP[2].w, TEMP[10].xxxx
>535:     DP4 TEMP[10].x, TEMP[2], IMM[1].wwww
>536:     ADD TEMP[2], TEMP[3], IMM[2].xyyy
>537:     TXL TEMP[12].x, TEMP[2], SAMP[3], SHADOW2D
>538:     MOV TEMP[2].x, TEMP[12].xxxx
>539:     ADD TEMP[9], TEMP[3], IMM[2].zyyy
>540:     TXL TEMP[9].x, TEMP[9], SAMP[3], SHADOW2D
>541:     ADD TEMP[4], TEMP[3], IMM[2].yzyy
>542:     TXL TEMP[4].x, TEMP[4], SAMP[3], SHADOW2D
>543:     ADD TEMP[7], TEMP[3], IMM[2].yxyy
>544:     TXL TEMP[7].x, TEMP[7], SAMP[3], SHADOW2D
>545:     MOV TEMP[2].y, TEMP[9].xxxx
>546:     MOV TEMP[2].z, TEMP[4].xxxx
>547:     MOV TEMP[2].w, TEMP[7].xxxx
>548:     DP4 TEMP[4].x, TEMP[2], IMM[2].wwww
>549:     MOV TEMP[1].w, TEMP[4].xxxx
>550:     MOV TEMP[7].xy, TEMP[3].xyyy
>551:     MOV TEMP[7].z, TEMP[5].xxxx
>552:     MOV TEMP[7].w, IMM[0].yyyy
>553:     TXL TEMP[5].x, TEMP[7], SAMP[3], SHADOW2D
>554:     ADD TEMP[4].x, TEMP[4].xxxx, TEMP[10].xxxx
>555:     MOV TEMP[1].z, TEMP[4].xxxx
>556:     MAD TEMP[4].x, TEMP[5].xxxx, IMM[3].xxxx, TEMP[4].xxxx
>557:     MOV TEMP[1].z, TEMP[4].xxxx
>558:     MOV TEMP[11].xw, TEMP[1]
>559:     FSGE TEMP[5].x, TEMP[8].zzzz, IMM[0].yyyy
>560:     UIF TEMP[5].xxxx :0
>561:       MOV TEMP[5].x, IMM[0].xxxx
>562:     ELSE :0
>563:       MOV TEMP[5].x, TEMP[4].xxxx
>564:     ENDIF
>565:     MOV TEMP[11].z, TEMP[5].xxxx
>566:     MOV TEMP[1].x, TEMP[11].xxzw
>567:     LRP TEMP[2].x, TEMP[6].xxxx, TEMP[1].yyyy, TEMP[5].xxxx
>568:     MOV TEMP[1].y, TEMP[2].xxxx
>569:   ENDIF
>570:   ADD TEMP[2].xyz, -CONST[89].xyzz, IN[5].xyzz
>571:   DP3 TEMP[4].x, TEMP[2].xyzz, TEMP[2].xyzz
>572:   MAD TEMP[4].x, TEMP[4].xxxx, CONST[68].yyyy, CONST[68].xxxx
>573:   MOV_SAT TEMP[4].x, TEMP[4].xxxx
>574:   LRP TEMP[2].x, TEMP[4].xxxx, IMM[0].xxxx, TEMP[1].yyyy
>575: ELSE :0
>576:   MOV TEMP[2].x, IMM[0].xxxx
>577: ENDIF
>578: MAD TEMP[4].xyz, IN[1].xyzz, TEMP[2].xxxx, IN[4].xyzz
>579: MOV TEMP[1].yzw, TEMP[4].yxyz
>580: UIF CONST[90].xxxx :0
>581:   DP3 TEMP[5].x, TEMP[4].xyzz, IMM[4].xyzz
>582:   RCP TEMP[5].x, TEMP[5].xxxx
>583:   MUL TEMP[5].x, TEMP[5].xxxx, IN[1].wwww
>584:   ADD TEMP[2].x, -TEMP[2].xxxx, IMM[0].xxxx
>585:   MAD TEMP[2].x, TEMP[5].xxxx, -TEMP[2].xxxx, IMM[0].xxxx
>586:   MUL TEMP[4].xyz, TEMP[4].zyxx, TEMP[2].xxxx
>587:   MAD TEMP[2].x, TEMP[2].xxxx, IMM[4].wwww, IMM[4].wwww
>588:   LRP TEMP[4].xyz, TEMP[2].xxxx, TEMP[4].zyxx, TEMP[4].xyzz
>589:   MOV TEMP[1].yzw, TEMP[4].yxyz
>590: ENDIF
>591: ADD TEMP[2].x, TEMP[0].wwww, IMM[1].yyyy
>592: MAD TEMP[2].x, CONST[20].wwww, TEMP[2].xxxx, IMM[0].xxxx
>593: ADD TEMP[4].x, TEMP[0].wwww, CONST[12].xxxx
>594: ADD TEMP[5].xyz, IMM[1].yyyy, CONST[1].xyzz
>595: MOV_SAT TEMP[4].x, TEMP[4].xxxx
>596: MAD TEMP[4].xyz, TEMP[4].xxxx, TEMP[5].xyzz, IMM[0].xxxx
>597: MUL TEMP[3].xyz, TEMP[1].yzww, TEMP[4].xyzz
>598: MUL TEMP[5].x, TEMP[2].xxxx, CONST[1].wwww
>599: MAD TEMP[2].x, TEMP[5].xxxx, IN[4].wwww, -TEMP[5].xxxx
>600: MAD TEMP[5].x, CONST[12].wwww, TEMP[2].xxxx, TEMP[5].xxxx
>601: DP3 TEMP[7].x, IN[3].xyzz, IN[2].xyzz
>602: ADD TEMP[2].x, TEMP[7].xxxx, TEMP[7].xxxx
>603: DP3 TEMP[7].x, IN[3].xyzz, IN[3].xyzz
>604: MUL TEMP[6].xyz, TEMP[7].xxxx, IN[2].xyzz
>605: MAD TEMP[6].xyz, TEMP[2].xxxx, IN[3].xyzz, -TEMP[6].xyzz
>606: MOV TEMP[7].xyz, TEMP[6].xyzz
>607: TEX TEMP[7].xyz, TEMP[7], SAMP[1], CUBE
>608: MUL TEMP[6].xyz, TEMP[7].xyzz, CONST[30].zzzz
>609: MUL TEMP[6].xyz, TEMP[1].xxxx, TEMP[6].xyzz
>610: MUL TEMP[6].xyz, TEMP[6].xyzz, CONST[0].xyzz
>611: MAD TEMP[1].xyz, TEMP[1].yzww, TEMP[4].xyzz, -CONST[19].zzzz
>612: MUL TEMP[1].xyz, TEMP[1].xyzz, CONST[19].wwww
>613: MOV_SAT TEMP[4].xyz, TEMP[1].xyzz
>614: MAD TEMP[1].xyz, TEMP[6].xyzz, TEMP[4].xyzz, -TEMP[6].xyzz
>615: MAD TEMP[1].xyz, CONST[19].yyyy, TEMP[1].xyzz, TEMP[6].xyzz
>616: MAD TEMP[2].xyz, TEMP[1].xyzz, TEMP[1].xyzz, -TEMP[1].xyzz
>617: MAD TEMP[1].xyz, CONST[19].xxxx, TEMP[2].xyzz, TEMP[1].xyzz
>618: DP3 TEMP[4].x, TEMP[1].xyzz, IMM[5].xyzz
>619: LRP TEMP[2].xyz, CONST[3].xyzz, TEMP[1].xyzz, TEMP[4].xxxx
>620: MAD TEMP[0].xyz, TEMP[0].xyzz, TEMP[3].xyzz, TEMP[2].xyzz
>621: ADD TEMP[1].xyz, CONST[20].xyzz, -IN[5].xyzz
>622: DP3 TEMP[1].x, TEMP[1].xyzz, TEMP[1].xyzz
>623: SQRT TEMP[1].x, TEMP[1].xxxx
>624: MAD TEMP[1].x, TEMP[1].xxxx, CONST[21].wwww, CONST[21].xxxx
>625: MOV_SAT TEMP[1].x, TEMP[1].xxxx
>626: MIN TEMP[1].x, TEMP[1].xxxx, CONST[21].zzzz
>627: ABS TEMP[2].x, CONST[12].yyyy
>628: MUL TEMP[3].xyz, TEMP[0].xyzz, CONST[30].xxxx
>629: MUL TEMP[4].x, CONST[29].wwww, IN[5].wwww
>630: FSGE TEMP[6].x, -TEMP[2].xxxx, IMM[0].yyyy
>631: UIF TEMP[6].xxxx :0
>632:   MOV TEMP[5].x, TEMP[5].xxxx
>633: ELSE :0
>634:   MOV TEMP[5].x, TEMP[4].xxxx
>635: ENDIF
>636: MOV TEMP[2].w, TEMP[5].xxxx
>637: MUL TEMP[1].x, TEMP[1].xxxx, TEMP[1].xxxx
>638: MAD TEMP[0].xyz, TEMP[0].xyzz, -CONST[30].xxxx, CONST[29].xyzz
>639: MAD TEMP[2].xyz, TEMP[1].xxxx, TEMP[0].xyzz, TEMP[3].xyzz
>640: MOV OUT[0], TEMP[2]
>641: END
>; ModuleID = 'tgsi'
>
>define void @main([9 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, 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:
>  %23 = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %1, i64 0, i64 0
>  %24 = load <16 x i8>, <16 x i8> addrspace(2)* %23, align 16, !tbaa !0
>  %25 = call float @llvm.SI.load.const(<16 x i8> %24, i32 0)
>  %26 = call float @llvm.SI.load.const(<16 x i8> %24, i32 4)
>  %27 = call float @llvm.SI.load.const(<16 x i8> %24, i32 8)
>  %28 = call float @llvm.SI.load.const(<16 x i8> %24, i32 16)
>  %29 = call float @llvm.SI.load.const(<16 x i8> %24, i32 20)
>  %30 = call float @llvm.SI.load.const(<16 x i8> %24, i32 24)
>  %31 = call float @llvm.SI.load.const(<16 x i8> %24, i32 28)
>  %32 = call float @llvm.SI.load.const(<16 x i8> %24, i32 48)
>  %33 = call float @llvm.SI.load.const(<16 x i8> %24, i32 52)
>  %34 = call float @llvm.SI.load.const(<16 x i8> %24, i32 56)
>  %35 = call float @llvm.SI.load.const(<16 x i8> %24, i32 192)
>  %36 = call float @llvm.SI.load.const(<16 x i8> %24, i32 196)
>  %37 = call float @llvm.SI.load.const(<16 x i8> %24, i32 204)
>  %38 = call float @llvm.SI.load.const(<16 x i8> %24, i32 304)
>  %39 = call float @llvm.SI.load.const(<16 x i8> %24, i32 308)
>  %40 = call float @llvm.SI.load.const(<16 x i8> %24, i32 312)
>  %41 = call float @llvm.SI.load.const(<16 x i8> %24, i32 316)
>  %42 = call float @llvm.SI.load.const(<16 x i8> %24, i32 320)
>  %43 = call float @llvm.SI.load.const(<16 x i8> %24, i32 324)
>  %44 = call float @llvm.SI.load.const(<16 x i8> %24, i32 328)
>  %45 = call float @llvm.SI.load.const(<16 x i8> %24, i32 332)
>  %46 = call float @llvm.SI.load.const(<16 x i8> %24, i32 336)
>  %47 = call float @llvm.SI.load.const(<16 x i8> %24, i32 344)
>  %48 = call float @llvm.SI.load.const(<16 x i8> %24, i32 348)
>  %49 = call float @llvm.SI.load.const(<16 x i8> %24, i32 464)
>  %50 = call float @llvm.SI.load.const(<16 x i8> %24, i32 468)
>  %51 = call float @llvm.SI.load.const(<16 x i8> %24, i32 472)
>  %52 = call float @llvm.SI.load.const(<16 x i8> %24, i32 476)
>  %53 = call float @llvm.SI.load.const(<16 x i8> %24, i32 480)
>  %54 = call float @llvm.SI.load.const(<16 x i8> %24, i32 488)
>  %55 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1080)
>  %56 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1084)
>  %57 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1088)
>  %58 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1092)
>  %59 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1168)
>  %60 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1172)
>  %61 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1176)
>  %62 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1180)
>  %63 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1184)
>  %64 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1188)
>  %65 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1192)
>  %66 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1196)
>  %67 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1232)
>  %68 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1236)
>  %69 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1240)
>  %70 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1244)
>  %71 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1248)
>  %72 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1252)
>  %73 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1256)
>  %74 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1260)
>  %75 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1296)
>  %76 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1300)
>  %77 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1304)
>  %78 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1308)
>  %79 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1312)
>  %80 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1316)
>  %81 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1320)
>  %82 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1324)
>  %83 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1376)
>  %84 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1380)
>  %85 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1384)
>  %86 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1388)
>  %87 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1392)
>  %88 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1396)
>  %89 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1400)
>  %90 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1404)
>  %91 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1408)
>  %92 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1412)
>  %93 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1416)
>  %94 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1420)
>  %95 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1424)
>  %96 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1428)
>  %97 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1432)
>  %98 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1440)
>  %99 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 0
>  %100 = load <8 x i32>, <8 x i32> addrspace(2)* %99, align 32, !tbaa !0
>  %101 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 0
>  %102 = load <4 x i32>, <4 x i32> addrspace(2)* %101, align 16, !tbaa !0
>  %103 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 1
>  %104 = load <8 x i32>, <8 x i32> addrspace(2)* %103, align 32, !tbaa !0
>  %105 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 1
>  %106 = load <4 x i32>, <4 x i32> addrspace(2)* %105, align 16, !tbaa !0
>  %107 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 2
>  %108 = load <8 x i32>, <8 x i32> addrspace(2)* %107, align 32, !tbaa !0
>  %109 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 2
>  %110 = load <4 x i32>, <4 x i32> addrspace(2)* %109, align 16, !tbaa !0
>  %111 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %3, i64 0, i64 3
>  %112 = load <8 x i32>, <8 x i32> addrspace(2)* %111, align 32, !tbaa !0
>  %113 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %2, i64 0, i64 3
>  %114 = load <4 x i32>, <4 x i32> addrspace(2)* %113, align 16, !tbaa !0
>  %115 = call float @llvm.SI.fs.interp(i32 0, i32 0, i32 %6, <2 x i32> %8)
>  %116 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %6, <2 x i32> %8)
>  %117 = call float @llvm.SI.fs.interp(i32 0, i32 1, i32 %6, <2 x i32> %8)
>  %118 = call float @llvm.SI.fs.interp(i32 1, i32 1, i32 %6, <2 x i32> %8)
>  %119 = call float @llvm.SI.fs.interp(i32 2, i32 1, i32 %6, <2 x i32> %8)
>  %120 = call float @llvm.SI.fs.interp(i32 3, i32 1, i32 %6, <2 x i32> %8)
>  %121 = call float @llvm.SI.fs.interp(i32 0, i32 2, i32 %6, <2 x i32> %8)
>  %122 = call float @llvm.SI.fs.interp(i32 1, i32 2, i32 %6, <2 x i32> %8)
>  %123 = call float @llvm.SI.fs.interp(i32 2, i32 2, i32 %6, <2 x i32> %8)
>  %124 = call float @llvm.SI.fs.interp(i32 0, i32 3, i32 %6, <2 x i32> %8)
>  %125 = call float @llvm.SI.fs.interp(i32 1, i32 3, i32 %6, <2 x i32> %8)
>  %126 = call float @llvm.SI.fs.interp(i32 2, i32 3, i32 %6, <2 x i32> %8)
>  %127 = call float @llvm.SI.fs.interp(i32 0, i32 4, i32 %6, <2 x i32> %8)
>  %128 = call float @llvm.SI.fs.interp(i32 1, i32 4, i32 %6, <2 x i32> %8)
>  %129 = call float @llvm.SI.fs.interp(i32 2, i32 4, i32 %6, <2 x i32> %8)
>  %130 = call float @llvm.SI.fs.interp(i32 3, i32 4, i32 %6, <2 x i32> %8)
>  %131 = call float @llvm.SI.fs.interp(i32 0, i32 5, i32 %6, <2 x i32> %8)
>  %132 = call float @llvm.SI.fs.interp(i32 1, i32 5, i32 %6, <2 x i32> %8)
>  %133 = call float @llvm.SI.fs.interp(i32 2, i32 5, i32 %6, <2 x i32> %8)
>  %134 = call float @llvm.SI.fs.interp(i32 3, i32 5, i32 %6, <2 x i32> %8)
>  %135 = bitcast float %115 to i32
>  %136 = bitcast float %116 to i32
>  %137 = insertelement <2 x i32> undef, i32 %135, i32 0
>  %138 = insertelement <2 x i32> %137, i32 %136, i32 1
>  %139 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %138, <8 x i32> %100, <4 x i32> %102, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %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 = bitcast float %115 to i32
>  %145 = bitcast float %116 to i32
>  %146 = insertelement <2 x i32> undef, i32 %144, i32 0
>  %147 = insertelement <2 x i32> %146, i32 %145, i32 1
>  %148 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %147, <8 x i32> %108, <4 x i32> %110, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %149 = extractelement <4 x float> %148, i32 0
>  %150 = bitcast float %98 to i32
>  %151 = icmp eq i32 %150, 0
>  br i1 %151, label %ENDIF, label %IF
>
>IF:                                               ; preds = %main_body
>  %152 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1372)
>  %153 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1368)
>  %154 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1364)
>  %155 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1360)
>  %156 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1148)
>  %157 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1144)
>  %158 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1140)
>  %159 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1136)
>  %160 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1132)
>  %161 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1128)
>  %162 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1124)
>  %163 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1120)
>  %164 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1116)
>  %165 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1112)
>  %166 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1108)
>  %167 = call float @llvm.SI.load.const(<16 x i8> %24, i32 1104)
>  %168 = fadd float %131, 0.000000e+00
>  %169 = fadd float %132, 0.000000e+00
>  %170 = fadd float %133, 0.000000e+00
>  %171 = fmul float %131, 0.000000e+00
>  %172 = fadd float %171, 1.000000e+00
>  %173 = fmul float %168, %167
>  %174 = fmul float %169, %166
>  %175 = fadd float %173, %174
>  %176 = fmul float %170, %165
>  %177 = fadd float %175, %176
>  %178 = fmul float %172, %164
>  %179 = fadd float %177, %178
>  %180 = fmul float %168, %163
>  %181 = fmul float %169, %162
>  %182 = fadd float %180, %181
>  %183 = fmul float %170, %161
>  %184 = fadd float %182, %183
>  %185 = fmul float %172, %160
>  %186 = fadd float %184, %185
>  %187 = call float @llvm.AMDIL.clamp.(float %179, float 0.000000e+00, float 1.000000e+00)
>  %188 = call float @llvm.AMDIL.clamp.(float %186, float 0.000000e+00, float 1.000000e+00)
>  %189 = fsub float %187, %179
>  %190 = fsub float %188, %186
>  %191 = fadd float %189, %190
>  %192 = fmul float %168, %59
>  %193 = fmul float %169, %60
>  %194 = fadd float %192, %193
>  %195 = fmul float %170, %61
>  %196 = fadd float %194, %195
>  %197 = fmul float %172, %62
>  %198 = fadd float %196, %197
>  %199 = fmul float %168, %63
>  %200 = fmul float %169, %64
>  %201 = fadd float %199, %200
>  %202 = fmul float %170, %65
>  %203 = fadd float %201, %202
>  %204 = fmul float %172, %66
>  %205 = fadd float %203, %204
>  %206 = call float @llvm.AMDIL.clamp.(float %198, float 0.000000e+00, float 1.000000e+00)
>  %207 = call float @llvm.AMDIL.clamp.(float %205, float 0.000000e+00, float 1.000000e+00)
>  %208 = fsub float %206, %198
>  %209 = fsub float %207, %205
>  %210 = fadd float %208, %209
>  %211 = fmul float %168, %67
>  %212 = fmul float %169, %68
>  %213 = fadd float %211, %212
>  %214 = fmul float %170, %69
>  %215 = fadd float %213, %214
>  %216 = fmul float %172, %70
>  %217 = fadd float %215, %216
>  %218 = fmul float %168, %71
>  %219 = fmul float %169, %72
>  %220 = fadd float %218, %219
>  %221 = fmul float %170, %73
>  %222 = fadd float %220, %221
>  %223 = fmul float %172, %74
>  %224 = fadd float %222, %223
>  %225 = call float @llvm.fabs.f32(float %210)
>  %226 = fcmp ole float %225, -0.000000e+00
>  %. = select i1 %226, float %198, float %217
>  %227 = call float @llvm.fabs.f32(float %210)
>  %228 = fcmp ole float %227, -0.000000e+00
>  %temp28.0 = select i1 %228, float %205, float %224
>  %229 = call float @llvm.fabs.f32(float %210)
>  %230 = fcmp ole float %229, -0.000000e+00
>  %.230 = select i1 %230, float 1.000000e+00, float 2.000000e+00
>  %231 = call float @llvm.fabs.f32(float %191)
>  %232 = fcmp ole float %231, -0.000000e+00
>  %temp28.2 = select i1 %232, float %179, float %.
>  %233 = call float @llvm.fabs.f32(float %191)
>  %234 = fcmp ole float %233, -0.000000e+00
>  %.temp28.0 = select i1 %234, float %186, float %temp28.0
>  %235 = call float @llvm.fabs.f32(float %191)
>  %236 = fcmp ole float %235, -0.000000e+00
>  %temp16.1 = select i1 %236, float 0.000000e+00, float %.230
>  %237 = fmul float %168, %159
>  %238 = fmul float %169, %158
>  %239 = fadd float %237, %238
>  %240 = fmul float %170, %157
>  %241 = fadd float %239, %240
>  %242 = fmul float %172, %156
>  %243 = fadd float %241, %242
>  %244 = fadd float %temp28.2, -5.000000e-01
>  %245 = fadd float %.temp28.0, -5.000000e-01
>  %246 = call float @llvm.fabs.f32(float %244)
>  %247 = call float @llvm.fabs.f32(float %245)
>  %248 = fsub float %246, %55
>  %249 = fsub float %247, %55
>  %250 = fmul float %248, %56
>  %251 = fmul float %249, %56
>  %252 = call float @llvm.AMDIL.clamp.(float %250, float 0.000000e+00, float 1.000000e+00)
>  %253 = call float @llvm.AMDIL.clamp.(float %251, float 0.000000e+00, float 1.000000e+00)
>  %254 = fsub float 1.000000e+00, %252
>  %255 = fsub float 1.000000e+00, %253
>  %256 = fmul float %255, %254
>  %257 = call float @llvm.AMDIL.clamp.(float %temp28.2, float 0.000000e+00, float 1.000000e+00)
>  %258 = call float @llvm.AMDIL.clamp.(float %.temp28.0, float 0.000000e+00, float 1.000000e+00)
>  %259 = fadd float %temp16.1, -1.000000e+00
>  %260 = fadd float %temp16.1, -2.000000e+00
>  %261 = call float @llvm.fabs.f32(float %temp16.1)
>  %262 = fcmp ole float %261, -0.000000e+00
>  %.231 = select i1 %262, float %153, float 0.000000e+00
>  %263 = call float @llvm.fabs.f32(float %temp16.1)
>  %264 = fcmp ole float %263, -0.000000e+00
>  %temp40.0 = select i1 %264, float %152, float 0.000000e+00
>  %265 = call float @llvm.fabs.f32(float %temp16.1)
>  %266 = fcmp ole float %265, -0.000000e+00
>  %.232 = select i1 %266, float %155, float 0.000000e+00
>  %267 = call float @llvm.fabs.f32(float %temp16.1)
>  %268 = fcmp ole float %267, -0.000000e+00
>  %temp40.2 = select i1 %268, float %154, float 0.000000e+00
>  %269 = call float @llvm.fabs.f32(float %259)
>  %270 = fcmp ole float %269, -0.000000e+00
>  %..231 = select i1 %270, float %85, float %.231
>  %271 = call float @llvm.fabs.f32(float %259)
>  %272 = fcmp ole float %271, -0.000000e+00
>  %temp32.1 = select i1 %272, float %86, float %temp40.0
>  %273 = call float @llvm.fabs.f32(float %259)
>  %274 = fcmp ole float %273, -0.000000e+00
>  %..232 = select i1 %274, float %83, float %.232
>  %275 = call float @llvm.fabs.f32(float %259)
>  %276 = fcmp ole float %275, -0.000000e+00
>  %temp32.3 = select i1 %276, float %84, float %temp40.2
>  %277 = call float @llvm.fabs.f32(float %260)
>  %278 = fcmp ole float %277, -0.000000e+00
>  %...231 = select i1 %278, float %89, float %..231
>  %279 = call float @llvm.fabs.f32(float %260)
>  %280 = fcmp ole float %279, -0.000000e+00
>  %temp32.5 = select i1 %280, float %90, float %temp32.1
>  %281 = call float @llvm.fabs.f32(float %260)
>  %282 = fcmp ole float %281, -0.000000e+00
>  %...232 = select i1 %282, float %87, float %..232
>  %283 = call float @llvm.fabs.f32(float %260)
>  %284 = fcmp ole float %283, -0.000000e+00
>  %temp32.7 = select i1 %284, float %88, float %temp32.3
>  %285 = fmul float %257, %...231
>  %286 = fadd float %285, %...232
>  %287 = fmul float %258, %temp32.5
>  %288 = fadd float %287, %temp32.7
>  %289 = fadd float %286, 0x3F40000000000000
>  %290 = fadd float %288, 0x3F40000000000000
>  %291 = fadd float %243, 0.000000e+00
>  %292 = bitcast float %291 to i32
>  %293 = bitcast float %289 to i32
>  %294 = bitcast float %290 to i32
>  %295 = insertelement <4 x i32> undef, i32 %292, i32 0
>  %296 = insertelement <4 x i32> %295, i32 %293, i32 1
>  %297 = insertelement <4 x i32> %296, i32 %294, i32 2
>  %298 = insertelement <4 x i32> %297, i32 0, i32 3
>  %299 = call <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32> %298, <8 x i32> %112, <4 x i32> %114, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %300 = extractelement <4 x float> %299, i32 0
>  %301 = fadd float %286, 0xBF40000000000000
>  %302 = fadd float %288, 0x3F40000000000000
>  %303 = fadd float %243, 0.000000e+00
>  %304 = fadd float %286, 0x3F40000000000000
>  %305 = fadd float %288, 0xBF40000000000000
>  %306 = fadd float %243, 0.000000e+00
>  %307 = fadd float %286, 0xBF40000000000000
>  %308 = fadd float %288, 0xBF40000000000000
>  %309 = fadd float %243, 0.000000e+00
>  %310 = bitcast float %303 to i32
>  %311 = bitcast float %301 to i32
>  %312 = bitcast float %302 to i32
>  %313 = insertelement <4 x i32> undef, i32 %310, i32 0
>  %314 = insertelement <4 x i32> %313, i32 %311, i32 1
>  %315 = insertelement <4 x i32> %314, i32 %312, i32 2
>  %316 = insertelement <4 x i32> %315, i32 0, i32 3
>  %317 = call <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32> %316, <8 x i32> %112, <4 x i32> %114, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %318 = extractelement <4 x float> %317, i32 0
>  %319 = bitcast float %306 to i32
>  %320 = bitcast float %304 to i32
>  %321 = bitcast float %305 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 0, i32 3
>  %326 = call <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32> %325, <8 x i32> %112, <4 x i32> %114, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %327 = extractelement <4 x float> %326, i32 0
>  %328 = bitcast float %309 to i32
>  %329 = bitcast float %307 to i32
>  %330 = bitcast float %308 to i32
>  %331 = insertelement <4 x i32> undef, i32 %328, i32 0
>  %332 = insertelement <4 x i32> %331, i32 %329, i32 1
>  %333 = insertelement <4 x i32> %332, i32 %330, i32 2
>  %334 = insertelement <4 x i32> %333, i32 0, i32 3
>  %335 = call <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32> %334, <8 x i32> %112, <4 x i32> %114, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %336 = extractelement <4 x float> %335, i32 0
>  %337 = fmul float %300, 6.250000e-02
>  %338 = fmul float %318, 6.250000e-02
>  %339 = fadd float %337, %338
>  %340 = fmul float %327, 6.250000e-02
>  %341 = fadd float %339, %340
>  %342 = fmul float %336, 6.250000e-02
>  %343 = fadd float %341, %342
>  %344 = fadd float %286, 0x3F40000000000000
>  %345 = fadd float %288, 0.000000e+00
>  %346 = fadd float %243, 0.000000e+00
>  %347 = bitcast float %346 to i32
>  %348 = bitcast float %344 to i32
>  %349 = bitcast float %345 to i32
>  %350 = insertelement <4 x i32> undef, i32 %347, i32 0
>  %351 = insertelement <4 x i32> %350, i32 %348, i32 1
>  %352 = insertelement <4 x i32> %351, i32 %349, i32 2
>  %353 = insertelement <4 x i32> %352, i32 0, i32 3
>  %354 = call <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32> %353, <8 x i32> %112, <4 x i32> %114, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %355 = extractelement <4 x float> %354, i32 0
>  %356 = fadd float %286, 0xBF40000000000000
>  %357 = fadd float %288, 0.000000e+00
>  %358 = fadd float %243, 0.000000e+00
>  %359 = bitcast float %358 to i32
>  %360 = bitcast float %356 to i32
>  %361 = bitcast float %357 to i32
>  %362 = insertelement <4 x i32> undef, i32 %359, i32 0
>  %363 = insertelement <4 x i32> %362, i32 %360, i32 1
>  %364 = insertelement <4 x i32> %363, i32 %361, i32 2
>  %365 = insertelement <4 x i32> %364, i32 0, i32 3
>  %366 = call <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32> %365, <8 x i32> %112, <4 x i32> %114, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %367 = extractelement <4 x float> %366, i32 0
>  %368 = extractelement <4 x float> %366, i32 3
>  %369 = fadd float %286, 0.000000e+00
>  %370 = fadd float %288, 0xBF40000000000000
>  %371 = fadd float %243, 0.000000e+00
>  %372 = bitcast float %371 to i32
>  %373 = bitcast float %369 to i32
>  %374 = bitcast float %370 to i32
>  %375 = insertelement <4 x i32> undef, i32 %372, i32 0
>  %376 = insertelement <4 x i32> %375, i32 %373, i32 1
>  %377 = insertelement <4 x i32> %376, i32 %374, i32 2
>  %378 = insertelement <4 x i32> %377, i32 0, i32 3
>  %379 = call <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32> %378, <8 x i32> %112, <4 x i32> %114, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %380 = extractelement <4 x float> %379, i32 0
>  %381 = fadd float %286, 0.000000e+00
>  %382 = fadd float %288, 0x3F40000000000000
>  %383 = fadd float %243, 0.000000e+00
>  %384 = bitcast float %383 to i32
>  %385 = bitcast float %381 to i32
>  %386 = bitcast float %382 to i32
>  %387 = insertelement <4 x i32> undef, i32 %384, i32 0
>  %388 = insertelement <4 x i32> %387, i32 %385, i32 1
>  %389 = insertelement <4 x i32> %388, i32 %386, i32 2
>  %390 = insertelement <4 x i32> %389, i32 0, i32 3
>  %391 = call <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32> %390, <8 x i32> %112, <4 x i32> %114, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %392 = extractelement <4 x float> %391, i32 0
>  %393 = fmul float %355, 1.250000e-01
>  %394 = fmul float %367, 1.250000e-01
>  %395 = fadd float %393, %394
>  %396 = fmul float %380, 1.250000e-01
>  %397 = fadd float %395, %396
>  %398 = fmul float %392, 1.250000e-01
>  %399 = fadd float %397, %398
>  %400 = bitcast float %243 to i32
>  %401 = bitcast float %286 to i32
>  %402 = bitcast float %288 to i32
>  %403 = insertelement <4 x i32> undef, i32 %400, i32 0
>  %404 = insertelement <4 x i32> %403, i32 %401, i32 1
>  %405 = insertelement <4 x i32> %404, i32 %402, i32 2
>  %406 = insertelement <4 x i32> %405, i32 0, i32 3
>  %407 = call <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32> %406, <8 x i32> %112, <4 x i32> %114, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %408 = extractelement <4 x float> %407, i32 0
>  %409 = fadd float %399, %343
>  %410 = fmul float %408, 2.500000e-01
>  %411 = fadd float %410, %409
>  %412 = fcmp olt float %256, 1.000000e+00
>  br i1 %412, label %IF111, label %ENDIF110
>
>ENDIF:                                            ; preds = %main_body, %ENDIF110
>  %temp8.0 = phi float [ %656, %ENDIF110 ], [ 1.000000e+00, %main_body ]
>  %temp31.0 = phi float [ %temp31.1, %ENDIF110 ], [ undef, %main_body ]
>  %413 = fmul float %117, %temp8.0
>  %414 = fadd float %413, %127
>  %415 = fmul float %118, %temp8.0
>  %416 = fadd float %415, %128
>  %417 = fmul float %119, %temp8.0
>  %418 = fadd float %417, %129
>  %419 = bitcast float %98 to i32
>  %420 = icmp eq i32 %419, 0
>  br i1 %420, label %ENDIF224, label %IF225
>
>IF111:                                            ; preds = %IF
>  %421 = fadd float %temp16.1, 0.000000e+00
>  %422 = fadd float %temp16.1, -1.000000e+00
>  %423 = fadd float %temp16.1, -2.000000e+00
>  %424 = call float @llvm.fabs.f32(float %421)
>  %425 = fcmp ole float %424, -0.000000e+00
>  %.233 = select i1 %425, float %59, float 0.000000e+00
>  %426 = call float @llvm.fabs.f32(float %421)
>  %427 = fcmp ole float %426, -0.000000e+00
>  %temp40.5 = select i1 %427, float %60, float 0.000000e+00
>  %428 = call float @llvm.fabs.f32(float %421)
>  %429 = fcmp ole float %428, -0.000000e+00
>  %.234 = select i1 %429, float %61, float 0.000000e+00
>  %430 = call float @llvm.fabs.f32(float %421)
>  %431 = fcmp ole float %430, -0.000000e+00
>  %temp40.7 = select i1 %431, float %62, float 0.000000e+00
>  %432 = call float @llvm.fabs.f32(float %421)
>  %433 = fcmp ole float %432, -0.000000e+00
>  %.235 = select i1 %433, float %63, float 0.000000e+00
>  %434 = call float @llvm.fabs.f32(float %421)
>  %435 = fcmp ole float %434, -0.000000e+00
>  %temp40.9 = select i1 %435, float %64, float 0.000000e+00
>  %436 = call float @llvm.fabs.f32(float %421)
>  %437 = fcmp ole float %436, -0.000000e+00
>  %.236 = select i1 %437, float %65, float 0.000000e+00
>  %438 = call float @llvm.fabs.f32(float %421)
>  %439 = fcmp ole float %438, -0.000000e+00
>  %temp40.11 = select i1 %439, float %66, float 0.000000e+00
>  %440 = call float @llvm.fabs.f32(float %422)
>  %441 = fcmp ole float %440, -0.000000e+00
>  %..233 = select i1 %441, float %67, float %.233
>  %442 = call float @llvm.fabs.f32(float %422)
>  %443 = fcmp ole float %442, -0.000000e+00
>  %temp40.13 = select i1 %443, float %68, float %temp40.5
>  %444 = call float @llvm.fabs.f32(float %422)
>  %445 = fcmp ole float %444, -0.000000e+00
>  %..234 = select i1 %445, float %69, float %.234
>  %446 = call float @llvm.fabs.f32(float %422)
>  %447 = fcmp ole float %446, -0.000000e+00
>  %temp40.15 = select i1 %447, float %70, float %temp40.7
>  %448 = call float @llvm.fabs.f32(float %422)
>  %449 = fcmp ole float %448, -0.000000e+00
>  %..235 = select i1 %449, float %71, float %.235
>  %450 = call float @llvm.fabs.f32(float %422)
>  %451 = fcmp ole float %450, -0.000000e+00
>  %temp40.17 = select i1 %451, float %72, float %temp40.9
>  %452 = call float @llvm.fabs.f32(float %422)
>  %453 = fcmp ole float %452, -0.000000e+00
>  %..236 = select i1 %453, float %73, float %.236
>  %454 = call float @llvm.fabs.f32(float %422)
>  %455 = fcmp ole float %454, -0.000000e+00
>  %temp40.19 = select i1 %455, float %74, float %temp40.11
>  %456 = call float @llvm.fabs.f32(float %423)
>  %457 = fcmp ole float %456, -0.000000e+00
>  %...233 = select i1 %457, float %75, float %..233
>  %458 = call float @llvm.fabs.f32(float %423)
>  %459 = fcmp ole float %458, -0.000000e+00
>  %temp40.21 = select i1 %459, float %76, float %temp40.13
>  %460 = call float @llvm.fabs.f32(float %423)
>  %461 = fcmp ole float %460, -0.000000e+00
>  %...234 = select i1 %461, float %77, float %..234
>  %462 = call float @llvm.fabs.f32(float %423)
>  %463 = fcmp ole float %462, -0.000000e+00
>  %temp40.23 = select i1 %463, float %78, float %temp40.15
>  %464 = call float @llvm.fabs.f32(float %423)
>  %465 = fcmp ole float %464, -0.000000e+00
>  %...235 = select i1 %465, float %79, float %..235
>  %466 = call float @llvm.fabs.f32(float %423)
>  %467 = fcmp ole float %466, -0.000000e+00
>  %temp40.25 = select i1 %467, float %80, float %temp40.17
>  %468 = call float @llvm.fabs.f32(float %423)
>  %469 = fcmp ole float %468, -0.000000e+00
>  %...236 = select i1 %469, float %81, float %..236
>  %470 = call float @llvm.fabs.f32(float %423)
>  %471 = fcmp ole float %470, -0.000000e+00
>  %temp40.27 = select i1 %471, float %82, float %temp40.19
>  %472 = fmul float %168, %...233
>  %473 = fmul float %169, %temp40.21
>  %474 = fadd float %472, %473
>  %475 = fmul float %170, %...234
>  %476 = fadd float %474, %475
>  %477 = fmul float %172, %temp40.23
>  %478 = fadd float %476, %477
>  %479 = call float @llvm.AMDIL.clamp.(float %478, float 0.000000e+00, float 1.000000e+00)
>  %480 = fmul float %168, %...235
>  %481 = fmul float %169, %temp40.25
>  %482 = fadd float %480, %481
>  %483 = fmul float %170, %...236
>  %484 = fadd float %482, %483
>  %485 = fmul float %172, %temp40.27
>  %486 = fadd float %484, %485
>  %487 = call float @llvm.AMDIL.clamp.(float %486, float 0.000000e+00, float 1.000000e+00)
>  %488 = call float @llvm.fabs.f32(float %421)
>  %489 = fcmp ole float %488, -0.000000e+00
>  %.237 = select i1 %489, float %85, float 0.000000e+00
>  %490 = call float @llvm.fabs.f32(float %421)
>  %491 = fcmp ole float %490, -0.000000e+00
>  %temp48.1 = select i1 %491, float %86, float 0.000000e+00
>  %492 = call float @llvm.fabs.f32(float %421)
>  %493 = fcmp ole float %492, -0.000000e+00
>  %.238 = select i1 %493, float %83, float 0.000000e+00
>  %494 = call float @llvm.fabs.f32(float %421)
>  %495 = fcmp ole float %494, -0.000000e+00
>  %temp48.3 = select i1 %495, float %84, float 0.000000e+00
>  %496 = call float @llvm.fabs.f32(float %422)
>  %497 = fcmp ole float %496, -0.000000e+00
>  %..237 = select i1 %497, float %89, float %.237
>  %498 = call float @llvm.fabs.f32(float %422)
>  %499 = fcmp ole float %498, -0.000000e+00
>  %temp40.29 = select i1 %499, float %90, float %temp48.1
>  %500 = call float @llvm.fabs.f32(float %422)
>  %501 = fcmp ole float %500, -0.000000e+00
>  %..238 = select i1 %501, float %87, float %.238
>  %502 = call float @llvm.fabs.f32(float %422)
>  %503 = fcmp ole float %502, -0.000000e+00
>  %temp40.31 = select i1 %503, float %88, float %temp48.3
>  %504 = call float @llvm.fabs.f32(float %423)
>  %505 = fcmp ole float %504, -0.000000e+00
>  %...237 = select i1 %505, float %93, float %..237
>  %506 = call float @llvm.fabs.f32(float %423)
>  %507 = fcmp ole float %506, -0.000000e+00
>  %temp40.33 = select i1 %507, float %94, float %temp40.29
>  %508 = call float @llvm.fabs.f32(float %423)
>  %509 = fcmp ole float %508, -0.000000e+00
>  %...238 = select i1 %509, float %91, float %..238
>  %510 = call float @llvm.fabs.f32(float %423)
>  %511 = fcmp ole float %510, -0.000000e+00
>  %temp40.35 = select i1 %511, float %92, float %temp40.31
>  %512 = fmul float %479, %...237
>  %513 = fadd float %512, %...238
>  %514 = fmul float %487, %temp40.33
>  %515 = fadd float %514, %temp40.35
>  %516 = fadd float %513, 0x3F40000000000000
>  %517 = fadd float %515, 0x3F40000000000000
>  %518 = fadd float %243, 0.000000e+00
>  %519 = bitcast float %518 to i32
>  %520 = bitcast float %516 to i32
>  %521 = bitcast float %517 to i32
>  %522 = insertelement <4 x i32> undef, i32 %519, i32 0
>  %523 = insertelement <4 x i32> %522, i32 %520, i32 1
>  %524 = insertelement <4 x i32> %523, i32 %521, i32 2
>  %525 = insertelement <4 x i32> %524, i32 0, i32 3
>  %526 = call <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32> %525, <8 x i32> %112, <4 x i32> %114, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %527 = extractelement <4 x float> %526, i32 0
>  %528 = fadd float %513, 0xBF40000000000000
>  %529 = fadd float %515, 0x3F40000000000000
>  %530 = fadd float %243, 0.000000e+00
>  %531 = fadd float %513, 0x3F40000000000000
>  %532 = fadd float %515, 0xBF40000000000000
>  %533 = fadd float %243, 0.000000e+00
>  %534 = fadd float %513, 0xBF40000000000000
>  %535 = fadd float %515, 0xBF40000000000000
>  %536 = fadd float %243, 0.000000e+00
>  %537 = bitcast float %530 to i32
>  %538 = bitcast float %528 to i32
>  %539 = bitcast float %529 to i32
>  %540 = insertelement <4 x i32> undef, i32 %537, i32 0
>  %541 = insertelement <4 x i32> %540, i32 %538, i32 1
>  %542 = insertelement <4 x i32> %541, i32 %539, i32 2
>  %543 = insertelement <4 x i32> %542, i32 0, i32 3
>  %544 = call <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32> %543, <8 x i32> %112, <4 x i32> %114, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %545 = extractelement <4 x float> %544, i32 0
>  %546 = bitcast float %533 to i32
>  %547 = bitcast float %531 to i32
>  %548 = bitcast float %532 to i32
>  %549 = insertelement <4 x i32> undef, i32 %546, i32 0
>  %550 = insertelement <4 x i32> %549, i32 %547, i32 1
>  %551 = insertelement <4 x i32> %550, i32 %548, i32 2
>  %552 = insertelement <4 x i32> %551, i32 0, i32 3
>  %553 = call <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32> %552, <8 x i32> %112, <4 x i32> %114, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %554 = extractelement <4 x float> %553, i32 0
>  %555 = bitcast float %536 to i32
>  %556 = bitcast float %534 to i32
>  %557 = bitcast float %535 to i32
>  %558 = insertelement <4 x i32> undef, i32 %555, i32 0
>  %559 = insertelement <4 x i32> %558, i32 %556, i32 1
>  %560 = insertelement <4 x i32> %559, i32 %557, i32 2
>  %561 = insertelement <4 x i32> %560, i32 0, i32 3
>  %562 = call <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32> %561, <8 x i32> %112, <4 x i32> %114, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %563 = extractelement <4 x float> %562, i32 0
>  %564 = fmul float %527, 6.250000e-02
>  %565 = fmul float %545, 6.250000e-02
>  %566 = fadd float %564, %565
>  %567 = fmul float %554, 6.250000e-02
>  %568 = fadd float %566, %567
>  %569 = fmul float %563, 6.250000e-02
>  %570 = fadd float %568, %569
>  %571 = fadd float %513, 0x3F40000000000000
>  %572 = fadd float %515, 0.000000e+00
>  %573 = fadd float %243, 0.000000e+00
>  %574 = bitcast float %573 to i32
>  %575 = bitcast float %571 to i32
>  %576 = bitcast float %572 to i32
>  %577 = insertelement <4 x i32> undef, i32 %574, i32 0
>  %578 = insertelement <4 x i32> %577, i32 %575, i32 1
>  %579 = insertelement <4 x i32> %578, i32 %576, i32 2
>  %580 = insertelement <4 x i32> %579, i32 0, i32 3
>  %581 = call <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32> %580, <8 x i32> %112, <4 x i32> %114, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %582 = extractelement <4 x float> %581, i32 0
>  %583 = fadd float %513, 0xBF40000000000000
>  %584 = fadd float %515, 0.000000e+00
>  %585 = fadd float %243, 0.000000e+00
>  %586 = bitcast float %585 to i32
>  %587 = bitcast float %583 to i32
>  %588 = bitcast float %584 to i32
>  %589 = insertelement <4 x i32> undef, i32 %586, i32 0
>  %590 = insertelement <4 x i32> %589, i32 %587, i32 1
>  %591 = insertelement <4 x i32> %590, i32 %588, i32 2
>  %592 = insertelement <4 x i32> %591, i32 0, i32 3
>  %593 = call <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32> %592, <8 x i32> %112, <4 x i32> %114, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %594 = extractelement <4 x float> %593, i32 0
>  %595 = fadd float %513, 0.000000e+00
>  %596 = fadd float %515, 0xBF40000000000000
>  %597 = fadd float %243, 0.000000e+00
>  %598 = bitcast float %597 to i32
>  %599 = bitcast float %595 to i32
>  %600 = bitcast float %596 to i32
>  %601 = insertelement <4 x i32> undef, i32 %598, i32 0
>  %602 = insertelement <4 x i32> %601, i32 %599, i32 1
>  %603 = insertelement <4 x i32> %602, i32 %600, i32 2
>  %604 = insertelement <4 x i32> %603, i32 0, i32 3
>  %605 = call <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32> %604, <8 x i32> %112, <4 x i32> %114, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %606 = extractelement <4 x float> %605, i32 0
>  %607 = fadd float %513, 0.000000e+00
>  %608 = fadd float %515, 0x3F40000000000000
>  %609 = fadd float %243, 0.000000e+00
>  %610 = bitcast float %609 to i32
>  %611 = bitcast float %607 to i32
>  %612 = bitcast float %608 to i32
>  %613 = insertelement <4 x i32> undef, i32 %610, i32 0
>  %614 = insertelement <4 x i32> %613, i32 %611, i32 1
>  %615 = insertelement <4 x i32> %614, i32 %612, i32 2
>  %616 = insertelement <4 x i32> %615, i32 0, i32 3
>  %617 = call <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32> %616, <8 x i32> %112, <4 x i32> %114, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %618 = extractelement <4 x float> %617, i32 0
>  %619 = fmul float %582, 1.250000e-01
>  %620 = fmul float %594, 1.250000e-01
>  %621 = fadd float %619, %620
>  %622 = fmul float %606, 1.250000e-01
>  %623 = fadd float %621, %622
>  %624 = fmul float %618, 1.250000e-01
>  %625 = fadd float %623, %624
>  %626 = bitcast float %243 to i32
>  %627 = bitcast float %513 to i32
>  %628 = bitcast float %515 to i32
>  %629 = insertelement <4 x i32> undef, i32 %626, i32 0
>  %630 = insertelement <4 x i32> %629, i32 %627, i32 1
>  %631 = insertelement <4 x i32> %630, i32 %628, i32 2
>  %632 = insertelement <4 x i32> %631, i32 0, i32 3
>  %633 = call <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32> %632, <8 x i32> %112, <4 x i32> %114, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %634 = extractelement <4 x float> %633, i32 0
>  %635 = fadd float %625, %570
>  %636 = fmul float %634, 2.500000e-01
>  %637 = fadd float %636, %635
>  %638 = fcmp oge float %423, 0.000000e+00
>  %.239 = select i1 %638, float 1.000000e+00, float %637
>  %639 = fsub float 1.000000e+00, %256
>  %640 = fmul float %411, %256
>  %641 = fmul float %.239, %639
>  %642 = fadd float %640, %641
>  br label %ENDIF110
>
>ENDIF110:                                         ; preds = %IF, %IF111
>  %temp5.0 = phi float [ %642, %IF111 ], [ %411, %IF ]
>  %temp31.1 = phi float [ 0.000000e+00, %IF111 ], [ %368, %IF ]
>  %643 = fsub float %131, %95
>  %644 = fsub float %132, %96
>  %645 = fsub float %133, %97
>  %646 = fmul float %643, %643
>  %647 = fmul float %644, %644
>  %648 = fadd float %647, %646
>  %649 = fmul float %645, %645
>  %650 = fadd float %648, %649
>  %651 = fmul float %650, %58
>  %652 = fadd float %651, %57
>  %653 = call float @llvm.AMDIL.clamp.(float %652, float 0.000000e+00, float 1.000000e+00)
>  %654 = fsub float 1.000000e+00, %653
>  %655 = fmul float %temp5.0, %654
>  %656 = fadd float %653, %655
>  br label %ENDIF
>
>IF225:                                            ; preds = %ENDIF
>  %657 = fmul float %414, 0x3FCB333340000000
>  %658 = fmul float %416, 0x3FE6E48E80000000
>  %659 = fadd float %658, %657
>  %660 = fmul float %418, 0x3FB2752540000000
>  %661 = fadd float %659, %660
>  %662 = fdiv float 1.000000e+00, %661
>  %663 = fmul float %662, %120
>  %664 = fsub float 1.000000e+00, %temp8.0
>  %665 = fmul float %664, %663
>  %666 = fsub float 1.000000e+00, %665
>  %667 = fmul float %418, %666
>  %668 = fmul float %416, %666
>  %669 = fmul float %414, %666
>  %670 = fmul float %666, 5.000000e-01
>  %671 = fadd float %670, 5.000000e-01
>  %672 = fsub float 1.000000e+00, %671
>  %673 = fmul float %669, %671
>  %674 = fmul float %667, %672
>  %675 = fadd float %673, %674
>  %676 = fsub float 1.000000e+00, %671
>  %677 = fmul float %668, %671
>  %678 = fmul float %668, %676
>  %679 = fadd float %677, %678
>  %680 = fsub float 1.000000e+00, %671
>  %681 = fmul float %667, %671
>  %682 = fmul float %669, %680
>  %683 = fadd float %681, %682
>  br label %ENDIF224
>
>ENDIF224:                                         ; preds = %ENDIF, %IF225
>  %temp5.1 = phi float [ %675, %IF225 ], [ %414, %ENDIF ]
>  %temp6.0 = phi float [ %679, %IF225 ], [ %416, %ENDIF ]
>  %temp7.0 = phi float [ %683, %IF225 ], [ %418, %ENDIF ]
>  %684 = fadd float %143, -1.000000e+00
>  %685 = fmul float %45, %684
>  %686 = fadd float %685, 1.000000e+00
>  %687 = fadd float %143, %35
>  %688 = fadd float %28, -1.000000e+00
>  %689 = fadd float %29, -1.000000e+00
>  %690 = fadd float %30, -1.000000e+00
>  %691 = call float @llvm.AMDIL.clamp.(float %687, float 0.000000e+00, float 1.000000e+00)
>  %692 = fmul float %691, %688
>  %693 = fadd float %692, 1.000000e+00
>  %694 = fmul float %691, %689
>  %695 = fadd float %694, 1.000000e+00
>  %696 = fmul float %691, %690
>  %697 = fadd float %696, 1.000000e+00
>  %698 = fmul float %temp5.1, %693
>  %699 = fmul float %temp6.0, %695
>  %700 = fmul float %temp7.0, %697
>  %701 = fmul float %686, %31
>  %702 = fmul float %701, %130
>  %703 = fsub float %702, %701
>  %704 = fmul float %37, %703
>  %705 = fadd float %704, %701
>  %706 = fmul float %124, %121
>  %707 = fmul float %125, %122
>  %708 = fadd float %707, %706
>  %709 = fmul float %126, %123
>  %710 = fadd float %708, %709
>  %711 = fadd float %710, %710
>  %712 = fmul float %124, %124
>  %713 = fmul float %125, %125
>  %714 = fadd float %713, %712
>  %715 = fmul float %126, %126
>  %716 = fadd float %714, %715
>  %717 = fmul float %716, %121
>  %718 = fmul float %716, %122
>  %719 = fmul float %716, %123
>  %720 = fmul float %711, %124
>  %721 = fsub float %720, %717
>  %722 = fmul float %711, %125
>  %723 = fsub float %722, %718
>  %724 = fmul float %711, %126
>  %725 = fsub float %724, %719
>  %726 = insertelement <4 x float> undef, float %721, i32 0
>  %727 = insertelement <4 x float> %726, float %723, i32 1
>  %728 = insertelement <4 x float> %727, float %725, i32 2
>  %729 = insertelement <4 x float> %728, float %temp31.0, i32 3
>  %730 = call <4 x float> @llvm.AMDGPU.cube(<4 x float> %729)
>  %731 = extractelement <4 x float> %730, i32 0
>  %732 = extractelement <4 x float> %730, i32 1
>  %733 = extractelement <4 x float> %730, i32 2
>  %734 = extractelement <4 x float> %730, i32 3
>  %735 = call float @llvm.fabs.f32(float %733)
>  %736 = fdiv float 1.000000e+00, %735
>  %737 = fmul float %731, %736
>  %738 = fadd float %737, 1.500000e+00
>  %739 = fmul float %732, %736
>  %740 = fadd float %739, 1.500000e+00
>  %741 = bitcast float %740 to i32
>  %742 = bitcast float %738 to i32
>  %743 = bitcast float %734 to i32
>  %744 = insertelement <4 x i32> undef, i32 %741, i32 0
>  %745 = insertelement <4 x i32> %744, i32 %742, i32 1
>  %746 = insertelement <4 x i32> %745, i32 %743, i32 2
>  %747 = call <4 x float> @llvm.SI.image.sample.v4i32(<4 x i32> %746, <8 x i32> %104, <4 x i32> %106, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
>  %748 = extractelement <4 x float> %747, i32 0
>  %749 = extractelement <4 x float> %747, i32 1
>  %750 = extractelement <4 x float> %747, i32 2
>  %751 = fmul float %748, %54
>  %752 = fmul float %749, %54
>  %753 = fmul float %750, %54
>  %754 = fmul float %149, %751
>  %755 = fmul float %149, %752
>  %756 = fmul float %149, %753
>  %757 = fmul float %754, %25
>  %758 = fmul float %755, %26
>  %759 = fmul float %756, %27
>  %760 = fmul float %temp5.1, %693
>  %761 = fsub float %760, %40
>  %762 = fmul float %temp6.0, %695
>  %763 = fsub float %762, %40
>  %764 = fmul float %temp7.0, %697
>  %765 = fsub float %764, %40
>  %766 = fmul float %761, %41
>  %767 = fmul float %763, %41
>  %768 = fmul float %765, %41
>  %769 = call float @llvm.AMDIL.clamp.(float %766, float 0.000000e+00, float 1.000000e+00)
>  %770 = call float @llvm.AMDIL.clamp.(float %767, float 0.000000e+00, float 1.000000e+00)
>  %771 = call float @llvm.AMDIL.clamp.(float %768, float 0.000000e+00, float 1.000000e+00)
>  %772 = fmul float %757, %769
>  %773 = fsub float %772, %757
>  %774 = fmul float %758, %770
>  %775 = fsub float %774, %758
>  %776 = fmul float %759, %771
>  %777 = fsub float %776, %759
>  %778 = fmul float %39, %773
>  %779 = fadd float %778, %757
>  %780 = fmul float %39, %775
>  %781 = fadd float %780, %758
>  %782 = fmul float %39, %777
>  %783 = fadd float %782, %759
>  %784 = fmul float %779, %779
>  %785 = fsub float %784, %779
>  %786 = fmul float %781, %781
>  %787 = fsub float %786, %781
>  %788 = fmul float %783, %783
>  %789 = fsub float %788, %783
>  %790 = fmul float %38, %785
>  %791 = fadd float %790, %779
>  %792 = fmul float %38, %787
>  %793 = fadd float %792, %781
>  %794 = fmul float %38, %789
>  %795 = fadd float %794, %783
>  %796 = fmul float %791, 0x3FD322D0E0000000
>  %797 = fmul float %793, 0x3FE2C8B440000000
>  %798 = fadd float %797, %796
>  %799 = fmul float %795, 0x3FBD2F1AA0000000
>  %800 = fadd float %798, %799
>  %801 = fsub float 1.000000e+00, %32
>  %802 = fmul float %791, %32
>  %803 = fmul float %800, %801
>  %804 = fadd float %802, %803
>  %805 = fsub float 1.000000e+00, %33
>  %806 = fmul float %793, %33
>  %807 = fmul float %800, %805
>  %808 = fadd float %806, %807
>  %809 = fsub float 1.000000e+00, %34
>  %810 = fmul float %795, %34
>  %811 = fmul float %800, %809
>  %812 = fadd float %810, %811
>  %813 = fmul float %140, %698
>  %814 = fadd float %813, %804
>  %815 = fmul float %141, %699
>  %816 = fadd float %815, %808
>  %817 = fmul float %142, %700
>  %818 = fadd float %817, %812
>  %819 = fsub float %42, %131
>  %820 = fsub float %43, %132
>  %821 = fsub float %44, %133
>  %822 = fmul float %819, %819
>  %823 = fmul float %820, %820
>  %824 = fadd float %823, %822
>  %825 = fmul float %821, %821
>  %826 = fadd float %824, %825
>  %827 = call float @llvm.sqrt.f32(float %826)
>  %828 = fmul float %827, %48
>  %829 = fadd float %828, %46
>  %830 = call float @llvm.AMDIL.clamp.(float %829, float 0.000000e+00, float 1.000000e+00)
>  %831 = call float @llvm.minnum.f32(float %830, float %47)
>  %832 = call float @llvm.fabs.f32(float %36)
>  %833 = fmul float %814, %53
>  %834 = fmul float %816, %53
>  %835 = fmul float %818, %53
>  %836 = fmul float %52, %134
>  %837 = fcmp ole float %832, -0.000000e+00
>  %.240 = select i1 %837, float %705, float %836
>  %838 = fmul float %831, %831
>  %839 = fmul float %53, %814
>  %840 = fsub float %49, %839
>  %841 = fmul float %53, %816
>  %842 = fsub float %50, %841
>  %843 = fmul float %53, %818
>  %844 = fsub float %51, %843
>  %845 = fmul float %838, %840
>  %846 = fadd float %845, %833
>  %847 = fmul float %838, %842
>  %848 = fadd float %847, %834
>  %849 = fmul float %838, %844
>  %850 = fadd float %849, %835
>  %851 = call i32 @llvm.SI.packf16(float %846, float %848)
>  %852 = bitcast i32 %851 to float
>  %853 = call i32 @llvm.SI.packf16(float %850, float %.240)
>  %854 = bitcast i32 %853 to float
>  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %852, float %854, float %852, float %854)
>  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.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #1
>
>; Function Attrs: readnone
>declare float @llvm.AMDIL.clamp.(float, float, float) #2
>
>; Function Attrs: nounwind readnone
>declare float @llvm.fabs.f32(float) #1
>
>; Function Attrs: nounwind readnone
>declare <4 x float> @llvm.SI.image.sample.c.l.v4i32(<4 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #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.image.sample.v4i32(<4 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #1
>
>; Function Attrs: nounwind readnone
>declare float @llvm.sqrt.f32(float) #1
>
>; 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}
>crash_20151212041947_1.dmp[10518]: Uploading dump (out-of-process)
>/tmp/dumps/crash_20151212041947_1.dmp
>/mnt/multimedia2/Games/Steam/SteamApps/common/Counter-Strike Global Offensive/csgo.sh: Zeile 57:  9966 Speicherzugriffsfehler  ${DEBUGGER} "${GAMEROOT}"/${GAMEEXE} "$@"
>crash_20151212041947_1.dmp[10518]: Finished uploading minidump (out-of-process): success = yes
>crash_20151212041947_1.dmp[10518]: response: Discarded=1
>crash_20151212041947_1.dmp[10518]: file ''/tmp/dumps/crash_20151212041947_1.dmp'', upload yes: ''Discarded=1''

Actions: View

Attachments on bug 93360: 120470 | 120476 | 120488

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.