Move the transform and lighting code to two new directories

	math:  Provides basic matrix and vector functionality that
               might be useful to multiple software t&l
	       implementations, and is used by core mesa to
	       manage the Model, Project, etc matrices.

	tnl:   The real transform & lighting code from core mesa,
	       including everything from glVertex3f through vertex
	       buffer handling, transformation, clipping, lighting
	       and handoff to a driver for rasterization.

The interfaces of these can be further tightened up, but the basic
splitting up of state and code move is done.

15 files changed, 5845 insertions, 0 deletions

diff --git a/src/mesa/math/m_clip_tmp.h b/src/mesa/math/m_clip_tmp.h
new file mode 100644
index 00000000000..321d3a9e616
--- /dev/null
+++ b/src/mesa/math/m_clip_tmp.h
@@ -0,0 +1,175 @@
+/* $Id: m_clip_tmp.h,v 1.1 2000/11/16 21:05:41 keithw Exp $ */
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  3.1
+ * 
+ * Copyright (C) 1999  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"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ * 
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * New (3.1) transformation code written by Keith Whitwell.
+ */
+
+
+/* KW: a clever asm implementation would nestle integer versions
+ * of the outcode calculation underneath the division.  Gcc won't
+ * do this, strangely enough, so I only do the divide in
+ * the case where the cliptest passes.  This isn't essential,
+ * and an asm implementation needn't replicate that behaviour.
+ */
+static GLvector4f * _XFORMAPI TAG(cliptest_points4)( GLvector4f *clip_vec, 
+					   GLvector4f *proj_vec, 
+					   GLubyte clipMask[],
+					   GLubyte *orMask, 
+					   GLubyte *andMask )
+{
+   const GLuint stride = clip_vec->stride;
+   const GLfloat *from = (GLfloat *)clip_vec->start;
+   const GLuint count = clip_vec->count;
+   GLuint c = 0;
+   GLfloat (*vProj)[4] = (GLfloat (*)[4])proj_vec->start;
+   GLubyte tmpAndMask = *andMask;
+   GLubyte tmpOrMask = *orMask;
+   GLuint i;
+   STRIDE_LOOP {
+      const GLfloat cx = from[0];
+      const GLfloat cy = from[1];
+      const GLfloat cz = from[2];
+      const GLfloat cw = from[3];
+#if defined(macintosh)
+      /* on powerpc cliptest is 17% faster in this way. */
+      GLuint mask;
+      mask = (((cw < cx) << CLIP_RIGHT_SHIFT));
+      mask |= (((cw < -cx) << CLIP_LEFT_SHIFT));
+      mask |= (((cw < cy) << CLIP_TOP_SHIFT));
+      mask |= (((cw < -cy) << CLIP_BOTTOM_SHIFT));
+      mask |= (((cw < cz) << CLIP_FAR_SHIFT));
+      mask |= (((cw < -cz) << CLIP_NEAR_SHIFT));
+#else /* !defined(macintosh)) */
+      GLubyte mask = 0;
+      if (-cx + cw < 0) mask |= CLIP_RIGHT_BIT;
+      if ( cx + cw < 0) mask |= CLIP_LEFT_BIT;
+      if (-cy + cw < 0) mask |= CLIP_TOP_BIT;
+      if ( cy + cw < 0) mask |= CLIP_BOTTOM_BIT;
+      if (-cz + cw < 0) mask |= CLIP_FAR_BIT;
+      if ( cz + cw < 0) mask |= CLIP_NEAR_BIT;
+#endif /* defined(macintosh) */
+
+      clipMask[i] = mask;
+      if (mask) {
+	 c++;
+	 tmpAndMask &= mask;
+	 tmpOrMask |= mask;
+	 vProj[i][0] = 0;	/* no longer required? */
+	 vProj[i][1] = 0;
+	 vProj[i][2] = 0;
+	 vProj[i][3] = 1;
+      } else {
+	 GLfloat oow = 1.0F / cw;	 
+	 vProj[i][3] = oow;
+	 vProj[i][0] = cx * oow;
+	 vProj[i][1] = cy * oow;
+	 vProj[i][2] = cz * oow;      
+      }	 
+   }
+
+   *orMask = tmpOrMask;
+   *andMask = (GLubyte) (c < count ? 0 : tmpAndMask);
+
+   proj_vec->flags |= VEC_SIZE_4;
+   proj_vec->size = 3;
+   proj_vec->count = clip_vec->count;
+   return proj_vec;
+}
+
+static GLvector4f * _XFORMAPI TAG(cliptest_points3)( GLvector4f *clip_vec, 
+					   GLvector4f *proj_vec, 
+					   GLubyte clipMask[],
+					   GLubyte *orMask, 
+					   GLubyte *andMask )
+{
+   const GLuint stride = clip_vec->stride;
+   const GLuint count = clip_vec->count;
+   const GLfloat *from = (GLfloat *)clip_vec->start;
+
+   GLubyte tmpOrMask = *orMask;
+   GLubyte tmpAndMask = *andMask;
+   GLuint i;
+   STRIDE_LOOP {
+      const GLfloat cx = from[0], cy = from[1], cz = from[2];
+      GLubyte mask = 0;
+      if (cx >  1.0)       mask |= CLIP_RIGHT_BIT;
+      else if (cx < -1.0)  mask |= CLIP_LEFT_BIT;
+      if (cy >  1.0)       mask |= CLIP_TOP_BIT;
+      else if (cy < -1.0)  mask |= CLIP_BOTTOM_BIT;
+      if (cz >  1.0)       mask |= CLIP_FAR_BIT;
+      else if (cz < -1.0)  mask |= CLIP_NEAR_BIT;
+      clipMask[i] = mask;
+      tmpOrMask |= mask;
+      tmpAndMask &= mask;
+   }
+
+   gl_vector4f_clean_elem(proj_vec, count, 3);
+      
+   *orMask = tmpOrMask;
+   *andMask = tmpAndMask;
+   return clip_vec;
+}
+
+static GLvector4f * _XFORMAPI TAG(cliptest_points2)( GLvector4f *clip_vec, 
+					   GLvector4f *proj_vec, 
+					   GLubyte clipMask[],
+					   GLubyte *orMask, 
+					   GLubyte *andMask )
+{
+   const GLuint stride = clip_vec->stride;
+   const GLuint count = clip_vec->count;
+   const GLfloat *from = (GLfloat *)clip_vec->start;
+
+   GLubyte tmpOrMask = *orMask;
+   GLubyte tmpAndMask = *andMask;
+   GLuint i;
+   STRIDE_LOOP {
+      const GLfloat cx = from[0], cy = from[1];
+      GLubyte mask = 0;
+      if (cx >  1.0)       mask |= CLIP_RIGHT_BIT;
+      else if (cx < -1.0)  mask |= CLIP_LEFT_BIT;
+      if (cy >  1.0)       mask |= CLIP_TOP_BIT;
+      else if (cy < -1.0)  mask |= CLIP_BOTTOM_BIT;
+      clipMask[i] = mask;
+      tmpOrMask |= mask;
+      tmpAndMask &= mask;
+   }
+
+   gl_vector4f_clean_elem(proj_vec, count, 3);
+
+   *orMask = tmpOrMask;
+   *andMask = tmpAndMask;
+   return clip_vec;
+}
+
+
+static void TAG(init_c_cliptest)( void )
+{
+   gl_clip_tab[4] = TAG(cliptest_points4);
+   gl_clip_tab[3] = TAG(cliptest_points3);
+   gl_clip_tab[2] = TAG(cliptest_points2);
+}
diff --git a/src/mesa/math/m_copy_tmp.h b/src/mesa/math/m_copy_tmp.h
new file mode 100644
index 00000000000..b328537fafa
--- /dev/null
+++ b/src/mesa/math/m_copy_tmp.h
@@ -0,0 +1,126 @@
+/* $Id: m_copy_tmp.h,v 1.1 2000/11/16 21:05:41 keithw Exp $ */
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  3.1
+ *
+ * Copyright (C) 1999  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"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * New (3.1) transformation code written by Keith Whitwell.
+ */
+
+
+#define COPY_FUNC( BITS )						\
+static void TAG2(copy, BITS)(GLvector4f *to, const GLvector4f *f,	\
+			     const GLubyte mask[] )			\
+{									\
+   GLfloat (*t)[4] = (GLfloat (*)[4])to->start;				\
+   GLfloat *from = f->start;						\
+   GLuint stride = f->stride;				        	\
+   GLuint count = f->count;						\
+   GLuint i;								\
+   (void) mask;								\
+									\
+   if (BITS)								\
+      STRIDE_LOOP {							\
+	 CULL_CHECK {							\
+	    if (BITS&1) t[i][0] = from[0];				\
+	    if (BITS&2) t[i][1] = from[1];				\
+	    if (BITS&4) t[i][2] = from[2];				\
+	    if (BITS&8) t[i][3] = from[3];				\
+	 }								\
+      }									\
+}
+
+
+
+/* static void TAG2(clean, BITS)(GLvector4f *to ) */
+/* { */
+/*    GLfloat (*t)[4] = to->data; */
+/*    GLuint i; */
+
+/*    if (BITS) */
+/*       for (i = 0 ; i < VB_SIZE ; i++) { */
+/*          if (BITS&1) t[i][0] = 0; */
+/* 	 if (BITS&2) t[i][1] = 0; */
+/* 	 if (BITS&4) t[i][2] = 0; */
+/* 	 if (BITS&8) t[i][3] = 1; */
+/*       } */
+/*    to->flags &= ~BITS; */
+/* } */
+
+
+/* We got them all here:
+ */
+COPY_FUNC( 0x0 )		/* noop */
+COPY_FUNC( 0x1 )
+COPY_FUNC( 0x2 )
+COPY_FUNC( 0x3 )
+COPY_FUNC( 0x4 )
+COPY_FUNC( 0x5 )
+COPY_FUNC( 0x6 )
+COPY_FUNC( 0x7 )
+COPY_FUNC( 0x8 )
+COPY_FUNC( 0x9 )
+COPY_FUNC( 0xa )
+COPY_FUNC( 0xb )
+COPY_FUNC( 0xc )
+COPY_FUNC( 0xd )
+COPY_FUNC( 0xe )
+COPY_FUNC( 0xf )
+
+static void TAG2(init_copy, 0 ) ( void )
+{
+   gl_copy_tab[IDX][0x0] = TAG2(copy, 0x0);
+   gl_copy_tab[IDX][0x1] = TAG2(copy, 0x1);
+   gl_copy_tab[IDX][0x2] = TAG2(copy, 0x2);
+   gl_copy_tab[IDX][0x3] = TAG2(copy, 0x3);
+   gl_copy_tab[IDX][0x4] = TAG2(copy, 0x4);
+   gl_copy_tab[IDX][0x5] = TAG2(copy, 0x5);
+   gl_copy_tab[IDX][0x6] = TAG2(copy, 0x6);
+   gl_copy_tab[IDX][0x7] = TAG2(copy, 0x7);
+   gl_copy_tab[IDX][0x8] = TAG2(copy, 0x8);
+   gl_copy_tab[IDX][0x9] = TAG2(copy, 0x9);
+   gl_copy_tab[IDX][0xa] = TAG2(copy, 0xa);
+   gl_copy_tab[IDX][0xb] = TAG2(copy, 0xb);
+   gl_copy_tab[IDX][0xc] = TAG2(copy, 0xc);
+   gl_copy_tab[IDX][0xd] = TAG2(copy, 0xd);
+   gl_copy_tab[IDX][0xe] = TAG2(copy, 0xe);
+   gl_copy_tab[IDX][0xf] = TAG2(copy, 0xf);
+
+/*    gl_clean_tab[IDX][0x0] = TAG2(clean, 0x0); */
+/*    gl_clean_tab[IDX][0x1] = TAG2(clean, 0x1); */
+/*    gl_clean_tab[IDX][0x2] = TAG2(clean, 0x2); */
+/*    gl_clean_tab[IDX][0x3] = TAG2(clean, 0x3); */
+/*    gl_clean_tab[IDX][0x4] = TAG2(clean, 0x4); */
+/*    gl_clean_tab[IDX][0x5] = TAG2(clean, 0x5); */
+/*    gl_clean_tab[IDX][0x6] = TAG2(clean, 0x6); */
+/*    gl_clean_tab[IDX][0x7] = TAG2(clean, 0x7); */
+/*    gl_clean_tab[IDX][0x8] = TAG2(clean, 0x8); */
+/*    gl_clean_tab[IDX][0x9] = TAG2(clean, 0x9); */
+/*    gl_clean_tab[IDX][0xa] = TAG2(clean, 0xa); */
+/*    gl_clean_tab[IDX][0xb] = TAG2(clean, 0xb); */
+/*    gl_clean_tab[IDX][0xc] = TAG2(clean, 0xc); */
+/*    gl_clean_tab[IDX][0xd] = TAG2(clean, 0xd); */
+/*    gl_clean_tab[IDX][0xe] = TAG2(clean, 0xe); */
+/*    gl_clean_tab[IDX][0xf] = TAG2(clean, 0xf); */
+}
diff --git a/src/mesa/math/m_debug_xform.c b/src/mesa/math/m_debug_xform.c
new file mode 100644
index 00000000000..5041fc4ee07
--- /dev/null
+++ b/src/mesa/math/m_debug_xform.c
@@ -0,0 +1,930 @@
+/* $Id: m_debug_xform.c,v 1.1 2000/11/16 21:05:41 keithw Exp $ */
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  3.5
+ *
+ * Copyright (C) 1999-2000  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"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Updated for P6 architecture by Gareth Hughes.
+ */
+
+#include "glheader.h"
+#include "context.h"
+#include "mem.h"
+
+#include "m_debug_xform.h"
+#include "m_matrix.h"
+#include "m_xform.h"
+
+
+#ifdef DEBUG  /* This code only used for debugging */
+
+
+/* Comment this out to deactivate the cycle counter.
+ * NOTE: it works only on CPUs which know the 'rdtsc' command (586 or higher)
+ * (hope, you don't try to debug Mesa on a 386 ;)
+ */
+#if defined(__GNUC__) && defined(__i386__) && defined(USE_X86_ASM)
+#define  RUN_XFORM_BENCHMARK
+#endif
+
+#define TEST_COUNT		128	/* size of the tested vector array   */
+
+#define REQUIRED_PRECISION	10	/* allow 4 bits to miss              */
+#define MAX_PRECISION		24	/* max. precision possible           */
+
+
+#ifdef  RUN_XFORM_BENCHMARK
+/* Overhead of profiling counter in cycles.  Automatically adjusted to
+ * your machine at run time - counter initialization should give very
+ * consistent results.
+ */
+static int need_counter = 1;
+static long counter_overhead = 0;
+
+/* Modify the the number of tests if you like.
+ * We take the minimum of all results, because every error should be
+ * positive (time used by other processes, task switches etc).
+ * It is assumed that all calculations are done in the cache.
+ */
+
+#if 1 /* PPro, PII, PIII version */
+
+/* Profiling on the P6 architecture requires a little more work, due to
+ * the internal out-of-order execution.  We must perform a serializing
+ * 'cpuid' instruction before and after the 'rdtsc' instructions to make
+ * sure no other uops are executed when we sample the timestamp counter.
+ */
+#define  INIT_COUNTER()							\
+   do {									\
+      int cycle_i;							\
+      counter_overhead = LONG_MAX;					\
+      for ( cycle_i = 0 ; cycle_i < 4 ; cycle_i++ ) {			\
+	 long cycle_tmp1 = 0, cycle_tmp2 = 0;				\
+	 __asm__ ( "push %%ebx       \n"				\
+		   "xor %%eax, %%eax \n"				\
+		   "cpuid            \n"				\
+		   "rdtsc            \n"				\
+		   "mov %%eax, %0    \n"				\
+		   "xor %%eax, %%eax \n"				\
+		   "cpuid            \n"				\
+		   "pop %%ebx        \n"				\
+		   "push %%ebx       \n"				\
+		   "xor %%eax, %%eax \n"				\
+		   "cpuid            \n"				\
+		   "rdtsc            \n"				\
+		   "mov %%eax, %1    \n"				\
+		   "xor %%eax, %%eax \n"				\
+		   "cpuid            \n"				\
+		   "pop %%ebx        \n"				\
+		   : "=m" (cycle_tmp1), "=m" (cycle_tmp2)		\
+		   : : "eax", "ecx", "edx" );				\
+	 if ( counter_overhead > (cycle_tmp2 - cycle_tmp1) ) {		\
+	    counter_overhead = cycle_tmp2 - cycle_tmp1;			\
+	 }								\
+      }									\
+   } while (0)
+
+#define  BEGIN_RACE(x)							\
+   x = LONG_MAX;							\
+   for ( cycle_i = 0 ; cycle_i < 10 ; cycle_i++ ) {			\
+      long cycle_tmp1 = 0, cycle_tmp2 = 0;				\
+      __asm__ ( "push %%ebx       \n"					\
+		"xor %%eax, %%eax \n"					\
+		"cpuid            \n"					\
+		"rdtsc            \n"					\
+		"mov %%eax, %0    \n"					\
+		"xor %%eax, %%eax \n"					\
+		"cpuid            \n"					\
+		"pop %%ebx        \n"					\
+		: "=m" (cycle_tmp1)					\
+		: : "eax", "ecx", "edx" );
+
+#define END_RACE(x)							\
+      __asm__ ( "push %%ebx       \n"					\
+		"xor %%eax, %%eax \n"					\
+		"cpuid            \n"					\
+		"rdtsc            \n"					\
+		"mov %%eax, %0    \n"					\
+		"xor %%eax, %%eax \n"					\
+		"cpuid            \n"					\
+		"pop %%ebx        \n"					\
+		: "=m" (cycle_tmp2)					\
+		: : "eax", "ecx", "edx" );				\
+      if ( x > (cycle_tmp2 - cycle_tmp1) ) {				\
+	 x = cycle_tmp2 - cycle_tmp1;					\
+      }									\
+   }									\
+   x -= counter_overhead;
+
+#else /* PPlain, PMMX version */
+
+/* To ensure accurate results, we stall the pipelines with the
+ * non-pairable 'cdq' instruction.  This ensures all the code being
+ * profiled is complete when the 'rdtsc' instruction executes.
+ */
+#define  INIT_COUNTER(x)						\
+   do {									\
+      int cycle_i;							\
+      x = LONG_MAX;							\
+      for ( cycle_i = 0 ; cycle_i < 32 ; cycle_i++ ) {			\
+	 long cycle_tmp1, cycle_tmp2, dummy;				\
+	 __asm__ ( "mov %%eax, %0" : "=a" (cycle_tmp1) );		\
+	 __asm__ ( "mov %%eax, %0" : "=a" (cycle_tmp2) );		\
+	 __asm__ ( "cdq" );						\
+	 __asm__ ( "cdq" );						\
+	 __asm__ ( "rdtsc" : "=a" (cycle_tmp1), "=d" (dummy) );		\
+	 __asm__ ( "cdq" );						\
+	 __asm__ ( "cdq" );						\
+	 __asm__ ( "rdtsc" : "=a" (cycle_tmp2), "=d" (dummy) );		\
+	 if ( x > (cycle_tmp2 - cycle_tmp1) )				\
+	    x = cycle_tmp2 - cycle_tmp1;				\
+      }									\
+   } while (0)
+
+#define  BEGIN_RACE(x)							\
+   x = LONG_MAX;							\
+   for ( cycle_i = 0 ; cycle_i < 16 ; cycle_i++ ) {			\
+      long cycle_tmp1, cycle_tmp2, dummy;				\
+      __asm__ ( "mov %%eax, %0" : "=a" (cycle_tmp1) );			\
+      __asm__ ( "mov %%eax, %0" : "=a" (cycle_tmp2) );			\
+      __asm__ ( "cdq" );						\
+      __asm__ ( "cdq" );						\
+      __asm__ ( "rdtsc" : "=a" (cycle_tmp1), "=d" (dummy) );
+
+
+#define END_RACE(x)							\
+      __asm__ ( "cdq" );						\
+      __asm__ ( "cdq" );						\
+      __asm__ ( "rdtsc" : "=a" (cycle_tmp2), "=d" (dummy) );		\
+      if ( x > (cycle_tmp2 - cycle_tmp1) )				\
+	 x = cycle_tmp2 - cycle_tmp1;					\
+   }									\
+   x -= counter_overhead;
+
+#endif
+
+#else
+
+#define BEGIN_RACE(x)
+#define END_RACE(x)
+
+#endif
+
+
+static char *mesa_profile = NULL;
+
+
+enum { NIL=0, ONE=1, NEG=-1, VAR=2 };
+
+static int m_general[16] = {
+   VAR, VAR, VAR, VAR,
+   VAR, VAR, VAR, VAR,
+   VAR, VAR, VAR, VAR,
+   VAR, VAR, VAR, VAR
+};
+static int m_identity[16] = {
+   ONE, NIL, NIL, NIL,
+   NIL, ONE, NIL, NIL,
+   NIL, NIL, ONE, NIL,
+   NIL, NIL, NIL, ONE
+};
+static int  m_2d[16]  = {
+   VAR, VAR, NIL, VAR,
+   VAR, VAR, NIL, VAR,
+   NIL, NIL, ONE, NIL,
+   NIL, NIL, NIL, ONE
+};
+static int m_2d_no_rot[16] = {
+   VAR, NIL, NIL, VAR,
+   NIL, VAR, NIL, VAR,
+   NIL, NIL, ONE, NIL,
+   NIL, NIL, NIL, ONE
+};
+static int m_3d[16] = {
+   VAR, VAR, VAR, VAR,
+   VAR, VAR, VAR, VAR,
+   VAR, VAR, VAR, VAR,
+   NIL, NIL, NIL, ONE
+};
+static int m_3d_no_rot[16] = {
+   VAR, NIL, NIL, VAR,
+   NIL, VAR, NIL, VAR,
+   NIL, NIL, VAR, VAR,
+   NIL, NIL, NIL, ONE
+};
+static int m_perspective[16] = {
+   VAR, NIL, VAR, NIL,
+   NIL, VAR, VAR, NIL,
+   NIL, NIL, VAR, VAR,
+   NIL, NIL, NEG, NIL
+};
+static int *templates[7] = {
+   m_general,
+   m_identity,
+   m_3d_no_rot,
+   m_perspective,
+   m_2d,
+   m_2d_no_rot,
+   m_3d
+};
+static int mtypes[7] = {
+   MATRIX_GENERAL,
+   MATRIX_IDENTITY,
+   MATRIX_3D_NO_ROT,
+   MATRIX_PERSPECTIVE,
+   MATRIX_2D,
+   MATRIX_2D_NO_ROT,
+   MATRIX_3D
+};
+static char *mstrings[7] = {
+   "MATRIX_GENERAL",
+   "MATRIX_IDENTITY",
+   "MATRIX_3D_NO_ROT",
+   "MATRIX_PERSPECTIVE",
+   "MATRIX_2D",
+   "MATRIX_2D_NO_ROT",
+   "MATRIX_3D"
+};
+
+
+
+static int m_norm_identity[16] = {
+   ONE, NIL, NIL, NIL,
+   NIL, ONE, NIL, NIL,
+   NIL, NIL, ONE, NIL,
+   NIL, NIL, NIL, NIL
+};
+static int m_norm_general[16] = {
+   VAR, VAR, VAR, NIL,
+   VAR, VAR, VAR, NIL,
+   VAR, VAR, VAR, NIL,
+   NIL, NIL, NIL, NIL
+};
+static int m_norm_no_rot[16] = {
+   VAR, NIL, NIL, NIL,
+   NIL, VAR, NIL, NIL,
+   NIL, NIL, VAR, NIL,
+   NIL, NIL, NIL, NIL
+};
+static int *norm_templates[8] = {
+   m_norm_no_rot,
+   m_norm_no_rot,
+   m_norm_no_rot,
+   m_norm_general,
+   m_norm_general,
+   m_norm_general,
+   m_norm_identity,
+   m_norm_identity
+};
+static int norm_types[8] = {
+   NORM_TRANSFORM_NO_ROT,
+   NORM_TRANSFORM_NO_ROT | NORM_RESCALE,
+   NORM_TRANSFORM_NO_ROT | NORM_NORMALIZE,
+   NORM_TRANSFORM,
+   NORM_TRANSFORM | NORM_RESCALE,
+   NORM_TRANSFORM | NORM_NORMALIZE,
+   NORM_RESCALE,
+   NORM_NORMALIZE
+};
+static int norm_scale_types[8] = {               /*  rescale factor          */
+   NIL,                                          /*  NIL disables rescaling  */
+   VAR,
+   NIL,
+   NIL,
+   VAR,
+   NIL,
+   VAR,
+   NIL
+};
+static int norm_normalize_types[8] = {           /*  normalizing ?? (no = 0) */
+   0,
+   0,
+   1,
+   0,
+   0,
+   1,
+   0,
+   1
+};
+static char *norm_strings[8] = {
+   "NORM_TRANSFORM_NO_ROT",
+   "NORM_TRANSFORM_NO_ROT | NORM_RESCALE",
+   "NORM_TRANSFORM_NO_ROT | NORM_NORMALIZE",
+   "NORM_TRANSFORM",
+   "NORM_TRANSFORM | NORM_RESCALE",
+   "NORM_TRANSFORM | NORM_NORMALIZE",
+   "NORM_RESCALE",
+   "NORM_NORMALIZE"
+};
+
+
+
+/* ================================================================
+ * Helper functions
+ */
+
+static GLfloat rnd( void )
+{
+   GLfloat f = (GLfloat)rand() / (GLfloat)RAND_MAX;
+   GLfloat gran = (GLfloat)(1 << 13);
+
+   f = (GLfloat)(GLint)(f * gran) / gran;
+
+   return f * 2.0 - 1.0;
+}
+
+static int significand_match( GLfloat a, GLfloat b )
+{
+   GLfloat d = a - b;
+   int a_ex, b_ex, d_ex;
+
+   if ( d == 0.0F ) {
+      return MAX_PRECISION;   /* Exact match */
+   }
+
+   if ( a == 0.0F || b == 0.0F ) {
+      /* It would probably be better to check if the
+       * non-zero number is denormalized and return
+       * the index of the highest set bit here.
+       */
+      return 0;
+   }
+
+   frexp( a, &a_ex );
+   frexp( b, &b_ex );
+   frexp( d, &d_ex );
+
+   if ( a_ex < b_ex )
+      return a_ex - d_ex;
+   else
+      return b_ex - d_ex;
+}
+
+
+
+/* ================================================================
+ * Reference transformations
+ */
+
+static void ref_transform( GLvector4f *dst,
+                           const GLmatrix *mat,
+                           const GLvector4f *src,
+                           const GLubyte *clipmask,
+                           const GLubyte flag )
+{
+   int i;
+   GLfloat *s = (GLfloat *)src->start;
+   GLfloat (*d)[4] = (GLfloat (*)[4])dst->start;
+   const GLfloat *m = mat->m;
+
+   (void) clipmask;
+   (void) flag;
+
+   for ( i = 0 ; i < src->count ; i++ ) {
+      GLfloat x = s[0], y = s[1], z = s[2], w = s[3];
+      d[i][0] = m[0]*x + m[4]*y + m[ 8]*z + m[12]*w;
+      d[i][1] = m[1]*x + m[5]*y + m[ 9]*z + m[13]*w;
+      d[i][2] = m[2]*x + m[6]*y + m[10]*z + m[14]*w;
+      d[i][3] = m[3]*x + m[7]*y + m[11]*z + m[15]*w;
+      s = (GLfloat *)((char *)s + src->stride);
+   }
+}
+
+static void ref_norm_transform_rescale( const GLmatrix *mat,
+					GLfloat scale,
+					const GLvector3f *in,
+					const GLfloat *lengths,
+					const GLubyte mask[],
+					GLvector3f *dest )
+{
+   int i;
+   const GLfloat *s = in->start;
+   const GLfloat *m = mat->inv;
+   GLfloat (*out)[3] = (GLfloat (*)[3])dest->start;
+
+   (void) mask;
+   (void) lengths;
+
+   for ( i = 0 ; i < in->count ; i++ ) {
+      GLfloat x = s[0], y = s[1], z = s[2] ;
+      GLfloat tx = m[0]*x + m[1]*y + m[ 2]*z ;
+      GLfloat ty = m[4]*x + m[5]*y + m[ 6]*z ;
+      GLfloat tz = m[8]*x + m[9]*y + m[10]*z ;
+
+      out[i][0] = tx * scale;
+      out[i][1] = ty * scale;
+      out[i][2] = tz * scale;
+
+      s = (GLfloat *)((char *)s + in->stride);
+   }
+}
+
+static void ref_norm_transform_normalize( const GLmatrix *mat,
+					  GLfloat scale,
+					  const GLvector3f *in,
+					  const GLfloat *lengths,
+					  const GLubyte mask[],
+					  GLvector3f *dest )
+{
+   int i;
+   const GLfloat *s = in->start;
+   const GLfloat *m = mat->inv;
+   GLfloat (*out)[3] = (GLfloat (*)[3])dest->start;
+
+   (void) mask;
+
+   for ( i = 0 ; i < in->count ; i++ ) {
+      GLfloat x = s[0], y = s[1], z = s[2] ;
+      GLfloat tx = m[0]*x + m[1]*y + m[ 2]*z ;
+      GLfloat ty = m[4]*x + m[5]*y + m[ 6]*z ;
+      GLfloat tz = m[8]*x + m[9]*y + m[10]*z ;
+
+      if ( !lengths ) {
+         GLfloat len = tx*tx + ty*ty + tz*tz;
+         if ( len > 1e-20 ) {
+	    /* Hmmm, don't know how we could test the precalculated
+	     * length case...
+	     */
+            scale = 1.0 / sqrt( len );
+            out[i][0] = tx * scale;
+            out[i][1] = ty * scale;
+            out[i][2] = tz * scale;
+         } else {
+            out[i][0] = out[i][1] = out[i][2] = 0;
+         }
+      } else {
+         scale = lengths[i];;
+         out[i][0] = tx * scale;
+         out[i][1] = ty * scale;
+         out[i][2] = tz * scale;
+      }
+
+      s = (GLfloat *)((char *)s + in->stride);
+   }
+}
+
+
+
+/* ================================================================
+ * Vertex transformation tests
+ */
+
+/* Ensure our arrays are correctly aligned.
+ */
+#if defined(__GNUC__)
+#define ALIGN16(x)		x __attribute__ ((aligned (16)))
+#else
+#define ALIGN16(x)		x
+#endif
+static GLfloat ALIGN16(s[TEST_COUNT][5]);
+static GLfloat ALIGN16(d[TEST_COUNT][4]);
+static GLfloat ALIGN16(r[TEST_COUNT][4]);
+
+static int test_transform_function( transform_func func, int psize, int mtype,
+                                    int masked, long *cycles )
+{
+   GLvector4f source[1], dest[1], ref[1];
+   GLmatrix mat[1];
+   GLfloat *m;
+   GLubyte mask[TEST_COUNT];
+   int i, j;
+#ifdef  RUN_XFORM_BENCHMARK
+   int cycle_i;                /* the counter for the benchmarks we run */
+#endif
+
+   (void) cycles;
+
+   if ( psize > 4 ) {
+      gl_problem( NULL, "test_transform_function called with psize > 4\n" );
+      return 0;
+   }
+
+   mat->m = (GLfloat *) ALIGN_MALLOC( 16 * sizeof(GLfloat), 16 );
+   mat->type = mtypes[mtype];
+
+   m = mat->m;
+
+   m[0] = 63.0; m[4] = 43.0; m[ 8] = 29.0; m[12] = 43.0;
+   m[1] = 55.0; m[5] = 17.0; m[ 9] = 31.0; m[13] =  7.0;
+   m[2] = 44.0; m[6] =  9.0; m[10] =  7.0; m[14] =  3.0;
+   m[3] = 11.0; m[7] = 23.0; m[11] = 91.0; m[15] =  9.0;
+
+   for ( i = 0 ; i < 4 ; i++ ) {
+      for ( j = 0 ; j < 4 ; j++ ) {
+         switch ( templates[mtype][i * 4 + j] ) {
+         case NIL:
+            m[j * 4 + i] = 0.0;
+            break;
+         case ONE:
+            m[j * 4 + i] = 1.0;
+            break;
+         case NEG:
+            m[j * 4 + i] = -1.0;
+            break;
+         case VAR:
+            break;
+         default:
+            abort();
+         }
+      }
+   }
+
+   for ( i = 0 ; i < TEST_COUNT ; i++) {
+      mask[i] = i % 2;				/* mask every 2nd element */
+      d[i][0] = s[i][0] = 0.0;
+      d[i][1] = s[i][1] = 0.0;
+      d[i][2] = s[i][2] = 0.0;
+      d[i][3] = s[i][3] = 1.0;
+      for ( j = 0 ; j < psize ; j++ )
+         s[i][j] = rnd();
+   }
+
+   source->data = (GLfloat(*)[4])s;
+   source->start = (GLfloat *)s;
+   source->count = TEST_COUNT;
+   source->stride = sizeof(s[0]);
+   source->size = 4;
+   source->flags = 0;
+
+   dest->data = (GLfloat(*)[4])d;
+   dest->start = (GLfloat *)d;
+   dest->count = TEST_COUNT;
+   dest->stride = sizeof(float[4]);
+   dest->size = 0;
+   dest->flags = 0;
+
+   ref->data = (GLfloat(*)[4])r;
+   ref->start = (GLfloat *)r;
+   ref->count = TEST_COUNT;
+   ref->stride = sizeof(float[4]);
+   ref->size = 0;
+   ref->flags = 0;
+
+   ref_transform( ref, mat, source, NULL, 0 );
+
+   if ( mesa_profile ) {
+      if ( masked ) {
+         BEGIN_RACE( *cycles );
+         func( dest, mat->m, source, mask, 1 );
+         END_RACE( *cycles );
+      } else {
+         BEGIN_RACE( *cycles );
+         func( dest, mat->m, source, NULL, 0 );
+         END_RACE( *cycles );
+     }
+   }
+   else {
+      if ( masked ) {
+         func( dest, mat->m, source, mask, 1 );
+      } else {
+         func( dest, mat->m, source, NULL, 0 );
+      }
+   }
+
+   for ( i = 0 ; i < TEST_COUNT ; i++ ) {
+      if ( masked && (mask[i] & 1) )
+         continue;
+
+      for ( j = 0 ; j < 4 ; j++ ) {
+         if ( significand_match( d[i][j], r[i][j] ) < REQUIRED_PRECISION ) {
+            printf( "-----------------------------\n" );
+            printf( "(i = %i, j = %i)\n", i, j );
+            printf( "%f \t %f \t [diff = %e - %i bit missed]\n",
+		    d[i][0], r[i][0], r[i][0]-d[i][0],
+		    MAX_PRECISION - significand_match( d[i][0], r[i][0] ) );
+            printf( "%f \t %f \t [diff = %e - %i bit missed]\n",
+		    d[i][1], r[i][1], r[i][1]-d[i][1],
+		    MAX_PRECISION - significand_match( d[i][1], r[i][1] ) );
+            printf( "%f \t %f \t [diff = %e - %i bit missed]\n",
+		    d[i][2], r[i][2], r[i][2]-d[i][2],
+		    MAX_PRECISION - significand_match( d[i][2], r[i][2] ) );
+            printf( "%f \t %f \t [diff = %e - %i bit missed]\n",
+		    d[i][3], r[i][3], r[i][3]-d[i][3],
+		    MAX_PRECISION - significand_match( d[i][3], r[i][3] ) );
+            return 0;
+         }
+      }
+   }
+
+   ALIGN_FREE( mat->m );
+   return 1;
+}
+
+void gl_test_all_transform_functions( char *description )
+{
+   int masked, psize, mtype;
+   long benchmark_tab[2][4][7];
+   static int first_time = 1;
+
+   if ( first_time ) {
+      first_time = 0;
+      mesa_profile = getenv( "MESA_PROFILE" );
+   }
+
+#ifdef RUN_XFORM_BENCHMARK
+   if ( mesa_profile ) {
+      if ( need_counter ) {
+	 need_counter = 0;
+	 INIT_COUNTER();
+	 printf( "counter overhead: %ld cycles\n\n", counter_overhead );
+      }
+      printf( "transform results after hooking in %s functions:\n", description );
+   }
+#endif
+
+   for ( masked = 0 ; masked <= 1 ; masked++ ) {
+      int cma = masked ? 1 : 0;
+      char *cmastring = masked ? "CULL_MASK_ACTIVE" : "0";
+
+#ifdef RUN_XFORM_BENCHMARK
+      if ( mesa_profile ) {
+         printf( "\n culling: %s \n", masked ? "CULL_MASK_ACTIVE" : "0" );
+         for ( psize = 1 ; psize <= 4 ; psize++ ) {
+            printf( " p%d\t", psize );
+         }
+         printf( "\n--------------------------------------------------------\n" );
+      }
+#endif
+
+      for ( mtype = 0 ; mtype < 7 ; mtype++ ) {
+         for ( psize = 1 ; psize <= 4 ; psize++ ) {
+            transform_func func = gl_transform_tab[cma][psize][mtypes[mtype]];
+            long *cycles = &(benchmark_tab[cma][psize-1][mtype]);
+
+            if ( test_transform_function( func, psize, mtype,
+					  masked, cycles ) == 0 ) {
+               char buf[100];
+               sprintf( buf, "gl_transform_tab[%s][%d][%s] failed test (%s)",
+                        cmastring, psize, mstrings[mtype], description );
+               gl_problem( NULL, buf );
+            }
+#ifdef RUN_XFORM_BENCHMARK
+            if ( mesa_profile )
+               printf( " %li\t", benchmark_tab[cma][psize-1][mtype] );
+#endif
+         }
+#ifdef RUN_XFORM_BENCHMARK
+         if ( mesa_profile )
+            printf( " | [%s]\n", mstrings[mtype] );
+#endif
+      }
+#ifdef RUN_XFORM_BENCHMARK
+      if ( mesa_profile )
+         printf( "\n" );
+#endif
+   }
+}
+
+
+
+/* ================================================================
+ * Normal transformation tests
+ */
+
+static int test_norm_function( normal_func func, int mtype,
+			       int masked, long *cycles )
+{
+   GLvector3f source[1], dest[1], dest2[1], ref[1], ref2[1];
+   GLmatrix mat[1];
+   GLfloat s[TEST_COUNT][5], d[TEST_COUNT][3], r[TEST_COUNT][3];
+   GLfloat d2[TEST_COUNT][3], r2[TEST_COUNT][3], length[TEST_COUNT];
+   GLfloat scale;
+   GLfloat *m;
+   GLubyte mask[TEST_COUNT];
+   int i, j;
+#ifdef  RUN_XFORM_BENCHMARK
+   int cycle_i;		/* the counter for the benchmarks we run */
+#endif
+
+   (void) cycles;
+
+   mat->m = (GLfloat *) ALIGN_MALLOC( 16 * sizeof(GLfloat), 16 );
+   mat->inv = m = mat->m;
+
+   m[0] = 63.0; m[4] = 43.0; m[ 8] = 29.0; m[12] = 43.0;
+   m[1] = 55.0; m[5] = 17.0; m[ 9] = 31.0; m[13] =  7.0;
+   m[2] = 44.0; m[6] =  9.0; m[10] =  7.0; m[14] =  3.0;
+   m[3] = 11.0; m[7] = 23.0; m[11] = 91.0; m[15] =  9.0;
+
+   scale = 1.0F + rnd () * norm_scale_types[mtype];
+
+   for ( i = 0 ; i < 4 ; i++ ) {
+      for ( j = 0 ; j < 4 ; j++ ) {
+         switch ( norm_templates[mtype][i * 4 + j] ) {
+         case NIL:
+            m[j * 4 + i] = 0.0;
+            break;
+         case ONE:
+            m[j * 4 + i] = 1.0;
+            break;
+         case NEG:
+            m[j * 4 + i] = -1.0;
+            break;
+         case VAR:
+            break;
+         default:
+            abort();
+         }
+      }
+   }
+
+   for ( i = 0 ; i < TEST_COUNT ; i++ ) {
+      mask[i] = i % 2;				/* mask every 2nd element */
+      d[i][0] = s[i][0] = d2[i][0] = 0.0;
+      d[i][1] = s[i][1] = d2[i][1] = 0.0;
+      d[i][2] = s[i][2] = d2[i][2] = 0.0;
+      for ( j = 0 ; j < 3 ; j++ )
+         s[i][j] = rnd();
+      length[i] = 1 / sqrt( s[i][0]*s[i][0] +
+			    s[i][1]*s[i][1] +
+			    s[i][2]*s[i][2] );
+   }
+
+   source->data = (GLfloat(*)[3])s;
+   source->start = (GLfloat *)s;
+   source->count = TEST_COUNT;
+   source->stride = sizeof(s[0]);
+   source->flags = 0;
+
+   dest->data = (GLfloat(*)[3])d;
+   dest->start = (GLfloat *)d;
+   dest->count = TEST_COUNT;
+   dest->stride = sizeof(float[3]);
+   dest->flags = 0;
+
+   dest2->data = (GLfloat(*)[3])d2;
+   dest2->start = (GLfloat *)d2;
+   dest2->count = TEST_COUNT;
+   dest2->stride = sizeof(float[3]);
+   dest2->flags = 0;
+
+   ref->data = (GLfloat(*)[3])r;
+   ref->start = (GLfloat *)r;
+   ref->count = TEST_COUNT;
+   ref->stride = sizeof(float[3]);
+   ref->flags = 0;
+
+   ref2->data = (GLfloat(*)[3])r2;
+   ref2->start = (GLfloat *)r2;
+   ref2->count = TEST_COUNT;
+   ref2->stride = sizeof(float[3]);
+   ref2->flags = 0;
+
+   if ( norm_normalize_types[mtype] == 0 ) {
+      ref_norm_transform_rescale( mat, scale, source, NULL, NULL, ref );
+   } else {
+      ref_norm_transform_normalize( mat, scale, source, NULL, NULL, ref );
+      ref_norm_transform_normalize( mat, scale, source, length, NULL, ref2 );
+   }
+
+   if ( mesa_profile ) {
+      if ( masked ) {
+         BEGIN_RACE( *cycles );
+         func( mat, scale, source, NULL, mask, dest );
+         END_RACE( *cycles );
+         func( mat, scale, source, length, mask, dest2 );
+      } else {
+         BEGIN_RACE( *cycles );
+         func( mat, scale, source, NULL, NULL, dest );
+         END_RACE( *cycles );
+         func( mat, scale, source, length, NULL, dest2 );
+      }
+   } else {
+      if ( masked ) {
+         func( mat, scale, source, NULL, mask, dest );
+         func( mat, scale, source, length, mask, dest2 );
+      } else {
+         func( mat, scale, source, NULL, NULL, dest );
+         func( mat, scale, source, length, NULL, dest2 );
+      }
+   }
+
+   for ( i = 0 ; i < TEST_COUNT ; i++ ) {
+      if ( masked && !(mask[i] & 1) )
+         continue;
+
+      for ( j = 0 ; j < 3 ; j++ ) {
+         if ( significand_match( d[i][j], r[i][j] ) < REQUIRED_PRECISION ) {
+            printf( "-----------------------------\n" );
+            printf( "(i = %i, j = %i)\n", i, j );
+            printf( "%f \t %f \t [ratio = %e - %i bit missed]\n",
+		    d[i][0], r[i][0], r[i][0]/d[i][0],
+		    MAX_PRECISION - significand_match( d[i][0], r[i][0] ) );
+            printf( "%f \t %f \t [ratio = %e - %i bit missed]\n",
+		    d[i][1], r[i][1], r[i][1]/d[i][1],
+		    MAX_PRECISION - significand_match( d[i][1], r[i][1] ) );
+            printf( "%f \t %f \t [ratio = %e - %i bit missed]\n",
+		    d[i][2], r[i][2], r[i][2]/d[i][2],
+		    MAX_PRECISION - significand_match( d[i][2], r[i][2] ) );
+            return 0;
+         }
+
+         if ( norm_normalize_types[mtype] != 0 ) {
+            if ( significand_match( d2[i][j], r2[i][j] ) < REQUIRED_PRECISION ) {
+               printf( "------------------- precalculated length case ------\n" );
+               printf( "(i = %i, j = %i)\n", i, j );
+               printf( "%f \t %f \t [ratio = %e - %i bit missed]\n",
+		       d2[i][0], r2[i][0], r2[i][0]/d2[i][0],
+		       MAX_PRECISION - significand_match( d2[i][0], r2[i][0] ) );
+               printf( "%f \t %f \t [ratio = %e - %i bit missed]\n",
+		       d2[i][1], r2[i][1], r2[i][1]/d2[i][1],
+		       MAX_PRECISION - significand_match( d2[i][1], r2[i][1] ) );
+               printf( "%f \t %f \t [ratio = %e - %i bit missed]\n",
+		       d2[i][2], r2[i][2], r2[i][2]/d2[i][2],
+		       MAX_PRECISION - significand_match( d2[i][2], r2[i][2] ) );
+               return 0;
+            }
+         }
+      }
+   }
+
+   ALIGN_FREE( mat->m );
+   return 1;
+}
+
+void gl_test_all_normal_transform_functions( char *description )
+{
+   int masked;
+   int mtype;
+   long benchmark_tab[0xf][0x4];
+   static int first_time = 1;
+
+   if ( first_time ) {
+      first_time = 0;
+      mesa_profile = getenv( "MESA_PROFILE" );
+   }
+
+#ifdef RUN_XFORM_BENCHMARK
+   if ( mesa_profile ) {
+      if ( need_counter ) {
+	 need_counter = 0;
+	 INIT_COUNTER();
+	 printf( "counter overhead: %ld cycles\n\n", counter_overhead );
+      }
+      printf( "normal transform results after hooking in %s functions:\n",
+	      description );
+   }
+#endif
+
+   for ( masked = 0 ; masked <= 1 ; masked++ ) {
+      int cma = masked ? 1 : 0;
+      char *cmastring = masked ? "CULL_MASK_ACTIVE" : "0";
+
+#ifdef RUN_XFORM_BENCHMARK
+      if ( mesa_profile ) {
+         printf( "\n culling: %s \n", masked ? "CULL_MASK_ACTIVE" : "0" );
+         printf( "\n-------------------------------------------------------\n" );
+      }
+#endif
+
+      for ( mtype = 0 ; mtype < 8 ; mtype++ ) {
+         normal_func func = gl_normal_tab[norm_types[mtype]][cma];
+         long *cycles = &(benchmark_tab[mtype][cma]);
+
+         if ( test_norm_function( func, mtype, masked, cycles ) == 0 ) {
+            char buf[100];
+            sprintf( buf, "gl_normal_tab[%s][%s] failed test (%s)",
+                     cmastring, norm_strings[mtype], description );
+            gl_problem( NULL, buf );
+	 }
+
+#ifdef RUN_XFORM_BENCHMARK
+         if ( mesa_profile ) {
+            printf( " %li\t", benchmark_tab[mtype][cma] );
+            printf( " | [%s]\n", norm_strings[mtype] );
+         }
+      }
+      if ( mesa_profile )
+	 printf( "\n" );
+#else
+      }
+#endif
+   }
+#ifdef RUN_XFORM_BENCHMARK
+   if ( mesa_profile )
+      fflush( stdout );
+#endif
+}
+
+#endif /* DEBUG */
diff --git a/src/mesa/math/m_dotprod_tmp.h b/src/mesa/math/m_dotprod_tmp.h
new file mode 100644
index 00000000000..637e35fd587
--- /dev/null
+++ b/src/mesa/math/m_dotprod_tmp.h
@@ -0,0 +1,128 @@
+/* $Id: m_dotprod_tmp.h,v 1.1 2000/11/16 21:05:41 keithw Exp $ */
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  3.1
+ * 
+ * Copyright (C) 1999  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"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ * 
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * New (3.1) transformation code written by Keith Whitwell.
+ */
+
+
+/* Note - respects the stride of the output vector.
+ */
+static void TAG(dotprod_vec2)( GLvector4f *out_vec, 
+			     GLuint elt,
+			     const GLvector4f *coord_vec, 
+			     const GLfloat plane[4], 
+			     const GLubyte mask[])
+{
+   GLuint stride = coord_vec->stride;
+   GLfloat *coord = coord_vec->start;
+   GLuint count = coord_vec->count;
+
+   GLuint outstride = out_vec->stride;
+   GLfloat *out = out_vec->start + elt;
+   GLuint i;
+   
+   const GLfloat plane0 = plane[0], plane1 = plane[1], plane3 = plane[3];
+      
+   (void) mask;
+
+   for (i=0;i<count;i++,STRIDE_F(coord,stride),STRIDE_F(out,outstride)) {
+      CULL_CHECK {
+	 *out = (coord[0] * plane0 +
+		 coord[1] * plane1 +
+		            plane3);
+      }
+   }   
+   out_vec->count = coord_vec->count;
+}
+
+static void TAG(dotprod_vec3)( GLvector4f *out_vec, 
+			     GLuint elt,
+			     const GLvector4f *coord_vec, 
+			     const GLfloat plane[4], 
+			     const GLubyte mask[])
+{
+   GLuint stride = coord_vec->stride;
+   GLfloat *coord = coord_vec->start;
+   GLuint count = coord_vec->count;
+
+   GLuint outstride = out_vec->stride;
+   GLfloat *out = out_vec->start + elt;
+   GLuint i;
+
+   const GLfloat plane0 = plane[0], plane1 = plane[1], plane2 = plane[2];
+   const GLfloat plane3 = plane[3];
+
+   (void) mask;
+
+   for (i=0;i<count;i++,STRIDE_F(coord,stride),STRIDE_F(out,outstride)) {
+      CULL_CHECK {
+	 *out = (coord[0] * plane0 +
+		 coord[1] * plane1 +
+		 coord[2] * plane2 +
+		            plane3);
+      }
+   }   
+   out_vec->count = coord_vec->count;
+}
+
+static void TAG(dotprod_vec4)( GLvector4f *out_vec, 
+			     GLuint elt,
+			     const GLvector4f *coord_vec, 
+			     const GLfloat plane[4], 
+			     const GLubyte mask[])
+{
+   GLuint stride = coord_vec->stride;
+   GLfloat *coord = coord_vec->start;
+   GLuint count = coord_vec->count;
+
+   GLuint outstride = out_vec->stride;
+   GLfloat *out = out_vec->start + elt;
+   GLuint i;
+
+   const GLfloat plane0 = plane[0], plane1 = plane[1], plane2 = plane[2];
+   const GLfloat plane3 = plane[3];
+
+   (void) mask;
+
+   for (i=0;i<count;i++,STRIDE_F(coord,stride),STRIDE_F(out,outstride)) {
+      CULL_CHECK {
+	 *out = (coord[0] * plane0 +
+		 coord[1] * plane1 +
+		 coord[2] * plane2 +
+		 coord[3] * plane3);
+      }
+   }   
+   out_vec->count = coord_vec->count;
+}
+
+
+static void TAG(init_dotprod)( void )
+{
+   gl_dotprod_tab[IDX&1][2] = TAG(dotprod_vec2);
+   gl_dotprod_tab[IDX&1][3] = TAG(dotprod_vec3);
+   gl_dotprod_tab[IDX&1][4] = TAG(dotprod_vec4);
+}
diff --git a/src/mesa/math/m_matrix.c b/src/mesa/math/m_matrix.c
new file mode 100644
index 00000000000..ae55c946d9c
--- /dev/null
+++ b/src/mesa/math/m_matrix.c
@@ -0,0 +1,1113 @@
+/* $Id: m_matrix.c,v 1.1 2000/11/16 21:05:41 keithw Exp $ */
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  3.5
+ * 
+ * Copyright (C) 1999-2000  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"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ * 
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+
+/*
+ * Matrix operations
+ *
+ * NOTES:
+ * 1. 4x4 transformation matrices are stored in memory in column major order.
+ * 2. Points/vertices are to be thought of as column vectors.
+ * 3. Transformation of a point p by a matrix M is: p' = M * p
+ */
+
+
+#include "glheader.h"
+#include "macros.h"
+#include "mem.h"
+#include "mmath.h"
+
+#include "m_matrix.h"
+
+
+static const char *types[] = {
+   "MATRIX_GENERAL",
+   "MATRIX_IDENTITY",
+   "MATRIX_3D_NO_ROT",
+   "MATRIX_PERSPECTIVE",
+   "MATRIX_2D",
+   "MATRIX_2D_NO_ROT",
+   "MATRIX_3D"
+};
+
+
+static GLfloat Identity[16] = {
+   1.0, 0.0, 0.0, 0.0,
+   0.0, 1.0, 0.0, 0.0,
+   0.0, 0.0, 1.0, 0.0,
+   0.0, 0.0, 0.0, 1.0
+};
+
+
+
+
+/*
+ * This matmul was contributed by Thomas Malik 
+ *
+ * Perform a 4x4 matrix multiplication  (product = a x b).
+ * Input:  a, b - matrices to multiply
+ * Output:  product - product of a and b
+ * WARNING: (product != b) assumed
+ * NOTE:    (product == a) allowed    
+ *
+ * KW: 4*16 = 64 muls
+ */
+#define A(row,col)  a[(col<<2)+row]
+#define B(row,col)  b[(col<<2)+row]
+#define P(row,col)  product[(col<<2)+row]
+
+static void matmul4( GLfloat *product, const GLfloat *a, const GLfloat *b )
+{
+   GLint i;
+   for (i = 0; i < 4; i++) {
+      const GLfloat ai0=A(i,0),  ai1=A(i,1),  ai2=A(i,2),  ai3=A(i,3);
+      P(i,0) = ai0 * B(0,0) + ai1 * B(1,0) + ai2 * B(2,0) + ai3 * B(3,0);
+      P(i,1) = ai0 * B(0,1) + ai1 * B(1,1) + ai2 * B(2,1) + ai3 * B(3,1);
+      P(i,2) = ai0 * B(0,2) + ai1 * B(1,2) + ai2 * B(2,2) + ai3 * B(3,2);
+      P(i,3) = ai0 * B(0,3) + ai1 * B(1,3) + ai2 * B(2,3) + ai3 * B(3,3);
+   }
+}
+
+
+/* Multiply two matrices known to occupy only the top three rows, such
+ * as typical model matrices, and ortho matrices.  
+ */
+static void matmul34( GLfloat *product, const GLfloat *a, const GLfloat *b )
+{
+   GLint i;
+   for (i = 0; i < 3; i++) {
+      const GLfloat ai0=A(i,0),  ai1=A(i,1),  ai2=A(i,2),  ai3=A(i,3);
+      P(i,0) = ai0 * B(0,0) + ai1 * B(1,0) + ai2 * B(2,0);
+      P(i,1) = ai0 * B(0,1) + ai1 * B(1,1) + ai2 * B(2,1);
+      P(i,2) = ai0 * B(0,2) + ai1 * B(1,2) + ai2 * B(2,2);
+      P(i,3) = ai0 * B(0,3) + ai1 * B(1,3) + ai2 * B(2,3) + ai3;
+   }
+   P(3,0) = 0;
+   P(3,1) = 0;
+   P(3,2) = 0;
+   P(3,3) = 1;
+}
+
+
+#undef A
+#undef B
+#undef P
+
+
+/*
+ * Multiply a matrix by an array of floats with known properties.
+ */
+static void matrix_multf( GLmatrix *mat, const GLfloat *m, GLuint flags )
+{
+   mat->flags |= (flags | MAT_DIRTY_TYPE | MAT_DIRTY_INVERSE);
+
+   if (TEST_MAT_FLAGS(mat, MAT_FLAGS_3D))
+      matmul34( mat->m, mat->m, m );
+   else 
+      matmul4( mat->m, mat->m, m ); 
+}
+
+
+static void print_matrix_floats( const GLfloat m[16] )
+{
+   int i;
+   for (i=0;i<4;i++) {
+      fprintf(stderr,"\t%f %f %f %f\n", m[i], m[4+i], m[8+i], m[12+i] );
+   }
+}
+
+void 
+_math_matrix_print( const GLmatrix *m )
+{
+   fprintf(stderr, "Matrix type: %s, flags: %x\n", types[m->type], m->flags);
+   print_matrix_floats(m->m);
+   fprintf(stderr, "Inverse: \n");
+   if (m->inv) {
+      GLfloat prod[16];
+      print_matrix_floats(m->inv);
+      matmul4(prod, m->m, m->inv);
+      fprintf(stderr, "Mat * Inverse:\n");
+      print_matrix_floats(prod);
+   }
+   else {
+      fprintf(stderr, "  - not available\n");
+   }
+}
+
+
+
+
+#define SWAP_ROWS(a, b) { GLfloat *_tmp = a; (a)=(b); (b)=_tmp; }
+#define MAT(m,r,c) (m)[(c)*4+(r)]
+
+/*
+ * Compute inverse of 4x4 transformation matrix.
+ * Code contributed by Jacques Leroy [email protected]
+ * Return GL_TRUE for success, GL_FALSE for failure (singular matrix)
+ */
+static GLboolean invert_matrix_general( GLmatrix *mat )
+{
+   const GLfloat *m = mat->m;
+   GLfloat *out = mat->inv;
+   GLfloat wtmp[4][8];
+   GLfloat m0, m1, m2, m3, s;
+   GLfloat *r0, *r1, *r2, *r3;
+  
+   r0 = wtmp[0], r1 = wtmp[1], r2 = wtmp[2], r3 = wtmp[3];
+  
+   r0[0] = MAT(m,0,0), r0[1] = MAT(m,0,1),
+   r0[2] = MAT(m,0,2), r0[3] = MAT(m,0,3),
+   r0[4] = 1.0, r0[5] = r0[6] = r0[7] = 0.0,
+  
+   r1[0] = MAT(m,1,0), r1[1] = MAT(m,1,1),
+   r1[2] = MAT(m,1,2), r1[3] = MAT(m,1,3),
+   r1[5] = 1.0, r1[4] = r1[6] = r1[7] = 0.0,
+  
+   r2[0] = MAT(m,2,0), r2[1] = MAT(m,2,1),
+   r2[2] = MAT(m,2,2), r2[3] = MAT(m,2,3),
+   r2[6] = 1.0, r2[4] = r2[5] = r2[7] = 0.0,
+  
+   r3[0] = MAT(m,3,0), r3[1] = MAT(m,3,1),
+   r3[2] = MAT(m,3,2), r3[3] = MAT(m,3,3),
+   r3[7] = 1.0, r3[4] = r3[5] = r3[6] = 0.0;
+  
+   /* choose pivot - or die */
+   if (fabs(r3[0])>fabs(r2[0])) SWAP_ROWS(r3, r2);
+   if (fabs(r2[0])>fabs(r1[0])) SWAP_ROWS(r2, r1);
+   if (fabs(r1[0])>fabs(r0[0])) SWAP_ROWS(r1, r0);
+   if (0.0 == r0[0])  return GL_FALSE;
+  
+   /* eliminate first variable     */
+   m1 = r1[0]/r0[0]; m2 = r2[0]/r0[0]; m3 = r3[0]/r0[0];
+   s = r0[1]; r1[1] -= m1 * s; r2[1] -= m2 * s; r3[1] -= m3 * s;
+   s = r0[2]; r1[2] -= m1 * s; r2[2] -= m2 * s; r3[2] -= m3 * s;
+   s = r0[3]; r1[3] -= m1 * s; r2[3] -= m2 * s; r3[3] -= m3 * s;
+   s = r0[4];
+   if (s != 0.0) { r1[4] -= m1 * s; r2[4] -= m2 * s; r3[4] -= m3 * s; }
+   s = r0[5];
+   if (s != 0.0) { r1[5] -= m1 * s; r2[5] -= m2 * s; r3[5] -= m3 * s; }
+   s = r0[6];
+   if (s != 0.0) { r1[6] -= m1 * s; r2[6] -= m2 * s; r3[6] -= m3 * s; }
+   s = r0[7];
+   if (s != 0.0) { r1[7] -= m1 * s; r2[7] -= m2 * s; r3[7] -= m3 * s; }
+  
+   /* choose pivot - or die */
+   if (fabs(r3[1])>fabs(r2[1])) SWAP_ROWS(r3, r2);
+   if (fabs(r2[1])>fabs(r1[1])) SWAP_ROWS(r2, r1);
+   if (0.0 == r1[1])  return GL_FALSE;
+  
+   /* eliminate second variable */
+   m2 = r2[1]/r1[1]; m3 = r3[1]/r1[1];
+   r2[2] -= m2 * r1[2]; r3[2] -= m3 * r1[2];
+   r2[3] -= m2 * r1[3]; r3[3] -= m3 * r1[3];
+   s = r1[4]; if (0.0 != s) { r2[4] -= m2 * s; r3[4] -= m3 * s; }
+   s = r1[5]; if (0.0 != s) { r2[5] -= m2 * s; r3[5] -= m3 * s; }
+   s = r1[6]; if (0.0 != s) { r2[6] -= m2 * s; r3[6] -= m3 * s; }
+   s = r1[7]; if (0.0 != s) { r2[7] -= m2 * s; r3[7] -= m3 * s; }
+  
+   /* choose pivot - or die */
+   if (fabs(r3[2])>fabs(r2[2])) SWAP_ROWS(r3, r2);
+   if (0.0 == r2[2])  return GL_FALSE;
+  
+   /* eliminate third variable */
+   m3 = r3[2]/r2[2];
+   r3[3] -= m3 * r2[3], r3[4] -= m3 * r2[4],
+   r3[5] -= m3 * r2[5], r3[6] -= m3 * r2[6],
+   r3[7] -= m3 * r2[7];
+  
+   /* last check */
+   if (0.0 == r3[3]) return GL_FALSE;
+  
+   s = 1.0/r3[3];              /* now back substitute row 3 */
+   r3[4] *= s; r3[5] *= s; r3[6] *= s; r3[7] *= s;
+  
+   m2 = r2[3];                 /* now back substitute row 2 */
+   s  = 1.0/r2[2];
+   r2[4] = s * (r2[4] - r3[4] * m2), r2[5] = s * (r2[5] - r3[5] * m2),
+   r2[6] = s * (r2[6] - r3[6] * m2), r2[7] = s * (r2[7] - r3[7] * m2);
+   m1 = r1[3];
+   r1[4] -= r3[4] * m1, r1[5] -= r3[5] * m1,
+   r1[6] -= r3[6] * m1, r1[7] -= r3[7] * m1;
+   m0 = r0[3];
+   r0[4] -= r3[4] * m0, r0[5] -= r3[5] * m0,
+   r0[6] -= r3[6] * m0, r0[7] -= r3[7] * m0;
+  
+   m1 = r1[2];                 /* now back substitute row 1 */
+   s  = 1.0/r1[1];
+   r1[4] = s * (r1[4] - r2[4] * m1), r1[5] = s * (r1[5] - r2[5] * m1),
+   r1[6] = s * (r1[6] - r2[6] * m1), r1[7] = s * (r1[7] - r2[7] * m1);
+   m0 = r0[2];
+   r0[4] -= r2[4] * m0, r0[5] -= r2[5] * m0,
+   r0[6] -= r2[6] * m0, r0[7] -= r2[7] * m0;
+  
+   m0 = r0[1];                 /* now back substitute row 0 */
+   s  = 1.0/r0[0];
+   r0[4] = s * (r0[4] - r1[4] * m0), r0[5] = s * (r0[5] - r1[5] * m0),
+   r0[6] = s * (r0[6] - r1[6] * m0), r0[7] = s * (r0[7] - r1[7] * m0);
+  
+   MAT(out,0,0) = r0[4]; MAT(out,0,1) = r0[5],
+   MAT(out,0,2) = r0[6]; MAT(out,0,3) = r0[7],
+   MAT(out,1,0) = r1[4]; MAT(out,1,1) = r1[5],
+   MAT(out,1,2) = r1[6]; MAT(out,1,3) = r1[7],
+   MAT(out,2,0) = r2[4]; MAT(out,2,1) = r2[5],
+   MAT(out,2,2) = r2[6]; MAT(out,2,3) = r2[7],
+   MAT(out,3,0) = r3[4]; MAT(out,3,1) = r3[5],
+   MAT(out,3,2) = r3[6]; MAT(out,3,3) = r3[7]; 
+  
+   return GL_TRUE;
+}
+#undef SWAP_ROWS
+
+
+/* Adapted from graphics gems II.
+ */  
+static GLboolean invert_matrix_3d_general( GLmatrix *mat )
+{
+   const GLfloat *in = mat->m;
+   GLfloat *out = mat->inv;
+   GLfloat pos, neg, t;
+   GLfloat det;
+
+   /* Calculate the determinant of upper left 3x3 submatrix and
+    * determine if the matrix is singular. 
+    */
+   pos = neg = 0.0;
+   t =  MAT(in,0,0) * MAT(in,1,1) * MAT(in,2,2);
+   if (t >= 0.0) pos += t; else neg += t;
+
+   t =  MAT(in,1,0) * MAT(in,2,1) * MAT(in,0,2);
+   if (t >= 0.0) pos += t; else neg += t;
+
+   t =  MAT(in,2,0) * MAT(in,0,1) * MAT(in,1,2);
+   if (t >= 0.0) pos += t; else neg += t;
+
+   t = -MAT(in,2,0) * MAT(in,1,1) * MAT(in,0,2);
+   if (t >= 0.0) pos += t; else neg += t;
+
+   t = -MAT(in,1,0) * MAT(in,0,1) * MAT(in,2,2);
+   if (t >= 0.0) pos += t; else neg += t;
+
+   t = -MAT(in,0,0) * MAT(in,2,1) * MAT(in,1,2);
+   if (t >= 0.0) pos += t; else neg += t;
+
+   det = pos + neg;
+
+   if (det*det < 1e-25) 
+      return GL_FALSE;
+   
+   det = 1.0 / det;
+   MAT(out,0,0) = (  (MAT(in,1,1)*MAT(in,2,2) - MAT(in,2,1)*MAT(in,1,2) )*det);
+   MAT(out,0,1) = (- (MAT(in,0,1)*MAT(in,2,2) - MAT(in,2,1)*MAT(in,0,2) )*det);
+   MAT(out,0,2) = (  (MAT(in,0,1)*MAT(in,1,2) - MAT(in,1,1)*MAT(in,0,2) )*det);
+   MAT(out,1,0) = (- (MAT(in,1,0)*MAT(in,2,2) - MAT(in,2,0)*MAT(in,1,2) )*det);
+   MAT(out,1,1) = (  (MAT(in,0,0)*MAT(in,2,2) - MAT(in,2,0)*MAT(in,0,2) )*det);
+   MAT(out,1,2) = (- (MAT(in,0,0)*MAT(in,1,2) - MAT(in,1,0)*MAT(in,0,2) )*det);
+   MAT(out,2,0) = (  (MAT(in,1,0)*MAT(in,2,1) - MAT(in,2,0)*MAT(in,1,1) )*det);
+   MAT(out,2,1) = (- (MAT(in,0,0)*MAT(in,2,1) - MAT(in,2,0)*MAT(in,0,1) )*det);
+   MAT(out,2,2) = (  (MAT(in,0,0)*MAT(in,1,1) - MAT(in,1,0)*MAT(in,0,1) )*det);
+
+   /* Do the translation part */
+   MAT(out,0,3) = - (MAT(in,0,3) * MAT(out,0,0) +
+		     MAT(in,1,3) * MAT(out,0,1) +
+		     MAT(in,2,3) * MAT(out,0,2) );
+   MAT(out,1,3) = - (MAT(in,0,3) * MAT(out,1,0) +
+		     MAT(in,1,3) * MAT(out,1,1) +
+		     MAT(in,2,3) * MAT(out,1,2) );
+   MAT(out,2,3) = - (MAT(in,0,3) * MAT(out,2,0) +
+		     MAT(in,1,3) * MAT(out,2,1) +
+		     MAT(in,2,3) * MAT(out,2,2) );
+    
+   return GL_TRUE;
+}
+
+
+static GLboolean invert_matrix_3d( GLmatrix *mat )
+{
+   const GLfloat *in = mat->m;
+   GLfloat *out = mat->inv;
+
+   if (!TEST_MAT_FLAGS(mat, MAT_FLAGS_ANGLE_PRESERVING)) {
+      return invert_matrix_3d_general( mat );
+   }
+   
+   if (mat->flags & MAT_FLAG_UNIFORM_SCALE) {
+      GLfloat scale = (MAT(in,0,0) * MAT(in,0,0) +
+                       MAT(in,0,1) * MAT(in,0,1) +
+                       MAT(in,0,2) * MAT(in,0,2));
+
+      if (scale == 0.0) 
+         return GL_FALSE;
+
+      scale = 1.0 / scale;
+
+      /* Transpose and scale the 3 by 3 upper-left submatrix. */
+      MAT(out,0,0) = scale * MAT(in,0,0);
+      MAT(out,1,0) = scale * MAT(in,0,1);
+      MAT(out,2,0) = scale * MAT(in,0,2);
+      MAT(out,0,1) = scale * MAT(in,1,0);
+      MAT(out,1,1) = scale * MAT(in,1,1);
+      MAT(out,2,1) = scale * MAT(in,1,2);
+      MAT(out,0,2) = scale * MAT(in,2,0);
+      MAT(out,1,2) = scale * MAT(in,2,1);
+      MAT(out,2,2) = scale * MAT(in,2,2);
+   }
+   else if (mat->flags & MAT_FLAG_ROTATION) {
+      /* Transpose the 3 by 3 upper-left submatrix. */
+      MAT(out,0,0) = MAT(in,0,0);
+      MAT(out,1,0) = MAT(in,0,1);
+      MAT(out,2,0) = MAT(in,0,2);
+      MAT(out,0,1) = MAT(in,1,0);
+      MAT(out,1,1) = MAT(in,1,1);
+      MAT(out,2,1) = MAT(in,1,2);
+      MAT(out,0,2) = MAT(in,2,0);
+      MAT(out,1,2) = MAT(in,2,1);
+      MAT(out,2,2) = MAT(in,2,2);
+   }
+   else {
+      /* pure translation */
+      MEMCPY( out, Identity, sizeof(Identity) );
+      MAT(out,0,3) = - MAT(in,0,3);
+      MAT(out,1,3) = - MAT(in,1,3);
+      MAT(out,2,3) = - MAT(in,2,3);
+      return GL_TRUE;
+   }
+    
+   if (mat->flags & MAT_FLAG_TRANSLATION) {
+      /* Do the translation part */
+      MAT(out,0,3) = - (MAT(in,0,3) * MAT(out,0,0) +
+			MAT(in,1,3) * MAT(out,0,1) +
+			MAT(in,2,3) * MAT(out,0,2) );
+      MAT(out,1,3) = - (MAT(in,0,3) * MAT(out,1,0) +
+			MAT(in,1,3) * MAT(out,1,1) +
+			MAT(in,2,3) * MAT(out,1,2) );
+      MAT(out,2,3) = - (MAT(in,0,3) * MAT(out,2,0) +
+			MAT(in,1,3) * MAT(out,2,1) +
+			MAT(in,2,3) * MAT(out,2,2) );
+   }
+   else {
+      MAT(out,0,3) = MAT(out,1,3) = MAT(out,2,3) = 0.0;
+   }
+    
+   return GL_TRUE;
+}
+
+  
+
+static GLboolean invert_matrix_identity( GLmatrix *mat )
+{
+   MEMCPY( mat->inv, Identity, sizeof(Identity) );
+   return GL_TRUE;
+}
+
+
+static GLboolean invert_matrix_3d_no_rot( GLmatrix *mat )
+{
+   const GLfloat *in = mat->m;
+   GLfloat *out = mat->inv;
+
+   if (MAT(in,0,0) == 0 || MAT(in,1,1) == 0 || MAT(in,2,2) == 0 )       
+      return GL_FALSE;
+  
+   MEMCPY( out, Identity, 16 * sizeof(GLfloat) );
+   MAT(out,0,0) = 1.0 / MAT(in,0,0);
+   MAT(out,1,1) = 1.0 / MAT(in,1,1);
+   MAT(out,2,2) = 1.0 / MAT(in,2,2);
+
+   if (mat->flags & MAT_FLAG_TRANSLATION) {
+      MAT(out,0,3) = - (MAT(in,0,3) * MAT(out,0,0));
+      MAT(out,1,3) = - (MAT(in,1,3) * MAT(out,1,1));
+      MAT(out,2,3) = - (MAT(in,2,3) * MAT(out,2,2));
+   }
+
+   return GL_TRUE;
+}
+
+
+static GLboolean invert_matrix_2d_no_rot( GLmatrix *mat )
+{
+   const GLfloat *in = mat->m;
+   GLfloat *out = mat->inv;
+
+   if (MAT(in,0,0) == 0 || MAT(in,1,1) == 0)       
+      return GL_FALSE;
+  
+   MEMCPY( out, Identity, 16 * sizeof(GLfloat) );
+   MAT(out,0,0) = 1.0 / MAT(in,0,0);
+   MAT(out,1,1) = 1.0 / MAT(in,1,1);
+
+   if (mat->flags & MAT_FLAG_TRANSLATION) {
+      MAT(out,0,3) = - (MAT(in,0,3) * MAT(out,0,0));
+      MAT(out,1,3) = - (MAT(in,1,3) * MAT(out,1,1));
+   }
+
+   return GL_TRUE;
+}
+
+
+static GLboolean invert_matrix_perspective( GLmatrix *mat )
+{
+   const GLfloat *in = mat->m;
+   GLfloat *out = mat->inv;
+
+   if (MAT(in,2,3) == 0)
+      return GL_FALSE;
+
+   MEMCPY( out, Identity, 16 * sizeof(GLfloat) );
+
+   MAT(out,0,0) = 1.0 / MAT(in,0,0);
+   MAT(out,1,1) = 1.0 / MAT(in,1,1);
+
+   MAT(out,0,3) = MAT(in,0,2);
+   MAT(out,1,3) = MAT(in,1,2);
+
+   MAT(out,2,2) = 0;
+   MAT(out,2,3) = -1;
+
+   MAT(out,3,2) = 1.0 / MAT(in,2,3);
+   MAT(out,3,3) = MAT(in,2,2) * MAT(out,3,2);
+
+   return GL_TRUE;
+}
+
+
+typedef GLboolean (*inv_mat_func)( GLmatrix *mat );
+
+
+static inv_mat_func inv_mat_tab[7] = {
+   invert_matrix_general,
+   invert_matrix_identity,
+   invert_matrix_3d_no_rot,
+   invert_matrix_perspective,
+   invert_matrix_3d,		/* lazy! */
+   invert_matrix_2d_no_rot,
+   invert_matrix_3d
+};
+
+
+static GLboolean matrix_invert( GLmatrix *mat )
+{
+   if (inv_mat_tab[mat->type](mat)) {
+      mat->flags &= ~MAT_FLAG_SINGULAR;
+      return GL_TRUE;
+   } else {
+      mat->flags |= MAT_FLAG_SINGULAR;
+      MEMCPY( mat->inv, Identity, sizeof(Identity) );
+      return GL_FALSE;
+   }  
+}
+
+
+
+
+
+
+/*
+ * Generate a 4x4 transformation matrix from glRotate parameters, and
+ * postmultiply the input matrix by it.
+ */
+void 
+_math_matrix_rotate( GLmatrix *mat, 
+		     GLfloat angle, GLfloat x, GLfloat y, GLfloat z )
+{
+   /* This function contributed by Erich Boleyn ([email protected]) */
+   GLfloat mag, s, c;
+   GLfloat xx, yy, zz, xy, yz, zx, xs, ys, zs, one_c;
+   GLfloat m[16];
+   
+   s = sin( angle * DEG2RAD );
+   c = cos( angle * DEG2RAD );
+
+   mag = GL_SQRT( x*x + y*y + z*z );
+
+   if (mag <= 1.0e-4) {
+      /* generate an identity matrix and return */
+      MEMCPY(m, Identity, sizeof(GLfloat)*16);
+      return;
+   }
+
+   x /= mag;
+   y /= mag;
+   z /= mag;
+
+#define M(row,col)  m[col*4+row]
+
+   /*
+    *     Arbitrary axis rotation matrix.
+    *
+    *  This is composed of 5 matrices, Rz, Ry, T, Ry', Rz', multiplied
+    *  like so:  Rz * Ry * T * Ry' * Rz'.  T is the final rotation
+    *  (which is about the X-axis), and the two composite transforms
+    *  Ry' * Rz' and Rz * Ry are (respectively) the rotations necessary
+    *  from the arbitrary axis to the X-axis then back.  They are
+    *  all elementary rotations.
+    *
+    *  Rz' is a rotation about the Z-axis, to bring the axis vector
+    *  into the x-z plane.  Then Ry' is applied, rotating about the
+    *  Y-axis to bring the axis vector parallel with the X-axis.  The
+    *  rotation about the X-axis is then performed.  Ry and Rz are
+    *  simply the respective inverse transforms to bring the arbitrary
+    *  axis back to it's original orientation.  The first transforms
+    *  Rz' and Ry' are considered inverses, since the data from the
+    *  arbitrary axis gives you info on how to get to it, not how
+    *  to get away from it, and an inverse must be applied.
+    *
+    *  The basic calculation used is to recognize that the arbitrary
+    *  axis vector (x, y, z), since it is of unit length, actually
+    *  represents the sines and cosines of the angles to rotate the
+    *  X-axis to the same orientation, with theta being the angle about
+    *  Z and phi the angle about Y (in the order described above)
+    *  as follows:
+    *
+    *  cos ( theta ) = x / sqrt ( 1 - z^2 )
+    *  sin ( theta ) = y / sqrt ( 1 - z^2 )
+    *
+    *  cos ( phi ) = sqrt ( 1 - z^2 )
+    *  sin ( phi ) = z
+    *
+    *  Note that cos ( phi ) can further be inserted to the above
+    *  formulas:
+    *
+    *  cos ( theta ) = x / cos ( phi )
+    *  sin ( theta ) = y / sin ( phi )
+    *
+    *  ...etc.  Because of those relations and the standard trigonometric
+    *  relations, it is pssible to reduce the transforms down to what
+    *  is used below.  It may be that any primary axis chosen will give the
+    *  same results (modulo a sign convention) using thie method.
+    *
+    *  Particularly nice is to notice that all divisions that might
+    *  have caused trouble when parallel to certain planes or
+    *  axis go away with care paid to reducing the expressions.
+    *  After checking, it does perform correctly under all cases, since
+    *  in all the cases of division where the denominator would have
+    *  been zero, the numerator would have been zero as well, giving
+    *  the expected result.
+    */
+
+   xx = x * x;
+   yy = y * y;
+   zz = z * z;
+   xy = x * y;
+   yz = y * z;
+   zx = z * x;
+   xs = x * s;
+   ys = y * s;
+   zs = z * s;
+   one_c = 1.0F - c;
+
+   M(0,0) = (one_c * xx) + c;
+   M(0,1) = (one_c * xy) - zs;
+   M(0,2) = (one_c * zx) + ys;
+   M(0,3) = 0.0F;
+
+   M(1,0) = (one_c * xy) + zs;
+   M(1,1) = (one_c * yy) + c;
+   M(1,2) = (one_c * yz) - xs;
+   M(1,3) = 0.0F;
+
+   M(2,0) = (one_c * zx) - ys;
+   M(2,1) = (one_c * yz) + xs;
+   M(2,2) = (one_c * zz) + c;
+   M(2,3) = 0.0F;
+
+   M(3,0) = 0.0F;
+   M(3,1) = 0.0F;
+   M(3,2) = 0.0F;
+   M(3,3) = 1.0F;
+
+#undef M
+
+   matrix_multf( mat, m, MAT_FLAG_ROTATION );
+}
+
+
+void
+_math_matrix_frustrum( GLmatrix *mat, 
+		       GLfloat left, GLfloat right,
+		       GLfloat bottom, GLfloat top, 
+		       GLfloat nearval, GLfloat farval )
+{
+   GLfloat x, y, a, b, c, d;
+   GLfloat m[16];
+
+   x = (2.0*nearval) / (right-left);
+   y = (2.0*nearval) / (top-bottom);
+   a = (right+left) / (right-left);
+   b = (top+bottom) / (top-bottom);
+   c = -(farval+nearval) / ( farval-nearval);
+   d = -(2.0*farval*nearval) / (farval-nearval);  /* error? */
+
+#define M(row,col)  m[col*4+row]
+   M(0,0) = x;     M(0,1) = 0.0F;  M(0,2) = a;      M(0,3) = 0.0F;
+   M(1,0) = 0.0F;  M(1,1) = y;     M(1,2) = b;      M(1,3) = 0.0F;
+   M(2,0) = 0.0F;  M(2,1) = 0.0F;  M(2,2) = c;      M(2,3) = d;
+   M(3,0) = 0.0F;  M(3,1) = 0.0F;  M(3,2) = -1.0F;  M(3,3) = 0.0F;
+#undef M
+
+   matrix_multf( mat, m, MAT_FLAG_PERSPECTIVE );
+}
+
+void
+_math_matrix_ortho( GLmatrix *mat, 	
+		    GLfloat left, GLfloat right,
+		    GLfloat bottom, GLfloat top, 
+		    GLfloat nearval, GLfloat farval )
+{
+   GLfloat x, y, z;
+   GLfloat tx, ty, tz;
+   GLfloat m[16];
+
+   x = 2.0 / (right-left);
+   y = 2.0 / (top-bottom);
+   z = -2.0 / (farval-nearval);
+   tx = -(right+left) / (right-left);
+   ty = -(top+bottom) / (top-bottom);
+   tz = -(farval+nearval) / (farval-nearval);
+
+#define M(row,col)  m[col*4+row]
+   M(0,0) = x;     M(0,1) = 0.0F;  M(0,2) = 0.0F;  M(0,3) = tx;
+   M(1,0) = 0.0F;  M(1,1) = y;     M(1,2) = 0.0F;  M(1,3) = ty;
+   M(2,0) = 0.0F;  M(2,1) = 0.0F;  M(2,2) = z;     M(2,3) = tz;
+   M(3,0) = 0.0F;  M(3,1) = 0.0F;  M(3,2) = 0.0F;  M(3,3) = 1.0F;
+#undef M
+
+   matrix_multf( mat, m, (MAT_FLAG_GENERAL_SCALE|MAT_FLAG_TRANSLATION));
+}
+
+
+#define ZERO(x) (1<<x)
+#define ONE(x)  (1<<(x+16))
+
+#define MASK_NO_TRX      (ZERO(12) | ZERO(13) | ZERO(14))
+#define MASK_NO_2D_SCALE ( ONE(0)  | ONE(5))
+
+#define MASK_IDENTITY    ( ONE(0)  | ZERO(4)  | ZERO(8)  | ZERO(12) |\
+			  ZERO(1)  |  ONE(5)  | ZERO(9)  | ZERO(13) |\
+			  ZERO(2)  | ZERO(6)  |  ONE(10) | ZERO(14) |\
+			  ZERO(3)  | ZERO(7)  | ZERO(11) |  ONE(15) )
+
+#define MASK_2D_NO_ROT   (           ZERO(4)  | ZERO(8)  |           \
+			  ZERO(1)  |            ZERO(9)  |           \
+			  ZERO(2)  | ZERO(6)  |  ONE(10) | ZERO(14) |\
+			  ZERO(3)  | ZERO(7)  | ZERO(11) |  ONE(15) )
+
+#define MASK_2D          (                      ZERO(8)  |           \
+			                        ZERO(9)  |           \
+			  ZERO(2)  | ZERO(6)  |  ONE(10) | ZERO(14) |\
+			  ZERO(3)  | ZERO(7)  | ZERO(11) |  ONE(15) )
+
+
+#define MASK_3D_NO_ROT   (           ZERO(4)  | ZERO(8)  |           \
+			  ZERO(1)  |            ZERO(9)  |           \
+			  ZERO(2)  | ZERO(6)  |                      \
+			  ZERO(3)  | ZERO(7)  | ZERO(11) |  ONE(15) )
+
+#define MASK_3D          (                                           \
+			                                             \
+			                                             \
+			  ZERO(3)  | ZERO(7)  | ZERO(11) |  ONE(15) )
+
+
+#define MASK_PERSPECTIVE (           ZERO(4)  |            ZERO(12) |\
+			  ZERO(1)  |                       ZERO(13) |\
+			  ZERO(2)  | ZERO(6)  |                      \
+			  ZERO(3)  | ZERO(7)  |            ZERO(15) )
+
+#define SQ(x) ((x)*(x))
+  
+/* Determine type and flags from scratch.  This is expensive enough to
+ * only want to do it once.
+ */
+static void analyze_from_scratch( GLmatrix *mat )
+{
+   const GLfloat *m = mat->m;
+   GLuint mask = 0;
+   GLuint i;
+
+   for (i = 0 ; i < 16 ; i++) {
+      if (m[i] == 0.0) mask |= (1<<i);
+   }
+   
+   if (m[0] == 1.0F) mask |= (1<<16);
+   if (m[5] == 1.0F) mask |= (1<<21);
+   if (m[10] == 1.0F) mask |= (1<<26);
+   if (m[15] == 1.0F) mask |= (1<<31);
+
+   mat->flags &= ~MAT_FLAGS_GEOMETRY;
+
+   /* Check for translation - no-one really cares 
+    */
+   if ((mask & MASK_NO_TRX) != MASK_NO_TRX) 
+      mat->flags |= MAT_FLAG_TRANSLATION;      
+
+   /* Do the real work
+    */
+   if (mask == MASK_IDENTITY) {
+      mat->type = MATRIX_IDENTITY;
+   }
+   else if ((mask & MASK_2D_NO_ROT) == MASK_2D_NO_ROT) {
+      mat->type = MATRIX_2D_NO_ROT;
+      
+      if ((mask & MASK_NO_2D_SCALE) != MASK_NO_2D_SCALE)
+	 mat->flags = MAT_FLAG_GENERAL_SCALE;
+   }
+   else if ((mask & MASK_2D) == MASK_2D) {
+      GLfloat mm = DOT2(m, m);
+      GLfloat m4m4 = DOT2(m+4,m+4);
+      GLfloat mm4 = DOT2(m,m+4);
+
+      mat->type = MATRIX_2D;
+
+      /* Check for scale */
+      if (SQ(mm-1) > SQ(1e-6) ||
+	  SQ(m4m4-1) > SQ(1e-6)) 
+	 mat->flags |= MAT_FLAG_GENERAL_SCALE;
+
+      /* Check for rotation */
+      if (SQ(mm4) > SQ(1e-6))
+	 mat->flags |= MAT_FLAG_GENERAL_3D;
+      else
+	 mat->flags |= MAT_FLAG_ROTATION;
+
+   }
+   else if ((mask & MASK_3D_NO_ROT) == MASK_3D_NO_ROT) {
+      mat->type = MATRIX_3D_NO_ROT;
+
+      /* Check for scale */
+      if (SQ(m[0]-m[5]) < SQ(1e-6) && 
+	  SQ(m[0]-m[10]) < SQ(1e-6)) {
+	 if (SQ(m[0]-1.0) > SQ(1e-6)) {
+	    mat->flags |= MAT_FLAG_UNIFORM_SCALE;
+         }
+      }
+      else {
+	 mat->flags |= MAT_FLAG_GENERAL_SCALE;
+      }
+   }
+   else if ((mask & MASK_3D) == MASK_3D) {
+      GLfloat c1 = DOT3(m,m);
+      GLfloat c2 = DOT3(m+4,m+4);
+      GLfloat c3 = DOT3(m+8,m+8);
+      GLfloat d1 = DOT3(m, m+4);
+      GLfloat cp[3];
+
+      mat->type = MATRIX_3D;
+
+      /* Check for scale */
+      if (SQ(c1-c2) < SQ(1e-6) && SQ(c1-c3) < SQ(1e-6)) {
+	 if (SQ(c1-1.0) > SQ(1e-6))
+	    mat->flags |= MAT_FLAG_UNIFORM_SCALE;
+	 /* else no scale at all */
+      }
+      else {
+	 mat->flags |= MAT_FLAG_GENERAL_SCALE;
+      }
+
+      /* Check for rotation */
+      if (SQ(d1) < SQ(1e-6)) {
+	 CROSS3( cp, m, m+4 );
+	 SUB_3V( cp, cp, (m+8) );
+	 if (LEN_SQUARED_3FV(cp) < SQ(1e-6)) 
+	    mat->flags |= MAT_FLAG_ROTATION;
+	 else
+	    mat->flags |= MAT_FLAG_GENERAL_3D;
+      }
+      else {
+	 mat->flags |= MAT_FLAG_GENERAL_3D; /* shear, etc */
+      }
+   }
+   else if ((mask & MASK_PERSPECTIVE) == MASK_PERSPECTIVE && m[11]==-1.0F) {
+      mat->type = MATRIX_PERSPECTIVE;
+      mat->flags |= MAT_FLAG_GENERAL;
+   }
+   else {
+      mat->type = MATRIX_GENERAL;
+      mat->flags |= MAT_FLAG_GENERAL;
+   }
+}
+
+
+/* Analyse a matrix given that its flags are accurate - this is the
+ * more common operation, hopefully. 
+ */
+static void analyze_from_flags( GLmatrix *mat )
+{
+   const GLfloat *m = mat->m;
+
+   if (TEST_MAT_FLAGS(mat, 0)) {
+      mat->type = MATRIX_IDENTITY;
+   }
+   else if (TEST_MAT_FLAGS(mat, (MAT_FLAG_TRANSLATION |
+				 MAT_FLAG_UNIFORM_SCALE |
+				 MAT_FLAG_GENERAL_SCALE))) {
+      if ( m[10]==1.0F && m[14]==0.0F ) {
+	 mat->type = MATRIX_2D_NO_ROT;
+      }
+      else {
+	 mat->type = MATRIX_3D_NO_ROT;
+      }
+   }
+   else if (TEST_MAT_FLAGS(mat, MAT_FLAGS_3D)) {
+      if (                                 m[ 8]==0.0F               
+            &&                             m[ 9]==0.0F
+            && m[2]==0.0F && m[6]==0.0F && m[10]==1.0F && m[14]==0.0F) {
+	 mat->type = MATRIX_2D;
+      }
+      else {
+	 mat->type = MATRIX_3D;
+      }
+   }
+   else if (                 m[4]==0.0F                 && m[12]==0.0F
+            && m[1]==0.0F                               && m[13]==0.0F
+            && m[2]==0.0F && m[6]==0.0F
+            && m[3]==0.0F && m[7]==0.0F && m[11]==-1.0F && m[15]==0.0F) {
+      mat->type = MATRIX_PERSPECTIVE;
+   }
+   else {
+      mat->type = MATRIX_GENERAL;
+   }
+}
+
+
+void 
+_math_matrix_analyze( GLmatrix *mat ) 
+{
+   if (mat->flags & MAT_DIRTY_TYPE) {
+      if (mat->flags & MAT_DIRTY_FLAGS) 
+	 analyze_from_scratch( mat );
+      else
+	 analyze_from_flags( mat );
+   }
+
+   if (mat->inv && (mat->flags & MAT_DIRTY_INVERSE)) {
+      matrix_invert( mat );
+   }
+
+   mat->flags &= ~(MAT_DIRTY_FLAGS|
+		   MAT_DIRTY_TYPE|
+		   MAT_DIRTY_INVERSE);
+}
+
+
+void 
+_math_matrix_copy( GLmatrix *to, const GLmatrix *from )
+{
+   MEMCPY( to->m, from->m, sizeof(Identity) );
+   to->flags = from->flags;
+   to->type = from->type;
+
+   if (to->inv != 0) {
+      if (from->inv == 0) {
+	 matrix_invert( to );
+      }
+      else {
+	 MEMCPY(to->inv, from->inv, sizeof(GLfloat)*16);
+      }
+   }
+}
+
+
+void 
+_math_matrix_scale( GLmatrix *mat, GLfloat x, GLfloat y, GLfloat z )
+{
+   GLfloat *m = mat->m;
+   m[0] *= x;   m[4] *= y;   m[8]  *= z;
+   m[1] *= x;   m[5] *= y;   m[9]  *= z;
+   m[2] *= x;   m[6] *= y;   m[10] *= z;
+   m[3] *= x;   m[7] *= y;   m[11] *= z;
+
+   if (fabs(x - y) < 1e-8 && fabs(x - z) < 1e-8)
+      mat->flags |= MAT_FLAG_UNIFORM_SCALE;
+   else
+      mat->flags |= MAT_FLAG_GENERAL_SCALE;
+
+   mat->flags |= (MAT_DIRTY_TYPE | 
+		  MAT_DIRTY_INVERSE);
+}
+
+
+void 
+_math_matrix_translate( GLmatrix *mat, GLfloat x, GLfloat y, GLfloat z )
+{
+   GLfloat *m = mat->m;
+   m[12] = m[0] * x + m[4] * y + m[8]  * z + m[12];
+   m[13] = m[1] * x + m[5] * y + m[9]  * z + m[13];
+   m[14] = m[2] * x + m[6] * y + m[10] * z + m[14];
+   m[15] = m[3] * x + m[7] * y + m[11] * z + m[15];
+
+   mat->flags |= (MAT_FLAG_TRANSLATION | 
+		  MAT_DIRTY_TYPE | 
+		  MAT_DIRTY_INVERSE);
+}
+
+
+void 
+_math_matrix_loadf( GLmatrix *mat, const GLfloat *m )
+{
+   MEMCPY( mat->m, m, 16*sizeof(GLfloat) );
+   mat->flags = (MAT_FLAG_GENERAL | MAT_DIRTY);
+}
+
+void 
+_math_matrix_ctr( GLmatrix *m )
+{
+   if ( m->m == 0 ) {
+      m->m = (GLfloat *) ALIGN_MALLOC( 16 * sizeof(GLfloat), 16 );
+   }
+   MEMCPY( m->m, Identity, sizeof(Identity) );
+   m->inv = 0;
+   m->type = MATRIX_IDENTITY;
+   m->flags = 0;
+}
+
+void 
+_math_matrix_dtr( GLmatrix *m )
+{
+   if ( m->m != 0 ) {
+      ALIGN_FREE( m->m );
+      m->m = 0;
+   }
+   if ( m->inv != 0 ) {
+      ALIGN_FREE( m->inv );
+      m->inv = 0;
+   }
+}
+
+
+void 
+_math_matrix_alloc_inv( GLmatrix *m )
+{
+   if ( m->inv == 0 ) {
+      m->inv = (GLfloat *) ALIGN_MALLOC( 16 * sizeof(GLfloat), 16 );
+      MEMCPY( m->inv, Identity, 16 * sizeof(GLfloat) );
+   }
+}
+
+
+void 
+_math_matrix_mul_matrix( GLmatrix *dest, const GLmatrix *a, const GLmatrix *b )
+{
+   dest->flags = (a->flags |
+		  b->flags |
+		  MAT_DIRTY_TYPE | 
+		  MAT_DIRTY_INVERSE);
+
+   if (TEST_MAT_FLAGS(dest, MAT_FLAGS_3D))
+      matmul34( dest->m, a->m, b->m );
+   else 
+      matmul4( dest->m, a->m, b->m );
+}
+
+
+void 
+_math_matrix_mul_floats( GLmatrix *dest, const GLfloat *m )
+{
+   dest->flags |= (MAT_FLAG_GENERAL |
+		   MAT_DIRTY_TYPE | 
+		   MAT_DIRTY_INVERSE);
+
+   matmul4( dest->m, dest->m, m );   
+}
+
+void 
+_math_matrix_set_identity( GLmatrix *mat )
+{
+   MEMCPY( mat->m, Identity, 16*sizeof(GLfloat) );
+
+   if (mat->inv)
+      MEMCPY( mat->inv, Identity, 16*sizeof(GLfloat) );
+
+   mat->type = MATRIX_IDENTITY;
+   mat->flags &= ~(MAT_DIRTY_FLAGS|
+		   MAT_DIRTY_TYPE|
+		   MAT_DIRTY_INVERSE);
+}
+
+
+
+void 
+_math_transposef( GLfloat to[16], const GLfloat from[16] )
+{
+   to[0] = from[0];
+   to[1] = from[4];
+   to[2] = from[8];
+   to[3] = from[12];
+   to[4] = from[1];
+   to[5] = from[5];
+   to[6] = from[9];
+   to[7] = from[13];
+   to[8] = from[2];
+   to[9] = from[6];
+   to[10] = from[10];
+   to[11] = from[14];
+   to[12] = from[3];
+   to[13] = from[7];
+   to[14] = from[11];
+   to[15] = from[15];
+}
+
+
+void 
+_math_transposed( GLdouble to[16], const GLdouble from[16] )
+{
+   to[0] = from[0];
+   to[1] = from[4];
+   to[2] = from[8];
+   to[3] = from[12];
+   to[4] = from[1];
+   to[5] = from[5];
+   to[6] = from[9];
+   to[7] = from[13];
+   to[8] = from[2];
+   to[9] = from[6];
+   to[10] = from[10];
+   to[11] = from[14];
+   to[12] = from[3];
+   to[13] = from[7];
+   to[14] = from[11];
+   to[15] = from[15];
+}
+
+void 
+_math_transposefd( GLfloat to[16], const GLdouble from[16] )
+{
+   to[0] = from[0];
+   to[1] = from[4];
+   to[2] = from[8];
+   to[3] = from[12];
+   to[4] = from[1];
+   to[5] = from[5];
+   to[6] = from[9];
+   to[7] = from[13];
+   to[8] = from[2];
+   to[9] = from[6];
+   to[10] = from[10];
+   to[11] = from[14];
+   to[12] = from[3];
+   to[13] = from[7];
+   to[14] = from[11];
+   to[15] = from[15];
+}
diff --git a/src/mesa/math/m_matrix.h b/src/mesa/math/m_matrix.h
new file mode 100644
index 00000000000..8eedbdb9424
--- /dev/null
+++ b/src/mesa/math/m_matrix.h
@@ -0,0 +1,176 @@
+/* $Id: m_matrix.h,v 1.1 2000/11/16 21:05:41 keithw Exp $ */
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  3.5
+ * 
+ * Copyright (C) 1999-2000  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"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ * 
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+
+#ifndef _M_MATRIX_H
+#define _M_MATRIX_H
+
+
+
+/* Give symbolic names to some of the entries in the matrix to help
+ * out with the rework of the viewport_map as a matrix transform.
+ */
+#define MAT_SX 0
+#define MAT_SY 5
+#define MAT_SZ 10
+#define MAT_TX 12
+#define MAT_TY 13
+#define MAT_TZ 14
+
+/*
+ * Different kinds of 4x4 transformation matrices:
+ */
+#define MATRIX_GENERAL		0	/* general 4x4 matrix */
+#define MATRIX_IDENTITY		1	/* identity matrix */
+#define MATRIX_3D_NO_ROT	2	/* ortho projection and others... */
+#define MATRIX_PERSPECTIVE	3	/* perspective projection matrix */
+#define MATRIX_2D		4	/* 2-D transformation */
+#define MATRIX_2D_NO_ROT	5	/* 2-D scale & translate only */
+#define MATRIX_3D		6	/* 3-D transformation */
+
+#define MAT_FLAG_IDENTITY       0
+#define MAT_FLAG_GENERAL        0x1
+#define MAT_FLAG_ROTATION       0x2
+#define MAT_FLAG_TRANSLATION    0x4
+#define MAT_FLAG_UNIFORM_SCALE  0x8
+#define MAT_FLAG_GENERAL_SCALE  0x10
+#define MAT_FLAG_GENERAL_3D     0x20
+#define MAT_FLAG_PERSPECTIVE    0x40
+#define MAT_FLAG_SINGULAR       0x80
+#define MAT_DIRTY_TYPE          0x100
+#define MAT_DIRTY_FLAGS         0x200
+#define MAT_DIRTY_INVERSE       0x400
+
+#define MAT_FLAGS_ANGLE_PRESERVING (MAT_FLAG_ROTATION | \
+				    MAT_FLAG_TRANSLATION | \
+				    MAT_FLAG_UNIFORM_SCALE)
+
+#define MAT_FLAGS_LENGTH_PRESERVING (MAT_FLAG_ROTATION | \
+				     MAT_FLAG_TRANSLATION)
+
+#define MAT_FLAGS_3D (MAT_FLAG_ROTATION | \
+		      MAT_FLAG_TRANSLATION | \
+		      MAT_FLAG_UNIFORM_SCALE | \
+		      MAT_FLAG_GENERAL_SCALE | \
+		      MAT_FLAG_GENERAL_3D)
+
+#define MAT_FLAGS_GEOMETRY (MAT_FLAG_GENERAL | \
+			    MAT_FLAG_ROTATION | \
+			    MAT_FLAG_TRANSLATION | \
+			    MAT_FLAG_UNIFORM_SCALE | \
+			    MAT_FLAG_GENERAL_SCALE | \
+			    MAT_FLAG_GENERAL_3D | \
+			    MAT_FLAG_PERSPECTIVE | \
+	                    MAT_FLAG_SINGULAR)
+
+#define MAT_DIRTY          (MAT_DIRTY_TYPE | \
+			    MAT_DIRTY_FLAGS | \
+			    MAT_DIRTY_INVERSE)
+
+#define TEST_MAT_FLAGS(mat, a)  \
+    ((MAT_FLAGS_GEOMETRY & (~(a)) & ((mat)->flags) ) == 0)
+
+
+typedef struct {
+   GLfloat *m;		/* 16-byte aligned */
+   GLfloat *inv;	/* optional, 16-byte aligned */
+   GLuint flags;
+   GLuint type;		/* one of the MATRIX_* values */
+} GLmatrix;
+
+
+
+
+extern void
+_math_matrix_ctr( GLmatrix *m );
+
+extern void
+_math_matrix_dtr( GLmatrix *m );
+
+extern void
+_math_matrix_alloc_inv( GLmatrix *m );
+
+extern void
+_math_matrix_mul_matrix( GLmatrix *dest, const GLmatrix *a, const GLmatrix *b );
+
+extern void
+_math_matrix_mul_floats( GLmatrix *dest, const GLfloat *b );
+
+extern void 
+_math_matrix_loadf( GLmatrix *mat, const GLfloat *m );
+
+extern void 
+_math_matrix_translate( GLmatrix *mat, GLfloat x, GLfloat y, GLfloat z );
+
+extern void
+_math_matrix_rotate( GLmatrix *m, GLfloat angle, 
+		     GLfloat x, GLfloat y, GLfloat z );
+
+extern void 
+_math_matrix_scale( GLmatrix *mat, GLfloat x, GLfloat y, GLfloat z );
+
+extern void
+_math_matrix_ortho( GLmatrix *mat, 	
+		    GLfloat left, GLfloat right,
+		    GLfloat bottom, GLfloat top, 
+		    GLfloat nearval, GLfloat farval );
+
+extern void
+_math_matrix_frustrum( GLmatrix *mat, 
+		       GLfloat left, GLfloat right,
+		       GLfloat bottom, GLfloat top, 
+		       GLfloat nearval, GLfloat farval );
+
+extern void
+_math_matrix_set_identity( GLmatrix *dest );
+
+extern void 
+_math_matrix_copy( GLmatrix *to, const GLmatrix *from );
+
+extern void
+_math_matrix_analyze( GLmatrix *mat );
+
+extern void
+_math_matrix_print( const GLmatrix *m );
+
+
+
+
+/* Related functions that don't actually operate on GLmatrix structs:
+ */
+extern void
+_math_transposef( GLfloat to[16], const GLfloat from[16] );
+
+extern void
+_math_transposed( GLdouble to[16], const GLdouble from[16] );
+
+extern void
+_math_transposefd( GLfloat to[16], const GLdouble from[16] );
+
+
+
+
+#endif
diff --git a/src/mesa/math/m_norm_tmp.h b/src/mesa/math/m_norm_tmp.h
new file mode 100644
index 00000000000..72770c2075e
--- /dev/null
+++ b/src/mesa/math/m_norm_tmp.h
@@ -0,0 +1,413 @@
+/* $Id: m_norm_tmp.h,v 1.1 2000/11/16 21:05:41 keithw Exp $ */
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  3.3
+ * 
+ * Copyright (C) 1999-2000  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"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ * 
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * New (3.1) transformation code written by Keith Whitwell.
+ */
+
+
+static void _XFORMAPI
+TAG(transform_normalize_normals)( const GLmatrix *mat,
+                                  GLfloat scale,
+                                  const GLvector3f *in,
+                                  const GLfloat *lengths,
+                                  const GLubyte mask[],
+                                  GLvector3f *dest )
+{
+   GLuint i;  
+   const GLfloat *from = in->start;
+   GLuint stride = in->stride;
+   GLuint count = in->count;
+   GLfloat (*out)[3] = (GLfloat (*)[3])dest->start;
+   GLfloat *m = mat->inv;
+   GLfloat m0 = m[0],  m4 = m[4],  m8 = m[8];
+   GLfloat m1 = m[1],  m5 = m[5],  m9 = m[9];
+   GLfloat m2 = m[2],  m6 = m[6],  m10 = m[10];
+
+   (void) mask;
+   if (!lengths) {
+      STRIDE_LOOP {
+	 CULL_CHECK {
+	    GLfloat tx, ty, tz;
+	    {
+	       const GLfloat ux = from[0],  uy = from[1],  uz = from[2];
+	       tx = ux * m0 + uy * m1 + uz * m2;
+	       ty = ux * m4 + uy * m5 + uz * m6;
+	       tz = ux * m8 + uy * m9 + uz * m10;
+	    }
+	    {
+	       GLdouble len = tx*tx + ty*ty + tz*tz;
+	       if (len > 1e-20) {
+		  GLdouble scale = 1.0 / GL_SQRT(len);
+		  out[i][0] = (GLfloat) (tx * scale);
+		  out[i][1] = (GLfloat) (ty * scale);
+		  out[i][2] = (GLfloat) (tz * scale);
+	       }
+	       else
+	       {
+		  out[i][0] = out[i][1] = out[i][2] = 0;
+	       }
+	    }
+	 }
+      }
+   }
+   else {
+      /* scale has been snapped to 1.0 if it is close.
+       */
+      if (scale != 1.0) {
+	 m0 *= scale,  m4 *= scale,  m8 *= scale;
+	 m1 *= scale,  m5 *= scale,  m9 *= scale;
+	 m2 *= scale,  m6 *= scale,  m10 *= scale;
+      }
+
+      STRIDE_LOOP {
+	 CULL_CHECK {
+	    GLfloat tx, ty, tz;
+	    {
+	       const GLfloat ux = from[0],  uy = from[1],  uz = from[2];
+	       tx = ux * m0 + uy * m1 + uz * m2;
+	       ty = ux * m4 + uy * m5 + uz * m6;
+	       tz = ux * m8 + uy * m9 + uz * m10;
+	    }
+	    {
+	       GLfloat len = lengths[i];
+	       out[i][0] = tx * len;
+	       out[i][1] = ty * len;
+	       out[i][2] = tz * len;
+	    }
+	 }
+      }
+   }
+   dest->count = in->count;
+}
+
+
+static void _XFORMAPI
+TAG(transform_normalize_normals_no_rot)( const GLmatrix *mat,
+                                         GLfloat scale,
+                                         const GLvector3f *in,
+                                         const GLfloat *lengths,
+                                         const GLubyte mask[],
+                                         GLvector3f *dest )
+{
+   GLuint i;  
+   const GLfloat *from = in->start;
+   GLuint stride = in->stride;
+   GLuint count = in->count;
+   GLfloat (*out)[3] = (GLfloat (*)[3])dest->start;
+   GLfloat *m = mat->inv;
+   GLfloat m0 = m[0];
+   GLfloat m5 = m[5];
+   GLfloat m10 = m[10];
+   (void) mask;
+   if (!lengths) {
+      STRIDE_LOOP {
+	 CULL_CHECK {
+	    GLfloat tx, ty, tz;
+	    {
+	       const GLfloat ux = from[0],  uy = from[1],  uz = from[2];
+	       tx = ux * m0                    ;
+	       ty =           uy * m5          ;
+	       tz =                     uz * m10;
+	    }
+	    {
+	       GLdouble len = tx*tx + ty*ty + tz*tz;
+	       if (len > 1e-20) {
+		  GLdouble scale = 1.0 / GL_SQRT(len);
+		  out[i][0] = (GLfloat) (tx * scale);
+		  out[i][1] = (GLfloat) (ty * scale);
+		  out[i][2] = (GLfloat) (tz * scale);
+	       }
+	       else
+	       {
+		  out[i][0] = out[i][1] = out[i][2] = 0;
+	       }
+	    }
+	 }
+      }
+   }
+   else {
+      /* scale has been snapped to 1.0 if it is close.
+       */
+      if (scale != 1.0) {
+	 m0 *= scale; 
+	 m5 *= scale;
+	 m10 *= scale;
+      }
+
+      STRIDE_LOOP {
+	 CULL_CHECK {
+	    GLfloat tx, ty, tz;
+	    {
+	       const GLfloat ux = from[0],  uy = from[1],  uz = from[2];
+	       tx = ux * m0                    ;
+	       ty =           uy * m5          ;
+	       tz =                     uz * m10;
+	    }
+	    {
+	       GLfloat len = lengths[i];
+	       out[i][0] = tx * len;
+	       out[i][1] = ty * len;
+	       out[i][2] = tz * len;
+	    }
+	 }
+      }
+   }
+   dest->count = in->count;
+}
+
+
+static void _XFORMAPI
+TAG(transform_rescale_normals_no_rot)( const GLmatrix *mat,
+                                       GLfloat scale,
+                                       const GLvector3f *in,
+                                       const GLfloat *lengths,
+                                       const GLubyte mask[],
+                                       GLvector3f *dest )
+{
+   GLuint i;  
+   const GLfloat *from = in->start;
+   GLuint stride = in->stride;
+   GLuint count = in->count;
+   GLfloat (*out)[3] = (GLfloat (*)[3])dest->start;
+   const GLfloat *m = mat->inv;
+   GLfloat m0 = scale*m[0];
+   GLfloat m5 = scale*m[5];
+   GLfloat m10 = scale*m[10];
+   (void) lengths;
+   (void) mask;
+   STRIDE_LOOP {
+      CULL_CHECK {
+	 GLfloat ux = from[0],  uy = from[1],  uz = from[2];
+	 out[i][0] = ux * m0;
+	 out[i][1] =           uy * m5;
+	 out[i][2] =                     uz * m10;
+      }
+   }
+   dest->count = in->count;
+}
+
+static void _XFORMAPI
+TAG(transform_rescale_normals)( const GLmatrix *mat,
+                                GLfloat scale,
+                                const GLvector3f *in,
+                                const GLfloat *lengths,
+                                const GLubyte mask[],
+                                GLvector3f *dest )
+{
+   GLuint i;  
+   const GLfloat *from = in->start;
+   GLuint stride = in->stride;
+   GLuint count = in->count;
+   GLfloat (*out)[3] = (GLfloat (*)[3])dest->start;
+   /* Since we are unlikely to have < 3 vertices in the buffer,
+    * it makes sense to pre-multiply by scale.
+    */
+   const GLfloat *m = mat->inv;
+   GLfloat m0 = scale*m[0],  m4 = scale*m[4],  m8 = scale*m[8];
+   GLfloat m1 = scale*m[1],  m5 = scale*m[5],  m9 = scale*m[9];
+   GLfloat m2 = scale*m[2],  m6 = scale*m[6],  m10 = scale*m[10];
+   (void) lengths;
+   (void) mask;
+   STRIDE_LOOP {
+      CULL_CHECK {
+	 GLfloat ux = from[0],  uy = from[1],  uz = from[2];
+	 out[i][0] = ux * m0 + uy * m1 + uz * m2;
+	 out[i][1] = ux * m4 + uy * m5 + uz * m6;
+	 out[i][2] = ux * m8 + uy * m9 + uz * m10;
+      }
+   }
+   dest->count = in->count;
+}
+
+
+static void _XFORMAPI
+TAG(transform_normals_no_rot)(const GLmatrix *mat,
+                              GLfloat scale,
+                              const GLvector3f *in,
+                              const GLfloat *lengths,
+                              const GLubyte mask[],
+                              GLvector3f *dest )
+{
+   GLuint i;  
+   const GLfloat *from = in->start;
+   GLuint stride = in->stride;
+   GLuint count = in->count;
+   GLfloat (*out)[3] = (GLfloat (*)[3])dest->start;
+   const GLfloat *m = mat->inv;
+   GLfloat m0 = m[0];
+   GLfloat m5 = m[5];
+   GLfloat m10 = m[10];
+   (void) scale;
+   (void) lengths;
+   (void) mask;
+   STRIDE_LOOP {
+      CULL_CHECK {
+	 GLfloat ux = from[0],  uy = from[1],  uz = from[2];
+	 out[i][0] = ux * m0;
+	 out[i][1] =           uy * m5;
+	 out[i][2] =                     uz * m10;
+      }
+   }
+   dest->count = in->count;
+}
+
+
+static void _XFORMAPI
+TAG(transform_normals)( const GLmatrix *mat,
+                        GLfloat scale,
+                        const GLvector3f *in,
+                        const GLfloat *lengths,
+                        const GLubyte mask[],
+                        GLvector3f *dest )
+{
+   GLuint i;  
+   const GLfloat *from = in->start;
+   GLuint stride = in->stride;
+   GLuint count = in->count;
+   GLfloat (*out)[3] = (GLfloat (*)[3])dest->start;
+   const GLfloat *m = mat->inv;
+   GLfloat m0 = m[0],  m4 = m[4],  m8 = m[8];
+   GLfloat m1 = m[1],  m5 = m[5],  m9 = m[9];
+   GLfloat m2 = m[2],  m6 = m[6],  m10 = m[10];
+   (void) scale;
+   (void) lengths;
+   (void) mask;
+   STRIDE_LOOP {
+      CULL_CHECK {
+	 GLfloat ux = from[0],  uy = from[1],  uz = from[2];
+	 out[i][0] = ux * m0 + uy * m1 + uz * m2;
+	 out[i][1] = ux * m4 + uy * m5 + uz * m6;
+	 out[i][2] = ux * m8 + uy * m9 + uz * m10;
+      }
+   }
+   dest->count = in->count;
+}
+
+
+static void _XFORMAPI
+TAG(normalize_normals)( const GLmatrix *mat,
+                        GLfloat scale,
+                        const GLvector3f *in,
+                        const GLfloat *lengths,
+                        const GLubyte mask[],
+                        GLvector3f *dest )
+{
+   GLuint i;  
+   const GLfloat *from = in->start;
+   GLuint stride = in->stride;
+   GLuint count = in->count;
+   GLfloat (*out)[3] = (GLfloat (*)[3])dest->start;
+   (void) mat;
+   (void) mask;
+   (void) scale;
+   if (lengths) {
+      STRIDE_LOOP {
+	 CULL_CHECK {
+	    const GLfloat x = from[0], y = from[1], z = from[2];
+	    GLfloat invlen = lengths[i];
+	    out[i][0] = x * invlen;
+	    out[i][1] = y * invlen;
+	    out[i][2] = z * invlen;
+	 }	 
+      }
+   }
+   else {
+      STRIDE_LOOP {
+	 CULL_CHECK {
+	    const GLfloat x = from[0], y = from[1], z = from[2];
+	    GLdouble len = x * x + y * y + z * z;
+	    if (len > 1e-50) {
+	       len = 1.0 / GL_SQRT(len);
+	       out[i][0] = (GLfloat) (x * len);
+	       out[i][1] = (GLfloat) (y * len);
+	       out[i][2] = (GLfloat) (z * len);
+	    }
+	    else {
+	       out[i][0] = x;
+	       out[i][1] = y;
+	       out[i][2] = z;
+	    }
+	 }	 
+      }
+   }
+   dest->count = in->count;
+}
+
+
+static void _XFORMAPI
+TAG(rescale_normals)( const GLmatrix *mat,
+                      GLfloat scale,
+                      const GLvector3f *in,
+                      const GLfloat *lengths,
+                      const GLubyte mask[],
+                      GLvector3f *dest )
+{
+   GLuint i;  
+   const GLfloat *from = in->start;
+   GLuint stride = in->stride;
+   GLuint count = in->count;
+   GLfloat (*out)[3] = (GLfloat (*)[3])dest->start;
+   (void) mat;
+   (void) lengths;
+   (void) mask;
+
+   STRIDE_LOOP {
+      CULL_CHECK {
+	 SCALE_SCALAR_3V( out[i], scale, from );
+      }
+   }
+   dest->count = in->count;
+}
+
+
+static void _XFORMAPI
+TAG(init_c_norm_transform)( void )
+{
+   gl_normal_tab[NORM_TRANSFORM_NO_ROT][IDX] = 
+      TAG(transform_normals_no_rot);
+
+   gl_normal_tab[NORM_TRANSFORM_NO_ROT | NORM_RESCALE][IDX] = 
+      TAG(transform_rescale_normals_no_rot);
+
+   gl_normal_tab[NORM_TRANSFORM_NO_ROT | NORM_NORMALIZE][IDX] = 
+      TAG(transform_normalize_normals_no_rot);
+
+   gl_normal_tab[NORM_TRANSFORM][IDX] = 
+      TAG(transform_normals);
+
+   gl_normal_tab[NORM_TRANSFORM | NORM_RESCALE][IDX] = 
+      TAG(transform_rescale_normals);
+
+   gl_normal_tab[NORM_TRANSFORM | NORM_NORMALIZE][IDX] = 
+      TAG(transform_normalize_normals);
+
+   gl_normal_tab[NORM_RESCALE][IDX] = 
+      TAG(rescale_normals);
+
+   gl_normal_tab[NORM_NORMALIZE][IDX] = 
+      TAG(normalize_normals);
+}
diff --git a/src/mesa/math/m_trans_tmp.h b/src/mesa/math/m_trans_tmp.h
new file mode 100644
index 00000000000..952fde55942
--- /dev/null
+++ b/src/mesa/math/m_trans_tmp.h
@@ -0,0 +1,210 @@
+/* $Id: m_trans_tmp.h,v 1.1 2000/11/16 21:05:41 keithw Exp $ */
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  3.1
+ * 
+ * Copyright (C) 1999  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"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ * 
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * New (3.1) transformation code written by Keith Whitwell.
+ */
+
+
+/* KW: This file also included by tnl/trans_elt.c to build code
+ *     specific to the implementation of array-elements in the
+ *     tnl module.
+ */
+
+
+#ifdef DEST_4F
+static void DEST_4F( GLfloat (*t)[4],
+		     CONST void *ptr,
+		     GLuint stride,
+		     ARGS)
+{
+   const GLubyte *f = (GLubyte *) ptr + SRC_START * stride;
+   const GLubyte *first = f;
+   GLuint i;
+
+   (void) first;
+   (void) start;
+   for (i = DST_START ; i < n ; i++, NEXT_F) {
+      CHECK { 
+         NEXT_F2;
+	 if (SZ >= 1) t[i][0] = TRX_4F(f, 0);
+	 if (SZ >= 2) t[i][1] = TRX_4F(f, 1);
+	 if (SZ >= 3) t[i][2] = TRX_4F(f, 2);
+	 if (SZ == 4) t[i][3] = TRX_4F(f, 3);
+      }
+   }
+}
+#endif
+
+
+#ifdef DEST_3F
+static void DEST_3F( GLfloat (*t)[3],
+		     CONST void *ptr,
+		     GLuint stride,
+		     ARGS)
+{
+   const GLubyte *f = (GLubyte *) ptr + SRC_START * stride;
+   const GLubyte *first = f;
+   GLuint i;
+   (void) first;
+   (void) start;
+   for (i = DST_START ; i < n ; i++, NEXT_F) {
+      CHECK {
+         NEXT_F2;
+	 t[i][0] = TRX_3F(f, 0);
+	 t[i][1] = TRX_3F(f, 1);
+	 t[i][2] = TRX_3F(f, 2);
+      }
+   }
+}
+#endif
+
+#ifdef DEST_1F
+static void DEST_1F( GLfloat *t,
+		     CONST void *ptr,
+		     GLuint stride,
+		     ARGS)
+{
+   const GLubyte *f = (GLubyte *) ptr + SRC_START * stride;
+   const GLubyte *first = f;
+   GLuint i;
+   (void) first;
+   (void) start;
+   for (i = DST_START ; i < n ; i++, NEXT_F) {
+      CHECK {
+         NEXT_F2;
+	 t[i] = TRX_1F(f, 0);
+      }
+   }
+}
+#endif
+
+#ifdef DEST_4UB
+static void DEST_4UB( GLubyte (*t)[4],
+		      CONST void *ptr,
+		      GLuint stride,
+		      ARGS)
+{
+   const GLubyte *f = (GLubyte *) ptr + SRC_START * stride;
+   const GLubyte *first = f;
+   GLuint i;
+   (void) start;
+   (void) first;
+   for (i = DST_START ; i < n ; i++, NEXT_F) {
+      CHECK {
+         NEXT_F2;
+	 if (SZ >= 1) TRX_UB(t[i][0], f, 0);
+	 if (SZ >= 2) TRX_UB(t[i][1], f, 1);
+	 if (SZ >= 3) TRX_UB(t[i][2], f, 2); 
+	 if (SZ == 4) TRX_UB(t[i][3], f, 3); else t[i][3] = 255;
+      }
+   }
+}
+#endif
+
+
+#ifdef DEST_1UB
+static void DEST_1UB( GLubyte *t,
+		      CONST void *ptr,
+		      GLuint stride,
+		      ARGS)
+{
+   const GLubyte *f = (GLubyte *) ptr + SRC_START * stride;
+   const GLubyte *first = f;
+   GLuint i;
+   (void) start;
+   (void) first;
+   for (i = DST_START ; i < n ; i++, NEXT_F) {
+      CHECK {
+         NEXT_F2;
+	  TRX_UB(t[i], f, 0);
+      }
+   }
+}
+#endif
+
+
+#ifdef DEST_1UI
+static void DEST_1UI( GLuint *t,
+		      CONST void *ptr,
+		      GLuint stride,
+		      ARGS)
+{
+   const GLubyte *f = (GLubyte *) ptr + SRC_START * stride;
+   const GLubyte *first = f;
+   GLuint i;
+   (void) start;
+   (void) first;
+
+   for (i = DST_START ; i < n ; i++, NEXT_F) {
+      CHECK {
+         NEXT_F2;
+	 t[i] = TRX_UI(f, 0);
+      }
+   }
+}
+#endif
+
+
+static void INIT(void)
+{
+#ifdef DEST_1UI
+   ASSERT(SZ == 1);
+   TAB(_1ui)[SRC_IDX] = DEST_1UI;
+#endif
+#ifdef DEST_1UB
+   ASSERT(SZ == 1);
+   TAB(_1ub)[SRC_IDX] = DEST_1UB;
+#endif
+#ifdef DEST_1F
+   ASSERT(SZ == 1);
+   TAB(_1f)[SRC_IDX] = DEST_1F;
+#endif
+#ifdef DEST_3F
+   ASSERT(SZ == 3);
+   TAB(_3f)[SRC_IDX] = DEST_3F;
+#endif
+#ifdef DEST_4UB
+   TAB(_4ub)[SZ][SRC_IDX] = DEST_4UB;
+#endif
+#ifdef DEST_4F
+   TAB(_4f)[SZ][SRC_IDX] = DEST_4F;
+#endif
+   
+}
+
+
+#undef INIT
+#undef DEST_1UI
+#undef DEST_1UB
+#undef DEST_4UB
+#undef DEST_3F
+#undef DEST_4F
+#undef DEST_1F
+#undef SZ
+#undef TAG
+
+
diff --git a/src/mesa/math/m_translate.c b/src/mesa/math/m_translate.c
new file mode 100644
index 00000000000..945e35706c4
--- /dev/null
+++ b/src/mesa/math/m_translate.c
@@ -0,0 +1,478 @@
+/* $Id: m_translate.c,v 1.1 2000/11/16 21:05:41 keithw Exp $ */
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  3.3
+ * 
+ * Copyright (C) 1999  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"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ * 
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * New (3.1) transformation code written by Keith Whitwell.
+ */
+
+
+#include "glheader.h"
+#include "colormac.h"
+#include "mem.h"
+#include "mmath.h"
+
+#include "m_translate.h"
+
+/* This macro is used on other systems, so undefine it for this module */
+
+#undef	CHECK
+
+trans_1f_func gl_trans_1f_tab[MAX_TYPES];
+trans_1ui_func gl_trans_1ui_tab[MAX_TYPES];
+trans_1ub_func gl_trans_1ub_tab[MAX_TYPES];
+trans_3f_func  gl_trans_3f_tab[MAX_TYPES];
+trans_4ub_func gl_trans_4ub_tab[5][MAX_TYPES];
+trans_4f_func  gl_trans_4f_tab[5][MAX_TYPES];
+
+
+#define PTR_ELT(ptr, elt) (((SRC *)ptr)[elt])
+
+
+#define TAB(x) gl_trans##x##_tab
+#define ARGS   GLuint start, GLuint n
+#define SRC_START  start
+#define DST_START  0
+#define STRIDE stride
+#define NEXT_F f += stride
+#define NEXT_F2
+#define CHECK
+
+
+
+
+/* GL_BYTE
+ */
+#define SRC GLbyte
+#define SRC_IDX TYPE_IDX(GL_BYTE)
+#define TRX_3F(f,n)   BYTE_TO_FLOAT( PTR_ELT(f,n) )
+#define TRX_4F(f,n)   (GLfloat)( PTR_ELT(f,n) )
+#define TRX_UB(ub, f,n)  ub = BYTE_TO_UBYTE( PTR_ELT(f,n) )
+#define TRX_UI(f,n)  (PTR_ELT(f,n) < 0 ? 0 : (GLuint)  PTR_ELT(f,n))
+
+
+#define SZ 4
+#define INIT init_trans_4_GLbyte_raw
+#define DEST_4F trans_4_GLbyte_4f_raw
+#define DEST_4UB trans_4_GLbyte_4ub_raw
+#include "m_trans_tmp.h"
+
+#define SZ 3
+#define INIT init_trans_3_GLbyte_raw
+#define DEST_4F trans_3_GLbyte_4f_raw
+#define DEST_4UB trans_3_GLbyte_4ub_raw
+#define DEST_3F trans_3_GLbyte_3f_raw
+#include "m_trans_tmp.h"
+
+#define SZ 2
+#define INIT init_trans_2_GLbyte_raw
+#define DEST_4F trans_2_GLbyte_4f_raw
+#include "m_trans_tmp.h"
+
+#define SZ 1
+#define INIT init_trans_1_GLbyte_raw
+#define DEST_4F trans_1_GLbyte_4f_raw
+#define DEST_1UB trans_1_GLbyte_1ub_raw
+#define DEST_1UI trans_1_GLbyte_1ui_raw
+#include "m_trans_tmp.h"
+
+#undef SRC
+#undef TRX_3F
+#undef TRX_4F
+#undef TRX_UB
+#undef TRX_UI
+#undef SRC_IDX
+
+/* GL_UNSIGNED_BYTE
+ */
+#define SRC GLubyte
+#define SRC_IDX TYPE_IDX(GL_UNSIGNED_BYTE)
+#define TRX_3F(f,n)		/* unused */
+#define TRX_4F(f,n)		/* unused */
+#define TRX_UB(ub, f,n)	     ub = PTR_ELT(f,n)
+#define TRX_UI(f,n)          (GLuint)PTR_ELT(f,n)
+
+/* 4ub->4ub handled in special case below.
+ */
+
+#define SZ 3
+#define INIT init_trans_3_GLubyte_raw
+#define DEST_4UB trans_3_GLubyte_4ub_raw
+#include "m_trans_tmp.h"
+
+
+#define SZ 1
+#define INIT init_trans_1_GLubyte_raw
+#define DEST_1UI trans_1_GLubyte_1ui_raw
+#define DEST_1UB trans_1_GLubyte_1ub_raw
+#include "m_trans_tmp.h"
+
+#undef SRC
+#undef SRC_IDX
+#undef TRX_3F
+#undef TRX_4F
+#undef TRX_UB
+#undef TRX_UI
+
+
+/* GL_SHORT
+ */
+#define SRC GLshort
+#define SRC_IDX TYPE_IDX(GL_SHORT)
+#define TRX_3F(f,n)   SHORT_TO_FLOAT( PTR_ELT(f,n) )
+#define TRX_4F(f,n)   (GLfloat)( PTR_ELT(f,n) )
+#define TRX_UB(ub, f,n)  ub = SHORT_TO_UBYTE(PTR_ELT(f,n))
+#define TRX_UI(f,n)  (PTR_ELT(f,n) < 0 ? 0 : (GLuint)  PTR_ELT(f,n))
+
+
+#define SZ  4
+#define INIT init_trans_4_GLshort_raw
+#define DEST_4F trans_4_GLshort_4f_raw
+#define DEST_4UB trans_4_GLshort_4ub_raw
+#include "m_trans_tmp.h"
+
+#define SZ 3
+#define INIT init_trans_3_GLshort_raw
+#define DEST_4F trans_3_GLshort_4f_raw
+#define DEST_4UB trans_3_GLshort_4ub_raw
+#define DEST_3F trans_3_GLshort_3f_raw
+#include "m_trans_tmp.h"
+
+#define SZ 2
+#define INIT init_trans_2_GLshort_raw
+#define DEST_4F trans_2_GLshort_4f_raw
+#include "m_trans_tmp.h"
+
+#define SZ 1
+#define INIT init_trans_1_GLshort_raw
+#define DEST_4F trans_1_GLshort_4f_raw
+#define DEST_1UB trans_1_GLshort_1ub_raw
+#define DEST_1UI trans_1_GLshort_1ui_raw
+#include "m_trans_tmp.h"
+
+
+#undef SRC
+#undef SRC_IDX
+#undef TRX_3F
+#undef TRX_4F
+#undef TRX_UB
+#undef TRX_UI
+
+
+/* GL_UNSIGNED_SHORT
+ */
+#define SRC GLushort
+#define SRC_IDX TYPE_IDX(GL_UNSIGNED_SHORT)
+#define TRX_3F(f,n)   USHORT_TO_FLOAT( PTR_ELT(f,n) )
+#define TRX_4F(f,n)   (GLfloat)( PTR_ELT(f,n) )
+#define TRX_UB(ub,f,n)  ub = (GLubyte) (PTR_ELT(f,n) >> 8)
+#define TRX_UI(f,n)  (GLuint)   PTR_ELT(f,n)
+
+
+#define SZ 4
+#define INIT init_trans_4_GLushort_raw
+#define DEST_4F trans_4_GLushort_4f_raw
+#define DEST_4UB trans_4_GLushort_4ub_raw
+#include "m_trans_tmp.h"
+
+#define SZ 3
+#define INIT init_trans_3_GLushort_raw
+#define DEST_4F trans_3_GLushort_4f_raw
+#define DEST_4UB trans_3_GLushort_4ub_raw
+#define DEST_3F trans_3_GLushort_3f_raw
+#include "m_trans_tmp.h"
+
+#define SZ 2
+#define INIT init_trans_2_GLushort_raw
+#define DEST_4F trans_2_GLushort_4f_raw
+#include "m_trans_tmp.h"
+
+#define SZ 1
+#define INIT init_trans_1_GLushort_raw
+#define DEST_4F trans_1_GLushort_4f_raw
+#define DEST_1UB trans_1_GLushort_1ub_raw
+#define DEST_1UI trans_1_GLushort_1ui_raw
+#include "m_trans_tmp.h"
+
+#undef SRC
+#undef SRC_IDX
+#undef TRX_3F
+#undef TRX_4F
+#undef TRX_UB
+#undef TRX_UI
+
+
+/* GL_INT
+ */
+#define SRC GLint
+#define SRC_IDX TYPE_IDX(GL_INT)
+#define TRX_3F(f,n)   INT_TO_FLOAT( PTR_ELT(f,n) )
+#define TRX_4F(f,n)   (GLfloat)( PTR_ELT(f,n) )
+#define TRX_UB(ub, f,n)  ub = INT_TO_UBYTE(PTR_ELT(f,n))
+#define TRX_UI(f,n)  (PTR_ELT(f,n) < 0 ? 0 : (GLuint)  PTR_ELT(f,n))
+
+
+#define SZ 4
+#define INIT init_trans_4_GLint_raw
+#define DEST_4F trans_4_GLint_4f_raw
+#define DEST_4UB trans_4_GLint_4ub_raw
+#include "m_trans_tmp.h"
+
+#define SZ 3
+#define INIT init_trans_3_GLint_raw
+#define DEST_4F trans_3_GLint_4f_raw
+#define DEST_4UB trans_3_GLint_4ub_raw
+#define DEST_3F trans_3_GLint_3f_raw
+#include "m_trans_tmp.h"
+
+#define SZ 2
+#define INIT init_trans_2_GLint_raw
+#define DEST_4F trans_2_GLint_4f_raw
+#include "m_trans_tmp.h"
+
+#define SZ 1
+#define INIT init_trans_1_GLint_raw
+#define DEST_4F trans_1_GLint_4f_raw
+#define DEST_1UB trans_1_GLint_1ub_raw
+#define DEST_1UI trans_1_GLint_1ui_raw
+#include "m_trans_tmp.h"
+
+
+#undef SRC
+#undef SRC_IDX
+#undef TRX_3F
+#undef TRX_4F
+#undef TRX_UB
+#undef TRX_UI
+
+
+/* GL_UNSIGNED_INT
+ */
+#define SRC GLuint
+#define SRC_IDX TYPE_IDX(GL_UNSIGNED_INT)
+#define TRX_3F(f,n)   INT_TO_FLOAT( PTR_ELT(f,n) )
+#define TRX_4F(f,n)   (GLfloat)( PTR_ELT(f,n) )
+#define TRX_UB(ub, f,n)  ub = (GLubyte) (PTR_ELT(f,n) >> 24)
+#define TRX_UI(f,n)		PTR_ELT(f,n)
+
+
+#define SZ 4
+#define INIT init_trans_4_GLuint_raw
+#define DEST_4F trans_4_GLuint_4f_raw
+#define DEST_4UB trans_4_GLuint_4ub_raw
+#include "m_trans_tmp.h"
+
+#define SZ 3
+#define INIT init_trans_3_GLuint_raw
+#define DEST_4F trans_3_GLuint_4f_raw
+#define DEST_4UB trans_3_GLuint_4ub_raw
+#define DEST_3F trans_3_GLuint_3f_raw
+#include "m_trans_tmp.h"
+
+#define SZ 2
+#define INIT init_trans_2_GLuint_raw
+#define DEST_4F trans_2_GLuint_4f_raw
+#include "m_trans_tmp.h"
+
+#define SZ 1
+#define INIT init_trans_1_GLuint_raw
+#define DEST_4F trans_1_GLuint_4f_raw
+#define DEST_1UB trans_1_GLuint_1ub_raw
+#define DEST_1UI trans_1_GLuint_1ui_raw
+#include "m_trans_tmp.h"
+
+#undef SRC
+#undef SRC_IDX
+#undef TRX_3F
+#undef TRX_4F
+#undef TRX_UB
+#undef TRX_UI
+
+
+/* GL_DOUBLE
+ */
+#define SRC GLdouble
+#define SRC_IDX TYPE_IDX(GL_DOUBLE)
+#define TRX_3F(f,n)   PTR_ELT(f,n)
+#define TRX_4F(f,n)   PTR_ELT(f,n)
+#define TRX_UB(ub,f,n) FLOAT_COLOR_TO_CHAN(ub, PTR_ELT(f,n))
+#define TRX_UI(f,n)  (GLuint) (GLint) PTR_ELT(f,n)
+#define TRX_1F(f,n)   PTR_ELT(f,n)
+
+
+#define SZ 4
+#define INIT init_trans_4_GLdouble_raw
+#define DEST_4F trans_4_GLdouble_4f_raw
+#define DEST_4UB trans_4_GLdouble_4ub_raw
+#include "m_trans_tmp.h"
+
+#define SZ 3
+#define INIT init_trans_3_GLdouble_raw
+#define DEST_4F trans_3_GLdouble_4f_raw
+#define DEST_4UB trans_3_GLdouble_4ub_raw
+#define DEST_3F trans_3_GLdouble_3f_raw
+#include "m_trans_tmp.h"
+
+#define SZ 2
+#define INIT init_trans_2_GLdouble_raw
+#define DEST_4F trans_2_GLdouble_4f_raw
+#include "m_trans_tmp.h"
+
+#define SZ 1
+#define INIT init_trans_1_GLdouble_raw
+#define DEST_4F trans_1_GLdouble_4f_raw
+#define DEST_1UB trans_1_GLdouble_1ub_raw
+#define DEST_1UI trans_1_GLdouble_1ui_raw
+#define DEST_1F trans_1_GLdouble_1f_raw
+#include "m_trans_tmp.h"
+
+#undef SRC
+#undef SRC_IDX
+
+/* GL_FLOAT
+ */
+#define SRC GLfloat
+#define SRC_IDX TYPE_IDX(GL_FLOAT)
+#define SZ 4
+#define INIT init_trans_4_GLfloat_raw 
+#define DEST_4UB trans_4_GLfloat_4ub_raw 
+#define DEST_4F  trans_4_GLfloat_4f_raw
+#include "m_trans_tmp.h"
+
+#define SZ 3
+#define INIT init_trans_3_GLfloat_raw
+#define DEST_4F  trans_3_GLfloat_4f_raw
+#define DEST_4UB trans_3_GLfloat_4ub_raw
+#define DEST_3F trans_3_GLfloat_3f_raw
+#include "m_trans_tmp.h"
+
+#define SZ 2
+#define INIT init_trans_2_GLfloat_raw
+#define DEST_4F trans_2_GLfloat_4f_raw
+#include "m_trans_tmp.h"
+
+#define SZ 1
+#define INIT init_trans_1_GLfloat_raw
+#define DEST_4F  trans_1_GLfloat_4f_raw
+#define DEST_1UB trans_1_GLfloat_1ub_raw
+#define DEST_1UI trans_1_GLfloat_1ui_raw
+#define DEST_1F trans_1_GLfloat_1f_raw
+
+#include "m_trans_tmp.h"
+
+#undef SRC
+#undef SRC_IDX
+#undef TRX_3F
+#undef TRX_4F
+#undef TRX_UB
+#undef TRX_UI
+
+
+static void trans_4_GLubyte_4ub_raw (GLubyte (*t)[4],
+				     CONST void *Ptr,
+				     GLuint stride,
+				     ARGS )
+{
+   const GLubyte *f = (GLubyte *) Ptr + SRC_START * stride;
+   GLuint i;
+
+   if (((((long) f | (long) stride)) & 3L) == 0L) { 
+      /* Aligned.
+       */
+      for (i = DST_START ; i < n ; i++, f += stride) {
+	 COPY_4UBV( t[i], f );
+      }
+   } else {
+      for (i = DST_START ; i < n ; i++, f += stride) {
+	 t[i][0] = f[0];
+	 t[i][1] = f[1];
+	 t[i][2] = f[2];
+	 t[i][3] = f[3];
+      }
+   }
+}
+
+
+static void init_translate_raw(void)
+{
+   MEMSET( TAB(_1ui), 0, sizeof(TAB(_1ui)) );
+   MEMSET( TAB(_1ub), 0, sizeof(TAB(_1ub)) );
+   MEMSET( TAB(_3f),  0, sizeof(TAB(_3f)) );
+   MEMSET( TAB(_4ub), 0, sizeof(TAB(_4ub)) );
+   MEMSET( TAB(_4f),  0, sizeof(TAB(_4f)) );
+
+   TAB(_4ub)[4][TYPE_IDX(GL_UNSIGNED_BYTE)] = trans_4_GLubyte_4ub_raw;
+
+   init_trans_4_GLbyte_raw();
+   init_trans_3_GLbyte_raw();
+   init_trans_2_GLbyte_raw();
+   init_trans_1_GLbyte_raw();
+   init_trans_1_GLubyte_raw();
+   init_trans_3_GLubyte_raw();
+   init_trans_4_GLshort_raw();
+   init_trans_3_GLshort_raw();
+   init_trans_2_GLshort_raw();
+   init_trans_1_GLshort_raw();
+   init_trans_4_GLushort_raw();
+   init_trans_3_GLushort_raw();
+   init_trans_2_GLushort_raw();
+   init_trans_1_GLushort_raw();
+   init_trans_4_GLint_raw();
+   init_trans_3_GLint_raw();
+   init_trans_2_GLint_raw();
+   init_trans_1_GLint_raw();
+   init_trans_4_GLuint_raw();
+   init_trans_3_GLuint_raw();
+   init_trans_2_GLuint_raw();
+   init_trans_1_GLuint_raw();
+   init_trans_4_GLdouble_raw();
+   init_trans_3_GLdouble_raw();
+   init_trans_2_GLdouble_raw();
+   init_trans_1_GLdouble_raw();
+   init_trans_4_GLfloat_raw();
+   init_trans_3_GLfloat_raw();
+   init_trans_2_GLfloat_raw();
+   init_trans_1_GLfloat_raw();
+}
+
+
+#undef TAB
+#undef CLASS
+#undef ARGS
+#undef CHECK
+#undef SRC_START
+#undef DST_START
+#undef NEXT_F
+#undef NEXT_F2
+
+
+
+
+
+void 
+_math_init_translate( void )
+{
+   init_translate_raw();
+}
diff --git a/src/mesa/math/m_translate.h b/src/mesa/math/m_translate.h
new file mode 100644
index 00000000000..5c60b972ff5
--- /dev/null
+++ b/src/mesa/math/m_translate.h
@@ -0,0 +1,92 @@
+/* $Id: m_translate.h,v 1.1 2000/11/16 21:05:41 keithw Exp $ */
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  3.1
+ * 
+ * Copyright (C) 1999  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"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ * 
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+
+#ifndef _M_TRANSLATE_H_
+#define _M_TRANSLATE_H_
+
+
+typedef void (*trans_1f_func)(GLfloat *to,
+			      CONST void *ptr,
+			      GLuint stride,
+			      GLuint start, 
+			      GLuint n );
+
+typedef void (*trans_1ui_func)(GLuint *to,
+			       CONST void *ptr,
+			       GLuint stride,
+			       GLuint start, 
+			       GLuint n );
+
+typedef void (*trans_1ub_func)(GLubyte *to,
+			       CONST void *ptr,
+			       GLuint stride,
+			       GLuint start,
+			       GLuint n );
+
+typedef void (*trans_4ub_func)(GLubyte (*to)[4],
+			       CONST void *ptr,
+			       GLuint stride,
+			       GLuint start,
+			       GLuint n );
+
+typedef void (*trans_4f_func)(GLfloat (*to)[4],
+			      CONST void *ptr,
+			      GLuint stride,
+			      GLuint start, 
+			      GLuint n );
+
+typedef void (*trans_3f_func)(GLfloat (*to)[3],
+			      CONST void *ptr,
+			      GLuint stride,
+			      GLuint start, 
+			      GLuint n );
+
+
+
+
+/* Translate GL_UNSIGNED_BYTE, etc to the indexes used in the arrays
+ * below.
+ */
+#define TYPE_IDX(t) ((t) & 0xf)
+
+#define MAX_TYPES TYPE_IDX(GL_DOUBLE)+1      /* 0xa + 1 */
+
+/* Only useful combinations are defined, thus there is no function to
+ * translate eg, ubyte->float or ubyte->ubyte, which are never used.  
+ */
+extern trans_1f_func gl_trans_1f_tab[MAX_TYPES];
+extern trans_1ui_func gl_trans_1ui_tab[MAX_TYPES];
+extern trans_1ub_func gl_trans_1ub_tab[MAX_TYPES];
+extern trans_3f_func  gl_trans_3f_tab[MAX_TYPES];
+extern trans_4ub_func gl_trans_4ub_tab[5][MAX_TYPES];
+extern trans_4f_func  gl_trans_4f_tab[5][MAX_TYPES];
+
+
+extern void gl_init_translate( void );
+
+
+#endif
diff --git a/src/mesa/math/m_vector.c b/src/mesa/math/m_vector.c
new file mode 100644
index 00000000000..4dbf68f8e0f
--- /dev/null
+++ b/src/mesa/math/m_vector.c
@@ -0,0 +1,367 @@
+/* $Id: m_vector.c,v 1.1 2000/11/16 21:05:41 keithw Exp $ */
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  3.5
+ * 
+ * Copyright (C) 1999-2000  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"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ * 
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * New (3.1) transformation code written by Keith Whitwell.
+ */
+
+
+#include "glheader.h"
+#include "macros.h"
+#include "mem.h"
+
+#include "m_vector.h"
+
+
+
+/*
+ * Given a vector [count][4] of floats, set all the [][elt] values
+ * to 0 (if elt = 0, 1, 2) or 1.0 (if elt = 3).
+ */
+void gl_vector4f_clean_elem( GLvector4f *vec, GLuint count, GLuint elt )
+{
+   static const GLubyte elem_bits[4] = {
+      VEC_DIRTY_0,
+      VEC_DIRTY_1, 
+      VEC_DIRTY_2, 
+      VEC_DIRTY_3
+   };
+   static const GLfloat clean[4] = { 0, 0, 0, 1 };
+   const GLfloat v = clean[elt];
+   GLfloat (*data)[4] = (GLfloat (*)[4])vec->start;
+   GLuint i; 
+
+   for (i = 0 ; i < count ; i++)
+      data[i][elt] = v;
+
+   vec->flags &= ~elem_bits[elt];
+}
+
+static const GLubyte size_bits[5] = {
+   0,
+   VEC_SIZE_1,
+   VEC_SIZE_2,
+   VEC_SIZE_3,
+   VEC_SIZE_4,
+};
+
+
+
+/*
+ * Initialize GLvector objects.
+ * Input: v - the vector object to initialize.
+ *        flags - bitwise-OR of VEC_* flags
+ *        storage - pointer to storage for the vector's data
+ */
+
+
+void gl_vector4f_init( GLvector4f *v, GLuint flags, GLfloat (*storage)[4] )
+{
+   v->stride = 4 * sizeof(GLfloat);
+   v->size = 2;   /* may change: 2-4 for vertices and 1-4 for texcoords */
+   v->data = storage;
+   v->start = (GLfloat *) storage;
+   v->count = 0;
+   v->flags = size_bits[4] | flags | VEC_GOOD_STRIDE;
+}
+
+void gl_vector3f_init( GLvector3f *v, GLuint flags, GLfloat (*storage)[3] )
+{
+   v->stride = 3 * sizeof(GLfloat);
+   v->data = storage;
+   v->start = (GLfloat *) storage;
+   v->count = 0;
+   v->flags = flags | VEC_GOOD_STRIDE;
+}
+
+void gl_vector1f_init( GLvector1f *v, GLuint flags, GLfloat *storage )
+{
+   v->stride = 1*sizeof(GLfloat);
+   v->data = storage;
+   v->start = (GLfloat *)storage;
+   v->count = 0;
+   v->flags = flags | VEC_GOOD_STRIDE;
+}
+
+void gl_vector4ub_init( GLvector4ub *v, GLuint flags, GLubyte (*storage)[4] )
+{
+   v->stride = 4 * sizeof(GLubyte);
+   v->data = storage;
+   v->start = (GLubyte *) storage;
+   v->count = 0;
+   v->flags = flags | VEC_GOOD_STRIDE;
+}
+
+void gl_vector1ub_init( GLvector1ub *v, GLuint flags, GLubyte *storage )
+{
+   v->stride = 1 * sizeof(GLubyte);
+   v->data = storage;
+   v->start = (GLubyte *) storage;
+   v->count = 0;
+   v->flags = flags | VEC_GOOD_STRIDE;
+}
+
+void gl_vector1ui_init( GLvector1ui *v, GLuint flags, GLuint *storage )
+{
+   v->stride = 1 * sizeof(GLuint);
+   v->data = storage;
+   v->start = (GLuint *) storage;
+   v->count = 0;
+   v->flags = flags | VEC_GOOD_STRIDE;
+}
+
+
+/*
+ * Initialize GLvector objects and allocate storage.
+ * Input: v - the vector object
+ *        sz - unused????
+ *        flags - bitwise-OR of VEC_* flags
+ *        count - number of elements to allocate in vector
+ *        alignment - desired memory alignment for the data (in bytes)
+ */
+
+
+void gl_vector4f_alloc( GLvector4f *v, GLuint flags, GLuint count,
+			GLuint alignment )
+{
+   v->stride = 4 * sizeof(GLfloat);
+   v->size = 2;
+   v->storage = ALIGN_MALLOC( count * 4 * sizeof(GLfloat), alignment );
+   v->start = (GLfloat *) v->storage;
+   v->data = (GLfloat (*)[4]) v->storage;
+   v->count = 0;
+   v->flags = size_bits[4] | flags | VEC_MALLOC | VEC_GOOD_STRIDE;
+}
+
+void gl_vector3f_alloc( GLvector3f *v, GLuint flags, GLuint count,
+			GLuint alignment )
+{
+   v->stride = 3 * sizeof(GLfloat);
+   v->storage = ALIGN_MALLOC( count * 3 * sizeof(GLfloat), alignment );
+   v->start = (GLfloat *) v->storage;
+   v->data = (GLfloat (*)[3]) v->storage;
+   v->count = 0;
+   v->flags = flags | VEC_MALLOC | VEC_GOOD_STRIDE;
+}
+
+void gl_vector1f_alloc( GLvector1f *v, GLuint flags, GLuint count,
+			GLuint alignment )
+{
+   v->stride = sizeof(GLfloat);
+   v->storage = v->start = (GLfloat *)
+      ALIGN_MALLOC( count * sizeof(GLfloat), alignment );
+   v->data = v->start;
+   v->count = 0;
+   v->flags = flags | VEC_MALLOC | VEC_GOOD_STRIDE;
+}
+
+void gl_vector4ub_alloc( GLvector4ub *v, GLuint flags, GLuint count,
+			 GLuint alignment )
+{
+   v->stride = 4 * sizeof(GLubyte);
+   v->storage = ALIGN_MALLOC( count * 4 * sizeof(GLubyte), alignment );
+   v->start = (GLubyte *) v->storage;
+   v->data = (GLubyte (*)[4]) v->storage;
+   v->count = 0;
+   v->flags = flags | VEC_MALLOC | VEC_GOOD_STRIDE;
+}
+
+void gl_vector1ub_alloc( GLvector1ub *v, GLuint flags, GLuint count,
+			 GLuint alignment )
+{
+   v->stride = 1 * sizeof(GLubyte);
+   v->storage = ALIGN_MALLOC( count * sizeof(GLubyte), alignment );
+   v->start = (GLubyte *) v->storage;      
+   v->data = (GLubyte *) v->storage;
+   v->count = 0;
+   v->flags = flags | VEC_MALLOC | VEC_GOOD_STRIDE;
+}
+
+void gl_vector1ui_alloc( GLvector1ui *v, GLuint flags, GLuint count,
+			 GLuint alignment )
+{
+   v->stride = 1 * sizeof(GLuint);
+   v->storage = ALIGN_MALLOC( count * sizeof(GLuint), alignment );
+   v->start = (GLuint *) v->storage;      
+   v->data = (GLuint *) v->storage;
+   v->count = 0;
+   v->flags = flags | VEC_MALLOC | VEC_GOOD_STRIDE;
+}
+
+
+
+/*
+ * Vector deallocation.  Free whatever memory is pointed to by the
+ * vector's storage field if the VEC_MALLOC flag is set.
+ * DO NOT free the GLvector object itself, though.
+ */
+
+
+void gl_vector4f_free( GLvector4f *v )
+{
+   if (v->flags & VEC_MALLOC) {
+      ALIGN_FREE( v->storage );
+      v->data = NULL;
+      v->start = NULL;
+      v->storage = NULL;
+      v->flags &= ~VEC_MALLOC;
+   }
+}
+
+void gl_vector3f_free( GLvector3f *v )
+{
+   if (v->flags & VEC_MALLOC) {
+      ALIGN_FREE( v->storage );
+      v->data = 0;
+      v->start = 0;
+      v->storage = 0;
+      v->flags &= ~VEC_MALLOC;
+   }
+}
+
+void gl_vector1f_free( GLvector1f *v )
+{
+   if (v->flags & VEC_MALLOC) {
+      ALIGN_FREE( v->storage );
+      v->data = NULL;
+      v->start = NULL;
+      v->storage = NULL;
+      v->flags &= ~VEC_MALLOC;
+   }
+}
+
+void gl_vector4ub_free( GLvector4ub *v )
+{
+   if (v->flags & VEC_MALLOC) {
+      ALIGN_FREE( v->storage );
+      v->data = NULL;
+      v->start = NULL;
+      v->storage = NULL;
+      v->flags &= ~VEC_MALLOC;
+   }
+}
+
+void gl_vector1ub_free( GLvector1ub *v )
+{
+   if (v->flags & VEC_MALLOC) {
+      ALIGN_FREE( v->storage );
+      v->data = NULL;
+      v->start = NULL;
+      v->storage = NULL;
+      v->flags &= ~VEC_MALLOC;
+   }
+}
+
+void gl_vector1ui_free( GLvector1ui *v )
+{
+   if (v->flags & VEC_MALLOC) {
+      ALIGN_FREE( v->storage );
+      v->data = NULL;
+      v->start = NULL;
+      v->storage = NULL;
+      v->flags &= ~VEC_MALLOC;
+   }
+}
+
+
+/*
+ * For debugging
+ */
+void gl_vector4f_print( GLvector4f *v, GLubyte *cullmask, GLboolean culling )
+{
+   GLfloat c[4] = { 0, 0, 0, 1 };
+   const char *templates[5] = {
+      "%d:\t0, 0, 0, 1\n",
+      "%d:\t%f, 0, 0, 1\n",
+      "%d:\t%f, %f, 0, 1\n",
+      "%d:\t%f, %f, %f, 1\n",
+      "%d:\t%f, %f, %f, %f\n"
+   };
+
+   const char *t = templates[v->size];
+   GLfloat *d = (GLfloat *)v->data;
+   GLuint j, i = 0, count;
+
+   printf("data-start\n");
+   for ( ; d != v->start ; STRIDE_F(d, v->stride), i++) 
+      printf( t, i, d[0], d[1], d[2], d[3]);
+   
+   printf("start-count(%u)\n", v->count);
+   count = i + v->count;
+
+   if (culling) {
+      for ( ; i < count ; STRIDE_F(d, v->stride), i++) 
+	 if (cullmask[i]) 
+	    printf( t, i, d[0], d[1], d[2], d[3]);      
+   }
+   else {
+      for ( ; i < count ; STRIDE_F(d, v->stride), i++) 
+	 printf( t, i, d[0], d[1], d[2], d[3]);      
+   }
+
+   for (j = v->size ; j < 4; j++) {
+      if ((v->flags & (1<<j)) == 0) {
+
+	 printf("checking col %u is clean as advertised ", j);
+
+	 for (i = 0, d = (GLfloat *) v->data ; 
+	      i < count && d[j] == c[j] ; 
+	      i++, STRIDE_F(d, v->stride)) {};
+
+	 if (i == count) 
+	    printf(" --> ok\n");
+	 else 
+	    printf(" --> Failed at %u ******\n", i);
+      }
+   }
+}
+
+
+/*
+ * For debugging
+ */
+void gl_vector3f_print( GLvector3f *v, GLubyte *cullmask, GLboolean culling )
+{
+   GLfloat *d = (GLfloat *)v->data;
+   GLuint i = 0, count;
+
+   printf("data-start\n");
+   for ( ; d != v->start ; STRIDE_F(d,v->stride), i++) 
+      printf( "%u:\t%f, %f, %f\n", i, d[0], d[1], d[2]);
+   
+   printf("start-count(%u)\n", v->count);
+   count = i + v->count;
+
+   if (culling) {
+      for ( ; i < count ; STRIDE_F(d,v->stride), i++) 
+	 if (cullmask[i])
+	    printf( "%u:\t%f, %f, %f\n", i, d[0], d[1], d[2]);      
+   }
+   else {
+      for ( ; i < count ; STRIDE_F(d,v->stride), i++) 
+	 printf( "%u:\t%f, %f, %f\n", i, d[0], d[1], d[2]);      
+   }
+}
diff --git a/src/mesa/math/m_vector.h b/src/mesa/math/m_vector.h
new file mode 100644
index 00000000000..c4af1eaade5
--- /dev/null
+++ b/src/mesa/math/m_vector.h
@@ -0,0 +1,188 @@
+/* $Id: m_vector.h,v 1.1 2000/11/16 21:05:41 keithw Exp $ */
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  3.5
+ * 
+ * Copyright (C) 1999-2000  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"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ * 
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * New (3.1) transformation code written by Keith Whitwell.
+ */
+
+
+#ifndef _M_VECTOR_H_
+#define _M_VECTOR_H_
+
+#include "glheader.h"
+
+
+#define VEC_DIRTY_0        0x1	/* dirty flags not really used any more */
+#define VEC_DIRTY_1        0x2
+#define VEC_DIRTY_2        0x4
+#define VEC_DIRTY_3        0x8
+#define VEC_MALLOC         0x10 /* storage field points to self-allocated mem*/
+#define VEC_WRITABLE       0x20	/* keep both + and - bits for easy testing */
+#define VEC_NOT_WRITABLE   0x40
+#define VEC_GOOD_STRIDE    0x80	
+#define VEC_BAD_STRIDE     0x100
+
+#define VEC_WRITABLE_FLAGS (VEC_WRITABLE|VEC_NOT_WRITABLE)
+#define VEC_STRIDE_FLAGS   (VEC_GOOD_STRIDE|VEC_BAD_STRIDE)
+
+
+#define VEC_SIZE_1   VEC_DIRTY_0
+#define VEC_SIZE_2   (VEC_DIRTY_0|VEC_DIRTY_1)
+#define VEC_SIZE_3   (VEC_DIRTY_0|VEC_DIRTY_1|VEC_DIRTY_2)
+#define VEC_SIZE_4   (VEC_DIRTY_0|VEC_DIRTY_1|VEC_DIRTY_2|VEC_DIRTY_3)
+
+
+
+/* Wrap all the information about vectors up in a struct.  Has
+ * additional fields compared to the other vectors to help us track of
+ * different vertex sizes, and whether we need to clean columns out
+ * because they contain non-(0,0,0,1) values.  
+ *
+ * The start field is used to reserve data for copied vertices at the
+ * end of gl_transform_vb, and avoids the need for a multiplication in
+ * the transformation routines.
+ */
+typedef struct {
+   GLfloat (*data)[4];	/* may be malloc'd or point to client data */
+   GLfloat *start;	/* points somewhere inside of <data> */
+   GLuint count;	/* size of the vector (in elements) */
+   GLuint stride;	/* stride from one element to the next (in bytes) */
+   GLuint size;		/* 2-4 for vertices and 1-4 for texcoords */
+   GLuint flags;	/* which columns are dirty */
+   void *storage;	/* self-allocated storage */
+} GLvector4f;
+
+
+extern void gl_vector4f_init( GLvector4f *v, GLuint flags,
+			      GLfloat (*storage)[4] );
+extern void gl_vector4f_alloc( GLvector4f *v, GLuint flags,
+			       GLuint count, GLuint alignment );
+extern void gl_vector4f_free( GLvector4f *v );
+extern void gl_vector4f_print( GLvector4f *v, GLubyte *, GLboolean );
+extern void gl_vector4f_clean_elem( GLvector4f *vec, GLuint nr, GLuint elt );
+
+
+/* Could use a single vector type for normals and vertices, but
+ * this way avoids some casts.  
+ */
+typedef struct {
+   GLfloat (*data)[3];       
+   GLfloat *start;
+   GLuint count;
+   GLuint stride;
+   GLuint flags;
+   void *storage;
+} GLvector3f;
+
+extern void gl_vector3f_init( GLvector3f *v, GLuint flags, GLfloat (*)[3] );
+extern void gl_vector3f_alloc( GLvector3f *v, GLuint flags, GLuint count,
+			       GLuint alignment );
+extern void gl_vector3f_free( GLvector3f *v );
+extern void gl_vector3f_print( GLvector3f *v, GLubyte *, GLboolean );
+
+
+typedef struct {
+   GLfloat *data;       
+   GLfloat *start;
+   GLuint count;
+   GLuint stride;
+   GLuint flags;
+   void *storage;
+} GLvector1f;
+
+extern void gl_vector1f_free( GLvector1f *v );
+extern void gl_vector1f_init( GLvector1f *v, GLuint flags, GLfloat * );
+extern void gl_vector1f_alloc( GLvector1f *v, GLuint flags, GLuint count,
+			       GLuint alignment );
+
+
+/* For 4ub rgba values.
+ */
+typedef struct {
+   GLubyte (*data)[4];
+   GLubyte *start;
+   GLuint count;
+   GLuint stride;		 
+   GLuint flags; 
+   void *storage;
+} GLvector4ub;
+
+extern void gl_vector4ub_init( GLvector4ub *v, GLuint flags, 
+			       GLubyte (*storage)[4] );
+extern void gl_vector4ub_alloc( GLvector4ub *v, GLuint flags, GLuint count,
+				GLuint alignment );
+extern void gl_vector4ub_free( GLvector4ub * );
+
+
+
+
+/* For 1ub values, eg edgeflag.
+ */
+typedef struct {
+   GLubyte *data;
+   GLubyte *start;
+   GLuint count;
+   GLuint stride;		 
+   GLuint flags; 
+   void *storage;
+} GLvector1ub;
+
+extern void gl_vector1ub_init( GLvector1ub *v, GLuint flags, GLubyte *storage);
+extern void gl_vector1ub_alloc( GLvector1ub *v, GLuint flags, GLuint count,
+				GLuint alignment );
+extern void gl_vector1ub_free( GLvector1ub * );
+
+
+
+
+/* For, eg Index, Array element.
+ */
+typedef struct {
+   GLuint *data;
+   GLuint *start;
+   GLuint count;
+   GLuint stride; 
+   GLuint flags; 
+   void *storage;
+} GLvector1ui;
+
+extern void gl_vector1ui_init( GLvector1ui *v, GLuint flags, GLuint *storage );
+extern void gl_vector1ui_alloc( GLvector1ui *v, GLuint flags, GLuint count, 
+				GLuint alignment );
+extern void gl_vector1ui_free( GLvector1ui * );
+
+
+
+/*
+ * Given vector <v>, return a pointer (cast to <type *> to the <i>-th element.
+ *
+ * End up doing a lot of slow imuls if not careful.
+ */
+#define VEC_ELT( v, type, i ) \
+       ( (type *)  ( ((GLbyte *) ((v)->data)) + (i) * (v)->stride) )
+
+
+#endif
diff --git a/src/mesa/math/m_xform.c b/src/mesa/math/m_xform.c
new file mode 100644
index 00000000000..6bc6a9cea3b
--- /dev/null
+++ b/src/mesa/math/m_xform.c
@@ -0,0 +1,251 @@
+/* $Id: m_xform.c,v 1.1 2000/11/16 21:05:41 keithw Exp $ */
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  3.5
+ *
+ * Copyright (C) 1999-2000  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"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+
+/*
+ * Matrix/vertex/vector transformation stuff
+ *
+ *
+ * NOTES:
+ * 1. 4x4 transformation matrices are stored in memory in column major order.
+ * 2. Points/vertices are to be thought of as column vectors.
+ * 3. Transformation of a point p by a matrix M is: p' = M * p
+ */
+
+
+#include "glheader.h"
+#include "macros.h"
+#include "mmath.h"
+
+#include "m_matrix.h"
+#include "m_xform.h"
+
+
+#ifdef DEBUG
+#include "m_debug_xform.h"
+#endif
+
+#ifdef USE_X86_ASM
+#include "X86/common_x86_asm.h"
+#endif
+
+clip_func gl_clip_tab[5];
+dotprod_func gl_dotprod_tab[2][5];
+vec_copy_func gl_copy_tab[2][0x10];
+normal_func gl_normal_tab[0xf][0x4];
+transform_func **(gl_transform_tab[2]);
+static transform_func *cull_transform_tab[5];
+static transform_func *raw_transform_tab[5];
+
+
+/* Raw data format used for:
+ *    - Object-to-eye transform prior to culling, although this too
+ *      could be culled under some circumstances.
+ *    - Eye-to-clip transform (via the function above).
+ *    - Cliptesting
+ *    - And everything else too, if culling happens to be disabled.
+ */
+#define TAG(x) x##_raw
+#define TAG2(x,y) x##y##_raw
+#define IDX 0
+#define STRIDE_LOOP for (i=0;i<count;i++, STRIDE_F(from, stride))
+#define LOOP for (i=0;i<n;i++)
+#define CULL_CHECK
+#define CLIP_CHECK
+#define ARGS
+#include "m_xform_tmp.h"
+#include "m_clip_tmp.h"
+#include "m_norm_tmp.h"
+#include "m_dotprod_tmp.h"
+#include "m_copy_tmp.h"
+#undef TAG
+#undef TAG2
+#undef LOOP
+#undef CULL_CHECK
+#undef CLIP_CHECK
+#undef ARGS
+#undef IDX
+
+/* Culled data used for:
+ *    - texture transformations
+ *    - viewport map transformation
+ *    - normal transformations prior to lighting
+ *    - user cliptests
+ */
+#define TAG(x) x##_masked
+#define TAG2(x,y) x##y##_masked
+#define IDX 1
+#define STRIDE_LOOP for (i=0;i<count;i++, STRIDE_F(from, stride))
+#define LOOP for (i=0;i<n;i++)
+#define CULL_CHECK if (mask[i])
+#define CLIP_CHECK if ((mask[i] & flag) == 0)
+#define ARGS , const GLubyte mask[]
+#include "m_xform_tmp.h"
+#include "m_norm_tmp.h"
+#include "m_dotprod_tmp.h"
+#include "m_copy_tmp.h"
+#undef TAG
+#undef TAG2
+#undef LOOP
+#undef CULL_CHECK
+#undef CLIP_CHECK
+#undef ARGS
+#undef IDX
+
+
+
+
+
+
+GLvector4f *gl_project_points( GLvector4f *proj_vec,
+			       const GLvector4f *clip_vec )
+{
+   const GLuint stride = clip_vec->stride;
+   const GLfloat *from = (GLfloat *)clip_vec->start;
+   const GLuint count = clip_vec->count;
+   GLfloat (*vProj)[4] = (GLfloat (*)[4])proj_vec->start;
+   GLuint i;
+
+   for (i = 0 ; i < count ; i++, STRIDE_F(from, stride))
+   {
+	 GLfloat oow = 1.0F / from[3];
+	 vProj[i][3] = oow;
+	 vProj[i][0] = from[0] * oow;
+	 vProj[i][1] = from[1] * oow;
+	 vProj[i][2] = from[2] * oow;
+   }
+
+   proj_vec->flags |= VEC_SIZE_4;
+   proj_vec->size = 3;
+   proj_vec->count = clip_vec->count;
+   return proj_vec;
+}
+
+
+
+
+
+
+/*
+ * Transform a 4-element row vector (1x4 matrix) by a 4x4 matrix.  This
+ * function is used for transforming clipping plane equations and spotlight
+ * directions.
+ * Mathematically,  u = v * m.
+ * Input:  v - input vector
+ *         m - transformation matrix
+ * Output:  u - transformed vector
+ */
+void gl_transform_vector( GLfloat u[4], const GLfloat v[4], const GLfloat m[16] )
+{
+   GLfloat v0=v[0], v1=v[1], v2=v[2], v3=v[3];
+#define M(row,col)  m[row + col*4]
+   u[0] = v0 * M(0,0) + v1 * M(1,0) + v2 * M(2,0) + v3 * M(3,0);
+   u[1] = v0 * M(0,1) + v1 * M(1,1) + v2 * M(2,1) + v3 * M(3,1);
+   u[2] = v0 * M(0,2) + v1 * M(1,2) + v2 * M(2,2) + v3 * M(3,2);
+   u[3] = v0 * M(0,3) + v1 * M(1,3) + v2 * M(2,3) + v3 * M(3,3);
+#undef M
+}
+
+
+/* Useful for one-off point transformations, as in clipping.
+ * Note that because the matrix isn't analyzed we do too many
+ * multiplies, and that the result is always 4-clean.
+ */
+void gl_transform_point_sz( GLfloat Q[4], const GLfloat M[16],
+			    const GLfloat P[4], GLuint sz )
+{
+   if (Q == P)
+      return;
+
+   if (sz == 4)
+   {
+      Q[0] = M[0] * P[0] + M[4] * P[1] + M[8] *  P[2] + M[12] * P[3];
+      Q[1] = M[1] * P[0] + M[5] * P[1] + M[9] *  P[2] + M[13] * P[3];
+      Q[2] = M[2] * P[0] + M[6] * P[1] + M[10] * P[2] + M[14] * P[3];
+      Q[3] = M[3] * P[0] + M[7] * P[1] + M[11] * P[2] + M[15] * P[3];
+   }
+   else if (sz == 3)
+   {
+      Q[0] = M[0] * P[0] + M[4] * P[1] + M[8] *  P[2] + M[12];
+      Q[1] = M[1] * P[0] + M[5] * P[1] + M[9] *  P[2] + M[13];
+      Q[2] = M[2] * P[0] + M[6] * P[1] + M[10] * P[2] + M[14];
+      Q[3] = M[3] * P[0] + M[7] * P[1] + M[11] * P[2] + M[15];
+   }
+   else if (sz == 2)
+   {
+      Q[0] = M[0] * P[0] + M[4] * P[1] +                M[12];
+      Q[1] = M[1] * P[0] + M[5] * P[1] +                M[13];
+      Q[2] = M[2] * P[0] + M[6] * P[1] +                M[14];
+      Q[3] = M[3] * P[0] + M[7] * P[1] +                M[15];
+   }
+   else if (sz == 1)
+   {
+      Q[0] = M[0] * P[0] +                              M[12];
+      Q[1] = M[1] * P[0] +                              M[13];
+      Q[2] = M[2] * P[0] +                              M[14];
+      Q[3] = M[3] * P[0] +                              M[15];
+   }
+}
+
+
+/*
+ * This is called only once.  It initializes several tables with pointers
+ * to optimized transformation functions.  This is where we can test for
+ * AMD 3Dnow! capability, Intel Katmai, etc. and hook in the right code.
+ */
+void 
+_math_init_transformation( void )
+{
+   gl_transform_tab[0] = raw_transform_tab;
+   gl_transform_tab[1] = cull_transform_tab;
+
+   init_c_transformations_raw();
+   init_c_transformations_masked();
+   init_c_norm_transform_raw();
+   init_c_norm_transform_masked();
+   init_c_cliptest_raw();
+   init_copy0_raw();
+   init_copy0_masked();
+   init_dotprod_raw();
+   init_dotprod_masked();
+
+#ifdef DEBUG
+   gl_test_all_transform_functions( "default" );
+   gl_test_all_normal_transform_functions( "default" );
+#endif
+
+#ifdef USE_X86_ASM
+   gl_init_all_x86_transform_asm();
+#endif
+}
+
+void
+_math_init( void )
+{
+   _math_init_transformation();
+   _math_init_translate();
+   _math_init_vertices();
+}
diff --git a/src/mesa/math/m_xform.h b/src/mesa/math/m_xform.h
new file mode 100644
index 00000000000..1c6ac461e82
--- /dev/null
+++ b/src/mesa/math/m_xform.h
@@ -0,0 +1,224 @@
+/* $Id: m_xform.h,v 1.1 2000/11/16 21:05:41 keithw Exp $ */
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  3.3
+ * 
+ * Copyright (C) 1999  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"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ * 
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+
+
+
+
+#ifndef _M_XFORM_H
+#define _M_XFORM_H
+
+
+#include "glheader.h"
+#include "config.h"
+#include "math/m_vector.h"
+#include "math/m_matrix.h"
+
+#ifdef USE_X86_ASM
+#define _XFORMAPI _ASMAPI
+#define _XFORMAPIP _ASMAPIP
+#else
+#define _XFORMAPI
+#define _XFORMAPIP *
+#endif
+
+/*
+ * Transform a point (column vector) by a matrix:   Q = M * P
+ */
+#define TRANSFORM_POINT( Q, M, P )					\
+   Q[0] = M[0] * P[0] + M[4] * P[1] + M[8] *  P[2] + M[12] * P[3];	\
+   Q[1] = M[1] * P[0] + M[5] * P[1] + M[9] *  P[2] + M[13] * P[3];	\
+   Q[2] = M[2] * P[0] + M[6] * P[1] + M[10] * P[2] + M[14] * P[3];	\
+   Q[3] = M[3] * P[0] + M[7] * P[1] + M[11] * P[2] + M[15] * P[3];
+
+
+#define TRANSFORM_POINT3( Q, M, P )				\
+   Q[0] = M[0] * P[0] + M[4] * P[1] + M[8] *  P[2] + M[12];	\
+   Q[1] = M[1] * P[0] + M[5] * P[1] + M[9] *  P[2] + M[13];	\
+   Q[2] = M[2] * P[0] + M[6] * P[1] + M[10] * P[2] + M[14];	\
+   Q[3] = M[3] * P[0] + M[7] * P[1] + M[11] * P[2] + M[15];
+
+
+/*
+ * Transform a normal (row vector) by a matrix:  [NX NY NZ] = N * MAT
+ */
+#define TRANSFORM_NORMAL( TO, N, MAT )				\
+do {								\
+   TO[0] = N[0] * MAT[0] + N[1] * MAT[1] + N[2] * MAT[2];	\
+   TO[1] = N[0] * MAT[4] + N[1] * MAT[5] + N[2] * MAT[6];	\
+   TO[2] = N[0] * MAT[8] + N[1] * MAT[9] + N[2] * MAT[10];	\
+} while (0)
+
+
+extern void gl_transform_vector( GLfloat u[4],
+				 const GLfloat v[4],
+                                 const GLfloat m[16] );
+
+
+extern void gl_init_transformation( void );
+
+
+/* KW: Clip functions now do projective divide as well.  The projected
+ * coordinates are very useful to us because they let us cull
+ * backfaces and eliminate vertices from lighting, fogging, etc
+ * calculations.  Despite the fact that this divide could be done one
+ * day in hardware, we would still have a reason to want to do it here
+ * as long as those other calculations remain in software.
+ *
+ * Clipping is a convenient place to do the divide on x86 as it should be
+ * possible to overlap with integer outcode calculations.
+ *
+ * There are two cases where we wouldn't want to do the divide in cliptest:
+ *    - When we aren't clipping.  We still might want to cull backfaces
+ *      so the divide should be done elsewhere.  This currently never 
+ *      happens.
+ *
+ *    - When culling isn't likely to help us, such as when the GL culling
+ *      is disabled and we not lighting or are only lighting
+ *      one-sided.  In this situation, backface determination provides
+ *      us with no useful information.  A tricky case to detect is when
+ *      all input data is already culled, although hopefully the
+ *      application wouldn't turn on culling in such cases.
+ *
+ * We supply a buffer to hold the [x/w,y/w,z/w,1/w] values which
+ * are the result of the projection.  This is only used in the 
+ * 4-vector case - in other cases, we just use the clip coordinates
+ * as the projected coordinates - they are identical.
+ * 
+ * This is doubly convenient because it means the Win[] array is now
+ * of the same stride as all the others, so I can now turn map_vertices
+ * into a straight-forward matrix transformation, with asm acceleration
+ * automatically available.  
+ */
+
+/* Vertex buffer clipping flags 
+ */
+#define CLIP_RIGHT_SHIFT 	0
+#define CLIP_LEFT_SHIFT 	1
+#define CLIP_TOP_SHIFT  	2
+#define CLIP_BOTTOM_SHIFT       3
+#define CLIP_NEAR_SHIFT  	4
+#define CLIP_FAR_SHIFT  	5
+
+#define CLIP_RIGHT_BIT   0x01
+#define CLIP_LEFT_BIT    0x02
+#define CLIP_TOP_BIT     0x04
+#define CLIP_BOTTOM_BIT  0x08
+#define CLIP_NEAR_BIT    0x10
+#define CLIP_FAR_BIT     0x20
+#define CLIP_USER_BIT    0x40
+#define CLIP_CULLED_BIT  0x80	/* Vertex has been culled */
+#define CLIP_ALL_BITS    0x3f
+
+
+typedef GLvector4f * (_XFORMAPIP clip_func)( GLvector4f *vClip,
+				  GLvector4f *vProj, 
+				  GLubyte clipMask[],
+				  GLubyte *orMask, 
+				  GLubyte *andMask );
+
+typedef void (*dotprod_func)( GLvector4f *out_vec, 
+			      GLuint elt,
+			      const GLvector4f *coord_vec, 
+			      const GLfloat plane[4], 
+			      const GLubyte mask[]);
+
+typedef void (*vec_copy_func)( GLvector4f *to, 
+			       const GLvector4f *from, 
+			       const GLubyte mask[]);
+
+
+
+/*
+ * Functions for transformation of normals in the VB.
+ */
+typedef void (_NORMAPIP normal_func)( const GLmatrix *mat,
+			     GLfloat scale,
+			     const GLvector3f *in,
+			     const GLfloat lengths[],
+			     const GLubyte mask[],
+			     GLvector3f *dest );
+
+
+/* Flags for selecting a normal transformation function. 
+ */
+#define NORM_RESCALE   0x1	/* apply the scale factor */
+#define NORM_NORMALIZE 0x2	/* normalize */
+#define NORM_TRANSFORM 0x4	/* apply the transformation matrix */
+#define NORM_TRANSFORM_NO_ROT 0x8	/* apply the transformation matrix */
+
+
+
+
+/* KW: New versions of the transform function allow a mask array
+ *     specifying that individual vector transform should be skipped
+ *     when the mask byte is zero.  This is always present as a
+ *     parameter, to allow a unified interface.  
+ */
+typedef void (_XFORMAPIP transform_func)( GLvector4f *to_vec,
+					  const GLfloat m[16],
+				const GLvector4f *from_vec, 
+				const GLubyte *clipmask,
+				const GLubyte flag );
+
+
+extern GLvector4f *gl_project_points( GLvector4f *to,
+			       const GLvector4f *from );
+
+extern void gl_transform_bounds3( GLubyte *orMask, GLubyte *andMask,
+				  const GLfloat m[16],
+			   CONST GLfloat src[][3] );
+
+extern void gl_transform_bounds2( GLubyte *orMask, GLubyte *andMask,
+				  const GLfloat m[16],
+			   CONST GLfloat src[][3] );
+
+
+extern dotprod_func  gl_dotprod_tab[2][5];
+extern vec_copy_func gl_copy_tab[2][0x10];
+extern clip_func     gl_clip_tab[5];
+extern normal_func   gl_normal_tab[0xf][0x4];
+
+/* Use of 3 layers of linked 1-dimensional arrays to reduce
+ * cost of lookup.
+ */
+extern transform_func **(gl_transform_tab[2]);
+
+
+extern void gl_transform_point_sz( GLfloat Q[4], const GLfloat M[16],
+				   const GLfloat P[4], GLuint sz );
+
+
+#define TransformRaw( to, mat, from ) \
+      ( (*gl_transform_tab[0][(from)->size][(mat)->type])( to, (mat)->m, from, 0, 0 ), \
+        (to) )
+
+#define Transform( to, mat, from, mask, cull ) \
+      ( (*gl_transform_tab[cull!=0][(from)->size][(mat)->type])( to, (mat)->m, from, mask, cull ), \
+	(to) )
+
+
+#endif
diff --git a/src/mesa/math/m_xform_tmp.h b/src/mesa/math/m_xform_tmp.h
new file mode 100644
index 00000000000..289255a9cf5
--- /dev/null
+++ b/src/mesa/math/m_xform_tmp.h
@@ -0,0 +1,974 @@
+/* $Id: m_xform_tmp.h,v 1.1 2000/11/16 21:05:41 keithw Exp $ */
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  3.1
+ *
+ * Copyright (C) 1999  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"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * New (3.1) transformation code written by Keith Whitwell.
+ */
+
+
+/*----------------------------------------------------------------------
+ * Begin Keith's new code
+ *
+ *----------------------------------------------------------------------
+ */
+
+/* KW: Fixed stride, now measured in bytes as is the OpenGL array stride.
+ */
+
+/* KW: These are now parameterized to produce two versions, one
+ *     which transforms all incoming points, and a second which
+ *     takes notice of a cullmask array, and only transforms
+ *     unculled vertices.
+ */
+
+/* KW: 1-vectors can sneak into the texture pipeline via the array
+ *     interface.  These functions are here because I want consistant
+ *     treatment of the vertex sizes and a lazy strategy for
+ *     cleaning unused parts of the vector, and so as not to exclude
+ *     them from the vertex array interface.
+ *
+ *     Under our current analysis of matrices, there is no way that
+ *     the product of a matrix and a 1-vector can remain a 1-vector,
+ *     with the exception of the identity transform.
+ */
+
+/* KW: No longer zero-pad outgoing vectors.  Now that external
+ *     vectors can get into the pipeline we cannot ever assume
+ *     that there is more to a vector than indicated by its
+ *     size.
+ */
+
+/* KW: Now uses clipmask and a flag to allow us to skip both/either
+ *     cliped and/or culled vertices.
+ */
+
+static void _XFORMAPI
+TAG(transform_points1_general)( GLvector4f *to_vec,
+				const GLfloat m[16],
+					    const GLvector4f *from_vec,
+					    const GLubyte *mask,
+					    const GLubyte flag )
+{
+   const GLuint stride = from_vec->stride;
+   GLfloat *from = from_vec->start;
+   GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
+   GLuint count = from_vec->count;
+   const GLfloat m0 = m[0],  m12 = m[12];
+   const GLfloat m1 = m[1],  m13 = m[13];
+   const GLfloat m2 = m[2],  m14 = m[14];
+   const GLfloat m3 = m[3],  m15 = m[15];
+   GLuint i;
+   (void) mask;
+   (void) flag;
+   STRIDE_LOOP {
+      CLIP_CHECK {
+	 const GLfloat ox = from[0];
+	 to[i][0] = m0 * ox + m12;
+	 to[i][1] = m1 * ox + m13;
+	 to[i][2] = m2 * ox + m14;
+	 to[i][3] = m3 * ox + m15;
+      }
+   }
+
+   to_vec->size = 4;
+   to_vec->flags |= VEC_SIZE_4;
+   to_vec->count = from_vec->count;
+}
+
+static void _XFORMAPI
+TAG(transform_points1_identity)( GLvector4f *to_vec,
+				 const GLfloat m[16],
+					     const GLvector4f *from_vec,
+					     const GLubyte *mask,
+					     const GLubyte flag )
+{
+   const GLuint stride = from_vec->stride;
+   GLfloat *from = from_vec->start;
+   GLuint count = from_vec->count;
+   GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
+   GLuint i;
+   (void) mask;
+   (void) flag;
+   if (to_vec == from_vec) return;
+   STRIDE_LOOP {
+      CLIP_CHECK {
+	 to[i][0] = from[0];
+      }
+   }
+
+   to_vec->size = 1;
+   to_vec->flags |= VEC_SIZE_1;
+   to_vec->count = from_vec->count;
+}
+
+static void _XFORMAPI
+TAG(transform_points1_2d)( GLvector4f *to_vec,
+			   const GLfloat m[16],
+				       const GLvector4f *from_vec,
+				       const GLubyte *mask,
+				       const GLubyte flag )
+{
+   const GLuint stride = from_vec->stride;
+   GLfloat *from = from_vec->start;
+   GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
+   GLuint count = from_vec->count;
+   const GLfloat m0 = m[0], m1 = m[1];
+   const GLfloat m12 = m[12], m13 = m[13];
+   GLuint i;
+   (void) mask;
+   (void) flag;
+   STRIDE_LOOP {
+      CLIP_CHECK {
+	 const GLfloat ox = from[0];
+	 to[i][0] = m0 * ox + m12;
+	 to[i][1] = m1 * ox + m13;
+      }
+   }
+   to_vec->size = 2;
+   to_vec->flags |= VEC_SIZE_2;
+   to_vec->count = from_vec->count;
+}
+
+static void _XFORMAPI
+TAG(transform_points1_2d_no_rot)( GLvector4f *to_vec,
+				  const GLfloat m[16],
+					      const GLvector4f *from_vec,
+					      const GLubyte *mask,
+					      const GLubyte flag )
+{
+   const GLuint stride = from_vec->stride;
+   GLfloat *from = from_vec->start;
+   GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
+   GLuint count = from_vec->count;
+   const GLfloat m0 = m[0], m12 = m[12], m13 = m[13];
+   GLuint i;
+   (void) mask;
+   (void) flag;
+   STRIDE_LOOP {
+      CLIP_CHECK {
+	 const GLfloat ox = from[0];
+	 to[i][0] = m0 * ox + m12;
+	 to[i][1] =           m13;
+      }
+   }
+
+   to_vec->size = 2;
+   to_vec->flags |= VEC_SIZE_2;
+   to_vec->count = from_vec->count;
+}
+
+static void _XFORMAPI
+TAG(transform_points1_3d)( GLvector4f *to_vec,
+			   const GLfloat m[16],
+				       const GLvector4f *from_vec,
+				       const GLubyte *mask,
+				       const GLubyte flag )
+{
+   const GLuint stride = from_vec->stride;
+   GLfloat *from = from_vec->start;
+   GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
+   GLuint count = from_vec->count;
+   const GLfloat m0 = m[0], m1 = m[1], m2 = m[2];
+   const GLfloat m12 = m[12], m13 = m[13], m14 = m[14];
+   GLuint i;
+   (void) mask;
+   (void) flag;
+   STRIDE_LOOP {
+      CLIP_CHECK {
+	 const GLfloat ox = from[0];
+	 to[i][0] = m0 * ox + m12;
+	 to[i][1] = m1 * ox + m13;
+	 to[i][2] = m2 * ox + m14;
+      }
+   }
+   to_vec->size = 3;
+   to_vec->flags |= VEC_SIZE_3;
+   to_vec->count = from_vec->count;
+}
+
+
+static void _XFORMAPI
+TAG(transform_points1_3d_no_rot)( GLvector4f *to_vec,
+				  const GLfloat m[16],
+					      const GLvector4f *from_vec,
+					      const GLubyte *mask,
+					      const GLubyte flag )
+{
+   const GLuint stride = from_vec->stride;
+   GLfloat *from = from_vec->start;
+   GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
+   GLuint count = from_vec->count;
+   const GLfloat m0 = m[0];
+   const GLfloat m12 = m[12], m13 = m[13], m14 = m[14];
+   GLuint i;
+   (void) mask;
+   (void) flag;
+   STRIDE_LOOP {
+      CLIP_CHECK {
+	 const GLfloat ox = from[0];
+	 to[i][0] = m0 * ox           + m12;
+	 to[i][1] =                     m13;
+	 to[i][2] =                     m14;
+      }
+   }
+   to_vec->size = 3;
+   to_vec->flags |= VEC_SIZE_3;
+   to_vec->count = from_vec->count;
+}
+
+static void _XFORMAPI
+TAG(transform_points1_perspective)( GLvector4f *to_vec,
+				    const GLfloat m[16],
+						const GLvector4f *from_vec,
+						const GLubyte *mask,
+						const GLubyte flag )
+{
+   const GLuint stride = from_vec->stride;
+   GLfloat *from = from_vec->start;
+   GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
+   GLuint count = from_vec->count;
+   const GLfloat m0 = m[0], m14 = m[14];
+   GLuint i;
+   (void) mask;
+   (void) flag;
+   STRIDE_LOOP {
+      CLIP_CHECK {
+	 const GLfloat ox = from[0];
+	 to[i][0] = m0 * ox                ;
+	 to[i][1] =           0            ;
+	 to[i][2] =                     m14;
+	 to[i][3] = 0;
+      }
+   }
+   to_vec->size = 4;
+   to_vec->flags |= VEC_SIZE_4;
+   to_vec->count = from_vec->count;
+}
+
+
+
+
+/* 2-vectors, which are a lot more relevant than 1-vectors, are
+ * present early in the geometry pipeline and throughout the
+ * texture pipeline.
+ */
+static void _XFORMAPI
+TAG(transform_points2_general)( GLvector4f *to_vec,
+				const GLfloat m[16],
+					    const GLvector4f *from_vec,
+					    const GLubyte *mask,
+					    const GLubyte flag )
+{
+   const GLuint stride = from_vec->stride;
+   GLfloat *from = from_vec->start;
+   GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
+   GLuint count = from_vec->count;
+   const GLfloat m0 = m[0],  m4 = m[4],  m12 = m[12];
+   const GLfloat m1 = m[1],  m5 = m[5],  m13 = m[13];
+   const GLfloat m2 = m[2],  m6 = m[6],  m14 = m[14];
+   const GLfloat m3 = m[3],  m7 = m[7],  m15 = m[15];
+   GLuint i;
+   (void) mask;
+   (void) flag;
+   STRIDE_LOOP {
+      CLIP_CHECK {
+	 const GLfloat ox = from[0], oy = from[1];
+	 to[i][0] = m0 * ox + m4 * oy + m12;
+	 to[i][1] = m1 * ox + m5 * oy + m13;
+	 to[i][2] = m2 * ox + m6 * oy + m14;
+	 to[i][3] = m3 * ox + m7 * oy + m15;
+      }
+   }
+   to_vec->size = 4;
+   to_vec->flags |= VEC_SIZE_4;
+   to_vec->count = from_vec->count;
+}
+
+static void _XFORMAPI
+TAG(transform_points2_identity)( GLvector4f *to_vec,
+				 const GLfloat m[16],
+					     const GLvector4f *from_vec,
+					     const GLubyte *mask,
+					     const GLubyte flag )
+{
+   const GLuint stride = from_vec->stride;
+   GLfloat *from = from_vec->start;
+   GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
+   GLuint count = from_vec->count;
+   GLuint i;
+   (void) mask;
+   (void) flag;
+   if (to_vec == from_vec) return;
+   STRIDE_LOOP {
+      CLIP_CHECK {
+	 to[i][0] = from[0];
+	 to[i][1] = from[1];
+      }
+   }
+   to_vec->size = 2;
+   to_vec->flags |= VEC_SIZE_2;
+   to_vec->count = from_vec->count;
+}
+
+static void _XFORMAPI
+TAG(transform_points2_2d)( GLvector4f *to_vec,
+			   const GLfloat m[16],
+				       const GLvector4f *from_vec,
+				       const GLubyte *mask,
+				       const GLubyte flag )
+{
+   const GLuint stride = from_vec->stride;
+   GLfloat *from = from_vec->start;
+   GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
+   GLuint count = from_vec->count;
+   const GLfloat m0 = m[0], m1 = m[1], m4 = m[4], m5 = m[5];
+   const GLfloat m12 = m[12], m13 = m[13];
+   GLuint i;
+   (void) mask;
+   (void) flag;
+   STRIDE_LOOP {
+      CLIP_CHECK {
+	 const GLfloat ox = from[0], oy = from[1];
+	 to[i][0] = m0 * ox + m4 * oy + m12;
+	 to[i][1] = m1 * ox + m5 * oy + m13;
+      }
+   }
+
+   to_vec->size = 2;
+   to_vec->flags |= VEC_SIZE_2;
+   to_vec->count = from_vec->count;
+}
+
+static void _XFORMAPI
+TAG(transform_points2_2d_no_rot)( GLvector4f *to_vec,
+				  const GLfloat m[16],
+					      const GLvector4f *from_vec,
+					      const GLubyte *mask,
+					      const GLubyte flag )
+{
+   const GLuint stride = from_vec->stride;
+   GLfloat *from = from_vec->start;
+   GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
+   GLuint count = from_vec->count;
+   const GLfloat m0 = m[0], m5 = m[5], m12 = m[12], m13 = m[13];
+   GLuint i;
+   (void) mask;
+   (void) flag;
+   STRIDE_LOOP {
+      CLIP_CHECK {
+	 const GLfloat ox = from[0], oy = from[1];
+	 to[i][0] = m0 * ox           + m12;
+	 to[i][1] =           m5 * oy + m13;
+      }
+   }
+
+   to_vec->size = 2;
+   to_vec->flags |= VEC_SIZE_2;
+   to_vec->count = from_vec->count;
+}
+
+static void _XFORMAPI
+TAG(transform_points2_3d)( GLvector4f *to_vec,
+			   const GLfloat m[16],
+				       const GLvector4f *from_vec,
+				       const GLubyte *mask,
+				       const GLubyte flag )
+{
+   const GLuint stride = from_vec->stride;
+   GLfloat *from = from_vec->start;
+   GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
+   GLuint count = from_vec->count;
+   const GLfloat m0 = m[0], m1 = m[1], m2 = m[2], m4 = m[4], m5 = m[5];
+   const GLfloat m6 = m[6], m12 = m[12], m13 = m[13], m14 = m[14];
+   GLuint i;
+   (void) mask;
+   (void) flag;
+   STRIDE_LOOP {
+      CLIP_CHECK {
+	 const GLfloat ox = from[0], oy = from[1];
+	 to[i][0] = m0 * ox + m4 * oy + m12;
+	 to[i][1] = m1 * ox + m5 * oy + m13;
+	 to[i][2] = m2 * ox + m6 * oy + m14;
+      }
+   }
+   to_vec->size = 3;
+   to_vec->flags |= VEC_SIZE_3;
+   to_vec->count = from_vec->count;
+}
+
+
+/* I would actually say this was a fairly important function, from
+ * a texture transformation point of view.
+ */
+static void _XFORMAPI
+TAG(transform_points2_3d_no_rot)( GLvector4f *to_vec,
+				  const GLfloat m[16],
+					      const GLvector4f *from_vec,
+					      const GLubyte *mask,
+					      const GLubyte flag )
+{
+   const GLuint stride = from_vec->stride;
+   GLfloat *from = from_vec->start;
+   GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
+   GLuint count = from_vec->count;
+   const GLfloat m0 = m[0], m5 = m[5];
+   const GLfloat m12 = m[12], m13 = m[13], m14 = m[14];
+   GLuint i;
+   (void) mask;
+   (void) flag;
+   STRIDE_LOOP {
+      CLIP_CHECK {
+	 const GLfloat ox = from[0], oy = from[1];
+	 to[i][0] = m0 * ox           + m12;
+	 to[i][1] =           m5 * oy + m13;
+	 to[i][2] =                     m14;
+      }
+   }
+   if (m14 == 0) {
+      to_vec->size = 2;
+      to_vec->flags |= VEC_SIZE_2;
+   } else {
+      to_vec->size = 3;
+      to_vec->flags |= VEC_SIZE_3;
+   }
+   to_vec->count = from_vec->count;
+}
+
+/* This may not be called too often, but I wouldn't say it was dead
+ * code.  It's also hard to remove any of these functions if you are
+ * attached to the assertions that have appeared in them.
+ */
+static void _XFORMAPI
+TAG(transform_points2_perspective)( GLvector4f *to_vec,
+				    const GLfloat m[16],
+						const GLvector4f *from_vec,
+						const GLubyte *mask,
+						const GLubyte flag )
+{
+   const GLuint stride = from_vec->stride;
+   GLfloat *from = from_vec->start;
+   GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
+   GLuint count = from_vec->count;
+   const GLfloat m0 = m[0], m5 = m[5], m14 = m[14];
+   GLuint i;
+   (void) mask;
+   (void) flag;
+   STRIDE_LOOP {
+      CLIP_CHECK {
+	 const GLfloat ox = from[0], oy = from[1];
+	 to[i][0] = m0 * ox                ;
+	 to[i][1] =           m5 * oy      ;
+	 to[i][2] =                     m14;
+	 to[i][3] = 0;
+      }
+   }
+   to_vec->size = 4;
+   to_vec->flags |= VEC_SIZE_4;
+   to_vec->count = from_vec->count;
+}
+
+
+
+static void _XFORMAPI
+TAG(transform_points3_general)( GLvector4f *to_vec,
+				const GLfloat m[16],
+					    const GLvector4f *from_vec,
+					    const GLubyte *mask,
+					    const GLubyte flag )
+{
+   const GLuint stride = from_vec->stride;
+   GLfloat *from = from_vec->start;
+   GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
+   GLuint count = from_vec->count;
+   const GLfloat m0 = m[0],  m4 = m[4],  m8 = m[8],  m12 = m[12];
+   const GLfloat m1 = m[1],  m5 = m[5],  m9 = m[9],  m13 = m[13];
+   const GLfloat m2 = m[2],  m6 = m[6],  m10 = m[10],  m14 = m[14];
+   const GLfloat m3 = m[3],  m7 = m[7],  m11 = m[11],  m15 = m[15];
+   GLuint i;
+   (void) mask;
+   (void) flag;
+   STRIDE_LOOP {
+      CLIP_CHECK {
+	 const GLfloat ox = from[0], oy = from[1], oz = from[2];
+	 to[i][0] = m0 * ox + m4 * oy + m8  * oz + m12;
+	 to[i][1] = m1 * ox + m5 * oy + m9  * oz + m13;
+	 to[i][2] = m2 * ox + m6 * oy + m10 * oz + m14;
+	 to[i][3] = m3 * ox + m7 * oy + m11 * oz + m15;
+      }
+   }
+   to_vec->size = 4;
+   to_vec->flags |= VEC_SIZE_4;
+   to_vec->count = from_vec->count;
+}
+
+static void _XFORMAPI
+TAG(transform_points3_identity)( GLvector4f *to_vec,
+				 const GLfloat m[16],
+					     const GLvector4f *from_vec,
+					     const GLubyte *mask,
+					     const GLubyte flag )
+{
+   const GLuint stride = from_vec->stride;
+   GLfloat *from = from_vec->start;
+   GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
+   GLuint count = from_vec->count;
+   GLuint i;
+   (void) mask;
+   (void) flag;
+   if (to_vec == from_vec) return;
+   STRIDE_LOOP {
+      CLIP_CHECK {
+	 to[i][0] = from[0];
+	 to[i][1] = from[1];
+	 to[i][2] = from[2];
+      }
+   }
+   to_vec->size = 3;
+   to_vec->flags |= VEC_SIZE_3;
+   to_vec->count = from_vec->count;
+}
+
+static void _XFORMAPI
+TAG(transform_points3_2d)( GLvector4f *to_vec,
+			   const GLfloat m[16],
+				       const GLvector4f *from_vec,
+				       const GLubyte *mask,
+				       const GLubyte flag )
+{
+   const GLuint stride = from_vec->stride;
+   GLfloat *from = from_vec->start;
+   GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
+   GLuint count = from_vec->count;
+   const GLfloat m0 = m[0], m1 = m[1], m4 = m[4], m5 = m[5];
+   const GLfloat m12 = m[12], m13 = m[13];
+   GLuint i;
+   (void) mask;
+   (void) flag;
+   STRIDE_LOOP {
+      CLIP_CHECK {
+	 const GLfloat ox = from[0], oy = from[1], oz = from[2];
+	 to[i][0] = m0 * ox + m4 * oy            + m12       ;
+	 to[i][1] = m1 * ox + m5 * oy            + m13       ;
+	 to[i][2] =                   +       oz             ;
+      }
+   }
+   to_vec->size = 3;
+   to_vec->flags |= VEC_SIZE_3;
+   to_vec->count = from_vec->count;
+}
+
+static void _XFORMAPI
+TAG(transform_points3_2d_no_rot)( GLvector4f *to_vec,
+				  const GLfloat m[16],
+					      const GLvector4f *from_vec,
+					      const GLubyte *mask,
+					      const GLubyte flag )
+{
+   const GLuint stride = from_vec->stride;
+   GLfloat *from = from_vec->start;
+   GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
+   GLuint count = from_vec->count;
+   const GLfloat m0 = m[0], m5 = m[5], m12 = m[12], m13 = m[13];
+   GLuint i;
+   (void) mask;
+   (void) flag;
+   STRIDE_LOOP {
+      CLIP_CHECK {
+	 const GLfloat ox = from[0], oy = from[1], oz = from[2];
+	 to[i][0] = m0 * ox                      + m12       ;
+	 to[i][1] =           m5 * oy            + m13       ;
+	 to[i][2] =                   +       oz             ;
+      }
+   }
+   to_vec->size = 3;
+   to_vec->flags |= VEC_SIZE_3;
+   to_vec->count = from_vec->count;
+}
+
+static void _XFORMAPI
+TAG(transform_points3_3d)( GLvector4f *to_vec,
+			   const GLfloat m[16],
+				       const GLvector4f *from_vec,
+				       const GLubyte *mask,
+				       const GLubyte flag )
+{
+   const GLuint stride = from_vec->stride;
+   GLfloat *from = from_vec->start;
+   GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
+   GLuint count = from_vec->count;
+   const GLfloat m0 = m[0], m1 = m[1], m2 = m[2], m4 = m[4], m5 = m[5];
+   const GLfloat m6 = m[6], m8 = m[8], m9 = m[9], m10 = m[10];
+   const GLfloat m12 = m[12], m13 = m[13], m14 = m[14];
+   GLuint i;
+   (void) mask;
+   (void) flag;
+   STRIDE_LOOP {
+      CLIP_CHECK {
+	 const GLfloat ox = from[0], oy = from[1], oz = from[2];
+	 to[i][0] = m0 * ox + m4 * oy +  m8 * oz + m12       ;
+	 to[i][1] = m1 * ox + m5 * oy +  m9 * oz + m13       ;
+	 to[i][2] = m2 * ox + m6 * oy + m10 * oz + m14       ;
+      }
+   }
+   to_vec->size = 3;
+   to_vec->flags |= VEC_SIZE_3;
+   to_vec->count = from_vec->count;
+}
+
+/* previously known as ortho...
+ */
+static void _XFORMAPI
+TAG(transform_points3_3d_no_rot)( GLvector4f *to_vec,
+				  const GLfloat m[16],
+					      const GLvector4f *from_vec,
+					      const GLubyte *mask,
+					      const GLubyte flag )
+{
+   const GLuint stride = from_vec->stride;
+   GLfloat *from = from_vec->start;
+   GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
+   GLuint count = from_vec->count;
+   const GLfloat m0 = m[0], m5 = m[5];
+   const GLfloat m10 = m[10], m12 = m[12], m13 = m[13], m14 = m[14];
+   GLuint i;
+   (void) mask;
+   (void) flag;
+   STRIDE_LOOP {
+      CLIP_CHECK {
+	 const GLfloat ox = from[0], oy = from[1], oz = from[2];
+	 to[i][0] = m0 * ox                      + m12       ;
+	 to[i][1] =           m5 * oy            + m13       ;
+	 to[i][2] =                     m10 * oz + m14       ;
+      }
+   }
+   to_vec->size = 3;
+   to_vec->flags |= VEC_SIZE_3;
+   to_vec->count = from_vec->count;
+}
+
+static void _XFORMAPI
+TAG(transform_points3_perspective)( GLvector4f *to_vec,
+				    const GLfloat m[16],
+						const GLvector4f *from_vec,
+						const GLubyte *mask,
+						const GLubyte flag )
+{
+   const GLuint stride = from_vec->stride;
+   GLfloat *from = from_vec->start;
+   GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
+   GLuint count = from_vec->count;
+   const GLfloat m0 = m[0], m5 = m[5], m8 = m[8], m9 = m[9];
+   const GLfloat m10 = m[10], m14 = m[14];
+   GLuint i;
+   (void) mask;
+   (void) flag;
+   STRIDE_LOOP {
+      CLIP_CHECK {
+	 const GLfloat ox = from[0], oy = from[1], oz = from[2];
+	 to[i][0] = m0 * ox           + m8  * oz       ;
+	 to[i][1] =           m5 * oy + m9  * oz       ;
+	 to[i][2] =                     m10 * oz + m14 ;
+	 to[i][3] =                          -oz       ;
+      }
+   }
+   to_vec->size = 4;
+   to_vec->flags |= VEC_SIZE_4;
+   to_vec->count = from_vec->count;
+}
+
+
+
+static void _XFORMAPI
+TAG(transform_points4_general)( GLvector4f *to_vec,
+				const GLfloat m[16],
+					    const GLvector4f *from_vec,
+					    const GLubyte *mask,
+					    const GLubyte flag )
+{
+   const GLuint stride = from_vec->stride;
+   GLfloat *from = from_vec->start;
+   GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
+   GLuint count = from_vec->count;
+   const GLfloat m0 = m[0],  m4 = m[4],  m8 = m[8],  m12 = m[12];
+   const GLfloat m1 = m[1],  m5 = m[5],  m9 = m[9],  m13 = m[13];
+   const GLfloat m2 = m[2],  m6 = m[6],  m10 = m[10],  m14 = m[14];
+   const GLfloat m3 = m[3],  m7 = m[7],  m11 = m[11],  m15 = m[15];
+   GLuint i;
+   (void) mask;
+   (void) flag;
+   STRIDE_LOOP {
+      CLIP_CHECK {
+	 const GLfloat ox = from[0], oy = from[1], oz = from[2], ow = from[3];
+	 to[i][0] = m0 * ox + m4 * oy + m8  * oz + m12 * ow;
+	 to[i][1] = m1 * ox + m5 * oy + m9  * oz + m13 * ow;
+	 to[i][2] = m2 * ox + m6 * oy + m10 * oz + m14 * ow;
+	 to[i][3] = m3 * ox + m7 * oy + m11 * oz + m15 * ow;
+      }
+   }
+   to_vec->size = 4;
+   to_vec->flags |= VEC_SIZE_4;
+   to_vec->count = from_vec->count;
+}
+
+static void _XFORMAPI
+TAG(transform_points4_identity)( GLvector4f *to_vec,
+				 const GLfloat m[16],
+					     const GLvector4f *from_vec,
+					     const GLubyte *mask,
+					     const GLubyte flag )
+{
+   const GLuint stride = from_vec->stride;
+   GLfloat *from = from_vec->start;
+   GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
+   GLuint count = from_vec->count;
+   GLuint i;
+   (void) mask;
+   (void) flag;
+   if (to_vec == from_vec) return;
+   STRIDE_LOOP {
+      CLIP_CHECK {
+	 to[i][0] = from[0];
+	 to[i][1] = from[1];
+	 to[i][2] = from[2];
+	 to[i][3] = from[3];
+      }
+   }
+   to_vec->size = 4;
+   to_vec->flags |= VEC_SIZE_4;
+   to_vec->count = from_vec->count;
+}
+
+static void _XFORMAPI
+TAG(transform_points4_2d)( GLvector4f *to_vec,
+			   const GLfloat m[16],
+				       const GLvector4f *from_vec,
+				       const GLubyte *mask,
+				       const GLubyte flag )
+{
+   const GLuint stride = from_vec->stride;
+   GLfloat *from = from_vec->start;
+   GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
+   GLuint count = from_vec->count;
+   const GLfloat m0 = m[0], m1 = m[1], m4 = m[4], m5 = m[5];
+   const GLfloat m12 = m[12], m13 = m[13];
+   GLuint i;
+   (void) mask;
+   (void) flag;
+   STRIDE_LOOP {
+      CLIP_CHECK {
+	 const GLfloat ox = from[0], oy = from[1], oz = from[2], ow = from[3];
+	 to[i][0] = m0 * ox + m4 * oy            + m12 * ow;
+	 to[i][1] = m1 * ox + m5 * oy            + m13 * ow;
+	 to[i][2] =                   +       oz           ;
+	 to[i][3] =                                      ow;
+      }
+   }
+   to_vec->size = 4;
+   to_vec->flags |= VEC_SIZE_4;
+   to_vec->count = from_vec->count;
+}
+
+static void _XFORMAPI
+TAG(transform_points4_2d_no_rot)( GLvector4f *to_vec,
+				  const GLfloat m[16],
+					      const GLvector4f *from_vec,
+					      const GLubyte *mask,
+					      const GLubyte flag )
+{
+   const GLuint stride = from_vec->stride;
+   GLfloat *from = from_vec->start;
+   GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
+   GLuint count = from_vec->count;
+   const GLfloat m0 = m[0], m5 = m[5], m12 = m[12], m13 = m[13];
+   GLuint i;
+   (void) mask;
+   (void) flag;
+   STRIDE_LOOP {
+      CLIP_CHECK {
+	 const GLfloat ox = from[0], oy = from[1], oz = from[2], ow = from[3];
+	 to[i][0] = m0 * ox                      + m12 * ow;
+	 to[i][1] =           m5 * oy            + m13 * ow;
+	 to[i][2] =                   +       oz           ;
+	 to[i][3] =                                      ow;
+      }
+   }
+   to_vec->size = 4;
+   to_vec->flags |= VEC_SIZE_4;
+   to_vec->count = from_vec->count;
+}
+
+static void _XFORMAPI
+TAG(transform_points4_3d)( GLvector4f *to_vec,
+			   const GLfloat m[16],
+				       const GLvector4f *from_vec,
+				       const GLubyte *mask,
+				       const GLubyte flag )
+{
+   const GLuint stride = from_vec->stride;
+   GLfloat *from = from_vec->start;
+   GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
+   GLuint count = from_vec->count;
+   const GLfloat m0 = m[0], m1 = m[1], m2 = m[2], m4 = m[4], m5 = m[5];
+   const GLfloat m6 = m[6], m8 = m[8], m9 = m[9], m10 = m[10];
+   const GLfloat m12 = m[12], m13 = m[13], m14 = m[14];
+   GLuint i;
+   (void) mask;
+   (void) flag;
+   STRIDE_LOOP {
+      CLIP_CHECK {
+	 const GLfloat ox = from[0], oy = from[1], oz = from[2], ow = from[3];
+	 to[i][0] = m0 * ox + m4 * oy +  m8 * oz + m12 * ow;
+	 to[i][1] = m1 * ox + m5 * oy +  m9 * oz + m13 * ow;
+	 to[i][2] = m2 * ox + m6 * oy + m10 * oz + m14 * ow;
+	 to[i][3] =                                      ow;
+      }
+   }
+   to_vec->size = 4;
+   to_vec->flags |= VEC_SIZE_4;
+   to_vec->count = from_vec->count;
+}
+
+static void _XFORMAPI
+TAG(transform_points4_3d_no_rot)( GLvector4f *to_vec,
+				  const GLfloat m[16],
+					      const GLvector4f *from_vec,
+					      const GLubyte *mask,
+					      const GLubyte flag )
+{
+   const GLuint stride = from_vec->stride;
+   GLfloat *from = from_vec->start;
+   GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
+   GLuint count = from_vec->count;
+   const GLfloat m0 = m[0], m5 = m[5];
+   const GLfloat m10 = m[10], m12 = m[12], m13 = m[13], m14 = m[14];
+   GLuint i;
+   (void) mask;
+   (void) flag;
+   STRIDE_LOOP {
+      CLIP_CHECK {
+	 const GLfloat ox = from[0], oy = from[1], oz = from[2], ow = from[3];
+	 to[i][0] = m0 * ox                      + m12 * ow;
+	 to[i][1] =           m5 * oy            + m13 * ow;
+	 to[i][2] =                     m10 * oz + m14 * ow;
+	 to[i][3] =                                      ow;
+      }
+   }
+   to_vec->size = 4;
+   to_vec->flags |= VEC_SIZE_4;
+   to_vec->count = from_vec->count;
+}
+
+static void _XFORMAPI
+TAG(transform_points4_perspective)( GLvector4f *to_vec,
+				    const GLfloat m[16],
+						const GLvector4f *from_vec,
+						const GLubyte *mask,
+						const GLubyte flag )
+{
+   const GLuint stride = from_vec->stride;
+   GLfloat *from = from_vec->start;
+   GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
+   GLuint count = from_vec->count;
+   const GLfloat m0 = m[0], m5 = m[5], m8 = m[8], m9 = m[9];
+   const GLfloat m10 = m[10], m14 = m[14];
+   GLuint i;
+   (void) mask;
+   (void) flag;
+   STRIDE_LOOP {
+      CLIP_CHECK {
+	 const GLfloat ox = from[0], oy = from[1], oz = from[2], ow = from[3];
+	 to[i][0] = m0 * ox           + m8  * oz            ;
+	 to[i][1] =           m5 * oy + m9  * oz            ;
+	 to[i][2] =                     m10 * oz + m14 * ow ;
+	 to[i][3] =                          -oz            ;
+      }
+   }
+
+   to_vec->size = 4;
+   to_vec->flags |= VEC_SIZE_4;
+   to_vec->count = from_vec->count;
+}
+
+static transform_func _XFORMAPI TAG(transform_tab_1)[7];
+static transform_func _XFORMAPI TAG(transform_tab_2)[7];
+static transform_func _XFORMAPI TAG(transform_tab_3)[7];
+static transform_func _XFORMAPI TAG(transform_tab_4)[7];
+
+/* Similar functions could be called several times, with more highly
+ * optimized routines overwriting the arrays.  This only occurs during
+ * startup.
+ */
+static void _XFORMAPI TAG(init_c_transformations)( void )
+{
+#define TAG_TAB   gl_transform_tab[IDX]
+#define TAG_TAB_1 TAG(transform_tab_1)
+#define TAG_TAB_2 TAG(transform_tab_2)
+#define TAG_TAB_3 TAG(transform_tab_3)
+#define TAG_TAB_4 TAG(transform_tab_4)
+
+   TAG_TAB[1] = TAG_TAB_1;
+   TAG_TAB[2] = TAG_TAB_2;
+   TAG_TAB[3] = TAG_TAB_3;
+   TAG_TAB[4] = TAG_TAB_4;
+
+   /* 1-D points (ie texcoords) */
+   TAG_TAB_1[MATRIX_GENERAL]     = TAG(transform_points1_general);
+   TAG_TAB_1[MATRIX_IDENTITY]    = TAG(transform_points1_identity);
+   TAG_TAB_1[MATRIX_3D_NO_ROT]   = TAG(transform_points1_3d_no_rot);
+   TAG_TAB_1[MATRIX_PERSPECTIVE] = TAG(transform_points1_perspective) ;
+   TAG_TAB_1[MATRIX_2D]          = TAG(transform_points1_2d);
+   TAG_TAB_1[MATRIX_2D_NO_ROT]   = TAG(transform_points1_2d_no_rot);
+   TAG_TAB_1[MATRIX_3D]          = TAG(transform_points1_3d);
+
+   /* 2-D points */
+   TAG_TAB_2[MATRIX_GENERAL]     = TAG(transform_points2_general);
+   TAG_TAB_2[MATRIX_IDENTITY]    = TAG(transform_points2_identity);
+   TAG_TAB_2[MATRIX_3D_NO_ROT]   = TAG(transform_points2_3d_no_rot);
+   TAG_TAB_2[MATRIX_PERSPECTIVE] = TAG(transform_points2_perspective) ;
+   TAG_TAB_2[MATRIX_2D]          = TAG(transform_points2_2d);
+   TAG_TAB_2[MATRIX_2D_NO_ROT]   = TAG(transform_points2_2d_no_rot);
+   TAG_TAB_2[MATRIX_3D]          = TAG(transform_points2_3d);
+
+   /* 3-D points */
+   TAG_TAB_3[MATRIX_GENERAL]     = TAG(transform_points3_general);
+   TAG_TAB_3[MATRIX_IDENTITY]    = TAG(transform_points3_identity);
+   TAG_TAB_3[MATRIX_3D_NO_ROT]   = TAG(transform_points3_3d_no_rot);
+   TAG_TAB_3[MATRIX_PERSPECTIVE] = TAG(transform_points3_perspective);
+   TAG_TAB_3[MATRIX_2D]          = TAG(transform_points3_2d);
+   TAG_TAB_3[MATRIX_2D_NO_ROT]   = TAG(transform_points3_2d_no_rot);
+   TAG_TAB_3[MATRIX_3D]          = TAG(transform_points3_3d);
+
+   /* 4-D points */
+   TAG_TAB_4[MATRIX_GENERAL]     = TAG(transform_points4_general);
+   TAG_TAB_4[MATRIX_IDENTITY]    = TAG(transform_points4_identity);
+   TAG_TAB_4[MATRIX_3D_NO_ROT]   = TAG(transform_points4_3d_no_rot);
+   TAG_TAB_4[MATRIX_PERSPECTIVE] = TAG(transform_points4_perspective);
+   TAG_TAB_4[MATRIX_2D]          = TAG(transform_points4_2d);
+   TAG_TAB_4[MATRIX_2D_NO_ROT]   = TAG(transform_points4_2d_no_rot);
+   TAG_TAB_4[MATRIX_3D]          = TAG(transform_points4_3d);
+
+#undef TAG_TAB
+}