Consolidation of asm code in 3.5

25 files changed, 7032 insertions, 342 deletions

diff --git a/Make-config b/Make-config
index d83f79c5686..3226209644e 100644
--- a/Make-config
+++ b/Make-config
@@ -1,4 +1,4 @@
-# $Id: Make-config,v 1.40 2001/03/29 03:41:39 gareth Exp $
+# $Id: Make-config,v 1.41 2001/03/29 06:46:15 gareth Exp $
 
 MESA_MAJOR=3
 MESA_MINOR=5
@@ -220,9 +220,7 @@ freebsd-386:
 	"MAKELIB = ../bin/mklib.freebsd" \
 	"APP_LIB_DEPS = -L/usr/X11R6/lib -lXext -lXmu -lXi -lX11 -lm" \
 	"ASM_SOURCES = X86/common_x86_asm.S X86/glapi_x86.S \
-		X86/x86_xform_raw2.S X86/x86_xform_masked2.S \
-		X86/x86_xform_raw3.S X86/x86_xform_masked3.S \
-		X86/x86_xform_raw4.S X86/x86_xform_masked4.S \
+		X86/x86_xform2.S X86/x86_xform3.S X86/x86_xform4.S \
 		X86/x86_cliptest.S"
 
 gcc:
@@ -556,26 +554,22 @@ linux-x86:
 	"OSMESA_LIB = libOSMesa.so" \
 	"CC = gcc -malign-loops=2 -malign-jumps=2 -malign-functions=2" \
 	"CPLUSPLUS = g++" \
-	"CFLAGS = -Wall -O3 -ansi -pedantic -fPIC -ffast-math -funroll-loops -fomit-frame-pointer -D_POSIX_SOURCE -D_POSIX_C_SOURCE=199309L -D_SVID_SOURCE -D_BSD_SOURCE -DUSE_XSHM -DUSE_X86_ASM -DUSE_MMX_ASM -DUSE_3DNOW_ASM -DUSE_KATMAI_ASM -DPTHREADS -I/usr/X11R6/include" \
+	"CFLAGS = -Wall -O3 -ansi -pedantic -fPIC -ffast-math -funroll-loops -fomit-frame-pointer -D_POSIX_SOURCE -D_POSIX_C_SOURCE=199309L -D_SVID_SOURCE -D_BSD_SOURCE -DUSE_XSHM -DUSE_X86_ASM -DUSE_MMX_ASM -DUSE_3DNOW_ASM -DUSE_SSE_ASM -DPTHREADS -I/usr/X11R6/include" \
 	"MAKELIB = ../bin/mklib.linux" \
 	"GL_LIB_DEPS = -L/usr/X11R6/lib -lX11 -lXext -lm -lm -lpthread" \
 	"GLU_LIB_DEPS = -L../lib -lGL -lm" \
 	"GLUT_LIB_DEPS = -L../lib -lGLU -lGL -L/usr/X11R6/lib -lX11 -lXmu -lXt -lXi -lm" \
 	"APP_LIB_DEPS = -lm" \
 	"ASM_SOURCES = X86/common_x86_asm.S X86/glapi_x86.S \
-		X86/x86_xform_raw2.S X86/x86_xform_masked2.S \
-		X86/x86_xform_raw3.S X86/x86_xform_masked3.S \
-		X86/x86_xform_raw4.S X86/x86_xform_masked4.S \
+		X86/x86_xform2.S X86/x86_xform3.S X86/x86_xform4.S \
 		X86/x86_cliptest.S X86/x86_vertex.S \
 		X86/mmx_blend.S \
-		X86/3dnow_xform_raw2.S X86/3dnow_xform_masked2.S \
-		X86/3dnow_xform_raw3.S X86/3dnow_xform_masked3.S \
-		X86/3dnow_xform_raw4.S X86/3dnow_xform_masked4.S \
-		X86/3dnow_norm_raw.S X86/3dnow_vertex.S \
-		X86/katmai_xform_raw2.S X86/katmai_xform_masked2.S \
-		X86/katmai_xform_raw3.S X86/katmai_xform_masked3.S \
-		X86/katmai_xform_raw4.S X86/katmai_xform_masked4.S \
-		X86/katmai_norm_raw.S X86/katmai_vertex.S"
+		X86/3dnow_xform1.S X86/3dnow_xform2.S \
+		X86/3dnow_xform3.S X86/3dnow_xform4.S \
+		X86/3dnow_normal.S X86/3dnow_vertex.S \
+		X86/sse_xform1.S X86/sse_xform2.S \
+		X86/sse_xform3.S X86/sse_xform4.S \
+		X86/sse_normal.S X86/sse_vertex.S"
 
 linux-x86-static:
 	$(MAKE) $(MFLAGS) -f Makefile.X11 targets \
@@ -586,26 +580,22 @@ linux-x86-static:
 	"OSMESA_LIB = libOSMesa.a" \
 	"CC = gcc -malign-loops=2 -malign-jumps=2 -malign-functions=2" \
 	"CPLUSPLUS = g++" \
-	"CFLAGS = -Wall -O3 -ansi -pedantic -fPIC -ffast-math -funroll-loops -fomit-frame-pointer -D_POSIX_SOURCE -D_POSIX_C_SOURCE=199309L -D_SVID_SOURCE -D_BSD_SOURCE -DUSE_XSHM -DUSE_X86_ASM -DUSE_MMX_ASM -DUSE_3DNOW_ASM -DUSE_KATMAI_ASM -DPTHREADS -I/usr/X11R6/include" \
+	"CFLAGS = -Wall -O3 -ansi -pedantic -fPIC -ffast-math -funroll-loops -fomit-frame-pointer -D_POSIX_SOURCE -D_POSIX_C_SOURCE=199309L -D_SVID_SOURCE -D_BSD_SOURCE -DUSE_XSHM -DUSE_X86_ASM -DUSE_MMX_ASM -DUSE_3DNOW_ASM -DUSE_SSE_ASM -DPTHREADS -I/usr/X11R6/include" \
 	"MAKELIB = ../bin/mklib.ar-ruv" \
 	"GL_LIB_DEPS = -L/usr/X11R6/lib -lX11 -lXext -lm -lm -lpthread" \
 	"GLU_LIB_DEPS = -L../lib -lGL -lm" \
 	"GLUT_LIB_DEPS = -L../lib -lGLU -lGL -L/usr/X11R6/lib -lX11 -lXmu -lXt -lXi -lm" \
 	"APP_LIB_DEPS = -lm" \
 	"ASM_SOURCES = X86/common_x86_asm.S X86/glapi_x86.S \
-		X86/x86_xform_raw2.S X86/x86_xform_masked2.S \
-		X86/x86_xform_raw3.S X86/x86_xform_masked3.S \
-		X86/x86_xform_raw4.S X86/x86_xform_masked4.S \
+		X86/x86_xform2.S X86/x86_xform3.S X86/x86_xform_raw4.S \
 		X86/x86_cliptest.S X86/x86_vertex.S \
 		X86/mmx_blend.S \
-		X86/3dnow_xform_raw2.S X86/3dnow_xform_masked2.S \
-		X86/3dnow_xform_raw3.S X86/3dnow_xform_masked3.S \
-		X86/3dnow_xform_raw4.S X86/3dnow_xform_masked4.S \
-		X86/3dnow_norm_raw.S X86/3dnow_vertex.S \
-		X86/katmai_xform_raw2.S X86/katmai_xform_masked2.S \
-		X86/katmai_xform_raw3.S X86/katmai_xform_masked3.S \
-		X86/katmai_xform_raw4.S X86/katmai_xform_masked4.S \
-		X86/katmai_norm_raw.S X86/katmai_vertex.S"
+		X86/3dnow_xform1.S X86/3dnow_xform2.S \
+		X86/3dnow_xform3.S X86/3dnow_xform4.S \
+		X86/3dnow_normal.S X86/3dnow_vertex.S \
+		X86/sse_xform1.S X86/sse_xform2.S \
+		X86/sse_xform3.S X86/sse_xform4.S \
+		X86/sse_normal.S X86/sse_vertex.S"
 
 # Contributed by Uwe_Maurer@t-online.de
 linux-ggi:
@@ -637,9 +627,7 @@ linux-386-ggi:
 	"MAKELIB = ../bin/mklib.ggi" \
 	"APP_LIB_DEPS = -L/usr/X11R6/lib -lX11 -lXext -lXmu -lXt -lXi -lggi -lm" \
 	"ASM_SOURCES = X86/common_x86_asm.S X86/glapi_x86.S \
-		X86/x86_xform_raw2.S X86/x86_xform_masked3.S \
-		X86/x86_xform_raw3.S X86/x86_xform_masked2.S \
-		X86/x86_xform_raw4.S X86/x86_xform_masked4.S \
+		X86/x86_xform2.S X86/x86_xform3.S X86/x86_xform4.S \
 		X86/x86_cliptest.S X86/x86_vertex.S"
 
 linux-glide:
@@ -676,9 +664,7 @@ linux-386-glide:
 	"GLUT_LIB_DEPS = -L../lib -lGLU -lGL -L/usr/X11R6/lib -lX11 -lXmu -lXt -lXi -lm" \
 	"APP_LIB_DEPS = -lm" \
 	"ASM_SOURCES = X86/common_x86_asm.S X86/glapi_x86.S \
-		X86/x86_xform_raw2.S X86/x86_xform_masked2.S \
-		X86/x86_xform_raw3.S X86/x86_xform_masked3.S \
-		X86/x86_xform_raw4.S X86/x86_xform_masked4.S \
+		X86/x86_xform2.S X86/x86_xform3.S X86/x86_xform4.S \
 		X86/x86_cliptest.S X86/x86_vertex.S"
 
 linux-386-opt-glide:
@@ -698,9 +684,7 @@ linux-386-opt-glide:
 	"GLUT_LIB_DEPS = -L../lib -lGLU -lGL -L/usr/X11R6/lib -lX11 -lXmu -lXt -lXi -lm" \
 	"APP_LIB_DEPS = -lm" \
 	"ASM_SOURCES = X86/common_x86_asm.S X86/glapi_x86.S \
-		X86/x86_xform_raw2.S X86/x86_xform_masked2.S \
-		X86/x86_xform_raw3.S X86/x86_xform_masked3.S \
-		X86/x86_xform_raw4.S X86/x86_xform_masked4.S \
+		X86/x86_xform2.S X86/x86_xform3.S X86/x86_xform4.S \
 		X86/x86_cliptest.S X86/x86_vertex.S"
 
 linux-386-opt-V2-glide:
@@ -720,9 +704,7 @@ linux-386-opt-V2-glide:
 	"GLUT_LIB_DEPS = -L../lib -lGLU -lGL -L/usr/X11R6/lib -lX11 -lXmu -lXt -lXi -lm" \
 	"APP_LIB_DEPS = -lm" \
 	"ASM_SOURCES = X86/common_x86_asm.S X86/glapi_x86.S \
-		X86/x86_xform_raw2.S X86/x86_xform_masked2.S \
-		X86/x86_xform_raw3.S X86/x86_xform_masked3.S \
-		X86/x86_xform_raw4.S X86/x86_xform_masked4.S \
+		X86/x86_xform2.S X86/x86_xform3.S X86/x86_xform4.S \
 		X86/x86_cliptest.S X86/x86_vertex.S"
 
 linux-x86-glide:
@@ -734,27 +716,23 @@ linux-x86-glide:
 	"OSMESA_LIB = libOSMesa.so" \
 	"CC = gcc -malign-loops=2 -malign-jumps=2 -malign-functions=2" \
 	"CPLUSPLUS = g++" \
-	"CFLAGS = -O3 -ansi -pedantic -fPIC -ffast-math -funroll-loops -fomit-frame-pointer -D_SVID_SOURCE -D_BSD_SOURCE -DUSE_X86_ASM -DUSE_MMX_ASM -DUSE_3DNOW_ASM -DUSE_KATMAI_ASM -DUSE_XSHM -DFX -DPTHREADS -I/usr/X11R6/include -I/usr/include/glide -I/usr/local/glide/include -I/usr/src/mesa-glx/src/FX/X86" \
+	"CFLAGS = -O3 -ansi -pedantic -fPIC -ffast-math -funroll-loops -fomit-frame-pointer -D_SVID_SOURCE -D_BSD_SOURCE -DUSE_X86_ASM -DUSE_MMX_ASM -DUSE_3DNOW_ASM -DUSE_SSE_ASM -DUSE_XSHM -DFX -DPTHREADS -I/usr/X11R6/include -I/usr/include/glide -I/usr/local/glide/include -I/usr/src/mesa-glx/src/FX/X86" \
 	"MAKELIB = ../bin/mklib.linux" \
 	"GL_LIB_DEPS = -L/usr/X11R6/lib -lX11 -lXext -lm -L/usr/local/glide/lib -lglide2x -lm -lpthread" \
 	"GLU_LIB_DEPS = -L../lib -lGL -lm" \
 	"GLUT_LIB_DEPS = -L../lib -lGLU -lGL -L/usr/X11R6/lib -lX11 -lXmu -lXt -lXi -lm" \
 	"APP_LIB_DEPS = -lm" \
 	"ASM_SOURCES = X86/common_x86_asm.S X86/glapi_x86.S \
-		X86/x86_xform_raw2.S X86/x86_xform_masked2.S \
-		X86/x86_xform_raw3.S X86/x86_xform_masked3.S \
-		X86/x86_xform_raw4.S X86/x86_xform_masked4.S \
+		X86/x86_xform2.S X86/x86_xform3.S X86/x86_xform4.S \
 		X86/x86_cliptest.S X86/x86_vertex.S \
 		X86/mmx_blend.S \
-		X86/3dnow_xform_raw2.S X86/3dnow_xform_raw2.S \
-		X86/3dnow_xform_raw3.S X86/3dnow_xform_raw3.S \
-		X86/3dnow_xform_raw4.S X86/3dnow_xform_raw4.S \
-		X86/3dnow_norm_raw.S X86/3dnow_vertex.S \
+		X86/3dnow_xform1.S X86/3dnow_xform2.S \
+		X86/3dnow_xform3.S X86/3dnow_xform4.S \
+		X86/3dnow_normal.S X86/3dnow_vertex.S \
 		FX/X86/fx_3dnow_fastpath.S \
-		X86/katmai_xform_raw2.S X86/katmai_xform_masked2.S \
-		X86/katmai_xform_raw3.S X86/katmai_xform_masked3.S \
-		X86/katmai_xform_raw4.S X86/katmai_xform_masked4.S \
-		X86/katmai_norm_raw.S X86/katmai_vertex.S"
+		X86/sse_xform1.S X86/sse_xform2.S \
+		X86/sse_xform3.S X86/sse_xform4.S \
+		X86/sse_normal.S X86/sse_vertex.S"
 
 linux-alpha:
 	$(MAKE) $(MFLAGS) -f Makefile.X11 targets \
@@ -999,9 +977,7 @@ os2-x11:
 	"MAKELIB = ..\\bin\\mklib-emx.cmd " \
 	"APP_LIB_DEPS = -Zmt -Zcrtdll -Zexe -L$(X11ROOT)/XFree86/lib -lXt -lX11 -lm" \
 	"ASM_SOURCES = X86/common_x86_asm.S X86/glapi_x86.S \
-		X86/x86_xform_raw2.S X86/x86_xform_masked2.S \
-		X86/x86_xform_raw3.S X86/x86_xform_masked3.S \
-		X86/x86_xform_raw4.S X86/x86_xform_masked4.S \
+		X86/x86_xform2.S X86/x86_xform3.S X86/x86_xform4.S \
 		X86/x86_cliptest.S X86/x86_vertex.S"
 
 osf1:
@@ -1503,26 +1479,22 @@ linux-x86-debug:
 	"OSMESA_LIB = libOSMesa.so" \
 	"CC = gcc -malign-loops=2 -malign-jumps=2 -malign-functions=2" \
 	"CPLUSPLUS = g++" \
-	"CFLAGS = -O2 -g -ansi -pedantic -Wall -Wmissing-prototypes -fPIC -ffast-math -D_POSIX_SOURCE -D_POSIX_C_SOURCE=199309L -D_SVID_SOURCE -D_BSD_SOURCE -DUSE_XSHM -DUSE_X86_ASM -DUSE_MMX_ASM -DUSE_3DNOW_ASM -DUSE_KATMAI_ASM -DPTHREADS -I/usr/X11R6/include -DDEBUG -DMESA_DEBUG" \
+	"CFLAGS = -O2 -g -ansi -pedantic -Wall -Wmissing-prototypes -fPIC -ffast-math -D_POSIX_SOURCE -D_POSIX_C_SOURCE=199309L -D_SVID_SOURCE -D_BSD_SOURCE -DUSE_XSHM -DUSE_X86_ASM -DUSE_MMX_ASM -DUSE_3DNOW_ASM -DUSE_SSE_ASM -DPTHREADS -I/usr/X11R6/include -DDEBUG -DMESA_DEBUG" \
 	"MAKELIB = ../bin/mklib.linux" \
 	"GL_LIB_DEPS = -L/usr/X11R6/lib -lX11 -lXext -lm -lm -lpthread" \
 	"GLU_LIB_DEPS = -L../lib -lGL -lm" \
 	"GLUT_LIB_DEPS = -L../lib -lGLU -lGL -L/usr/X11R6/lib -lX11 -lXmu -lXt -lXi -lm" \
 	"APP_LIB_DEPS = -lm" \
 	"ASM_SOURCES = X86/common_x86_asm.S X86/glapi_x86.S \
-		X86/x86_xform_raw2.S X86/x86_xform_masked2.S \
-		X86/x86_xform_raw3.S X86/x86_xform_masked3.S \
-		X86/x86_xform_raw4.S X86/x86_xform_masked4.S \
+		X86/x86_xform2.S X86/x86_xform3.S X86/x86_xform4.S \
 		X86/x86_cliptest.S X86/x86_vertex.S \
 		X86/mmx_blend.S \
-		X86/3dnow_xform_raw2.S X86/3dnow_xform_masked2.S \
-		X86/3dnow_xform_raw3.S X86/3dnow_xform_masked3.S \
-		X86/3dnow_xform_raw4.S X86/3dnow_xform_masked4.S \
-		X86/3dnow_norm_raw.S X86/3dnow_vertex.S \
-		X86/katmai_xform_raw2.S X86/katmai_xform_masked2.S \
-		X86/katmai_xform_raw3.S X86/katmai_xform_masked3.S \
-		X86/katmai_xform_raw4.S X86/katmai_xform_masked4.S \
-		X86/katmai_norm_raw.S X86/katmai_vertex.S"
+		X86/3dnow_xform1.S X86/3dnow_xform2.S \
+		X86/3dnow_xform3.S X86/3dnow_xform4.S \
+		X86/3dnow_normal.S X86/3dnow_vertex.S \
+		X86/sse_xform1.S X86/sse_xform2.S \
+		X86/sse_xform3.S X86/sse_xform4.S \
+		X86/sse_normal.S X86/sse_vertex.S"
 
 linux-glide-debug:
 	$(MAKE) $(MFLAGS) -f Makefile.X11 targets \
@@ -1548,25 +1520,21 @@ linux-prof:
 	"OSMESA_LIB = libOSMesa.a" \
 	"CC = gcc" \
 	"CPLUSPLUS = g++" \
-	"CFLAGS = -Wall -O2 -pg -ansi -pedantic -ffast-math -DUSE_XSHM -D_SVID_SOURCE -D_BSD_SOURCE -DUSE_X86_ASM -DUSE_MMX_ASM -DUSE_3DNOW_ASM -DUSE_KATMAI_ASM" \
+	"CFLAGS = -Wall -O2 -pg -ansi -pedantic -ffast-math -DUSE_XSHM -D_SVID_SOURCE -D_BSD_SOURCE -DUSE_X86_ASM -DUSE_MMX_ASM -DUSE_3DNOW_ASM -DUSE_SSE_ASM" \
 	"CCFLAGS = $(CFLAGS)" \
 	"MAKELIB = ../bin/mklib.ar-ruv" \
 	"APP_LIB_DEPS = -L/usr/X11/lib -lX11 -lXext -lXmu -lXt -lXi -lSM -lICE -lm" \
 	"ASM_SOURCES = X86/common_x86_asm.S X86/glapi_x86.S \
-		X86/x86_xform_raw2.S X86/x86_xform_masked2.S \
-		X86/x86_xform_raw2.S X86/x86_xform_masked3.S \
-		X86/x86_xform_raw4.S X86/x86_xform_masked4.S \
+		X86/x86_xform2.S X86/x86_xform3.S X86/x86_xform4.S \
 		X86/x86_cliptest.S X86/x86_vertex.S \
 		X86/mmx_blend.S \
-		X86/3dnow_xform_raw2.S X86/3dnow_xform_masked2.S \
-		X86/3dnow_xform_raw3.S X86/3dnow_xform_masked3.S \
-		X86/3dnow_xform_raw4.S X86/3dnow_xform_masked4.S \
-		X86/3dnow_norm_raw.S X86/3dnow_vertex.S \
+		X86/3dnow_xform1.S X86/3dnow_xform2.S \
+		X86/3dnow_xform3.S X86/3dnow_xform4.S \
+		X86/3dnow_normal.S X86/3dnow_vertex.S \
 		FX/X86/fx_3dnow_fastpath.S \
-		X86/katmai_xform_raw2.S X86/katmai_xform_masked2.S \
-		X86/katmai_xform_raw3.S X86/katmai_xform_masked3.S \
-		X86/katmai_xform_raw4.S X86/katmai_xform_masked4.S \
-		X86/katmai_norm_raw.S X86/katmai_vertex.S"
+		X86/sse_xform1.S X86/sse_xform2.S \
+		X86/sse_xform3.S X86/sse_xform4.S \
+		X86/sse_normal.S X86/sse_vertex.S"
 
 linux-glide-prof:
 	$(MAKE) $(MFLAGS) -f Makefile.X11 targets \
@@ -1577,22 +1545,18 @@ linux-glide-prof:
 	"OSMESA_LIB = libOSMesa.a" \
 	"CC = gcc" \
 	"CPLUSPLUS = g++" \
-	"CFLAGS = -O2 -pg -ansi -pedantic -Wall -DUSE_XSHM -D_SVID_SOURCE -D_BSD_SOURCE -DUSE_X86_ASM -DUSE_MMX_ASM -DUSE_3DNOW_ASM -DUSE_KATMAI_ASM -I/usr/include/glide -I/usr/local/glide/include" \
+	"CFLAGS = -O2 -pg -ansi -pedantic -Wall -DUSE_XSHM -D_SVID_SOURCE -D_BSD_SOURCE -DUSE_X86_ASM -DUSE_MMX_ASM -DUSE_3DNOW_ASM -DUSE_SSE_ASM -I/usr/include/glide -I/usr/local/glide/include" \
 	"CCFLAGS = $(CFLAGS)" \
 	"MAKELIB = ../bin/mklib.ar-ruv" \
 	"APP_LIB_DEPS = -L/usr/local/glide/lib -lglide2x -L/usr/X11/lib -lX11 -lXext -lXmu -lXt -lXi -lSM -lICE -lm" \
 	"ASM_SOURCES = X86/common_x86_asm.S X86/glapi_x86.S \
-		X86/x86_xform_raw2.S X86/x86_xform_masked2.S \
-		X86/x86_xform_raw3.S X86/x86_xform_masked3.S \
-		X86/x86_xform_raw4.S X86/x86_xform_masked4.S \
+		X86/x86_xform2.S X86/x86_xform3.S X86/x86_xform4.S \
 		X86/x86_cliptest.S X86/x86_vertex.S \
 		X86/mmx_blend.S \
-		X86/3dnow_xform_raw2.S X86/3dnow_xform_masked2.S \
-		X86/3dnow_xform_raw3.S X86/3dnow_xform_masked3.S \
-		X86/3dnow_xform_raw4.S X86/3dnow_xform_masked4.S \
-		X86/3dnow_norm_raw.S X86/3dnow_vertex.S \
+		X86/3dnow_xform1.S X86/3dnow_xform2.S \
+		X86/3dnow_xform3.S X86/3dnow_xform4.S \
+		X86/3dnow_normal.S X86/3dnow_vertex.S \
 		FX/X86/fx_3dnow_fastpath.S \
-		X86/katmai_xform_raw2.S X86/katmai_xform_masked2.S \
-		X86/katmai_xform_raw3.S X86/katmai_xform_masked3.S \
-		X86/katmai_xform_raw4.S X86/katmai_xform_masked4.S \
-		X86/katmai_norm_raw.S X86/katmai_vertex.S"
+		X86/sse_xform1.S X86/sse_xform2.S \
+		X86/sse_xform3.S X86/sse_xform4.S \
+		X86/sse_normal.S X86/sse_vertex.S"
diff --git a/src/mesa/Makefile.X11 b/src/mesa/Makefile.X11
index c514e927f02..7a7d18a2366 100644
--- a/src/mesa/Makefile.X11
+++ b/src/mesa/Makefile.X11
@@ -1,4 +1,4 @@
-# $Id: Makefile.X11,v 1.50 2001/03/29 03:41:40 gareth Exp $
+# $Id: Makefile.X11,v 1.51 2001/03/29 06:46:15 gareth Exp $
 
 # Mesa 3-D graphics library
 # Version:  3.5
@@ -108,7 +108,7 @@ CORE_SOURCES = \
 	X86/x86.c \
 	X86/common_x86.c \
 	X86/3dnow.c \
-	X86/katmai.c \
+	X86/sse.c \
 	math/m_debug_norm.c \
 	math/m_debug_vertex.c \
 	math/m_debug_xform.c \
@@ -280,7 +280,7 @@ X86/common_x86.o: X86/common_x86.c
 X86/common_x86_asm.o: X86/common_x86_asm.S X86/matypes.h
 X86/3dnow.o: X86/3dnow.c
 	$(CC) -c -I. -I$(INCDIR) $(CFLAGS) $< -o $@
-X86/katmai.o: X86/katmai.c
+X86/sse.o: X86/sse.c
 	$(CC) -c -I. -I$(INCDIR) $(CFLAGS) $< -o $@
 X86/matypes.h: mtypes.h tnl/t_context.h X86/gen_matypes.c
 	$(CC) -I. -I$(INCDIR) $(CFLAGS) X86/gen_matypes.c -o X86/gen_matypes
diff --git a/src/mesa/main/Makefile.X11 b/src/mesa/main/Makefile.X11
index c514e927f02..7a7d18a2366 100644
--- a/src/mesa/main/Makefile.X11
+++ b/src/mesa/main/Makefile.X11
@@ -1,4 +1,4 @@
-# $Id: Makefile.X11,v 1.50 2001/03/29 03:41:40 gareth Exp $
+# $Id: Makefile.X11,v 1.51 2001/03/29 06:46:15 gareth Exp $
 
 # Mesa 3-D graphics library
 # Version:  3.5
@@ -108,7 +108,7 @@ CORE_SOURCES = \
 	X86/x86.c \
 	X86/common_x86.c \
 	X86/3dnow.c \
-	X86/katmai.c \
+	X86/sse.c \
 	math/m_debug_norm.c \
 	math/m_debug_vertex.c \
 	math/m_debug_xform.c \
@@ -280,7 +280,7 @@ X86/common_x86.o: X86/common_x86.c
 X86/common_x86_asm.o: X86/common_x86_asm.S X86/matypes.h
 X86/3dnow.o: X86/3dnow.c
 	$(CC) -c -I. -I$(INCDIR) $(CFLAGS) $< -o $@
-X86/katmai.o: X86/katmai.c
+X86/sse.o: X86/sse.c
 	$(CC) -c -I. -I$(INCDIR) $(CFLAGS) $< -o $@
 X86/matypes.h: mtypes.h tnl/t_context.h X86/gen_matypes.c
 	$(CC) -I. -I$(INCDIR) $(CFLAGS) X86/gen_matypes.c -o X86/gen_matypes
diff --git a/src/mesa/math/m_debug_norm.c b/src/mesa/math/m_debug_norm.c
index a6f0cb76a75..45532475b3a 100644
--- a/src/mesa/math/m_debug_norm.c
+++ b/src/mesa/math/m_debug_norm.c
@@ -1,4 +1,4 @@
-/* $Id: m_debug_norm.c,v 1.5 2001/03/12 00:48:41 gareth Exp $ */
+/* $Id: m_debug_norm.c,v 1.6 2001/03/29 06:46:27 gareth Exp $ */
 
 /*
  * Mesa 3-D graphics library
@@ -186,8 +186,7 @@ static void ref_norm_transform_normalize( const GLmatrix *mat,
  * Normal transformation tests
  */
 
-static int test_norm_function( normal_func func, int mtype,
-			       int masked, long *cycles )
+static int test_norm_function( normal_func func, int mtype, long *cycles )
 {
    GLvector3f source[1], dest[1], dest2[1], ref[1], ref2[1];
    GLmatrix mat[1];
@@ -195,7 +194,6 @@ static int test_norm_function( normal_func func, int mtype,
    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_DEBUG_BENCHMARK
    int cycle_i;		/* the counter for the benchmarks we run */
@@ -231,7 +229,6 @@ static int test_norm_function( normal_func func, int mtype,
    }
 
    for ( i = 0 ; i < TEST_COUNT ; i++ ) {
-      mask[i] = i % 2;				/* mask every 2nd element */
       ASSIGN_3V( d[i],  0.0, 0.0, 0.0 );
       ASSIGN_3V( s[i],  0.0, 0.0, 0.0 );
       ASSIGN_3V( d2[i], 0.0, 0.0, 0.0 );
@@ -278,31 +275,16 @@ static int test_norm_function( normal_func func, int mtype,
    }
 
    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 );
-      }
+      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 );
-      }
+      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" );
@@ -344,7 +326,6 @@ static int test_norm_function( normal_func func, int mtype,
 
 void _math_test_all_normal_transform_functions( char *description )
 {
-   int masked;
    int mtype;
    long benchmark_tab[0xf][0x4];
    static int first_time = 1;
@@ -362,46 +343,33 @@ void _math_test_all_normal_transform_functions( char *description )
       }
       printf( "normal transform results after hooking in %s functions:\n",
 	      description );
+      printf( "\n-------------------------------------------------------\n" );
    }
 #endif
 
-   for ( masked = 0 ; masked <= 1 ; masked++ ) {
-      int cma = masked ? 1 : 0;
-      char *cmastring = masked ? "CULL_MASK_ACTIVE" : "0";
+   for ( mtype = 0 ; mtype < 8 ; mtype++ ) {
+      normal_func func = _mesa_normal_tab[norm_types[mtype]][0];
+      long *cycles = &(benchmark_tab[mtype][0]);
 
-#ifdef RUN_DEBUG_BENCHMARK
-      if ( mesa_profile ) {
-         printf( "\n culling: %s \n", masked ? "CULL_MASK_ACTIVE" : "0" );
-         printf( "\n-------------------------------------------------------\n" );
+      if ( test_norm_function( func, mtype, cycles ) == 0 ) {
+	 char buf[100];
+	 sprintf( buf, "_mesa_normal_tab[0][%s] failed test (%s)",
+		  norm_strings[mtype], description );
+	 _mesa_problem( NULL, buf );
       }
-#endif
-
-      for ( mtype = 0 ; mtype < 8 ; mtype++ ) {
-         normal_func func = _mesa_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, "_mesa_normal_tab[%s][%s] failed test (%s)",
-                     cmastring, norm_strings[mtype], description );
-            _mesa_problem( NULL, buf );
-	 }
 
 #ifdef RUN_DEBUG_BENCHMARK
-         if ( mesa_profile ) {
-            printf( " %li\t", benchmark_tab[mtype][cma] );
-            printf( " | [%s]\n", norm_strings[mtype] );
-         }
-      }
-      if ( mesa_profile )
-	 printf( "\n" );
-#else
+      if ( mesa_profile ) {
+	 printf( " %li\t", benchmark_tab[mtype][0] );
+	 printf( " | [%s]\n", norm_strings[mtype] );
       }
 #endif
    }
 #ifdef RUN_DEBUG_BENCHMARK
-   if ( mesa_profile )
+   if ( mesa_profile ) {
+      printf( "\n" );
       fflush( stdout );
+   }
 #endif
 }
 
diff --git a/src/mesa/math/m_debug_xform.c b/src/mesa/math/m_debug_xform.c
index 9636fc439d2..6e19adea384 100644
--- a/src/mesa/math/m_debug_xform.c
+++ b/src/mesa/math/m_debug_xform.c
@@ -1,4 +1,4 @@
-/* $Id: m_debug_xform.c,v 1.6 2001/03/12 02:02:36 gareth Exp $ */
+/* $Id: m_debug_xform.c,v 1.7 2001/03/29 06:46:27 gareth Exp $ */
 
 /*
  * Mesa 3-D graphics library
@@ -159,8 +159,8 @@ static GLfloat s[TEST_COUNT][4] ALIGN16;
 static GLfloat d[TEST_COUNT][4] ALIGN16;
 static GLfloat r[TEST_COUNT][4] ALIGN16;
 
-static int test_transform_function( transform_func func, int psize, int mtype,
-                                    int masked, long *cycles )
+static int test_transform_function( transform_func func, int psize,
+				    int mtype, long *cycles )
 {
    GLvector4f source[1], dest[1], ref[1];
    GLmatrix mat[1];
@@ -238,28 +238,15 @@ static int test_transform_function( transform_func func, int psize, int mtype,
    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 );
-     }
+      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 );
-      }
+      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" );
@@ -287,7 +274,7 @@ static int test_transform_function( transform_func func, int psize, int mtype,
 
 void _math_test_all_transform_functions( char *description )
 {
-   int masked, psize, mtype;
+   int psize, mtype;
    long benchmark_tab[2][4][7];
    static int first_time = 1;
 
@@ -306,47 +293,41 @@ void _math_test_all_transform_functions( char *description )
    }
 #endif
 
-   for ( masked = 0 ; masked <= 1 ; masked++ ) {
-      int cma = masked ? 1 : 0;
-      char *cmastring = masked ? "CULL_MASK_ACTIVE" : "0";
-
 #ifdef RUN_DEBUG_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" );
+   if ( mesa_profile ) {
+      printf( "\n" );
+      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 = _mesa_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, "_mesa_transform_tab[%s][%d][%s] failed test (%s)",
-                        cmastring, psize, mstrings[mtype], description );
-               _mesa_problem( NULL, buf );
-            }
+   for ( mtype = 0 ; mtype < 7 ; mtype++ ) {
+      for ( psize = 1 ; psize <= 4 ; psize++ ) {
+	 transform_func func = _mesa_transform_tab[0][psize][mtypes[mtype]];
+	 long *cycles = &(benchmark_tab[0][psize-1][mtype]);
+
+	 if ( test_transform_function( func, psize, mtype, cycles ) == 0 ) {
+	    char buf[100];
+	    sprintf( buf, "_mesa_transform_tab[0][%d][%s] failed test (%s)",
+		     psize, mstrings[mtype], description );
+	    _mesa_problem( NULL, buf );
+	 }
 #ifdef RUN_DEBUG_BENCHMARK
-            if ( mesa_profile )
-               printf( " %li\t", benchmark_tab[cma][psize-1][mtype] );
-#endif
-         }
-#ifdef RUN_DEBUG_BENCHMARK
-         if ( mesa_profile )
-            printf( " | [%s]\n", mstrings[mtype] );
+	 if ( mesa_profile )
+	    printf( " %li\t", benchmark_tab[0][psize-1][mtype] );
 #endif
       }
 #ifdef RUN_DEBUG_BENCHMARK
       if ( mesa_profile )
-         printf( "\n" );
+	 printf( " | [%s]\n", mstrings[mtype] );
 #endif
    }
+#ifdef RUN_DEBUG_BENCHMARK
+   if ( mesa_profile )
+      printf( "\n" );
+#endif
 }
 
 
diff --git a/src/mesa/x86/3dnow.c b/src/mesa/x86/3dnow.c
index b6ef3739361..821a33d76cd 100644
--- a/src/mesa/x86/3dnow.c
+++ b/src/mesa/x86/3dnow.c
@@ -1,4 +1,4 @@
-/* $Id: 3dnow.c,v 1.17 2001/03/28 20:44:43 gareth Exp $ */
+/* $Id: 3dnow.c,v 1.18 2001/03/29 06:46:15 gareth Exp $ */
 
 /*
  * Mesa 3-D graphics library
@@ -51,31 +51,31 @@
 			const GLubyte flag
 
 
-#define DECLARE_XFORM_GROUP( pfx, sz, masked ) \
-extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_general_##masked( XFORM_ARGS );	\
-extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_identity_##masked( XFORM_ARGS );	\
-extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_3d_no_rot_##masked( XFORM_ARGS );	\
-extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_perspective_##masked( XFORM_ARGS );	\
-extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_2d_##masked( XFORM_ARGS );		\
-extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_2d_no_rot_##masked( XFORM_ARGS );	\
-extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_3d_##masked( XFORM_ARGS );
+#define DECLARE_XFORM_GROUP( pfx, sz ) \
+extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_general( XFORM_ARGS );		\
+extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_identity( XFORM_ARGS );	\
+extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_3d_no_rot( XFORM_ARGS );	\
+extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_perspective( XFORM_ARGS );	\
+extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_2d( XFORM_ARGS );		\
+extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_2d_no_rot( XFORM_ARGS );	\
+extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_3d( XFORM_ARGS );
 
 
-#define ASSIGN_XFORM_GROUP( pfx, cma, sz, masked )			\
-   _mesa_transform_tab[cma][sz][MATRIX_GENERAL] =			\
-      _mesa_##pfx##_transform_points##sz##_general_##masked;		\
-   _mesa_transform_tab[cma][sz][MATRIX_IDENTITY] =			\
-      _mesa_##pfx##_transform_points##sz##_identity_##masked;		\
-   _mesa_transform_tab[cma][sz][MATRIX_3D_NO_ROT] =			\
-      _mesa_##pfx##_transform_points##sz##_3d_no_rot_##masked;		\
-   _mesa_transform_tab[cma][sz][MATRIX_PERSPECTIVE] =			\
-      _mesa_##pfx##_transform_points##sz##_perspective_##masked;	\
-   _mesa_transform_tab[cma][sz][MATRIX_2D] =				\
-      _mesa_##pfx##_transform_points##sz##_2d_##masked;			\
-   _mesa_transform_tab[cma][sz][MATRIX_2D_NO_ROT] =			\
-      _mesa_##pfx##_transform_points##sz##_2d_no_rot_##masked;		\
-   _mesa_transform_tab[cma][sz][MATRIX_3D] =				\
-      _mesa_##pfx##_transform_points##sz##_3d_##masked;
+#define ASSIGN_XFORM_GROUP( pfx, sz )					\
+   _mesa_transform_tab[0][sz][MATRIX_GENERAL] =				\
+      _mesa_##pfx##_transform_points##sz##_general;			\
+   _mesa_transform_tab[0][sz][MATRIX_IDENTITY] =			\
+      _mesa_##pfx##_transform_points##sz##_identity;			\
+   _mesa_transform_tab[0][sz][MATRIX_3D_NO_ROT] =			\
+      _mesa_##pfx##_transform_points##sz##_3d_no_rot;			\
+   _mesa_transform_tab[0][sz][MATRIX_PERSPECTIVE] =			\
+      _mesa_##pfx##_transform_points##sz##_perspective;			\
+   _mesa_transform_tab[0][sz][MATRIX_2D] =				\
+      _mesa_##pfx##_transform_points##sz##_2d;				\
+   _mesa_transform_tab[0][sz][MATRIX_2D_NO_ROT] =			\
+      _mesa_##pfx##_transform_points##sz##_2d_no_rot;			\
+   _mesa_transform_tab[0][sz][MATRIX_3D] =				\
+      _mesa_##pfx##_transform_points##sz##_3d;
 
 
 
@@ -87,47 +87,42 @@ extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_3d_##masked( XFORM_ARGS
 			GLvector3f *dest
 
 
-#define DECLARE_NORM_GROUP( pfx, masked ) \
-extern void _ASMAPI _mesa_##pfx##_rescale_normals_##masked( NORM_ARGS );			\
-extern void _ASMAPI _mesa_##pfx##_normalize_normals_##masked( NORM_ARGS );			\
-extern void _ASMAPI _mesa_##pfx##_transform_normals_##masked( NORM_ARGS );			\
-extern void _ASMAPI _mesa_##pfx##_transform_normals_no_rot_##masked( NORM_ARGS );		\
-extern void _ASMAPI _mesa_##pfx##_transform_rescale_normals_##masked( NORM_ARGS );		\
-extern void _ASMAPI _mesa_##pfx##_transform_rescale_normals_no_rot_##masked( NORM_ARGS );	\
-extern void _ASMAPI _mesa_##pfx##_transform_normalize_normals_##masked( NORM_ARGS );		\
-extern void _ASMAPI _mesa_##pfx##_transform_normalize_normals_no_rot_##masked( NORM_ARGS );
-
-
-#define ASSIGN_NORM_GROUP( pfx, cma, masked )				\
-   _mesa_normal_tab[NORM_RESCALE][cma] =				\
-      _mesa_##pfx##_rescale_normals_##masked;				\
-   _mesa_normal_tab[NORM_NORMALIZE][cma] =				\
-      _mesa_##pfx##_normalize_normals_##masked;				\
-   _mesa_normal_tab[NORM_TRANSFORM][cma] =				\
-      _mesa_##pfx##_transform_normals_##masked;				\
-   _mesa_normal_tab[NORM_TRANSFORM_NO_ROT][cma] =			\
-      _mesa_##pfx##_transform_normals_no_rot_##masked;			\
-   _mesa_normal_tab[NORM_TRANSFORM | NORM_RESCALE][cma] =		\
-      _mesa_##pfx##_transform_rescale_normals_##masked;			\
-   _mesa_normal_tab[NORM_TRANSFORM_NO_ROT | NORM_RESCALE][cma] =	\
-      _mesa_##pfx##_transform_rescale_normals_no_rot_##masked;		\
-   _mesa_normal_tab[NORM_TRANSFORM | NORM_NORMALIZE][cma] =		\
-      _mesa_##pfx##_transform_normalize_normals_##masked;		\
-   _mesa_normal_tab[NORM_TRANSFORM_NO_ROT | NORM_NORMALIZE][cma] =	\
-      _mesa_##pfx##_transform_normalize_normals_no_rot_##masked;
+#define DECLARE_NORM_GROUP( pfx ) \
+extern void _ASMAPI _mesa_##pfx##_rescale_normals( NORM_ARGS );				\
+extern void _ASMAPI _mesa_##pfx##_normalize_normals( NORM_ARGS );			\
+extern void _ASMAPI _mesa_##pfx##_transform_normals( NORM_ARGS );			\
+extern void _ASMAPI _mesa_##pfx##_transform_normals_no_rot( NORM_ARGS );		\
+extern void _ASMAPI _mesa_##pfx##_transform_rescale_normals( NORM_ARGS );		\
+extern void _ASMAPI _mesa_##pfx##_transform_rescale_normals_no_rot( NORM_ARGS );	\
+extern void _ASMAPI _mesa_##pfx##_transform_normalize_normals( NORM_ARGS );		\
+extern void _ASMAPI _mesa_##pfx##_transform_normalize_normals_no_rot( NORM_ARGS );
+
+
+#define ASSIGN_NORM_GROUP( pfx )					\
+   _mesa_normal_tab[NORM_RESCALE][0] =					\
+      _mesa_##pfx##_rescale_normals;					\
+   _mesa_normal_tab[NORM_NORMALIZE][0] =				\
+      _mesa_##pfx##_normalize_normals;					\
+   _mesa_normal_tab[NORM_TRANSFORM][0] =				\
+      _mesa_##pfx##_transform_normals;					\
+   _mesa_normal_tab[NORM_TRANSFORM_NO_ROT][0] =				\
+      _mesa_##pfx##_transform_normals_no_rot;				\
+   _mesa_normal_tab[NORM_TRANSFORM|NORM_RESCALE][0] =			\
+      _mesa_##pfx##_transform_rescale_normals;				\
+   _mesa_normal_tab[NORM_TRANSFORM_NO_ROT|NORM_RESCALE][0] =		\
+      _mesa_##pfx##_transform_rescale_normals_no_rot;			\
+   _mesa_normal_tab[NORM_TRANSFORM|NORM_NORMALIZE][0] =			\
+      _mesa_##pfx##_transform_normalize_normals;			\
+   _mesa_normal_tab[NORM_TRANSFORM_NO_ROT|NORM_NORMALIZE][0] =		\
+      _mesa_##pfx##_transform_normalize_normals_no_rot;
 
 
 #ifdef USE_3DNOW_ASM
-DECLARE_XFORM_GROUP( 3dnow, 2, raw )
-DECLARE_XFORM_GROUP( 3dnow, 3, raw )
-DECLARE_XFORM_GROUP( 3dnow, 4, raw )
+DECLARE_XFORM_GROUP( 3dnow, 2 )
+DECLARE_XFORM_GROUP( 3dnow, 3 )
+DECLARE_XFORM_GROUP( 3dnow, 4 )
 
-DECLARE_XFORM_GROUP( 3dnow, 2, masked )
-DECLARE_XFORM_GROUP( 3dnow, 3, masked )
-DECLARE_XFORM_GROUP( 3dnow, 4, masked )
-
-DECLARE_NORM_GROUP( 3dnow, raw )
-/*DECLARE_NORM_GROUP( 3dnow, masked )*/
+DECLARE_NORM_GROUP( 3dnow )
 
 
 extern void _ASMAPI
@@ -155,16 +150,11 @@ _mesa_3dnow_project_clipped_vertices( GLfloat *first,
 void _mesa_init_3dnow_transform_asm( void )
 {
 #ifdef USE_3DNOW_ASM
-   ASSIGN_XFORM_GROUP( 3dnow, 0, 2, raw );
-   ASSIGN_XFORM_GROUP( 3dnow, 0, 3, raw );
-   ASSIGN_XFORM_GROUP( 3dnow, 0, 4, raw );
-
-/*     ASSIGN_XFORM_GROUP( 3dnow, CULL_MASK_ACTIVE, 2, masked ); */
-/*     ASSIGN_XFORM_GROUP( 3dnow, CULL_MASK_ACTIVE, 3, masked ); */
-/*     ASSIGN_XFORM_GROUP( 3dnow, CULL_MASK_ACTIVE, 4, masked ); */
+   ASSIGN_XFORM_GROUP( 3dnow, 2 );
+   ASSIGN_XFORM_GROUP( 3dnow, 3 );
+   ASSIGN_XFORM_GROUP( 3dnow, 4 );
 
-   ASSIGN_NORM_GROUP( 3dnow, 0, raw );
-/* ASSIGN_NORM_GROUP( 3dnow, CULL_MASK_ACTIVE, masked ); */
+   ASSIGN_NORM_GROUP( 3dnow );
 
 #ifdef DEBUG
    _math_test_all_transform_functions( "3DNow!" );
@@ -177,6 +167,7 @@ void _mesa_init_3dnow_vertex_asm( void )
 {
 #ifdef USE_3DNOW_ASM
    _mesa_xform_points3_v16_general = _mesa_v16_3dnow_general_xform;
+
    _mesa_project_v16 = _mesa_3dnow_project_vertices;
    _mesa_project_clipped_v16 = _mesa_3dnow_project_clipped_vertices;
 
diff --git a/src/mesa/x86/3dnow_normal.S b/src/mesa/x86/3dnow_normal.S
new file mode 100644
index 00000000000..f7cc069b165
--- /dev/null
+++ b/src/mesa/x86/3dnow_normal.S
@@ -0,0 +1,858 @@
+/* $Id: 3dnow_normal.S,v 1.1 2001/03/29 06:46:16 gareth Exp $ */
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  3.5
+ *
+ * Copyright (C) 1999-2001  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.
+ */
+
+/*
+ * 3Dnow assembly code by Holger Waechtler
+ */
+
+#include "matypes.h"
+#include "norm_args.h"
+
+        SEG_TEXT
+
+#define M(i)    REGOFF(i * 4, ECX)
+#define STRIDE  REGOFF(12, ESI)
+
+
+ALIGNTEXT16
+GLOBL GLNAME(_mesa_3dnow_transform_normalize_normals)
+GLNAME(_mesa_3dnow_transform_normalize_normals):
+
+ #define FRAME_OFFSET 12
+
+    PUSH_L     ( EDI )
+    PUSH_L     ( ESI )
+    PUSH_L     ( EBP )
+
+    MOV_L      ( ARG_LENGTHS, EDI )
+    MOV_L      ( ARG_IN, ESI )
+    MOV_L      ( ARG_DEST, EAX )
+    MOV_L      ( REGOFF(V3F_COUNT, ESI), EBP ) /*  dest->count = in->count   */
+    MOV_L      ( EBP, REGOFF(V3F_COUNT, EAX) )
+    MOV_L      ( REGOFF(V3F_START, ESI), EDX ) /*  in->start    */
+    MOV_L      ( REGOFF(V3F_START, EAX), EAX ) /*  dest->start  */
+    MOV_L      ( ARG_MAT, ECX )
+    MOV_L      ( REGOFF(MATRIX_INV, ECX), ECX ) /*  mat->inv     */
+
+    CMP_L      ( CONST(0), EBP )        /*   count > 0 ??  */
+    JE         ( LLBL (G3TN_end) )
+
+    MOV_L      ( REGOFF (V3F_COUNT, ESI), EBP )
+    FEMMS
+
+    PUSH_L     ( EBP )
+    PUSH_L     ( EAX )
+    PUSH_L     ( EDX )                  /*  save counter & pointer for   */
+                                        /*  the normalize pass           */
+ #undef  FRAME_OFFSET
+ #define FRAME_OFFSET 24
+
+    MOVQ       ( M(0), MM3 )            /*  m1              | m0         */
+    MOVQ       ( M(4), MM4 )            /*  m5              | m4         */
+
+    MOVD       ( M(2), MM5 )            /*                  | m2         */
+    PUNPCKLDQ  ( M(6), MM5 )            /*  m6              | m2         */
+
+    MOVQ       ( M(8), MM6 )            /*  m9              | m8         */
+    MOVQ       ( M(10), MM7 )           /*                  | m10        */
+
+    CMP_L      ( CONST(0), EDI )        /*  lengths == 0 ?                 */
+    JNE        ( LLBL (G3TN_scale_end ) )
+
+    MOVD       ( ARG_SCALE, MM0 )       /*               | scale           */
+    PUNPCKLDQ  ( MM0, MM0 )             /* scale         | scale           */
+
+    PFMUL      ( MM0, MM3 )             /* scale * m1    | scale * m0      */
+    PFMUL      ( MM0, MM4 )             /* scale * m5    | scale * m4      */
+    PFMUL      ( MM0, MM5 )             /* scale * m6    | scale * m2      */
+    PFMUL      ( MM0, MM6 )             /* scale * m9    | scale * m8      */
+    PFMUL      ( MM0, MM7 )             /*               | scale * m10     */
+
+LLBL (G3TN_scale_end):
+    MOVQ       ( REGIND (EDX), MM0 )    /*  x1              | x0         */
+    MOVD       ( REGOFF (8, EDX), MM2 ) /*                  | x2         */
+
+ALIGNTEXT32
+LLBL (G3TN_transform):
+    MOVQ       ( MM0, MM1 )             /*  x1              | x0           */
+    PUNPCKLDQ  ( MM2, MM2 )             /*  x2              | x2           */
+
+    PFMUL      ( MM3, MM0 )             /*  x1*m1           | x0*m0        */
+    ADD_L      ( CONST(12), EAX )       /*  next r                         */
+
+    PREFETCHW  ( REGIND(EAX) )
+
+    PFMUL      ( MM4, MM1 )             /*  x1*m5           | x0*m4        */
+    PFACC      ( MM1, MM0 )             /*  x0*m4+x1*m5     | x0*m0+x1*m1  */
+
+    PFMUL      ( MM5, MM2 )             /*  x2*m6           | x2*m2        */
+    PFADD      ( MM2, MM0 )             /* x0*m4+x1*m5+x2*m6| x0*m0+...+x2**/
+
+    MOVQ       ( REGIND (EDX), MM1 )    /*  x1           | x0              */
+    MOVQ       ( MM0, REGOFF(-12, EAX) ) /* write r0, r1                   */
+
+    PFMUL      ( MM6, MM1 )             /* x1*m9         | x0*m8           */
+    MOVD       ( REGOFF (8, EDX), MM2 ) /*               | x2              */
+
+    PFMUL      ( MM7, MM2 )             /*               | x2*m10          */
+    PFACC      ( MM1, MM1 )             /*  *not used*   | x0*m8+x1*m9     */
+
+    PFADD      ( MM2, MM1 )             /*  *not used*   | x0*m8+x1*m9+x2*m*/
+    ADD_L      ( STRIDE, EDX )          /*  next normal                    */
+
+    PREFETCH   ( REGIND(EDX) )
+
+    MOVD       ( MM1, REGOFF(-4, EAX) ) /*  write r2                       */
+    MOVQ       ( REGIND (EDX), MM0 )    /*  x1           | x0              */
+
+    MOVD       ( REGOFF (8, EDX), MM2 ) /*               | x2              */
+    DEC_L      ( EBP )                  /*  decrement normal counter       */
+    JA         ( LLBL (G3TN_transform) )
+
+
+    POP_L      ( EDX )                  /*  end of transform ---           */
+    POP_L      ( EAX )                  /*    now normalizing ...          */
+    POP_L      ( EBP )
+
+    MOVQ       ( REGIND(EAX), MM0 )     /*  x1              | x0           */
+    MOVD       ( REGOFF(8, EAX), MM1 )  /*                  | x2           */
+
+    CMP_L      ( CONST(0), EDI )        /*  lengths == 0 ?                 */
+    JE         ( LLBL (G3TN_norm ) )    /*  calculate lengths              */
+
+
+ALIGNTEXT32
+LLBL (G3TN_norm_w_lengths):
+
+    PREFETCHW  ( REGOFF(12,EAX) )
+
+    MOVD       ( REGIND (EDI), MM3 )    /*                  | length (x)   */
+    PFMUL      ( MM3, MM1 )             /*                  | x2 (normalize*/
+
+    PUNPCKLDQ  ( MM3, MM3 )             /*  length (x)      | length (x)   */
+    PFMUL      ( MM3, MM0 )             /*  x1 (normalized) | x0 (normalize*/
+
+    ADD_L      ( STRIDE, EDX )          /*  next normal                    */
+    ADD_L      ( CONST(4), EDI )        /*  next length                    */
+
+    PREFETCH   ( REGIND(EDI) )
+
+    MOVQ       ( MM0, REGIND(EAX) )     /*  write new x0, x1               */
+    MOVD       ( MM1, REGOFF(8, EAX) )  /*  write new x2                   */
+
+    ADD_L      ( CONST(12), EAX )       /*  next r                         */
+    DEC_L      ( EBP )                  /*  decrement normal counter       */
+
+    MOVQ       ( REGIND(EAX), MM0 )     /*  x1              | x0           */
+    MOVD       ( REGOFF(8, EAX), MM1 )  /*                  | x2           */
+    JA         ( LLBL (G3TN_norm_w_lengths) )
+    JMP        ( LLBL (G3TN_exit_3dnow) )
+
+ALIGNTEXT32
+LLBL (G3TN_norm):
+
+    PREFETCHW  ( REGIND(EAX) )
+
+    MOVQ       ( MM0, MM3 )             /*  x1              | x0           */
+    MOVQ       ( MM1, MM4 )             /*                  | x2           */
+
+    PFMUL      ( MM0, MM3 )             /*  x1*x1           | x0*x0        */
+    ADD_L      ( CONST(12), EAX )       /*  next r                         */
+
+    PFMUL      ( MM1, MM4 )             /*                  | x2*x2        */
+    PFADD      ( MM4, MM3 )             /*                  | x0*x0+x2*x2  */
+
+    PFACC      ( MM3, MM3 )             /* **not used**    | x0*x0+x1*x1+x2**/
+    PFRSQRT    ( MM3, MM5 )             /*  1/sqrt (x0*x0+x1*x1+x2*x2)     */
+
+    MOVQ       ( MM5, MM4 )
+    PUNPCKLDQ  ( MM3, MM3 )
+
+    DEC_L      ( EBP )                  /*  decrement normal counter       */
+    PFMUL      ( MM5, MM5 )
+
+    PFRSQIT1   ( MM3, MM5 )
+    PFRCPIT2   ( MM4, MM5 )
+
+    PFMUL      ( MM5, MM0 )             /*  x1 (normalized) | x0 (normalize*/
+
+    MOVQ       ( MM0, REGOFF(-12, EAX) ) /*  write new x0, x1              */
+    PFMUL      ( MM5, MM1 )             /*                 | x2 (normalize*/
+
+    MOVD       ( MM1, REGOFF(-4, EAX) ) /*  write new x2                  */
+    MOVQ       ( REGIND (EAX), MM0 )    /*  x1             | x0           */
+
+    MOVD       ( REGOFF(8, EAX), MM1 )  /*                 | x2           */
+    JA         ( LLBL (G3TN_norm) )
+
+LLBL (G3TN_exit_3dnow):
+    FEMMS
+
+LLBL (G3TN_end):
+    POP_L      ( EBP )
+    POP_L      ( ESI )
+    POP_L      ( EDI )
+    RET
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME(_mesa_3dnow_transform_normalize_normals_no_rot)
+GLNAME(_mesa_3dnow_transform_normalize_normals_no_rot):
+
+ #undef FRAME_OFFSET
+ #define FRAME_OFFSET 12
+
+    PUSH_L     ( EDI )
+    PUSH_L     ( ESI )
+    PUSH_L     ( EBP )
+
+    MOV_L      ( ARG_LENGTHS, EDI )
+    MOV_L      ( ARG_IN, ESI )
+    MOV_L      ( ARG_DEST, EAX )
+    MOV_L      ( REGOFF(V3F_COUNT, ESI), EBP ) /*  dest->count = in->count   */
+    MOV_L      ( EBP, REGOFF(V3F_COUNT, EAX) )
+    MOV_L      ( ARG_MAT, ECX )
+    MOV_L      ( REGOFF(V3F_START, EAX), EAX ) /*  dest->start  */
+    MOV_L      ( REGOFF(MATRIX_INV, ECX), ECX ) /*  mat->inv     */
+    MOV_L      ( REGOFF(V3F_START, ESI), EDX ) /*  in->start    */
+
+    CMP_L      ( CONST(0), EBP ) /*   count > 0 ??  */
+    JE         ( LLBL (G3TNNR_end) )
+
+    FEMMS
+
+    MOVD       ( M(0), MM0 )            /*               | m0                 */
+    PUNPCKLDQ  ( M(5), MM0 )            /* m5            | m0                 */
+
+    MOVD       ( M(10), MM2 )           /*               | m10                */
+    PUNPCKLDQ  ( MM2, MM2 )             /* m10           | m10                */
+
+    CMP_L      ( CONST(0), EDI )        /*  lengths == 0 ?                    */
+    JNE        ( LLBL (G3TNNR_scale_end ) )
+
+    MOVD       ( ARG_SCALE, MM7 )       /*               | scale              */
+    PUNPCKLDQ  ( MM7, MM7 )             /* scale         | scale              */
+
+    PFMUL      ( MM7, MM0 )             /* scale * m5    | scale * m0         */
+    PFMUL      ( MM7, MM2 )             /* scale * m10   | scale * m10        */
+
+ALIGNTEXT32
+LLBL (G3TNNR_scale_end):
+    MOVQ       ( REGIND(EDX), MM6 )     /* x1            | x0                 */
+    MOVD       ( REGOFF(8, EDX), MM7 )  /*               | x2                 */
+
+    CMP_L      ( CONST(0), EDI )        /* lengths == 0 ?                     */
+    JE         ( LLBL (G3TNNR_norm) )   /* need to calculate lengths          */
+
+    MOVD       ( REGIND(EDI), MM3 )     /*                 | length (x)       */
+
+
+ALIGNTEXT32
+LLBL (G3TNNR_norm_w_lengths):           /* use precalculated lengths          */
+
+    PREFETCHW  ( REGIND(EAX) )
+
+    PFMUL      ( MM0, MM6 )             /* x1*m5         | x0*m0              */
+    ADD_L      ( STRIDE, EDX )          /* next normal                        */
+
+    PREFETCH   ( REGIND(EDX) )
+
+    PFMUL      ( MM2, MM7 )             /*               | x2*m10             */
+    ADD_L      ( CONST(12), EAX )       /* next r                             */
+
+    PFMUL      ( MM3, MM7 )             /*               | x2 (normalized)  */
+    PUNPCKLDQ  ( MM3, MM3 )             /* length (x)    | length (x)       */
+
+    ADD_L      ( CONST(4), EDI )        /* next length                        */
+    PFMUL      ( MM3, MM6 )             /* x1 (normalized) | x0 (normalized)  */
+
+    DEC_L      ( EBP )                  /* decrement normal counter           */
+    MOVQ       ( MM6, REGOFF(-12, EAX) ) /* write r0, r1                      */
+
+    MOVD       ( MM7, REGOFF(-4, EAX) ) /* write r2                           */
+    MOVD       ( REGIND(EDI), MM3 )     /*                 | length (x)       */
+
+    MOVQ       ( REGIND(EDX), MM6 )     /* x1              | x0               */
+    MOVD       ( REGOFF(8, EDX), MM7 )  /*                 | x2               */
+
+    JA         ( LLBL (G3TNNR_norm_w_lengths) )
+    JMP        ( LLBL (G3TNNR_exit_3dnow) )
+
+ALIGNTEXT32
+LLBL (G3TNNR_norm):                     /* need to calculate lengths          */
+
+    PREFETCHW  ( REGIND(EAX) )
+
+    PFMUL      ( MM0, MM6 )             /* x1*m5           | x0*m0            */
+    ADD_L      ( CONST(12), EAX )       /* next r                             */
+
+    PFMUL      ( MM2, MM7 )             /*                 | x2*m10           */
+    MOVQ       ( MM6, MM3 )             /* x1 (transformed)| x0 (transformed) */
+
+    MOVQ       ( MM7, MM4 )             /*                 | x2 (transformed) */
+    PFMUL      ( MM6, MM3 )             /* x1*x1           | x0*x0            */
+
+
+    PFMUL      ( MM7, MM4 )             /*                 | x2*x2            */
+    PFACC      ( MM3, MM3 )             /* **not used**    | x0*x0+x1*x1      */
+
+    PFADD      ( MM4, MM3 )             /*                 | x0*x0+x1*x1+x2*x2*/
+    ADD_L      ( STRIDE, EDX )          /* next normal            */
+
+    PREFETCH   ( REGIND(EDX) )
+
+    PFRSQRT    ( MM3, MM5 )             /* 1/sqrt (x0*x0+x1*x1+x2*x2)         */
+    MOVQ       ( MM5, MM4 )
+
+    PUNPCKLDQ  ( MM3, MM3 )
+    PFMUL      ( MM5, MM5 )
+
+    PFRSQIT1   ( MM3, MM5 )
+    DEC_L      ( EBP )                  /* decrement normal counter           */
+
+    PFRCPIT2   ( MM4, MM5 )
+    PFMUL      ( MM5, MM6 )             /* x1 (normalized) | x0 (normalized)  */
+
+    MOVQ       ( MM6, REGOFF(-12, EAX) ) /* write r0, r1                      */
+    PFMUL      ( MM5, MM7 )             /*                 | x2 (normalized)  */
+
+    MOVD       ( MM7, REGOFF(-4, EAX) ) /* write r2                           */
+    MOVQ       ( REGIND(EDX), MM6 )     /* x1              | x0               */
+
+    MOVD       ( REGOFF(8, EDX), MM7 )  /*                 | x2               */
+    JA         ( LLBL (G3TNNR_norm) )
+
+
+LLBL (G3TNNR_exit_3dnow):
+    FEMMS
+
+LLBL (G3TNNR_end):
+    POP_L      ( EBP )
+    POP_L      ( ESI )
+    POP_L      ( EDI )
+    RET
+
+
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME(_mesa_3dnow_transform_rescale_normals_no_rot)
+GLNAME(_mesa_3dnow_transform_rescale_normals_no_rot):
+
+ #undef FRAME_OFFSET
+ #define FRAME_OFFSET 12
+
+    PUSH_L     ( EDI )
+    PUSH_L     ( ESI )
+    PUSH_L     ( EBP )
+
+    MOV_L      ( ARG_IN, EAX )
+    MOV_L      ( ARG_DEST, EDX )
+    MOV_L      ( REGOFF(V3F_COUNT, EAX), EBP ) /*  dest->count = in->count   */
+    MOV_L      ( EBP, REGOFF(V3F_COUNT, EDX) )
+    MOV_L      ( ARG_IN, ESI )
+    MOV_L      ( ARG_MAT, ECX )
+    MOV_L      ( REGOFF(MATRIX_INV, ECX), ECX ) /*  mat->inv     */
+    MOV_L      ( REGOFF(V3F_START, EDX), EAX ) /*  dest->start  */
+    MOV_L      ( REGOFF(V3F_START, ESI), EDX ) /*  in->start    */
+
+    CMP_L      ( CONST(0), EBP )
+    JE         ( LLBL (G3TRNR_end) )
+
+    FEMMS
+
+    MOVD       ( ARG_SCALE, MM6 )       /*               | scale              */
+    PUNPCKLDQ  ( MM6, MM6 )             /* scale         | scale              */
+
+    MOVD       ( REGIND(ECX), MM0 )     /*               | m0                 */
+    PUNPCKLDQ  ( REGOFF(20, ECX), MM0 ) /* m5            | m0                 */
+
+    PFMUL      ( MM6, MM0 )             /* scale*m5      | scale*m0           */
+    MOVD       ( REGOFF(40, ECX), MM2 ) /*               | m10                */
+
+    MOVQ       ( REGIND(EDX), MM4 )     /* x1            | x0                 */
+    PFMUL      ( MM6, MM2 )             /*               | scale*m10          */
+
+    MOVD       ( REGOFF(8, EDX), MM5 )  /*               | x2                 */
+
+ALIGNTEXT32
+LLBL (G3TRNR_rescale):
+
+    PREFETCHW  ( REGIND(EAX) )
+
+    PFMUL      ( MM0, MM4 )             /* x1*m5         | x0*m0              */
+    ADD_L      ( STRIDE, EDX )          /* next normal                        */
+
+    PREFETCH   ( REGIND(EDX) )
+
+    PFMUL      ( MM2, MM5 )             /*               | x2*m10             */
+    ADD_L      ( CONST(12), EAX )       /* next r                             */
+
+    DEC_L      ( EBP )                  /* decrement normal counter           */
+    MOVQ       ( MM4, REGOFF(-12, EAX) ) /* write r0, r1                      */
+
+    MOVD       ( MM5, REGOFF(-4, EAX) ) /* write r2                           */
+    MOVQ       ( REGIND(EDX), MM4 )     /* x1            | x0                 */
+
+    MOVD       ( REGOFF(8, EDX), MM5 )  /*               | x2                 */
+    JA         ( LLBL (G3TRNR_rescale) ) /* cnt > 0 ? -> process next normal  */
+
+    FEMMS
+
+LLBL (G3TRNR_end):
+    POP_L      ( EBP )
+    POP_L      ( ESI )
+    POP_L      ( EDI )
+    RET
+
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME(_mesa_3dnow_transform_rescale_normals)
+GLNAME(_mesa_3dnow_transform_rescale_normals):
+
+ #undef  FRAME_OFFSET
+ #define FRAME_OFFSET 8
+
+    PUSH_L     ( EDI )
+    PUSH_L     ( ESI )
+
+    MOV_L      ( ARG_IN, ESI )
+    MOV_L      ( ARG_DEST, EAX )
+    MOV_L      ( ARG_MAT, ECX )
+    MOV_L      ( REGOFF(V3F_COUNT, ESI), EDI ) /*  dest->count = in->count   */
+    MOV_L      ( EDI, REGOFF(V3F_COUNT, EAX) )
+    MOV_L      ( REGOFF(V3F_START, EAX), EAX ) /*  dest->start  */
+    MOV_L      ( REGOFF(V3F_START, ESI), EDX ) /*  in->start    */
+    MOV_L      ( REGOFF(MATRIX_INV, ECX), ECX ) /*  mat->inv     */
+
+    CMP_L      ( CONST(0), EDI )
+    JE         ( LLBL (G3TR_end) )
+
+    FEMMS
+
+    MOVQ       ( REGIND(ECX), MM3 )     /* m1            | m0                 */
+
+    MOVQ       ( REGOFF(16,ECX), MM4 )  /* m5            | m4                 */
+    MOVD       ( ARG_SCALE, MM0 )       /* scale       */
+
+    MOVD       ( REGOFF(8,ECX), MM5 )   /*               | m2                 */
+    PUNPCKLDQ  ( MM0, MM0 )             /* scale         | scale              */
+
+    PUNPCKLDQ  ( REGOFF(24, ECX), MM5 )
+    PFMUL      ( MM0, MM3 )             /* scale*m1      | scale*m0           */
+
+    MOVQ       ( REGOFF(32, ECX), MM6 ) /* m9            | m8*/
+    PFMUL      ( MM0, MM4 )             /* scale*m5      | scale*m4           */
+
+    MOVD       ( REGOFF(40, ECX), MM7 ) /*               | m10                */
+    PFMUL      ( MM0, MM5 )             /* scale*m6      | scale*m2           */
+
+    PFMUL      ( MM0, MM6 )             /* scale*m9      | scale*m8           */
+    MOVD       ( REGOFF(8, EDX), MM2 )  /*               | x2                 */
+
+    PFMUL      ( MM0, MM7 )             /*               | scale*m10          */
+    MOVQ       ( REGIND(EDX), MM0 )     /* x1            | x0                 */
+
+ALIGNTEXT32
+LLBL (G3TR_rescale):
+
+    PREFETCHW  ( REGIND(EAX) )
+
+    MOVQ       ( MM0, MM1 )             /* x1            | x0                 */
+    PUNPCKLDQ  ( MM2, MM2 )             /* x2            | x2                 */
+
+    PFMUL      ( MM3, MM0 )             /* x1*m1         | x0*m0              */
+    ADD_L      ( CONST(12), EAX )       /* next r                             */
+
+    PFMUL      ( MM4, MM1 )             /* x1*m5         | x0*m4              */
+    PFACC      ( MM1, MM0 )             /* x0*m4+x1*m5   | x0*m0+x1*m1        */
+
+    MOVQ       ( REGIND(EDX), MM1 )     /* x1            | x0                 */
+
+    PFMUL      ( MM5, MM2 )             /* x2*m6         | x2*m2              */
+    PFADD      ( MM2, MM0 )             /* x0*m4...+x2*m6| x0*m0+x1*m1+x2*m2  */
+
+    MOVD       ( REGOFF(8, EDX), MM2 )  /*               | x2                 */
+    ADD_L      ( STRIDE, EDX )          /* next normal                    */
+
+    PREFETCH   ( REGIND(EDX) )
+
+    MOVQ       ( MM0, REGOFF(-12, EAX) ) /* write r0, r1                      */
+    PFMUL      ( MM6, MM1 )             /* x1*m9         | x0*m8              */
+
+    PFMUL      ( MM7, MM2 )             /*               | x2*m10             */
+    PFACC      ( MM1, MM1 )             /* *not used*    | x0*m8+x1*m9        */
+
+    PFADD      ( MM2, MM1 )             /* *not used*    | x0*m8+x1*m9+x2*m10 */
+    MOVD       ( MM1, REGOFF(-4, EAX) ) /* write r2                           */
+
+    MOVQ       ( REGIND(EDX), MM0 )     /* x1            | x0                 */
+    MOVD       ( REGOFF(8, EDX), MM2 )  /*               | x2                 */
+
+    DEC_L      ( EDI )                  /* decrement normal counter           */
+    JA         ( LLBL (G3TR_rescale) )
+
+    FEMMS
+
+LLBL (G3TR_end):
+    POP_L       ( ESI )
+    POP_L       ( EDI )
+    RET
+
+
+
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME(_mesa_3dnow_transform_normals_no_rot)
+GLNAME(_mesa_3dnow_transform_normals_no_rot):
+
+ #undef  FRAME_OFFSET
+ #define FRAME_OFFSET 8
+
+    PUSH_L     ( EDI )
+    PUSH_L     ( ESI )
+
+    MOV_L      ( ARG_IN, ESI )
+    MOV_L      ( ARG_DEST, EAX )
+    MOV_L      ( ARG_MAT, ECX )
+    MOV_L      ( REGOFF(V3F_COUNT, ESI), EDI ) /*  dest->count = in->count   */
+    MOV_L      ( EDI, REGOFF(V3F_COUNT, EAX) )
+    MOV_L      ( REGOFF(V3F_START, EAX), EAX ) /*  dest->start  */
+    MOV_L      ( REGOFF(V3F_START, ESI), EDX ) /*  in->start    */
+    MOV_L      ( REGOFF(MATRIX_INV, ECX), ECX ) /*  mat->inv     */
+
+    CMP_L      ( CONST(0), EDI )
+    JE         ( LLBL (G3TNR_end) )
+
+    FEMMS
+
+    MOVD       ( REGIND(ECX), MM0 )     /*               | m0                 */
+    PUNPCKLDQ  ( REGOFF(20, ECX), MM0 ) /* m5            | m0                 */
+
+    MOVD       ( REGOFF(40, ECX), MM2 ) /*               | m10                */
+    PUNPCKLDQ  ( MM2, MM2 )             /* m10           | m10                */
+
+    MOVQ       ( REGIND(EDX), MM4 )     /* x1            | x0                 */
+    MOVD       ( REGOFF(8, EDX), MM5 )  /*               | x2                 */
+
+ALIGNTEXT32
+LLBL (G3TNR_transform):
+
+    PREFETCHW  ( REGIND(EAX) )
+
+    PFMUL      ( MM0, MM4 )             /* x1*m5         | x0*m0              */
+    ADD_L      ( STRIDE, EDX)           /* next normal      */
+
+    PREFETCH   ( REGIND(EDX) )
+
+    PFMUL      ( MM2, MM5 )             /*               | x2*m10             */
+    ADD_L      ( CONST(12), EAX )       /* next r                             */
+
+    DEC_L      ( EDI )                  /* decrement normal counter           */
+    MOVQ       ( MM4, REGOFF(-12, EAX) ) /* write r0, r1                      */
+
+    MOVD       ( MM5, REGOFF(-4, EAX) ) /* write r2                           */
+    MOVQ       ( REGIND(EDX), MM4 )     /* x1            | x0                 */
+
+    MOVD       ( REGOFF(8, EDX), MM5 )  /*               | x2                 */
+    JA         ( LLBL (G3TNR_transform) )
+
+    FEMMS
+
+LLBL (G3TNR_end):
+    POP_L       ( ESI )
+    POP_L       ( EDI )
+    RET
+
+
+
+
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME(_mesa_3dnow_transform_normals)
+GLNAME(_mesa_3dnow_transform_normals):
+
+ #undef  FRAME_OFFSET
+ #define FRAME_OFFSET 8
+
+    PUSH_L     ( EDI )
+    PUSH_L     ( ESI )
+
+    MOV_L      ( ARG_IN, ESI )
+    MOV_L      ( ARG_DEST, EAX )
+    MOV_L      ( ARG_MAT, ECX )
+    MOV_L      ( REGOFF(V3F_COUNT, ESI), EDI ) /*  dest->count = in->count   */
+    MOV_L      ( EDI, REGOFF(V3F_COUNT, EAX) )
+    MOV_L      ( REGOFF(V3F_START, EAX), EAX ) /*  dest->start  */
+    MOV_L      ( REGOFF(V3F_START, ESI), EDX ) /*  in->start    */
+    MOV_L      ( REGOFF(MATRIX_INV, ECX), ECX ) /*  mat->inv     */
+
+    CMP_L      ( CONST(0), EDI )        /* count > 0 ??                       */
+    JE         ( LLBL (G3T_end) )
+
+    FEMMS
+
+    MOVQ       ( REGIND(ECX), MM3 )     /* m1            | m0                 */
+    MOVQ       ( REGOFF(16, ECX), MM4 ) /* m5            | m4                 */
+
+    MOVD       ( REGOFF(8, ECX), MM5 )  /*               | m2                 */
+    PUNPCKLDQ  ( REGOFF(24, ECX), MM5 ) /* m6            | m2                 */
+
+    MOVQ       ( REGOFF(32, ECX), MM6 ) /* m9            | m8                 */
+    MOVD       ( REGOFF(40, ECX), MM7 ) /*               | m10                */
+
+    MOVQ       ( REGIND(EDX), MM0 )     /* x1            | x0                 */
+    MOVD       ( REGOFF(8, EDX), MM2 )  /*               | x2                 */
+
+ALIGNTEXT32
+LLBL (G3T_transform):
+
+    PREFETCHW  ( REGIND(EAX) )
+
+    MOVQ       ( MM0, MM1 )             /* x1            | x0                 */
+    PUNPCKLDQ  ( MM2, MM2 )             /* x2            | x2                 */
+
+    PFMUL      ( MM3, MM0 )             /* x1*m1         | x0*m0              */
+    ADD_L      ( CONST(12), EAX )       /* next r                             */
+
+    PFMUL      ( MM4, MM1 )             /* x1*m5         | x0*m4              */
+    PFACC      ( MM1, MM0 )             /* x0*m4+x1*m5   | x0*m0+x1*m1        */
+
+    PFMUL      ( MM5, MM2 )             /* x2*m6         | x2*m2              */
+    PFADD      ( MM2, MM0 )             /* x0*m4...+x2*m6| x0*m0+x1*m1+x2*m2  */
+
+    MOVQ       ( REGIND(EDX), MM1 )     /* x1            | x0                 */
+    MOVQ       ( MM0, REGOFF(-12, EAX) ) /* write r0, r1                      */
+
+    PFMUL      ( MM6, MM1 )             /* x1*m9         | x0*m8              */
+    MOVD       ( REGOFF(8, EDX), MM2 )  /*               | x2                 */
+
+    PFMUL      ( MM7, MM2 )             /*               | x2*m10             */
+    ADD_L      ( STRIDE, EDX )          /* next normal               */
+
+    PREFETCH   ( REGIND(EDX) )
+
+    PFACC      ( MM1, MM1 )             /* *not used*    | x0*m8+x1*m9        */
+    PFADD      ( MM2, MM1 )             /* *not used*    | x0*m8+x1*m9+x2*m10 */
+
+    MOVD       ( MM1, REGOFF(-4, EAX) ) /* write r2                           */
+    MOVQ       ( REGIND(EDX), MM0 )     /* x1            | x0                 */
+
+    MOVD       ( REGOFF(8, EDX), MM2 )  /*               | x2                 */
+    DEC_L      ( EDI )                  /* decrement normal counter           */
+    JA         ( LLBL (G3T_transform) )
+
+    FEMMS
+
+LLBL (G3T_end):
+    POP_L  ( ESI )
+    POP_L  ( EDI )
+    RET
+
+
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME(_mesa_3dnow_normalize_normals)
+GLNAME(_mesa_3dnow_normalize_normals):
+
+ #undef  FRAME_OFFSET
+ #define FRAME_OFFSET 12
+
+    PUSH_L     ( EDI )
+    PUSH_L     ( ESI )
+    PUSH_L     ( EBP )
+
+    MOV_L      ( ARG_IN, ESI )
+    MOV_L      ( ARG_DEST, EAX )
+    MOV_L      ( REGOFF(V3F_COUNT, ESI), EBP ) /*  dest->count = in->count   */
+    MOV_L      ( EBP, REGOFF(V3F_COUNT, EAX) )
+    MOV_L      ( REGOFF(V3F_START, EAX), EAX ) /*  dest->start  */
+    MOV_L      ( REGOFF(V3F_START, ESI), ECX ) /*  in->start    */
+    MOV_L      ( ARG_LENGTHS, EDX )
+
+    CMP_L      ( CONST(0), EBP ) /* count > 0 ?? */
+    JE         ( LLBL (G3N_end) )
+
+    FEMMS
+
+    MOVQ       ( REGIND(ECX), MM0 )     /* x1              | x0               */
+    MOVD       ( REGOFF(8, ECX), MM1 )  /*                 | x2               */
+
+    CMP_L      ( CONST(0), EDX )        /* lengths == 0 ?                     */
+    JE         ( LLBL (G3N_norm2) )     /* calculate lengths                  */
+
+ALIGNTEXT32
+LLBL (G3N_norm1):                       /* use precalculated lengths          */
+
+    PREFETCH   ( REGIND(EAX) )
+
+    MOVD       ( REGIND(EDX), MM3 )     /*                 | length (x)       */
+    PFMUL      ( MM3, MM1 )             /*                 | x2 (normalized)  */
+
+    PUNPCKLDQ  ( MM3, MM3 )             /* length (x)      | length (x)       */
+    ADD_L      ( STRIDE, ECX )          /* next normal            */
+
+    PREFETCH   ( REGIND(ECX) )
+
+    PFMUL      ( MM3, MM0 )             /* x1 (normalized) | x0 (normalized)  */
+    MOVQ       ( MM0, REGIND(EAX) )     /* write new x0, x1                   */
+
+    MOVD       ( MM1, REGOFF(8, EAX) )  /* write new x2                       */
+    ADD_L      ( CONST(12), EAX )       /* next r                             */
+
+    ADD_L      ( CONST(4), EDX )        /* next length                        */
+    DEC_L      ( EBP )                  /* decrement normal counter           */
+
+    MOVQ       ( REGIND(ECX), MM0 )     /* x1              | x0               */
+    MOVD       ( REGOFF(8, ECX), MM1 )  /*                 | x2               */
+    JA         ( LLBL (G3N_norm1) )
+
+    JMP        ( LLBL (G3N_end1) )
+
+ALIGNTEXT32
+LLBL (G3N_norm2):                       /* need to calculate lengths          */
+
+    PREFETCHW  ( REGIND(EAX) )
+
+    MOVQ       ( MM0, MM3 )             /* x1              | x0               */
+    ADD_L      ( STRIDE, ECX )          /* next normal    */
+
+    PREFETCH   ( REGIND(ECX) )
+
+    PFMUL      ( MM0, MM3 )             /* x1*x1           | x0*x0            */
+    MOVQ       ( MM1, MM4 )             /*                 | x2               */
+
+    ADD_L      ( CONST(12), EAX )       /* next r                             */
+    PFMUL      ( MM1, MM4 )             /*                 | x2*x2            */
+
+    PFADD      ( MM4, MM3 )             /*                 | x0*x0+x2*x2      */
+    PFACC      ( MM3, MM3 )             /* x0*x0+...+x2*x2 | x0*x0+x1*x1+x2*x2*/
+
+    PFRSQRT    ( MM3, MM5 )             /* 1/sqrt (x0*x0+x1*x1+x2*x2)         */
+    MOVQ       ( MM5, MM4 )
+
+    PUNPCKLDQ  ( MM3, MM3 )
+    PFMUL      ( MM5, MM5 )
+
+    PFRSQIT1   ( MM3, MM5 )
+    DEC_L      ( EBP )                  /* decrement normal counter           */
+
+    PFRCPIT2   ( MM4, MM5 )
+
+    PFMUL      ( MM5, MM0 )             /* x1 (normalized) | x0 (normalized)  */
+    MOVQ       ( MM0, REGOFF(-12, EAX) ) /* write new x0, x1                  */
+
+    PFMUL      ( MM5, MM1 )             /*                 | x2 (normalized)  */
+    MOVD       ( MM1, REGOFF(-4, EAX) ) /* write new x2                       */
+
+    MOVQ       ( REGIND(ECX), MM0 )     /* x1              | x0               */
+    MOVD       ( REGOFF(8, ECX), MM1 )  /*                 | x2               */
+    JA         ( LLBL (G3N_norm2) )
+
+LLBL (G3N_end1):
+    FEMMS
+
+LLBL (G3N_end):
+    POP_L      ( EBP )
+    POP_L      ( ESI )
+    POP_L      ( EDI )
+    RET
+
+
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME(_mesa_3dnow_rescale_normals)
+GLNAME(_mesa_3dnow_rescale_normals):
+
+ #undef  FRAME_OFFSET
+ #define FRAME_OFFSET 8
+    PUSH_L     ( EDI )
+    PUSH_L     ( ESI )
+
+    MOV_L      ( ARG_IN, ESI )
+    MOV_L      ( ARG_DEST, EAX )
+    MOV_L      ( REGOFF(V3F_COUNT, ESI), EDX ) /*  dest->count = in->count   */
+    MOV_L      ( EDX, REGOFF(V3F_COUNT, EAX) )
+    MOV_L      ( REGOFF(V3F_START, EAX), EAX ) /*  dest->start  */
+    MOV_L      ( REGOFF(V3F_START, ESI), ECX ) /*  in->start    */
+
+    CMP_L      ( CONST(0), EDX )
+    JE         ( LLBL (G3R_end) )
+
+    FEMMS
+
+    MOVD       ( ARG_SCALE, MM0 )       /* scale                              */
+    PUNPCKLDQ  ( MM0, MM0 )
+
+    MOVQ       ( REGIND(ECX), MM1 )     /* x1            | x0                 */
+    MOVD       ( REGOFF(8, ECX), MM2 )  /*               | x2                 */
+
+ALIGNTEXT32
+LLBL (G3R_rescale):
+
+    PREFETCHW  ( REGIND(EAX) )
+
+    PFMUL      ( MM0, MM1 )             /* x1*scale      | x0*scale           */
+    ADD_L      ( STRIDE, ECX )          /* next normal                  */
+
+    PREFETCH   ( REGIND(ECX) )
+
+    PFMUL      ( MM0, MM2 )             /*               | x2*scale           */
+    ADD_L      ( CONST(12), EAX )       /* next r                             */
+
+    MOVQ       ( MM1, REGOFF(-12, EAX) ) /* write r0, r1                      */
+    MOVD       ( MM2, REGOFF(-4, EAX) ) /* write r2                           */
+
+    DEC_L      ( EDX )                  /* decrement normal counter           */
+    MOVQ       ( REGIND(ECX), MM1 )     /* x1            | x0                 */
+
+    MOVD       ( REGOFF(8, ECX), MM2 )  /*               | x2                 */
+    JA         ( LLBL (G3R_rescale) )
+
+    FEMMS
+
+LLBL (G3R_end):
+    POP_L      ( ESI )
+    POP_L      ( EDI )
+    RET
diff --git a/src/mesa/x86/3dnow_xform1.S b/src/mesa/x86/3dnow_xform1.S
new file mode 100644
index 00000000000..9957676ee96
--- /dev/null
+++ b/src/mesa/x86/3dnow_xform1.S
@@ -0,0 +1,423 @@
+/* $Id: 3dnow_xform1.S,v 1.1 2001/03/29 06:46:16 gareth Exp $ */
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  3.5
+ *
+ * Copyright (C) 1999-2001  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.
+ */
+
+#include "matypes.h"
+#include "xform_args.h"
+
+    SEG_TEXT
+
+#define FRAME_OFFSET	4
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_3dnow_transform_points1_general )
+GLNAME( _mesa_3dnow_transform_points1_general ):
+
+    PUSH_L    ( ESI )
+
+    MOV_L     ( ARG_DEST, ECX )
+    MOV_L     ( ARG_MATRIX, ESI )
+    MOV_L     ( ARG_SOURCE, EAX )
+    MOV_L     ( CONST(4), REGOFF(V4F_SIZE, ECX) )
+    OR_B      ( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, ECX) )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), EDX )
+    MOV_L     ( EDX, REGOFF(V4F_COUNT, ECX) )
+
+    PUSH_L    ( EDI )
+
+    MOV_L     ( REGOFF(4, ECX), EDX )
+    MOV_L     ( ESI, ECX )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), ESI )
+    MOV_L     ( REGOFF(V4F_STRIDE, EAX), EDI )
+    MOV_L     ( REGOFF(V4F_START, EAX), EAX )
+
+    TEST_L    ( ESI, ESI )
+    JZ        ( LLBL( G3TPGR_3 ) )
+
+    MOVQ      ( REGIND(ECX), MM0 )	/* m01             | m00             */
+    MOVQ      ( REGOFF(8, ECX), MM1 )	/* m03             | m02             */
+
+    MOVQ      ( REGOFF(48, ECX), MM2 )	/* m31             | m30             */
+    MOVQ      ( REGOFF(56, ECX), MM3 )	/* m33             | m32             */
+
+ALIGNTEXT16
+LLBL( G3TPGR_2 ):
+
+    MOVD      ( REGIND(EAX), MM4 )	/*                 | x0              */
+    PUNPCKLDQ ( MM4, MM4 )		/* x0              | x0              */
+
+    MOVQ      ( MM4, MM5 )		/* x0              | x0              */
+    PFMUL     ( MM0, MM4 )		/* x0*m01          | x0*m00          */
+
+    PFMUL     ( MM1, MM5 )		/* x0*m03          | x0*m02          */
+    PFADD     ( MM2, MM4 )		/* x0*m01+m31      | x0*m00+m30      */
+
+    PFADD     ( MM3, MM5 )		/* x0*m03+m33      | x0*m02+m32      */
+    MOVQ      ( MM4, REGIND(EDX) )	/* write r1, r0                      */
+
+    MOVQ      ( MM5, REGOFF(8, EDX) )	/* write r3, r2                      */
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+
+    ADD_L     ( CONST(16), EDX )	/* next r                            */
+    DEC_L     ( ESI )			/* decrement vertex counter          */
+
+    JNZ       ( LLBL( G3TPGR_2 ) )	/* cnt > 0 ? -> process next vertex  */
+
+LLBL( G3TPGR_3 ):
+
+    FEMMS
+    POP_L     ( EDI )
+    POP_L     ( ESI )
+    RET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_3dnow_transform_points1_identity )
+GLNAME( _mesa_3dnow_transform_points1_identity ):
+
+    PUSH_L    ( ESI )
+
+    MOV_L     ( ARG_DEST, ECX )
+    MOV_L     ( ARG_MATRIX, ESI )
+    MOV_L     ( ARG_SOURCE, EAX )
+    MOV_L     ( CONST(1), REGOFF(V4F_SIZE, ECX) )
+    OR_B      ( CONST(VEC_SIZE_1), REGOFF(V4F_FLAGS, ECX) )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), EDX )
+    MOV_L     ( EDX, REGOFF(V4F_COUNT, ECX) )
+
+    PUSH_L    ( EDI )
+
+    MOV_L     ( REGOFF(4, ECX), EDX )
+    MOV_L     ( ESI, ECX )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), ESI )
+    MOV_L     ( REGOFF(V4F_STRIDE, EAX), EDI )
+    MOV_L     ( REGOFF(V4F_START, EAX), EAX )
+
+    TEST_L    ( ESI, ESI )
+    JZ        ( LLBL( G3TPIR_4) )
+
+ALIGNTEXT16
+LLBL( G3TPIR_3 ):
+
+    MOVD      ( REGIND(EAX), MM0 )	/*                 | x0              */
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+
+    MOVD      ( MM0, REGIND(EDX) )	/*                 | r0              */
+    ADD_L     ( CONST(16), EDX )	/* next r                            */
+
+    DEC_L     ( ESI )			/* decrement vertex counter          */
+    JNZ       ( LLBL( G3TPIR_3 ) )	/* cnt > 0 ? -> process next vertex  */
+
+LLBL( G3TPIR_4 ):
+
+    FEMMS
+    POP_L     ( EDI )
+    POP_L     ( ESI )
+    RET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_3dnow_transform_points1_3d_no_rot )
+GLNAME( _mesa_3dnow_transform_points1_3d_no_rot ):
+
+    PUSH_L    ( ESI )
+
+    MOV_L     ( ARG_DEST, ECX )
+    MOV_L     ( ARG_MATRIX, ESI )
+    MOV_L     ( ARG_SOURCE, EAX )
+    MOV_L     ( CONST(3), REGOFF(V4F_SIZE, ECX) )
+    OR_B      ( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, ECX) )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), EDX )
+    MOV_L     ( EDX, REGOFF(V4F_COUNT, ECX) )
+
+    PUSH_L    ( EDI )
+
+    MOV_L     ( REGOFF(4, ECX), EDX )
+    MOV_L     ( ESI, ECX )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), ESI )
+    MOV_L     ( REGOFF(V4F_STRIDE, EAX), EDI )
+    MOV_L     ( REGOFF(V4F_START, EAX), EAX )
+
+    TEST_L    ( ESI, ESI )
+    JZ        ( LLBL( G3TP3NRR_3 ) )
+
+    MOVD      ( REGIND(ECX), MM0 )	/*                 | m00             */
+    MOVQ      ( REGOFF(48, ECX), MM2 )	/* m31             | m30             */
+
+    MOVD      ( REGOFF(56, ECX), MM3 )	/*                 | m32             */
+
+ALIGNTEXT16
+LLBL( G3TP3NRR_2 ):
+
+    MOVD      ( REGIND(EAX), MM4 )	/*                 | x0              */
+    PFMUL     ( MM0, MM4 )		/*                 | x0*m00          */
+
+    PFADD     ( MM2, MM4 )		/* m31             | x0*m00+m30      */
+    MOVQ      ( MM4, REGIND(EDX) )	/* write r1, r0                      */
+
+    MOVD      ( MM3, REGOFF(8, EDX) )	/* write r2                          */
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+
+    ADD_L     ( CONST(16), EDX )	/* next r                            */
+    DEC_L     ( ESI )			/* decrement vertex counter          */
+
+    JNZ       ( LLBL( G3TP3NRR_2 ) )	/* cnt > 0 ? -> process next vertex  */
+
+LLBL( G3TP3NRR_3 ):
+
+    FEMMS
+    POP_L     ( EDI )
+    POP_L     ( ESI )
+    RET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_3dnow_transform_points1_perspective )
+GLNAME( _mesa_3dnow_transform_points1_perspective ):
+
+    PUSH_L    ( ESI )
+
+    MOV_L     ( ARG_DEST, ECX )
+    MOV_L     ( ARG_MATRIX, ESI )
+    MOV_L     ( ARG_SOURCE, EAX )
+    MOV_L     ( CONST(4), REGOFF(V4F_SIZE, ECX) )
+    OR_B      ( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, ECX) )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), EDX )
+    MOV_L     ( EDX, REGOFF(V4F_COUNT, ECX) )
+
+    PUSH_L    ( EDI )
+
+    MOV_L     ( REGOFF(4, ECX), EDX )
+    MOV_L     ( ESI, ECX )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), ESI )
+    MOV_L     ( REGOFF(V4F_STRIDE, EAX), EDI )
+    MOV_L     ( REGOFF(V4F_START, EAX), EAX )
+
+    TEST_L    ( ESI, ESI )
+    JZ        ( LLBL( G3TPPR_3 ) )
+
+    MOVD      ( REGIND(ECX), MM0 )	/*                 | m00             */
+    MOVD      ( REGOFF(56, ECX), MM3 )	/*                 | m32             */
+
+ALIGNTEXT16
+LLBL( G3TPPR_2 ):
+
+    MOVD      ( REGIND(EAX), MM4 )	/* 0               | x0              */
+    PFMUL     ( MM0, MM4 )		/* 0               | x0*m00          */
+
+    MOVQ      ( MM4, REGIND(EDX) )	/* write r1, r0                      */
+    MOVQ      ( MM3, REGOFF(8, EDX) )	/* write r2  (=m32), r3 (=0)         */
+
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+    ADD_L     ( CONST(16), EDX )	/* next r                            */
+
+    DEC_L     ( ESI )			/* decrement vertex counter          */
+    JNZ       ( LLBL( G3TPPR_2 ) )	/* cnt > 0 ? -> process next vertex  */
+
+LLBL( G3TPPR_3 ):
+
+    FEMMS
+    POP_L     ( EDI )
+    POP_L     ( ESI )
+    RET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_3dnow_transform_points1_2d )
+GLNAME( _mesa_3dnow_transform_points1_2d ):
+
+    PUSH_L    ( ESI )
+
+    MOV_L     ( ARG_DEST, ECX )
+    MOV_L     ( ARG_MATRIX, ESI )
+    MOV_L     ( ARG_SOURCE, EAX )
+    MOV_L     ( CONST(2), REGOFF(V4F_SIZE, ECX) )
+    OR_B      ( CONST(VEC_SIZE_2), REGOFF(V4F_FLAGS, ECX) )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), EDX )
+    MOV_L     ( EDX, REGOFF(V4F_COUNT, ECX) )
+
+    PUSH_L    ( EDI )
+
+    MOV_L     ( REGOFF(4, ECX), EDX )
+    MOV_L     ( ESI, ECX )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), ESI )
+    MOV_L     ( REGOFF(V4F_STRIDE, EAX), EDI )
+    MOV_L     ( REGOFF(V4F_START, EAX), EAX )
+
+    TEST_L    ( ESI, ESI )
+    JZ        ( LLBL( G3TP2R_3 ) )
+
+    MOVQ      ( REGIND(ECX), MM0 )	/* m01             | m00             */
+    MOVQ      ( REGOFF(48, ECX), MM2 )	/* m31             | m30             */
+
+ALIGNTEXT16
+LLBL( G3TP2R_2 ):
+
+    MOVD      ( REGIND(EAX), MM4 )	/*                 | x0              */
+    PUNPCKLDQ ( MM4, MM4 )		/* x0              | x0              */
+
+    PFMUL     ( MM0, MM4 )		/* x0*m01          | x0*m00          */
+    PFADD     ( MM2, MM4 )		/* x0*m01+m31      | x0*m00+m30      */
+
+    MOVQ      ( MM4, REGIND(EDX) )	/* write r1, r0                      */
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+
+    ADD_L     ( CONST(16), EDX )	/* next r                            */
+    DEC_L     ( ESI )			/* decrement vertex counter          */
+
+    JNZ       ( LLBL( G3TP2R_2 ) )	/* cnt > 0 ? -> process next vertex  */
+
+LLBL( G3TP2R_3 ):
+
+    FEMMS
+    POP_L     ( EDI )
+    POP_L     ( ESI )
+    RET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_3dnow_transform_points1_2d_no_rot )
+GLNAME( _mesa_3dnow_transform_points1_2d_no_rot ):
+
+    PUSH_L    ( ESI )
+
+    MOV_L     ( ARG_DEST, ECX )
+    MOV_L     ( ARG_MATRIX, ESI )
+    MOV_L     ( ARG_SOURCE, EAX )
+    MOV_L     ( CONST(2), REGOFF(V4F_SIZE, ECX) )
+    OR_B      ( CONST(VEC_SIZE_2), REGOFF(V4F_FLAGS, ECX) )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), EDX )
+    MOV_L     ( EDX, REGOFF(V4F_COUNT, ECX) )
+
+    PUSH_L    ( EDI )
+
+    MOV_L     ( REGOFF(4, ECX), EDX )
+    MOV_L     ( ESI, ECX )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), ESI )
+    MOV_L     ( REGOFF(V4F_STRIDE, EAX), EDI )
+    MOV_L     ( REGOFF(V4F_START, EAX), EAX )
+
+    TEST_L    ( ESI, ESI )
+    JZ        ( LLBL( G3TP2NRR_3 ) )
+
+    MOVD      ( REGIND(ECX), MM0 )	/*                 | m00             */
+    MOVQ      ( REGOFF(48, ECX), MM2 )	/* m31             | m30             */
+
+ALIGNTEXT16
+LLBL( G3TP2NRR_2 ):
+
+    MOVD      ( REGIND(EAX), MM4 )	/*                 | x0              */
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+
+    PFMUL     ( MM0, MM4 )		/*                 | x0*m00          */
+    PFADD     ( MM2, MM4 )		/* m31             | x0*m00+m30      */
+
+    MOVQ      ( MM4, REGIND(EDX) )	/* write r1, r0                      */
+    ADD_L     ( CONST(16), EDX )	/* next r                            */
+
+    DEC_L     ( ESI )			/* decrement vertex counter          */
+    JNZ       ( LLBL( G3TP2NRR_2 ) )	/* cnt > 0 ? -> process next vertex  */
+
+LLBL( G3TP2NRR_3 ):
+
+    FEMMS
+    POP_L     ( EDI )
+    POP_L     ( ESI )
+    RET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_3dnow_transform_points1_3d )
+GLNAME( _mesa_3dnow_transform_points1_3d ):
+
+    PUSH_L    ( ESI )
+
+    MOV_L     ( ARG_DEST, ECX )
+    MOV_L     ( ARG_MATRIX, ESI )
+    MOV_L     ( ARG_SOURCE, EAX )
+    MOV_L     ( CONST(3), REGOFF(V4F_SIZE, ECX) )
+    OR_B      ( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, ECX) )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), EDX )
+    MOV_L     ( EDX, REGOFF(V4F_COUNT, ECX) )
+
+    PUSH_L    ( EDI )
+
+    MOV_L     ( REGOFF(4, ECX), EDX )
+    MOV_L     ( ESI, ECX )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), ESI )
+    MOV_L     ( REGOFF(V4F_STRIDE, EAX), EDI )
+    MOV_L     ( REGOFF(V4F_START, EAX), EAX )
+
+    TEST_L    ( ESI, ESI )
+    JZ        ( LLBL( G3TP3R_3 ) )
+
+    MOVQ      ( REGIND(ECX), MM0 )	/* m01             | m00             */
+    MOVD      ( REGOFF(8, ECX), MM1 )	/*                 | m02             */
+
+    MOVQ      ( REGOFF(48, ECX), MM2 )	/* m31             | m30             */
+    MOVD      ( REGOFF(56, ECX), MM3 )	/*                 | m32             */
+
+ALIGNTEXT16
+LLBL( G3TP3R_2 ):
+
+    MOVD      ( REGIND(EAX), MM4 )	/*                 | x0              */
+    PUNPCKLDQ ( MM4, MM4 )		/* x0              | x0              */
+
+    MOVQ      ( MM4, MM5 )		/*                 | x0              */
+    PFMUL     ( MM0, MM4 )		/* x0*m01          | x0*m00          */
+
+    PFMUL     ( MM1, MM5 )		/*                 | x0*m02          */
+    PFADD     ( MM2, MM4 )		/* x0*m01+m31      | x0*m00+m30      */
+
+    PFADD     ( MM3, MM5 )		/*                 | x0*m02+m32      */
+    MOVQ      ( MM4, REGIND(EDX) )	/* write r1, r0                      */
+
+    MOVD      ( MM5, REGOFF(8, EDX) )	/* write r2                          */
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+
+    ADD_L     ( CONST(16), EDX )	/* next r                            */
+    DEC_L     ( ESI )			/* decrement vertex counter          */
+
+    JNZ       ( LLBL( G3TP3R_2 ) )	/* cnt > 0 ? -> process next vertex  */
+
+LLBL( G3TP3R_3 ):
+
+    FEMMS
+    POP_L     ( EDI )
+    POP_L     ( ESI )
+    RET
diff --git a/src/mesa/x86/3dnow_xform2.S b/src/mesa/x86/3dnow_xform2.S
new file mode 100644
index 00000000000..cb3a8531df7
--- /dev/null
+++ b/src/mesa/x86/3dnow_xform2.S
@@ -0,0 +1,464 @@
+/* $Id: 3dnow_xform2.S,v 1.1 2001/03/29 06:46:16 gareth Exp $ */
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  3.5
+ *
+ * Copyright (C) 1999-2001  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.
+ */
+
+#include "matypes.h"
+#include "xform_args.h"
+
+    SEG_TEXT
+
+#define FRAME_OFFSET	4
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_3dnow_transform_points2_general )
+GLNAME( _mesa_3dnow_transform_points2_general ):
+
+    PUSH_L    ( ESI )
+
+    MOV_L     ( ARG_DEST, ECX )
+    MOV_L     ( ARG_MATRIX, ESI )
+    MOV_L     ( ARG_SOURCE, EAX )
+    MOV_L     ( CONST(4), REGOFF(V4F_SIZE, ECX) )
+    OR_B      ( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, ECX) )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), EDX )
+    MOV_L     ( EDX, REGOFF(V4F_COUNT, ECX) )
+
+    PUSH_L    ( EDI )
+
+    MOV_L     ( REGOFF(V4F_START, ECX), EDX )
+    MOV_L     ( ESI, ECX )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), ESI )
+    MOV_L     ( REGOFF(V4F_STRIDE, EAX), EDI )
+    MOV_L     ( REGOFF(V4F_START, EAX), EAX )
+
+    TEST_L    ( ESI, ESI )
+    JZ        ( LLBL( G3TPGR_3 ) )
+
+    MOVD      ( REGIND(ECX), MM0 )	/*                 | m00             */
+    PUNPCKLDQ ( REGOFF(16, ECX), MM0 )	/* m10             | m00             */
+
+    MOVD      ( REGOFF(4, ECX), MM1 )	/*                 | m01             */
+    PUNPCKLDQ ( REGOFF(20, ECX), MM1 )	/* m11             | m01             */
+
+    MOVD      ( REGOFF(8, ECX), MM2 )	/*                 | m02             */
+    PUNPCKLDQ ( REGOFF(24, ECX), MM2 )	/* m12             | m02             */
+
+    MOVD      ( REGOFF(12, ECX), MM3 )	/*                 | m03             */
+    PUNPCKLDQ ( REGOFF(28, ECX), MM3 )	/* m13             | m03             */
+
+    MOVQ      ( REGOFF(48, ECX), MM4 )	/* m31             | m30             */
+    MOVQ      ( REGOFF(56, ECX), MM5 )	/* m33             | m32             */
+
+ALIGNTEXT16
+LLBL( G3TPGR_2 ):
+
+    MOVQ      ( REGIND(EAX), MM6 )	/* x1              | x0              */
+    MOVQ      ( MM6, MM7 )		/* x1              | x0              */
+
+    PFMUL     ( MM0, MM6 )		/* x1*m10          | x0*m00          */
+    PFMUL     ( MM1, MM7 )		/* x1*m11          | x0*m01          */
+
+    PFACC     ( MM7, MM6 )		/* x0*m01+x1*m11   | x0*x00+x1*m10   */
+    PFADD     ( MM4, MM6 )		/* x0*...*m11+m31  | x0*...*m10+m30  */
+
+    MOVQ      ( MM6, REGIND(EDX) )	/* write r1, r0                      */
+    MOVQ      ( REGIND(EAX), MM6 )	/* x1              | x0              */
+
+    MOVQ      ( MM6, MM7 )		/* x1              | x0              */
+    PFMUL     ( MM2, MM6 )		/* x1*m12          | x0*m02          */
+
+    PFMUL     ( MM3, MM7 )		/* x1*m13          | x0*m03          */
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+
+    PFACC     ( MM7, MM6 )		/* x0*m03+x1*m13   | x0*x02+x1*m12   */
+    PFADD     ( MM5, MM6 )		/* x0*...*m13+m33  | x0*...*m12+m32  */
+
+    MOVQ      ( MM6, REGOFF(8, EDX) )	/* write r3, r2                      */
+    ADD_L     ( CONST(16), EDX )	/* next r                            */
+
+    DEC_L     ( ESI )			/* decrement vertex counter          */
+    JNZ       ( LLBL( G3TPGR_2 ) )	/* cnt > 0 ? -> process next vertex  */
+
+LLBL( G3TPGR_3 ):
+
+    FEMMS
+    POP_L     ( EDI )
+    POP_L     ( ESI )
+    RET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_3dnow_transform_points2_perspective )
+GLNAME( _mesa_3dnow_transform_points2_perspective ):
+
+    PUSH_L    ( ESI )
+
+    MOV_L     ( ARG_DEST, ECX )
+    MOV_L     ( ARG_MATRIX, ESI )
+    MOV_L     ( ARG_SOURCE, EAX )
+    MOV_L     ( CONST(4), REGOFF(V4F_SIZE, ECX) )
+    OR_B      ( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, ECX) )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), EDX )
+    MOV_L     ( EDX, REGOFF(V4F_COUNT, ECX) )
+
+    PUSH_L    ( EDI )
+
+    MOV_L     ( REGOFF(V4F_START, ECX), EDX )
+    MOV_L     ( ESI, ECX )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), ESI )
+    MOV_L     ( REGOFF(V4F_STRIDE, EAX), EDI )
+    MOV_L     ( REGOFF(V4F_START, EAX), EAX )
+
+    TEST_L    ( ESI, ESI )
+    JZ        ( LLBL( G3TPPR_3 ) )
+
+    MOVD      ( REGIND(ECX), MM0 )	/*                 | m00             */
+    PUNPCKLDQ ( REGOFF(20, ECX), MM0 )	/* m11             | m00             */
+
+    MOVD      ( REGOFF(56, ECX), MM3 )	/*                 | m32             */
+
+ALIGNTEXT16
+LLBL( G3TPPR_2 ):
+
+    MOVQ      ( REGIND(EAX), MM4 )	/* x1              | x0              */
+    PFMUL     ( MM0, MM4 )		/* x1*m11          | x0*m00          */
+
+    MOVQ      ( MM4, REGIND(EDX) )	/* write r1, r0                      */
+    MOVQ      ( MM3, REGOFF(8, EDX) )	/* write r2  (=m32), r3 (=0)         */
+
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+    ADD_L     ( CONST(16), EDX )	/* next r                            */
+
+    DEC_L     ( ESI )			/* decrement vertex counter          */
+    JNZ       ( LLBL( G3TPPR_2 ) )	/* cnt > 0 ? -> process next vertex  */
+
+LLBL( G3TPPR_3 ):
+
+    FEMMS
+    POP_L     ( EDI )
+    POP_L     ( ESI )
+    RET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_3dnow_transform_points2_3d )
+GLNAME( _mesa_3dnow_transform_points2_3d ):
+
+    PUSH_L    ( ESI )
+
+    MOV_L     ( ARG_DEST, ECX )
+    MOV_L     ( ARG_MATRIX, ESI )
+    MOV_L     ( ARG_SOURCE, EAX )
+    MOV_L     ( CONST(3), REGOFF(V4F_SIZE, ECX) )
+    OR_B      ( CONST(VEC_SIZE_3 ), REGOFF(V4F_FLAGS, ECX) )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), EDX )
+    MOV_L     ( EDX, REGOFF(V4F_COUNT, ECX) )
+
+    PUSH_L    ( EDI )
+
+    MOV_L     ( REGOFF(V4F_START, ECX), EDX )
+    MOV_L     ( ESI, ECX )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), ESI )
+    MOV_L     ( REGOFF(V4F_STRIDE, EAX), EDI )
+    MOV_L     ( REGOFF(V4F_START, EAX), EAX )
+
+    TEST_L    ( ESI, ESI )
+    JZ        ( LLBL( G3TP3R_3 ) )
+
+    MOVD      ( REGIND(ECX), MM0 )	/*                 | m00             */
+    PUNPCKLDQ ( REGOFF(16, ECX), MM0 )	/* m10             | m00             */
+
+    MOVD      ( REGOFF(4, ECX), MM1 )	/*                 | m01             */
+    PUNPCKLDQ ( REGOFF(20, ECX), MM1 )	/* m11             | m01             */
+
+    MOVD      ( REGOFF(8, ECX), MM2 )	/*                 | m02             */
+    PUNPCKLDQ ( REGOFF(24, ECX), MM2 )	/* m12             | m02             */
+
+    MOVQ      ( REGOFF(48, ECX), MM4 )	/* m31             | m30             */
+    MOVD      ( REGOFF(56, ECX), MM5 )	/*                 | m32             */
+
+ALIGNTEXT16
+LLBL( G3TP3R_2 ):
+
+    MOVQ      ( REGIND(EAX), MM6 )	/* x1              | x0              */
+    MOVQ      ( MM6, MM7 )		/* x1              | x0              */
+
+    PFMUL     ( MM0, MM6 )		/* x1*m10          | x0*m00          */
+    PFMUL     ( MM1, MM7 )		/* x1*m11          | x0*m01          */
+
+    PFACC     ( MM7, MM6 )		/* x0*m01+x1*m11   | x0*x00+x1*m10   */
+    PFADD     ( MM4, MM6 )		/* x0*...*m11+m31  | x0*...*m10+m30  */
+
+    MOVQ      ( MM6, REGIND(EDX) )	/* write r1, r0                      */
+    MOVQ      ( REGIND(EAX), MM6 )	/* x1              | x0              */
+
+    MOVQ      ( MM6, MM7 )		/* x1              | x0              */
+    PFMUL     ( MM2, MM6 )		/* x1*m12          | x0*m02          */
+
+    PFACC     ( MM7, MM6 )		/* ***trash***     | x0*x02+x1*m12   */
+    PFADD     ( MM5, MM6 )		/* ***trash***     | x0*...*m12+m32  */
+
+    MOVD      ( MM6, REGOFF(8, EDX) )	/* write r2                          */
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+
+    ADD_L     ( CONST(16), EDX )	/* next r                            */
+    DEC_L     ( ESI )			/* decrement vertex counter          */
+
+    JNZ       ( LLBL( G3TP3R_2 ) )	/* cnt > 0 ? -> process next vertex  */
+
+LLBL( G3TP3R_3 ):
+
+    FEMMS
+    POP_L     ( EDI )
+    POP_L     ( ESI )
+    RET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_3dnow_transform_points2_3d_no_rot )
+GLNAME( _mesa_3dnow_transform_points2_3d_no_rot ):
+
+    PUSH_L    ( ESI )
+
+    MOV_L     ( ARG_DEST, ECX )
+    MOV_L     ( ARG_MATRIX, ESI )
+    MOV_L     ( ARG_SOURCE, EAX )
+    MOV_L     ( CONST(3), REGOFF(V4F_SIZE, ECX) )
+    OR_B      ( CONST(VEC_SIZE_3 ), REGOFF(V4F_FLAGS, ECX) )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), EDX )
+    MOV_L     ( EDX, REGOFF(V4F_COUNT, ECX) )
+
+    PUSH_L    ( EDI )
+
+    MOV_L     ( REGOFF(V4F_START, ECX), EDX )
+    MOV_L     ( ESI, ECX )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), ESI )
+    MOV_L     ( REGOFF(V4F_STRIDE, EAX), EDI )
+    MOV_L     ( REGOFF(V4F_START, EAX), EAX )
+
+    TEST_L    ( ESI, ESI )
+    JZ        ( LLBL( G3TP3NRR_3 ) )
+
+    MOVD      ( REGIND(ECX), MM0 )	/*                 | m00             */
+    PUNPCKLDQ ( REGOFF(20, ECX), MM0 )	/* m11             | m00             */
+
+    MOVQ      ( REGOFF(48, ECX), MM2 )	/* m31             | m30             */
+    MOVD      ( REGOFF(56, ECX), MM3 )	/*                 | m32             */
+
+ALIGNTEXT16
+LLBL( G3TP3NRR_2 ):
+
+    MOVQ      ( REGIND(EAX), MM4 )	/* x1              | x0              */
+    PFMUL     ( MM0, MM4 )		/* x1*m11          | x0*m00          */
+
+    PFADD     ( MM2, MM4 )		/* x1*m11+m31      | x0*m00+m30      */
+    MOVQ      ( MM4, REGIND(EDX) )	/* write r1, r0                      */
+
+    MOVD      ( MM3, REGOFF(8, EDX) )	/* write r2                          */
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+
+    ADD_L     ( CONST(16), EDX )	/* next r                            */
+    DEC_L     ( ESI )			/* decrement vertex counter          */
+
+    JNZ       ( LLBL( G3TP3NRR_2 ) )	/* cnt > 0 ? -> process next vertex  */
+
+LLBL( G3TP3NRR_3 ):
+
+    FEMMS
+    POP_L     ( EDI )
+    POP_L     ( ESI )
+    RET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_3dnow_transform_points2_2d )
+GLNAME( _mesa_3dnow_transform_points2_2d ):
+
+    PUSH_L    ( ESI )
+
+    MOV_L     ( ARG_DEST, ECX )
+    MOV_L     ( ARG_MATRIX, ESI )
+    MOV_L     ( ARG_SOURCE, EAX )
+    MOV_L     ( CONST(2), REGOFF(V4F_SIZE, ECX) )
+    OR_B      ( CONST(VEC_SIZE_2), REGOFF(V4F_FLAGS, ECX) )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), EDX )
+    MOV_L     ( EDX, REGOFF(V4F_COUNT, ECX) )
+
+    PUSH_L    ( EDI )
+
+    MOV_L     ( REGOFF(V4F_START, ECX), EDX )
+    MOV_L     ( ESI, ECX )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), ESI )
+    MOV_L     ( REGOFF(V4F_STRIDE, EAX), EDI )
+    MOV_L     ( REGOFF(V4F_START, EAX), EAX )
+
+    TEST_L    ( ESI, ESI )
+    JZ        ( LLBL( G3TP2R_3 ) )
+
+    MOVQ      ( REGIND(ECX), MM0 )	/* m01             | m00             */
+    MOVQ      ( REGOFF(16, ECX), MM1 )	/* m11             | m10             */
+
+    MOVQ      ( REGOFF(48, ECX), MM2 )	/* m31             | m30             */
+
+ALIGNTEXT16
+LLBL( G3TP2R_2 ):
+
+    MOVD      ( REGIND(EAX), MM4 )	/*                 | x0              */
+    MOVD      ( REGOFF(4, EAX), MM5 )	/*                 | x1              */
+
+    PUNPCKLDQ ( MM4, MM4 )		/* x0              | x0              */
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+
+    PFMUL     ( MM0, MM4 )		/* x0*m01          | x0*m00          */
+    PUNPCKLDQ ( MM5, MM5 )		/* x1              | x1              */
+
+    PFMUL     ( MM1, MM5 )		/* x1*m11          | x1*m10          */
+    PFADD     ( MM2, MM4 )		/* x...x1*m11+31   | x0*..*m10+m30   */
+
+    PFADD     ( MM5, MM4 )		/* x0*m01+x1*m11   | x0*m00+x1*m10   */
+    MOVQ      ( MM4, REGIND(EDX) )	/* write r1, r0                      */
+
+    ADD_L     ( CONST(16), EDX )	/* next r                            */
+    DEC_L     ( ESI )			/* decrement vertex counter          */
+
+    JNZ       ( LLBL( G3TP2R_2 ) )	/* cnt > 0 ? -> process next vertex  */
+
+LLBL( G3TP2R_3 ):
+
+    FEMMS
+    POP_L     ( EDI )
+    POP_L     ( ESI )
+    RET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_3dnow_transform_points2_2d_no_rot )
+GLNAME( _mesa_3dnow_transform_points2_2d_no_rot ):
+
+    PUSH_L    ( ESI )
+
+    MOV_L     ( ARG_DEST, ECX )
+    MOV_L     ( ARG_MATRIX, ESI )
+    MOV_L     ( ARG_SOURCE, EAX )
+    MOV_L     ( CONST(2), REGOFF(V4F_SIZE, ECX) )
+    OR_B      ( CONST(VEC_SIZE_2), REGOFF(V4F_FLAGS, ECX) )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), EDX )
+    MOV_L     ( EDX, REGOFF(V4F_COUNT, ECX) )
+
+    PUSH_L    ( EDI )
+
+    MOV_L     ( REGOFF(V4F_START, ECX), EDX )
+    MOV_L     ( ESI, ECX )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), ESI )
+    MOV_L     ( REGOFF(V4F_STRIDE, EAX), EDI )
+    MOV_L     ( REGOFF(V4F_START, EAX), EAX )
+
+    TEST_L    ( ESI, ESI )
+    JZ        ( LLBL( G3TP2NRR_3 ) )
+
+    MOVD      ( REGIND(ECX), MM0 )	/*                 | m00             */
+    PUNPCKLDQ ( REGOFF(20, ECX), MM0 )	/* m11             | m00             */
+
+    MOVQ      ( REGOFF(48, ECX), MM2 )	/* m31             | m30             */
+
+ALIGNTEXT16
+LLBL( G3TP2NRR_2 ):
+
+    MOVQ      ( REGIND(EAX), MM4 )	/* x1              | x0              */
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+
+    PFMUL     ( MM0, MM4 )		/* x1*m11          | x0*m00          */
+    PFADD     ( MM2, MM4 )		/* m31             | x0*m00+m30      */
+
+    MOVQ      ( MM4, REGIND(EDX) )	/* write r1, r0                      */
+    ADD_L     ( CONST(16), EDX )	/* next r                            */
+
+    DEC_L     ( ESI )			/* decrement vertex counter          */
+    JNZ       ( LLBL( G3TP2NRR_2 ) )	/* cnt > 0 ? -> process next vertex  */
+
+LLBL( G3TP2NRR_3 ):
+
+    FEMMS
+    POP_L     ( EDI )
+    POP_L     ( ESI )
+    RET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_3dnow_transform_points2_identity )
+GLNAME( _mesa_3dnow_transform_points2_identity ):
+
+    PUSH_L    ( ESI )
+
+    MOV_L     ( ARG_DEST, ECX )
+    MOV_L     ( ARG_MATRIX, ESI )
+    MOV_L     ( ARG_SOURCE, EAX )
+    MOV_L     ( CONST(2), REGOFF(V4F_SIZE, ECX) )
+    OR_B      ( CONST(VEC_SIZE_2), REGOFF(V4F_FLAGS, ECX) )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), EDX )
+    MOV_L     ( EDX, REGOFF(V4F_COUNT, ECX) )
+
+    PUSH_L    ( EDI )
+
+    MOV_L     ( REGOFF(V4F_START, ECX), EDX )
+    MOV_L     ( ESI, ECX )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), ESI )
+    MOV_L     ( REGOFF(V4F_STRIDE, EAX), EDI )
+    MOV_L     ( REGOFF(V4F_START, EAX), EAX )
+
+    TEST_L    ( ESI, ESI )
+    JZ        ( LLBL( G3TPIR_3 ) )
+
+ALIGNTEXT16
+LLBL( G3TPIR_3 ):
+
+    MOVQ      ( REGIND(EAX), MM0 )	/* x1              | x0              */
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+
+    MOVQ      ( MM0, REGIND(EDX) )	/* r1              | r0              */
+    ADD_L     ( CONST(16), EDX )	/* next r                            */
+
+    DEC_L     ( ESI )			/* decrement vertex counter          */
+    JNZ       ( LLBL( G3TPIR_3 ) )	/* cnt > 0 ? -> process next vertex  */
+
+LLBL( G3TPIR_4 ):
+
+    FEMMS
+    POP_L     ( EDI )
+    POP_L     ( ESI )
+    RET
diff --git a/src/mesa/x86/3dnow_xform3.S b/src/mesa/x86/3dnow_xform3.S
new file mode 100644
index 00000000000..7d225b33bc4
--- /dev/null
+++ b/src/mesa/x86/3dnow_xform3.S
@@ -0,0 +1,570 @@
+/* $Id: 3dnow_xform3.S,v 1.1 2001/03/29 06:46:16 gareth Exp $ */
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  3.5
+ *
+ * Copyright (C) 1999-2001  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.
+ */
+
+#include "matypes.h"
+#include "xform_args.h"
+
+    SEG_TEXT
+
+#define FRAME_OFFSET	4
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_3dnow_transform_points3_general )
+GLNAME( _mesa_3dnow_transform_points3_general ):
+
+    PUSH_L    ( ESI )
+
+    MOV_L     ( ARG_DEST, ECX )
+    MOV_L     ( ARG_MATRIX, ESI )
+    MOV_L     ( ARG_SOURCE, EAX )
+    MOV_L     ( CONST(4), REGOFF(V4F_SIZE, ECX) )
+    OR_B      ( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, ECX) )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), EDX )
+    MOV_L     ( EDX, REGOFF(V4F_COUNT, ECX) )
+
+    PUSH_L    ( EDI )
+
+    MOV_L     ( REGOFF(V4F_START, ECX), EDX )
+    MOV_L     ( ESI, ECX )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), ESI )
+    MOV_L     ( REGOFF(V4F_STRIDE, EAX), EDI )
+    MOV_L     ( REGOFF(V4F_START, EAX), EAX )
+
+    TEST_L    ( ESI, ESI )
+    JZ        ( LLBL( G3TPGR_2 ) )
+
+    PREFETCHW ( REGIND(EDX) )
+
+    MOVQ      ( REGIND(EAX), MM0 )	/* x1              | x0              */
+    MOVD      ( REGOFF(8, EAX), MM2 )	/*                 | x2              */
+
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+
+ALIGNTEXT16
+LLBL( G3TPGR_1 ):
+
+    PREFETCHW ( REGOFF(32, EDX) )	/* prefetch 2 vertices ahead         */
+    PREFETCH  ( REGIND(EAX) )
+
+    MOVQ      ( MM0, MM1 )		/* x1              | x0              */
+    PUNPCKLDQ ( MM2, MM2 )		/* x2              | x2              */
+
+    PUNPCKLDQ ( MM0, MM0 )		/* x0              | x0              */
+    MOVQ      ( MM2, MM5 )		/* x2              | x2              */
+
+    PUNPCKHDQ ( MM1, MM1 )		/* x1              | x1              */
+    PFMUL     ( REGOFF(32, ECX), MM2 )	/* x2*m9           | x2*m8           */
+
+    MOVQ      ( MM0, MM3 )		/* x0              | x0              */
+    PFMUL     ( REGOFF(40, ECX), MM5 )	/* x2*m11          | x2*m10          */
+
+    MOVQ      ( MM1, MM4 )		/* x1              | x1              */
+    PFMUL     ( REGIND(ECX), MM0 )	/* x0*m1           | x0*m0           */
+
+    PFADD     ( REGOFF(48, ECX), MM2 )	/* x2*m9+m13       | x2*m8+m12       */
+    PFMUL     ( REGOFF(16, ECX), MM1 )	/* x1*m5           | x1*m4           */
+
+    PFADD     ( REGOFF(56, ECX), MM5 )	/* x2*m11+m15      | x2*m10+m14      */
+    PFADD     ( MM0, MM1 )		/* x0*m1+x1*m5     | x0*m0+x1*m4     */
+
+    PFMUL     ( REGOFF(8, ECX), MM3 )	/* x0*m3           | x0*m2           */
+    PFADD     ( MM1, MM2 )		/* r1              | r0              */
+
+    PFMUL     ( REGOFF(24, ECX), MM4 )	/* x1*m7           | x1*m6           */
+    ADD_L     ( CONST(16), EDX )	/* next output vertex                */
+
+    PFADD     ( MM3, MM4 )		/* x0*m3+x1*m7     | x0*m2+x1*m6     */
+    MOVQ      ( MM2, REGOFF(-16, EDX) )	/* write r0, r1                      */
+
+    PFADD     ( MM4, MM5 )		/* r3              | r2              */
+    MOVQ      ( MM5, REGOFF(-8, EDX) )	/* write r2, r3                      */
+
+    MOVQ      ( REGIND(EAX), MM0 )	/* x1              | x0              */
+    MOVD      ( REGOFF(8, EAX), MM2 )	/*                 | x2              */
+
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+    DEC_L     ( ESI )			/* decrement vertex counter          */
+    JNZ       ( LLBL( G3TPGR_1 ) )	/* cnt > 0 ? -> process next vertex  */
+
+LLBL( G3TPGR_2 ):
+
+    FEMMS
+    POP_L     ( EDI )
+    POP_L     ( ESI )
+    RET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_3dnow_transform_points3_perspective )
+GLNAME( _mesa_3dnow_transform_points3_perspective ):
+
+    PUSH_L    ( ESI )
+
+    MOV_L     ( ARG_DEST, ECX )
+    MOV_L     ( ARG_MATRIX, ESI )
+    MOV_L     ( ARG_SOURCE, EAX )
+    MOV_L     ( CONST(4), REGOFF(V4F_SIZE, ECX) )
+    OR_B      ( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, ECX) )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), EDX )
+    MOV_L     ( EDX, REGOFF(V4F_COUNT, ECX) )
+
+    PUSH_L    ( EDI )
+
+    MOV_L     ( REGOFF(V4F_START, ECX), EDX )
+    MOV_L     ( ESI, ECX )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), ESI )
+    MOV_L     ( REGOFF(V4F_STRIDE, EAX), EDI )
+    MOV_L     ( REGOFF(V4F_START, EAX), EAX )
+
+    TEST_L    ( ESI, ESI )
+    JZ        ( LLBL( G3TPPR_2 ) )
+
+    PREFETCH  ( REGIND(EAX) )
+    PREFETCHW ( REGIND(EDX) )
+
+    MOVD      ( REGIND(ECX), MM0 )	/*                 | m00             */
+    PUNPCKLDQ ( REGOFF(20, ECX), MM0 )	/* m11             | m00             */
+
+    MOVQ      ( REGOFF(32, ECX), MM1 )	/* m21             | m20             */
+    MOVD      ( REGOFF(40, ECX), MM2 )	/*                 | m22             */
+
+    MOVD      ( REGOFF(56, ECX), MM3 )	/*                 | m32             */
+
+    MOVD      ( REGOFF(8, EAX), MM5 )	/*                 | x2              */
+    MOVQ      ( REGIND(EAX), MM4 )	/* x1              | x0              */
+
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+
+ALIGNTEXT16
+LLBL( G3TPPR_1 ):
+
+    PREFETCHW ( REGOFF(32, EDX) )	/* prefetch 2 vertices ahead         */
+    PREFETCH  ( REGIND(EAX) )
+
+    PXOR      ( MM7, MM7 )		/* 0               | 0               */
+    MOVQ      ( MM5, MM6 )		/*                 | x2              */
+
+    PFMUL     ( MM0, MM4 )		/* x1*m11          | x0*m00          */
+    PFSUB     ( MM5, MM7 )		/*                 | -x2             */
+
+    PFMUL     ( MM2, MM6 )		/*                 | x2*m22          */
+    PUNPCKLDQ ( MM5, MM5 )		/* x2              | x2              */
+
+    ADD_L     ( CONST(16), EDX )	/* next r                            */
+    PFMUL     ( MM1, MM5 )		/* x2*m21          | x2*m20          */
+
+    PFADD     ( MM3, MM6 )		/*                 | x2*m22+m32      */
+    PFADD     ( MM4, MM5 )		/* x1*m11+x2*m21   | x0*m00+x2*m20   */
+
+    MOVQ      ( MM5, REGOFF(-16, EDX) )	/* write r0, r1                      */
+    MOVD      ( MM6, REGOFF(-8, EDX) )	/* write r2                          */
+
+    MOVD      ( MM7, REGOFF(-4, EDX) )	/* write r3                          */
+    MOVD      ( REGOFF(8, EAX), MM5 )	/*                 | x2              */
+
+    MOVQ      ( REGIND(EAX), MM4 )	/* x1              | x0              */
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+
+    DEC_L     ( ESI )			/* decrement vertex counter          */
+    JNZ       ( LLBL( G3TPPR_1 ) )	/* cnt > 0 ? -> process next vertex  */
+
+LLBL( G3TPPR_2 ):
+
+    FEMMS
+    POP_L     ( EDI )
+    POP_L     ( ESI )
+    RET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_3dnow_transform_points3_3d )
+GLNAME( _mesa_3dnow_transform_points3_3d ):
+
+    PUSH_L    ( ESI )
+
+    MOV_L     ( ARG_DEST, ECX )
+    MOV_L     ( ARG_MATRIX, ESI )
+    MOV_L     ( ARG_SOURCE, EAX )
+    MOV_L     ( CONST(3), REGOFF(V4F_SIZE, ECX) )
+    OR_B      ( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, ECX) )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), EDX )
+    MOV_L     ( EDX, REGOFF(V4F_COUNT, ECX) )
+
+    PUSH_L    ( EDI )
+
+    MOV_L     ( REGOFF(V4F_START, ECX), EDX )
+    MOV_L     ( ESI, ECX )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), ESI )
+    MOV_L     ( REGOFF(V4F_STRIDE, EAX), EDI )
+    MOV_L     ( REGOFF(V4F_START, EAX), EAX )
+
+    TEST_L    ( ESI, ESI )
+    JZ        ( LLBL( G3TP3R_2 ) )
+
+    PREFETCH  ( REGIND(EAX) )
+    PREFETCH  ( REGIND(EDX) )
+
+    MOVD      ( REGOFF(8, ECX), MM7 )	/*                 | m2              */
+    PUNPCKLDQ ( REGOFF(24, ECX), MM7 )	/* m6              | m2              */
+
+    MOVQ      ( REGIND(EAX), MM0 )	/* x1              | x0              */
+    MOVD      ( REGOFF(8, EAX), MM1 )	/*                 | x2              */
+
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+
+ALIGNTEXT16
+LLBL( G3TP3R_1 ):
+
+    PREFETCHW ( REGOFF(32, EDX) )	/* prefetch 2 vertices ahead         */
+    PREFETCH  ( REGIND(EAX) )
+
+    MOVQ      ( MM0, MM2 )		/* x1              | x0              */
+    ADD_L     ( CONST(16), EDX )	/* next r                            */
+
+    PUNPCKLDQ ( MM2, MM2 )		/* x0              | x0              */
+    MOVQ      ( MM0, MM3 )		/* x1              | x0              */
+
+    PFMUL     ( REGIND(ECX), MM2 )	/* x0*m1           | x0*m0           */
+    PUNPCKHDQ ( MM3, MM3 )		/* x1              | x1              */
+
+    MOVQ      ( MM1, MM4 )		/*                 | x2              */
+    PFMUL     ( REGOFF(16, ECX), MM3 )	/* x1*m5           | x1*m4           */
+
+    PUNPCKLDQ ( MM4, MM4 )		/* x2              | x2              */
+    PFADD     ( MM2, MM3 )		/* x0*m1+x1*m5     | x0*m0+x1*m4     */
+
+    PFMUL     ( REGOFF(32, ECX), MM4 )	/* x2*m9           | x2*m8           */
+    PFADD     ( REGOFF(48, ECX), MM3 )	/* x0*m1+...+m11   | x0*m0+x1*m4+m12 */
+
+    PFMUL     ( MM7, MM0 )		/* x1*m6           | x0*m2           */
+    PFADD     ( MM4, MM3 )		/* r1              | r0              */
+
+    PFMUL     ( REGOFF(40, ECX), MM1 )	/*                 | x2*m10          */
+    PUNPCKLDQ ( REGOFF(56, ECX), MM1 )	/* m14             | x2*m10          */
+
+    PFACC     ( MM0, MM1 )
+
+    MOVQ      ( MM3, REGOFF(-16, EDX) )	/* write r0, r1                      */
+    PFACC     ( MM1, MM1 )		/*                 | r2              */
+
+    MOVD      ( MM1, REGOFF(-8, EDX) )	/* write r2                          */
+    MOVQ      ( REGIND(EAX), MM0 )	/* x1              | x0              */
+
+    MOVD      ( REGOFF(8, EAX), MM1 )	/*                 | x2              */
+
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+    DEC_L     ( ESI )			/* decrement vertex counter          */
+    JNZ       ( LLBL( G3TP3R_1 ) )	/* cnt > 0 ? -> process next vertex  */
+
+LLBL( G3TP3R_2 ):
+
+    FEMMS
+    POP_L     ( EDI )
+    POP_L     ( ESI )
+    RET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_3dnow_transform_points3_3d_no_rot )
+GLNAME( _mesa_3dnow_transform_points3_3d_no_rot ):
+
+    PUSH_L    ( ESI )
+
+    MOV_L     ( ARG_DEST, ECX )
+    MOV_L     ( ARG_MATRIX, ESI )
+    MOV_L     ( ARG_SOURCE, EAX )
+    MOV_L     ( CONST(3), REGOFF(V4F_SIZE, ECX) )
+    OR_B      ( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, ECX) )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), EDX )
+    MOV_L     ( EDX, REGOFF(V4F_COUNT, ECX) )
+
+    PUSH_L    ( EDI )
+
+    MOV_L     ( REGOFF(V4F_START, ECX), EDX )
+    MOV_L     ( ESI, ECX )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), ESI )
+    MOV_L     ( REGOFF(V4F_STRIDE, EAX), EDI )
+    MOV_L     ( REGOFF(V4F_START, EAX), EAX )
+
+    TEST_L    ( ESI, ESI )
+    JZ        ( LLBL( G3TP3NRR_2 ) )
+
+    PREFETCH  ( REGIND(EAX) )
+    PREFETCHW ( REGIND(EDX) )
+
+    MOVD      ( REGIND(ECX), MM0 )	/*                 | m00             */
+    PUNPCKLDQ ( REGOFF(20, ECX), MM0 )	/* m11             | m00             */
+
+    MOVD      ( REGOFF(40, ECX), MM2 )	/*                 | m22             */
+    PUNPCKLDQ ( MM2, MM2 )		/* m22             | m22             */
+
+    MOVQ      ( REGOFF(48, ECX), MM1 )	/* m31             | m30             */
+    MOVD      ( REGOFF(56, ECX), MM3 )	/*                 | m32             */
+
+    PUNPCKLDQ ( MM3, MM3 )		/* m32             | m32             */
+    MOVQ      ( REGIND(EAX), MM4 )	/* x1              | x0              */
+
+    MOVD      ( REGOFF(8, EAX), MM5 )	/*                 | x2              */
+
+
+ALIGNTEXT16
+LLBL( G3TP3NRR_1 ):
+
+    PREFETCHW ( REGOFF(32, EDX) )	/* prefetch 2 vertices ahead         */
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+
+    PREFETCHW ( REGIND(EAX) )
+    PFMUL     ( MM0, MM4 )		/* x1*m11          | x0*m00          */
+
+    PFADD     ( MM1, MM4 )		/* x1*m11+m31      | x0*m00+m30      */
+    PFMUL     ( MM2, MM5 )		/*                 | x2*m22          */
+
+    PFADD     ( MM3, MM5 )		/*                 | x2*m22+m32      */
+    MOVQ      ( MM4, REGIND(EDX) )	/* write r0, r1                      */
+
+    ADD_L     ( CONST(16), EDX )	/* next r                            */
+    DEC_L     ( ESI )			/* decrement vertex counter          */
+
+    MOVD      ( MM5, REGOFF(-8, EDX) )	/* write r2                          */
+    MOVQ      ( REGIND(EAX), MM4 )	/* x1              | x0              */
+
+    MOVD      ( REGOFF(8, EAX), MM5 )	/*                 | x2              */
+    JNZ       ( LLBL( G3TP3NRR_1 ) )	/* cnt > 0 ? -> process next vertex  */
+
+LLBL( G3TP3NRR_2 ):
+
+    FEMMS
+    POP_L     ( EDI )
+    POP_L     ( ESI )
+    RET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_3dnow_transform_points3_2d )
+GLNAME( _mesa_3dnow_transform_points3_2d ):
+
+    PUSH_L    ( ESI )
+
+    MOV_L     ( ARG_DEST, ECX )
+    MOV_L     ( ARG_MATRIX, ESI )
+    MOV_L     ( ARG_SOURCE, EAX )
+    MOV_L     ( CONST(3), REGOFF(V4F_SIZE, ECX) )
+    OR_B      ( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, ECX) )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), EDX )
+    MOV_L     ( EDX, REGOFF(V4F_COUNT, ECX) )
+
+    PUSH_L    ( EDI )
+
+    MOV_L     ( REGOFF(V4F_START, ECX), EDX )
+    MOV_L     ( ESI, ECX )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), ESI )
+    MOV_L     ( REGOFF(V4F_STRIDE, EAX), EDI )
+    MOV_L     ( REGOFF(V4F_START, EAX), EAX )
+
+    TEST_L    ( ESI, ESI )
+    JZ        ( LLBL( G3TP2R_3) )
+
+    PREFETCH  ( REGIND(EAX) )
+    PREFETCHW ( REGIND(EDX) )
+
+    MOVD      ( REGIND(ECX), MM0 )	/*                 | m00             */
+    PUNPCKLDQ ( REGOFF(16, ECX), MM0 )	/* m10             | m00             */
+
+    MOVD      ( REGOFF(4, ECX), MM1 )	/*                 | m01             */
+    PUNPCKLDQ ( REGOFF(20, ECX), MM1 )	/* m11             | m01             */
+
+    MOVQ      ( REGOFF(48, ECX), MM2 )	/* m31             | m30             */
+    MOVQ      ( REGIND(EAX), MM3 )	/* x1              | x0              */
+
+    MOVD      ( REGOFF(8, EAX), MM5 )	/*                 | x2              */
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+
+ALIGNTEXT16
+LLBL( G3TP2R_2 ):
+
+    PREFETCHW ( REGOFF(32, EDX) )	/* prefetch 2 vertices ahead         */
+    PREFETCH  ( REGIND(EAX) )
+
+    MOVQ      ( MM3, MM4 )		/* x1              | x0              */
+    PFMUL     ( MM0, MM3 )		/* x1*m10          | x0*m00          */
+
+    ADD_L     ( CONST(16), EDX )	/* next r                            */
+    PFMUL     ( MM1, MM4 )		/* x1*m11          | x0*m01          */
+
+    PFACC     ( MM4, MM3 )		/* x0*m00+x1*m10   | x0*m01+x1*m11   */
+    MOVD      ( MM5, REGOFF(-8, EDX) )	/* write r2 (=x2)                    */
+
+    PFADD     ( MM2, MM3 )		/* x0*...*m10+m30  | x0*...*m11+m31  */
+    MOVQ      ( MM3, REGOFF(-16, EDX) )	/* write r0, r1                      */
+
+    MOVQ      ( REGIND(EAX), MM3 )	/* x1              | x0              */
+    MOVD      ( REGOFF(8, EAX), MM5 )	/*                 | x2              */
+
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+    DEC_L     ( ESI )			/* decrement vertex counter          */
+    JNZ       ( LLBL( G3TP2R_2 ) )	/* cnt > 0 ? -> process next vertex  */
+
+LLBL( G3TP2R_3 ):
+
+    FEMMS
+    POP_L     ( EDI )
+    POP_L     ( ESI )
+    RET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_3dnow_transform_points3_2d_no_rot )
+GLNAME( _mesa_3dnow_transform_points3_2d_no_rot ):
+
+    PUSH_L    ( ESI )
+
+    MOV_L     ( ARG_DEST, ECX )
+    MOV_L     ( ARG_MATRIX, ESI )
+    MOV_L     ( ARG_SOURCE, EAX )
+    MOV_L     ( CONST(3), REGOFF(V4F_SIZE, ECX) )
+    OR_B      ( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, ECX) )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), EDX )
+    MOV_L     ( EDX, REGOFF(V4F_COUNT, ECX) )
+
+    PUSH_L    ( EDI )
+
+    MOV_L     ( REGOFF(V4F_START, ECX), EDX )
+    MOV_L     ( ESI, ECX )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), ESI )
+    MOV_L     ( REGOFF(V4F_STRIDE, EAX), EDI )
+    MOV_L     ( REGOFF(V4F_START, EAX), EAX )
+
+    TEST_L    ( ESI, ESI )
+    JZ        ( LLBL( G3TP2NRR_2 ) )
+
+    PREFETCH  ( REGIND(EAX) )
+    PREFETCHW ( REGIND(EDX) )
+
+    MOVD      ( REGIND(ECX), MM0 )	/*                 | m00             */
+    PUNPCKLDQ ( REGOFF(20, ECX), MM0 )	/* m11             | m00             */
+
+    MOVQ      ( REGOFF(48, ECX), MM1 )	/* m31             | m30             */
+    MOVQ      ( REGIND(EAX), MM4 )	/* x1              | x0              */
+
+    MOVD      ( REGOFF(8, EAX), MM5 )	/*                 | x2              */
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+
+
+ALIGNTEXT16
+LLBL( G3TP2NRR_1 ):
+
+    PREFETCHW ( REGOFF(32, EDX) )	/* prefetch 2 vertices ahead         */
+    PREFETCH  ( REGIND(EAX) )
+
+    PFMUL     ( MM0, MM4 )		/* x1*m11          | x0*m00          */
+    ADD_L     ( CONST(16), EDX )	/* next r                            */
+
+    PFADD     ( MM1, MM4 )		/* x1*m11+m31      | x0*m00+m30      */
+
+    MOVQ      ( MM4, REGOFF(-16, EDX) )	/* write r0, r1                      */
+    MOVD      ( MM5, REGOFF(-8, EDX) )	/* write r2 (=x2)                    */
+
+    MOVQ      ( REGIND(EAX), MM4 )	/* x1              | x0              */
+    MOVD      ( REGOFF(8, EAX), MM5 )	/*                 | x2              */
+
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+    DEC_L     ( ESI )			/* decrement vertex counter          */
+    JNZ       ( LLBL( G3TP2NRR_1 ) )	/* cnt > 0 ? -> process next vertex  */
+
+LLBL( G3TP2NRR_2 ):
+
+    FEMMS
+    POP_L     ( EDI )
+    POP_L     ( ESI )
+    RET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_3dnow_transform_points3_identity )
+GLNAME( _mesa_3dnow_transform_points3_identity ):
+
+    PUSH_L    ( ESI )
+
+    MOV_L     ( ARG_DEST, ECX )
+    MOV_L     ( ARG_MATRIX, ESI )
+    MOV_L     ( ARG_SOURCE, EAX )
+    MOV_L     ( CONST(3), REGOFF(V4F_SIZE, ECX) )
+    OR_B      ( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, ECX) )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), EDX )
+    MOV_L     ( EDX, REGOFF(V4F_COUNT, ECX) )
+
+    PUSH_L    ( EDI )
+
+    MOV_L     ( REGOFF(V4F_START, ECX), EDX )
+    MOV_L     ( ESI, ECX )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), ESI )
+    MOV_L     ( REGOFF(V4F_STRIDE, EAX), EDI )
+    MOV_L     ( REGOFF(V4F_START, EAX), EAX )
+
+    TEST_L    ( ESI, ESI )
+    JZ        ( LLBL( G3TPIR_2 ) )
+
+    PREFETCHW ( REGIND(EDX) )
+
+    MOVQ      ( REGIND(EAX), MM0 )	/* x1              | x0              */
+    MOVD      ( REGOFF(8, EAX), MM1 )	/*                 | x2              */
+
+ALIGNTEXT16
+LLBL( G3TPIR_1 ):
+
+    PREFETCHW ( REGOFF(32, EDX) )
+
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+    ADD_L     ( CONST(16), EDX )	/* next r                            */
+
+    DEC_L     ( ESI )			/* decrement vertex counter          */
+    MOVQ      ( MM0, REGOFF(-16, EDX) )	/* r1              | r0              */
+
+    MOVD      ( MM1, REGOFF(-8, EDX) )	/*                 | r2              */
+    MOVQ      ( REGIND(EAX), MM0 )	/* x1              | x0              */
+
+    MOVD      ( REGOFF(8, EAX), MM1 )	/*                 | x2              */
+    JNZ       ( LLBL( G3TPIR_1 ) )	/* cnt > 0 ? -> process next vertex  */
+
+LLBL( G3TPIR_2 ):
+
+    FEMMS
+    POP_L     ( EDI )
+    POP_L     ( ESI )
+    RET
diff --git a/src/mesa/x86/3dnow_xform4.S b/src/mesa/x86/3dnow_xform4.S
new file mode 100644
index 00000000000..06d7c6950a2
--- /dev/null
+++ b/src/mesa/x86/3dnow_xform4.S
@@ -0,0 +1,588 @@
+/* $Id: 3dnow_xform4.S,v 1.1 2001/03/29 06:46:16 gareth Exp $ */
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  3.5
+ *
+ * Copyright (C) 1999-2001  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.
+ */
+
+#include "matypes.h"
+#include "xform_args.h"
+
+    SEG_TEXT
+
+#define FRAME_OFFSET	4
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_3dnow_transform_points4_general )
+GLNAME( _mesa_3dnow_transform_points4_general ):
+
+    PUSH_L    ( ESI )
+
+    MOV_L     ( ARG_DEST, ECX )
+    MOV_L     ( ARG_MATRIX, ESI )
+    MOV_L     ( ARG_SOURCE, EAX )
+    MOV_L     ( CONST(4), REGOFF(V4F_SIZE, ECX) )
+    OR_B      ( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, ECX) )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), EDX )
+    MOV_L     ( EDX, REGOFF(V4F_COUNT, ECX) )
+
+    PUSH_L    ( EDI )
+
+    MOV_L     ( REGOFF(V4F_START, ECX), EDX )
+    MOV_L     ( ESI, ECX )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), ESI )
+    MOV_L     ( REGOFF(V4F_STRIDE, EAX), EDI )
+    MOV_L     ( REGOFF(V4F_START, EAX), EAX )
+
+    TEST_L    ( ESI, ESI )
+    JZ        ( LLBL( G3TPGR_2 ) )
+
+    PREFETCHW ( REGIND(EDX) )
+
+    MOVQ      ( REGIND(EAX), MM0 )	/* x1              | x0              */
+    MOVQ      ( REGOFF(8, EAX), MM4 )	/* x3              | x2              */
+
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+
+ALIGNTEXT16
+LLBL( G3TPGR_1 ):
+
+    PREFETCHW ( REGOFF(32, EDX) )	/* prefetch 2 vertices ahead         */
+    PREFETCH  ( REGIND(EAX) )
+
+    MOVQ      ( MM0, MM2 )		/* x1              | x0              */
+    MOVQ      ( MM4, MM6 )		/* x3              | x2              */
+
+    PUNPCKLDQ ( MM0, MM0 )		/* x0              | x0              */
+    PUNPCKHDQ ( MM2, MM2 )		/* x1              | x1              */
+
+    MOVQ      ( MM0, MM1 )		/* x0              | x0              */
+    ADD_L     ( CONST(16), EDX )	/* next r                            */
+
+    PFMUL     ( REGIND(ECX), MM0 )	/* x0*m1           | x0*m0           */
+    MOVQ      ( MM2, MM3 )		/* x1              | x1              */
+
+    PFMUL     ( REGOFF(8, ECX), MM1 )	/* x0*m3           | x0*m2           */
+    PUNPCKLDQ ( MM4, MM4 )		/* x2              | x2              */
+
+    PFMUL     ( REGOFF(16, ECX), MM2 )	/* x1*m5           | x1*m4           */
+    MOVQ      ( MM4, MM5 )		/* x2              | x2              */
+
+    PFMUL     ( REGOFF(24, ECX), MM3 )	/* x1*m7           | x1*m6           */
+    PUNPCKHDQ ( MM6, MM6 )		/* x3              | x3              */
+
+    PFMUL     ( REGOFF(32, ECX), MM4 )	/* x2*m9           | x2*m8           */
+    MOVQ      ( MM6, MM7 )		/* x3              | x3              */
+
+    PFMUL     ( REGOFF(40, ECX), MM5 )	/* x2*m11          | x2*m10          */
+    PFADD     ( MM0, MM2 )
+
+    PFMUL     ( REGOFF(48, ECX), MM6 )	/* x3*m13          | x3*m12          */
+    PFADD     ( MM1, MM3 )
+
+    PFMUL     ( REGOFF(56, ECX), MM7 )	/* x3*m15          | x3*m14          */
+    PFADD     ( MM4, MM6 )
+
+    PFADD     ( MM5, MM7 )
+    PFADD     ( MM2, MM6 )
+
+    PFADD     ( MM3, MM7 )
+    MOVQ      ( MM6, REGOFF(-16, EDX) )
+
+    MOVQ      ( MM7, REGOFF(-8, EDX) )
+    MOVQ      ( REGIND(EAX), MM0 )	/* x1            | x0                */
+
+    MOVQ      ( REGOFF(8, EAX), MM4 )	/* x3            | x2                */
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+
+    DEC_L     ( ESI )			/* decrement vertex counter          */
+    JNZ       ( LLBL( G3TPGR_1 ) )	/* cnt > 0 ? -> process next vertex  */
+
+LLBL( G3TPGR_2 ):
+
+    FEMMS
+    POP_L     ( EDI )
+    POP_L     ( ESI )
+    RET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_3dnow_transform_points4_perspective )
+GLNAME( _mesa_3dnow_transform_points4_perspective ):
+
+    PUSH_L    ( ESI )
+
+    MOV_L     ( ARG_DEST, ECX )
+    MOV_L     ( ARG_MATRIX, ESI )
+    MOV_L     ( ARG_SOURCE, EAX )
+    MOV_L     ( CONST(4), REGOFF(V4F_SIZE, ECX) )
+    OR_B      ( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, ECX) )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), EDX )
+    MOV_L     ( EDX, REGOFF(V4F_COUNT, ECX) )
+
+    PUSH_L    ( EDI )
+
+    MOV_L     ( REGOFF(V4F_START, ECX), EDX )
+    MOV_L     ( ESI, ECX )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), ESI )
+    MOV_L     ( REGOFF(V4F_STRIDE, EAX), EDI )
+    MOV_L     ( REGOFF(V4F_START, EAX), EAX )
+
+    TEST_L    ( ESI, ESI )
+    JZ        ( LLBL( G3TPPR_2 ) )
+
+    PREFETCH  ( REGIND(EAX) )
+    PREFETCHW ( REGIND(EDX) )
+
+    MOVD      ( REGIND(ECX), MM0 )	/*                 | m00             */
+    PUNPCKLDQ ( REGOFF(20, ECX), MM0 )	/* m11             | m00             */
+
+    MOVD      ( REGOFF(40, ECX), MM1 )	/*                 | m22             */
+    PUNPCKLDQ ( REGOFF(56, ECX), MM1 )	/* m32             | m22             */
+
+    MOVQ      ( REGOFF(32, ECX), MM2 )	/* m21             | m20             */
+    PXOR      ( MM7, MM7 )		/* 0               | 0               */
+
+    MOVQ      ( REGIND(EAX), MM4 )	/* x1              | x0              */
+    MOVQ      ( REGOFF(8, EAX), MM5 )	/* x3              | x2              */
+    MOVD      ( REGOFF(8, EAX), MM3 )	/*                 | x2              */
+
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+
+ALIGNTEXT16
+LLBL( G3TPPR_1 ):
+
+    PREFETCHW ( REGOFF(32, EDX) )	/* prefetch 2 vertices ahead         */
+    PREFETCH  ( REGOFF(32, EAX) )	/* hopefully stride is zero          */
+
+    MOVQ      ( MM5, MM6 )		/* x3              | x2              */
+    PFMUL     ( MM0, MM4 )		/* x1*m11          | x0*m00          */
+
+    PUNPCKLDQ ( MM5, MM5 )		/* x2              | x2              */
+    ADD_L     ( CONST(16), EDX )	/* next r                            */
+
+    PFMUL     ( MM2, MM5 )		/* x2*m21          | x2*m20          */
+    PFSUBR    ( MM7, MM3 )		/*                 | -x2             */
+
+    PFMUL     ( MM1, MM6 )		/* x3*m32          | x2*m22          */
+    PFADD     ( MM4, MM5 )		/* x1*m11+x2*m21   | x0*m00+x2*m20   */
+
+    PFACC     ( MM3, MM6 )		/* -x2             | x2*m22+x3*m32   */
+    MOVQ      ( MM5, REGOFF(-16, EDX) )	/* write r0, r1                      */
+
+    MOVQ      ( MM6, REGOFF(-8, EDX) )	/* write r2, r3                      */
+    MOVQ      ( REGIND(EAX), MM4 )	/* x1              | x0              */
+
+    MOVQ      ( REGOFF(8, EAX), MM5 )	/* x3              | x2              */
+    MOVD      ( REGOFF(8, EAX), MM3 )	/*                 | x2              */
+
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+    DEC_L     ( ESI )			/* decrement vertex counter          */
+
+    JNZ       ( LLBL( G3TPPR_1 ) )	/* cnt > 0 ? -> process next vertex  */
+
+LLBL( G3TPPR_2 ):
+
+    FEMMS
+    POP_L     ( EDI )
+    POP_L     ( ESI )
+    RET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_3dnow_transform_points4_3d )
+GLNAME( _mesa_3dnow_transform_points4_3d ):
+
+    PUSH_L    ( ESI )
+
+    MOV_L     ( ARG_DEST, ECX )
+    MOV_L     ( ARG_MATRIX, ESI )
+    MOV_L     ( ARG_SOURCE, EAX )
+    MOV_L     ( CONST(4), REGOFF(V4F_SIZE, ECX) )
+    OR_B      ( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, ECX) )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), EDX )
+    MOV_L     ( EDX, REGOFF(V4F_COUNT, ECX) )
+
+    PUSH_L    ( EDI )
+
+    MOV_L     ( REGOFF(V4F_START, ECX), EDX )
+    MOV_L     ( ESI, ECX )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), ESI )
+    MOV_L     ( REGOFF(V4F_STRIDE, EAX), EDI )
+    MOV_L     ( REGOFF(V4F_START, EAX), EAX )
+
+    TEST_L    ( ESI, ESI )
+    JZ        ( LLBL( G3TP3R_2 ) )
+
+    MOVD      ( REGOFF(8, ECX), MM6 )	/*                 | m2              */
+    PUNPCKLDQ ( REGOFF(24, ECX), MM6 )	/* m6              | m2              */
+
+    MOVD      ( REGOFF(40, ECX), MM7 )	/*                 | m10             */
+    PUNPCKLDQ ( REGOFF(56, ECX), MM7 )	/* m14             | m10             */
+
+    MOVQ      ( REGIND(EAX), MM2 )	/* x1              | x0              */
+    MOVQ      ( REGOFF(8, EAX), MM3 )	/* x3              | x2              */
+
+ALIGNTEXT16
+LLBL( G3TP3R_1 ):
+
+    PREFETCHW ( REGOFF(32, EDX) )	/* prefetch 2 vertices ahead         */
+    PREFETCH  ( REGOFF(32, EAX) )	/* hopefully array is tightly packed */
+
+    MOVQ      ( MM2, MM0 )		/* x1              | x0              */
+    MOVQ      ( MM3, MM4 )		/* x3              | x2              */
+
+    MOVQ      ( MM0, MM1 )		/* x1              | x0              */
+    MOVQ      ( MM4, MM5 )		/* x3              | x2              */
+
+    PUNPCKLDQ ( MM0, MM0 )		/* x0              | x0              */
+    PUNPCKHDQ ( MM1, MM1 )		/* x1              | x1              */
+
+    PFMUL     ( REGIND(ECX), MM0 )	/* x0*m1           | x0*m0           */
+    PUNPCKLDQ ( MM3, MM3 )		/* x2              | x2              */
+
+    PFMUL     ( REGOFF(16, ECX), MM1 )	/* x1*m5           | x1*m4           */
+    PUNPCKHDQ ( MM4, MM4 )		/* x3              | x3              */
+
+    PFMUL     ( MM6, MM2 )		/* x1*m6           | x0*m2           */
+    PFADD     ( MM0, MM1 )		/* x0*m1+x1*m5     | x0*m0+x1*m4     */
+
+    PFMUL     ( REGOFF(32, ECX), MM3 )	/* x2*m9           | x2*m8           */
+    ADD_L     ( CONST(16), EDX )	/* next r                            */
+
+    PFMUL     ( REGOFF(48, ECX), MM4 )	/* x3*m13          | x3*m12          */
+    PFADD     ( MM1, MM3 )		/* x0*m1+..+x2*m9  | x0*m0+...+x2*m8 */
+
+    PFMUL     ( MM7, MM5 )		/* x3*m14          | x2*m10          */
+    PFADD     ( MM3, MM4 )		/* r1              | r0              */
+
+    PFACC     ( MM2, MM5 )		/* x0*m2+x1*m6     | x2*m10+x3*m14   */
+    MOVD      ( REGOFF(12, EAX), MM0 )	/*                 | x3              */
+
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+    PFACC     ( MM0, MM5 )		/* r3              | r2              */
+
+    MOVQ      ( MM4, REGOFF(-16, EDX) )	/* write r0, r1                      */
+    MOVQ      ( MM5, REGOFF(-8, EDX) )	/* write r2, r3                      */
+
+    MOVQ      ( REGIND(EAX), MM2 )	/* x1              | x0              */
+    MOVQ      ( REGOFF(8, EAX), MM3 )	/* x3              | x2              */
+
+    DEC_L     ( ESI )			/* decrement vertex counter          */
+    JNZ       ( LLBL( G3TP3R_1 ) )	/* cnt > 0 ? -> process next vertex  */
+
+LLBL( G3TP3R_2 ):
+
+    FEMMS
+    POP_L     ( EDI )
+    POP_L     ( ESI )
+    RET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_3dnow_transform_points4_3d_no_rot )
+GLNAME( _mesa_3dnow_transform_points4_3d_no_rot ):
+
+    PUSH_L    ( ESI )
+    MOV_L     ( ARG_DEST, ECX )
+    MOV_L     ( ARG_MATRIX, ESI )
+    MOV_L     ( ARG_SOURCE, EAX )
+    MOV_L     ( CONST(4), REGOFF(V4F_SIZE, ECX) )
+    OR_B      ( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, ECX) )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), EDX )
+    MOV_L     ( EDX, REGOFF(V4F_COUNT, ECX) )
+
+    PUSH_L    ( EDI )
+
+    MOV_L     ( REGOFF(V4F_START, ECX), EDX )
+    MOV_L     ( ESI, ECX )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), ESI )
+    MOV_L     ( REGOFF(V4F_STRIDE, EAX), EDI )
+    MOV_L     ( REGOFF(V4F_START, EAX), EAX )
+
+    TEST_L    ( ESI, ESI )
+    JZ        ( LLBL( G3TP3NRR_2 ) )
+
+    MOVD      ( REGIND(ECX), MM0 )	/*                 | m00             */
+    PUNPCKLDQ ( REGOFF(20, ECX), MM0 )	/* m11             | m00             */
+
+    MOVD      ( REGOFF(40, ECX), MM2 )	/*                 | m22             */
+    PUNPCKLDQ ( REGOFF(56, ECX), MM2 )	/* m32             | m22             */
+
+    MOVQ      ( REGOFF(48, ECX), MM1 )	/* m31             | m30             */
+
+    MOVQ      ( REGIND(EAX), MM4 )	/* x1              | x0              */
+    MOVQ      ( REGOFF(8, EAX), MM5 )	/* x3              | x2              */
+    MOVD      ( REGOFF(12, EAX), MM7 )	/*                 | x3              */
+
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+
+ALIGNTEXT16
+LLBL( G3TP3NRR_1 ):
+
+    PREFETCHW ( REGOFF(32, EDX) )	/* prefetch 2 vertices ahead         */
+    PREFETCH  ( REGOFF(32, EAX) )	/* hopefully stride is zero          */
+
+    MOVQ      ( MM5, MM6 )		/* x3              | x2              */
+    PFMUL     ( MM0, MM4 )		/* x1*m11          | x0*m00          */
+
+    PUNPCKHDQ ( MM6, MM6 )		/* x3              | x3              */
+    PFMUL     ( MM2, MM5 )		/* x3*m32          | x2*m22          */
+
+    PFMUL     ( MM1, MM6 )		/* x3*m31          | x3*m30          */
+    PFACC     ( MM7, MM5 )		/* x3              | x2*m22+x3*m32   */
+
+    PFADD     ( MM6, MM4 )		/* x1*m11+x3*m31   | x0*m00+x3*m30   */
+    ADD_L     ( CONST(16), EDX )	/* next r                            */
+
+    MOVQ      ( MM4, REGOFF(-16, EDX) )	/* write r0, r1                      */
+    MOVQ      ( MM5, REGOFF(-8, EDX) )	/* write r2, r3                      */
+
+    MOVQ      ( REGIND(EAX), MM4 )	/* x1              | x0              */
+    MOVQ      ( REGOFF(8, EAX), MM5 )	/* x3              | x2              */
+
+    MOVD      ( REGOFF(12, EAX), MM7 )	/*                 | x3              */
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+
+    DEC_L     ( ESI )			/* decrement vertex counter          */
+    JNZ       ( LLBL( G3TP3NRR_1 ) )	/* cnt > 0 ? -> process next vertex  */
+
+LLBL( G3TP3NRR_2 ):
+
+    FEMMS
+    POP_L     ( EDI )
+    POP_L     ( ESI )
+    RET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_3dnow_transform_points4_2d )
+GLNAME( _mesa_3dnow_transform_points4_2d ):
+
+    PUSH_L    ( ESI )
+
+    MOV_L     ( ARG_DEST, ECX )
+    MOV_L     ( ARG_MATRIX, ESI )
+    MOV_L     ( ARG_SOURCE, EAX )
+    MOV_L     ( CONST(4), REGOFF(V4F_SIZE, ECX) )
+    OR_B      ( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, ECX) )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), EDX )
+    MOV_L     ( EDX, REGOFF(V4F_COUNT, ECX) )
+
+    PUSH_L    ( EDI )
+
+    MOV_L     ( REGOFF(V4F_START, ECX), EDX )
+    MOV_L     ( ESI, ECX )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), ESI )
+    MOV_L     ( REGOFF(V4F_STRIDE, EAX), EDI )
+    MOV_L     ( REGOFF(V4F_START, EAX), EAX )
+
+    TEST_L    ( ESI, ESI )
+    JZ        ( LLBL( G3TP2R_2 ) )
+
+    MOVD      ( REGIND(ECX), MM0 )	/*                 | m00             */
+    PUNPCKLDQ ( REGOFF(16, ECX), MM0 )	/* m10             | m00             */
+
+    MOVD      ( REGOFF(4, ECX), MM1 )	/*                 | m01             */
+    PUNPCKLDQ ( REGOFF(20, ECX), MM1 )	/* m11             | m01             */
+
+    MOVQ      ( REGOFF(48, ECX), MM2 )	/* m31             | m30             */
+
+    MOVQ      ( REGIND(EAX), MM3 )	/* x1              | x0              */
+    MOVQ      ( REGOFF(8, EAX), MM5 )	/* x3              | x2              */
+
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+
+ALIGNTEXT16
+LLBL( G3TP2R_1 ):
+
+    PREFETCHW ( REGOFF(32, EDX) )       /* prefetch 2 vertices ahead         */
+    PREFETCH  ( REGIND(EAX) )
+
+    MOVQ      ( MM3, MM4 )		/* x1              | x0              */
+    MOVQ      ( MM5, MM6 )		/* x3              | x2              */
+
+    PFMUL     ( MM1, MM4 )		/* x1*m11          | x0*m01          */
+    PUNPCKHDQ ( MM6, MM6 )		/* x3              | x3              */
+
+    PFMUL     ( MM0, MM3 )		/* x1*m10          | x0*m00          */
+    ADD_L     ( CONST(16), EDX )	/* next r                            */
+
+    PFACC     ( MM4, MM3 )		/* x0*m01+x1*m11   | x0*m00+x1*m10   */
+    PFMUL     ( MM2, MM6 )		/* x3*m31          | x3*m30          */
+
+    PFADD     ( MM6, MM3 )		/* r1              | r0              */
+    MOVQ      ( MM5, REGOFF(-8, EDX) )	/* write r2, r3                      */
+
+    MOVQ      ( MM3, REGOFF(-16, EDX) )	/* write r0, r1                      */
+    MOVQ      ( REGIND(EAX), MM3 )	/* x1              | x0              */
+
+    MOVQ      ( REGOFF(8, EAX), MM5 )	/* x3              | x2              */
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+
+    DEC_L     ( ESI )			/* decrement vertex counter          */
+    JNZ       ( LLBL( G3TP2R_1 ) )	/* cnt > 0 ? -> process next vertex  */
+
+LLBL( G3TP2R_2 ):
+
+    FEMMS
+    POP_L     ( EDI )
+    POP_L     ( ESI )
+    RET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_3dnow_transform_points4_2d_no_rot )
+GLNAME( _mesa_3dnow_transform_points4_2d_no_rot ):
+
+    PUSH_L    ( ESI )
+
+    MOV_L     ( ARG_DEST, ECX )
+    MOV_L     ( ARG_MATRIX, ESI )
+    MOV_L     ( ARG_SOURCE, EAX )
+    MOV_L     ( CONST(4), REGOFF(V4F_SIZE, ECX) )
+    OR_B      ( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, ECX) )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), EDX )
+    MOV_L     ( EDX, REGOFF(V4F_COUNT, ECX) )
+
+    PUSH_L    ( EDI )
+
+    MOV_L     ( REGOFF(V4F_START, ECX), EDX )
+    MOV_L     ( ESI, ECX )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), ESI )
+    MOV_L     ( REGOFF(V4F_STRIDE, EAX), EDI )
+    MOV_L     ( REGOFF(V4F_START, EAX), EAX )
+
+    TEST_L    ( ESI, ESI )
+    JZ        ( LLBL( G3TP2NRR_3 ) )
+
+    MOVD      ( REGIND(ECX), MM0 )	/*                 | m00             */
+    PUNPCKLDQ ( REGOFF(20, ECX), MM0 )	/* m11             | m00             */
+
+    MOVQ      ( REGOFF(48, ECX), MM1 )	/* m31             | m30             */
+
+    MOVQ      ( REGIND(EAX), MM4 )	/* x1              | x0              */
+    MOVQ      ( REGOFF(8, EAX), MM5 )	/* x3              | x2              */
+
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+
+ALIGNTEXT16
+LLBL( G3TP2NRR_2 ):
+
+    PREFETCHW ( REGOFF(32, EDX) )	/* prefetch 2 vertices ahead         */
+    PREFETCH  ( REGIND(EAX) )
+
+    PFMUL     ( MM0, MM4 )		/* x1*m11          | x0*m00          */
+    MOVQ      ( MM5, MM6 )		/* x3              | x2              */
+
+    ADD_L     ( CONST(16), EDX )	/* next r                            */
+    PUNPCKHDQ ( MM6, MM6 )		/* x3              | x3              */
+
+    PFMUL     ( MM1, MM6 )		/* x3*m31          | x3*m30          */
+    PFADD     ( MM4, MM6 )		/* x1*m11+x3*m31   | x0*m00+x3*m30   */
+
+    MOVQ      ( MM6, REGOFF(-16, EDX) )	/* write r0, r1                      */
+    MOVQ      ( MM5, REGOFF(-8, EDX) )	/* write r2, r3                      */
+
+    MOVQ      ( REGIND(EAX), MM4 )	/* x1              | x0              */
+    MOVQ      ( REGOFF(8, EAX), MM5 )	/* x3              | x2              */
+
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+    DEC_L     ( ESI )			/* decrement vertex counter          */
+
+    JNZ       ( LLBL( G3TP2NRR_2 ) )	/* cnt > 0 ? -> process next vertex  */
+
+LLBL( G3TP2NRR_3 ):
+
+    FEMMS
+    POP_L     ( EDI )
+    POP_L     ( ESI )
+    RET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_3dnow_transform_points4_identity )
+GLNAME( _mesa_3dnow_transform_points4_identity ):
+
+    PUSH_L    ( ESI )
+
+    MOV_L     ( ARG_DEST, ECX )
+    MOV_L     ( ARG_MATRIX, ESI )
+    MOV_L     ( ARG_SOURCE, EAX )
+    MOV_L     ( CONST(4), REGOFF(V4F_SIZE, ECX) )
+    OR_B      ( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, ECX) )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), EDX )
+    MOV_L     ( EDX, REGOFF(V4F_COUNT, ECX) )
+
+    PUSH_L    ( EDI )
+
+    MOV_L     ( REGOFF(V4F_START, ECX), EDX )
+    MOV_L     ( ESI, ECX )
+    MOV_L     ( REGOFF(V4F_COUNT, EAX), ESI )
+    MOV_L     ( REGOFF(V4F_STRIDE, EAX), EDI )
+    MOV_L     ( REGOFF(V4F_START, EAX), EAX )
+
+    TEST_L    ( ESI, ESI )
+    JZ        ( LLBL( G3TPIR_2 ) )
+
+    MOVQ      ( REGIND(EAX), MM0 )	/* x1              | x0              */
+    MOVQ      ( REGOFF(8, EAX), MM1 )	/* x3              | x2              */
+
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+
+ALIGNTEXT16
+LLBL( G3TPIR_1 ):
+
+    PREFETCHW ( REGOFF(32, EDX) )       /* prefetch 2 vertices ahead         */
+    PREFETCH  ( REGIND(EAX) )
+
+    ADD_L     ( CONST(16), EDX )	/* next r                            */
+    MOVQ      ( MM0, REGOFF(-16, EDX) )	/* r1              | r0              */
+
+    MOVQ      ( MM1, REGOFF(-8, EDX) )	/* r3              | r2              */
+    MOVQ      ( REGIND(EAX), MM0 )	/* x1              | x0              */
+
+    MOVQ      ( REGOFF(8, EAX), MM1 )	/* x3              | x2              */
+    ADD_L     ( EDI, EAX )		/* next vertex                       */
+
+    DEC_L     ( ESI )			/* decrement vertex counter          */
+    JNZ       ( LLBL( G3TPIR_1 ) )	/* cnt > 0 ? -> process next vertex  */
+
+LLBL( G3TPIR_2 ):
+
+    FEMMS
+    POP_L     ( EDI )
+    POP_L     ( ESI )
+    RET
diff --git a/src/mesa/x86/common_x86.c b/src/mesa/x86/common_x86.c
index 7d893917185..15028197fb8 100644
--- a/src/mesa/x86/common_x86.c
+++ b/src/mesa/x86/common_x86.c
@@ -1,4 +1,4 @@
-/* $Id: common_x86.c,v 1.14 2001/03/28 20:44:43 gareth Exp $ */
+/* $Id: common_x86.c,v 1.15 2001/03/29 06:46:16 gareth Exp $ */
 
 /*
  * Mesa 3-D graphics library
@@ -35,7 +35,7 @@
 
 #include <stdlib.h>
 #include <stdio.h>
-#if defined(USE_KATMAI_ASM) && defined(__linux__)
+#if defined(USE_SSE_ASM) && defined(__linux__)
 #include <signal.h>
 #endif
 
@@ -67,7 +67,7 @@ static void message( const char *msg )
    }
 }
 
-#if defined(USE_KATMAI_ASM)
+#if defined(USE_SSE_ASM)
 /*
  * We must verify that the Streaming SIMD Extensions are truly supported
  * on this processor before we go ahead and hook out the optimized code.
@@ -84,8 +84,8 @@ static void message( const char *msg )
  * not good.
  */
 
-extern void _mesa_test_os_katmai_support( void );
-extern void _mesa_test_os_katmai_exception_support( void );
+extern void _mesa_test_os_sse_support( void );
+extern void _mesa_test_os_sse_exception_support( void );
 
 #if defined(__linux__) && defined(_POSIX_SOURCE)
 static void sigill_handler( int signal, struct sigcontext sc )
@@ -135,7 +135,7 @@ static void sigfpe_handler( int signal, struct sigcontext sc )
  *
  * GH: Isn't this just awful?
  */
-static void check_os_katmai_support( void )
+static void check_os_sse_support( void )
 {
 #if defined(__linux__)
 #if defined(_POSIX_SOURCE)
@@ -159,7 +159,7 @@ static void check_os_katmai_support( void )
    if ( cpu_has_xmm ) {
       message( "Testing OS support for SSE... " );
 
-      _mesa_test_os_katmai_support();
+      _mesa_test_os_sse_support();
 
       if ( cpu_has_xmm ) {
 	 message( "yes.\n" );
@@ -184,7 +184,7 @@ static void check_os_katmai_support( void )
    if ( cpu_has_xmm ) {
       message( "Testing OS support for SSE unmasked exceptions... " );
 
-      _mesa_test_os_katmai_exception_support();
+      _mesa_test_os_sse_exception_support();
 
       if ( cpu_has_xmm ) {
 	 message( "yes.\n" );
@@ -220,7 +220,7 @@ static void check_os_katmai_support( void )
 #endif /* __linux__ */
 }
 
-#endif /* USE_KATMAI_ASM */
+#endif /* USE_SSE_ASM */
 
 
 void _mesa_init_all_x86_transform_asm( void )
@@ -257,14 +257,14 @@ void _mesa_init_all_x86_transform_asm( void )
    }
 #endif
 
-#ifdef USE_KATMAI_ASM
-   if ( cpu_has_xmm && getenv( "MESA_FORCE_KATMAI" ) == 0 ) {
-      check_os_katmai_support();
+#ifdef USE_SSE_ASM
+   if ( cpu_has_xmm && getenv( "MESA_FORCE_SSE" ) == 0 ) {
+      check_os_sse_support();
    }
    if ( cpu_has_xmm ) {
-      if ( getenv( "MESA_NO_KATMAI" ) == 0 ) {
-         message( "Katmai cpu detected.\n" );
-         _mesa_init_katmai_transform_asm();
+      if ( getenv( "MESA_NO_SSE" ) == 0 ) {
+         message( "SSE cpu detected.\n" );
+         _mesa_init_sse_transform_asm();
       } else {
          _mesa_x86_cpu_features &= ~(X86_FEATURE_XMM);
       }
@@ -289,9 +289,9 @@ void _mesa_init_all_x86_vertex_asm( void )
    }
 #endif
 
-#ifdef USE_KATMAI_ASM
-   if ( cpu_has_xmm && getenv( "MESA_NO_KATMAI" ) == 0 ) {
-      _mesa_init_katmai_vertex_asm();
+#ifdef USE_SSE_ASM
+   if ( cpu_has_xmm && getenv( "MESA_NO_SSE" ) == 0 ) {
+      _mesa_init_sse_vertex_asm();
    }
 #endif
 #endif
diff --git a/src/mesa/x86/common_x86_asm.S b/src/mesa/x86/common_x86_asm.S
index 69cc83f6cd2..8a0a6477afa 100644
--- a/src/mesa/x86/common_x86_asm.S
+++ b/src/mesa/x86/common_x86_asm.S
@@ -1,4 +1,4 @@
-/* $Id: common_x86_asm.S,v 1.6 2001/03/28 20:44:43 gareth Exp $ */
+/* $Id: common_x86_asm.S,v 1.7 2001/03/29 06:46:16 gareth Exp $ */
 
 /*
  * Mesa 3-D graphics library
@@ -58,8 +58,8 @@
 GLNAME( found_intel ):	STRING( "Genuine Intel processor found\n\0" )
 GLNAME( found_amd ):	STRING( "Authentic AMD processor found\n\0" )
 
-#ifdef USE_KATMAI_ASM
-GLNAME( katmai_test_dummy ):
+#ifdef USE_SSE_ASM
+GLNAME( sse_test_dummy ):
 	D_LONG 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000
 #endif
 
@@ -157,15 +157,15 @@ LLBL ( cpuid_done ):
 	RET
 
 
-#ifdef USE_KATMAI_ASM
+#ifdef USE_SSE_ASM
 /* Execute an SSE instruction to see if the operating system correctly
  * supports SSE.  A signal handler for SIGILL should have been set
  * before calling this function, otherwise this could kill the client
  * application.
  */
 ALIGNTEXT4
-GLOBL GLNAME( _mesa_test_os_katmai_support )
-GLNAME( _mesa_test_os_katmai_support ):
+GLOBL GLNAME( _mesa_test_os_sse_support )
+GLNAME( _mesa_test_os_sse_support ):
 
 	XORPS	( XMM0, XMM0 )
 
@@ -178,8 +178,8 @@ GLNAME( _mesa_test_os_katmai_support ):
  * otherwise this could kill the client application.
  */
 ALIGNTEXT4
-GLOBL GLNAME( _mesa_test_os_katmai_exception_support )
-GLNAME( _mesa_test_os_katmai_exception_support ):
+GLOBL GLNAME( _mesa_test_os_sse_exception_support )
+GLNAME( _mesa_test_os_sse_exception_support ):
 
 	PUSH_L	( EBP )
 	MOV_L	( ESP, EBP )
@@ -196,7 +196,7 @@ GLNAME( _mesa_test_os_katmai_exception_support ):
 	LDMXCSR	( REGOFF( -8, EBP ) )
 
 	XORPS	( XMM0, XMM0 )
-	MOVUPS	( GLNAME( katmai_test_dummy ), XMM1 )
+	MOVUPS	( GLNAME( sse_test_dummy ), XMM1 )
 
 	DIVPS	( XMM0, XMM1 )
 
diff --git a/src/mesa/x86/common_x86_asm.h b/src/mesa/x86/common_x86_asm.h
index 312855dc652..ba965e6c088 100644
--- a/src/mesa/x86/common_x86_asm.h
+++ b/src/mesa/x86/common_x86_asm.h
@@ -1,4 +1,4 @@
-/* $Id: common_x86_asm.h,v 1.7 2001/03/28 20:44:44 gareth Exp $ */
+/* $Id: common_x86_asm.h,v 1.8 2001/03/29 06:46:16 gareth Exp $ */
 
 /*
  * Mesa 3-D graphics library
@@ -52,8 +52,8 @@
 #ifdef USE_3DNOW_ASM
 #include "3dnow.h"
 #endif
-#ifdef USE_KATMAI_ASM
-#include "katmai.h"
+#ifdef USE_SSE_ASM
+#include "sse.h"
 #endif
 #endif
 
diff --git a/src/mesa/x86/sse.c b/src/mesa/x86/sse.c
new file mode 100644
index 00000000000..a986636edc7
--- /dev/null
+++ b/src/mesa/x86/sse.c
@@ -0,0 +1,208 @@
+/* $Id: sse.c,v 1.1 2001/03/29 06:46:16 gareth Exp $ */
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  3.5
+ *
+ * Copyright (C) 1999-2001  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.
+ */
+
+/*
+ * PentiumIII-SIMD (SSE) optimizations contributed by
+ * Andre Werthmann <wertmann@cs.uni-potsdam.de>
+ */
+
+#include "glheader.h"
+#include "context.h"
+#include "mtypes.h"
+#include "sse.h"
+
+#include "math/m_vertices.h"
+#include "math/m_xform.h"
+
+#include "tnl/t_context.h"
+
+#ifdef DEBUG
+#include "math/m_debug.h"
+#endif
+
+
+#define XFORM_ARGS	GLvector4f *to_vec,				\
+			const GLfloat m[16],				\
+			const GLvector4f *from_vec,			\
+			const GLubyte *mask,				\
+			const GLubyte flag
+
+
+#define DECLARE_XFORM_GROUP( pfx, sz ) \
+extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_general( XFORM_ARGS );		\
+extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_identity( XFORM_ARGS );	\
+extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_3d_no_rot( XFORM_ARGS );	\
+extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_perspective( XFORM_ARGS );	\
+extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_2d( XFORM_ARGS );		\
+extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_2d_no_rot( XFORM_ARGS );	\
+extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_3d( XFORM_ARGS );
+
+
+#define ASSIGN_XFORM_GROUP( pfx, sz )					\
+   _mesa_transform_tab[0][sz][MATRIX_GENERAL] =				\
+      _mesa_##pfx##_transform_points##sz##_general;			\
+   _mesa_transform_tab[0][sz][MATRIX_IDENTITY] =			\
+      _mesa_##pfx##_transform_points##sz##_identity;			\
+   _mesa_transform_tab[0][sz][MATRIX_3D_NO_ROT] =			\
+      _mesa_##pfx##_transform_points##sz##_3d_no_rot;			\
+   _mesa_transform_tab[0][sz][MATRIX_PERSPECTIVE] =			\
+      _mesa_##pfx##_transform_points##sz##_perspective;			\
+   _mesa_transform_tab[0][sz][MATRIX_2D] =				\
+      _mesa_##pfx##_transform_points##sz##_2d;				\
+   _mesa_transform_tab[0][sz][MATRIX_2D_NO_ROT] =			\
+      _mesa_##pfx##_transform_points##sz##_2d_no_rot;			\
+   _mesa_transform_tab[0][sz][MATRIX_3D] =				\
+      _mesa_##pfx##_transform_points##sz##_3d;
+
+
+
+#define NORM_ARGS	const GLmatrix *mat,				\
+			GLfloat scale,					\
+			const GLvector3f *in,				\
+			const GLfloat *lengths,				\
+			const GLubyte mask[],				\
+			GLvector3f *dest
+
+
+#define DECLARE_NORM_GROUP( pfx ) \
+extern void _ASMAPI _mesa_##pfx##_rescale_normals( NORM_ARGS );				\
+extern void _ASMAPI _mesa_##pfx##_normalize_normals( NORM_ARGS );			\
+extern void _ASMAPI _mesa_##pfx##_transform_normals( NORM_ARGS );			\
+extern void _ASMAPI _mesa_##pfx##_transform_normals_no_rot( NORM_ARGS );		\
+extern void _ASMAPI _mesa_##pfx##_transform_rescale_normals( NORM_ARGS );		\
+extern void _ASMAPI _mesa_##pfx##_transform_rescale_normals_no_rot( NORM_ARGS );	\
+extern void _ASMAPI _mesa_##pfx##_transform_normalize_normals( NORM_ARGS );		\
+extern void _ASMAPI _mesa_##pfx##_transform_normalize_normals_no_rot( NORM_ARGS );
+
+
+#define ASSIGN_NORM_GROUP( pfx )					\
+   _mesa_normal_tab[NORM_RESCALE][0] =					\
+      _mesa_##pfx##_rescale_normals;					\
+   _mesa_normal_tab[NORM_NORMALIZE][0] =				\
+      _mesa_##pfx##_normalize_normals;					\
+   _mesa_normal_tab[NORM_TRANSFORM][0] =				\
+      _mesa_##pfx##_transform_normals;					\
+   _mesa_normal_tab[NORM_TRANSFORM_NO_ROT][0] =				\
+      _mesa_##pfx##_transform_normals_no_rot;				\
+   _mesa_normal_tab[NORM_TRANSFORM|NORM_RESCALE][0] =			\
+      _mesa_##pfx##_transform_rescale_normals;				\
+   _mesa_normal_tab[NORM_TRANSFORM_NO_ROT|NORM_RESCALE][0] =		\
+      _mesa_##pfx##_transform_rescale_normals_no_rot;			\
+   _mesa_normal_tab[NORM_TRANSFORM|NORM_NORMALIZE][0] =			\
+      _mesa_##pfx##_transform_normalize_normals;			\
+   _mesa_normal_tab[NORM_TRANSFORM_NO_ROT|NORM_NORMALIZE][0] =		\
+      _mesa_##pfx##_transform_normalize_normals_no_rot;
+
+
+#ifdef USE_SSE_ASM
+DECLARE_XFORM_GROUP( sse, 2 )
+DECLARE_XFORM_GROUP( sse, 3 )
+
+#if 1
+/* Some functions are not written in SSE-assembly, because the fpu ones are faster */
+extern void _mesa_sse_transform_normals_no_rot( NORM_ARGS );
+extern void _mesa_sse_transform_rescale_normals( NORM_ARGS );
+extern void _mesa_sse_transform_rescale_normals_no_rot( NORM_ARGS );
+
+extern void _mesa_sse_transform_points4_general( XFORM_ARGS );
+extern void _mesa_sse_transform_points4_3d( XFORM_ARGS );
+extern void _mesa_sse_transform_points4_identity( XFORM_ARGS );
+#else
+DECLARE_NORM_GROUP( sse )
+#endif
+
+
+extern void _ASMAPI
+_mesa_v16_sse_general_xform( GLfloat *first_vert,
+			     const GLfloat *m,
+			     const GLfloat *src,
+			     GLuint src_stride,
+			     GLuint count );
+
+extern void _ASMAPI
+_mesa_sse_project_vertices( GLfloat *first,
+			    GLfloat *last,
+			    const GLfloat *m,
+			    GLuint stride );
+
+extern void _ASMAPI
+_mesa_sse_project_clipped_vertices( GLfloat *first,
+				    GLfloat *last,
+				    const GLfloat *m,
+				    GLuint stride,
+				    const GLubyte *clipmask );
+#endif
+
+
+void _mesa_init_sse_transform_asm( void )
+{
+#ifdef USE_SSE_ASM
+   ASSIGN_XFORM_GROUP( sse, 2 );
+   ASSIGN_XFORM_GROUP( sse, 3 );
+
+#if 1
+   /* TODO: Finish these off.
+    */
+   _mesa_normal_tab[NORM_TRANSFORM_NO_ROT][0] =
+      _mesa_sse_transform_normals_no_rot;
+   _mesa_normal_tab[NORM_TRANSFORM|NORM_RESCALE][0] =
+      _mesa_sse_transform_rescale_normals;
+   _mesa_normal_tab[NORM_TRANSFORM_NO_ROT|NORM_RESCALE][0] =
+      _mesa_sse_transform_rescale_normals_no_rot;
+
+   _mesa_transform_tab[0][4][MATRIX_GENERAL] =
+      _mesa_sse_transform_points4_general;
+   _mesa_transform_tab[0][4][MATRIX_3D] =
+      _mesa_sse_transform_points4_3d;
+   _mesa_transform_tab[0][4][MATRIX_IDENTITY] =
+      _mesa_sse_transform_points4_identity;
+#else
+   ASSIGN_NORM_GROUP( sse );
+#endif
+
+#ifdef DEBUG
+   _math_test_all_transform_functions( "SSE" );
+   _math_test_all_normal_transform_functions( "SSE" );
+#endif
+#endif
+}
+
+void _mesa_init_sse_vertex_asm( void )
+{
+#ifdef USE_SSE_ASM
+   _mesa_xform_points3_v16_general = _mesa_v16_sse_general_xform;
+#if 0
+   /* GH: These are broken.  I'm fixing them now.
+    */
+   _mesa_project_v16 = _mesa_sse_project_vertices;
+   _mesa_project_clipped_v16 = _mesa_sse_project_clipped_vertices;
+#endif
+
+#ifdef DEBUG_NOT
+   _math_test_all_vertex_functions( "SSE" );
+#endif
+#endif
+}
diff --git a/src/mesa/x86/sse.h b/src/mesa/x86/sse.h
new file mode 100644
index 00000000000..3ed1a95892e
--- /dev/null
+++ b/src/mesa/x86/sse.h
@@ -0,0 +1,40 @@
+/* $Id: sse.h,v 1.1 2001/03/29 06:46:16 gareth Exp $ */
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  3.5
+ *
+ * Copyright (C) 1999-2001  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.
+ */
+
+/*
+ * PentiumIII-SIMD (SSE) optimizations contributed by
+ * Andre Werthmann <wertmann@cs.uni-potsdam.de>
+ */
+
+#ifndef __SSE_H__
+#define __SSE_H__
+
+#include "math/m_xform.h"
+
+void _mesa_init_sse_transform_asm( void );
+void _mesa_init_sse_vertex_asm( void );
+
+#endif
diff --git a/src/mesa/x86/sse_normal.S b/src/mesa/x86/sse_normal.S
new file mode 100644
index 00000000000..8dd84967154
--- /dev/null
+++ b/src/mesa/x86/sse_normal.S
@@ -0,0 +1,252 @@
+/* $Id: sse_normal.S,v 1.1 2001/03/29 06:46:16 gareth Exp $ */
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  3.5
+ *
+ * Copyright (C) 1999-2001  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.
+ */
+
+/** TODO:
+  * - insert PREFETCH instructions to avoid cache-misses !
+  * - some more optimizations are possible...
+  * - for 40-50% more performance in the SSE-functions, the
+  *   data (trans-matrix, src_vert, dst_vert) needs to be 16byte aligned !
+  */
+
+#include "matypes.h"
+#include "norm_args.h"
+
+   SEG_TEXT
+
+#define M(i)    REGOFF(i * 4, EDX)
+#define S(i)	REGOFF(i * 4, ESI)
+#define D(i)	REGOFF(i * 4, EDI)
+#define STRIDE  REGOFF(12, ESI)
+
+
+ALIGNTEXT16
+GLOBL GLNAME(_mesa_sse_transform_rescale_normals_no_rot)
+GLNAME(_mesa_sse_transform_rescale_normals_no_rot):
+
+#define FRAME_OFFSET 8
+	PUSH_L  ( ESI )
+	PUSH_L  ( EDI )
+
+	MOV_L	( ARG_IN, ESI )				/* ptr to source GLvector3f */
+	MOV_L	( ARG_DEST, EDI )			/* ptr to dest GLvector3f */
+
+	MOV_L	( ARG_MAT, EDX )			/* ptr to matrix */
+	MOV_L	( REGOFF(MATRIX_INV, EDX), EDX)		/* matrix->inv */
+
+	MOV_L	( REGOFF(V3F_COUNT, ESI), ECX ) 	/* source count */
+
+	TEST_L	( ECX, ECX )
+	JZ( LLBL(K_G3TRNNRR_finish) )			/* count was zero; go to finish */
+
+	MOV_L	( STRIDE, EAX )				/* stride */
+	MOV_L	( ECX, REGOFF(V3F_COUNT, EDI) )		/* set dest-count */
+
+	IMUL_L( CONST(12), ECX )			/* count *= 12 */
+	MOV_L( REGOFF(V3F_START, ESI), ESI )		/* ptr to first source vertex */
+
+	MOV_L( REGOFF(V3F_START, EDI), EDI )		/* ptr to first dest vertex */
+	ADD_L( EDI, ECX ) 				/* count += dest ptr */
+
+ALIGNTEXT32
+	MOVSS	( M(0), XMM1 )				/* m0 */
+	MOVSS	( M(5), XMM2 )				/* m5 */
+	UNPCKLPS( XMM2, XMM1 )				/* m5 | m0 */
+	MOVSS	( ARG_SCALE, XMM0 )			/* scale */
+	SHUFPS	( CONST(0x0), XMM0, XMM0 )		/* scale | scale */
+	MULPS	( XMM0, XMM1 )				/* m5*scale | m0*scale */
+	MULSS	( M(10), XMM0 )				/* m10*scale */
+
+ALIGNTEXT32
+LLBL(K_G3TRNNRR_top):
+	MOVLPS	( S(0), XMM2 )				/* uy | ux */
+	MULPS	( XMM1, XMM2 )				/* uy*m5*scale | ux*m0*scale */
+	MOVLPS	( XMM2, D(0) )				/* ->D(1) | D(0) */
+
+	MOVSS	( S(2), XMM2 )				/* uz */
+	MULSS	( XMM0, XMM2 )				/* uz*m10*scale */
+	MOVSS	( XMM2, D(2) )				/* ->D(2) */
+
+LLBL(K_G3TRNNRR_skip):
+	ADD_L	( CONST(12), EDI )
+	ADD_L	( EAX, ESI )
+	CMP_L	( ECX, EDI )
+	JNE	( LLBL(K_G3TRNNRR_top) )
+
+LLBL(K_G3TRNNRR_finish):
+	POP_L	( EDI )
+	POP_L	( ESI )
+	RET
+#undef FRAME_OFFSET
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME(_mesa_sse_transform_rescale_normals)
+GLNAME(_mesa_sse_transform_rescale_normals):
+
+#define FRAME_OFFSET 8
+	PUSH_L  ( ESI )
+	PUSH_L  ( EDI )
+
+	MOV_L	( ARG_IN, ESI )				/* ptr to source GLvector3f */
+	MOV_L	( ARG_DEST, EDI )			/* ptr to dest GLvector3f */
+
+	MOV_L	( ARG_MAT, EDX )			/* ptr to matrix */
+	MOV_L	( REGOFF(MATRIX_INV, EDX), EDX)		/* matrix->inv */
+
+	MOV_L	( REGOFF(V3F_COUNT, ESI), ECX ) 	/* source count */
+
+	TEST_L	( ECX, ECX )
+	JZ( LLBL(K_G3TRNR_finish) )			/* count was zero; go to finish */
+
+	MOV_L	( STRIDE, EAX )				/* stride */
+	MOV_L	( ECX, REGOFF(V3F_COUNT, EDI) )		/* set dest-count */
+
+	IMUL_L( CONST(12), ECX )			/* count *= 12 */
+	MOV_L( REGOFF(V3F_START, ESI), ESI )		/* ptr to first source vertex */
+
+	MOV_L( REGOFF(V3F_START, EDI), EDI )		/* ptr to first dest vertex */
+	ADD_L( EDI, ECX ) 				/* count += dest ptr */
+
+ALIGNTEXT32
+	MOVSS	( M(0), XMM0 )				/* m0 */
+	MOVSS	( M(4), XMM1 )				/* m4 */
+	UNPCKLPS( XMM1, XMM0 )				/* m4 | m0 */
+
+	MOVSS	( ARG_SCALE, XMM4 )			/* scale */
+	SHUFPS	( CONST(0x0), XMM4, XMM4 )		/* scale | scale */
+
+	MULPS	( XMM4, XMM0 )				/* m4*scale | m0*scale */
+	MOVSS	( M(1), XMM1 )				/* m1 */
+	MOVSS	( M(5), XMM2 )				/* m5 */
+	UNPCKLPS( XMM2, XMM1 )				/* m5 | m1 */
+	MULPS	( XMM4, XMM1 )				/* m5*scale | m1*scale */
+	MOVSS	( M(2), XMM2 )				/* m2 */
+	MOVSS	( M(6), XMM3 )				/* m6 */
+	UNPCKLPS( XMM3, XMM2 )				/* m6 | m2 */
+	MULPS	( XMM4, XMM2 )				/* m6*scale | m2*scale */
+
+	MOVSS	( M(8), XMM6 )				/* m8 */
+	MULSS	( ARG_SCALE, XMM6 )			/* m8*scale */
+	MOVSS	( M(9), XMM7 )				/* m9 */
+	MULSS	( ARG_SCALE, XMM7 )			/* m9*scale */
+
+ALIGNTEXT32
+LLBL(K_G3TRNR_top):
+	MOVSS	( S(0), XMM3 )				/* ux */
+	SHUFPS	( CONST(0x0), XMM3, XMM3 )		/* ux | ux */
+	MULPS	( XMM0, XMM3 )				/* ux*m4 | ux*m0 */
+	MOVSS	( S(1), XMM4 )				/* uy */
+	SHUFPS	( CONST(0x0), XMM4, XMM4 )		/* uy | uy */
+	MULPS	( XMM1, XMM4 )				/* uy*m5 | uy*m1 */
+	MOVSS	( S(2), XMM5 )				/* uz */
+	SHUFPS	( CONST(0x0), XMM5, XMM5 )		/* uz | uz */
+	MULPS	( XMM2, XMM5 )				/* uz*m6 | uz*m2 */
+
+	ADDPS	( XMM4, XMM3 )
+	ADDPS	( XMM5, XMM3 )
+	MOVLPS	( XMM3, D(0) )
+
+	MOVSS	( M(10), XMM3 )				/* m10 */
+	MULSS	( ARG_SCALE, XMM3 )			/* m10*scale */
+	MULSS	( S(2), XMM3 )				/* m10*scale*uz */
+	MOVSS	( S(1), XMM4 )				/* uy */
+	MULSS	( XMM7, XMM4 )				/* uy*m9*scale */
+	MOVSS	( S(0), XMM5 )				/* ux */
+	MULSS	( XMM6, XMM5 )				/* ux*m8*scale */
+
+	ADDSS	( XMM4, XMM3 )
+	ADDSS	( XMM5, XMM3 )
+	MOVSS	( XMM3, D(2) )
+
+LLBL(K_G3TRNR_skip):
+	ADD_L	( CONST(12), EDI )
+	ADD_L	( EAX, ESI )
+	CMP_L	( ECX, EDI )
+	JNE	( LLBL(K_G3TRNR_top) )
+
+LLBL(K_G3TRNR_finish):
+	POP_L	( EDI )
+	POP_L	( ESI )
+	RET
+#undef FRAME_OFFSET
+
+
+ALIGNTEXT16
+GLOBL GLNAME(_mesa_sse_transform_normals_no_rot)
+GLNAME(_mesa_sse_transform_normals_no_rot):
+
+#define FRAME_OFFSET 8
+	PUSH_L  ( ESI )
+	PUSH_L  ( EDI )
+
+	MOV_L	( ARG_IN, ESI )				/* ptr to source GLvector3f */
+	MOV_L	( ARG_DEST, EDI )			/* ptr to dest GLvector3f */
+
+	MOV_L	( ARG_MAT, EDX )			/* ptr to matrix */
+	MOV_L	( REGOFF(MATRIX_INV, EDX), EDX)		/* matrix->inv */
+
+	MOV_L	( REGOFF(V3F_COUNT, ESI), ECX ) 	/* source count */
+
+	TEST_L	( ECX, ECX )
+	JZ( LLBL(K_G3TNNRR_finish) )			/* count was zero; go to finish */
+
+	MOV_L	( STRIDE, EAX )				/* stride */
+	MOV_L	( ECX, REGOFF(V3F_COUNT, EDI) )		/* set dest-count */
+
+	IMUL_L( CONST(12), ECX )			/* count *= 12 */
+	MOV_L( REGOFF(V3F_START, ESI), ESI )		/* ptr to first source vertex */
+
+	MOV_L( REGOFF(V3F_START, EDI), EDI )		/* ptr to first dest vertex */
+	ADD_L( EDI, ECX ) 				/* count += dest ptr */
+
+ALIGNTEXT32
+	MOVSS( M(0), XMM0 )				/* m0 */
+	MOVSS( M(5), XMM1 )				/* m5 */
+	UNPCKLPS( XMM1, XMM0 )				/* m5 | m0 */
+	MOVSS( M(10), XMM1 )				/* m10 */
+
+ALIGNTEXT32
+LLBL(K_G3TNNRR_top):
+	MOVLPS( S(0), XMM2 )				/* uy | ux */
+	MULPS( XMM0, XMM2 )				/* uy*m5 | ux*m0 */
+	MOVLPS( XMM2, D(0) )
+
+	MOVSS( S(2), XMM2 )				/* uz */
+	MULSS( XMM1, XMM2 )				/* uz*m10 */
+	MOVSS( XMM2, D(2) )
+
+LLBL(K_G3TNNRR_skip):
+	ADD_L	( CONST(12), EDI )
+	ADD_L	( EAX, ESI )
+	CMP_L	( ECX, EDI )
+	JNE	( LLBL(K_G3TNNRR_top) )
+
+LLBL(K_G3TNNRR_finish):
+	POP_L	( EDI )
+	POP_L	( ESI )
+	RET
+#undef FRAME_OFFSET
diff --git a/src/mesa/x86/sse_xform1.S b/src/mesa/x86/sse_xform1.S
new file mode 100644
index 00000000000..89c55673311
--- /dev/null
+++ b/src/mesa/x86/sse_xform1.S
@@ -0,0 +1,433 @@
+/* $Id: sse_xform1.S,v 1.1 2001/03/29 06:46:27 gareth Exp $ */
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  3.5
+ *
+ * Copyright (C) 1999-2001  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.
+ */
+
+/** TODO:
+  * - insert PREFETCH instructions to avoid cache-misses !
+  * - some more optimizations are possible...
+  * - for 40-50% more performance in the SSE-functions, the
+  *   data (trans-matrix, src_vert, dst_vert) needs to be 16byte aligned !
+  */
+
+#include "matypes.h"
+#include "xform_args.h"
+
+   SEG_TEXT
+
+#define S(i) 	REGOFF(i * 4, ESI)
+#define D(i) 	REGOFF(i * 4, EDI)
+#define M(i) 	REGOFF(i * 4, EDX)
+
+
+ALIGNTEXT4
+GLOBL GLNAME(_mesa_sse_transform_points1_general)
+GLNAME( _mesa_sse_transform_points1_general ):
+
+#define FRAME_OFFSET 8
+    PUSH_L    ( ESI )
+    PUSH_L    ( EDI )
+
+    MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) 	/* ptr to source GLvector4f */
+    MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) 	/* ptr to dest GLvector4f */
+
+    MOV_L( ARG_MATRIX, EDX ) 			/* ptr to matrix */
+    MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) 	/* source count */
+
+    CMP_L( CONST(0), ECX )			/* count == 0 ? */
+    JE( LLBL(K_GTP1GR_finish) )			/* yes -> nothing to do. */
+
+    MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) 	/* stride */
+    OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) ) 	/* set dest flags */
+
+    MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) 	/* set dest count */
+    MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) ) 	/* set dest size */
+
+    SHL_L( CONST(4), ECX ) 			/* count *= 16 */
+    MOV_L( REGOFF(V4F_START, ESI), ESI ) 	/* ptr to first source vertex */
+
+    MOV_L( REGOFF(V4F_START, EDI), EDI ) 	/* ptr to first dest vertex */
+    ADD_L( EDI, ECX ) 				/* count += dest ptr */
+
+
+ALIGNTEXT32
+    MOVAPS( M(0), XMM0 )			/* m3  | m2  | m1  | m0  */
+    MOVAPS( M(12), XMM1 )			/* m15 | m14 | m13 | m12 */
+
+ALIGNTEXT32
+LLBL(K_GTP1GR_top):
+    MOVSS( S(0), XMM2 )				/* ox */
+    SHUFPS( CONST(0x0), XMM2, XMM2 )		/* ox | ox | ox | ox */
+    MULPS( XMM0, XMM2 )				/* ox*m3 | ox*m2 | ox*m1 | ox*m0 */
+    ADDPS( XMM1, XMM2 )				/* + | + | + | + */
+    MOVUPS( XMM2, D(0) )
+
+LLBL(K_GTP1GR_skip):
+    ADD_L     ( CONST(16), EDI )
+    ADD_L     ( EAX, ESI )
+    CMP_L     ( ECX, EDI )
+    JNE       ( LLBL(K_GTP1GR_top) )
+
+LLBL(K_GTP1GR_finish):
+    POP_L     ( EDI )
+    POP_L     ( ESI )
+    RET
+#undef FRAME_OFFSET
+
+
+
+ALIGNTEXT4
+GLOBL GLNAME(_mesa_sse_transform_points1_identity)
+GLNAME( _mesa_sse_transform_points1_identity ):
+
+#define FRAME_OFFSET 8
+    PUSH_L    ( ESI )
+    PUSH_L    ( EDI )
+
+    MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) 	/* ptr to source GLvector4f */
+    MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) 	/* ptr to dest GLvector4f */
+
+    MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) 	/* source count */
+
+    TEST_L( ECX, ECX)
+    JZ( LLBL(K_GTP1IR_finish) ) 		/* count was zero; go to finish */
+
+    MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) 	/* stride */
+    OR_L( CONST(VEC_SIZE_1), REGOFF(V4F_FLAGS, EDI) ) 	/* set dest flags */
+
+    MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) 	/* set dest count */
+    MOV_L( CONST(1), REGOFF(V4F_SIZE, EDI) ) 	/* set dest size */
+
+    SHL_L( CONST(4), ECX ) 			/* count *= 16 */
+    MOV_L( REGOFF(V4F_START, ESI), ESI ) 	/* ptr to first source vertex */
+
+    MOV_L( REGOFF(V4F_START, EDI), EDI ) 	/* ptr to first dest vertex */
+    ADD_L( EDI, ECX ) 				/* count += dest ptr */
+
+    CMP_L( ESI, EDI )
+    JE( LLBL(K_GTP1IR_finish) )
+
+
+ALIGNTEXT32
+LLBL(K_GTP1IR_top):
+    MOV_L( S(0), EDX )
+    MOV_L( EDX, D(0) )
+
+LLBL(K_GTP1IR_skip):
+    ADD_L     ( CONST(16), EDI )
+    ADD_L     ( EAX, ESI )
+    CMP_L     ( ECX, EDI )
+    JNE       ( LLBL(K_GTP1IR_top) )
+
+LLBL(K_GTP1IR_finish):
+    POP_L     ( EDI )
+    POP_L     ( ESI )
+    RET
+#undef FRAME_OFFSET
+
+
+
+ALIGNTEXT4
+GLOBL GLNAME(_mesa_sse_transform_points1_3d_no_rot)
+GLNAME(_mesa_sse_transform_points1_3d_no_rot):
+
+#define FRAME_OFFSET 8
+    PUSH_L( ESI )
+    PUSH_L( EDI )
+
+    MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI )	/* ptr to source GLvector4f */
+    MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) 	/* ptr to dest GLvector4f */
+
+    MOV_L( ARG_MATRIX, EDX ) 			/* ptr to matrix */
+    MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) 	/* source count */
+
+    TEST_L( ECX, ECX)
+    JZ( LLBL(K_GTP13DNRR_finish) ) 		/* count was zero; go to finish */
+
+    MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) 	/* stride */
+    OR_L( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, EDI) ) 	/* set dest flags */
+
+    MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) 	/* set dest count */
+    MOV_L( CONST(3), REGOFF(V4F_SIZE, EDI) ) 	/* set dest size */
+
+    SHL_L( CONST(4), ECX ) 			/* count *= 16 */
+    MOV_L( REGOFF(V4F_START, ESI), ESI ) 	/* ptr to first source vertex */
+
+    MOV_L( REGOFF(V4F_START, EDI), EDI ) 	/* ptr to first dest vertex */
+    ADD_L( EDI, ECX ) 				/* count += dest ptr */
+
+
+ALIGNTEXT32
+    MOVSS( M(0), XMM0 )				/* m0 */
+    MOVSS( M(12), XMM1 )			/* m12 */
+    MOVSS( M(13), XMM2 )			/* m13 */
+    MOVSS( M(14), XMM3 )			/* m14 */
+
+ALIGNTEXT32
+LLBL(K_GTP13DNRR_top):
+    MOVSS( S(0), XMM4 )				/* ox */
+    MULSS( XMM0, XMM4 )				/* ox*m0 */
+    ADDSS( XMM1, XMM4 )				/* ox*m0+m12 */
+    MOVSS( XMM4, D(0) )
+
+    MOVSS( XMM2, D(1) )
+    MOVSS( XMM3, D(2) )
+
+LLBL(K_GTP13DNRR_skip):
+    ADD_L    ( CONST(16), EDI )
+    ADD_L    ( EAX, ESI )
+    CMP_L    ( ECX, EDI )
+    JNE      ( LLBL(K_GTP13DNRR_top) )
+
+LLBL(K_GTP13DNRR_finish):
+    POP_L    ( EDI )
+    POP_L    ( ESI )
+    RET
+#undef FRAME_OFFSET
+
+
+
+ALIGNTEXT4
+GLOBL GLNAME(_mesa_sse_transform_points1_perspective)
+GLNAME(_mesa_sse_transform_points1_perspective):
+
+#define FRAME_OFFSET 8
+    PUSH_L   ( ESI )
+    PUSH_L   ( EDI )
+
+    MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) 	/* ptr to source GLvector4f */
+    MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) 	/* ptr to dest GLvector4f */
+
+    MOV_L( ARG_MATRIX, EDX ) 			/* ptr to matrix */
+    MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) 	/* source count */
+
+    TEST_L( ECX, ECX)
+    JZ( LLBL(K_GTP13PR_finish) )		/* count was zero; go to finish */
+
+    MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) 	/* stride */
+    OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) ) 	/* set dest flags */
+
+    MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) 	/* set dest count */
+    MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) ) 	/* set dest size */
+
+    SHL_L( CONST(4), ECX ) 			/* count *= 16 */
+    MOV_L( REGOFF(V4F_START, ESI), ESI ) 	/* ptr to first source vertex */
+
+    MOV_L( REGOFF(V4F_START, EDI), EDI ) 	/* ptr to first dest vertex */
+    ADD_L( EDI, ECX ) 				/* count += dest ptr */
+
+
+ALIGNTEXT32
+    XORPS( XMM0, XMM0 )				/* 0 | 0 | 0 | 0 */
+    MOVSS( M(0), XMM1 )				/* m0 */
+    MOVSS( M(14), XMM2 )			/* m14 */
+
+ALIGNTEXT32
+LLBL(K_GTP13PR_top):
+    MOVSS( S(0), XMM3 )				/* ox */
+    MULSS( XMM1, XMM3 )				/* ox*m0 */
+    MOVSS( XMM3, D(0) )				/* ox*m0->D(0) */
+    MOVSS( XMM2, D(2) )				/* m14->D(2) */
+
+    MOVSS( XMM0, D(1) )
+    MOVSS( XMM0, D(3) )
+
+LLBL(K_GTP13PR_skip):
+    ADD_L( CONST(16), EDI )
+    ADD_L( EAX, ESI )
+    CMP_L( ECX, EDI )
+    JNE( LLBL(K_GTP13PR_top) )
+
+LLBL(K_GTP13PR_finish):
+    POP_L    ( EDI )
+    POP_L    ( ESI )
+    RET
+#undef FRAME_OFFSET
+
+
+ALIGNTEXT4
+GLOBL GLNAME(_mesa_sse_transform_points1_2d)
+GLNAME(_mesa_sse_transform_points1_2d):
+
+#define FRAME_OFFSET 8
+    PUSH_L( ESI )
+    PUSH_L( EDI )
+
+    MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) 	/* ptr to source GLvector4f */
+    MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) 	/* ptr to dest GLvector4f */
+
+    MOV_L( ARG_MATRIX, EDX ) 			/* ptr to matrix */
+    MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) 	/* source count */
+
+    TEST_L( ECX, ECX)
+    JZ( LLBL(K_GTP13P2DR_finish) ) 		/* count was zero; go to finish */
+
+    MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) 	/* stride */
+    OR_L( CONST(VEC_SIZE_2), REGOFF(V4F_FLAGS, EDI) ) 	/* set dest flags */
+
+    MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) 	/* set dest count */
+    MOV_L( CONST(2), REGOFF(V4F_SIZE, EDI) ) 	/* set dest size */
+
+    SHL_L( CONST(4), ECX ) 			/* count *= 16 */
+    MOV_L( REGOFF(V4F_START, ESI), ESI ) 	/* ptr to first source vertex */
+
+    MOV_L( REGOFF(V4F_START, EDI), EDI ) 	/* ptr to first dest vertex */
+    ADD_L( EDI, ECX ) 				/* count += dest ptr */
+
+ALIGNTEXT32
+    MOVLPS( M(0), XMM0 )			/* m1  | m0  */
+    MOVLPS( M(12), XMM1 )			/* m13 | m12 */
+
+ALIGNTEXT32
+LLBL(K_GTP13P2DR_top):
+    MOVSS( S(0), XMM2 )				/* ox */
+    SHUFPS( CONST(0x0), XMM2, XMM2 )		/* ox | ox | ox | ox */
+    MULPS( XMM0, XMM2 )				/* - | - | ox*m1 | ox*m0 */
+    ADDPS( XMM1, XMM2 )				/* - | - | ox*m1+m13 | ox*m0+m12 */
+    MOVLPS( XMM2, D(0) )
+
+LLBL(K_GTP13P2DR_skip):
+    ADD_L    ( CONST(16), EDI )
+    ADD_L    ( EAX, ESI )
+    CMP_L    ( ECX, EDI )
+    JNE      ( LLBL(K_GTP13P2DR_top) )
+
+LLBL(K_GTP13P2DR_finish):
+    POP_L    ( EDI )
+    POP_L    ( ESI )
+    RET
+#undef FRAME_OFFSET
+
+
+ALIGNTEXT4
+GLOBL GLNAME(_mesa_sse_transform_points1_2d_no_rot)
+GLNAME(_mesa_sse_transform_points1_2d_no_rot):
+
+#define FRAME_OFFSET 8
+	PUSH_L( ESI )
+	PUSH_L( EDI )
+
+	MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) 	/* ptr to source GLvector4f */
+	MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) 	/* ptr to dest GLvector4f */
+
+	MOV_L( ARG_MATRIX, EDX ) 		/* ptr to matrix */
+	MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) 	/* source count */
+
+	TEST_L( ECX, ECX)
+	JZ( LLBL(K_GTP13P2DNRR_finish) ) 	/* count was zero; go to finish */
+
+	MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) 	/* stride */
+	OR_L( CONST(VEC_SIZE_2), REGOFF(V4F_FLAGS, EDI) ) 	/* set dest flags */
+
+	MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) 	/* set dest count */
+	MOV_L( CONST(2), REGOFF(V4F_SIZE, EDI) ) 	/* set dest size */
+
+	SHL_L( CONST(4), ECX ) 			/* count *= 16 */
+	MOV_L( REGOFF(V4F_START, ESI), ESI ) 	/* ptr to first source vertex */
+
+	MOV_L( REGOFF(V4F_START, EDI), EDI ) 	/* ptr to first dest vertex */
+	ADD_L( EDI, ECX ) 			/* count += dest ptr */
+
+ALIGNTEXT32
+	MOVSS( M(0), XMM0 )			/* m0 */
+	MOVSS( M(12), XMM1 )			/* m12 */
+	MOVSS( M(13), XMM2 )			/* m13 */
+
+ALIGNTEXT32
+LLBL(K_GTP13P2DNRR_top):
+	MOVSS( S(0), XMM3 )			/* ox */
+	MULSS( XMM0, XMM3 )			/* ox*m0 */
+	ADDSS( XMM1, XMM3 )			/* ox*m0+m12 */
+	MOVSS( XMM3, D(0) )
+	MOVSS( XMM2, D(1) )
+
+LLBL(K_GTP13P2DNRR_skip):
+	ADD_L( CONST(16), EDI )
+	ADD_L( EAX, ESI )
+	CMP_L( ECX, EDI )
+	JNE( LLBL(K_GTP13P2DNRR_top) )
+
+LLBL(K_GTP13P2DNRR_finish):
+	POP_L( EDI )
+	POP_L( ESI )
+	RET
+#undef FRAME_OFFSET
+
+
+
+ALIGNTEXT4
+GLOBL GLNAME(_mesa_sse_transform_points1_3d)
+GLNAME(_mesa_sse_transform_points1_3d):
+
+#define FRAME_OFFSET 8
+	PUSH_L( ESI )
+	PUSH_L( EDI )
+
+	MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) 	/* ptr to source GLvector4f */
+	MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) 	/* ptr to dest GLvector4f */
+
+	MOV_L( ARG_MATRIX, EDX ) 		/* ptr to matrix */
+	MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) 	/* source count */
+
+	TEST_L( ECX, ECX)
+	JZ( LLBL(K_GTP13P3DR_finish) ) 	/* count was zero; go to finish */
+
+	MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) 	/* stride */
+	OR_L( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, EDI) ) 	/* set dest flags */
+
+	MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) 	/* set dest count */
+	MOV_L( CONST(3), REGOFF(V4F_SIZE, EDI) ) 	/* set dest size */
+
+	SHL_L( CONST(4), ECX ) 			/* count *= 16 */
+	MOV_L( REGOFF(V4F_START, ESI), ESI ) 	/* ptr to first source vertex */
+
+	MOV_L( REGOFF(V4F_START, EDI), EDI ) 	/* ptr to first dest vertex */
+	ADD_L( EDI, ECX ) 			/* count += dest ptr */
+
+
+ALIGNTEXT32
+	MOVAPS( M(0), XMM0 )			/* m3  | m2  | m1  |  m0 */
+	MOVAPS( M(12), XMM1 )			/* m15 | m14 | m13 | m12 */
+
+ALIGNTEXT32
+LLBL(K_GTP13P3DR_top):
+	MOVSS( S(0), XMM2 )			/* ox */
+	SHUFPS( CONST(0x0), XMM2, XMM2 )	/* ox | ox | ox | ox */
+	MULPS( XMM0, XMM2 )			/* ox*m3 | ox*m2 | ox*m1 | ox*m0 */
+	ADDPS( XMM1, XMM2 )			/* +m15  | +m14  | +m13  | +m12  */
+	MOVLPS( XMM2, D(0) )			/*   -   |   -   | ->D(1)| ->D(0)*/
+	UNPCKHPS( XMM2, XMM2 )			/* ox*m3+m15 | ox*m3+m15 | ox*m2+m14 | ox*m2+m14 */
+	MOVSS( XMM2, D(2) )
+
+LLBL(K_GTP13P3DR_skip):
+	ADD_L( CONST(16), EDI )
+	ADD_L( EAX, ESI )
+	CMP_L( ECX, EDI )
+	JNE( LLBL(K_GTP13P3DR_top) )
+
+LLBL(K_GTP13P3DR_finish):
+	POP_L( EDI )
+	POP_L( ESI )
+	RET
+#undef FRAME_OFFSET
diff --git a/src/mesa/x86/sse_xform2.S b/src/mesa/x86/sse_xform2.S
new file mode 100644
index 00000000000..f936088f917
--- /dev/null
+++ b/src/mesa/x86/sse_xform2.S
@@ -0,0 +1,452 @@
+/* $Id: sse_xform2.S,v 1.1 2001/03/29 06:46:27 gareth Exp $ */
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  3.5
+ *
+ * Copyright (C) 1999-2001  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.
+ */
+
+/** TODO:
+  * - insert PREFETCH instructions to avoid cache-misses !
+  * - some more optimizations are possible...
+  * - for 40-50% more performance in the SSE-functions, the
+  *   data (trans-matrix, src_vert, dst_vert) needs to be 16byte aligned !
+  */
+
+#include "matypes.h"
+#include "xform_args.h"
+
+   SEG_TEXT
+
+#define S(i) 	REGOFF(i * 4, ESI)
+#define D(i) 	REGOFF(i * 4, EDI)
+#define M(i) 	REGOFF(i * 4, EDX)
+
+
+ALIGNTEXT4
+GLOBL GLNAME(_mesa_sse_transform_points2_general)
+GLNAME( _mesa_sse_transform_points2_general ):
+
+#define FRAME_OFFSET 8
+    PUSH_L    ( ESI )
+    PUSH_L    ( EDI )
+
+    MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) 	/* ptr to source GLvector4f */
+    MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) 	/* ptr to dest GLvector4f */
+
+    MOV_L( ARG_MATRIX, EDX ) 			/* ptr to matrix */
+    MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) 	/* source count */
+
+    TEST_L( ECX, ECX )
+    JZ( LLBL(K_GTP2GR_finish) )			/* count was zero; go to finish */
+
+    MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) 	/* stride */
+    OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) ) 	/* set dest flags */
+
+    MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) 	/* set dest count */
+    MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) ) 	/* set dest size */
+
+    SHL_L( CONST(4), ECX ) 			/* count *= 16 */
+    MOV_L( REGOFF(V4F_START, ESI), ESI ) 	/* ptr to first source vertex */
+
+    MOV_L( REGOFF(V4F_START, EDI), EDI ) 	/* ptr to first dest vertex */
+    ADD_L( EDI, ECX ) 				/* count += dest ptr */
+
+ALIGNTEXT32
+    MOVAPS( M(0), XMM0 )			/* m3  | m2  | m1  | m0 */
+    MOVAPS( M(4), XMM1 )			/* m7  | m6  | m5  | m4 */
+    MOVAPS( M(12), XMM2 )			/* m15 | m14 | m13 | m12 */
+
+ALIGNTEXT32
+LLBL(K_GTP2GR_top):
+    MOVSS( S(0), XMM3 )				/* ox */
+    SHUFPS( CONST(0x0), XMM3, XMM3 )		/* ox | ox | ox | ox */
+    MULPS( XMM0, XMM3 )				/* ox*m3 | ox*m2 | ox*m1 | ox*m0 */
+    MOVSS( S(1), XMM4 )				/* oy */
+    SHUFPS( CONST(0x0), XMM4, XMM4 )		/* oy | oy | oy | oy */
+    MULPS( XMM1, XMM4 )				/* oy*m7 | oy*m6 | oy*m5 | oy*m4 */
+
+    ADDPS( XMM4, XMM3 )
+    ADDPS( XMM2, XMM3 )
+    MOVAPS( XMM3, D(0) )
+
+LLBL(K_GTP2GR_skip):
+    ADD_L     ( CONST(16), EDI )
+    ADD_L     ( EAX, ESI )
+    CMP_L     ( ECX, EDI )
+    JNE       ( LLBL(K_GTP2GR_top) )
+
+LLBL(K_GTP2GR_finish):
+    POP_L     ( EDI )
+    POP_L     ( ESI )
+    RET
+#undef FRAME_OFFSET
+
+
+ALIGNTEXT4
+GLOBL GLNAME(_mesa_sse_transform_points2_identity)
+GLNAME( _mesa_sse_transform_points2_identity ):
+
+#define FRAME_OFFSET 8
+    PUSH_L    ( ESI )
+    PUSH_L    ( EDI )
+
+    MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) /* ptr to source GLvector4f */
+    MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) 	/* ptr to dest GLvector4f */
+
+    MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) 	/* source count */
+
+    TEST_L( ECX, ECX)
+    JZ( LLBL(K_GTP2IR_finish) )			/* count was zero; go to finish */
+
+    MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) 	/* stride */
+    OR_L( CONST(VEC_SIZE_2), REGOFF(V4F_FLAGS, EDI) ) 	/* set dest flags */
+
+    MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) 	/* set dest count */
+    MOV_L( CONST(2), REGOFF(V4F_SIZE, EDI) ) 	/* set dest size */
+
+    SHL_L( CONST(4), ECX ) 			/* count *= 16 */
+    MOV_L( REGOFF(V4F_START, ESI), ESI ) 	/* ptr to first source vertex */
+
+    MOV_L( REGOFF(V4F_START, EDI), EDI ) 	/* ptr to first dest vertex */
+    ADD_L( EDI, ECX ) 				/* count += dest ptr */
+
+    CMP_L( ESI, EDI )
+    JE( LLBL(K_GTP2IR_finish) )
+
+
+ALIGNTEXT32
+LLBL(K_GTP2IR_top):
+    MOV_L     ( S(0), EDX )
+    MOV_L     ( EDX, D(0) )
+    MOV_L     ( S(1), EDX )
+    MOV_L     ( EDX, D(1) )
+
+LLBL(K_GTP2IR_skip):
+    ADD_L     ( CONST(16), EDI )
+    ADD_L     ( EAX, ESI )
+    CMP_L     ( ECX, EDI )
+    JNE       ( LLBL(K_GTP2IR_top) )
+
+LLBL(K_GTP2IR_finish):
+    POP_L     ( EDI )
+    POP_L     ( ESI )
+    RET
+#undef FRAME_OFFSET
+
+
+ALIGNTEXT4
+GLOBL GLNAME(_mesa_sse_transform_points2_3d_no_rot)
+GLNAME(_mesa_sse_transform_points2_3d_no_rot):
+
+#define FRAME_OFFSET 8
+    PUSH_L( ESI )
+    PUSH_L( EDI )
+
+    MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) 	/* ptr to source GLvector4f */
+    MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) 	/* ptr to dest GLvector4f */
+
+    MOV_L( ARG_MATRIX, EDX ) 			/* ptr to matrix */
+    MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) 	/* source count */
+
+    TEST_L( ECX, ECX)
+    JZ( LLBL(K_GTP23DNRR_finish) ) 		/* count was zero; go to finish */
+
+    MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) 	/* stride */
+    OR_L( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, EDI) ) 	/* set dest flags */
+
+    MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) 	/* set dest count */
+    MOV_L( CONST(3), REGOFF(V4F_SIZE, EDI) ) 	/* set dest size */
+
+    SHL_L( CONST(4), ECX ) 			/* count *= 16 */
+    MOV_L( REGOFF(V4F_START, ESI), ESI ) 	/* ptr to first source vertex */
+
+    MOV_L( REGOFF(V4F_START, EDI), EDI ) 	/* ptr to first dest vertex */
+    ADD_L( EDI, ECX ) 				/* count += dest ptr */
+
+
+ALIGNTEXT32
+    MOVSS    ( M(0), XMM1 )			/* - | - |  -  | m0  */
+    MOVSS    ( M(5), XMM2 )			/* - | - |  -  | m5  */
+    UNPCKLPS ( XMM2, XMM1 )			/* - | - | m5  | m0  */
+    MOVLPS   ( M(12), XMM2 )			/* - | - | m13 | m12 */
+    MOVSS    ( M(14), XMM3 )			/* - | - |  -  | m14 */
+
+ALIGNTEXT32
+LLBL(K_GTP23DNRR_top):
+    MOVLPS   ( S(0), XMM0 )			/* - | - |  oy   | ox */
+    MULPS    ( XMM1, XMM0 )			/* - | - | oy*m5 | ox*m0 */
+    ADDPS    ( XMM2, XMM0 )			/* - | - | +m13  | +m12 */
+    MOVLPS   ( XMM0, D(0) )			/* -> D(1) | -> D(0) */
+
+    MOVSS    ( XMM3, D(2) )			/* -> D(2) */
+
+LLBL(K_GTP23DNRR_skip):
+    ADD_L    ( CONST(16), EDI )
+    ADD_L    ( EAX, ESI )
+    CMP_L    ( ECX, EDI )
+    JNE      ( LLBL(K_GTP23DNRR_top) )
+
+LLBL(K_GTP23DNRR_finish):
+    POP_L    ( EDI )
+    POP_L    ( ESI )
+    RET
+#undef FRAME_OFFSET
+
+
+ALIGNTEXT4
+GLOBL GLNAME(_mesa_sse_transform_points2_perspective)
+GLNAME(_mesa_sse_transform_points2_perspective):
+
+#define FRAME_OFFSET 8
+    PUSH_L   ( ESI )
+    PUSH_L   ( EDI )
+
+    MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI )	/* ptr to source GLvector4f */
+    MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) 	/* ptr to dest GLvector4f */
+
+    MOV_L( ARG_MATRIX, EDX ) 			/* ptr to matrix */
+    MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) 	/* source count */
+
+    TEST_L( ECX, ECX)
+    JZ( LLBL(K_GTP23PR_finish) )		/* count was zero; go to finish */
+
+    MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) 	/* stride */
+    OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) ) 	/* set dest flags */
+
+    MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) 	/* set dest count */
+    MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) ) 	/* set dest size */
+
+    SHL_L( CONST(4), ECX ) 			/* count *= 16 */
+    MOV_L( REGOFF(V4F_START, ESI), ESI ) 	/* ptr to first source vertex */
+
+    MOV_L( REGOFF(V4F_START, EDI), EDI ) 	/* ptr to first dest vertex */
+    ADD_L( EDI, ECX ) 				/* count += dest ptr */
+
+ALIGNTEXT32
+    MOVSS    ( M(0), XMM1 )			/* -  | -  |  -  | m0  */
+    MOVSS    ( M(5), XMM2 )			/* -  | -  |  -  | m5  */
+    UNPCKLPS ( XMM2, XMM1 )			/* -  | -  | m5  | m0  */
+    MOVSS    ( M(14), XMM3 )			/* m14 */
+    XORPS    ( XMM0, XMM0 )			/* 0 | 0 | 0 | 0 */
+
+ALIGNTEXT32
+LLBL(K_GTP23PR_top):
+    MOVLPS( S(0), XMM4 )			/* oy | ox */
+    MULPS( XMM1, XMM4 )				/* oy*m5 | ox*m0 */
+    MOVLPS( XMM4, D(0) )			/* ->D(1) | ->D(0) */
+    MOVSS( XMM3, D(2) )				/* ->D(2) */
+    MOVSS( XMM0, D(3) )				/* ->D(3) */
+
+LLBL(K_GTP23PR_skip):
+    ADD_L( CONST(16), EDI )
+    ADD_L( EAX, ESI )
+    CMP_L( ECX, EDI )
+    JNE( LLBL(K_GTP23PR_top) )
+
+LLBL(K_GTP23PR_finish):
+    POP_L    ( EDI )
+    POP_L    ( ESI )
+    RET
+#undef FRAME_OFFSET
+
+
+
+ALIGNTEXT4
+GLOBL GLNAME(_mesa_sse_transform_points2_2d)
+GLNAME(_mesa_sse_transform_points2_2d):
+
+#define FRAME_OFFSET 8
+    PUSH_L( ESI )
+    PUSH_L( EDI )
+
+    MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) 	/* ptr to source GLvector4f */
+    MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) 	/* ptr to dest GLvector4f */
+
+    MOV_L( ARG_MATRIX, EDX ) 			/* ptr to matrix */
+    MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) 	/* source count */
+
+    TEST_L( ECX, ECX)
+    JZ( LLBL(K_GTP23P2DR_finish) ) 		/* count was zero; go to finish */
+
+    MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) 	/* stride */
+    OR_L( CONST(VEC_SIZE_2), REGOFF(V4F_FLAGS, EDI) ) 	/* set dest flags */
+
+    MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) 	/* set dest count */
+    MOV_L( CONST(2), REGOFF(V4F_SIZE, EDI) ) 	/* set dest size */
+
+    SHL_L( CONST(4), ECX ) 			/* count *= 16 */
+    MOV_L( REGOFF(V4F_START, ESI), ESI ) 	/* ptr to first source vertex */
+
+    MOV_L( REGOFF(V4F_START, EDI), EDI ) 	/* ptr to first dest vertex */
+    ADD_L( EDI, ECX ) 				/* count += dest ptr */
+
+ALIGNTEXT32
+    MOVLPS( M(0), XMM0 )			/* m1  | m0 */
+    MOVLPS( M(4), XMM1 )			/* m5  | m4 */
+    MOVLPS( M(12), XMM2 )			/* m13 | m12 */
+
+ALIGNTEXT32
+LLBL(K_GTP23P2DR_top):
+    MOVSS( S(0), XMM3 )				/* ox */
+    SHUFPS( CONST(0x0), XMM3, XMM3 )		/* ox | ox */
+    MULPS( XMM0, XMM3 )				/* ox*m1 | ox*m0 */
+
+    MOVSS( S(1), XMM4 )				/* oy */
+    SHUFPS( CONST(0x0), XMM4, XMM4 )		/* oy | oy */
+    MULPS( XMM1, XMM4 )				/* oy*m5 | oy*m4 */
+
+    ADDPS( XMM4, XMM3 )
+    ADDPS( XMM2, XMM3 )
+    MOVLPS( XMM3, D(0) )			/* ->D(1) | ->D(0) */
+
+LLBL(K_GTP23P2DR_skip):
+    ADD_L    ( CONST(16), EDI )
+    ADD_L    ( EAX, ESI )
+    CMP_L    ( ECX, EDI )
+    JNE      ( LLBL(K_GTP23P2DR_top) )
+
+LLBL(K_GTP23P2DR_finish):
+    POP_L    ( EDI )
+    POP_L    ( ESI )
+    RET
+#undef FRAME_OFFSET
+
+
+
+ALIGNTEXT4
+GLOBL GLNAME(_mesa_sse_transform_points2_2d_no_rot)
+GLNAME(_mesa_sse_transform_points2_2d_no_rot):
+
+#define FRAME_OFFSET 8
+	PUSH_L( ESI )
+	PUSH_L( EDI )
+
+	MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) 	/* ptr to source GLvector4f */
+	MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) 	/* ptr to dest GLvector4f */
+
+	MOV_L( ARG_MATRIX, EDX ) 		/* ptr to matrix */
+	MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) 	/* source count */
+
+	TEST_L( ECX, ECX)
+	JZ( LLBL(K_GTP23P2DNRR_finish) ) 	/* count was zero; go to finish */
+
+	MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) 	/* stride */
+	OR_L( CONST(VEC_SIZE_2), REGOFF(V4F_FLAGS, EDI) ) 	/* set dest flags */
+
+	MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) 	/* set dest count */
+	MOV_L( CONST(2), REGOFF(V4F_SIZE, EDI) ) 	/* set dest size */
+
+	SHL_L( CONST(4), ECX ) 			/* count *= 16 */
+	MOV_L( REGOFF(V4F_START, ESI), ESI ) 	/* ptr to first source vertex */
+
+	MOV_L( REGOFF(V4F_START, EDI), EDI ) 	/* ptr to first dest vertex */
+	ADD_L( EDI, ECX ) 			/* count += dest ptr */
+
+ALIGNTEXT32
+	MOVSS    ( M(0), XMM1 )			/* m0 */
+	MOVSS    ( M(5), XMM2 )			/* m5 */
+	UNPCKLPS ( XMM2, XMM1 )			/* m5 | m0 */
+	MOVLPS   ( M(12), XMM2 )		/* m13 | m12 */
+
+ALIGNTEXT32
+LLBL(K_GTP23P2DNRR_top):
+	MOVLPS( S(0), XMM0 )			/* oy | ox */
+	MULPS( XMM1, XMM0 )			/* oy*m5 | ox*m0 */
+	ADDPS( XMM2, XMM0 )			/* +m13 | +m12 */
+	MOVLPS( XMM0, D(0) )			/* ->D(1) | ->D(0) */
+
+LLBL(K_GTP23P2DNRR_skip):
+	ADD_L( CONST(16), EDI )
+	ADD_L( EAX, ESI )
+	CMP_L( ECX, EDI )
+	JNE( LLBL(K_GTP23P2DNRR_top) )
+
+LLBL(K_GTP23P2DNRR_finish):
+	POP_L( EDI )
+	POP_L( ESI )
+	RET
+#undef FRAME_OFFSET
+
+
+
+ALIGNTEXT4
+GLOBL GLNAME(_mesa_sse_transform_points2_3d)
+GLNAME(_mesa_sse_transform_points2_3d):
+
+#define FRAME_OFFSET 8
+	PUSH_L( ESI )
+	PUSH_L( EDI )
+
+	MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) 	/* ptr to source GLvector4f */
+	MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) 	/* ptr to dest GLvector4f */
+
+	MOV_L( ARG_MATRIX, EDX ) 		/* ptr to matrix */
+	MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) 	/* source count */
+
+	TEST_L( ECX, ECX)
+	JZ( LLBL(K_GTP23P3DR_finish) ) 	/* count was zero; go to finish */
+
+	MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) 	/* stride */
+	OR_L( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, EDI) ) 	/* set dest flags */
+
+	MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) 	/* set dest count */
+	MOV_L( CONST(3), REGOFF(V4F_SIZE, EDI) ) 	/* set dest size */
+
+	SHL_L( CONST(4), ECX ) 			/* count *= 16 */
+	MOV_L( REGOFF(V4F_START, ESI), ESI ) 	/* ptr to first source vertex */
+
+	MOV_L( REGOFF(V4F_START, EDI), EDI ) 	/* ptr to first dest vertex */
+	ADD_L( EDI, ECX ) 			/* count += dest ptr */
+
+ALIGNTEXT32
+	MOVAPS( M(0), XMM0 )			/* m2  | m1  | m0 */
+	MOVAPS( M(4), XMM1 )			/* m6  | m5  | m4 */
+	MOVAPS( M(12), XMM2 )			/* m14 | m13 | m12 */
+
+ALIGNTEXT32
+LLBL(K_GTP23P3DR_top):
+	MOVSS( S(0), XMM3 )			/* ox */
+	SHUFPS( CONST(0x0), XMM3, XMM3 )	/* ox | ox | ox */
+	MULPS( XMM0, XMM3 )			/* ox*m2 | ox*m1 | ox*m0 */
+
+	MOVSS( S(1), XMM4 )			/* oy */
+	SHUFPS( CONST(0x0), XMM4, XMM4 )	/* oy | oy | oy */
+	MULPS( XMM1, XMM4 )			/* oy*m6 | oy*m5 | oy*m4 */
+
+	ADDPS( XMM4, XMM3 )
+	ADDPS( XMM2, XMM3 )
+
+	MOVLPS( XMM3, D(0) )			/* ->D(1) | ->D(0) */
+	UNPCKHPS( XMM3, XMM3 )
+	MOVSS( XMM3, D(2) )			/* ->D(2) */
+
+LLBL(K_GTP23P3DR_skip):
+	ADD_L( CONST(16), EDI )
+	ADD_L( EAX, ESI )
+	CMP_L( ECX, EDI )
+	JNE( LLBL(K_GTP23P3DR_top) )
+
+LLBL(K_GTP23P3DR_finish):
+	POP_L( EDI )
+	POP_L( ESI )
+	RET
+#undef FRAME_OFFSET
diff --git a/src/mesa/x86/sse_xform3.S b/src/mesa/x86/sse_xform3.S
new file mode 100644
index 00000000000..ff989c337df
--- /dev/null
+++ b/src/mesa/x86/sse_xform3.S
@@ -0,0 +1,498 @@
+/* $Id: sse_xform3.S,v 1.1 2001/03/29 06:46:27 gareth Exp $ */
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  3.5
+ *
+ * Copyright (C) 1999-2001  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.
+ */
+
+/** TODO:
+  * - insert PREFETCH instructions to avoid cache-misses !
+  * - some more optimizations are possible...
+  * - for 40-50% more performance in the SSE-functions, the
+  *   data (trans-matrix, src_vert, dst_vert) needs to be 16byte aligned !
+  */
+
+#include "matypes.h"
+#include "xform_args.h"
+
+   SEG_TEXT
+
+#define S(i) 	REGOFF(i * 4, ESI)
+#define D(i) 	REGOFF(i * 4, EDI)
+#define M(i) 	REGOFF(i * 4, EDX)
+
+
+ALIGNTEXT4
+GLOBL GLNAME(_mesa_sse_transform_points3_general)
+GLNAME( _mesa_sse_transform_points3_general ):
+
+#define FRAME_OFFSET 8
+    PUSH_L    ( ESI )
+    PUSH_L    ( EDI )
+
+    MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) 	/* ptr to source GLvector4f */
+    MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) 	/* ptr to dest GLvector4f */
+
+    MOV_L( ARG_MATRIX, EDX ) 			/* ptr to matrix */
+    MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) 	/* source count */
+
+    CMP_L     ( CONST(0), ECX )			/* count == 0 ? */
+    JE        ( LLBL(K_GTPGR_finish) )		/* yes -> nothing to do. */
+
+    MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) 	/* stride */
+    OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) ) 	/* set dest flags */
+
+    MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) 	/* set dest count */
+    MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) ) 	/* set dest size */
+
+    SHL_L( CONST(4), ECX ) 			/* count *= 16 */
+    MOV_L( REGOFF(V4F_START, ESI), ESI ) 	/* ptr to first source vertex */
+
+    MOV_L( REGOFF(V4F_START, EDI), EDI ) 	/* ptr to first dest vertex */
+    ADD_L( EDI, ECX ) 				/* count += dest ptr */
+
+
+ALIGNTEXT32
+    MOVAPS    ( REGOFF(0, EDX), XMM0 )	/* m0  | m1  | m2  | m3 */
+    MOVAPS    ( REGOFF(16, EDX), XMM1 )	/* m4  | m5  | m6  | m7 */
+    MOVAPS    ( REGOFF(32, EDX), XMM2 )	/* m8  | m9  | m10 | m11 */
+    MOVAPS    ( REGOFF(48, EDX), XMM3 )	/* m12 | m13 | m14 | m15 */
+
+
+ALIGNTEXT32
+LLBL(K_GTPGR_top):
+    MOVSS     ( REGOFF(0, ESI), XMM4 )		/*    |    |    | ox */
+    SHUFPS    ( CONST(0x0), XMM4, XMM4 )	/* ox | ox | ox | ox */
+    MOVSS     ( REGOFF(4, ESI), XMM5 )		/*    |    |    | oy */
+    SHUFPS    ( CONST(0x0), XMM5, XMM5 )	/* oy | oy | oy | oy */
+    MOVSS     ( REGOFF(8, ESI), XMM6 )		/*    |    |    | oz */
+    SHUFPS    ( CONST(0x0), XMM6, XMM6 )	/* oz | oz | oz | oz */
+
+    MULPS     ( XMM0, XMM4 )		/* m3*ox  | m2*ox  | m1*ox | m0*ox */
+    MULPS     ( XMM1, XMM5 )		/* m7*oy  | m6*oy  | m5*oy | m4*oy */
+    MULPS     ( XMM2, XMM6 )		/* m11*oz | m10*oz | m9*oz | m8*oz */
+
+    ADDPS     ( XMM5, XMM4 )
+    ADDPS     ( XMM6, XMM4 )
+    ADDPS     ( XMM3, XMM4 )
+
+    MOVAPS    ( XMM4, REGOFF(0, EDI) )
+
+LLBL(K_GTPGR_skip):
+    ADD_L     ( CONST(16), EDI )
+    ADD_L     ( EAX, ESI )
+    CMP_L     ( ECX, EDI )
+    JNE       ( LLBL(K_GTPGR_top) )
+
+LLBL(K_GTPGR_finish):
+    POP_L     ( EDI )
+    POP_L     ( ESI )
+    RET
+#undef FRAME_OFFSET
+
+
+ALIGNTEXT4
+GLOBL GLNAME(_mesa_sse_transform_points3_identity)
+GLNAME( _mesa_sse_transform_points3_identity ):
+
+#define FRAME_OFFSET 8
+    PUSH_L    ( ESI )
+    PUSH_L    ( EDI )
+
+    MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) 	/* ptr to source GLvector4f */
+    MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) 	/* ptr to dest GLvector4f */
+
+    MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) 	/* source count */
+
+    TEST_L( ECX, ECX)
+    JZ( LLBL(K_GTPIR_finish) ) 			/* count was zero; go to finish */
+
+    MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) 	/* stride */
+    OR_L( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, EDI) ) 	/* set dest flags */
+
+    MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) 	/* set dest count */
+    MOV_L( CONST(3), REGOFF(V4F_SIZE, EDI) ) 	/* set dest size */
+
+    SHL_L( CONST(4), ECX ) 			/* count *= 16 */
+    MOV_L( REGOFF(V4F_START, ESI), ESI ) 	/* ptr to first source vertex */
+
+    MOV_L( REGOFF(V4F_START, EDI), EDI ) 	/* ptr to first dest vertex */
+    ADD_L( EDI, ECX ) 				/* count += dest ptr */
+
+    CMP_L( ESI, EDI )
+    JE( LLBL(K_GTPIR_finish) )
+
+
+ALIGNTEXT32
+LLBL(K_GTPIR_top):
+    MOVLPS    ( S(0), XMM0 )
+    MOVLPS    ( XMM0, D(0) )
+    MOVSS     ( S(2), XMM0 )
+    MOVSS     ( XMM0, D(2) )
+
+LLBL(K_GTPIR_skip):
+    ADD_L     ( CONST(16), EDI )
+    ADD_L     ( EAX, ESI )
+    CMP_L     ( ECX, EDI )
+    JNE       ( LLBL(K_GTPIR_top) )
+
+LLBL(K_GTPIR_finish):
+    POP_L     ( EDI )
+    POP_L     ( ESI )
+    RET
+#undef FRAME_OFFSET
+
+
+
+
+ALIGNTEXT4
+GLOBL GLNAME(_mesa_sse_transform_points3_3d_no_rot)
+GLNAME(_mesa_sse_transform_points3_3d_no_rot):
+
+#define FRAME_OFFSET 8
+    PUSH_L( ESI )
+    PUSH_L( EDI )
+
+    MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) 	/* ptr to source GLvector4f */
+    MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) 	/* ptr to dest GLvector4f */
+
+
+    MOV_L( ARG_MATRIX, EDX ) 			/* ptr to matrix */
+    MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) 	/* source count */
+
+    TEST_L( ECX, ECX)
+    JZ( LLBL(K_GTP3DNRR_finish) ) 		/* count was zero; go to finish */
+
+    MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) 	/* stride */
+    OR_L( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, EDI) ) 	/* set dest flags */
+
+    MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) 	/* set dest count */
+    MOV_L( CONST(3), REGOFF(V4F_SIZE, EDI) ) 	/* set dest size */
+
+    SHL_L( CONST(4), ECX ) 			/* count *= 16 */
+    MOV_L( REGOFF(V4F_START, ESI), ESI ) 	/* ptr to first source vertex */
+
+    MOV_L( REGOFF(V4F_START, EDI), EDI ) 	/* ptr to first dest vertex */
+    ADD_L( EDI, ECX ) 				/* count += dest ptr */
+
+
+ALIGNTEXT32
+    MOVSS    ( M(0), XMM1 )			/* - | - |  -  | m0  */
+    MOVSS    ( M(5), XMM2 )			/* - | - |  -  | m5  */
+    UNPCKLPS ( XMM2, XMM1 )			/* - | - | m5  | m0  */
+    MOVLPS   ( M(12), XMM2 )			/* - | - | m13 | m12 */
+    MOVSS    ( M(10), XMM3 )			/* - | - |  -  | m10 */
+    MOVSS    ( M(14), XMM4 )			/* - | - |  -  | m14 */
+
+ALIGNTEXT32
+LLBL(K_GTP3DNRR_top):
+
+    MOVLPS   ( S(0), XMM0 )			/* - | - |  s1   | s0 */
+    MULPS    ( XMM1, XMM0 )			/* - | - | s1*m5 | s0*m0 */
+    ADDPS    ( XMM2, XMM0 )			/* - | - | +m13  | +m12 */
+    MOVLPS   ( XMM0, D(0) )			/* -> D(1) | -> D(0) */
+
+    MOVSS    ( S(2), XMM0 )			/* sz */
+    MULSS    ( XMM3, XMM0 )			/* sz*m10 */
+    ADDSS    ( XMM4, XMM0 )			/* +m14 */
+    MOVSS    ( XMM0, D(2) )			/* -> D(2) */
+
+LLBL(K_GTP3DNRR_skip):
+    ADD_L    ( CONST(16), EDI )
+    ADD_L    ( EAX, ESI )
+    CMP_L    ( ECX, EDI )
+    JNE      ( LLBL(K_GTP3DNRR_top) )
+
+LLBL(K_GTP3DNRR_finish):
+    POP_L    ( EDI )
+    POP_L    ( ESI )
+    RET
+#undef FRAME_OFFSET
+
+
+
+ALIGNTEXT4
+GLOBL GLNAME(_mesa_sse_transform_points3_perspective)
+GLNAME(_mesa_sse_transform_points3_perspective):
+
+#define FRAME_OFFSET 8
+    PUSH_L   ( ESI )
+    PUSH_L   ( EDI )
+
+    MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) 	/* ptr to source GLvector4f */
+    MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) 	/* ptr to dest GLvector4f */
+
+    MOV_L( ARG_MATRIX, EDX ) 			/* ptr to matrix */
+    MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) 	/* source count */
+
+    TEST_L( ECX, ECX)
+    JZ( LLBL(K_GTP3PR_finish) )			/* count was zero; go to finish */
+
+    MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) 	/* stride */
+    OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) ) 	/* set dest flags */
+
+    MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) 	/* set dest count */
+    MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) ) 	/* set dest size */
+
+    SHL_L( CONST(4), ECX ) 			/* count *= 16 */
+    MOV_L( REGOFF(V4F_START, ESI), ESI ) 	/* ptr to first source vertex */
+
+    MOV_L( REGOFF(V4F_START, EDI), EDI ) 	/* ptr to first dest vertex */
+    ADD_L( EDI, ECX ) 				/* count += dest ptr */
+
+ALIGNTEXT32
+    MOVSS    ( M(0), XMM1 )			/* -  | -  |  -  | m0  */
+    MOVSS    ( M(5), XMM2 )			/* -  | -  |  -  | m5  */
+    UNPCKLPS ( XMM2, XMM1 )			/* -  | -  | m5  | m0  */
+    MOVLPS   ( M(8), XMM2 )			/* -  | -  | m9  | m8  */
+    MOVSS    ( M(10), XMM3 )			/* m10 */
+    MOVSS    ( M(14), XMM4 )			/* m14 */
+    XORPS    ( XMM6, XMM6 )			/* 0 */
+
+ALIGNTEXT32
+LLBL(K_GTP3PR_top):
+    MOVLPS   ( S(0), XMM0 )			/* oy | ox */
+    MULPS    ( XMM1, XMM0 )			/* oy*m5 | ox*m0 */
+    MOVSS    ( S(2), XMM5 )			/* oz */
+    SHUFPS   ( CONST(0x0), XMM5, XMM5 )		/* oz | oz */
+    MULPS    ( XMM2, XMM5 )			/* oz*m9 | oz*m8 */
+    ADDPS    ( XMM5, XMM0 )			/* +oy*m5 | +ox*m0 */
+    MOVLPS   ( XMM0, D(0) )			/* ->D(1) | ->D(0) */
+
+    MOVSS    ( S(2), XMM0 )			/* oz */
+    MULSS    ( XMM3, XMM0 )			/* oz*m10 */
+    ADDSS    ( XMM4, XMM0 )			/* +m14 */
+    MOVSS    ( XMM0, D(2) )			/* ->D(2) */
+
+    MOVSS    ( S(2), XMM0 )			/* oz */
+    MOVSS    ( XMM6, XMM5 )			/* 0 */
+    SUBPS    ( XMM0, XMM5 )			/* -oz */
+    MOVSS    ( XMM5, D(3) )			/* ->D(3) */
+
+LLBL(K_GTP3PR_skip):
+    ADD_L( CONST(16), EDI )
+    ADD_L( EAX, ESI )
+    CMP_L( ECX, EDI )
+    JNE( LLBL(K_GTP3PR_top) )
+
+LLBL(K_GTP3PR_finish):
+    POP_L    ( EDI )
+    POP_L    ( ESI )
+    RET
+#undef FRAME_OFFSET
+
+
+
+ALIGNTEXT4
+GLOBL GLNAME(_mesa_sse_transform_points3_2d)
+GLNAME(_mesa_sse_transform_points3_2d):
+
+#define FRAME_OFFSET 8
+    PUSH_L( ESI )
+    PUSH_L( EDI )
+
+    MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) 	/* ptr to source GLvector4f */
+    MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) 	/* ptr to dest GLvector4f */
+
+    MOV_L( ARG_MATRIX, EDX ) 			/* ptr to matrix */
+    MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) 	/* source count */
+
+    TEST_L( ECX, ECX)
+    JZ( LLBL(K_GTP3P2DR_finish) ) 		/* count was zero; go to finish */
+
+    MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) 	/* stride */
+    OR_L( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, EDI) ) 	/* set dest flags */
+
+    MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) 	/* set dest count */
+    MOV_L( CONST(3), REGOFF(V4F_SIZE, EDI) ) 	/* set dest size */
+
+    SHL_L( CONST(4), ECX ) 			/* count *= 16 */
+    MOV_L( REGOFF(V4F_START, ESI), ESI ) 	/* ptr to first source vertex */
+
+    MOV_L( REGOFF(V4F_START, EDI), EDI ) 	/* ptr to first dest vertex */
+    ADD_L( EDI, ECX ) 				/* count += dest ptr */
+
+ALIGNTEXT32
+    MOVLPS( M(0), XMM0 )			/* m1  | m0 */
+    MOVLPS( M(4), XMM1 )			/* m5  | m4 */
+    MOVLPS( M(12), XMM2 )			/* m13 | m12 */
+
+ALIGNTEXT32
+LLBL(K_GTP3P2DR_top):
+    MOVSS    ( S(0), XMM3 )			/* ox */
+    SHUFPS   ( CONST(0x0), XMM3, XMM3 )		/* ox | ox */
+    MULPS    ( XMM0, XMM3 )			/* ox*m1 | ox*m0 */
+    MOVSS    ( S(1), XMM4 )			/* oy */
+    SHUFPS   ( CONST(0x0), XMM4, XMM4 )		/* oy | oy */
+    MULPS    ( XMM1, XMM4 )			/* oy*m5 | oy*m4 */
+
+    ADDPS    ( XMM4, XMM3 )
+    ADDPS    ( XMM2, XMM3 )
+    MOVLPS   ( XMM3, D(0) )
+
+    MOVSS    ( S(2), XMM3 )
+    MOVSS    ( XMM3, D(2) )
+
+LLBL(K_GTP3P2DR_skip):
+    ADD_L    ( CONST(16), EDI )
+    ADD_L    ( EAX, ESI )
+    CMP_L    ( ECX, EDI )
+    JNE      ( LLBL(K_GTP3P2DR_top) )
+
+LLBL(K_GTP3P2DR_finish):
+    POP_L    ( EDI )
+    POP_L    ( ESI )
+    RET
+#undef FRAME_OFFSET
+
+
+
+ALIGNTEXT4
+GLOBL GLNAME(_mesa_sse_transform_points3_2d_no_rot)
+GLNAME(_mesa_sse_transform_points3_2d_no_rot):
+
+#define FRAME_OFFSET 8
+	PUSH_L( ESI )
+	PUSH_L( EDI )
+
+	MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) 	/* ptr to source GLvector4f */
+	MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) 	/* ptr to dest GLvector4f */
+
+	MOV_L( ARG_MATRIX, EDX ) 		/* ptr to matrix */
+	MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) 	/* source count */
+
+	TEST_L( ECX, ECX)
+	JZ( LLBL(K_GTP3P2DNRR_finish) ) 	/* count was zero; go to finish */
+
+	MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) 	/* stride */
+	OR_L( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, EDI) ) 	/* set dest flags */
+
+	MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) 	/* set dest count */
+	MOV_L( CONST(3), REGOFF(V4F_SIZE, EDI) ) 	/* set dest size */
+
+	SHL_L( CONST(4), ECX ) 			/* count *= 16 */
+	MOV_L( REGOFF(V4F_START, ESI), ESI ) 	/* ptr to first source vertex */
+
+	MOV_L( REGOFF(V4F_START, EDI), EDI ) 	/* ptr to first dest vertex */
+	ADD_L( EDI, ECX ) 			/* count += dest ptr */
+
+ALIGNTEXT32
+	MOVSS    ( M(0), XMM1 )			/* m0 */
+	MOVSS    ( M(5), XMM2 )			/* m5 */
+	UNPCKLPS ( XMM2, XMM1 )			/* m5 | m0 */
+	MOVLPS   ( M(12), XMM2 )		/* m13 | m12 */
+
+ALIGNTEXT32
+LLBL(K_GTP3P2DNRR_top):
+	MOVLPS( S(0), XMM0 )			/* oy | ox */
+	MULPS( XMM1, XMM0 )			/* oy*m5 | ox*m0 */
+	ADDPS( XMM2, XMM0 )			/* +m13 | +m12 */
+	MOVLPS( XMM0, D(0) )			/* ->D(1) | ->D(0) */
+
+	MOVSS( S(2), XMM0 )
+	MOVSS( XMM0, D(2) )
+
+LLBL(K_GTP3P2DNRR_skip):
+	ADD_L( CONST(16), EDI )
+	ADD_L( EAX, ESI )
+	CMP_L( ECX, EDI )
+	JNE( LLBL(K_GTP3P2DNRR_top) )
+
+LLBL(K_GTP3P2DNRR_finish):
+	POP_L( EDI )
+	POP_L( ESI )
+	RET
+#undef FRAME_OFFSET
+
+
+
+
+ALIGNTEXT4
+GLOBL GLNAME(_mesa_sse_transform_points3_3d)
+GLNAME(_mesa_sse_transform_points3_3d):
+
+#define FRAME_OFFSET 8
+	PUSH_L( ESI )
+	PUSH_L( EDI )
+
+	MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) 	/* ptr to source GLvector4f */
+	MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) 	/* ptr to dest GLvector4f */
+
+
+	MOV_L( ARG_MATRIX, EDX ) 		/* ptr to matrix */
+	MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) 	/* source count */
+
+	TEST_L( ECX, ECX)
+	JZ( LLBL(K_GTP3P3DR_finish) ) 	/* count was zero; go to finish */
+
+	MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) 	/* stride */
+	OR_L( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, EDI) ) 	/* set dest flags */
+
+	MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) 	/* set dest count */
+	MOV_L( CONST(3), REGOFF(V4F_SIZE, EDI) ) 	/* set dest size */
+
+	SHL_L( CONST(4), ECX ) 			/* count *= 16 */
+	MOV_L( REGOFF(V4F_START, ESI), ESI ) 	/* ptr to first source vertex */
+
+	MOV_L( REGOFF(V4F_START, EDI), EDI ) 	/* ptr to first dest vertex */
+	ADD_L( EDI, ECX ) 			/* count += dest ptr */
+
+
+ALIGNTEXT32
+	MOVAPS( M(0), XMM0 )			/* m2  | m1  | m0 */
+	MOVAPS( M(4), XMM1 )			/* m6  | m5  | m4 */
+	MOVAPS( M(8), XMM2 )			/* m10 | m9  | m8 */
+	MOVAPS( M(12), XMM3 )			/* m14 | m13 | m12 */
+
+ALIGNTEXT32
+LLBL(K_GTP3P3DR_top):
+	MOVSS( S(0), XMM4 )
+	SHUFPS( CONST(0x0), XMM4, XMM4 )	/* ox | ox | ox */
+	MULPS( XMM0, XMM4 )			/* ox*m2 | ox*m1 | ox*m0 */
+
+	MOVSS( S(1), XMM5 )
+	SHUFPS( CONST(0x0), XMM5, XMM5 )	/* oy | oy | oy */
+	MULPS( XMM1, XMM5 )			/* oy*m6 | oy*m5 | oy*m4 */
+
+	MOVSS( S(2), XMM6 )
+	SHUFPS( CONST(0x0), XMM6, XMM6 )	/* oz | oz | oz */
+	MULPS( XMM2, XMM6 )			/* oz*m10 | oz*m9 | oz*m8 */
+
+	ADDPS( XMM5, XMM4 )			/* + | + | + */
+	ADDPS( XMM6, XMM4 )			/* + | + | + */
+	ADDPS( XMM3, XMM4 )			/* + | + | + */
+
+	MOVLPS( XMM4, D(0) )			/* => D(1) | => D(0) */
+	UNPCKHPS( XMM4, XMM4 )
+	MOVSS( XMM4, D(2) )
+
+LLBL(K_GTP3P3DR_skip):
+	ADD_L( CONST(16), EDI )
+	ADD_L( EAX, ESI )
+	CMP_L( ECX, EDI )
+	JNE( LLBL(K_GTP3P3DR_top) )
+
+LLBL(K_GTP3P3DR_finish):
+	POP_L( EDI )
+	POP_L( ESI )
+	RET
+#undef FRAME_OFFSET
diff --git a/src/mesa/x86/sse_xform4.S b/src/mesa/x86/sse_xform4.S
new file mode 100644
index 00000000000..13cb2a3ecb5
--- /dev/null
+++ b/src/mesa/x86/sse_xform4.S
@@ -0,0 +1,226 @@
+/* $Id: sse_xform4.S,v 1.1 2001/03/29 06:46:27 gareth Exp $ */
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  3.5
+ *
+ * Copyright (C) 1999-2001  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.
+ */
+
+#include "matypes.h"
+#include "xform_args.h"
+
+	SEG_TEXT
+
+#define FRAME_OFFSET	8
+
+#define SRC(i)		REGOFF(i * 4, ESI)
+#define DST(i)		REGOFF(i * 4, EDI)
+#define MAT(i)		REGOFF(i * 4, EDX)
+
+#define SELECT(r0, r1, r2, r3)	CONST( r0 * 64 + r1 * 16 + r2 * 4 + r3 )
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_sse_transform_points4_general )
+GLNAME( _mesa_sse_transform_points4_general ):
+
+	PUSH_L( ESI )
+	PUSH_L( EDI )
+
+	MOV_L( ARG_SOURCE, ESI )
+	MOV_L( ARG_DEST, EDI )
+
+	MOV_L( ARG_MATRIX, EDX )
+	MOV_L( REGOFF(V4F_COUNT, ESI), ECX )
+
+	TEST_L( ECX, ECX )			/* verify non-zero count */
+	JE( LLBL( sse_general_done ) )
+
+	MOV_L( REGOFF(V4F_STRIDE, ESI), EAX )	/* stride */
+	OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) ) /* set dest flags */
+
+	MOV_L( ECX, REGOFF(V4F_COUNT, EDI) )	/* set dest count */
+	MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) )/* set dest size */
+
+	MOV_L( REGOFF(V4F_START, ESI), ESI )	/* ptr to first source vertex */
+	MOV_L( REGOFF(V4F_START, EDI), EDI )	/* ptr to first dest vertex */
+
+	PREFETCHT0( REGIND(ESI) )
+
+	MOVAPS( MAT(0), XMM4 )			/* m3  | m2  | m1  | m0  */
+	MOVAPS( MAT(4), XMM5 )			/* m7  | m6  | m5  | m4  */
+	MOVAPS( MAT(8), XMM6 )			/* m11 | m10 | m9  | m8  */
+	MOVAPS( MAT(12), XMM7 )			/* m15 | m14 | m13 | m12 */
+
+ALIGNTEXT16
+LLBL( sse_general_loop ):
+
+	MOVSS( SRC(0), XMM0 )			/* ox */
+	SHUFPS( CONST(0x0), XMM0, XMM0 )	/* ox | ox | ox | ox */
+	MULPS( XMM4, XMM0 )			/* ox*m3 | ox*m2 | ox*m1 | ox*m0 */
+
+	MOVSS( SRC(1), XMM1 )			/* oy */
+	SHUFPS( CONST(0x0), XMM1, XMM1 )	/* oy | oy | oy | oy */
+	MULPS( XMM5, XMM1 )			/* oy*m7 | oy*m6 | oy*m5 | oy*m4 */
+
+	MOVSS( SRC(2), XMM2 )			/* oz */
+	SHUFPS( CONST(0x0), XMM2, XMM2 )	/* oz | oz | oz | oz */
+	MULPS( XMM6, XMM2 )			/* oz*m11 | oz*m10 | oz*m9 | oz*m8 */
+
+	MOVSS( SRC(3), XMM3 )			/* ow */
+	SHUFPS( CONST(0x0), XMM3, XMM3 )	/* ow | ow | ow | ow */
+	MULPS( XMM7, XMM3 )			/* ow*m15 | ow*m14 | ow*m13 | ow*m12 */
+
+	ADDPS( XMM1, XMM0 )			/* ox*m3+oy*m7 | ... */
+	ADDPS( XMM2, XMM0 )			/* ox*m3+oy*m7+oz*m11 | ... */
+	ADDPS( XMM3, XMM0 )			/* ox*m3+oy*m7+oz*m11+ow*m15 | ... */
+	MOVAPS( XMM0, DST(0) )			/* ->D(3) | ->D(2) | ->D(1) | ->D(0) */
+
+	ADD_L( CONST(16), EDI )
+	ADD_L( EAX, ESI )
+
+	DEC_L( ECX )
+	JNZ( LLBL( sse_general_loop ) )
+
+LLBL( sse_general_done ):
+
+	POP_L( EDI )
+	POP_L( ESI )
+	RET
+
+
+
+
+ALIGNTEXT4
+GLOBL GLNAME( _mesa_sse_transform_points4_3d )
+GLNAME( _mesa_sse_transform_points4_3d ):
+
+	PUSH_L( ESI )
+	PUSH_L( EDI )
+
+	MOV_L( ARG_SOURCE, ESI )		/* ptr to source GLvector4f */
+	MOV_L( ARG_DEST, EDI )			/* ptr to dest GLvector4f */
+
+	MOV_L( ARG_MATRIX, EDX )		/* ptr to matrix */
+	MOV_L( REGOFF(V4F_COUNT, ESI), ECX )	/* source count */
+
+	TEST_L( ECX, ECX)
+	JZ( LLBL(K_GTP43P3DR_finish) )		/* count was zero; go to finish */
+
+	MOV_L( REGOFF(V4F_STRIDE, ESI), EAX )	/* stride */
+	OR_L( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, EDI) ) /* set dest flags */
+
+	MOV_L( ECX, REGOFF(V4F_COUNT, EDI) )	/* set dest count */
+	MOV_L( CONST(3), REGOFF(V4F_SIZE, EDI) )/* set dest size */
+
+	SHL_L( CONST(4), ECX )			/* count *= 16 */
+	MOV_L( REGOFF(V4F_START, ESI), ESI )	/* ptr to first source vertex */
+
+	MOV_L( REGOFF(V4F_START, EDI), EDI )	/* ptr to first dest vertex */
+	ADD_L( EDI, ECX )			/* count += dest ptr */
+
+	MOVAPS( MAT(0), XMM0 )			/* m3  | m2  | m1  |  m0 */
+	MOVAPS( MAT(4), XMM1 )			/* m7  | m6  | m5  |  m4 */
+	MOVAPS( MAT(8), XMM2 )			/* m11 | m10 | m9  |  m8 */
+	MOVAPS( MAT(12), XMM3 )			/* m15 | m14 | m13 | m12 */
+
+ALIGNTEXT32
+LLBL( K_GTP43P3DR_top ):
+	MOVSS( SRC(0), XMM4 )			/* ox */
+	SHUFPS( CONST(0x0), XMM4, XMM4 )	/* ox | ox | ox | ox */
+	MULPS( XMM0, XMM4 )			/* ox*m3 | ox*m2 | ox*m1 | ox*m0 */
+
+	MOVSS( SRC(1), XMM5 )			/* oy */
+	SHUFPS( CONST(0x0), XMM5, XMM5 )	/* oy | oy | oy | oy */
+	MULPS( XMM1, XMM5 )			/* oy*m7 | oy*m6 | oy*m5 | oy*m4 */
+
+	MOVSS( SRC(2), XMM6 )			/* oz */
+	SHUFPS( CONST(0x0), XMM6, XMM6 )	/* oz | oz | oz | oz */
+	MULPS( XMM2, XMM6 )			/* oz*m11 | oz*m10 | oz*m9 | oz*m8 */
+
+	MOVSS( SRC(3), XMM7 )			/* ow */
+	SHUFPS( CONST(0x0), XMM7, XMM7 )	/* ow | ow | ow | ow */
+	MULPS( XMM3, XMM7 )			/* ow*m15 | ow*m14 | ow*m13 | ow*m12 */
+
+	ADDPS( XMM5, XMM4 )			/* ox*m3+oy*m7 | ... */
+	ADDPS( XMM6, XMM4 )			/* ox*m3+oy*m7+oz*m11 | ... */
+	ADDPS( XMM7, XMM4 )			/* ox*m3+oy*m7+oz*m11+ow*m15 | ... */
+	MOVAPS( XMM4, DST(0) )			/* ->D(3) | ->D(2) | ->D(1) | ->D(0) */
+
+	MOVSS( SRC(3), XMM4 )			/* ow */
+	MOVSS( XMM4, DST(3) )			/* ->D(3) */
+
+LLBL( K_GTP43P3DR_skip ):
+	ADD_L( CONST(16), EDI )
+	ADD_L( EAX, ESI )
+	CMP_L( ECX, EDI )
+	JNE( LLBL(K_GTP43P3DR_top) )
+
+LLBL( K_GTP43P3DR_finish ):
+	POP_L( EDI )
+	POP_L( ESI )
+	RET
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_sse_transform_points4_identity )
+GLNAME( _mesa_sse_transform_points4_identity ):
+
+	PUSH_L( ESI )
+	PUSH_L( EDI )
+
+	MOV_L( ARG_SOURCE, ESI )
+	MOV_L( ARG_DEST, EDI )
+
+	MOV_L( ARG_MATRIX, EDX )
+	MOV_L( REGOFF(V4F_COUNT, ESI), ECX )
+
+	TEST_L( ECX, ECX )			/* verify non-zero count */
+	JE( LLBL( sse_identity_done ) )
+
+	MOV_L( REGOFF(V4F_STRIDE, ESI), EAX )	/* stride */
+	OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) ) /* set dest flags */
+
+	MOV_L( ECX, REGOFF(V4F_COUNT, EDI) )	/* set dest count */
+	MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) )/* set dest size */
+
+	MOV_L( REGOFF(V4F_START, ESI), ESI )	/* ptr to first source vertex */
+	MOV_L( REGOFF(V4F_START, EDI), EDI )	/* ptr to first dest vertex */
+
+ALIGNTEXT16
+LLBL( sse_identity_loop ):
+
+	PREFETCHNTA( REGOFF(32, ESI) )
+
+	MOVAPS( REGIND(ESI), XMM0 )
+	ADD_L( EAX, ESI )
+
+	MOVAPS( XMM0, REGIND(EDI) )
+	ADD_L( CONST(16), EDI )
+
+	DEC_L( ECX )
+	JNZ( LLBL( sse_identity_loop ) )
+
+LLBL( sse_identity_done ):
+
+	POP_L( EDI )
+	POP_L( ESI )
+	RET
diff --git a/src/mesa/x86/x86.c b/src/mesa/x86/x86.c
index 4f35a2ffc50..a51f181916d 100644
--- a/src/mesa/x86/x86.c
+++ b/src/mesa/x86/x86.c
@@ -1,4 +1,4 @@
-/* $Id: x86.c,v 1.19 2001/03/28 20:44:44 gareth Exp $ */
+/* $Id: x86.c,v 1.20 2001/03/29 06:46:27 gareth Exp $ */
 
 /*
  * Mesa 3-D graphics library
@@ -50,40 +50,37 @@
 			const GLubyte flag
 
 
-#define DECLARE_XFORM_GROUP( pfx, sz, masked ) \
-extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_general_##masked( XFORM_ARGS );	\
-extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_identity_##masked( XFORM_ARGS );	\
-extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_3d_no_rot_##masked( XFORM_ARGS );	\
-extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_perspective_##masked( XFORM_ARGS );	\
-extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_2d_##masked( XFORM_ARGS );		\
-extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_2d_no_rot_##masked( XFORM_ARGS );	\
-extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_3d_##masked( XFORM_ARGS );
-
-
-#define ASSIGN_XFORM_GROUP( pfx, cma, sz, masked )			\
-   _mesa_transform_tab[cma][sz][MATRIX_GENERAL] =			\
-      _mesa_##pfx##_transform_points##sz##_general_##masked;		\
-   _mesa_transform_tab[cma][sz][MATRIX_IDENTITY] =			\
-      _mesa_##pfx##_transform_points##sz##_identity_##masked;		\
-   _mesa_transform_tab[cma][sz][MATRIX_3D_NO_ROT] =			\
-      _mesa_##pfx##_transform_points##sz##_3d_no_rot_##masked;		\
-   _mesa_transform_tab[cma][sz][MATRIX_PERSPECTIVE] =			\
-      _mesa_##pfx##_transform_points##sz##_perspective_##masked;	\
-   _mesa_transform_tab[cma][sz][MATRIX_2D] =				\
-      _mesa_##pfx##_transform_points##sz##_2d_##masked;			\
-   _mesa_transform_tab[cma][sz][MATRIX_2D_NO_ROT] =			\
-      _mesa_##pfx##_transform_points##sz##_2d_no_rot_##masked;		\
-   _mesa_transform_tab[cma][sz][MATRIX_3D] =				\
-      _mesa_##pfx##_transform_points##sz##_3d_##masked;
+#define DECLARE_XFORM_GROUP( pfx, sz ) \
+extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_general( XFORM_ARGS );		\
+extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_identity( XFORM_ARGS );	\
+extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_3d_no_rot( XFORM_ARGS );	\
+extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_perspective( XFORM_ARGS );	\
+extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_2d( XFORM_ARGS );		\
+extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_2d_no_rot( XFORM_ARGS );	\
+extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_3d( XFORM_ARGS );
+
+
+#define ASSIGN_XFORM_GROUP( pfx, sz )					\
+   _mesa_transform_tab[0][sz][MATRIX_GENERAL] =				\
+      _mesa_##pfx##_transform_points##sz##_general;			\
+   _mesa_transform_tab[0][sz][MATRIX_IDENTITY] =			\
+      _mesa_##pfx##_transform_points##sz##_identity;			\
+   _mesa_transform_tab[0][sz][MATRIX_3D_NO_ROT] =			\
+      _mesa_##pfx##_transform_points##sz##_3d_no_rot;			\
+   _mesa_transform_tab[0][sz][MATRIX_PERSPECTIVE] =			\
+      _mesa_##pfx##_transform_points##sz##_perspective;			\
+   _mesa_transform_tab[0][sz][MATRIX_2D] =				\
+      _mesa_##pfx##_transform_points##sz##_2d;				\
+   _mesa_transform_tab[0][sz][MATRIX_2D_NO_ROT] =			\
+      _mesa_##pfx##_transform_points##sz##_2d_no_rot;			\
+   _mesa_transform_tab[0][sz][MATRIX_3D] =				\
+      _mesa_##pfx##_transform_points##sz##_3d;
 
 
 #ifdef USE_X86_ASM
-DECLARE_XFORM_GROUP( x86, 2, raw )
-DECLARE_XFORM_GROUP( x86, 3, raw )
-DECLARE_XFORM_GROUP( x86, 4, raw )
-DECLARE_XFORM_GROUP( x86, 2, masked )
-DECLARE_XFORM_GROUP( x86, 3, masked )
-DECLARE_XFORM_GROUP( x86, 4, masked )
+DECLARE_XFORM_GROUP( x86, 2 )
+DECLARE_XFORM_GROUP( x86, 3 )
+DECLARE_XFORM_GROUP( x86, 4 )
 
 
 extern GLvector4f * _ASMAPI
@@ -119,13 +116,9 @@ _mesa_v16_x86_general_xform( GLfloat *dest,
 void _mesa_init_x86_transform_asm( void )
 {
 #ifdef USE_X86_ASM
-   ASSIGN_XFORM_GROUP( x86, 0, 2, raw );
-   ASSIGN_XFORM_GROUP( x86, 0, 3, raw );
-   ASSIGN_XFORM_GROUP( x86, 0, 4, raw );
-
-/*     ASSIGN_XFORM_GROUP( x86, CULL_MASK_ACTIVE, 2, masked ); */
-/*     ASSIGN_XFORM_GROUP( x86, CULL_MASK_ACTIVE, 3, masked ); */
-/*     ASSIGN_XFORM_GROUP( x86, CULL_MASK_ACTIVE, 4, masked ); */
+   ASSIGN_XFORM_GROUP( x86, 2 );
+   ASSIGN_XFORM_GROUP( x86, 3 );
+   ASSIGN_XFORM_GROUP( x86, 4 );
 
    /* XXX this function has been found to cause FP overflow exceptions */
    _mesa_clip_tab[4] = _mesa_x86_cliptest_points4;
diff --git a/src/mesa/x86/x86_xform2.S b/src/mesa/x86/x86_xform2.S
new file mode 100644
index 00000000000..0e46a8411ec
--- /dev/null
+++ b/src/mesa/x86/x86_xform2.S
@@ -0,0 +1,536 @@
+/* $Id: x86_xform2.S,v 1.1 2001/03/29 06:46:27 gareth Exp $ */
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  3.5
+ *
+ * Copyright (C) 1999-2001  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.
+ */
+
+#include "matypes.h"
+#include "xform_args.h"
+
+	SEG_TEXT
+
+#define FP_ONE		1065353216
+#define FP_ZERO		0
+
+#define SRC(i)		REGOFF(i * 4, ESI)
+#define DST(i)		REGOFF(i * 4, EDI)
+#define MAT(i)		REGOFF(i * 4, EDX)
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_x86_transform_points2_general )
+GLNAME( _mesa_x86_transform_points2_general ):
+
+#define FRAME_OFFSET 8
+	PUSH_L( ESI )
+	PUSH_L( EDI )
+
+	MOV_L( ARG_SOURCE, ESI )
+	MOV_L( ARG_DEST, EDI )
+
+	MOV_L( ARG_MATRIX, EDX )
+	MOV_L( REGOFF(V4F_COUNT, ESI), ECX )
+
+	TEST_L( ECX, ECX )
+	JZ( LLBL( x86_p2_gr_done ) )
+
+	MOV_L( REGOFF(V4F_STRIDE, ESI), EAX )
+	OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) )
+
+	MOV_L( ECX, REGOFF(V4F_COUNT, EDI) )
+	MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) )
+
+	SHL_L( CONST(4), ECX )
+	MOV_L( REGOFF(V4F_START, ESI), ESI )
+
+	MOV_L( REGOFF(V4F_START, EDI), EDI )
+	ADD_L( EDI, ECX )
+
+ALIGNTEXT16
+LLBL( x86_p2_gr_loop ):
+
+	FLD_S( SRC(0) )			/* F4 */
+	FMUL_S( MAT(0) )
+	FLD_S( SRC(0) )			/* F5 F4 */
+	FMUL_S( MAT(1) )
+	FLD_S( SRC(0) )			/* F6 F5 F4 */
+	FMUL_S( MAT(2) )
+	FLD_S( SRC(0) )			/* F7 F6 F5 F4 */
+	FMUL_S( MAT(3) )
+
+	FLD_S( SRC(1) )			/* F0 F7 F6 F5 F4 */
+	FMUL_S( MAT(4) )
+	FLD_S( SRC(1) )			/* F1 F0 F7 F6 F5 F4 */
+	FMUL_S( MAT(5) )
+	FLD_S( SRC(1) )			/* F2 F1 F0 F7 F6 F5 F4 */
+	FMUL_S( MAT(6) )
+	FLD_S( SRC(1) )			/* F3 F2 F1 F0 F7 F6 F5 F4 */
+	FMUL_S( MAT(7) )
+
+	FXCH( ST(3) )			/* F0 F2 F1 F3 F7 F6 F5 F4 */
+	FADDP( ST(0), ST(7) )		/* F2 F1 F3 F7 F6 F5 F4 */
+	FXCH( ST(1) )			/* F1 F2 F3 F7 F6 F5 F4 */
+	FADDP( ST(0), ST(5) )		/* F2 F3 F7 F6 F5 F4 */
+	FADDP( ST(0), ST(3) )		/* F3 F7 F6 F5 F4 */
+	FADDP( ST(0), ST(1) )		/* F7 F6 F5 F4 */
+
+	FXCH( ST(3) )			/* F4 F6 F5 F7 */
+	FADD_S( MAT(12) )
+	FXCH( ST(2) )			/* F5 F6 F4 F7 */
+	FADD_S( MAT(13) )
+	FXCH( ST(1) )			/* F6 F5 F4 F7 */
+	FADD_S( MAT(14) )
+	FXCH( ST(3) )			/* F7 F5 F4 F6 */
+	FADD_S( MAT(15) )
+
+	FXCH( ST(2) )			/* F4 F5 F7 F6 */
+	FSTP_S( DST(0) )		/* F5 F7 F6 */
+	FSTP_S( DST(1) )		/* F7 F6 */
+	FXCH( ST(1) )			/* F6 F7 */
+	FSTP_S( DST(2) )		/* F7 */
+	FSTP_S( DST(3) )		/* */
+
+LLBL( x86_p2_gr_skip ):
+
+	ADD_L( CONST(16), EDI )
+	ADD_L( EAX, ESI )
+	CMP_L( ECX, EDI )
+	JNE( LLBL( x86_p2_gr_loop ) )
+
+LLBL( x86_p2_gr_done ):
+
+	POP_L( EDI )
+	POP_L( ESI )
+	RET
+#undef FRAME_OFFSET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_x86_transform_points2_perspective )
+GLNAME( _mesa_x86_transform_points2_perspective ):
+
+#define FRAME_OFFSET 12
+	PUSH_L( ESI )
+	PUSH_L( EDI )
+	PUSH_L( EBX )
+
+	MOV_L( ARG_SOURCE, ESI )
+	MOV_L( ARG_DEST, EDI )
+
+	MOV_L( ARG_MATRIX, EDX )
+	MOV_L( REGOFF(V4F_COUNT, ESI), ECX )
+
+	TEST_L( ECX, ECX )
+	JZ( LLBL( x86_p2_pr_done ) )
+
+	MOV_L( REGOFF(V4F_STRIDE, ESI), EAX )
+	OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) )
+
+	MOV_L( ECX, REGOFF(V4F_COUNT, EDI) )
+	MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) )
+
+	SHL_L( CONST(4), ECX )
+	MOV_L( REGOFF(V4F_START, ESI), ESI )
+
+	MOV_L( REGOFF(V4F_START, EDI), EDI )
+	ADD_L( EDI, ECX )
+
+	MOV_L( MAT(14), EBX )
+
+ALIGNTEXT16
+LLBL( x86_p2_pr_loop ):
+
+	FLD_S( SRC(0) )			/* F4 */
+	FMUL_S( MAT(0) )
+
+	FLD_S( SRC(1) )			/* F1 F4 */
+	FMUL_S( MAT(5) )
+
+	FXCH( ST(1) )			/* F4 F1 */
+	FSTP_S( DST(0)   )		/* F1 */
+	FSTP_S( DST(1)   )		/* */
+	MOV_L( EBX, DST(2) )
+	MOV_L( CONST(FP_ZERO), DST(3) )
+
+LLBL( x86_p2_pr_skip ):
+
+	ADD_L( CONST(16), EDI )
+	ADD_L( EAX, ESI )
+	CMP_L( ECX, EDI )
+	JNE( LLBL( x86_p2_pr_loop ) )
+
+LLBL( x86_p2_pr_done ):
+
+	POP_L( EBX )
+	POP_L( EDI )
+	POP_L( ESI )
+	RET
+#undef FRAME_OFFSET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_x86_transform_points2_3d )
+GLNAME( _mesa_x86_transform_points2_3d ):
+
+#define FRAME_OFFSET 8
+	PUSH_L( ESI )
+	PUSH_L( EDI )
+
+	MOV_L( ARG_SOURCE, ESI )
+	MOV_L( ARG_DEST, EDI )
+
+	MOV_L( ARG_MATRIX, EDX )
+	MOV_L( REGOFF(V4F_COUNT, ESI), ECX )
+
+	TEST_L( ECX, ECX )
+	JZ( LLBL( x86_p2_3dr_done ) )
+
+	MOV_L( REGOFF(V4F_STRIDE, ESI), EAX )
+	OR_L( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, EDI) )
+
+	MOV_L( ECX, REGOFF(V4F_COUNT, EDI) )
+	MOV_L( CONST(3), REGOFF(V4F_SIZE, EDI) )
+
+	SHL_L( CONST(4), ECX )
+	MOV_L( REGOFF(V4F_START, ESI), ESI )
+
+	MOV_L( REGOFF(V4F_START, EDI), EDI )
+	ADD_L( EDI, ECX )
+
+ALIGNTEXT16
+LLBL( x86_p2_3dr_loop ):
+
+	FLD_S( SRC(0) )			/* F4 */
+	FMUL_S( MAT(0) )
+	FLD_S( SRC(0) )			/* F5 F4 */
+	FMUL_S( MAT(1) )
+	FLD_S( SRC(0) )			/* F6 F5 F4 */
+	FMUL_S( MAT(2) )
+
+	FLD_S( SRC(1) )			/* F0 F6 F5 F4 */
+	FMUL_S( MAT(4) )
+	FLD_S( SRC(1) )			/* F1 F0 F6 F5 F4 */
+	FMUL_S( MAT(5) )
+	FLD_S( SRC(1) )			/* F2 F1 F0 F6 F5 F4 */
+	FMUL_S( MAT(6) )
+
+	FXCH( ST(2) )			/* F0 F1 F2 F6 F5 F4 */
+	FADDP( ST(0), ST(5) )		/* F1 F2 F6 F5 F4 */
+	FADDP( ST(0), ST(3) )		/* F2 F6 F5 F4 */
+	FADDP( ST(0), ST(1) )		/* F6 F5 F4 */
+
+	FXCH( ST(2) )			/* F4 F5 F6 */
+	FADD_S( MAT(12) )
+	FXCH( ST(1) )			/* F5 F4 F6 */
+	FADD_S( MAT(13) )
+	FXCH( ST(2) )			/* F6 F4 F5 */
+	FADD_S( MAT(14) )
+
+	FXCH( ST(1) )			/* F4 F6 F5 */
+	FSTP_S( DST(0) )		/* F6 F5 */
+	FXCH( ST(1) )			/* F5 F6 */
+	FSTP_S( DST(1) )		/* F6 */
+	FSTP_S( DST(2) )		/* */
+
+LLBL( x86_p2_3dr_skip ):
+
+	ADD_L( CONST(16), EDI )
+	ADD_L( EAX, ESI )
+	CMP_L( ECX, EDI )
+	JNE( LLBL( x86_p2_3dr_loop ) )
+
+LLBL( x86_p2_3dr_done ):
+
+	POP_L( EDI )
+	POP_L( ESI )
+	RET
+#undef FRAME_OFFSET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_x86_transform_points2_3d_no_rot )
+GLNAME( _mesa_x86_transform_points2_3d_no_rot ):
+
+#define FRAME_OFFSET 12
+	PUSH_L( ESI )
+	PUSH_L( EDI )
+	PUSH_L( EBX )
+
+	MOV_L( ARG_SOURCE, ESI )
+	MOV_L( ARG_DEST, EDI )
+
+	MOV_L( ARG_MATRIX, EDX )
+	MOV_L( REGOFF(V4F_COUNT, ESI), ECX )
+
+	TEST_L( ECX, ECX )
+	JZ( LLBL( x86_p2_3dnrr_done ) )
+
+	MOV_L( REGOFF(V4F_STRIDE, ESI), EAX )
+	OR_L( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, EDI) )
+
+	MOV_L( ECX, REGOFF(V4F_COUNT, EDI) )
+	MOV_L( CONST(3), REGOFF(V4F_SIZE, EDI) )
+
+	SHL_L( CONST(4), ECX )
+	MOV_L( REGOFF(V4F_START, ESI), ESI )
+
+	MOV_L( REGOFF(V4F_START, EDI), EDI )
+	ADD_L( EDI, ECX )
+
+	MOV_L( MAT(14), EBX )
+
+ALIGNTEXT16
+LLBL( x86_p2_3dnrr_loop ):
+
+	FLD_S( SRC(0) )			/* F4 */
+	FMUL_S( MAT(0) )
+
+	FLD_S( SRC(1) )			/* F1 F4 */
+	FMUL_S( MAT(5) )
+
+	FXCH( ST(1) )			/* F4 F1 */
+	FADD_S( MAT(12) )
+	FLD_S( MAT(13) )		/* F5 F4 F1 */
+	FXCH( ST(2) )			/* F1 F4 F5 */
+	FADDP( ST(0), ST(2) )		/* F4 F5 */
+
+	FSTP_S( DST(0) )		/* F5 */
+	FSTP_S( DST(1) )		/* */
+	MOV_L( EBX, DST(2) )
+
+LLBL( x86_p2_3dnrr_skip ):
+
+	ADD_L( CONST(16), EDI )
+	ADD_L( EAX, ESI )
+	CMP_L( ECX, EDI )
+	JNE( LLBL( x86_p2_3dnrr_loop ) )
+
+LLBL( x86_p2_3dnrr_done ):
+
+	POP_L( EBX )
+	POP_L( EDI )
+	POP_L( ESI )
+	RET
+#undef FRAME_OFFSET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_x86_transform_points2_2d )
+GLNAME( _mesa_x86_transform_points2_2d ):
+
+#define FRAME_OFFSET 8
+	PUSH_L( ESI )
+	PUSH_L( EDI )
+
+	MOV_L( ARG_SOURCE, ESI )
+	MOV_L( ARG_DEST, EDI )
+
+	MOV_L( ARG_MATRIX, EDX )
+	MOV_L( REGOFF(V4F_COUNT, ESI), ECX )
+
+	TEST_L( ECX, ECX )
+	JZ( LLBL( x86_p2_2dr_done ) )
+
+	MOV_L( REGOFF(V4F_STRIDE, ESI), EAX )
+	OR_L( CONST(VEC_SIZE_2), REGOFF(V4F_FLAGS, EDI) )
+
+	MOV_L( ECX, REGOFF(V4F_COUNT, EDI) )
+	MOV_L( CONST(2), REGOFF(V4F_SIZE, EDI) )
+
+	SHL_L( CONST(4), ECX )
+	MOV_L( REGOFF(V4F_START, ESI), ESI )
+
+	MOV_L( REGOFF(V4F_START, EDI), EDI )
+	ADD_L( EDI, ECX )
+
+ALIGNTEXT16
+LLBL( x86_p2_2dr_loop ):
+
+	FLD_S( SRC(0) )			/* F4 */
+	FMUL_S( MAT(0) )
+	FLD_S( SRC(0) )			/* F5 F4 */
+	FMUL_S( MAT(1) )
+
+	FLD_S( SRC(1) )			/* F0 F5 F4 */
+	FMUL_S( MAT(4) )
+	FLD_S( SRC(1) )			/* F1 F0 F5 F4 */
+	FMUL_S( MAT(5) )
+
+	FXCH( ST(1) )			/* F0 F1 F5 F4 */
+	FADDP( ST(0), ST(3) )		/* F1 F5 F4 */
+	FADDP( ST(0), ST(1) )		/* F5 F4 */
+
+	FXCH( ST(1) )			/* F4 F5 */
+	FADD_S( MAT(12) )
+	FXCH( ST(1) )			/* F5 F4 */
+	FADD_S( MAT(13) )
+
+	FXCH( ST(1) )			/* F4 F5 */
+	FSTP_S( DST(0) )		/* F5 */
+	FSTP_S( DST(1) )		/* */
+
+LLBL( x86_p2_2dr_skip ):
+
+	ADD_L( CONST(16), EDI )
+	ADD_L( EAX, ESI )
+	CMP_L( ECX, EDI )
+	JNE( LLBL( x86_p2_2dr_loop ) )
+
+LLBL( x86_p2_2dr_done ):
+
+	POP_L( EDI )
+	POP_L( ESI )
+	RET
+#undef FRAME_OFFSET
+
+
+
+
+ALIGNTEXT4
+GLOBL GLNAME( _mesa_x86_transform_points2_2d_no_rot )
+GLNAME( _mesa_x86_transform_points2_2d_no_rot ):
+
+#define FRAME_OFFSET 8
+	PUSH_L( ESI )
+	PUSH_L( EDI )
+
+	MOV_L( ARG_SOURCE, ESI )
+	MOV_L( ARG_DEST, EDI )
+
+	MOV_L( ARG_MATRIX, EDX )
+	MOV_L( REGOFF(V4F_COUNT, ESI), ECX )
+
+	TEST_L( ECX, ECX )
+	JZ( LLBL( x86_p2_2dnrr_done ) )
+
+	MOV_L( REGOFF(V4F_STRIDE, ESI), EAX )
+	OR_L( CONST(VEC_SIZE_2), REGOFF(V4F_FLAGS, EDI) )
+
+	MOV_L( ECX, REGOFF(V4F_COUNT, EDI) )
+	MOV_L( CONST(2), REGOFF(V4F_SIZE, EDI) )
+
+	SHL_L( CONST(4), ECX )
+	MOV_L( REGOFF(V4F_START, ESI), ESI )
+
+	MOV_L( REGOFF(V4F_START, EDI), EDI )
+	ADD_L( EDI, ECX )
+
+ALIGNTEXT16
+LLBL( x86_p2_2dnrr_loop ):
+
+	FLD_S( SRC(0) )			/* F4 */
+	FMUL_S( MAT(0) )
+
+	FLD_S( SRC(1) )			/* F1 F4 */
+	FMUL_S( MAT(5) )
+
+	FXCH( ST(1) )			/* F4 F1 */
+	FADD_S( MAT(12) )
+	FLD_S( MAT(13) )		/* F5 F4 F1 */
+	FXCH( ST(2) )			/* F1 F4 F5 */
+	FADDP( ST(0), ST(2) )		/* F4 F5 */
+
+	FSTP_S( DST(0)   )		/* F5 */
+	FSTP_S( DST(1)   )		/* */
+
+LLBL( x86_p2_2dnrr_skip ):
+
+	ADD_L( CONST(16), EDI )
+	ADD_L( EAX, ESI )
+	CMP_L( ECX, EDI )
+	JNE( LLBL( x86_p2_2dnrr_loop ) )
+
+LLBL( x86_p2_2dnrr_done ):
+
+	POP_L( EDI )
+	POP_L( ESI )
+	RET
+#undef FRAME_OFFSET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_x86_transform_points2_identity )
+GLNAME( _mesa_x86_transform_points2_identity ):
+
+#define FRAME_OFFSET 12
+	PUSH_L( ESI )
+	PUSH_L( EDI )
+	PUSH_L( EBX )
+
+	MOV_L( ARG_SOURCE, ESI )
+	MOV_L( ARG_DEST, EDI )
+
+	MOV_L( ARG_MATRIX, EDX )
+	MOV_L( REGOFF(V4F_COUNT, ESI), ECX )
+
+	TEST_L( ECX, ECX )
+	JZ( LLBL( x86_p2_ir_done ) )
+
+	MOV_L( REGOFF(V4F_STRIDE, ESI), EAX )
+	OR_L( CONST(VEC_SIZE_2), REGOFF(V4F_FLAGS, EDI) )
+
+	MOV_L( ECX, REGOFF(V4F_COUNT, EDI) )
+	MOV_L( CONST(2), REGOFF(V4F_SIZE, EDI) )
+
+	SHL_L( CONST(4), ECX )
+	MOV_L( REGOFF(V4F_START, ESI), ESI )
+
+	MOV_L( REGOFF(V4F_START, EDI), EDI )
+	ADD_L( EDI, ECX )
+
+	CMP_L( ESI, EDI )
+	JE( LLBL( x86_p2_ir_done ) )
+
+ALIGNTEXT16
+LLBL( x86_p2_ir_loop ):
+
+	MOV_L( SRC(0), EBX )
+	MOV_L( SRC(1), EDX )
+
+	MOV_L( EBX, DST(0) )
+	MOV_L( EDX, DST(1) )
+
+LLBL( x86_p2_ir_skip ):
+
+	ADD_L( CONST(16), EDI )
+	ADD_L( EAX, ESI )
+	CMP_L( ECX, EDI )
+	JNE( LLBL( x86_p2_ir_loop ) )
+
+LLBL( x86_p2_ir_done ):
+
+	POP_L( EBX )
+	POP_L( EDI )
+	POP_L( ESI )
+	RET
+#undef FRAME_OFFSET
diff --git a/src/mesa/x86/x86_xform3.S b/src/mesa/x86/x86_xform3.S
new file mode 100644
index 00000000000..1f798744e81
--- /dev/null
+++ b/src/mesa/x86/x86_xform3.S
@@ -0,0 +1,606 @@
+/* $Id: x86_xform3.S,v 1.1 2001/03/29 06:46:27 gareth Exp $ */
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  3.5
+ *
+ * Copyright (C) 1999-2001  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.
+ */
+
+#include "matypes.h"
+#include "xform_args.h"
+
+	SEG_TEXT
+
+#define FP_ONE		1065353216
+#define FP_ZERO		0
+
+#define SRC(i)		REGOFF(i * 4, ESI)
+#define DST(i)		REGOFF(i * 4, EDI)
+#define MAT(i)		REGOFF(i * 4, EDX)
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_x86_transform_points3_general )
+GLNAME( _mesa_x86_transform_points3_general ):
+
+#define FRAME_OFFSET 8
+	PUSH_L( ESI )
+	PUSH_L( EDI )
+
+	MOV_L( ARG_SOURCE, ESI )
+	MOV_L( ARG_DEST, EDI )
+
+	MOV_L( ARG_MATRIX, EDX )
+	MOV_L( REGOFF(V4F_COUNT, ESI), ECX )
+
+	TEST_L( ECX, ECX )
+	JZ( LLBL( x86_p3_gr_done ) )
+
+	MOV_L( REGOFF(V4F_STRIDE, ESI), EAX )
+	OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) )
+
+	MOV_L( ECX, REGOFF(V4F_COUNT, EDI) )
+	MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) )
+
+	SHL_L( CONST(4), ECX )
+	MOV_L( REGOFF(V4F_START, ESI), ESI )
+
+	MOV_L( REGOFF(V4F_START, EDI), EDI )
+	ADD_L( EDI, ECX )
+
+ALIGNTEXT16
+LLBL( x86_p3_gr_loop ):
+
+	FLD_S( SRC(0) )			/* F4 */
+	FMUL_S( MAT(0) )
+	FLD_S( SRC(0) )			/* F5 F4 */
+	FMUL_S( MAT(1) )
+	FLD_S( SRC(0) )			/* F6 F5 F4 */
+	FMUL_S( MAT(2) )
+	FLD_S( SRC(0) )			/* F7 F6 F5 F4 */
+	FMUL_S( MAT(3) )
+
+	FLD_S( SRC(1) )			/* F0 F7 F6 F5 F4 */
+	FMUL_S( MAT(4) )
+	FLD_S( SRC(1) )			/* F1 F0 F7 F6 F5 F4 */
+	FMUL_S( MAT(5) )
+	FLD_S( SRC(1) )			/* F2 F1 F0 F7 F6 F5 F4 */
+	FMUL_S( MAT(6) )
+	FLD_S( SRC(1) )			/* F3 F2 F1 F0 F7 F6 F5 F4 */
+	FMUL_S( MAT(7) )
+
+	FXCH( ST(3) )			/* F0 F2 F1 F3 F7 F6 F5 F4 */
+	FADDP( ST(0), ST(7) )		/* F2 F1 F3 F7 F6 F5 F4 */
+	FXCH( ST(1) )			/* F1 F2 F3 F7 F6 F5 F4 */
+	FADDP( ST(0), ST(5) )		/* F2 F3 F7 F6 F5 F4 */
+	FADDP( ST(0), ST(3) )		/* F3 F7 F6 F5 F4 */
+	FADDP( ST(0), ST(1) )		/* F7 F6 F5 F4 */
+
+	FLD_S( SRC(2) )			/* F0 F7 F6 F5 F4 */
+	FMUL_S( MAT(8) )
+	FLD_S( SRC(2) )			/* F1 F0 F7 F6 F5 F4 */
+	FMUL_S( MAT(9) )
+	FLD_S( SRC(2) )			/* F2 F1 F0 F7 F6 F5 F4 */
+	FMUL_S( MAT(10) )
+	FLD_S( SRC(2) )			/* F3 F2 F1 F0 F7 F6 F5 F4 */
+	FMUL_S( MAT(11) )
+
+	FXCH( ST(3) )			/* F0 F2 F1 F3 F7 F6 F5 F4 */
+	FADDP( ST(0), ST(7) )		/* F2 F1 F3 F7 F6 F5 F4 */
+	FXCH( ST(1) )			/* F1 F2 F3 F7 F6 F5 F4 */
+	FADDP( ST(0), ST(5) )		/* F2 F3 F7 F6 F5 F4 */
+	FADDP( ST(0), ST(3) )		/* F3 F7 F6 F5 F4 */
+	FADDP( ST(0), ST(1) )		/* F7 F6 F5 F4 */
+
+	FXCH( ST(3) )			/* F4 F6 F5 F7 */
+	FADD_S( MAT(12) )
+	FXCH( ST(2) )			/* F5 F6 F4 F7 */
+	FADD_S( MAT(13) )
+	FXCH( ST(1) )			/* F6 F5 F4 F7 */
+	FADD_S( MAT(14) )
+	FXCH( ST(3) )			/* F7 F5 F4 F6 */
+	FADD_S( MAT(15) )
+
+	FXCH( ST(2) )			/* F4 F5 F7 F6 */
+	FSTP_S( DST(0) )		/* F5 F7 F6 */
+	FSTP_S( DST(1) )		/* F7 F6 */
+	FXCH( ST(1) )			/* F6 F7 */
+	FSTP_S( DST(2) )		/* F7 */
+	FSTP_S( DST(3) )		/* */
+
+LLBL( x86_p3_gr_skip ):
+
+	ADD_L( CONST(16), EDI )
+	ADD_L( EAX, ESI )
+	CMP_L( ECX, EDI )
+	JNE( LLBL( x86_p3_gr_loop ) )
+
+LLBL( x86_p3_gr_done ):
+
+	POP_L( EDI )
+	POP_L( ESI )
+	RET
+#undef FRAME_OFFSET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_x86_transform_points3_perspective )
+GLNAME( _mesa_x86_transform_points3_perspective ):
+
+#define FRAME_OFFSET 12
+	PUSH_L( ESI )
+	PUSH_L( EDI )
+	PUSH_L( EBX )
+
+	MOV_L( ARG_SOURCE, ESI )
+	MOV_L( ARG_DEST, EDI )
+
+	MOV_L( ARG_MATRIX, EDX )
+	MOV_L( REGOFF(V4F_COUNT, ESI), ECX )
+
+	TEST_L( ECX, ECX )
+	JZ( LLBL( x86_p3_pr_done ) )
+
+	MOV_L( REGOFF(V4F_STRIDE, ESI), EAX )
+	OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) )
+
+	MOV_L( ECX, REGOFF(V4F_COUNT, EDI) )
+	MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) )
+
+	SHL_L( CONST(4), ECX )
+	MOV_L( REGOFF(V4F_START, ESI), ESI )
+
+	MOV_L( REGOFF(V4F_START, EDI), EDI )
+	ADD_L( EDI, ECX )
+
+ALIGNTEXT16
+LLBL( x86_p3_pr_loop ):
+
+	FLD_S( SRC(0) )			/* F4 */
+	FMUL_S( MAT(0) )
+
+	FLD_S( SRC(1) )			/* F5 F4 */
+	FMUL_S( MAT(5) )
+
+	FLD_S( SRC(2) )			/* F0 F5 F4 */
+	FMUL_S( MAT(8) )
+	FLD_S( SRC(2) )			/* F1 F0 F5 F4 */
+	FMUL_S( MAT(9) )
+	FLD_S( SRC(2) )			/* F2 F1 F0 F5 F4 */
+	FMUL_S( MAT(10) )
+
+	FXCH( ST(2) )			/* F0 F1 F2 F5 F4 */
+	FADDP( ST(0), ST(4) )		/* F1 F2 F5 F4 */
+	FADDP( ST(0), ST(2) )		/* F2 F5 F4 */
+	FLD_S( MAT(14) )		/* F6 F2 F5 F4 */
+	FXCH( ST(1) )			/* F2 F6 F5 F4 */
+	FADDP( ST(0), ST(1) )		/* F6 F5 F4 */
+
+	MOV_L( SRC(2), EBX )
+	XOR_L( CONST(-2147483648), EBX )/* change sign */
+
+	FXCH( ST(2) )			/* F4 F5 F6 */
+	FSTP_S( DST(0) )		/* F5 F6 */
+	FSTP_S( DST(1) )		/* F6 */
+	FSTP_S( DST(2) )		/* */
+	MOV_L( EBX, DST(3) )
+
+LLBL( x86_p3_pr_skip ):
+
+	ADD_L( CONST(16), EDI )
+	ADD_L( EAX, ESI )
+	CMP_L( ECX, EDI )
+	JNE( LLBL( x86_p3_pr_loop ) )
+
+LLBL( x86_p3_pr_done ):
+
+	POP_L( EBX )
+	POP_L( EDI )
+	POP_L( ESI )
+	RET
+#undef FRAME_OFFSET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_x86_transform_points3_3d )
+GLNAME( _mesa_x86_transform_points3_3d ):
+
+#define FRAME_OFFSET 8
+	PUSH_L( ESI )
+	PUSH_L( EDI )
+
+	MOV_L( ARG_SOURCE, ESI )
+	MOV_L( ARG_DEST, EDI )
+
+	MOV_L( ARG_MATRIX, EDX )
+	MOV_L( REGOFF(V4F_COUNT, ESI), ECX )
+
+	TEST_L( ECX, ECX )
+	JZ( LLBL( x86_p3_3dr_done ) )
+
+	MOV_L( REGOFF(V4F_STRIDE, ESI), EAX )
+	OR_L( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, EDI) )
+
+	MOV_L( ECX, REGOFF(V4F_COUNT, EDI) )
+	MOV_L( CONST(3), REGOFF(V4F_SIZE, EDI) )
+
+	SHL_L( CONST(4), ECX )
+	MOV_L( REGOFF(V4F_START, ESI), ESI )
+
+	MOV_L( REGOFF(V4F_START, EDI), EDI )
+	ADD_L( EDI, ECX )
+
+ALIGNTEXT16
+LLBL( x86_p3_3dr_loop ):
+
+	FLD_S( SRC(0) )			/* F4 */
+	FMUL_S( MAT(0) )
+	FLD_S( SRC(0) )			/* F5 F4 */
+	FMUL_S( MAT(1) )
+	FLD_S( SRC(0) )			/* F6 F5 F4 */
+	FMUL_S( MAT(2) )
+
+	FLD_S( SRC(1) )			/* F0 F6 F5 F4 */
+	FMUL_S( MAT(4) )
+	FLD_S( SRC(1) )			/* F1 F0 F6 F5 F4 */
+	FMUL_S( MAT(5) )
+	FLD_S( SRC(1) )			/* F2 F1 F0 F6 F5 F4 */
+	FMUL_S( MAT(6) )
+
+	FXCH( ST(2) )			/* F0 F1 F2 F6 F5 F4 */
+	FADDP( ST(0), ST(5) )		/* F1 F2 F6 F5 F4 */
+	FADDP( ST(0), ST(3) )		/* F2 F6 F5 F4 */
+	FADDP( ST(0), ST(1) )		/* F6 F5 F4 */
+
+	FLD_S( SRC(2) )			/* F0 F6 F5 F4 */
+	FMUL_S( MAT(8) )
+	FLD_S( SRC(2) )			/* F1 F0 F6 F5 F4 */
+	FMUL_S( MAT(9) )
+	FLD_S( SRC(2) )			/* F2 F1 F0 F6 F5 F4 */
+	FMUL_S( MAT(10) )
+
+	FXCH( ST(2) )			/* F0 F1 F2 F6 F5 F4 */
+	FADDP( ST(0), ST(5) )		/* F1 F2 F6 F5 F4 */
+	FADDP( ST(0), ST(3) )		/* F2 F6 F5 F4 */
+	FADDP( ST(0), ST(1) )		/* F6 F5 F4 */
+
+	FXCH( ST(2) )			/* F4 F5 F6 */
+	FADD_S( MAT(12) )
+	FXCH( ST(1) )			/* F5 F4 F6 */
+	FADD_S( MAT(13) )
+	FXCH( ST(2) )			/* F6 F4 F5 */
+	FADD_S( MAT(14) )
+
+	FXCH( ST(1) )			/* F4 F6 F5 */
+	FSTP_S( DST(0)   )		/* F6 F5 */
+	FXCH( ST(1) )			/* F5 F6 */
+	FSTP_S( DST(1)   )		/* F6 */
+	FSTP_S( DST(2)   )		/* */
+
+LLBL( x86_p3_3dr_skip ):
+
+	ADD_L( CONST(16), EDI )
+	ADD_L( EAX, ESI )
+	CMP_L( ECX, EDI )
+	JNE( LLBL( x86_p3_3dr_loop ) )
+
+LLBL( x86_p3_3dr_done ):
+
+	POP_L( EDI )
+	POP_L( ESI )
+	RET
+#undef FRAME_OFFSET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_x86_transform_points3_3d_no_rot )
+GLNAME( _mesa_x86_transform_points3_3d_no_rot ):
+
+#define FRAME_OFFSET 8
+	PUSH_L( ESI )
+	PUSH_L( EDI )
+
+	MOV_L( ARG_SOURCE, ESI )
+	MOV_L( ARG_DEST, EDI )
+
+
+	MOV_L( ARG_MATRIX, EDX )
+	MOV_L( REGOFF(V4F_COUNT, ESI), ECX )
+
+	TEST_L( ECX, ECX )
+	JZ( LLBL( x86_p3_3dnrr_done ) )
+
+	MOV_L( REGOFF(V4F_STRIDE, ESI), EAX )
+	OR_L( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, EDI) )
+
+	MOV_L( ECX, REGOFF(V4F_COUNT, EDI) )
+	MOV_L( CONST(3), REGOFF(V4F_SIZE, EDI) )
+
+	SHL_L( CONST(4), ECX )
+	MOV_L( REGOFF(V4F_START, ESI), ESI )
+
+	MOV_L( REGOFF(V4F_START, EDI), EDI )
+	ADD_L( EDI, ECX )
+
+ALIGNTEXT16
+LLBL( x86_p3_3dnrr_loop ):
+
+	FLD_S( SRC(0) )			/* F4 */
+	FMUL_S( MAT(0) )
+
+	FLD_S( SRC(1) )			/* F1 F4 */
+	FMUL_S( MAT(5) )
+
+	FLD_S( SRC(2) )			/* F2 F1 F4 */
+	FMUL_S( MAT(10) )
+
+	FXCH( ST(2) )			/* F4 F1 F2 */
+	FADD_S( MAT(12) )
+	FLD_S( MAT(13) )		/* F5 F4 F1 F2 */
+	FXCH( ST(2) )			/* F1 F4 F5 F2 */
+	FADDP( ST(0), ST(2) )		/* F4 F5 F2 */
+	FLD_S( MAT(14) )		/* F6 F4 F5 F2 */
+	FXCH( ST(3) )			/* F2 F4 F5 F6 */
+	FADDP( ST(0), ST(3) )		/* F4 F5 F6 */
+
+	FSTP_S( DST(0)   )		/* F5 F6 */
+	FSTP_S( DST(1)   )		/* F6 */
+	FSTP_S( DST(2)   )		/* */
+
+LLBL( x86_p3_3dnrr_skip ):
+
+	ADD_L( CONST(16), EDI )
+	ADD_L( EAX, ESI )
+	CMP_L( ECX, EDI )
+	JNE( LLBL( x86_p3_3dnrr_loop ) )
+
+LLBL( x86_p3_3dnrr_done ):
+
+	POP_L( EDI )
+	POP_L( ESI )
+	RET
+#undef FRAME_OFFSET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_x86_transform_points3_2d )
+GLNAME( _mesa_x86_transform_points3_2d ):
+
+#define FRAME_OFFSET 12
+	PUSH_L( ESI )
+	PUSH_L( EDI )
+	PUSH_L( EBX )
+
+	MOV_L( ARG_SOURCE, ESI )
+	MOV_L( ARG_DEST, EDI )
+
+	MOV_L( ARG_MATRIX, EDX )
+	MOV_L( REGOFF(V4F_COUNT, ESI), ECX )
+
+	TEST_L( ECX, ECX )
+	JZ( LLBL( x86_p3_2dr_done ) )
+
+	MOV_L( REGOFF(V4F_STRIDE, ESI), EAX )
+	OR_L( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, EDI) )
+
+	MOV_L( ECX, REGOFF(V4F_COUNT, EDI) )
+	MOV_L( CONST(3), REGOFF(V4F_SIZE, EDI) )
+
+	SHL_L( CONST(4), ECX )
+	MOV_L( REGOFF(V4F_START, ESI), ESI )
+
+	MOV_L( REGOFF(V4F_START, EDI), EDI )
+	ADD_L( EDI, ECX )
+
+ALIGNTEXT16
+LLBL( x86_p3_2dr_loop ):
+
+	FLD_S( SRC(0) )			/* F4 */
+	FMUL_S( MAT(0) )
+	FLD_S( SRC(0) )			/* F5 F4 */
+	FMUL_S( MAT(1) )
+
+	FLD_S( SRC(1) )			/* F0 F5 F4 */
+	FMUL_S( MAT(4) )
+	FLD_S( SRC(1) )			/* F1 F0 F5 F4 */
+	FMUL_S( MAT(5) )
+
+	FXCH( ST(1) )			/* F0 F1 F5 F4 */
+	FADDP( ST(0), ST(3) )		/* F1 F5 F4 */
+	FADDP( ST(0), ST(1) )		/* F5 F4 */
+
+	FXCH( ST(1) )			/* F4 F5 */
+	FADD_S( MAT(12) )
+	FXCH( ST(1) )			/* F5 F4 */
+	FADD_S( MAT(13) )
+
+	MOV_L( SRC(2), EBX )
+
+	FXCH( ST(1) )			/* F4 F5 */
+	FSTP_S( DST(0)   )		/* F5 */
+	FSTP_S( DST(1)   )		/* */
+	MOV_L( EBX, DST(2) )
+
+LLBL( x86_p3_2dr_skip ):
+
+	ADD_L( CONST(16), EDI )
+	ADD_L( EAX, ESI )
+	CMP_L( ECX, EDI )
+	JNE( LLBL( x86_p3_2dr_loop ) )
+
+LLBL( x86_p3_2dr_done ):
+
+	POP_L( EBX )
+	POP_L( EDI )
+	POP_L( ESI )
+	RET
+#undef FRAME_OFFSET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_x86_transform_points3_2d_no_rot )
+GLNAME( _mesa_x86_transform_points3_2d_no_rot ):
+
+#define FRAME_OFFSET 12
+	PUSH_L( ESI )
+	PUSH_L( EDI )
+	PUSH_L( EBX )
+
+	MOV_L( ARG_SOURCE, ESI )
+	MOV_L( ARG_DEST, EDI )
+
+	MOV_L( ARG_MATRIX, EDX )
+	MOV_L( REGOFF(V4F_COUNT, ESI), ECX )
+
+	TEST_L( ECX, ECX )
+	JZ( LLBL( x86_p3_2dnrr_done ) )
+
+	MOV_L( REGOFF(V4F_STRIDE, ESI), EAX )
+	OR_L( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, EDI) )
+
+	MOV_L( ECX, REGOFF(V4F_COUNT, EDI) )
+	MOV_L( CONST(3), REGOFF(V4F_SIZE, EDI) )
+
+	SHL_L( CONST(4), ECX )
+	MOV_L( REGOFF(V4F_START, ESI), ESI )
+
+	MOV_L( REGOFF(V4F_START, EDI), EDI )
+	ADD_L( EDI, ECX )
+
+ALIGNTEXT16
+LLBL( x86_p3_2dnrr_loop ):
+
+	FLD_S( SRC(0) )			/* F4 */
+	FMUL_S( MAT(0) )
+
+	FLD_S( SRC(1) )			/* F1 F4 */
+	FMUL_S( MAT(5) )
+
+	FXCH( ST(1) )			/* F4 F1 */
+	FADD_S( MAT(12) )
+	FLD_S( MAT(13) )		/* F5 F4 F1 */
+
+	FXCH( ST(2) )			/* F1 F4 F5 */
+	FADDP( ST(0), ST(2) )		/* F4 F5 */
+
+	MOV_L( SRC(2), EBX )
+
+	FSTP_S( DST(0) )		/* F5 */
+	FSTP_S( DST(1) )		/* */
+	MOV_L( EBX, DST(2) )
+
+LLBL( x86_p3_2dnrr_skip ):
+
+	ADD_L( CONST(16), EDI )
+	ADD_L( EAX, ESI )
+	CMP_L( ECX, EDI )
+	JNE( LLBL( x86_p3_2dnrr_loop ) )
+
+LLBL( x86_p3_2dnrr_done ):
+
+	POP_L( EBX )
+	POP_L( EDI )
+	POP_L( ESI )
+	RET
+#undef FRAME_OFFSET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_x86_transform_points3_identity )
+GLNAME(_mesa_x86_transform_points3_identity ):
+
+#define FRAME_OFFSET 16
+	PUSH_L( ESI )
+	PUSH_L( EDI )
+	PUSH_L( EBX )
+	PUSH_L( EBP )
+
+	MOV_L( ARG_SOURCE, ESI )
+	MOV_L( ARG_DEST, EDI )
+
+	MOV_L( ARG_MATRIX, EDX )
+	MOV_L( REGOFF(V4F_COUNT, ESI), ECX )
+
+	TEST_L( ECX, ECX )
+	JZ( LLBL( x86_p3_ir_done ) )
+
+	MOV_L( REGOFF(V4F_STRIDE, ESI), EAX )
+	OR_L( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, EDI) )
+
+	MOV_L( ECX, REGOFF(V4F_COUNT, EDI) )
+	MOV_L( CONST(3), REGOFF(V4F_SIZE, EDI) )
+
+	SHL_L( CONST(4), ECX )
+	MOV_L( REGOFF(V4F_START, ESI), ESI )
+
+	MOV_L( REGOFF(V4F_START, EDI), EDI )
+	ADD_L( EDI, ECX )
+
+	CMP_L( ESI, EDI )
+	JE( LLBL( x86_p3_ir_done ) )
+
+ALIGNTEXT16
+LLBL( x86_p3_ir_loop ):
+
+#if 1
+	MOV_L( SRC(0), EBX )
+	MOV_L( SRC(1), EBP )
+	MOV_L( SRC(2), EDX )
+
+	MOV_L( EBX, DST(0) )
+	MOV_L( EBP, DST(1) )
+	MOV_L( EDX, DST(2) )
+#else
+	FLD_S( SRC(0) )
+	FLD_S( SRC(1) )
+	FLD_S( SRC(2) )
+
+	FSTP_S( DST(2) )
+	FSTP_S( DST(1) )
+	FSTP_S( DST(0) )
+#endif
+
+LLBL( x86_p3_ir_skip ):
+
+	ADD_L( CONST(16), EDI )
+	ADD_L( EAX, ESI )
+	CMP_L( ECX, EDI )
+	JNE( LLBL( x86_p3_ir_loop ) )
+
+LLBL( x86_p3_ir_done ):
+
+	POP_L( EBP )
+	POP_L( EBX )
+	POP_L( EDI )
+	POP_L( ESI )
+	RET
diff --git a/src/mesa/x86/x86_xform4.S b/src/mesa/x86/x86_xform4.S
new file mode 100644
index 00000000000..1b6a8d8aec0
--- /dev/null
+++ b/src/mesa/x86/x86_xform4.S
@@ -0,0 +1,639 @@
+/* $Id: x86_xform4.S,v 1.1 2001/03/29 06:46:27 gareth Exp $ */
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  3.5
+ *
+ * Copyright (C) 1999-2001  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.
+ */
+
+#include "matypes.h"
+#include "xform_args.h"
+
+	SEG_TEXT
+
+#define FP_ONE		1065353216
+#define FP_ZERO		0
+
+#define SRC(i)		REGOFF(i * 4, ESI)
+#define DST(i)		REGOFF(i * 4, EDI)
+#define MAT(i)		REGOFF(i * 4, EDX)
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_x86_transform_points4_general )
+GLNAME( _mesa_x86_transform_points4_general ):
+
+#define FRAME_OFFSET 8
+	PUSH_L( ESI )
+	PUSH_L( EDI )
+
+	MOV_L( ARG_SOURCE, ESI )
+	MOV_L( ARG_DEST, EDI )
+
+	MOV_L( ARG_MATRIX, EDX )
+	MOV_L( REGOFF(V4F_COUNT, ESI), ECX )
+
+	TEST_L( ECX, ECX )
+	JZ( LLBL( x86_p4_gr_done ) )
+
+	MOV_L( REGOFF(V4F_STRIDE, ESI), EAX )
+	OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) )
+
+	MOV_L( ECX, REGOFF(V4F_COUNT, EDI) )
+	MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) )
+
+	SHL_L( CONST(4), ECX )
+	MOV_L( REGOFF(V4F_START, ESI), ESI )
+
+	MOV_L( REGOFF(V4F_START, EDI), EDI )
+	ADD_L( EDI, ECX )
+
+ALIGNTEXT16
+LLBL( x86_p4_gr_loop ):
+
+	FLD_S( SRC(0) )			/* F4 */
+	FMUL_S( MAT(0) )
+	FLD_S( SRC(0) )			/* F5 F4 */
+	FMUL_S( MAT(1) )
+	FLD_S( SRC(0) )			/* F6 F5 F4 */
+	FMUL_S( MAT(2) )
+	FLD_S( SRC(0) )			/* F7 F6 F5 F4 */
+	FMUL_S( MAT(3) )
+
+	FLD_S( SRC(1) )			/* F0 F7 F6 F5 F4 */
+	FMUL_S( MAT(4) )
+	FLD_S( SRC(1) )			/* F1 F0 F7 F6 F5 F4 */
+	FMUL_S( MAT(5) )
+	FLD_S( SRC(1) )			/* F2 F1 F0 F7 F6 F5 F4 */
+	FMUL_S( MAT(6) )
+	FLD_S( SRC(1) )			/* F3 F2 F1 F0 F7 F6 F5 F4 */
+	FMUL_S( MAT(7) )
+
+	FXCH( ST(3) )			/* F0 F2 F1 F3 F7 F6 F5 F4 */
+	FADDP( ST(0), ST(7) )		/* F2 F1 F3 F7 F6 F5 F4 */
+	FXCH( ST(1) )			/* F1 F2 F3 F7 F6 F5 F4 */
+	FADDP( ST(0), ST(5) )		/* F2 F3 F7 F6 F5 F4 */
+	FADDP( ST(0), ST(3) )		/* F3 F7 F6 F5 F4 */
+	FADDP( ST(0), ST(1) )		/* F7 F6 F5 F4 */
+
+	FLD_S( SRC(2) )			/* F0 F7 F6 F5 F4 */
+	FMUL_S( MAT(8) )
+	FLD_S( SRC(2) )			/* F1 F0 F7 F6 F5 F4 */
+	FMUL_S( MAT(9) )
+	FLD_S( SRC(2) )			/* F2 F1 F0 F7 F6 F5 F4 */
+	FMUL_S( MAT(10) )
+	FLD_S( SRC(2) )			/* F3 F2 F1 F0 F7 F6 F5 F4 */
+	FMUL_S( MAT(11) )
+
+	FXCH( ST(3) )			/* F0 F2 F1 F3 F7 F6 F5 F4 */
+	FADDP( ST(0), ST(7) )		/* F2 F1 F3 F7 F6 F5 F4 */
+	FXCH( ST(1) )			/* F1 F2 F3 F7 F6 F5 F4 */
+	FADDP( ST(0), ST(5) )		/* F2 F3 F7 F6 F5 F4 */
+	FADDP( ST(0), ST(3) )		/* F3 F7 F6 F5 F4 */
+	FADDP( ST(0), ST(1) )		/* F7 F6 F5 F4 */
+
+	FLD_S( SRC(3) )			/* F0 F7 F6 F5 F4 */
+	FMUL_S( MAT(12) )
+	FLD_S( SRC(3) )			/* F1 F0 F7 F6 F5 F4 */
+	FMUL_S( MAT(13) )
+	FLD_S( SRC(3) )			/* F2 F1 F0 F7 F6 F5 F4 */
+	FMUL_S( MAT(14) )
+	FLD_S( SRC(3) )			/* F3 F2 F1 F0 F7 F6 F5 F4 */
+	FMUL_S( MAT(15) )
+
+	FXCH( ST(3) )			/* F0 F2 F1 F3 F7 F6 F5 F4 */
+	FADDP( ST(0), ST(7) )		/* F2 F1 F3 F7 F6 F5 F4 */
+	FXCH( ST(1) )			/* F1 F2 F3 F7 F6 F5 F4 */
+	FADDP( ST(0), ST(5) )		/* F2 F3 F7 F6 F5 F4 */
+	FADDP( ST(0), ST(3) )		/* F3 F7 F6 F5 F4 */
+	FADDP( ST(0), ST(1) )		/* F7 F6 F5 F4 */
+
+	FXCH( ST(3) )			/* F4 F6 F5 F7 */
+	FSTP_S( DST(0) )		/* F6 F5 F7 */
+	FXCH( ST(1) )			/* F5 F6 F7 */
+	FSTP_S( DST(1) )		/* F6 F7 */
+	FSTP_S( DST(2) )		/* F7 */
+	FSTP_S( DST(3) )		/* */
+
+LLBL( x86_p4_gr_skip ):
+
+	ADD_L( CONST(16), EDI )
+	ADD_L( EAX, ESI )
+	CMP_L( ECX, EDI )
+	JNE( LLBL( x86_p4_gr_loop ) )
+
+LLBL( x86_p4_gr_done ):
+
+	POP_L( EDI )
+	POP_L( ESI )
+	RET
+#undef FRAME_OFFSET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_x86_transform_points4_perspective )
+GLNAME( _mesa_x86_transform_points4_perspective ):
+
+#define FRAME_OFFSET 12
+	PUSH_L( ESI )
+	PUSH_L( EDI )
+	PUSH_L( EBX )
+
+	MOV_L( ARG_SOURCE, ESI )
+	MOV_L( ARG_DEST, EDI )
+
+	MOV_L( ARG_MATRIX, EDX )
+	MOV_L( REGOFF(V4F_COUNT, ESI), ECX )
+
+	TEST_L( ECX, ECX )
+	JZ( LLBL( x86_p4_pr_done ) )
+
+	MOV_L( REGOFF(V4F_STRIDE, ESI), EAX )
+	OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) )
+
+	MOV_L( ECX, REGOFF(V4F_COUNT, EDI) )
+	MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) )
+
+	SHL_L( CONST(4), ECX )
+	MOV_L( REGOFF(V4F_START, ESI), ESI )
+
+	MOV_L( REGOFF(V4F_START, EDI), EDI )
+	ADD_L( EDI, ECX )
+
+ALIGNTEXT16
+LLBL( x86_p4_pr_loop ):
+
+	FLD_S( SRC(0) )			/* F4 */
+	FMUL_S( MAT(0) )
+
+	FLD_S( SRC(1) )			/* F5 F4 */
+	FMUL_S( MAT(5) )
+
+	FLD_S( SRC(2) )			/* F0 F5 F4 */
+	FMUL_S( MAT(8) )
+	FLD_S( SRC(2) )			/* F1 F0 F5 F4 */
+	FMUL_S( MAT(9) )
+	FLD_S( SRC(2) )			/* F6 F1 F0 F5 F4 */
+	FMUL_S( MAT(10) )
+
+	FXCH( ST(2) )			/* F0 F1 F6 F5 F4 */
+	FADDP( ST(0), ST(4) )		/* F1 F6 F5 F4 */
+	FADDP( ST(0), ST(2) )		/* F6 F5 F4 */
+
+	FLD_S( SRC(3) )			/* F2 F6 F5 F4 */
+	FMUL_S( MAT(14) )
+
+	FADDP( ST(0), ST(1) )		/* F6 F5 F4 */
+
+	MOV_L( SRC(2), EBX )
+	XOR_L( CONST(-2147483648), EBX )/* change sign */
+
+	FXCH( ST(2) )			/* F4 F5 F6 */
+	FSTP_S( DST(0) )		/* F5 F6 */
+	FSTP_S( DST(1) )		/* F6 */
+	FSTP_S( DST(2) )		/* */
+	MOV_L( EBX, DST(3) )
+
+LLBL( x86_p4_pr_skip ):
+
+	ADD_L( CONST(16), EDI )
+	ADD_L( EAX, ESI )
+	CMP_L( ECX, EDI )
+	JNE( LLBL( x86_p4_pr_loop ) )
+
+LLBL( x86_p4_pr_done ):
+
+	POP_L( EBX )
+	POP_L( EDI )
+	POP_L( ESI )
+	RET
+#undef FRAME_OFFSET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_x86_transform_points4_3d )
+GLNAME( _mesa_x86_transform_points4_3d ):
+
+#define FRAME_OFFSET 12
+	PUSH_L( ESI )
+	PUSH_L( EDI )
+	PUSH_L( EBX )
+
+	MOV_L( ARG_SOURCE, ESI )
+	MOV_L( ARG_DEST, EDI )
+
+	MOV_L( ARG_MATRIX, EDX )
+	MOV_L( REGOFF(V4F_COUNT, ESI), ECX )
+
+	TEST_L( ECX, ECX )
+	JZ( LLBL( x86_p4_3dr_done ) )
+
+	MOV_L( REGOFF(V4F_STRIDE, ESI), EAX )
+	OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) )
+
+	MOV_L( ECX, REGOFF(V4F_COUNT, EDI) )
+	MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) )
+
+	SHL_L( CONST(4), ECX )
+	MOV_L( REGOFF(V4F_START, ESI), ESI )
+
+	MOV_L( REGOFF(V4F_START, EDI), EDI )
+	ADD_L( EDI, ECX )
+
+ALIGNTEXT16
+LLBL( x86_p4_3dr_loop ):
+
+	FLD_S( SRC(0) )			/* F4 */
+	FMUL_S( MAT(0) )
+	FLD_S( SRC(0) )			/* F5 F4 */
+	FMUL_S( MAT(1) )
+	FLD_S( SRC(0) )			/* F6 F5 F4 */
+	FMUL_S( MAT(2) )
+
+	FLD_S( SRC(1) )			/* F0 F6 F5 F4 */
+	FMUL_S( MAT(4) )
+	FLD_S( SRC(1) )			/* F1 F0 F6 F5 F4 */
+	FMUL_S( MAT(5) )
+	FLD_S( SRC(1) )			/* F2 F1 F0 F6 F5 F4 */
+	FMUL_S( MAT(6) )
+
+	FXCH( ST(2) )			/* F0 F1 F2 F6 F5 F4 */
+	FADDP( ST(0), ST(5) )		/* F1 F2 F6 F5 F4 */
+	FADDP( ST(0), ST(3) )		/* F2 F6 F5 F4 */
+	FADDP( ST(0), ST(1) )		/* F6 F5 F4 */
+
+	FLD_S( SRC(2) )			/* F0 F6 F5 F4 */
+	FMUL_S( MAT(8) )
+	FLD_S( SRC(2) )			/* F1 F0 F6 F5 F4 */
+	FMUL_S( MAT(9) )
+	FLD_S( SRC(2) )			/* F2 F1 F0 F6 F5 F4 */
+	FMUL_S( MAT(10) )
+
+	FXCH( ST(2) )			/* F0 F1 F2 F6 F5 F4 */
+	FADDP( ST(0), ST(5) )		/* F1 F2 F6 F5 F4 */
+	FADDP( ST(0), ST(3) )		/* F2 F6 F5 F4 */
+	FADDP( ST(0), ST(1) )		/* F6 F5 F4 */
+
+	FLD_S( SRC(3) )			/* F0 F6 F5 F4 */
+	FMUL_S( MAT(12) )
+	FLD_S( SRC(3) )			/* F1 F0 F6 F5 F4 */
+	FMUL_S( MAT(13) )
+	FLD_S( SRC(3) )			/* F2 F1 F0 F6 F5 F4 */
+	FMUL_S( MAT(14) )
+
+	FXCH( ST(2) )			/* F0 F1 F2 F6 F5 F4 */
+	FADDP( ST(0), ST(5) )		/* F1 F2 F6 F5 F4 */
+	FADDP( ST(0), ST(3) )		/* F2 F6 F5 F4 */
+	FADDP( ST(0), ST(1) )		/* F6 F5 F4 */
+
+	MOV_L( SRC(3), EBX )
+
+	FXCH( ST(2) )			/* F4 F5 F6 */
+	FSTP_S( DST(0) )		/* F5 F6 */
+	FSTP_S( DST(1) )		/* F6 */
+	FSTP_S( DST(2) )		/* */
+	MOV_L( EBX, DST(3) )
+
+LLBL( x86_p4_3dr_skip ):
+
+	ADD_L( CONST(16), EDI )
+	ADD_L( EAX, ESI )
+	CMP_L( ECX, EDI )
+	JNE( LLBL( x86_p4_3dr_loop ) )
+
+LLBL( x86_p4_3dr_done ):
+
+	POP_L( EBX )
+	POP_L( EDI )
+	POP_L( ESI )
+	RET
+#undef FRAME_OFFSET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME(_mesa_x86_transform_points4_3d_no_rot)
+GLNAME(_mesa_x86_transform_points4_3d_no_rot):
+
+#define FRAME_OFFSET 12
+	PUSH_L( ESI )
+	PUSH_L( EDI )
+	PUSH_L( EBX )
+
+	MOV_L( ARG_SOURCE, ESI )
+	MOV_L( ARG_DEST, EDI )
+
+	MOV_L( ARG_MATRIX, EDX )
+	MOV_L( REGOFF(V4F_COUNT, ESI), ECX )
+
+	TEST_L( ECX, ECX )
+	JZ( LLBL( x86_p4_3dnrr_done ) )
+
+	MOV_L( REGOFF(V4F_STRIDE, ESI), EAX )
+	OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) )
+
+	MOV_L( ECX, REGOFF(V4F_COUNT, EDI) )
+	MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) )
+
+	SHL_L( CONST(4), ECX )
+	MOV_L( REGOFF(V4F_START, ESI), ESI )
+
+	MOV_L( REGOFF(V4F_START, EDI), EDI )
+	ADD_L( EDI, ECX )
+
+ALIGNTEXT16
+LLBL( x86_p4_3dnrr_loop ):
+
+	FLD_S( SRC(0) )			/* F4 */
+	FMUL_S( MAT(0) )
+
+	FLD_S( SRC(1) )			/* F5 F4 */
+	FMUL_S( MAT(5) )
+
+	FLD_S( SRC(2) )			/* F6 F5 F4 */
+	FMUL_S( MAT(10) )
+
+	FLD_S( SRC(3) )			/* F0 F6 F5 F4 */
+	FMUL_S( MAT(12) )
+	FLD_S( SRC(3) )			/* F1 F0 F6 F5 F4 */
+	FMUL_S( MAT(13) )
+	FLD_S( SRC(3) )			/* F2 F1 F0 F6 F5 F4 */
+	FMUL_S( MAT(14) )
+
+	FXCH( ST(2) )			/* F0 F1 F2 F6 F5 F4 */
+	FADDP( ST(0), ST(5) )		/* F1 F2 F6 F5 F4 */
+	FADDP( ST(0), ST(3) )		/* F2 F6 F5 F4 */
+	FADDP( ST(0), ST(1) )		/* F6 F5 F4 */
+
+	MOV_L( SRC(3), EBX )
+
+	FXCH( ST(2) )			/* F4 F5 F6 */
+	FSTP_S( DST(0)   )		/* F5 F6 */
+	FSTP_S( DST(1)   )		/* F6 */
+	FSTP_S( DST(2)   )		/* */
+	MOV_L( EBX, DST(3) )
+
+LLBL( x86_p4_3dnrr_skip ):
+
+	ADD_L( CONST(16), EDI )
+	ADD_L( EAX, ESI )
+	CMP_L( ECX, EDI )
+	JNE( LLBL( x86_p4_3dnrr_loop ) )
+
+LLBL( x86_p4_3dnrr_done ):
+
+	POP_L( EBX )
+	POP_L( EDI )
+	POP_L( ESI )
+	RET
+#undef FRAME_OFFSET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_x86_transform_points4_2d )
+GLNAME( _mesa_x86_transform_points4_2d ):
+
+#define FRAME_OFFSET 16
+	PUSH_L( ESI )
+	PUSH_L( EDI )
+	PUSH_L( EBX )
+	PUSH_L( EBP )
+
+	MOV_L( ARG_SOURCE, ESI )
+	MOV_L( ARG_DEST, EDI )
+
+	MOV_L( ARG_MATRIX, EDX )
+	MOV_L( REGOFF(V4F_COUNT, ESI), ECX )
+
+	TEST_L( ECX, ECX )
+	JZ( LLBL( x86_p4_2dr_done ) )
+
+	MOV_L( REGOFF(V4F_STRIDE, ESI), EAX )
+	OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) )
+
+	MOV_L( ECX, REGOFF(V4F_COUNT, EDI) )
+	MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) )
+
+	SHL_L( CONST(4), ECX )
+	MOV_L( REGOFF(V4F_START, ESI), ESI )
+
+	MOV_L( REGOFF(V4F_START, EDI), EDI )
+	ADD_L( EDI, ECX )
+
+ALIGNTEXT16
+LLBL( x86_p4_2dr_loop ):
+
+	FLD_S( SRC(0) )			/* F4 */
+	FMUL_S( MAT(0) )
+	FLD_S( SRC(0) )			/* F5 F4 */
+	FMUL_S( MAT(1) )
+
+	FLD_S( SRC(1) )			/* F0 F5 F4 */
+	FMUL_S( MAT(4) )
+	FLD_S( SRC(1) )			/* F1 F0 F5 F4 */
+	FMUL_S( MAT(5) )
+
+	FXCH( ST(1) )			/* F0 F1 F5 F4 */
+	FADDP( ST(0), ST(3) )		/* F1 F5 F4 */
+	FADDP( ST(0), ST(1) )		/* F5 F4 */
+
+	FLD_S( SRC(3) )			/* F0 F5 F4 */
+	FMUL_S( MAT(12) )
+	FLD_S( SRC(3) )			/* F1 F0 F5 F4 */
+	FMUL_S( MAT(13) )
+
+	FXCH( ST(1) )			/* F0 F1 F5 F4 */
+	FADDP( ST(0), ST(3) )		/* F1 F5 F4 */
+	FADDP( ST(0), ST(1) )		/* F5 F4 */
+
+	MOV_L( SRC(2), EBX )
+	MOV_L( SRC(3), EBP )
+
+	FXCH( ST(1) )			/* F4 F5 */
+	FSTP_S( DST(0) )		/* F5 */
+	FSTP_S( DST(1) )		/* */
+	MOV_L( EBX, DST(2) )
+	MOV_L( EBP, DST(3) )
+
+LLBL( x86_p4_2dr_skip ):
+
+	ADD_L( CONST(16), EDI )
+	ADD_L( EAX, ESI )
+	CMP_L( ECX, EDI )
+	JNE( LLBL( x86_p4_2dr_loop ) )
+
+LLBL( x86_p4_2dr_done ):
+
+	POP_L( EBP )
+	POP_L( EBX )
+	POP_L( EDI )
+	POP_L( ESI )
+	RET
+#undef FRAME_OFFSET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_x86_transform_points4_2d_no_rot )
+GLNAME( _mesa_x86_transform_points4_2d_no_rot ):
+
+#define FRAME_OFFSET 16
+	PUSH_L( ESI )
+	PUSH_L( EDI )
+	PUSH_L( EBX )
+	PUSH_L( EBP )
+
+	MOV_L( ARG_SOURCE, ESI )
+	MOV_L( ARG_DEST, EDI )
+
+	MOV_L( ARG_MATRIX, EDX )
+	MOV_L( REGOFF(V4F_COUNT, ESI), ECX )
+
+	TEST_L( ECX, ECX )
+	JZ( LLBL( x86_p4_2dnrr_done ) )
+
+	MOV_L( REGOFF(V4F_STRIDE, ESI), EAX )
+	OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) )
+
+	MOV_L( ECX, REGOFF(V4F_COUNT, EDI) )
+	MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) )
+
+	SHL_L( CONST(4), ECX )
+	MOV_L( REGOFF(V4F_START, ESI), ESI )
+
+	MOV_L( REGOFF(V4F_START, EDI), EDI )
+	ADD_L( EDI, ECX )
+
+ALIGNTEXT16
+LLBL( x86_p4_2dnrr_loop ):
+
+	FLD_S( SRC(0) )			/* F4 */
+	FMUL_S( MAT(0) )
+
+	FLD_S( SRC(1) )			/* F5 F4 */
+	FMUL_S( MAT(5) )
+
+	FLD_S( SRC(3) )			/* F0 F5 F4 */
+	FMUL_S( MAT(12) )
+	FLD_S( SRC(3) )			/* F1 F0 F5 F4 */
+	FMUL_S( MAT(13) )
+
+	FXCH( ST(1) )			/* F0 F1 F5 F4 */
+	FADDP( ST(0), ST(3) )		/* F1 F5 F4 */
+	FADDP( ST(0), ST(1) )		/* F5 F4 */
+
+	MOV_L( SRC(2), EBX )
+	MOV_L( SRC(3), EBP )
+
+	FXCH( ST(1) )			/* F4 F5 */
+	FSTP_S( DST(0)   )		/* F5 */
+	FSTP_S( DST(1)   )		/* */
+	MOV_L( EBX, DST(2) )
+	MOV_L( EBP, DST(3) )
+
+LLBL( x86_p4_2dnrr_skip ):
+
+	ADD_L( CONST(16), EDI )
+	ADD_L( EAX, ESI )
+	CMP_L( ECX, EDI )
+	JNE( LLBL( x86_p4_2dnrr_loop ) )
+
+LLBL( x86_p4_2dnrr_done ):
+
+	POP_L( EBP )
+	POP_L( EBX )
+	POP_L( EDI )
+	POP_L( ESI )
+	RET
+#undef FRAME_OFFSET
+
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME( _mesa_x86_transform_points4_identity )
+GLNAME( _mesa_x86_transform_points4_identity ):
+
+#define FRAME_OFFSET 12
+	PUSH_L( ESI )
+	PUSH_L( EDI )
+	PUSH_L( EBX )
+
+	MOV_L( ARG_SOURCE, ESI )
+	MOV_L( ARG_DEST, EDI )
+
+	MOV_L( ARG_MATRIX, EDX )
+	MOV_L( REGOFF(V4F_COUNT, ESI), ECX )
+
+	TEST_L( ECX, ECX )
+	JZ( LLBL( x86_p4_ir_done ) )
+
+	MOV_L( REGOFF(V4F_STRIDE, ESI), EAX )
+	OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) )
+
+	MOV_L( ECX, REGOFF(V4F_COUNT, EDI) )
+	MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) )
+
+	SHL_L( CONST(4), ECX )
+	MOV_L( REGOFF(V4F_START, ESI), ESI )
+
+	MOV_L( REGOFF(V4F_START, EDI), EDI )
+	ADD_L( EDI, ECX )
+
+	CMP_L( ESI, EDI )
+	JE( LLBL( x86_p4_ir_done ) )
+
+ALIGNTEXT16
+LLBL( x86_p4_ir_loop ):
+
+	MOV_L( SRC(0), EBX )
+	MOV_L( SRC(1), EDX )
+
+	MOV_L( EBX, DST(0) )
+	MOV_L( EDX, DST(1) )
+
+	MOV_L( SRC(2), EBX )
+	MOV_L( SRC(3), EDX )
+
+	MOV_L( EBX, DST(2) )
+	MOV_L( EDX, DST(3) )
+
+LLBL( x86_p4_ir_skip ):
+
+	ADD_L( CONST(16), EDI )
+	ADD_L( EAX, ESI )
+	CMP_L( ECX, EDI )
+	JNE( LLBL( x86_p4_ir_loop ) )
+
+LLBL( x86_p4_ir_done ):
+
+	POP_L( EBX )
+	POP_L( EDI )
+	POP_L( ESI )
+	RET