Attachment 5155 Details for Bug 6412 – big patch for mach64 DRI - try 4

[patch] big patch for mach64 DRI - try 4

mach64_dd_vb-4.patch (text/plain), 42.19 KB, created by George - on 2006-04-02 01:00:35 UTC

(hide)

Description:

Filename:

MIME Type:

Creator: George -

Created: 2006-04-02 01:00:35 UTC

Size: 42.19 KB

patch

obsolete

>Common subdirectories: mach64-vtx/CVS and mach64-vtx-1/CVS
>diff -au -x '*swp' -x '*o' -x 'depend*' mach64-vtx/mach64_context.h mach64-vtx-1/mach64_context.h
>--- mach64-vtx/mach64_context.h	2006-03-12 00:02:32.000000000 +0200
>+++ mach64-vtx-1/mach64_context.h	2006-04-01 16:59:07.000000000 +0300
>@@ -58,6 +58,9 @@
> /* Native vertex format */
> #define MACH64_NATIVE_VTXFMT		1
> 
>+/* premultiply texture coordinates by homogenous coordinate */
>+/* #define MACH64_PREMULT_TEXCOORDS */
>+
> /* Flags for what context state needs to be updated:
>  */
> #define MACH64_NEW_ALPHA		0x0001
>@@ -94,14 +97,49 @@
> 
> /* The size of this union is not of relevence:
>  */
>-union mach64_vertex_t {
>+#define TAG(x) mach64##x
>+typedef struct {
>+   GLubyte blue, green, red, alpha;
>+} TAG(_color_t);
>+
>+typedef union {
>+   struct {
>+      GLfloat u3, v3, q3;
>+      GLfloat u2, v2, q2;
>+      GLfloat u1, v1, q1;
>+      GLfloat u0, v0, q0;
>+      TAG(_color_t) specular;
>+      GLuint z;
>+      TAG(_color_t) color;
>+      GLshort y;
>+      GLshort x;
>+   } pv;			/* unused */
>+   struct {
>+      GLfloat u3, v3, q3;	/* unused */
>+      GLfloat u2, v2, q2;	/* unused */
>+      GLfloat u1, v1;		/* tex1 */
>+      GLfloat w1;
>+      GLfloat u0, v0;		/* tex0 */
>+      GLfloat w;
>+      TAG(_color_t) specular;	/* notex */
>+      GLuint z;			/* tiny */
>+      TAG(_color_t) color;
>+      GLshort y;
>+      GLshort x;
>+   } v;
>+   struct {
>+      GLfloat f[13];		/* pad to sizeof(struct v) */
>+      GLuint z;			/* tiny */
>+      TAG(_color_t) color;
>+      GLshort y;
>+      GLshort x;
>+   } tv;
>    GLfloat f[16];
>    GLuint ui[16];
>    GLushort us2[16][2];
>    GLubyte ub4[16][4];
>-};
>-
>-typedef union mach64_vertex_t mach64Vertex, *mach64VertexPtr;
>+} TAG(Vertex), *TAG(VertexPtr);
>+#undef TAG
> 
> #else
> 
>@@ -330,25 +368,10 @@
>  * Byte ordering
>  */
> #if MESA_LITTLE_ENDIAN == 1
>-#define LE32_IN( x )		( *(GLuint *)(x) )
>-#define LE32_IN_FLOAT( x )	( *(GLfloat *)(x) )
> #define LE32_OUT( x, y )	do { *(GLuint *)(x) = (y); } while (0)
>-#define LE32_OUT_FLOAT( x, y )	do { *(GLfloat *)(x) = (y); } while (0)
> #else
> #include <byteswap.h>
>-#define LE32_IN( x )		bswap_32( *(GLuint *)(x) )
>-#define LE32_IN_FLOAT( x )						\
>-({									\
>-   GLuint __tmp = bswap_32( *(GLuint *)(x) );				\
>-   *(GLfloat *)&__tmp;							\
>-})
> #define LE32_OUT( x, y )	do { *(GLuint *)(x) = bswap_32( y ); } while (0)
>-#define LE32_OUT_FLOAT( x, y )						\
>-do {									\
>-   GLuint __tmp;							\
>-   *(GLfloat *)&__tmp = (y);						\
>-   *(GLuint *)(x) = bswap_32( __tmp );					\
>-} while (0)
> #endif
> 
> /* ================================================================
>Only in mach64-vtx: mach64_native_vb.c
>Only in mach64-vtx: mach64_native_vbtmp.h
>diff -au -x '*swp' -x '*o' -x 'depend*' mach64-vtx/mach64_tris.c mach64-vtx-1/mach64_tris.c
>--- mach64-vtx/mach64_tris.c	2006-03-27 04:50:58.000000000 +0300
>+++ mach64-vtx-1/mach64_tris.c	2006-04-01 17:37:20.000000000 +0300
>@@ -70,10 +70,22 @@
>  *                    Emit primitives as inline vertices               *
>  ***********************************************************************/
> 
>+/*
>+ * TODO: LE32 with non-x86 is untested
>+ *
>+ * TODO: add comments in draw_triangle about vertex format (mach64
>+ * TODO: does not have VF_CNTL, VERTEX_FMT registers ...), and
>+ * TODO: about vbsize, ooa, and other mystic calculations
>+ *
>+ * TODO: order the fields of mach64Vertex from xy to w to sec_w and
>+ * TODO: submit them in reserse order (asm ? perf regression ?)
>+ */
>+
> #if defined(USE_X86_ASM)
>+/* x86 does not need LE32_OUT() because it is LE */
> #define DO_COPY_VERTEX( vb, vertsize, v, n, m )					\
> do {										\
>-   register const CARD32 *__p __asm__( "esi" ) = (CARD32 *)v + 10 - vertsize;	\
>+   register const CARD32 *__p __asm__( "esi" ) = (CARD32 *)v + 16 - vertsize;	\
>    register int __s __asm__( "ecx" ) = vertsize;				\
>    if ( vertsize > 7 ) {							\
>       *vb++ = (2 << 16) | ADRINDEX( MACH64_VERTEX_##n##_SECONDARY_S );		\
>@@ -91,20 +103,20 @@
> #else
> #define DO_COPY_VERTEX( vb, vertsize, v, n, m )				\
> do {									\
>-   CARD32 *__p = (CARD32 *)v + 10 - vertsize;				\
>+   CARD32 *__p = (CARD32 *)v + 16 - vertsize;				\
>    int __s = vertsize;							\
>    if ( vertsize > 7 ) {						\
>       LE32_OUT( vb++, (2 << 16) |					\
> 	    	      ADRINDEX( MACH64_VERTEX_##n##_SECONDARY_S ) );	\
>-      *vb++ = *__p++;							\
>-      *vb++ = *__p++;							\
>-      *vb++ = *__p++;							\
>+      LE32_OUT( vb++ , *__p++ );					\
>+      LE32_OUT( vb++ , *__p++ );					\
>+      LE32_OUT( vb++ , *__p++ );					\
>       __s -= 3;								\
>    }									\
>    LE32_OUT( vb++, ((__s - 1 + m) << 16) |				\
> 	           (ADRINDEX( MACH64_VERTEX_##n##_X_Y ) - (__s - 1)) );	\
>    while ( __s-- ) {							\
>-      *vb++ = *__p++;							\
>+      LE32_OUT( vb++, *__p++ );						\
>    }									\
> } while (0)
> #endif
>@@ -124,8 +136,6 @@
>    const GLuint vertsize = mmesa->vertex_size;
>    GLint a;
>    GLfloat ooa;
>-   GLuint xy;
>-   const GLuint xyoffset = 9;
>    GLint xx[3], yy[3]; /* 2 fractional bits for hardware */
>    unsigned vbsiz = (vertsize + (vertsize > 7 ? 2 : 1)) * 4 + 2;
>    CARD32 *vb, *vbchk;
>@@ -142,17 +152,14 @@
>       mach64_print_vertex( ctx, v3 );
>    }
>    
>-   xy = LE32_IN( &v0->ui[xyoffset] );
>-   xx[0] = (GLshort)( xy >> 16 );
>-   yy[0] = (GLshort)( xy & 0xffff );
>-   
>-   xy = LE32_IN( &v1->ui[xyoffset] );
>-   xx[1] = (GLshort)( xy >> 16 );
>-   yy[1] = (GLshort)( xy & 0xffff );
>-   
>-   xy = LE32_IN( &v3->ui[xyoffset] );
>-   xx[2] = (GLshort)( xy >> 16 );
>-   yy[2] = (GLshort)( xy & 0xffff );
>+   xx[0] = v0->v.x;
>+   yy[0] = v0->v.y;
>+   
>+   xx[1] = v1->v.x;
>+   yy[1] = v1->v.y;
>+   
>+   xx[2] = v3->v.x;
>+   yy[2] = v3->v.y;
> 	   
>    a = (xx[0] - xx[2]) * (yy[1] - yy[2]) -
>        (yy[0] - yy[2]) * (xx[1] - xx[2]);
>@@ -176,9 +183,8 @@
>    COPY_VERTEX_OOA( vb, vertsize, v3, 3 );
>    LE32_OUT( vb++, *(CARD32 *)&ooa );
> 
>-   xy = LE32_IN( &v2->ui[xyoffset] );
>-   xx[0] = (GLshort)( xy >> 16 );
>-   yy[0] = (GLshort)( xy & 0xffff );
>+   xx[0] = v2->v.x;
>+   yy[0] = v2->v.y;
> 	   
>    a = (xx[0] - xx[2]) * (yy[1] - yy[2]) -
>        (yy[0] - yy[2]) * (xx[1] - xx[2]);
>@@ -429,8 +435,6 @@
>    GLuint vertsize = mmesa->vertex_size;
>    GLint a;
>    GLfloat ooa;
>-   GLuint xy;
>-   const GLuint xyoffset = 9;
>    GLint xx[3], yy[3]; /* 2 fractional bits for hardware */
>    unsigned vbsiz = (vertsize + (vertsize > 7 ? 2 : 1)) * 3 + 1;
>    CARD32 *vb, *vbchk;
>@@ -445,17 +449,14 @@
>       mach64_print_vertex( ctx, v2 );
>    }
>    
>-   xy = LE32_IN( &v0->ui[xyoffset] );
>-   xx[0] = (GLshort)( xy >> 16 );
>-   yy[0] = (GLshort)( xy & 0xffff );
>-   
>-   xy = LE32_IN( &v1->ui[xyoffset] );
>-   xx[1] = (GLshort)( xy >> 16 );
>-   yy[1] = (GLshort)( xy & 0xffff );
>-   
>-   xy = LE32_IN( &v2->ui[xyoffset] );
>-   xx[2] = (GLshort)( xy >> 16 );
>-   yy[2] = (GLshort)( xy & 0xffff );
>+   xx[0] = v0->v.x;
>+   yy[0] = v0->v.y;
>+   
>+   xx[1] = v1->v.x;
>+   yy[1] = v1->v.y;
>+   
>+   xx[2] = v2->v.x;
>+   yy[2] = v2->v.y;
> 	   
>    a = (xx[0] - xx[2]) * (yy[1] - yy[2]) -
>        (yy[0] - yy[2]) * (xx[1] - xx[2]);
>@@ -675,9 +676,7 @@
>    const GLuint vertsize = mmesa->vertex_size;
>    GLint width = (GLint)(mmesa->glCtx->Line._Width * 2.0); /* 2 fractional bits for hardware */
>    GLfloat ooa;
>-   GLuint *pxy0, *pxy1;
>-   GLuint xy0old, xy0, xy1old, xy1;
>-   const GLuint xyoffset = 9;
>+   GLshort x0old, y0old, x1old, y1old;
>    GLint x0, y0, x1, y1;
>    GLint dx, dy, ix, iy;
>    unsigned vbsiz = (vertsize + (vertsize > 7 ? 2 : 1)) * 4 + 2;
>@@ -694,17 +693,11 @@
>    if( !width )
>       width = 1;	/* round to the nearest supported width */
>       
>-   pxy0 = &v0->ui[xyoffset];
>-   xy0old = *pxy0;
>-   xy0 = LE32_IN( &xy0old );
>-   x0 = (GLshort)( xy0 >> 16 );
>-   y0 = (GLshort)( xy0 & 0xffff );
>-   
>-   pxy1 = &v1->ui[xyoffset];
>-   xy1old = *pxy1;
>-   xy1 = LE32_IN( &xy1old );
>-   x1 = (GLshort)( xy1 >> 16 );
>-   y1 = (GLshort)( xy1 & 0xffff );
>+   x0 = x0old = v0->v.x;
>+   y0 = y0old = v0->v.y;
>+   
>+   x1 = x1old = v1->v.x;
>+   y1 = y1old = v1->v.y;
>    
>    if ( (dx = x1 - x0) < 0 ) {
>       dx = -dx;
>@@ -727,22 +720,31 @@
>    vb = (CARD32 *)mach64AllocDmaLow( mmesa, vbsiz * sizeof(CARD32) );
>    vbchk = vb + vbsiz;
> 
>-   LE32_OUT( pxy0, (( x0 - ix ) << 16) | (( y0 - iy ) & 0xffff) );
>+   v0->v.x = x0 - ix;
>+   v0->v.y = y0 - iy;
>    COPY_VERTEX( vb, vertsize, v0, 1 );
>-   LE32_OUT( pxy1, (( x1 - ix ) << 16) | (( y1 - iy ) & 0xffff) );
>+
>+   v1->v.x = x1 - ix;
>+   v1->v.y = y1 - iy;
>    COPY_VERTEX( vb, vertsize, v1, 2 );
>-   LE32_OUT( pxy0, (( x0 + ix ) << 16) | (( y0 + iy ) & 0xffff) );
>+
>+   v0->v.x = x0 + ix;
>+   v0->v.y = y0 + iy;
>    COPY_VERTEX_OOA( vb, vertsize, v0, 3 );
>    LE32_OUT( vb++, *(CARD32 *)&ooa );
> 
>    ooa = -ooa;
>    
>-   LE32_OUT( pxy1, (( x1 + ix ) << 16) | (( y1 + iy ) & 0xffff) );
>+   v1->v.x = x1 + ix;
>+   v1->v.y = y1 + iy;
>    COPY_VERTEX_OOA( vb, vertsize, v1, 1 );
>    LE32_OUT( vb++, *(CARD32 *)&ooa );
> 
>-   *pxy0 = xy0old;
>-   *pxy1 = xy1old;
>+   v0->v.x = x0old;
>+   v0->v.y = y0old;
>+   
>+   v1->v.x = x1old;
>+   v1->v.y = y1old;
> #else /* !MACH64_NATIVE_VTXFMT */
>    GLuint vertsize = mmesa->vertex_size;
>    GLint coloridx;
>@@ -963,9 +965,7 @@
>    const GLuint vertsize = mmesa->vertex_size;
>    GLint sz = (GLint)(mmesa->glCtx->Point._Size * 2.0); /* 2 fractional bits for hardware */
>    GLfloat ooa;
>-   GLuint *pxy;
>-   GLuint xyold, xy;
>-   const GLuint xyoffset = 9;
>+   GLshort xold, yold;
>    GLint x, y;
>    unsigned vbsiz = (vertsize + (vertsize > 7 ? 2 : 1)) * 4 + 2;
>    CARD32 *vb, *vbchk;
>@@ -979,32 +979,36 @@
>    if( !sz )
>       sz = 1;	/* round to the nearest supported size */
>       
>-   pxy = &v0->ui[xyoffset];
>-   xyold = *pxy;
>-   xy = LE32_IN( &xyold );
>-   x = (GLshort)( xy >> 16 );
>-   y = (GLshort)( xy & 0xffff );
>+   x = xold = v0->v.x;
>+   y = yold = v0->v.y;
>    
>    ooa = 4.0 / (sz * sz);
>    
>    vb = (CARD32 *)mach64AllocDmaLow( mmesa, vbsiz * sizeof(CARD32) );
>    vbchk = vb + vbsiz;
> 
>-   LE32_OUT( pxy, (( x - sz ) << 16) | (( y - sz ) & 0xffff) );
>+   v0->v.x = x - sz;
>+   v0->v.y = y - sz;
>    COPY_VERTEX( vb, vertsize, v0, 1 );
>-   LE32_OUT( pxy, (( x + sz ) << 16) | (( y - sz ) & 0xffff) );
>+
>+   v0->v.x = x + sz;
>+   v0->v.y = y - sz;
>    COPY_VERTEX( vb, vertsize, v0, 2 );
>-   LE32_OUT( pxy, (( x - sz ) << 16) | (( y + sz ) & 0xffff) );
>+
>+   v0->v.x = x - sz;
>+   v0->v.y = y + sz;
>    COPY_VERTEX_OOA( vb, vertsize, v0, 3 );
>    LE32_OUT( vb++, *(CARD32 *)&ooa );
> 
>    ooa = -ooa;
>    
>-   LE32_OUT( pxy, (( x + sz ) << 16) | (( y + sz ) & 0xffff) );
>+   v0->v.x = x + sz;
>+   v0->v.y = y + sz;
>    COPY_VERTEX_OOA( vb, vertsize, v0, 1 );
>    LE32_OUT( vb++, *(CARD32 *)&ooa );
> 
>-   *pxy = xyold;
>+   v0->v.x = xold;
>+   v0->v.y = yold;
> #else /* !MACH64_NATIVE_VTXFMT */
>    GLuint vertsize = mmesa->vertex_size; 
>    GLint coloridx;
>@@ -1254,57 +1258,18 @@
> 
> #if MACH64_NATIVE_VTXFMT
> 
>-/* #define DEPTH_SCALE 65536.0 */
>-#define DEPTH_SCALE 1
>+#define DEPTH_SCALE 1.0
> #define UNFILLED_TRI unfilled_tri
> #define UNFILLED_QUAD unfilled_quad
>-#define VERT_X(_v) ((GLfloat)(GLshort)(LE32_IN( &(_v)->ui[xyoffset] ) & 0xffff) / 4.0)
>-#define VERT_Y(_v) ((GLfloat)(GLshort)(LE32_IN( &(_v)->ui[xyoffset] ) >> 16) / 4.0)
>-#define VERT_Z(_v) ((GLfloat) LE32_IN( &(_v)->ui[zoffset] ))
>+#define VERT_X(_v) ((GLfloat)_v->v.x / 4.0)
>+#define VERT_Y(_v) ((GLfloat)_v->v.y / 4.0)
>+#define VERT_Z(_v) ((GLfloat)_v->v.z)
> #define INSANE_VERTICES
>-#define VERT_SET_Z(_v,val) LE32_OUT( &(_v)->ui[zoffset], (GLuint)(val) )
>-#define VERT_Z_ADD(_v,val) LE32_OUT( &(_v)->ui[zoffset], LE32_IN( &(_v)->ui[zoffset] ) + (GLuint)(val) )
>+#define VERT_SET_Z(_v,val) do { _v->v.z  = (GLuint)(val); } while (0)
>+#define VERT_Z_ADD(_v,val) do { _v->v.z += (GLuint)(val); } while (0)
> #define AREA_IS_CCW( a ) ((a) < 0)
> #define GET_VERTEX(e) (mmesa->verts + ((e) * mmesa->vertex_size * sizeof(int)))
> 
>-#define MACH64_COLOR( dst, src )                \
>-do {						\
>-   UNCLAMPED_FLOAT_TO_UBYTE(dst[0], src[2]);	\
>-   UNCLAMPED_FLOAT_TO_UBYTE(dst[1], src[1]);				\
>-   UNCLAMPED_FLOAT_TO_UBYTE(dst[2], src[0]);				\
>-   UNCLAMPED_FLOAT_TO_UBYTE(dst[3], src[3]);				\
>-} while (0)
>-
>-#define MACH64_SPEC( dst, src )			\
>-do {						\
>-   UNCLAMPED_FLOAT_TO_UBYTE(dst[0], src[2]);	\
>-   UNCLAMPED_FLOAT_TO_UBYTE(dst[1], src[1]);	\
>-   UNCLAMPED_FLOAT_TO_UBYTE(dst[2], src[0]);	\
>-} while (0)
>-
>-#define VERT_SET_RGBA( v, c )    MACH64_COLOR( v->ub4[coloroffset], c )
>-#define VERT_COPY_RGBA( v0, v1 ) v0->ui[coloroffset] = v1->ui[coloroffset]
>-#define VERT_SAVE_RGBA( idx )    color[idx] = v[idx]->ui[coloroffset]
>-#define VERT_RESTORE_RGBA( idx ) v[idx]->ui[coloroffset] = color[idx]
>-
>-#define VERT_SET_SPEC( v, c )    if (havespec) MACH64_SPEC( v->ub4[specoffset], c )
>-#define VERT_COPY_SPEC( v0, v1 ) if (havespec) COPY_3V( v0->ub4[specoffset], v1->ub4[specoffset] )
>-#define VERT_SAVE_SPEC( idx )    if (havespec) spec[idx] = v[idx]->ui[specoffset]
>-#define VERT_RESTORE_SPEC( idx ) if (havespec) v[idx]->ui[specoffset] = spec[idx]
>-
>-#define LOCAL_VARS(n)						\
>-   mach64ContextPtr mmesa = MACH64_CONTEXT(ctx);		\
>-   GLuint color[n], spec[n];					\
>-   GLuint vertex_size = mmesa->vertex_size;			\
>-   const GLuint xyoffset = 9;					\
>-   const GLuint coloroffset = 8;				\
>-   const GLuint zoffset = 7;					\
>-   const GLuint specoffset = 6;					\
>-   GLboolean havespec = vertex_size >= 4 ? 1 : 0;		\
>-   (void) color; (void) spec; (void) vertex_size; 		\
>-   (void) xyoffset; (void) coloroffset; (void) zoffset;		\
>-   (void) specoffset; (void) havespec;
>-
> #else
> 
> #define DEPTH_SCALE 1.0
>@@ -1316,6 +1281,8 @@
> #define AREA_IS_CCW( a ) (a > 0)
> #define GET_VERTEX(e) (mmesa->verts + ((e) * mmesa->vertex_size * sizeof(int)))
> 
>+#endif
>+
> #define MACH64_COLOR( dst, src )                \
> do {						\
>    UNCLAMPED_FLOAT_TO_UBYTE(dst[0], src[2]);	\
>@@ -1336,17 +1303,32 @@
> #define VERT_SAVE_RGBA( idx )    color[idx] = v[idx]->ui[coloroffset]
> #define VERT_RESTORE_RGBA( idx ) v[idx]->ui[coloroffset] = color[idx]
> 
>-#define VERT_SET_SPEC( v, c )    if (havespec) MACH64_SPEC( v->ub4[5], c )
>-#define VERT_COPY_SPEC( v0, v1 ) if (havespec) COPY_3V(v0->ub4[5], v1->ub4[5])
>-#define VERT_SAVE_SPEC( idx )    if (havespec) spec[idx] = v[idx]->ui[5]
>-#define VERT_RESTORE_SPEC( idx ) if (havespec) v[idx]->ui[5] = spec[idx]
>+#define VERT_SET_SPEC( v, c )    if (havespec) MACH64_SPEC( v->ub4[specoffset], c )
>+#define VERT_COPY_SPEC( v0, v1 ) if (havespec) COPY_3V(v0->ub4[specoffset], v1->ub4[specoffset])
>+#define VERT_SAVE_SPEC( idx )    if (havespec) spec[idx] = v[idx]->ui[specoffset]
>+#define VERT_RESTORE_SPEC( idx ) if (havespec) v[idx]->ui[specoffset] = spec[idx]
>+
>+#if MACH64_NATIVE_VTXFMT
>+
>+#define LOCAL_VARS(n)						\
>+   mach64ContextPtr mmesa = MACH64_CONTEXT(ctx);		\
>+   GLuint color[n], spec[n];					\
>+   const GLuint coloroffset = 14;				\
>+   const GLuint specoffset = 12;				\
>+   GLboolean havespec = mmesa->vertex_size >= 4 ? 1 : 0;	\
>+   (void) color; (void) spec;					\
>+   (void) coloroffset; (void) specoffset; (void) havespec;
>+
>+#else
> 
> #define LOCAL_VARS(n)						\
>    mach64ContextPtr mmesa = MACH64_CONTEXT(ctx);		\
>    GLuint color[n], spec[n];					\
>    GLuint coloroffset = (mmesa->vertex_size == 4 ? 3 : 4);	\
>+   const GLuint specoffset = 5;					\
>    GLboolean havespec = (mmesa->vertex_size == 4 ? 0 : 1);	\
>-   (void) color; (void) spec; (void) coloroffset; (void) havespec;
>+   (void) color; (void) spec;					\
>+   (void) coloroffset; (void) specoffset; (void) havespec;
> 
> #endif
> 
>@@ -1576,7 +1558,16 @@
>    tnl->Driver.Render.Line( ctx, ii, jj );
> }
> 
>-#if MACH64_NATIVE_VTXFMT
>+#if (0) && MACH64_NATIVE_VTXFMT
>+/*
>+ * Optimized version of mach64FastRenderClippedPoly(), it submits (n) vertices
>+ * instead of (3 * n). Does not seem to improve performance over the simple one.
>+ *
>+ * FIXME:
>+ * Further optimization is possible by consolidating the mach64AllocDmaLow()
>+ * calls. However, doing so results in clipping whole polygons if they only
>+ * have one vertex outside of the cliprect.
>+ */
> static void mach64FastRenderClippedPoly( GLcontext *ctx, const GLuint *elts,
> 					 GLuint n )
> {
>@@ -1584,30 +1575,26 @@
>    const GLuint vertsize = mmesa->vertex_size;
>    GLint a;
>    GLfloat ooa;
>-   GLuint xy;
>-   const GLuint xyoffset = 9;
>    GLint xx[3], yy[3]; /* 2 fractional bits for hardware */
>-   unsigned vbsiz = (vertsize + (vertsize > 7 ? 2 : 1)) * n + (n-2);
>+   unsigned vbsiz = (vertsize + (vertsize > 7 ? 2 : 1)) * 3 + 1;
>    CARD32 *vb, *vbchk;
>    GLubyte *mach64verts = (GLubyte *)mmesa->verts;
>    mach64VertexPtr v0, v1, v2;
>    int i;
>    
>+   /* draw triangle with vertices elts[0], elts[1], elts[2] */
>    v0 = (mach64VertexPtr)VERT(elts[1]);
>    v1 = (mach64VertexPtr)VERT(elts[2]);
>    v2 = (mach64VertexPtr)VERT(elts[0]);
>       
>-   xy = LE32_IN( &v0->ui[xyoffset] );
>-   xx[0] = (GLshort)( xy >> 16 );
>-   yy[0] = (GLshort)( xy & 0xffff );
>-   
>-   xy = LE32_IN( &v1->ui[xyoffset] );
>-   xx[1] = (GLshort)( xy >> 16 );
>-   yy[1] = (GLshort)( xy & 0xffff );
>-   
>-   xy = LE32_IN( &v2->ui[xyoffset] );
>-   xx[2] = (GLshort)( xy >> 16 );
>-   yy[2] = (GLshort)( xy & 0xffff );
>+   xx[0] = v0->v.x;
>+   yy[0] = v0->v.y;
>+   
>+   xx[1] = v1->v.x;
>+   yy[1] = v1->v.y;
>+   
>+   xx[2] = v2->v.x;
>+   yy[2] = v2->v.y;
> 	   
>    a = (xx[0] - xx[2]) * (yy[1] - yy[2]) -
>        (yy[0] - yy[2]) * (xx[1] - xx[2]);
>@@ -1631,42 +1618,38 @@
>    COPY_VERTEX_OOA( vb, vertsize, v2, 3 );
>    LE32_OUT( vb++, *(CARD32 *)&ooa );
> 
>-   i = 3;
>-   while (1) {
>-      if (i >= n)
>-	 break;
>-      v0 = (mach64VertexPtr)VERT(elts[i]);
>-      i++;
>+   assert( vb == vbchk );
> 
>-      xy = LE32_IN( &v0->ui[xyoffset] );
>-      xx[0] = (GLshort)( xy >> 16 );
>-      yy[0] = (GLshort)( xy & 0xffff );
>-	      
>-      a = (xx[0] - xx[2]) * (yy[1] - yy[2]) -
>-	  (yy[0] - yy[2]) * (xx[1] - xx[2]);
>-      ooa = 16.0 / a;
>+   /* draw triangles with vertices elts[0], elts[i-1], elts[i] */
>+   for (i = 3 ; i < n ; i++) {
>+      unsigned vbsiz = (vertsize + (vertsize > 7 ? 2 : 1)) * 1 + 1;
>+      mach64VertexPtr vk;
>+      int k;
>    
>-      COPY_VERTEX_OOA( vb, vertsize, v0, 1 );
>-      LE32_OUT( vb++, *(CARD32 *)&ooa );
>+      /* odd elements use v0, even elements use v1 */
>+      k = 1 - (i & 0x1);
>+   
>+      vk = VERT(elts[i]);
>+   
>+      xx[k] = vk->v.x;
>+      yy[k] = vk->v.y;
>       
>-      if (i >= n)
>-	 break;
>-      v1 = (mach64VertexPtr)VERT(elts[i]);
>-      i++;
>-
>-      xy = LE32_IN( &v1->ui[xyoffset] );
>-      xx[1] = (GLshort)( xy >> 16 );
>-      yy[1] = (GLshort)( xy & 0xffff );
>-	      
>       a = (xx[0] - xx[2]) * (yy[1] - yy[2]) -
> 	  (yy[0] - yy[2]) * (xx[1] - xx[2]);
>       ooa = 16.0 / a;
>+      
>+      vb = (CARD32 *)mach64AllocDmaLow( mmesa, vbsiz * sizeof(CARD32) );
>+      vbchk = vb + vbsiz;
>    
>-      COPY_VERTEX_OOA( vb, vertsize, v1, 2 );
>+      if (k == 0)
>+	 COPY_VERTEX_OOA( vb, vertsize, vk, 1 );
>+      else 
>+	 COPY_VERTEX_OOA( vb, vertsize, vk, 2 );
>       LE32_OUT( vb++, *(CARD32 *)&ooa );
>-   }
>+   
>+      assert( vb == vbchk );
> 
>-   assert( vb == vbchk );
>+   }
> }
> #else
> static void mach64FastRenderClippedPoly( GLcontext *ctx, const GLuint *elts,
>diff -au -x '*swp' -x '*o' -x 'depend*' mach64-vtx/mach64_vb.c mach64-vtx-1/mach64_vb.c
>--- mach64-vtx/mach64_vb.c	2006-03-26 22:53:19.000000000 +0300
>+++ mach64-vtx-1/mach64_vb.c	2006-04-01 17:33:34.000000000 +0300
>@@ -82,8 +82,9 @@
> #define DO_TEX3 0
> #define DO_PTEX (IND & MACH64_PTEX_BIT)
> 
>-#define VERTEX mach64Vertex
> #define LOCALVARS mach64ContextPtr mmesa = MACH64_CONTEXT(ctx);
>+#define VERTEX mach64Vertex
>+#define VERTEX_COLOR mach64_color_t
> #define GET_VIEWPORT_MAT() mmesa->hw_viewport
> #define GET_TEXSOURCE(n)  mmesa->tmu_source[n]
> #define GET_VERTEX_FORMAT() mmesa->vertex_format
>@@ -123,9 +124,6 @@
> 
> #define PTEX_FALLBACK() FALLBACK(MACH64_CONTEXT(ctx), MACH64_FALLBACK_TEXTURE, 1)
> 
>-#define IMPORT_FLOAT_COLORS mach64_import_float_colors
>-#define IMPORT_FLOAT_SPEC_COLORS mach64_import_float_spec_colors
>-
> #define INTERP_VERTEX setup_tab[mmesa->SetupIndex].interp
> #define COPY_PV_VERTEX setup_tab[mmesa->SetupIndex].copy_pv
> 
>@@ -133,167 +131,14 @@
>  *         Generate  pv-copying and translation functions              *
>  ***********************************************************************/
> 
>-#if MACH64_NATIVE_VTXFMT
>-
>-#define TAG(x) mach64_##x
>-#include "mach64_native_vb.c"
>-
>-#else
>-
> #define TAG(x) mach64_##x
> #include "tnl_dd/t_dd_vb.c"
> 
>-#endif
>-
> /***********************************************************************
>  *             Generate vertex emit and interp functions               *
>  ***********************************************************************/
> 
> 
>-#if MACH64_NATIVE_VTXFMT
>-
>-#define IND (MACH64_XYZW_BIT|MACH64_RGBA_BIT)
>-#define TAG(x) x##_wg
>-#include "mach64_native_vbtmp.h"
>-
>-#define IND (MACH64_XYZW_BIT|MACH64_RGBA_BIT|MACH64_SPEC_BIT)
>-#define TAG(x) x##_wgs
>-#include "mach64_native_vbtmp.h"
>-
>-#define IND (MACH64_XYZW_BIT|MACH64_RGBA_BIT|MACH64_TEX0_BIT)
>-#define TAG(x) x##_wgt0
>-#include "mach64_native_vbtmp.h"
>-
>-#define IND (MACH64_XYZW_BIT|MACH64_RGBA_BIT|MACH64_TEX0_BIT|MACH64_TEX1_BIT)
>-#define TAG(x) x##_wgt0t1
>-#include "mach64_native_vbtmp.h"
>-
>-#define IND (MACH64_XYZW_BIT|MACH64_RGBA_BIT|MACH64_TEX0_BIT|MACH64_PTEX_BIT)
>-#define TAG(x) x##_wgpt0
>-#include "mach64_native_vbtmp.h"
>-
>-#define IND (MACH64_XYZW_BIT|MACH64_RGBA_BIT|MACH64_SPEC_BIT|MACH64_TEX0_BIT)
>-#define TAG(x) x##_wgst0
>-#include "mach64_native_vbtmp.h"
>-
>-#define IND (MACH64_XYZW_BIT|MACH64_RGBA_BIT|MACH64_SPEC_BIT|MACH64_TEX0_BIT|\
>-             MACH64_TEX1_BIT)
>-#define TAG(x) x##_wgst0t1
>-#include "mach64_native_vbtmp.h"
>-
>-#define IND (MACH64_XYZW_BIT|MACH64_RGBA_BIT|MACH64_SPEC_BIT|MACH64_TEX0_BIT|\
>-             MACH64_PTEX_BIT)
>-#define TAG(x) x##_wgspt0
>-#include "mach64_native_vbtmp.h"
>-
>-#define IND (MACH64_XYZW_BIT|MACH64_RGBA_BIT|MACH64_FOG_BIT)
>-#define TAG(x) x##_wgf
>-#include "mach64_native_vbtmp.h"
>-
>-#define IND (MACH64_XYZW_BIT|MACH64_RGBA_BIT|MACH64_FOG_BIT|MACH64_SPEC_BIT)
>-#define TAG(x) x##_wgfs
>-#include "mach64_native_vbtmp.h"
>-
>-#define IND (MACH64_XYZW_BIT|MACH64_RGBA_BIT|MACH64_FOG_BIT|MACH64_TEX0_BIT)
>-#define TAG(x) x##_wgft0
>-#include "mach64_native_vbtmp.h"
>-
>-#define IND (MACH64_XYZW_BIT|MACH64_RGBA_BIT|MACH64_FOG_BIT|MACH64_TEX0_BIT|\
>-             MACH64_TEX1_BIT)
>-#define TAG(x) x##_wgft0t1
>-#include "mach64_native_vbtmp.h"
>-
>-#define IND (MACH64_XYZW_BIT|MACH64_RGBA_BIT|MACH64_FOG_BIT|MACH64_TEX0_BIT|\
>-             MACH64_PTEX_BIT)
>-#define TAG(x) x##_wgfpt0
>-#include "mach64_native_vbtmp.h"
>-
>-#define IND (MACH64_XYZW_BIT|MACH64_RGBA_BIT|MACH64_FOG_BIT|MACH64_SPEC_BIT|\
>-             MACH64_TEX0_BIT)
>-#define TAG(x) x##_wgfst0
>-#include "mach64_native_vbtmp.h"
>-
>-#define IND (MACH64_XYZW_BIT|MACH64_RGBA_BIT|MACH64_FOG_BIT|MACH64_SPEC_BIT|\
>-             MACH64_TEX0_BIT|MACH64_TEX1_BIT)
>-#define TAG(x) x##_wgfst0t1
>-#include "mach64_native_vbtmp.h"
>-
>-#define IND (MACH64_XYZW_BIT|MACH64_RGBA_BIT|MACH64_FOG_BIT|MACH64_SPEC_BIT|\
>-             MACH64_TEX0_BIT|MACH64_PTEX_BIT)
>-#define TAG(x) x##_wgfspt0
>-#include "mach64_native_vbtmp.h"
>-
>-#define IND (MACH64_TEX0_BIT)
>-#define TAG(x) x##_t0
>-#include "mach64_native_vbtmp.h"
>-
>-#define IND (MACH64_TEX0_BIT|MACH64_TEX1_BIT)
>-#define TAG(x) x##_t0t1
>-#include "mach64_native_vbtmp.h"
>-
>-#define IND (MACH64_FOG_BIT)
>-#define TAG(x) x##_f
>-#include "mach64_native_vbtmp.h"
>-
>-#define IND (MACH64_FOG_BIT|MACH64_TEX0_BIT)
>-#define TAG(x) x##_ft0
>-#include "mach64_native_vbtmp.h"
>-
>-#define IND (MACH64_FOG_BIT|MACH64_TEX0_BIT|MACH64_TEX1_BIT)
>-#define TAG(x) x##_ft0t1
>-#include "mach64_native_vbtmp.h"
>-
>-#define IND (MACH64_RGBA_BIT)
>-#define TAG(x) x##_g
>-#include "mach64_native_vbtmp.h"
>-
>-#define IND (MACH64_RGBA_BIT|MACH64_SPEC_BIT)
>-#define TAG(x) x##_gs
>-#include "mach64_native_vbtmp.h"
>-
>-#define IND (MACH64_RGBA_BIT|MACH64_TEX0_BIT)
>-#define TAG(x) x##_gt0
>-#include "mach64_native_vbtmp.h"
>-
>-#define IND (MACH64_RGBA_BIT|MACH64_TEX0_BIT|MACH64_TEX1_BIT)
>-#define TAG(x) x##_gt0t1
>-#include "mach64_native_vbtmp.h"
>-
>-#define IND (MACH64_RGBA_BIT|MACH64_SPEC_BIT|MACH64_TEX0_BIT)
>-#define TAG(x) x##_gst0
>-#include "mach64_native_vbtmp.h"
>-
>-#define IND (MACH64_RGBA_BIT|MACH64_SPEC_BIT|MACH64_TEX0_BIT|MACH64_TEX1_BIT)
>-#define TAG(x) x##_gst0t1
>-#include "mach64_native_vbtmp.h"
>-
>-#define IND (MACH64_RGBA_BIT|MACH64_FOG_BIT)
>-#define TAG(x) x##_gf
>-#include "mach64_native_vbtmp.h"
>-
>-#define IND (MACH64_RGBA_BIT|MACH64_FOG_BIT|MACH64_SPEC_BIT)
>-#define TAG(x) x##_gfs
>-#include "mach64_native_vbtmp.h"
>-
>-#define IND (MACH64_RGBA_BIT|MACH64_FOG_BIT|MACH64_TEX0_BIT)
>-#define TAG(x) x##_gft0
>-#include "mach64_native_vbtmp.h"
>-
>-#define IND (MACH64_RGBA_BIT|MACH64_FOG_BIT|MACH64_TEX0_BIT|MACH64_TEX1_BIT)
>-#define TAG(x) x##_gft0t1
>-#include "mach64_native_vbtmp.h"
>-
>-#define IND (MACH64_RGBA_BIT|MACH64_FOG_BIT|MACH64_SPEC_BIT|MACH64_TEX0_BIT)
>-#define TAG(x) x##_gfst0
>-#include "mach64_native_vbtmp.h"
>-
>-#define IND (MACH64_RGBA_BIT|MACH64_FOG_BIT|MACH64_SPEC_BIT|MACH64_TEX0_BIT|\
>-             MACH64_TEX1_BIT)
>-#define TAG(x) x##_gfst0t1
>-#include "mach64_native_vbtmp.h"
>-
>-#else
>-
> #define IND (MACH64_XYZW_BIT|MACH64_RGBA_BIT)
> #define TAG(x) x##_wg
> #include "mach64_vbtmp.h"
>@@ -434,8 +279,6 @@
> #define TAG(x) x##_gfst0t1
> #include "mach64_vbtmp.h"
> 
>-#endif
>-
> static void init_setup_tab( void )
> {
>    init_wg();
>diff -au -x '*swp' -x '*o' -x 'depend*' mach64-vtx/mach64_vb.h mach64-vtx-1/mach64_vb.h
>--- mach64-vtx/mach64_vb.h	2006-03-27 19:30:45.000000000 +0300
>+++ mach64-vtx-1/mach64_vb.h	2006-04-01 16:43:14.000000000 +0300
>@@ -36,9 +36,6 @@
> #include "swrast/swrast.h"
> #include "mach64_context.h"
> 
>-/* premultiply texture coordinates by homogenous coordinate */
>-#define MACH64_PREMULT_TEXCOORDS
>-
> #define _MACH64_NEW_VERTEX_STATE (_DD_NEW_SEPARATE_SPECULAR |          \
>                                _DD_NEW_TRI_LIGHT_TWOSIDE |             \
>                                _DD_NEW_TRI_UNFILLED |                  \
>diff -au -x '*swp' -x '*o' -x 'depend*' mach64-vtx/mach64_vbtmp.h mach64-vtx-1/mach64_vbtmp.h
>--- mach64-vtx/mach64_vbtmp.h	2004-06-09 08:04:46.000000000 +0300
>+++ mach64-vtx-1/mach64_vbtmp.h	2006-04-01 17:33:30.000000000 +0300
>@@ -1,9 +1,9 @@
>-/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */
>+
> /*
>  * Mesa 3-D graphics library
>- * Version:  3.5
>+ * Version:  5.0.1
>  *
>- * Copyright (C) 1999-2001  Brian Paul   All Rights Reserved.
>+ * Copyright (C) 1999-2003  Brian Paul   All Rights Reserved.
>  *
>  * Permission is hereby granted, free of charge, to any person obtaining a
>  * copy of this software and associated documentation files (the "Software"),
>@@ -23,11 +23,7 @@
>  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
>  *
>  * Authors:
>- *    Keith Whitwell <keithw@valinux.com>
>- *
>- * Modified for mach64 by:
>- *    Leif Delgass <ldelgass@retinalburn.net>
>- *    José Fonseca <j_r_fonseca@yahoo.co.uk>
>+ *    Keith Whitwell <keith@tungstengraphics.com>
>  */
> 
> 
>@@ -64,7 +60,10 @@
>  *    unsigned char ub4[4][16];
>  * }
>  *
>-
>+ 
>+ * VERTEX:   hw vertex type as above
>+ * VERTEX_COLOR: hw color struct type in VERTEX
>+ *
>  * DO_XYZW:  Emit xyz and maybe w coordinates.
>  * DO_RGBA:  Emit color.
>  * DO_SPEC:  Emit specular color.
>@@ -99,10 +98,16 @@
> #define VIEWPORT_Y(dst,y) dst = y
> #define VIEWPORT_Z(dst,z) dst = z
> #else
>+#if MACH64_NATIVE_VTXFMT
>+#define VIEWPORT_X(dst,x) dst =  ((GLshort)((s[0]  * x + s[12]) * 4.0))
>+#define VIEWPORT_Y(dst,y) dst =  ((GLshort)((s[5]  * y + s[13]) * 4.0))
>+#define VIEWPORT_Z(dst,z) dst = (((GLuint)  (s[10] * z + s[14])) << 15)
>+#else
> #define VIEWPORT_X(dst,x) dst = s[0]  * x + s[12]
> #define VIEWPORT_Y(dst,y) dst = s[5]  * y + s[13]
> #define VIEWPORT_Z(dst,z) dst = s[10] * z + s[14]
> #endif
>+#endif
> 
> #if (HAVE_HW_DIVIDE && !HAVE_PTEX_VERTICES)
> #error "can't cope with this combination" 
>@@ -124,15 +129,13 @@
> 		       GLuint stride )
> {
>    LOCALVARS
>-   struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
>+      struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
>    GLfloat (*tc0)[4], (*tc1)[4], (*fog)[4];
>    GLfloat (*tc2)[4], (*tc3)[4];
>-   GLfloat (*spec)[4];
>-   GLfloat (*col)[4];
>-   GLuint col_stride;
>-   GLuint tc0_stride, tc1_stride, spec_stride, fog_stride;
>+   GLfloat (*col)[4], (*spec)[4];
>+   GLuint tc0_stride, tc1_stride, col_stride, spec_stride, fog_stride;
>    GLuint tc2_stride, tc3_stride;
>-   GLuint tc0_size, tc1_size;
>+   GLuint tc0_size, tc1_size, col_size;
>    GLuint tc2_size, tc3_size;
>    GLfloat (*coord)[4];
>    GLuint coord_stride;
>@@ -187,23 +190,27 @@
>    }
> 
>    if (DO_RGBA) {
>-      col = VB->ColorPtr[0]->data;
>       col_stride = VB->ColorPtr[0]->stride;
>+      col = VB->ColorPtr[0]->data;
>+      col_size = VB->ColorPtr[0]->size;
>    }
> 
>    if (DO_SPEC) {
>-      spec = VB->SecondaryColorPtr[0]->data;
>-      spec_stride = VB->SecondaryColorPtr[0]->stride;
>-   } else {
>-      spec = (GLfloat (*)[4])ctx->Current.Attrib[VERT_ATTRIB_COLOR1];
>-      spec_stride = 0;
>+      if (VB->SecondaryColorPtr[0]) {
>+	 spec_stride = VB->SecondaryColorPtr[0]->stride;
>+	 spec = VB->SecondaryColorPtr[0]->data;
>+      } else {
>+	 spec = (GLfloat (*)[4])ctx->Current.Attrib[VERT_ATTRIB_COLOR1];
>+	 spec_stride = 0;
>+      }
>    }
> 
>    if (DO_FOG) {
>       if (VB->FogCoordPtr) {
> 	 fog = VB->FogCoordPtr->data;
> 	 fog_stride = VB->FogCoordPtr->stride;
>-      } else {
>+      }
>+      else {
> 	 static GLfloat tmp[4] = {0, 0, 0, 0};
> 	 fog = &tmp;
> 	 fog_stride = 0;
>@@ -213,73 +220,64 @@
>    /* May have nonstandard strides:
>     */
>    if (start) {
>-      coord =  (GLfloat (*)[4])((GLubyte *)coord + start * coord_stride);
>+      STRIDE_4F(coord, start * coord_stride);
>       if (DO_TEX0)
>-	 tc0 =  (GLfloat (*)[4])((GLubyte *)tc0 + start * tc0_stride);
>+	 STRIDE_4F(tc0, start * tc0_stride);
>       if (DO_TEX1) 
>-	 tc1 =  (GLfloat (*)[4])((GLubyte *)tc1 + start * tc1_stride);
>+	 STRIDE_4F(tc1, start * tc1_stride);
>       if (DO_TEX2) 
>-	 tc2 =  (GLfloat (*)[4])((GLubyte *)tc2 + start * tc2_stride);
>+	 STRIDE_4F(tc2, start * tc2_stride);
>       if (DO_TEX3) 
>-	 tc3 =  (GLfloat (*)[4])((GLubyte *)tc3 + start * tc3_stride);
>+	 STRIDE_4F(tc3, start * tc3_stride);
>       if (DO_RGBA) 
> 	 STRIDE_4F(col, start * col_stride);
>       if (DO_SPEC)
> 	 STRIDE_4F(spec, start * spec_stride);
>       if (DO_FOG)
> 	 STRIDE_4F(fog, start * fog_stride);
>-      //	 fog =  (GLfloat (*)[4])((GLubyte *)fog + start * fog_stride);
>-      /*  STRIDE_F(fog, start * fog_stride); */
>    }
>-   
>+
>    for (i=start; i < end; i++, v = (VERTEX *)((GLubyte *)v + stride)) {
>+      if (DO_TEX0 || DO_TEX1) {
>+	 if (mask[i] == 0) {
>+	    v->v.w = coord[0][3];
>+	 }
>+	 else {
>+	    v->v.w = 1.0;
>+	 }
>+      }
>       if (DO_XYZW) {
> 	 if (HAVE_HW_VIEWPORT || mask[i] == 0) {
>-	    /* unclipped */
> 	    VIEWPORT_X(v->v.x, coord[0][0]);
> 	    VIEWPORT_Y(v->v.y, coord[0][1]);
> 	    VIEWPORT_Z(v->v.z, coord[0][2]);
>-	    v->v.w = coord[0][3];
>-	 } else {
>-	    /* clipped */
>-	    v->v.w = 1.0;
>-	 }
>-	 if (MACH64_DEBUG & DEBUG_VERBOSE_PRIMS) {
>-	    fprintf(stderr, "%s: vert (importable) %d: %.2f %.2f %.2f %f\n", 
>-		    __FUNCTION__, i, v->v.x, v->v.y, v->v.z, v->v.w);
> 	 }
>-	 coord =  (GLfloat (*)[4])((GLubyte *)coord +  coord_stride);
>+	 STRIDE_4F(coord, coord_stride);
>       }
>       if (DO_RGBA) {
>-	 if (HAVE_RGBA_COLOR) {
>-	    *(GLuint *)&v->v.color = *(GLuint *)&col[0];
>-	    STRIDE_4F(col, col_stride);
>+	 UNCLAMPED_FLOAT_TO_UBYTE(v->v.color.red, col[0][0]);
>+	 UNCLAMPED_FLOAT_TO_UBYTE(v->v.color.green, col[0][1]);
>+	 UNCLAMPED_FLOAT_TO_UBYTE(v->v.color.blue, col[0][2]);
>+	 if (col_size == 4) {
>+	    UNCLAMPED_FLOAT_TO_UBYTE(v->v.color.alpha, col[0][3]);
> 	 } else {
>-	    v->v.color.blue  = col[0][2];
>-	    v->v.color.green = col[0][1];
>-	    v->v.color.red   = col[0][0];
>-	    v->v.color.alpha = col[0][3];
>-	    STRIDE_4F(col, col_stride);
>+	    v->v.color.alpha = CHAN_MAX;
> 	 }
>+	 STRIDE_4F(col, col_stride);
>       }
>       if (DO_SPEC) {
>-	 v->v.specular.red = spec[0][0];
>-	 v->v.specular.green = spec[0][1];
>-	 v->v.specular.blue = spec[0][2];
>+	 UNCLAMPED_FLOAT_TO_UBYTE(v->v.specular.red, spec[0][0]);
>+	 UNCLAMPED_FLOAT_TO_UBYTE(v->v.specular.green, spec[0][1]);
>+	 UNCLAMPED_FLOAT_TO_UBYTE(v->v.specular.blue, spec[0][2]);
> 	 STRIDE_4F(spec, spec_stride);
>       }
>       if (DO_FOG) {
>-	 v->v.specular.alpha = fog[0][0] * 255.0;
>-	 /*  STRIDE_F(fog, fog_stride); */
>-	 fog =  (GLfloat (*)[4])((GLubyte *)fog + fog_stride);
>+	 UNCLAMPED_FLOAT_TO_UBYTE(v->v.specular.alpha, fog[0][0]);
>+	 STRIDE_4F(fog, fog_stride);
>       }
>       if (DO_TEX0) {
> 	 v->v.u0 = tc0[0][0];
> 	 v->v.v0 = tc0[0][1];
>-	 if (MACH64_DEBUG & DEBUG_VERBOSE_PRIMS) {
>-	    fprintf(stderr, "%s: vert (importable) %d: u0: %.2f, v0: %.2f, w: %f\n", 
>-		    __FUNCTION__, i, v->v.u0, v->v.v0, v->v.w);
>-	 }
> #ifdef MACH64_PREMULT_TEXCOORDS
> 	 v->v.u0 *= v->v.w;
> 	 v->v.v0 *= v->v.w;
>@@ -302,7 +300,7 @@
> #endif
> 	    } 
> 	 } 
>-	 tc0 =  (GLfloat (*)[4])((GLubyte *)tc0 +  tc0_stride);
>+	 STRIDE_4F(tc0, tc0_stride);
>       }
>       if (DO_TEX1) {
> 	 if (DO_PTEX) {
>@@ -321,7 +319,7 @@
> 	 v->v.u1 *= v->v.w;
> 	 v->v.v1 *= v->v.w;
> #endif
>-	 tc1 =  (GLfloat (*)[4])((GLubyte *)tc1 +  tc1_stride);
>+	 STRIDE_4F(tc1, tc1_stride);
>       } 
>       else if (DO_PTEX) {
> 	 *(GLuint *)&v->pv.q1 = 0;	/* avoid culling on radeon */
>@@ -339,7 +337,7 @@
> 	    v->v.u2 = tc2[0][0];
> 	    v->v.v2 = tc2[0][1];
> 	 }
>-	 tc2 =  (GLfloat (*)[4])((GLubyte *)tc2 +  tc2_stride);
>+	 STRIDE_4F(tc2, tc2_stride);
>       } 
>       if (DO_TEX3) {
> 	 if (DO_PTEX) {
>@@ -349,18 +347,17 @@
> 	       v->pv.q3 = tc3[0][3];
> 	    else
> 	       v->pv.q3 = 1.0;
>-	    } 
>+	 } 
> 	 else {
> 	    v->v.u3 = tc3[0][0];
> 	    v->v.v3 = tc3[0][1];
> 	 }
>-	 tc3 =  (GLfloat (*)[4])((GLubyte *)tc3 +  tc3_stride);
>+	 STRIDE_4F(tc3, tc3_stride);
>       } 
>    }
>-}
> 
>+}
> #else
>-#if DO_XYZW
> 
> #if HAVE_HW_DIVIDE
> #error "cannot use tiny vertices with hw perspective divide"
>@@ -370,12 +367,12 @@
> 		       void *dest, GLuint stride )
> {
>    LOCALVARS
>-   struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
>+      struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
>    GLfloat (*col)[4];
>-   GLuint col_stride;
>+   GLuint col_stride, col_size;
>    GLfloat (*coord)[4] = VB->NdcPtr->data;
>    GLuint coord_stride = VB->NdcPtr->stride;
>-   GLfloat *v = (GLfloat *)dest;
>+   VERTEX *v = (VERTEX *)dest;
>    const GLubyte *mask = VB->ClipMask;
>    const GLfloat *s = GET_VIEWPORT_MAT();
>    int i;
>@@ -386,80 +383,36 @@
> 
>    col = VB->ColorPtr[0]->data;
>    col_stride = VB->ColorPtr[0]->stride;
>+   col_size = VB->ColorPtr[0]->size;
> 
>-   /* Pack what's left into a 4-dword vertex.  Color is in a different
>-    * place, and there is no 'w' coordinate.
>-    */
>    if (start) {
>-      coord =  (GLfloat (*)[4])((GLubyte *)coord + start * coord_stride);
>+      STRIDE_4F(coord, start * coord_stride);
>       STRIDE_4F(col, start * col_stride);
>    }
>-   
>-   for (i=start; i < end; i++, v+=4) {
>-      if (HAVE_HW_VIEWPORT || mask[i] == 0) {
>-	 VIEWPORT_X(v[0], coord[0][0]);
>-	 VIEWPORT_Y(v[1], coord[0][1]);
>-	 VIEWPORT_Z(v[2], coord[0][2]);
>+
>+   for (i=start; i < end; i++, v = (VERTEX *)((GLubyte *)v + stride)) {
>+      if (DO_XYZW) {
>+	 if (HAVE_HW_VIEWPORT || mask[i] == 0) {
>+	    VIEWPORT_X(v->tv.x, coord[0][0]);
>+	    VIEWPORT_Y(v->tv.y, coord[0][1]);
>+	    VIEWPORT_Z(v->tv.z, coord[0][2]);
>+	 }
>+	 STRIDE_4F( coord, coord_stride );
>       }
>-      coord =  (GLfloat (*)[4])((GLubyte *)coord +  coord_stride);
>       if (DO_RGBA) {
>-	 if (HAVE_RGBA_COLOR) {
>-	    *(GLuint *)&v[3] = *(GLuint *)col;
>-	 }
>-	 else {
>-	    GLubyte *b = (GLubyte *)&v[3];
>-	    UNCLAMPED_FLOAT_TO_UBYTE(b[0], col[0][2]);
>-	    UNCLAMPED_FLOAT_TO_UBYTE(b[1], col[0][1]);
>-	    UNCLAMPED_FLOAT_TO_UBYTE(b[2], col[0][0]);
>-	    UNCLAMPED_FLOAT_TO_UBYTE(b[3], col[0][3]);
>+	 UNCLAMPED_FLOAT_TO_UBYTE(v->tv.color.red, col[0][0]);
>+	 UNCLAMPED_FLOAT_TO_UBYTE(v->tv.color.green, col[0][1]);
>+	 UNCLAMPED_FLOAT_TO_UBYTE(v->tv.color.blue, col[0][2]);
>+	 if (col_size == 4) {
>+	    UNCLAMPED_FLOAT_TO_UBYTE(v->tv.color.alpha, col[0][3]);
>+	 } else {
>+	    v->tv.color.alpha = CHAN_MAX;
> 	 }
> 	 STRIDE_4F( col, col_stride );
>       }
>-      if (MACH64_DEBUG & DEBUG_VERBOSE_PRIMS) {
>-	 fprintf(stderr, "vert (importable) %d: %.2f %.2f %.2f %x\n",
>-		 i, v[0], v[1], v[2], *(int *)&v[3]);
>-      }
>    }
> }
>-#else
>-static void TAG(emit)( GLcontext *ctx, GLuint start, GLuint end,
>-		       void *dest, GLuint stride )
>-{
>-   LOCALVARS
>-   struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
>-   GLfloat (*col)[4];
>-   GLuint col_stride;
>-   GLfloat *v = (GLfloat *)dest;
>-   int i;
> 
>-   col = VB->ColorPtr[0]->data;
>-   col_stride = VB->ColorPtr[0]->stride;
>-
>-   if (start)
>-      STRIDE_4F(col, col_stride * start);
>-
>-   /* Need to figure out where color is:
>-    */
>-   if (GET_VERTEX_FORMAT() == TINY_VERTEX_FORMAT)
>-      v += 3;
>-   else
>-      v += 4;
>-
>-   for (i=start; i < end; i++, STRIDE_F(v, stride)) {
>-      if (HAVE_RGBA_COLOR) {
>-	 *(GLuint *)v = *(GLuint *)col[0];
>-      }
>-      else {
>-	 GLubyte *b = (GLubyte *)v;
>-	 UNCLAMPED_FLOAT_TO_UBYTE(b[0], col[0][2]);
>-	 UNCLAMPED_FLOAT_TO_UBYTE(b[1], col[0][1]);
>-	 UNCLAMPED_FLOAT_TO_UBYTE(b[2], col[0][0]);
>-	 UNCLAMPED_FLOAT_TO_UBYTE(b[3], col[0][3]);
>-      }
>-      STRIDE_4F( col, col_stride );
>-   }
>-}
>-#endif /* emit */
> #endif /* emit */
> 
> #if (DO_XYZW) && (DO_RGBA)
>@@ -561,39 +514,32 @@
>       w = dstclip[3];
>    }
>    else {
>-      w = (dstclip[3] == 0.0F) ? 1.0 : (1.0 / dstclip[3]);
>+      w = 1.0 / dstclip[3];
>       VIEWPORT_X( dst->v.x, dstclip[0] * w );
>       VIEWPORT_Y( dst->v.y, dstclip[1] * w );
>       VIEWPORT_Z( dst->v.z, dstclip[2] * w );
>    }
> 
>-   if (MACH64_DEBUG & DEBUG_VERBOSE_PRIMS) {
>-      fprintf( stderr, "%s: dst vert: %.2f %.2f %.2f %f\n",
>-	       __FUNCTION__,
>-	       dst->v.x,
>-	       dst->v.y,
>-	       dst->v.z,
>-	       w );
>-   }
>-
>    if ((HAVE_HW_DIVIDE && CHECK_HW_DIVIDE) || 
>        DO_FOG || DO_SPEC || DO_TEX0 || DO_TEX1 ||
>        DO_TEX2 || DO_TEX3 || !HAVE_TINY_VERTICES) {
> 
>-      dst->v.w = w;
>+      if (DO_TEX0 || DO_TEX1) {
>+	 dst->v.w = w;
>+      }
> 
>-      INTERP_UB( t, dst->ub4[4][0], out->ub4[4][0], in->ub4[4][0] );
>-      INTERP_UB( t, dst->ub4[4][1], out->ub4[4][1], in->ub4[4][1] );
>-      INTERP_UB( t, dst->ub4[4][2], out->ub4[4][2], in->ub4[4][2] );
>-      INTERP_UB( t, dst->ub4[4][3], out->ub4[4][3], in->ub4[4][3] );
>+      INTERP_UB( t, dst->v.color.red,   out->v.color.red,   in->v.color.red );
>+      INTERP_UB( t, dst->v.color.green, out->v.color.green, in->v.color.green );
>+      INTERP_UB( t, dst->v.color.blue,  out->v.color.blue,  in->v.color.blue );
>+      INTERP_UB( t, dst->v.color.alpha, out->v.color.alpha, in->v.color.alpha );
> 
>       if (DO_SPEC) {
>-	 INTERP_UB( t, dst->ub4[5][0], out->ub4[5][0], in->ub4[5][0] );
>-	 INTERP_UB( t, dst->ub4[5][1], out->ub4[5][1], in->ub4[5][1] );
>-	 INTERP_UB( t, dst->ub4[5][2], out->ub4[5][2], in->ub4[5][2] );
>+	 INTERP_UB( t, dst->v.specular.red,   out->v.specular.red,   in->v.specular.red );
>+	 INTERP_UB( t, dst->v.specular.green, out->v.specular.green, in->v.specular.green );
>+	 INTERP_UB( t, dst->v.specular.blue,  out->v.specular.blue,  in->v.specular.blue );
>       }
>       if (DO_FOG) {
>-	 INTERP_UB( t, dst->ub4[5][3], out->ub4[5][3], in->ub4[5][3] );
>+	 INTERP_UB( t, dst->v.specular.alpha, out->v.specular.alpha, in->v.specular.alpha );
>       }
>       if (DO_TEX0) {
> 	 if (DO_PTEX) {
>@@ -604,8 +550,8 @@
> 	    } else {
> 	       GLfloat wout = VB->NdcPtr->data[eout][3];
> 	       GLfloat win = VB->NdcPtr->data[ein][3];
>-	       GLfloat qout = out->pv.w / wout;
>-	       GLfloat qin = in->pv.w / win;
>+	       GLfloat qout = out->v.w / wout;
>+	       GLfloat qin = in->v.w / win;
> 	       GLfloat qdst, rqdst;
> 
> 	       ASSERT( !HAVE_HW_DIVIDE );
>@@ -658,7 +604,7 @@
> 	 }
>       }
>       else if (DO_PTEX) {
>-	 dst->pv.q0 = 0.0;	/* must be a valid float on radeon */
>+	 dst->pv.q1 = 0.0;	/* must be a valid float on radeon */
>       }
>       if (DO_TEX2) {
> 	 if (DO_PTEX) {
>@@ -683,16 +629,47 @@
>    } else {
>       /* 4-dword vertex.  Color is in v[3] and there is no oow coordinate.
>        */
>-      INTERP_UB( t, dst->ub4[3][0], out->ub4[3][0], in->ub4[3][0] );
>-      INTERP_UB( t, dst->ub4[3][1], out->ub4[3][1], in->ub4[3][1] );
>-      INTERP_UB( t, dst->ub4[3][2], out->ub4[3][2], in->ub4[3][2] );
>-      INTERP_UB( t, dst->ub4[3][3], out->ub4[3][3], in->ub4[3][3] );
>+      INTERP_UB( t, dst->tv.color.red,   out->tv.color.red,   in->tv.color.red );
>+      INTERP_UB( t, dst->tv.color.green, out->tv.color.green, in->tv.color.green );
>+      INTERP_UB( t, dst->tv.color.blue,  out->tv.color.blue,  in->tv.color.blue );
>+      INTERP_UB( t, dst->tv.color.alpha, out->tv.color.alpha, in->tv.color.alpha );
>    }
> }
> 
> #endif /* rgba && xyzw */
> 
> 
>+#if MACH64_NATIVE_VTXFMT
>+static void TAG(copy_pv)( GLcontext *ctx, GLuint edst, GLuint esrc )
>+{
>+#if DO_SPEC || DO_FOG || DO_RGBA
>+   LOCALVARS   
>+   GLubyte *verts = GET_VERTEX_STORE();
>+   GLuint size = GET_VERTEX_SIZE();
>+   GLuint *dst = (GLuint *)(verts + (edst * size));
>+   GLuint *src = (GLuint *)(verts + (esrc * size));
>+#endif
>+
>+#if DO_SPEC || DO_FOG
>+   dst[12] = src[12];			/* VERTEX_?_SPEC_ARGB */
>+#endif
>+
>+#if DO_RGBA
>+   dst[14] = src[14];			/* VERTEX_?_ARGB */
>+#endif
>+}
>+
>+#define TEX1_VERTEX_SIZE	10
>+#define TEX0_VERTEX_SIZE	 7
>+#define NOTEX_VERTEX_SIZE	 4
>+#define TINY_VERTEX_SIZE	 3
>+#else
>+#define TEX1_VERTEX_SIZE	10
>+#define TEX0_VERTEX_SIZE	 8
>+#define NOTEX_VERTEX_SIZE	 6
>+#define TINY_VERTEX_SIZE	 4
>+#endif
>+
> static void TAG(init)( void )
> {
>    setup_tab[IND].emit = TAG(emit);
>@@ -702,6 +679,9 @@
>    setup_tab[IND].interp = TAG(interp);
> #endif
> 
>+#if MACH64_NATIVE_VTXFMT
>+   setup_tab[IND].copy_pv = TAG(copy_pv);
>+#else
>    if (DO_SPEC)
>       setup_tab[IND].copy_pv = copy_pv_rgba4_spec5;
>    else if (HAVE_HW_DIVIDE || DO_SPEC || DO_FOG || DO_TEX0 || DO_TEX1 ||
>@@ -709,6 +689,7 @@
>       setup_tab[IND].copy_pv = copy_pv_rgba4;
>    else
>       setup_tab[IND].copy_pv = copy_pv_rgba3;
>+#endif
> 
>    if (DO_TEX3) {
>       if (DO_PTEX) {
>@@ -740,7 +721,7 @@
>       }
>       else {
> 	 setup_tab[IND].vertex_format = TEX1_VERTEX_FORMAT;
>-	 setup_tab[IND].vertex_size = 10;
>+	 setup_tab[IND].vertex_size = TEX1_VERTEX_SIZE;
>       }
>    }
>    else if (DO_TEX0) {
>@@ -749,20 +730,19 @@
> 	 setup_tab[IND].vertex_size = 12;
>       } else {
> 	 setup_tab[IND].vertex_format = TEX0_VERTEX_FORMAT;
>-	 setup_tab[IND].vertex_size = 8;
>+	 setup_tab[IND].vertex_size = TEX0_VERTEX_SIZE;
>       }
>    }
>    else if (!HAVE_HW_DIVIDE && !DO_SPEC && !DO_FOG && HAVE_TINY_VERTICES) {
>       setup_tab[IND].vertex_format = TINY_VERTEX_FORMAT;
>-      setup_tab[IND].vertex_size = 4;
>+      setup_tab[IND].vertex_size = TINY_VERTEX_SIZE;
>    } else if (HAVE_NOTEX_VERTICES) {
>       setup_tab[IND].vertex_format = NOTEX_VERTEX_FORMAT;
>-      setup_tab[IND].vertex_size = 6;
>+      setup_tab[IND].vertex_size = NOTEX_VERTEX_SIZE;
>    } else {
>       setup_tab[IND].vertex_format = TEX0_VERTEX_FORMAT;
>-      setup_tab[IND].vertex_size = 8;
>+      setup_tab[IND].vertex_size = TEX0_VERTEX_SIZE;
>    }
>-
> }
> 
> 
>Common subdirectories: mach64-vtx/server and mach64-vtx-1/server

Actions: View | Splinter Review

Attachments on bug 6412: 5096 | 5097 | 5098 | 5100 | 5106 | 5109 | 5126 | 5155 | 5164 | 5198 | 5206 | 5209 | 5282 | 5723

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.