Merge commit 'origin/master' into glsl-pp-rework-2

Conflicts:
	src/gallium/winsys/gdi/SConscript

configs/default

@@ -23,6 +23,7 @@ HOST_CC = $(CC)
 CFLAGS = -O
 CXXFLAGS = -O
 LDFLAGS =
+HOST_CFLAGS = $(CFLAGS)
 GLU_CFLAGS = 
 
 # Compiler for building demos/tests/etc

configure.ac

@@ -1202,7 +1202,7 @@ if test "x$enable_gallium_radeon" = xyes; then
 fi
 
 dnl
-dnl Gallium Radeon configuration
+dnl Gallium Nouveau configuration
 dnl
 AC_ARG_ENABLE([gallium-nouveau],
     [AS_HELP_STRING([--enable-gallium-nouveau],

docs/relnotes-7.5.2.html

@@ -45,6 +45,10 @@ tbd
 <ul>
 <li>Assorted bug fixes for i965/i945 drivers
 <li>Fixed Gallium glDrawPixels(GL_STENCIL_INDEX) failure.
+<li>Fixed GLSL linker/preprocessor version directive issue seen in Wine
+    (such as bug 23946)
+<li>glUseProgram() is now compiled into display lists (bug 23746).
+<li>glUniform functions are now compiled into display lists
 </ul>
 
 

docs/relnotes-7.6.html

@@ -57,6 +57,10 @@ This was written by Zack Rusin at Tungsten Graphics.
     GL_ARB_fragment_program.</li>
 <li>Added configure --with-max-width=W, --with-max-height=H options to specify
     max framebuffer, viewport size.
+<li>Initial version of Gallium llvmpipe driver.  This is a new driver based
+    on LLVM which makes exensive use of run-time code generation.  This is
+    an "alpha" stage driver.  See the src/gallium/drivers/llvmpipe/README
+    file for more information.
 </ul>
 
 

docs/relnotes-7.7.html

@@ -37,6 +37,7 @@ tbd
 <li>GL_ARB_draw_elements_base_vertex (supported in Intel i965 and software drivers)</li>
 <li>GL_ARB_depth_clamp (supported in Intel i965 DRI and software drivers)</li>
 <li>GL_NV_depth_clamp (supported in Intel i965 DRI and software drivers)</li>
+<li>GL_ARB_provoking_vertex (same as GL_EXT_provoking_vertex)</li>
 </ul>
 
 

include/GL/gl.h

@@ -1740,6 +1740,9 @@ GLAPI void GLAPIENTRY glSeparableFilter2D( GLenum target,
 GLAPI void GLAPIENTRY glGetSeparableFilter( GLenum target, GLenum format,
 	GLenum type, GLvoid *row, GLvoid *column, GLvoid *span );
 
+typedef void (APIENTRYP PFNGLBLENDCOLORPROC) (GLclampf red, GLclampf green, GLclampf blue, GLclampf alpha);
+typedef void (APIENTRYP PFNGLBLENDEQUATIONPROC) (GLenum mode);
+
 
 
 /*
@@ -1945,6 +1948,18 @@ GLAPI void GLAPIENTRY glMultTransposeMatrixf( const GLfloat m[16] );
 GLAPI void GLAPIENTRY glSampleCoverage( GLclampf value, GLboolean invert );
 
 
+typedef void (APIENTRYP PFNGLACTIVETEXTUREPROC) (GLenum texture);
+typedef void (APIENTRYP PFNGLSAMPLECOVERAGEPROC) (GLclampf value, GLboolean invert);
+typedef void (APIENTRYP PFNGLCOMPRESSEDTEXIMAGE3DPROC) (GLenum target, GLint level, GLenum internalformat, GLsizei width, GLsizei height, GLsizei depth, GLint border, GLsizei imageSize, const GLvoid *data);
+typedef void (APIENTRYP PFNGLCOMPRESSEDTEXIMAGE2DPROC) (GLenum target, GLint level, GLenum internalformat, GLsizei width, GLsizei height, GLint border, GLsizei imageSize, const GLvoid *data);
+typedef void (APIENTRYP PFNGLCOMPRESSEDTEXIMAGE1DPROC) (GLenum target, GLint level, GLenum internalformat, GLsizei width, GLint border, GLsizei imageSize, const GLvoid *data);
+typedef void (APIENTRYP PFNGLCOMPRESSEDTEXSUBIMAGE3DPROC) (GLenum target, GLint level, GLint xoffset, GLint yoffset, GLint zoffset, GLsizei width, GLsizei height, GLsizei depth, GLenum format, GLsizei imageSize, const GLvoid *data);
+typedef void (APIENTRYP PFNGLCOMPRESSEDTEXSUBIMAGE2DPROC) (GLenum target, GLint level, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height, GLenum format, GLsizei imageSize, const GLvoid *data);
+typedef void (APIENTRYP PFNGLCOMPRESSEDTEXSUBIMAGE1DPROC) (GLenum target, GLint level, GLint xoffset, GLsizei width, GLenum format, GLsizei imageSize, const GLvoid *data);
+typedef void (APIENTRYP PFNGLGETCOMPRESSEDTEXIMAGEPROC) (GLenum target, GLint level, GLvoid *img);
+
+
+
 /*
  * GL_ARB_multitexture (ARB extension 1 and OpenGL 1.2.1)
  */

progs/demos/cubemap.c

@@ -58,6 +58,7 @@ static GLint ClampIndex = 0;
 static GLboolean supportFBO = GL_FALSE;
 static GLboolean supportSeamless = GL_FALSE;
 static GLboolean seamless = GL_FALSE;
+static GLuint TexObj = 0;
 
 
 static struct {
@@ -543,6 +544,10 @@ static void init( GLboolean useImageFiles )
 
    printf("GL_RENDERER: %s\n", (char *) glGetString(GL_RENDERER));
 
+
+   glGenTextures(1, &TexObj);
+   glBindTexture(GL_TEXTURE_CUBE_MAP_ARB, TexObj);
+
    if (useImageFiles) {
       load_envmaps();
    }

progs/demos/lodbias.c

@@ -43,6 +43,7 @@ static GLboolean Anim = GL_TRUE;
 static GLint Bias = 0, BiasStepSign = +1; /* ints avoid fp precision problem */
 static GLint BiasMin = -400, BiasMax = 400;
 static int win = 0;
+static GLuint TexObj = 0;
 
 
 static void
@@ -214,6 +215,9 @@ static void Init( void )
 
    glPixelStorei(GL_UNPACK_ALIGNMENT, 1);
 
+   glGenTextures(1, &TexObj);
+   glBindTexture(GL_TEXTURE_2D, TexObj);
+
    if (glutExtensionSupported("GL_SGIS_generate_mipmap")) {
       /* test auto mipmap generation */
       GLint width, height, i;

progs/glsl/Makefile

@@ -10,16 +10,15 @@ LIB_DEP = \
 	$(TOP)/$(LIB_DIR)/$(GLU_LIB_NAME) \
 	$(TOP)/$(LIB_DIR)/$(GLUT_LIB_NAME)
 
-LIBS = -L$(TOP)/$(LIB_DIR) -l$(GLUT_LIB) -l$(GLEW_LIB) -l$(GLU_LIB) -l$(GL_LIB) $(APP_LIB_DEPS)
-
-INCLUDE_DIRS = -I$(TOP)/progs/util
+LIBS = -L$(TOP)/$(LIB_DIR) -l$(GLUT_LIB) -l$(GLEW_LIB) -l$(GLU_LIB) \
+	-l$(GL_LIB) $(APP_LIB_DEPS)
 
 # using : to avoid APP_CC pointing to CC loop
-CC:=$(APP_CC)
+CC := $(APP_CC)
 CFLAGS += -I$(INCDIR)
-LDLIBS=$(LIBS)
+LDLIBS = $(LIBS)
 
-DEMO_SOURCES = \
+PROG_SOURCES = \
 	array.c \
 	bitmap.c \
 	brick.c \
@@ -59,8 +58,8 @@ UTIL_SOURCES = \
 	readtex.c
 
 UTIL_OBJS = $(UTIL_SOURCES:.c=.o)
-PROG_OBJS = $(DEMO_SOURCES:.c=.o)
-PROGS = $(DEMO_SOURCES:%.c=%)
+PROG_OBJS = $(PROG_SOURCES:.c=.o)
+PROGS = $(PROG_SOURCES:%.c=%)
 
 ##### TARGETS #####
 

progs/perf/Makefile

@@ -0,0 +1,49 @@
+# progs/demos/Makefile
+
+TOP = ../..
+include $(TOP)/configs/current
+
+INCDIR = $(TOP)/include
+
+LIBS = -L$(TOP)/$(LIB_DIR) -l$(GLUT_LIB) -l$(GLEW_LIB) \
+	-l$(GLU_LIB) -l$(GL_LIB) $(APP_LIB_DEPS)
+
+# using : to avoid APP_CC pointing to CC loop
+CC := $(APP_CC)
+CFLAGS += -I$(INCDIR)
+LDLIBS = $(LIBS)
+
+PROG_SOURCES = \
+	drawoverhead.c \
+	teximage.c \
+	vbo.c \
+	vertexrate.c \
+
+PROG_OBJS = $(PROG_SOURCES:.c=.o)
+
+PROGS = $(PROG_SOURCES:%.c=%)
+
+
+UTIL_SOURCES = \
+	common.c \
+	glmain.c
+
+UTIL_HEADERS = \
+	common.h \
+	glmain.h
+
+UTIL_OBJS = $(UTIL_SOURCES:.c=.o)
+
+
+
+default: $(PROGS)
+
+$(PROG_OBJS): $(UTIL_HEADERS)
+
+$(PROGS): $(UTIL_OBJS)
+
+
+
+clean:
+	-rm -f $(PROGS)
+	-rm -f *.o *~

progs/perf/common.c

@@ -0,0 +1,93 @@
+/*
+ * Copyright (C) 2009  VMware, Inc.  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
+ * VMWARE 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.
+ */
+
+/**
+ * Common perf code.  This should be re-usable with other APIs.
+ */
+
+#include "common.h"
+#include "glmain.h"
+
+
+/**
+ * Run function 'f' for enough iterations to reach a steady state.
+ * Return the rate (iterations/second).
+ */
+double
+PerfMeasureRate(PerfRateFunc f)
+{
+   const double minDuration = 1.0;
+   double rate = 0.0, prevRate = 0.0;
+   unsigned subiters;
+
+   /* Compute initial number of iterations to try.
+    * If the test function is pretty slow this helps to avoid
+    * extraordarily long run times.
+    */
+   subiters = 2;
+   {
+      const double t0 = PerfGetTime();
+      double t1;
+      do {
+         f(subiters); /* call the rendering function */
+         t1 = PerfGetTime();
+         subiters *= 2;
+      } while (t1 - t0 < 0.1 * minDuration);
+   }
+   /*printf("initial subIters = %u\n", subiters);*/
+
+   while (1) {
+      const double t0 = PerfGetTime();
+      unsigned iters = 0;
+      double t1;
+
+      do {
+         f(subiters); /* call the rendering function */
+         t1 = PerfGetTime();
+         iters += subiters;
+      } while (t1 - t0 < minDuration);
+
+      rate = iters / (t1 - t0);
+
+      if (0)
+         printf("prevRate %f  rate  %f  ratio %f  iters %u\n",
+                prevRate, rate, rate/prevRate, iters);
+
+      /* Try and speed the search up by skipping a few steps:
+       */
+      if (rate > prevRate * 1.6)
+         subiters *= 8;
+      else if (rate > prevRate * 1.2)
+         subiters *= 4;
+      else if (rate > prevRate * 1.05)
+         subiters *= 2;
+      else
+         break;
+
+      prevRate = rate;
+   }
+
+   if (0)
+      printf("%s returning iters %u  rate %f\n", __FUNCTION__, subiters, rate);
+   return rate;
+}
+
+

progs/perf/common.h

@@ -0,0 +1,35 @@
+/*
+ * Copyright (C) 2009  VMware, Inc.  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
+ * VMWARE BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+
+#ifndef COMMON_H
+#define COMMON_H
+
+
+typedef void (*PerfRateFunc)(unsigned count);
+
+
+extern double
+PerfMeasureRate(PerfRateFunc f);
+
+
+#endif /* COMMON_H */
+

progs/perf/drawoverhead.c

@@ -0,0 +1,133 @@
+/*
+ * Copyright (C) 2009  VMware, Inc.  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
+ * VMWARE 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.
+ */
+
+/**
+ * Measure drawing overhead
+ *
+ * This is the first in a series of simple performance benchmarks.
+ * The code in this file should be as simple as possible to make it
+ * easily portable to other APIs.
+ *
+ * All the window-system stuff should be contained in glmain.c (or TBDmain.c).
+ * All the re-usable, generic code should be in common.c (XXX not done yet).
+ *
+ * Brian Paul
+ * 15 Sep 2009
+ */
+
+#include "glmain.h"
+#include "common.h"
+
+
+int WinWidth = 100, WinHeight = 100;
+
+static GLuint VBO;
+
+struct vertex
+{
+   GLfloat x, y;
+};
+
+static const struct vertex vertices[4] = {
+   { -1.0, -1.0 },
+   {  1.0, -1.0 },
+   {  1.0,  1.0 },
+   { -1.0,  1.0 }
+};
+
+
+/** Called from test harness/main */
+void
+PerfInit(void)
+{
+   /* setup VBO w/ vertex data */
+   glGenBuffersARB(1, &VBO);
+   glBindBufferARB(GL_ARRAY_BUFFER_ARB, VBO);
+   glBufferDataARB(GL_ARRAY_BUFFER_ARB,
+                   sizeof(vertices), vertices, GL_STATIC_DRAW_ARB);
+   glVertexPointer(2, GL_FLOAT, sizeof(struct vertex), (void *) 0);
+   glEnableClientState(GL_VERTEX_ARRAY);
+
+   /* misc GL state */
+   glAlphaFunc(GL_ALWAYS, 0.0);
+}
+
+
+static void
+DrawNoStateChange(unsigned count)
+{
+   unsigned i;
+   for (i = 0; i < count; i++) {
+      glDrawArrays(GL_POINTS, 0, 4);
+   }
+   glFinish();
+}
+
+
+static void
+DrawNopStateChange(unsigned count)
+{
+   unsigned i;
+   for (i = 0; i < count; i++) {
+      glDisable(GL_ALPHA_TEST);
+      glDrawArrays(GL_POINTS, 0, 4);
+   }
+   glFinish();
+}
+
+
+static void
+DrawStateChange(unsigned count)
+{
+   unsigned i;
+   for (i = 0; i < count; i++) {
+      if (i & 1)
+         glEnable(GL_TEXTURE_GEN_S);
+      else
+         glDisable(GL_TEXTURE_GEN_S);
+      glDrawArrays(GL_POINTS, 0, 4);
+   }
+   glFinish();
+}
+
+
+/** Called from test harness/main */
+void
+PerfDraw(void)
+{
+   double rate0, rate1, rate2, overhead;
+
+   rate0 = PerfMeasureRate(DrawNoStateChange);
+   printf("   Draw only: %.1f draws/second\n", rate0);
+
+   rate1 = PerfMeasureRate(DrawNopStateChange);
+   overhead = 1000.0 * (1.0 / rate1 - 1.0 / rate0);
+   printf("   Draw w/ nop state change: %.1f draws/sec (overhead: %f ms/draw)\n",
+          rate1, overhead);
+
+   rate2 = PerfMeasureRate(DrawStateChange);
+   overhead = 1000.0 * (1.0 / rate2 - 1.0 / rate0);
+   printf("   Draw w/ state change: %.1f draws/sec (overhead: %f ms/draw)\n",
+          rate2, overhead);
+
+   exit(0);
+}
+

progs/perf/glmain.c

@@ -0,0 +1,156 @@
+/*
+ * Copyright (C) 2009  VMware, Inc.  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
+ * VMWARE 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.
+ */
+
+/**
+ * OpenGL/GLUT common code for perf programs.
+ * Brian Paul
+ * 15 Sep 2009
+ */
+
+
+#include "glmain.h"
+#include <GL/glut.h>
+
+
+static int Win;
+static GLfloat Xrot = 0, Yrot = 0, Zrot = 0;
+static GLboolean Anim = GL_FALSE;
+
+
+/** Return time in seconds */
+double
+PerfGetTime(void)
+{
+   return glutGet(GLUT_ELAPSED_TIME) * 0.001;
+}
+
+
+void
+PerfSwapBuffers(void)
+{
+   glutSwapBuffers();
+}
+
+
+static void
+Idle(void)
+{
+   Xrot += 3.0;
+   Yrot += 4.0;
+   Zrot += 2.0;
+   glutPostRedisplay();
+}
+
+
+static void
+Draw(void)
+{
+   PerfDraw();
+   glutSwapBuffers();
+}
+
+
+static void
+Reshape(int width, int height)
+{
+   WinWidth = width;
+   WinHeight = height;
+   glViewport(0, 0, width, height);
+   glMatrixMode(GL_PROJECTION);
+   glLoadIdentity();
+   glFrustum(-1.0, 1.0, -1.0, 1.0, 5.0, 25.0);
+   glMatrixMode(GL_MODELVIEW);
+   glLoadIdentity();
+   glTranslatef(0.0, 0.0, -15.0);
+}
+
+
+static void
+Key(unsigned char key, int x, int y)
+{
+   const GLfloat step = 3.0;
+   (void) x;
+   (void) y;
+   switch (key) {
+   case 'a':
+      Anim = !Anim;
+      if (Anim)
+         glutIdleFunc(Idle);
+      else
+         glutIdleFunc(NULL);
+      break;
+   case 'z':
+      Zrot -= step;
+      break;
+   case 'Z':
+      Zrot += step;
+      break;
+   case 27:
+      glutDestroyWindow(Win);
+      exit(0);
+      break;
+   }
+   glutPostRedisplay();
+}
+
+
+static void
+SpecialKey(int key, int x, int y)
+{
+   const GLfloat step = 3.0;
+   (void) x;
+   (void) y;
+   switch (key) {
+   case GLUT_KEY_UP:
+      Xrot -= step;
+      break;
+   case GLUT_KEY_DOWN:
+      Xrot += step;
+      break;
+   case GLUT_KEY_LEFT:
+      Yrot -= step;
+      break;
+   case GLUT_KEY_RIGHT:
+      Yrot += step;
+      break;
+   }
+   glutPostRedisplay();
+}
+
+
+int
+main(int argc, char *argv[])
+{
+   glutInit(&argc, argv);
+   glutInitWindowSize(WinWidth, WinHeight);
+   glutInitDisplayMode(GLUT_RGB | GLUT_DOUBLE | GLUT_DEPTH);
+   Win = glutCreateWindow(argv[0]);
+   glewInit();
+   glutReshapeFunc(Reshape);
+   glutKeyboardFunc(Key);
+   glutSpecialFunc(SpecialKey);
+   glutDisplayFunc(Draw);
+   if (Anim)
+      glutIdleFunc(Idle);
+   PerfInit();
+   glutMainLoop();
+   return 0;
+}

progs/perf/glmain.h

@@ -0,0 +1,54 @@
+/*
+ * Copyright (C) 2009  VMware, Inc.  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
+ * VMWARE BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+
+#ifndef GLMAIN_H
+#define GLMAIN_H
+
+
+#define GL_GLEXT_PROTOTYPES
+#include <GL/glew.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <math.h>
+
+
+/** Test programs can use these vars/functions */
+
+extern int WinWidth, WinHeight;
+
+extern double
+PerfGetTime(void);
+
+extern void
+PerfSwapBuffers(void);
+
+
+/** Test programs must implement these functions **/
+
+extern void
+PerfInit(void);
+
+extern void
+PerfDraw(void);
+
+
+#endif /* GLMAIN_H */

progs/perf/teximage.c

@@ -0,0 +1,210 @@
+/*
+ * Copyright (C) 2009  VMware, Inc.  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
+ * VMWARE 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.
+ */
+
+/**
+ * Measure glTexSubImage2D rate
+ *
+ * Brian Paul
+ * 16 Sep 2009
+ */
+
+#include "glmain.h"
+#include "common.h"
+
+
+int WinWidth = 100, WinHeight = 100;
+
+static GLuint VBO;
+static GLuint TexObj = 0;
+static GLubyte *TexImage = NULL;
+static GLsizei TexSize;
+static GLenum TexSrcFormat, TexSrcType;
+
+static const GLboolean DrawPoint = GL_TRUE;
+static const GLboolean TexSubImage4 = GL_TRUE;
+
+struct vertex
+{
+   GLfloat x, y, s, t;
+};
+
+static const struct vertex vertices[1] = {
+   { 0.0, 0.0, 0.5, 0.5 },
+};
+
+
+#define VOFFSET(F) ((void *) offsetof(struct vertex, F))
+
+/** Called from test harness/main */
+void
+PerfInit(void)
+{
+   /* setup VBO w/ vertex data */
+   glGenBuffersARB(1, &VBO);
+   glBindBufferARB(GL_ARRAY_BUFFER_ARB, VBO);
+   glBufferDataARB(GL_ARRAY_BUFFER_ARB,
+                   sizeof(vertices), vertices, GL_STATIC_DRAW_ARB);
+   glVertexPointer(2, GL_FLOAT, sizeof(struct vertex), VOFFSET(x));
+   glTexCoordPointer(2, GL_FLOAT, sizeof(struct vertex), VOFFSET(s));
+   glEnableClientState(GL_VERTEX_ARRAY);
+   glEnableClientState(GL_TEXTURE_COORD_ARRAY);
+
+   /* texture */
+   glGenTextures(1, &TexObj);
+   glBindTexture(GL_TEXTURE_2D, TexObj);
+   glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
+   glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
+   glEnable(GL_TEXTURE_2D);
+}
+
+
+static void
+UploadTexImage2D(unsigned count)
+{
+   unsigned i;
+   for (i = 0; i < count; i++) {
+      /* XXX is this equivalent to a glTexSubImage call since we're
+       * always specifying the same image size?  That case isn't optimized
+       * in Mesa but may be optimized in other drivers.  Note sure how
+       * much difference that might make.
+       */
+      glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA,
+                   TexSize, TexSize, 0,
+                   TexSrcFormat, TexSrcType, TexImage);
+      if (DrawPoint)
+         glDrawArrays(GL_POINTS, 0, 1);
+   }
+   glFinish();
+}
+
+
+static void
+UploadTexSubImage2D(unsigned count)
+{
+   unsigned i;
+   for (i = 0; i < count; i++) {
+      if (TexSubImage4) {
+         GLsizei halfSize = (TexSize == 1) ? 1 : TexSize / 2;
+         GLsizei halfPos = TexSize - halfSize;
+         /* do glTexSubImage2D in four pieces */
+         /* lower-left */
+         glPixelStorei(GL_UNPACK_ROW_LENGTH, TexSize);
+         glTexSubImage2D(GL_TEXTURE_2D, 0,
+                         0, 0, halfSize, halfSize,
+                         TexSrcFormat, TexSrcType, TexImage);
+         /* lower-right */
+         glPixelStorei(GL_UNPACK_SKIP_PIXELS, halfPos);
+         glTexSubImage2D(GL_TEXTURE_2D, 0,
+                         halfPos, 0, halfSize, halfSize,
+                         TexSrcFormat, TexSrcType, TexImage);
+         /* upper-left */
+         glPixelStorei(GL_UNPACK_SKIP_PIXELS, 0);
+         glPixelStorei(GL_UNPACK_SKIP_ROWS, halfPos);
+         glTexSubImage2D(GL_TEXTURE_2D, 0,
+                         0, halfPos, halfSize, halfSize,
+                         TexSrcFormat, TexSrcType, TexImage);
+         /* upper-right */
+         glPixelStorei(GL_UNPACK_SKIP_PIXELS, halfPos);
+         glPixelStorei(GL_UNPACK_SKIP_ROWS, halfPos);
+         glTexSubImage2D(GL_TEXTURE_2D, 0,
+                         halfPos, halfPos, halfSize, halfSize,
+                         TexSrcFormat, TexSrcType, TexImage);
+         /* reset the unpacking state */
+         glPixelStorei(GL_UNPACK_SKIP_PIXELS, 0);
+         glPixelStorei(GL_UNPACK_SKIP_PIXELS, 0);
+         glPixelStorei(GL_UNPACK_SKIP_ROWS, 0);
+      }
+      else {
+         /* replace whole texture image at once */
+         glTexSubImage2D(GL_TEXTURE_2D, 0,
+                         0, 0, TexSize, TexSize,
+                         TexSrcFormat, TexSrcType, TexImage);
+      }
+      if (DrawPoint)
+         glDrawArrays(GL_POINTS, 0, 1);
+   }
+   glFinish();
+}
+
+
+/* XXX any other formats to measure? */
+static const struct {
+   GLenum format, type;
+   const char *name;
+} SrcFormats[] = {
+   { GL_RGBA, GL_UNSIGNED_BYTE, "GL_RGBA/GLubyte" },
+   { GL_BGRA, GL_UNSIGNED_BYTE, "GL_BGRA/GLubyte" },
+   { 0, 0, NULL }
+};
+
+
+
+/** Called from test harness/main */
+void
+PerfDraw(void)
+{
+   GLint maxSize;
+   double rate;
+   GLint fmt, subImage;
+
+   glGetIntegerv(GL_MAX_TEXTURE_SIZE, &maxSize);
+
+   /* loop over source data formats */
+   for (fmt = 0; SrcFormats[fmt].format; fmt++) {
+      TexSrcFormat = SrcFormats[fmt].format;
+      TexSrcType = SrcFormats[fmt].type;
+
+      /* loop over glTexImage, glTexSubImage */
+      for (subImage = 0; subImage < 2; subImage++) {
+
+         /* loop over texture sizes */
+         for (TexSize = 16; TexSize <= maxSize; TexSize *= 2) {
+            GLint bytesPerImage;
+            double mbPerSec;
+
+            bytesPerImage = TexSize * TexSize * 4;
+            TexImage = malloc(bytesPerImage);
+
+            if (subImage) {
+               /* create initial, empty texture */
+               glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA,
+                            TexSize, TexSize, 0,
+                            TexSrcFormat, TexSrcType, NULL);
+               rate = PerfMeasureRate(UploadTexSubImage2D);
+            }
+            else {
+               rate = PerfMeasureRate(UploadTexImage2D);
+            }
+
+            mbPerSec = rate * bytesPerImage / (1024.0 * 1024.0);
+
+            printf("  glTex%sImage2D(%s %d x %d): "
+                   "%.1f images/sec, %.1f MB/sec\n",
+                   (subImage ? "Sub" : ""),
+                   SrcFormats[fmt].name, TexSize, TexSize, rate, mbPerSec);
+
+            free(TexImage);
+         }
+      }
+   }
+
+   exit(0);
+}

progs/perf/vbo.c

@@ -0,0 +1,138 @@
+/*
+ * Copyright (C) 2009  VMware, Inc.  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
+ * VMWARE 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.
+ */
+
+/**
+ * Measure VBO upload speed.
+ * That is, measure glBufferDataARB() and glBufferSubDataARB().
+ *
+ * Brian Paul
+ * 16 Sep 2009
+ */
+
+#include <string.h>
+#include "glmain.h"
+#include "common.h"
+
+
+int WinWidth = 100, WinHeight = 100;
+
+static GLuint VBO;
+
+static GLsizei VBOSize = 0;
+static GLubyte *VBOData = NULL;
+
+static const GLboolean DrawPoint = GL_TRUE;
+static const GLboolean BufferSubDataInHalves = GL_TRUE;
+
+static const GLfloat Vertex0[2] = { 0.0, 0.0 };
+
+
+/** Called from test harness/main */
+void
+PerfInit(void)
+{
+   /* setup VBO */
+   glGenBuffersARB(1, &VBO);
+   glBindBufferARB(GL_ARRAY_BUFFER_ARB, VBO);
+   glVertexPointer(2, GL_FLOAT, sizeof(Vertex0), (void *) 0);
+   glEnableClientState(GL_VERTEX_ARRAY);
+}
+
+
+static void
+UploadVBO(unsigned count)
+{
+   unsigned i;
+   for (i = 0; i < count; i++) {
+      glBufferDataARB(GL_ARRAY_BUFFER, VBOSize, VBOData, GL_STREAM_DRAW_ARB);
+
+      if (DrawPoint)
+         glDrawArrays(GL_POINTS, 0, 1);
+   }
+   glFinish();
+}
+
+
+static void
+UploadSubVBO(unsigned count)
+{
+   unsigned i;
+   for (i = 0; i < count; i++) {
+      if (BufferSubDataInHalves) {
+         GLsizei half = VBOSize / 2;
+         glBufferSubDataARB(GL_ARRAY_BUFFER, 0, half, VBOData);
+         glBufferSubDataARB(GL_ARRAY_BUFFER, half, half, VBOData + half);
+      }
+      else {
+         glBufferSubDataARB(GL_ARRAY_BUFFER, 0, VBOSize, VBOData);
+      }
+
+      if (DrawPoint)
+         glDrawArrays(GL_POINTS, 0, 1);
+   }
+   glFinish();
+}
+
+
+static const GLsizei Sizes[] = {
+   64,
+   1024,
+   16*1024,
+   256*1024,
+   1024*1024,
+   16*1024*1024,
+   0 /* end of list */
+};
+
+
+/** Called from test harness/main */
+void
+PerfDraw(void)
+{
+   double rate, mbPerSec;
+   int sub, sz;
+
+   /* loop over whole/sub buffer upload */
+   for (sub = 0; sub < 2; sub++) {
+
+      /* loop over VBO sizes */
+      for (sz = 0; Sizes[sz]; sz++) {
+         VBOSize = Sizes[sz];
+
+         VBOData = malloc(VBOSize);
+         memcpy(VBOData, Vertex0, sizeof(Vertex0));
+
+         if (sub)
+            rate = PerfMeasureRate(UploadSubVBO);
+         else
+            rate = PerfMeasureRate(UploadVBO);
+
+         mbPerSec = rate * VBOSize / (1024.0 * 1024.0);
+
+         printf("  glBuffer%sDataARB(size = %d): %.1f MB/sec\n",
+                (sub ? "Sub" : ""), VBOSize, mbPerSec);
+
+         free(VBOData);
+      }
+   }
+
+   exit(0);
+}

progs/perf/vertexrate.c

@@ -0,0 +1,271 @@
+/*
+ * Copyright (C) 2009  VMware, Inc.  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
+ * VMWARE 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.
+ */
+
+/**
+ * Measure simple vertex processing rate via:
+ *  - immediate mode
+ *  - vertex arrays
+ *  - VBO vertex arrays
+ *  - glDrawElements
+ *  - VBO glDrawElements
+ *  - glDrawRangeElements
+ *  - VBO glDrawRangeElements
+ *
+ * Brian Paul
+ * 16 Sep 2009
+ */
+
+#include <assert.h>
+#include <string.h>
+#include "glmain.h"
+#include "common.h"
+
+
+#define MAX_VERTS (100 * 100)
+
+/** glVertex2/3/4 size */
+#define VERT_SIZE 4
+
+int WinWidth = 500, WinHeight = 500;
+
+static GLuint VertexBO, ElementBO;
+
+static unsigned NumVerts = MAX_VERTS;
+static unsigned VertBytes = VERT_SIZE * sizeof(float);
+static float *VertexData = NULL;
+
+static unsigned NumElements = MAX_VERTS;
+static GLuint *Elements = NULL;
+
+
+/**
+ * Load VertexData buffer with a 2-D grid of points in the range [-1,1]^2.
+ */
+static void
+InitializeVertexData(void)
+{
+   unsigned i;
+   float x = -1.0, y = -1.0;
+   float dx = 2.0 / 100;
+   float dy = 2.0 / 100;
+
+   VertexData = (float *) malloc(NumVerts * VertBytes);
+
+   for (i = 0; i < NumVerts; i++) {
+      VertexData[i * VERT_SIZE + 0] = x;
+      VertexData[i * VERT_SIZE + 1] = y;
+      VertexData[i * VERT_SIZE + 2] = 0.0;
+      VertexData[i * VERT_SIZE + 3] = 1.0;
+      x += dx;
+      if (x > 1.0) {
+         x = -1.0;
+         y += dy;
+      }
+   }
+
+   Elements = (GLuint *) malloc(NumVerts * sizeof(GLuint));
+
+   for (i = 0; i < NumVerts; i++) {
+      Elements[i] = NumVerts - i - 1;
+   }
+}
+
+
+/** Called from test harness/main */
+void
+PerfInit(void)
+{
+   InitializeVertexData();
+
+   /* setup VertexBO */
+   glGenBuffersARB(1, &VertexBO);
+   glBindBufferARB(GL_ARRAY_BUFFER_ARB, VertexBO);
+   glBufferDataARB(GL_ARRAY_BUFFER_ARB,
+                   NumVerts * VertBytes, VertexData, GL_STATIC_DRAW_ARB);
+   glEnableClientState(GL_VERTEX_ARRAY);
+
+   /* setup ElementBO */
+   glGenBuffersARB(1, &ElementBO);
+   glBindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, ElementBO);
+   glBufferDataARB(GL_ELEMENT_ARRAY_BUFFER_ARB,
+                   NumElements * sizeof(GLuint), Elements, GL_STATIC_DRAW_ARB);
+}
+
+
+static void
+DrawImmediate(unsigned count)
+{
+   unsigned i;
+   glBindBufferARB(GL_ELEMENT_ARRAY_BUFFER, 0);
+   glBindBufferARB(GL_ARRAY_BUFFER, 0);
+   for (i = 0; i < count; i++) {
+      unsigned j;
+      glBegin(GL_POINTS);
+      for (j = 0; j < NumVerts; j++) {
+#if VERT_SIZE == 4
+         glVertex4fv(VertexData + j * 4);
+#elif VERT_SIZE == 3
+         glVertex3fv(VertexData + j * 3);
+#elif VERT_SIZE == 2
+         glVertex2fv(VertexData + j * 2);
+#else
+         abort();
+#endif
+      }
+      glEnd();
+   }
+   glFinish();
+   PerfSwapBuffers();
+}
+
+
+static void
+DrawArraysMem(unsigned count)
+{
+   unsigned i;
+   glBindBufferARB(GL_ELEMENT_ARRAY_BUFFER, 0);
+   glBindBufferARB(GL_ARRAY_BUFFER, 0);
+   glVertexPointer(VERT_SIZE, GL_FLOAT, VertBytes, VertexData);
+   for (i = 0; i < count; i++) {
+      glDrawArrays(GL_POINTS, 0, NumVerts);
+   }
+   glFinish();
+   PerfSwapBuffers();
+}
+
+
+static void
+DrawArraysVBO(unsigned count)
+{
+   unsigned i;
+   glBindBufferARB(GL_ELEMENT_ARRAY_BUFFER, 0);
+   glBindBufferARB(GL_ARRAY_BUFFER, VertexBO);
+   glVertexPointer(VERT_SIZE, GL_FLOAT, VertBytes, (void *) 0);
+   for (i = 0; i < count; i++) {
+      glDrawArrays(GL_POINTS, 0, NumVerts);
+   }
+   glFinish();
+   PerfSwapBuffers();
+}
+
+
+static void
+DrawElementsMem(unsigned count)
+{
+   unsigned i;
+   glBindBufferARB(GL_ELEMENT_ARRAY_BUFFER, 0);
+   glBindBufferARB(GL_ARRAY_BUFFER, 0);
+   glVertexPointer(VERT_SIZE, GL_FLOAT, VertBytes, VertexData);
+   for (i = 0; i < count; i++) {
+      glDrawElements(GL_POINTS, NumVerts, GL_UNSIGNED_INT, Elements);
+   }
+   glFinish();
+   PerfSwapBuffers();
+}
+
+
+static void
+DrawElementsBO(unsigned count)
+{
+   unsigned i;
+   glBindBufferARB(GL_ELEMENT_ARRAY_BUFFER, ElementBO);
+   glBindBufferARB(GL_ARRAY_BUFFER, VertexBO);
+   glVertexPointer(VERT_SIZE, GL_FLOAT, VertBytes, (void *) 0);
+   for (i = 0; i < count; i++) {
+      glDrawElements(GL_POINTS, NumVerts, GL_UNSIGNED_INT, (void *) 0);
+   }
+   glFinish();
+   PerfSwapBuffers();
+}
+
+
+static void
+DrawRangeElementsMem(unsigned count)
+{
+   unsigned i;
+   glBindBufferARB(GL_ELEMENT_ARRAY_BUFFER, 0);
+   glBindBufferARB(GL_ARRAY_BUFFER, 0);
+   glVertexPointer(VERT_SIZE, GL_FLOAT, VertBytes, VertexData);
+   for (i = 0; i < count; i++) {
+      glDrawRangeElements(GL_POINTS, 0, NumVerts - 1,
+                          NumVerts, GL_UNSIGNED_INT, Elements);
+   }
+   glFinish();
+   PerfSwapBuffers();
+}
+
+
+static void
+DrawRangeElementsBO(unsigned count)
+{
+   unsigned i;
+   glBindBufferARB(GL_ELEMENT_ARRAY_BUFFER, ElementBO);
+   glBindBufferARB(GL_ARRAY_BUFFER, VertexBO);
+   glVertexPointer(VERT_SIZE, GL_FLOAT, VertBytes, (void *) 0);
+   for (i = 0; i < count; i++) {
+      glDrawRangeElements(GL_POINTS, 0, NumVerts - 1,
+                          NumVerts, GL_UNSIGNED_INT, (void *) 0);
+   }
+   glFinish();
+   PerfSwapBuffers();
+}
+
+
+/** Called from test harness/main */
+void
+PerfDraw(void)
+{
+   double rate;
+
+   glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
+
+   printf("Vertex rate (%d x Vertex%df)\n", NumVerts, VERT_SIZE);
+
+   rate = PerfMeasureRate(DrawImmediate);
+   rate *= NumVerts;
+   printf("  Immediate mode: %.1f verts/sec\n", rate);
+
+   rate = PerfMeasureRate(DrawArraysMem);
+   rate *= NumVerts;
+   printf("  glDrawArrays: %.1f verts/sec\n", rate);
+
+   rate = PerfMeasureRate(DrawArraysVBO);
+   rate *= NumVerts;
+   printf("  VBO glDrawArrays: %.1f verts/sec\n", rate);
+
+   rate = PerfMeasureRate(DrawElementsMem);
+   rate *= NumVerts;
+   printf("  glDrawElements: %.1f verts/sec\n", rate);
+
+   rate = PerfMeasureRate(DrawElementsBO);
+   rate *= NumVerts;
+   printf("  VBO glDrawElements: %.1f verts/sec\n", rate);
+
+   rate = PerfMeasureRate(DrawRangeElementsMem);
+   rate *= NumVerts;
+   printf("  glDrawRangeElements: %.1f verts/sec\n", rate);
+
+   rate = PerfMeasureRate(DrawRangeElementsBO);
+   rate *= NumVerts;
+   printf("  VBO glDrawRangeElements: %.1f verts/sec\n", rate);
+
+   exit(0);
+}

progs/tests/zreaddraw.c

@@ -13,14 +13,16 @@
 
 static GLint WinWidth = 500, WinHeight = 500;
 static GLboolean Invert = GL_FALSE;
+static GLboolean TestPacking = GL_FALSE;
+static GLboolean TestList = GL_FALSE;
 
 
 static void Display(void)
 {
-   GLfloat depth[100 * 100];
-   GLfloat depth2[400 * 400];
-   GLfloat min, max;
-   int i;
+   GLfloat depth[100 * 100 * 2];
+   GLfloat depth2[400 * 400]; /* *2 to test pixelstore stuff */
+   GLuint list;
+   GLenum depthType = GL_FLOAT;
 
    glClearColor(0.5, 0.5, 0.5, 1.0);
    glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
@@ -36,16 +38,32 @@ static void Display(void)
    glLoadIdentity();
    glutSolidSphere(1.0, 20, 10);
 
-   /* read the depth image */
-   glReadPixels(0, 0, 100, 100, GL_DEPTH_COMPONENT, GL_FLOAT, depth);
-   min = max = depth[0];
-   for (i = 1; i < 100 * 100; i++) {
-      if (depth[i] < min)
-         min = depth[i];
-      if (depth[i] > max)
-         max = depth[i];
+   if (TestPacking) {
+      glPixelStorei(GL_PACK_ROW_LENGTH, 120);
+      glPixelStorei(GL_PACK_SKIP_PIXELS, 5);
+   }
+
+   /* read the depth image */
+   glReadPixels(0, 0, 100, 100, GL_DEPTH_COMPONENT, depthType, depth);
+   if (depthType == GL_FLOAT) {
+      GLfloat min, max;
+      int i;
+      min = max = depth[0];
+      for (i = 1; i < 100 * 100; i++) {
+         if (depth[i] < min)
+            min = depth[i];
+         if (depth[i] > max)
+            max = depth[i];
+      }
+      printf("Depth value range: [%f, %f]\n", min, max);
+   }
+
+   if (TestPacking) {
+      glPixelStorei(GL_PACK_ROW_LENGTH, 0);
+      glPixelStorei(GL_PACK_SKIP_PIXELS, 0);
+      glPixelStorei(GL_UNPACK_ROW_LENGTH, 120);
+      glPixelStorei(GL_UNPACK_SKIP_PIXELS, 5);
    }
-   printf("Depth value range: [%f, %f]\n", min, max);
 
    /* draw depth image with scaling (into z buffer) */
    glPixelZoom(4.0, 4.0);
@@ -55,12 +73,27 @@ static void Display(void)
       glPixelTransferf(GL_DEPTH_SCALE, -1.0);
       glPixelTransferf(GL_DEPTH_BIAS, 1.0);
    }
-   glDrawPixels(100, 100, GL_DEPTH_COMPONENT, GL_FLOAT, depth);
+   if (TestList) {
+      list = glGenLists(1);
+      glNewList(list, GL_COMPILE);
+      glDrawPixels(100, 100, GL_DEPTH_COMPONENT, depthType, depth);
+      glEndList();
+      glCallList(list);
+      glDeleteLists(list, 1);
+   }
+   else {
+      glDrawPixels(100, 100, GL_DEPTH_COMPONENT, depthType, depth);
+   }
    if (Invert) {
       glPixelTransferf(GL_DEPTH_SCALE, 1.0);
       glPixelTransferf(GL_DEPTH_BIAS, 0.0);
    }
 
+   if (TestPacking) {
+      glPixelStorei(GL_UNPACK_ROW_LENGTH, 0);
+      glPixelStorei(GL_UNPACK_SKIP_PIXELS, 0);
+   }
+
    glDisable(GL_DEPTH_TEST);
 
    /* read back scaled depth image */
@@ -89,6 +122,14 @@ static void Key(unsigned char key, int x, int y)
       case 'i':
          Invert = !Invert;
          break;
+      case 'p':
+         TestPacking = !TestPacking;
+         printf("Test pixel pack/unpack: %d\n", TestPacking);
+         break;
+      case 'l':
+         TestList = !TestList;
+         printf("Test dlist: %d\n", TestList);
+         break;
       case 27:
          exit(0);
          break;

progs/vp/vp-tris.c

@@ -119,6 +119,12 @@ static void Init( void )
       glBindProgramARB(GL_VERTEX_PROGRAM_ARB, prognum);
       glProgramStringARB(GL_VERTEX_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB,
 		        sz, (const GLubyte *) buf);
+      if (glGetError()) {
+         printf("Program failed to compile:\n%s\n", buf);
+         printf("Error: %s\n",
+                (char *) glGetString(GL_PROGRAM_ERROR_STRING_ARB));
+         exit(1);
+      }
       assert(glIsProgramARB(prognum));
    }
 

scons/dxsdk.py

@@ -52,11 +52,20 @@ def generate(env):
         target_cpu = 'x64'
     else:
         raise SCons.Errors.InternalError, "Unsupported target machine"
-    include_dir = 'Include'
+
+    include_dir = os.path.join(dxsdk_root, 'Include')
+    lib_dir = os.path.join(dxsdk_root, 'Lib', target_cpu)
 
     env.Append(CPPDEFINES = [('HAVE_DXSDK', '1')])
-    env.Prepend(CPPPATH = [os.path.join(dxsdk_root, 'Include')])
-    env.Prepend(LIBPATH = [os.path.join(dxsdk_root, 'Lib', target_cpu)])
+
+    gcc = 'gcc' in os.path.basename(env['CC']).split('-')
+    if gcc:
+        # Make GCC more forgiving towards Microsoft's headers
+        env.Prepend(CPPFLAGS = ['-isystem', include_dir])
+    else:
+        env.Prepend(CPPPATH = [include_dir])
+
+    env.Prepend(LIBPATH = [lib_dir])
 
 def exists(env):
     return get_dxsdk_root(env) is not None

scons/llvm.py

@@ -51,7 +51,9 @@ def generate(env):
 
         llvm_bin_dir = os.path.join(llvm_dir, llvm_subdir, 'bin')
         if not os.path.isdir(llvm_bin_dir):
-            raise SCons.Errors.InternalError, "LLVM build directory not found"
+            llvm_bin_dir = os.path.join(llvm_dir, 'bin')
+            if not os.path.isdir(llvm_bin_dir):
+                raise SCons.Errors.InternalError, "LLVM binary directory not found"
 
         env.PrependENVPath('PATH', llvm_bin_dir)
 
@@ -65,6 +67,8 @@ def generate(env):
         except OSError:
             print 'llvm-config version %s failed' % version
         else:
+            if env['platform'] == 'windows':
+                env.Append(LIBS = ['imagehlp', 'psapi'])
             env['LINK'] = env['CXX']
             env['LLVM_VERSION'] = version
 

scons/winddk.py

@@ -85,8 +85,6 @@ def get_winddk_paths(env, version, root):
         else:
             # TODO: take in consideration the host cpu
             bin_dir = os.path.join(root, 'bin', 'win64', 'x86', cpu_bin(target_cpu))
-
-    env.PrependENVPath('PATH', [bin_dir])
     
     crt_inc_dir = os.path.join(root, 'inc', 'crt')
     if version_major >= 6000:
@@ -98,17 +96,33 @@ def get_winddk_paths(env, version, root):
         sdk_inc_dir = os.path.join(root, 'inc', target_os)
         wdm_inc_dir = os.path.join(root, 'inc', 'ddk', 'wdm', target_os)
 
-    env.PrependENVPath('INCLUDE', [
-        wdm_inc_dir,
-        ddk_inc_dir,
-        crt_inc_dir,
-        sdk_inc_dir,
-    ])
+    if env['toolchain'] == 'winddk':
+        env.PrependENVPath('PATH', [bin_dir])
+        env.PrependENVPath('INCLUDE', [
+            wdm_inc_dir,
+            ddk_inc_dir,
+            crt_inc_dir,
+            sdk_inc_dir,
+        ])
+        env.PrependENVPath('LIB', [
+            os.path.join(root, 'lib', 'crt', target_cpu),
+            os.path.join(root, 'lib', target_os, target_cpu),
+        ])
+    elif env['toolchain'] == 'crossmingw':
+        env.Prepend(CPPFLAGS = [
+            '-isystem', ddk_inc_dir,
+            '-isystem', sdk_inc_dir,
+        ])
+    else:
+        env.Prepend(CPPPATH = [
+            wdm_inc_dir,
+            ddk_inc_dir,
+            sdk_inc_dir,
+        ])
+        env.Prepend(LIBPATH = [
+            os.path.join(root, 'lib', target_os, target_cpu),
+        ])
 
-    env.PrependENVPath('LIB', [
-        os.path.join(root, 'lib', 'crt', target_cpu),
-        os.path.join(root, 'lib', target_os, target_cpu),
-    ])
 
 def generate(env):
     if not env.has_key('ENV'):
@@ -120,9 +134,10 @@ def generate(env):
             get_winddk_paths(env, version, root)
             break
 
-    msvc_sa.generate(env)
-    mslib_sa.generate(env)
-    mslink_sa.generate(env)
+    if env['toolchain'] == 'winddk':
+        msvc_sa.generate(env)
+        mslib_sa.generate(env)
+        mslink_sa.generate(env)
 
 def exists(env):
     for version in versions:

src/gallium/auxiliary/util/u_format.csv

@@ -2,7 +2,7 @@ PIPE_FORMAT_A8R8G8B8_UNORM        , arith , 1, 1, un8 , un8 , un8 , un8 , zyxw,
 PIPE_FORMAT_X8R8G8B8_UNORM        , arith , 1, 1, un8 , un8 , un8 , un8 , zyx1, rgb
 PIPE_FORMAT_B8G8R8A8_UNORM        , arith , 1, 1, un8 , un8 , un8 , un8 , yzwx, rgb
 PIPE_FORMAT_B8G8R8X8_UNORM        , arith , 1, 1, un8 , un8 , un8 , un8 , yzw1, rgb
-PIPE_FORMAT_A1R5G5B5_UNORM        , arith , 1, 1, un1 , un5 , un5 , un5 , zyxw, rgb
+PIPE_FORMAT_A1R5G5B5_UNORM        , arith , 1, 1, un5 , un5 , un5 , un1 , zyxw, rgb
 PIPE_FORMAT_A4R4G4B4_UNORM        , arith , 1, 1, un4 , un4 , un4 , un4 , zyxw, rgb
 PIPE_FORMAT_R5G6B5_UNORM          , arith , 1, 1, un5 , un6 , un5 ,     , zyx1, rgb
 PIPE_FORMAT_A2B10G10R10_UNORM     , arith , 1, 1, un10, un10, un10, un2 , xyzw, rgb

src/gallium/auxiliary/util/u_math.h

@@ -354,7 +354,9 @@ util_is_pot(unsigned x)
  * Find first bit set in word.  Least significant bit is 1.
  * Return 0 if no bits set.
  */
-#if defined(_MSC_VER) && _MSC_VER >= 1300
+#if defined(_MSC_VER) && _MSC_VER >= 1300 && (_M_IX86 || _M_AMD64 || _M_IA64)
+unsigned char _BitScanForward(unsigned long* Index, unsigned long Mask);
+#pragma intrinsic(_BitScanForward)
 static INLINE
 unsigned long ffs( unsigned long u )
 {

src/gallium/drivers/cell/ppu/cell_screen.c

@@ -64,8 +64,6 @@ cell_get_param(struct pipe_screen *screen, int param)
       return 1;
    case PIPE_CAP_GLSL:
       return 1;
-   case PIPE_CAP_S3TC:
-      return 0;
    case PIPE_CAP_ANISOTROPIC_FILTER:
       return 0;
    case PIPE_CAP_POINT_SPRITE:

src/gallium/drivers/i915simple/i915_screen.c

@@ -101,8 +101,6 @@ i915_get_param(struct pipe_screen *screen, int param)
       return 1;
    case PIPE_CAP_GLSL:
       return 0;
-   case PIPE_CAP_S3TC:
-      return 0;
    case PIPE_CAP_ANISOTROPIC_FILTER:
       return 0;
    case PIPE_CAP_POINT_SPRITE:

src/gallium/drivers/i965simple/brw_screen.c

@@ -85,8 +85,6 @@ brw_get_param(struct pipe_screen *screen, int param)
       return 1;
    case PIPE_CAP_GLSL:
       return 0;
-   case PIPE_CAP_S3TC:
-      return 0;
    case PIPE_CAP_ANISOTROPIC_FILTER:
       return 0;
    case PIPE_CAP_POINT_SPRITE:

src/gallium/drivers/llvmpipe/Makefile

@@ -3,6 +3,8 @@ include $(TOP)/configs/current
 
 LIBNAME = llvmpipe
 
+CFLAGS += -D__STDC_CONSTANT_MACROS -D__STDC_LIMIT_MACROS
+
 C_SOURCES = \
 	lp_bld_alpha.c \
 	lp_bld_arit.c \

src/gallium/drivers/llvmpipe/SConscript

@@ -3,7 +3,7 @@ Import('*')
 env = env.Clone()
 
 env.Tool('llvm')
-if env.has_key('LLVM_VERSION') is False:
+if not env.has_key('LLVM_VERSION'):
     print 'warning: LLVM not found: not building llvmpipe'
     Return()
 

src/gallium/drivers/llvmpipe/lp_bld_alpha.c

@@ -45,7 +45,7 @@
 void
 lp_build_alpha_test(LLVMBuilderRef builder,
                     const struct pipe_alpha_state *state,
-                    union lp_type type,
+                    struct lp_type type,
                     struct lp_build_mask_context *mask,
                     LLVMValueRef alpha,
                     LLVMValueRef ref)

src/gallium/drivers/llvmpipe/lp_bld_alpha.h

@@ -38,14 +38,14 @@
 #include <llvm-c/Core.h>  
 
 struct pipe_alpha_state;
-union lp_type;
+struct lp_type;
 struct lp_build_mask_context;
 
 
 void
 lp_build_alpha_test(LLVMBuilderRef builder,
                     const struct pipe_alpha_state *state,
-                    union lp_type type,
+                    struct lp_type type,
                     struct lp_build_mask_context *mask,
                     LLVMValueRef alpha,
                     LLVMValueRef ref);

src/gallium/drivers/llvmpipe/lp_bld_arit.c

@@ -65,7 +65,7 @@ lp_build_min_simple(struct lp_build_context *bld,
                     LLVMValueRef a,
                     LLVMValueRef b)
 {
-   const union lp_type type = bld->type;
+   const struct lp_type type = bld->type;
    const char *intrinsic = NULL;
    LLVMValueRef cond;
 
@@ -113,7 +113,7 @@ lp_build_max_simple(struct lp_build_context *bld,
                     LLVMValueRef a,
                     LLVMValueRef b)
 {
-   const union lp_type type = bld->type;
+   const struct lp_type type = bld->type;
    const char *intrinsic = NULL;
    LLVMValueRef cond;
 
@@ -159,7 +159,7 @@ LLVMValueRef
 lp_build_comp(struct lp_build_context *bld,
               LLVMValueRef a)
 {
-   const union lp_type type = bld->type;
+   const struct lp_type type = bld->type;
 
    if(a == bld->one)
       return bld->zero;
@@ -188,7 +188,7 @@ lp_build_add(struct lp_build_context *bld,
              LLVMValueRef a,
              LLVMValueRef b)
 {
-   const union lp_type type = bld->type;
+   const struct lp_type type = bld->type;
    LLVMValueRef res;
 
    if(a == bld->zero)
@@ -241,7 +241,7 @@ lp_build_sub(struct lp_build_context *bld,
              LLVMValueRef a,
              LLVMValueRef b)
 {
-   const union lp_type type = bld->type;
+   const struct lp_type type = bld->type;
    LLVMValueRef res;
 
    if(b == bld->zero)
@@ -405,7 +405,7 @@ lp_build_mul(struct lp_build_context *bld,
              LLVMValueRef a,
              LLVMValueRef b)
 {
-   const union lp_type type = bld->type;
+   const struct lp_type type = bld->type;
 
    if(a == bld->zero)
       return bld->zero;
@@ -477,7 +477,7 @@ lp_build_div(struct lp_build_context *bld,
              LLVMValueRef a,
              LLVMValueRef b)
 {
-   const union lp_type type = bld->type;
+   const struct lp_type type = bld->type;
 
    if(a == bld->zero)
       return bld->zero;
@@ -590,7 +590,7 @@ LLVMValueRef
 lp_build_abs(struct lp_build_context *bld,
              LLVMValueRef a)
 {
-   const union lp_type type = bld->type;
+   const struct lp_type type = bld->type;
    LLVMTypeRef vec_type = lp_build_vec_type(type);
 
    if(!type.sign)
@@ -627,7 +627,7 @@ LLVMValueRef
 lp_build_sgn(struct lp_build_context *bld,
              LLVMValueRef a)
 {
-   const union lp_type type = bld->type;
+   const struct lp_type type = bld->type;
    LLVMTypeRef vec_type = lp_build_vec_type(type);
    LLVMValueRef cond;
    LLVMValueRef res;
@@ -678,7 +678,7 @@ lp_build_round_sse41(struct lp_build_context *bld,
                      LLVMValueRef a,
                      enum lp_build_round_sse41_mode mode)
 {
-   const union lp_type type = bld->type;
+   const struct lp_type type = bld->type;
    LLVMTypeRef vec_type = lp_build_vec_type(type);
    const char *intrinsic;
 
@@ -706,7 +706,7 @@ LLVMValueRef
 lp_build_round(struct lp_build_context *bld,
                LLVMValueRef a)
 {
-   const union lp_type type = bld->type;
+   const struct lp_type type = bld->type;
 
    assert(type.floating);
 
@@ -724,7 +724,7 @@ LLVMValueRef
 lp_build_floor(struct lp_build_context *bld,
                LLVMValueRef a)
 {
-   const union lp_type type = bld->type;
+   const struct lp_type type = bld->type;
 
    assert(type.floating);
 
@@ -742,7 +742,7 @@ LLVMValueRef
 lp_build_ceil(struct lp_build_context *bld,
               LLVMValueRef a)
 {
-   const union lp_type type = bld->type;
+   const struct lp_type type = bld->type;
 
    assert(type.floating);
 
@@ -760,7 +760,7 @@ LLVMValueRef
 lp_build_trunc(struct lp_build_context *bld,
                LLVMValueRef a)
 {
-   const union lp_type type = bld->type;
+   const struct lp_type type = bld->type;
 
    assert(type.floating);
 
@@ -782,7 +782,7 @@ LLVMValueRef
 lp_build_int(struct lp_build_context *bld,
              LLVMValueRef a)
 {
-   const union lp_type type = bld->type;
+   const struct lp_type type = bld->type;
    LLVMTypeRef int_vec_type = lp_build_int_vec_type(type);
 
    assert(type.floating);
@@ -805,7 +805,7 @@ LLVMValueRef
 lp_build_sqrt(struct lp_build_context *bld,
               LLVMValueRef a)
 {
-   const union lp_type type = bld->type;
+   const struct lp_type type = bld->type;
    LLVMTypeRef vec_type = lp_build_vec_type(type);
    char intrinsic[32];
 
@@ -823,7 +823,7 @@ LLVMValueRef
 lp_build_rcp(struct lp_build_context *bld,
              LLVMValueRef a)
 {
-   const union lp_type type = bld->type;
+   const struct lp_type type = bld->type;
 
    if(a == bld->zero)
       return bld->undef;
@@ -854,7 +854,7 @@ LLVMValueRef
 lp_build_rsqrt(struct lp_build_context *bld,
                LLVMValueRef a)
 {
-   const union lp_type type = bld->type;
+   const struct lp_type type = bld->type;
 
    assert(type.floating);
 
@@ -875,7 +875,7 @@ LLVMValueRef
 lp_build_cos(struct lp_build_context *bld,
               LLVMValueRef a)
 {
-   const union lp_type type = bld->type;
+   const struct lp_type type = bld->type;
    LLVMTypeRef vec_type = lp_build_vec_type(type);
    char intrinsic[32];
 
@@ -895,7 +895,7 @@ LLVMValueRef
 lp_build_sin(struct lp_build_context *bld,
               LLVMValueRef a)
 {
-   const union lp_type type = bld->type;
+   const struct lp_type type = bld->type;
    LLVMTypeRef vec_type = lp_build_vec_type(type);
    char intrinsic[32];
 
@@ -966,7 +966,7 @@ lp_build_polynomial(struct lp_build_context *bld,
                     const double *coeffs,
                     unsigned num_coeffs)
 {
-   const union lp_type type = bld->type;
+   const struct lp_type type = bld->type;
    LLVMValueRef res = NULL;
    unsigned i;
 
@@ -1014,7 +1014,7 @@ lp_build_exp2_approx(struct lp_build_context *bld,
                      LLVMValueRef *p_frac_part,
                      LLVMValueRef *p_exp2)
 {
-   const union lp_type type = bld->type;
+   const struct lp_type type = bld->type;
    LLVMTypeRef vec_type = lp_build_vec_type(type);
    LLVMTypeRef int_vec_type = lp_build_int_vec_type(type);
    LLVMValueRef ipart = NULL;
@@ -1107,7 +1107,7 @@ lp_build_log2_approx(struct lp_build_context *bld,
                      LLVMValueRef *p_floor_log2,
                      LLVMValueRef *p_log2)
 {
-   const union lp_type type = bld->type;
+   const struct lp_type type = bld->type;
    LLVMTypeRef vec_type = lp_build_vec_type(type);
    LLVMTypeRef int_vec_type = lp_build_int_vec_type(type);
 

src/gallium/drivers/llvmpipe/lp_bld_arit.h

@@ -40,7 +40,7 @@
 #include <llvm-c/Core.h>  
 
 
-union lp_type type;
+struct lp_type type;
 struct lp_build_context;
 
 

src/gallium/drivers/llvmpipe/lp_bld_blend.h

@@ -46,7 +46,7 @@
 
 
 struct pipe_blend_state;
-union lp_type;
+struct lp_type;
 struct lp_build_context;
 
 
@@ -74,7 +74,7 @@ lp_build_blend_func(struct lp_build_context *bld,
 LLVMValueRef
 lp_build_blend_aos(LLVMBuilderRef builder,
                    const struct pipe_blend_state *blend,
-                   union lp_type type,
+                   struct lp_type type,
                    LLVMValueRef src,
                    LLVMValueRef dst,
                    LLVMValueRef const_,
@@ -84,7 +84,7 @@ lp_build_blend_aos(LLVMBuilderRef builder,
 void
 lp_build_blend_soa(LLVMBuilderRef builder,
                    const struct pipe_blend_state *blend,
-                   union lp_type type,
+                   struct lp_type type,
                    LLVMValueRef src[4],
                    LLVMValueRef dst[4],
                    LLVMValueRef const_[4],

src/gallium/drivers/llvmpipe/lp_bld_blend_aos.c

@@ -303,7 +303,7 @@ lp_build_blend_func(struct lp_build_context *bld,
 LLVMValueRef
 lp_build_blend_aos(LLVMBuilderRef builder,
                    const struct pipe_blend_state *blend,
-                   union lp_type type,
+                   struct lp_type type,
                    LLVMValueRef src,
                    LLVMValueRef dst,
                    LLVMValueRef const_,

src/gallium/drivers/llvmpipe/lp_bld_blend_soa.c

@@ -199,7 +199,7 @@ lp_build_blend_soa_factor(struct lp_build_blend_soa_context *bld,
 void
 lp_build_blend_soa(LLVMBuilderRef builder,
                    const struct pipe_blend_state *blend,
-                   union lp_type type,
+                   struct lp_type type,
                    LLVMValueRef src[4],
                    LLVMValueRef dst[4],
                    LLVMValueRef con[4],

src/gallium/drivers/llvmpipe/lp_bld_const.c

@@ -42,7 +42,7 @@
 
 
 unsigned
-lp_mantissa(union lp_type type)
+lp_mantissa(struct lp_type type)
 {
    assert(type.floating);
 
@@ -72,7 +72,7 @@ lp_mantissa(union lp_type type)
  * Same as lp_const_scale(), but in terms of shifts.
  */
 unsigned
-lp_const_shift(union lp_type type)
+lp_const_shift(struct lp_type type)
 {
    if(type.floating)
       return 0;
@@ -86,7 +86,7 @@ lp_const_shift(union lp_type type)
 
 
 unsigned
-lp_const_offset(union lp_type type)
+lp_const_offset(struct lp_type type)
 {
    if(type.floating || type.fixed)
       return 0;
@@ -104,7 +104,7 @@ lp_const_offset(union lp_type type)
  * else for the fixed points types and normalized integers.
  */
 double
-lp_const_scale(union lp_type type)
+lp_const_scale(struct lp_type type)
 {
    unsigned long long llscale;
    double dscale;
@@ -122,7 +122,7 @@ lp_const_scale(union lp_type type)
  * Minimum value representable by the type.
  */
 double
-lp_const_min(union lp_type type)
+lp_const_min(struct lp_type type)
 {
    unsigned bits;
 
@@ -158,7 +158,7 @@ lp_const_min(union lp_type type)
  * Maximum value representable by the type.
  */
 double
-lp_const_max(union lp_type type)
+lp_const_max(struct lp_type type)
 {
    unsigned bits;
 
@@ -190,7 +190,7 @@ lp_const_max(union lp_type type)
 
 
 double
-lp_const_eps(union lp_type type)
+lp_const_eps(struct lp_type type)
 {
    if (type.floating) {
       switch(type.width) {
@@ -211,7 +211,7 @@ lp_const_eps(union lp_type type)
 
 
 LLVMValueRef
-lp_build_undef(union lp_type type)
+lp_build_undef(struct lp_type type)
 {
    LLVMTypeRef vec_type = lp_build_vec_type(type);
    return LLVMGetUndef(vec_type);
@@ -219,7 +219,7 @@ lp_build_undef(union lp_type type)
                
 
 LLVMValueRef
-lp_build_zero(union lp_type type)
+lp_build_zero(struct lp_type type)
 {
    LLVMTypeRef vec_type = lp_build_vec_type(type);
    return LLVMConstNull(vec_type);
@@ -227,7 +227,7 @@ lp_build_zero(union lp_type type)
                
 
 LLVMValueRef
-lp_build_one(union lp_type type)
+lp_build_one(struct lp_type type)
 {
    LLVMTypeRef elem_type;
    LLVMValueRef elems[LP_MAX_VECTOR_LENGTH];
@@ -269,7 +269,7 @@ lp_build_one(union lp_type type)
                
 
 LLVMValueRef
-lp_build_const_scalar(union lp_type type,
+lp_build_const_scalar(struct lp_type type,
                       double val)
 {
    LLVMTypeRef elem_type = lp_build_elem_type(type);
@@ -295,7 +295,7 @@ lp_build_const_scalar(union lp_type type,
 
 
 LLVMValueRef
-lp_build_int_const_scalar(union lp_type type,
+lp_build_int_const_scalar(struct lp_type type,
                           long long val)
 {
    LLVMTypeRef elem_type = lp_build_int_elem_type(type);
@@ -312,7 +312,7 @@ lp_build_int_const_scalar(union lp_type type,
 
 
 LLVMValueRef
-lp_build_const_aos(union lp_type type, 
+lp_build_const_aos(struct lp_type type, 
                    double r, double g, double b, double a, 
                    const unsigned char *swizzle)
 {
@@ -352,8 +352,8 @@ lp_build_const_aos(union lp_type type,
 
 
 LLVMValueRef
-lp_build_const_mask_aos(union lp_type type,
-                        boolean cond[4])
+lp_build_const_mask_aos(struct lp_type type,
+                        const boolean cond[4])
 {
    LLVMTypeRef elem_type = LLVMIntType(type.width);
    LLVMValueRef masks[LP_MAX_VECTOR_LENGTH];

src/gallium/drivers/llvmpipe/lp_bld_const.h

@@ -42,67 +42,67 @@
 #include <pipe/p_compiler.h>
 
 
-union lp_type type;
+struct lp_type type;
 
 
 unsigned
-lp_mantissa(union lp_type type);
+lp_mantissa(struct lp_type type);
 
 
 unsigned
-lp_const_shift(union lp_type type);
+lp_const_shift(struct lp_type type);
 
 
 unsigned
-lp_const_offset(union lp_type type);
+lp_const_offset(struct lp_type type);
 
 
 double
-lp_const_scale(union lp_type type);
+lp_const_scale(struct lp_type type);
 
 double
-lp_const_min(union lp_type type);
+lp_const_min(struct lp_type type);
 
 
 double
-lp_const_max(union lp_type type);
+lp_const_max(struct lp_type type);
 
 
 double
-lp_const_eps(union lp_type type);
+lp_const_eps(struct lp_type type);
 
 
 LLVMValueRef
-lp_build_undef(union lp_type type);
+lp_build_undef(struct lp_type type);
 
 
 LLVMValueRef
-lp_build_zero(union lp_type type);
+lp_build_zero(struct lp_type type);
 
 
 LLVMValueRef
-lp_build_one(union lp_type type);
+lp_build_one(struct lp_type type);
 
 
 LLVMValueRef
-lp_build_const_scalar(union lp_type type,
+lp_build_const_scalar(struct lp_type type,
                       double val);
 
 
 LLVMValueRef
-lp_build_int_const_scalar(union lp_type type,
+lp_build_int_const_scalar(struct lp_type type,
                           long long val);
 
 
 LLVMValueRef
-lp_build_const_aos(union lp_type type, 
+lp_build_const_aos(struct lp_type type, 
                    double r, double g, double b, double a, 
                    const unsigned char *swizzle);
 
 
 LLVMValueRef
-lp_build_const_mask_aos(union lp_type type,
-                        boolean cond[4]);
+lp_build_const_mask_aos(struct lp_type type,
+                        const boolean cond[4]);
 
 
 #endif /* !LP_BLD_CONST_H */

src/gallium/drivers/llvmpipe/lp_bld_conv.c

@@ -86,7 +86,7 @@
  */
 LLVMValueRef
 lp_build_clamped_float_to_unsigned_norm(LLVMBuilderRef builder,
-                                        union lp_type src_type,
+                                        struct lp_type src_type,
                                         unsigned dst_width,
                                         LLVMValueRef src)
 {
@@ -152,7 +152,7 @@ lp_build_clamped_float_to_unsigned_norm(LLVMBuilderRef builder,
 LLVMValueRef
 lp_build_unsigned_norm_to_float(LLVMBuilderRef builder,
                                 unsigned src_width,
-                                union lp_type dst_type,
+                                struct lp_type dst_type,
                                 LLVMValueRef src)
 {
    LLVMTypeRef vec_type = lp_build_vec_type(dst_type);
@@ -248,8 +248,8 @@ lp_build_const_pack_shuffle(unsigned n)
  */
 static void
 lp_build_expand(LLVMBuilderRef builder,
-               union lp_type src_type,
-               union lp_type dst_type,
+               struct lp_type src_type,
+               struct lp_type dst_type,
                LLVMValueRef src,
                LLVMValueRef *dst, unsigned num_dsts)
 {
@@ -266,7 +266,7 @@ lp_build_expand(LLVMBuilderRef builder,
    dst[0] = src;
 
    while(src_type.width < dst_type.width) {
-      union lp_type new_type = src_type;
+      struct lp_type new_type = src_type;
       LLVMTypeRef new_vec_type;
 
       new_type.width *= 2;
@@ -314,8 +314,8 @@ lp_build_expand(LLVMBuilderRef builder,
  */
 static LLVMValueRef
 lp_build_pack2(LLVMBuilderRef builder,
-               union lp_type src_type,
-               union lp_type dst_type,
+               struct lp_type src_type,
+               struct lp_type dst_type,
                boolean clamped,
                LLVMValueRef lo,
                LLVMValueRef hi)
@@ -392,8 +392,8 @@ lp_build_pack2(LLVMBuilderRef builder,
  */
 static LLVMValueRef
 lp_build_pack(LLVMBuilderRef builder,
-              union lp_type src_type,
-              union lp_type dst_type,
+              struct lp_type src_type,
+              struct lp_type dst_type,
               boolean clamped,
               const LLVMValueRef *src, unsigned num_srcs)
 {
@@ -410,7 +410,7 @@ lp_build_pack(LLVMBuilderRef builder,
       tmp[i] = src[i];
 
    while(src_type.width > dst_type.width) {
-      union lp_type new_type = src_type;
+      struct lp_type new_type = src_type;
 
       new_type.width /= 2;
       new_type.length *= 2;
@@ -442,12 +442,12 @@ lp_build_pack(LLVMBuilderRef builder,
  */
 void
 lp_build_conv(LLVMBuilderRef builder,
-              union lp_type src_type,
-              union lp_type dst_type,
+              struct lp_type src_type,
+              struct lp_type dst_type,
               const LLVMValueRef *src, unsigned num_srcs,
               LLVMValueRef *dst, unsigned num_dsts)
 {
-   union lp_type tmp_type;
+   struct lp_type tmp_type;
    LLVMValueRef tmp[LP_MAX_VECTOR_LENGTH];
    unsigned num_tmps;
    unsigned i;
@@ -470,7 +470,7 @@ lp_build_conv(LLVMBuilderRef builder,
     * Clamp if necessary
     */
 
-   if(src_type.value != dst_type.value) {
+   if(memcmp(&src_type, &dst_type, sizeof src_type) != 0) {
       struct lp_build_context bld;
       double src_min = lp_const_min(src_type);
       double dst_min = lp_const_min(dst_type);
@@ -656,8 +656,8 @@ lp_build_conv(LLVMBuilderRef builder,
  */
 void
 lp_build_conv_mask(LLVMBuilderRef builder,
-                   union lp_type src_type,
-                   union lp_type dst_type,
+                   struct lp_type src_type,
+                   struct lp_type dst_type,
                    const LLVMValueRef *src, unsigned num_srcs,
                    LLVMValueRef *dst, unsigned num_dsts)
 {

src/gallium/drivers/llvmpipe/lp_bld_conv.h

@@ -40,33 +40,33 @@
 #include <llvm-c/Core.h>  
 
 
-union lp_type type;
+struct lp_type type;
 
 
 LLVMValueRef
 lp_build_clamped_float_to_unsigned_norm(LLVMBuilderRef builder,
-                                        union lp_type src_type,
+                                        struct lp_type src_type,
                                         unsigned dst_width,
                                         LLVMValueRef src);
 
 LLVMValueRef
 lp_build_unsigned_norm_to_float(LLVMBuilderRef builder,
                                 unsigned src_width,
-                                union lp_type dst_type,
+                                struct lp_type dst_type,
                                 LLVMValueRef src);
 
 
 void
 lp_build_conv(LLVMBuilderRef builder,
-              union lp_type src_type,
-              union lp_type dst_type,
+              struct lp_type src_type,
+              struct lp_type dst_type,
               const LLVMValueRef *srcs, unsigned num_srcs,
               LLVMValueRef *dsts, unsigned num_dsts);
 
 void
 lp_build_conv_mask(LLVMBuilderRef builder,
-                   union lp_type src_type,
-                   union lp_type dst_type,
+                   struct lp_type src_type,
+                   struct lp_type dst_type,
                    const LLVMValueRef *src, unsigned num_srcs,
                    LLVMValueRef *dst, unsigned num_dsts);
 

src/gallium/drivers/llvmpipe/lp_bld_depth.c

@@ -71,11 +71,11 @@
 /**
  * Return a type appropriate for depth/stencil testing.
  */
-union lp_type
+struct lp_type
 lp_depth_type(const struct util_format_description *format_desc,
               unsigned length)
 {
-   union lp_type type;
+   struct lp_type type;
    unsigned swizzle;
 
    assert(format_desc->colorspace == UTIL_FORMAT_COLORSPACE_ZS);
@@ -85,7 +85,7 @@ lp_depth_type(const struct util_format_description *format_desc,
    swizzle = format_desc->swizzle[0];
    assert(swizzle < 4);
 
-   type.value = 0;
+   memset(&type, 0, sizeof type);
    type.width = format_desc->block.bits;
 
    if(format_desc->channel[swizzle].type == UTIL_FORMAT_TYPE_FLOAT) {
@@ -114,7 +114,7 @@ lp_depth_type(const struct util_format_description *format_desc,
 void
 lp_build_depth_test(LLVMBuilderRef builder,
                     const struct pipe_depth_state *state,
-                    union lp_type type,
+                    struct lp_type type,
                     const struct util_format_description *format_desc,
                     struct lp_build_mask_context *mask,
                     LLVMValueRef src,

src/gallium/drivers/llvmpipe/lp_bld_depth.h

@@ -41,11 +41,11 @@
  
 struct pipe_depth_state;
 struct util_format_description;
-union lp_type;
+struct lp_type;
 struct lp_build_mask_context;
 
 
-union lp_type
+struct lp_type
 lp_depth_type(const struct util_format_description *format_desc,
               unsigned length);
 
@@ -53,7 +53,7 @@ lp_depth_type(const struct util_format_description *format_desc,
 void
 lp_build_depth_test(LLVMBuilderRef builder,
                     const struct pipe_depth_state *state,
-                    union lp_type type,
+                    struct lp_type type,
                     const struct util_format_description *format_desc,
                     struct lp_build_mask_context *mask,
                     LLVMValueRef src,

src/gallium/drivers/llvmpipe/lp_bld_flow.c

@@ -405,7 +405,7 @@ lp_build_mask_check(struct lp_build_mask_context *mask)
 void
 lp_build_mask_begin(struct lp_build_mask_context *mask,
                     struct lp_build_flow_context *flow,
-                    union lp_type type,
+                    struct lp_type type,
                     LLVMValueRef value)
 {
    memset(mask, 0, sizeof *mask);

src/gallium/drivers/llvmpipe/lp_bld_flow.h

@@ -38,7 +38,7 @@
 #include <llvm-c/Core.h>  
 
 
-union lp_type;
+struct lp_type;
 
 
 struct lp_build_flow_context;
@@ -84,7 +84,7 @@ struct lp_build_mask_context
 void
 lp_build_mask_begin(struct lp_build_mask_context *mask,
                     struct lp_build_flow_context *flow,
-                    union lp_type type,
+                    struct lp_type type,
                     LLVMValueRef value);
 
 /**

src/gallium/drivers/llvmpipe/lp_bld_format.h

@@ -39,7 +39,7 @@
 #include "pipe/p_format.h"
 
 struct util_format_description;
-union lp_type;
+struct lp_type;
 
 
 /**
@@ -103,7 +103,7 @@ lp_build_gather(LLVMBuilderRef builder,
 void
 lp_build_unpack_rgba_soa(LLVMBuilderRef builder,
                          const struct util_format_description *format_desc,
-                         union lp_type type,
+                         struct lp_type type,
                          LLVMValueRef packed,
                          LLVMValueRef *rgba);
 
@@ -111,7 +111,7 @@ lp_build_unpack_rgba_soa(LLVMBuilderRef builder,
 void
 lp_build_load_rgba_soa(LLVMBuilderRef builder,
                        const struct util_format_description *format_desc,
-                       union lp_type type,
+                       struct lp_type type,
                        LLVMValueRef base_ptr,
                        LLVMValueRef offsets,
                        LLVMValueRef *rgba);

src/gallium/drivers/llvmpipe/lp_bld_format_soa.c

@@ -84,7 +84,7 @@ lp_build_gather(LLVMBuilderRef builder,
 
 
 static LLVMValueRef
-lp_build_format_swizzle(union lp_type type,
+lp_build_format_swizzle(struct lp_type type,
                         const LLVMValueRef *inputs,
                         enum util_format_swizzle swizzle)
 {
@@ -110,7 +110,7 @@ lp_build_format_swizzle(union lp_type type,
 void
 lp_build_unpack_rgba_soa(LLVMBuilderRef builder,
                          const struct util_format_description *format_desc,
-                         union lp_type type,
+                         struct lp_type type,
                          LLVMValueRef packed,
                          LLVMValueRef *rgba)
 {
@@ -188,7 +188,7 @@ lp_build_unpack_rgba_soa(LLVMBuilderRef builder,
 void
 lp_build_load_rgba_soa(LLVMBuilderRef builder,
                        const struct util_format_description *format_desc,
-                       union lp_type type,
+                       struct lp_type type,
                        LLVMValueRef base_ptr,
                        LLVMValueRef offsets,
                        LLVMValueRef *rgba)

src/gallium/drivers/llvmpipe/lp_bld_interp.c

@@ -292,7 +292,7 @@ void
 lp_build_interp_soa_init(struct lp_build_interp_soa_context *bld,
                          const struct tgsi_token *tokens,
                          LLVMBuilderRef builder,
-                         union lp_type type,
+                         struct lp_type type,
                          LLVMValueRef a0_ptr,
                          LLVMValueRef dadx_ptr,
                          LLVMValueRef dady_ptr,

src/gallium/drivers/llvmpipe/lp_bld_interp.h

@@ -83,7 +83,7 @@ void
 lp_build_interp_soa_init(struct lp_build_interp_soa_context *bld,
                          const struct tgsi_token *tokens,
                          LLVMBuilderRef builder,
-                         union lp_type type,
+                         struct lp_type type,
                          LLVMValueRef a0_ptr,
                          LLVMValueRef dadx_ptr,
                          LLVMValueRef dady_ptr,

src/gallium/drivers/llvmpipe/lp_bld_logic.c

@@ -45,7 +45,7 @@ lp_build_cmp(struct lp_build_context *bld,
              LLVMValueRef a,
              LLVMValueRef b)
 {
-   const union lp_type type = bld->type;
+   const struct lp_type type = bld->type;
    LLVMTypeRef vec_type = lp_build_vec_type(type);
    LLVMTypeRef int_vec_type = lp_build_int_vec_type(type);
    LLVMValueRef zeros = LLVMConstNull(int_vec_type);
@@ -301,7 +301,7 @@ lp_build_select(struct lp_build_context *bld,
                 LLVMValueRef a,
                 LLVMValueRef b)
 {
-   union lp_type type = bld->type;
+   struct lp_type type = bld->type;
    LLVMValueRef res;
 
    if(a == b)
@@ -339,7 +339,7 @@ lp_build_select_aos(struct lp_build_context *bld,
                     LLVMValueRef b,
                     const boolean cond[4])
 {
-   const union lp_type type = bld->type;
+   const struct lp_type type = bld->type;
    const unsigned n = type.length;
    unsigned i, j;
 

src/gallium/drivers/llvmpipe/lp_bld_logic.h

@@ -42,7 +42,7 @@
 #include "pipe/p_defines.h" /* For PIPE_FUNC_xxx */
 
 
-union lp_type type;
+struct lp_type type;
 struct lp_build_context;
 
 

src/gallium/drivers/llvmpipe/lp_bld_sample.h

@@ -40,7 +40,7 @@
 
 struct pipe_texture;
 struct pipe_sampler_state;
-union lp_type;
+struct lp_type;
 
 
 /**
@@ -123,7 +123,7 @@ void
 lp_build_sample_soa(LLVMBuilderRef builder,
                     const struct lp_sampler_static_state *static_state,
                     struct lp_sampler_dynamic_state *dynamic_state,
-                    union lp_type fp_type,
+                    struct lp_type fp_type,
                     unsigned unit,
                     unsigned num_coords,
                     const LLVMValueRef *coords,

src/gallium/drivers/llvmpipe/lp_bld_sample_soa.c

@@ -97,15 +97,15 @@ struct lp_build_sample_context
    const struct util_format_description *format_desc;
 
    /** Incoming coordinates type and build context */
-   union lp_type coord_type;
+   struct lp_type coord_type;
    struct lp_build_context coord_bld;
 
    /** Integer coordinates */
-   union lp_type int_coord_type;
+   struct lp_type int_coord_type;
    struct lp_build_context int_coord_bld;
 
    /** Output texels type and build context */
-   union lp_type texel_type;
+   struct lp_type texel_type;
    struct lp_build_context texel_bld;
 };
 
@@ -208,6 +208,11 @@ lp_build_sample_wrap(struct lp_build_sample_context *bld,
    case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE:
    case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER:
       /* FIXME */
+      _debug_printf("warning: failed to translate texture wrap mode %u\n", wrap_mode);
+      coord = lp_build_max(int_coord_bld, coord, int_coord_bld->zero);
+      coord = lp_build_min(int_coord_bld, coord, length_minus_one);
+      break;
+
    default:
       assert(0);
    }
@@ -337,7 +342,7 @@ void
 lp_build_sample_soa(LLVMBuilderRef builder,
                     const struct lp_sampler_static_state *static_state,
                     struct lp_sampler_dynamic_state *dynamic_state,
-                    union lp_type type,
+                    struct lp_type type,
                     unsigned unit,
                     unsigned num_coords,
                     const LLVMValueRef *coords,

src/gallium/drivers/llvmpipe/lp_bld_swizzle.c

@@ -64,7 +64,7 @@ LLVMValueRef
 lp_build_broadcast_scalar(struct lp_build_context *bld,
                           LLVMValueRef scalar)
 {
-   const union lp_type type = bld->type;
+   const struct lp_type type = bld->type;
    LLVMValueRef res;
    unsigned i;
 
@@ -83,7 +83,7 @@ lp_build_broadcast_aos(struct lp_build_context *bld,
                        LLVMValueRef a,
                        unsigned channel)
 {
-   const union lp_type type = bld->type;
+   const struct lp_type type = bld->type;
    const unsigned n = type.length;
    unsigned i, j;
 
@@ -115,7 +115,7 @@ lp_build_broadcast_aos(struct lp_build_context *bld,
        *   YY00 YY00 .... YY00
        *   YYYY YYYY .... YYYY  <= output
        */
-      union lp_type type4 = type;
+      struct lp_type type4 = type;
       const char shifts[4][2] = {
          { 1,  2},
          {-1,  2},

src/gallium/drivers/llvmpipe/lp_bld_swizzle.h

@@ -40,7 +40,7 @@
 #include <llvm-c/Core.h>  
 
 
-union lp_type type;
+struct lp_type type;
 struct lp_build_context;
 
 

src/gallium/drivers/llvmpipe/lp_bld_tgsi.h

@@ -39,7 +39,7 @@
 
 
 struct tgsi_token;
-union lp_type;
+struct lp_type;
 struct lp_build_context;
 struct lp_build_mask_context;
 
@@ -60,7 +60,7 @@ struct lp_build_sampler_soa
    void
    (*emit_fetch_texel)( struct lp_build_sampler_soa *sampler,
                         LLVMBuilderRef builder,
-                        union lp_type type,
+                        struct lp_type type,
                         unsigned unit,
                         unsigned num_coords,
                         const LLVMValueRef *coords,
@@ -72,7 +72,7 @@ struct lp_build_sampler_soa
 void
 lp_build_tgsi_soa(LLVMBuilderRef builder,
                   const struct tgsi_token *tokens,
-                  union lp_type type,
+                  struct lp_type type,
                   struct lp_build_mask_context *mask,
                   LLVMValueRef consts_ptr,
                   const LLVMValueRef *pos,

src/gallium/drivers/llvmpipe/lp_bld_tgsi_soa.c

@@ -1415,7 +1415,7 @@ emit_instruction(
 void
 lp_build_tgsi_soa(LLVMBuilderRef builder,
                   const struct tgsi_token *tokens,
-                  union lp_type type,
+                  struct lp_type type,
                   struct lp_build_mask_context *mask,
                   LLVMValueRef consts_ptr,
                   const LLVMValueRef *pos,

src/gallium/drivers/llvmpipe/lp_bld_type.c

@@ -33,7 +33,7 @@
 
 
 LLVMTypeRef
-lp_build_elem_type(union lp_type type)
+lp_build_elem_type(struct lp_type type)
 {
    if (type.floating) {
       switch(type.width) {
@@ -55,7 +55,7 @@ lp_build_elem_type(union lp_type type)
 
 
 LLVMTypeRef
-lp_build_vec_type(union lp_type type)
+lp_build_vec_type(struct lp_type type)
 {
    LLVMTypeRef elem_type = lp_build_elem_type(type);
    return LLVMVectorType(elem_type, type.length);
@@ -69,7 +69,7 @@ lp_build_vec_type(union lp_type type)
  * type and check for identity.
  */
 boolean
-lp_check_elem_type(union lp_type type, LLVMTypeRef elem_type) 
+lp_check_elem_type(struct lp_type type, LLVMTypeRef elem_type) 
 {
    LLVMTypeKind elem_kind;
 
@@ -107,7 +107,7 @@ lp_check_elem_type(union lp_type type, LLVMTypeRef elem_type)
 
 
 boolean
-lp_check_vec_type(union lp_type type, LLVMTypeRef vec_type) 
+lp_check_vec_type(struct lp_type type, LLVMTypeRef vec_type) 
 {
    LLVMTypeRef elem_type;
 
@@ -128,7 +128,7 @@ lp_check_vec_type(union lp_type type, LLVMTypeRef vec_type)
 
 
 boolean
-lp_check_value(union lp_type type, LLVMValueRef val) 
+lp_check_value(struct lp_type type, LLVMValueRef val) 
 {
    LLVMTypeRef vec_type;
 
@@ -143,25 +143,26 @@ lp_check_value(union lp_type type, LLVMValueRef val)
 
 
 LLVMTypeRef
-lp_build_int_elem_type(union lp_type type)
+lp_build_int_elem_type(struct lp_type type)
 {
    return LLVMIntType(type.width);
 }
 
 
 LLVMTypeRef
-lp_build_int_vec_type(union lp_type type)
+lp_build_int_vec_type(struct lp_type type)
 {
    LLVMTypeRef elem_type = lp_build_int_elem_type(type);
    return LLVMVectorType(elem_type, type.length);
 }
 
 
-union lp_type
-lp_int_type(union lp_type type)
+struct lp_type
+lp_int_type(struct lp_type type)
 {
-   union lp_type int_type;
-   int_type.value = 0;
+   struct lp_type int_type;
+
+   memset(&int_type, 0, sizeof int_type);
    int_type.width = type.width;
    int_type.length = type.length;
    return int_type;
@@ -171,7 +172,7 @@ lp_int_type(union lp_type type)
 void
 lp_build_context_init(struct lp_build_context *bld,
                       LLVMBuilderRef builder,
-                      union lp_type type)
+                      struct lp_type type)
 {
    bld->builder = builder;
    bld->type = type;

src/gallium/drivers/llvmpipe/lp_bld_type.h

@@ -56,58 +56,55 @@
  * on the types used for intermediate computations, such as signed vs unsigned,
  * normalized values, or fixed point.
  */
-union lp_type {
-   struct {
-      /** 
-       * Floating-point. Cannot be used with fixed. Integer numbers are
-       * represented by this zero.
-       */
-      unsigned floating:1;
+struct lp_type {
+   /**
+    * Floating-point. Cannot be used with fixed. Integer numbers are
+    * represented by this zero.
+    */
+   unsigned floating:1;
 
-      /** 
-       * Fixed-point. Cannot be used with floating. Integer numbers are
-       * represented by this zero.
-       */
-      unsigned fixed:1;
-      
-      /** 
-       * Whether it can represent negative values or not.
-       *
-       * If this is not set for floating point, it means that all values are
-       * assumed to be positive.
-       */
-      unsigned sign:1;
+   /**
+    * Fixed-point. Cannot be used with floating. Integer numbers are
+    * represented by this zero.
+    */
+   unsigned fixed:1;
 
-      /**
-       * Whether values are normalized to fit [0, 1] interval, or [-1, 1]
-       * interval for signed types.
-       *
-       * For integer types it means the representable integer range should be
-       * interpreted as the interval above.
-       *
-       * For floating and fixed point formats it means the values should be
-       * clamped to the interval above.
-       */
-      unsigned norm:1;
+   /**
+    * Whether it can represent negative values or not.
+    *
+    * If this is not set for floating point, it means that all values are
+    * assumed to be positive.
+    */
+   unsigned sign:1;
 
-      /**
-       * Element width.
-       *
-       * For fixed point values, the fixed point is assumed to be at half the
-       * width.
-       */
-      unsigned width:14;
+   /**
+    * Whether values are normalized to fit [0, 1] interval, or [-1, 1]
+    * interval for signed types.
+    *
+    * For integer types it means the representable integer range should be
+    * interpreted as the interval above.
+    *
+    * For floating and fixed point formats it means the values should be
+    * clamped to the interval above.
+    */
+   unsigned norm:1;
 
-      /** 
-       * Vector length.
-       *
-       * width*length should be a power of two greater or equal to eight.
-       *
-       * @sa LP_MAX_VECTOR_LENGTH
-       */
-      unsigned length:14;
-   };
-   uint32_t value;
+   /**
+    * Element width.
+    *
+    * For fixed point values, the fixed point is assumed to be at half the
+    * width.
+    */
+   unsigned width:14;
+
+   /**
+    * Vector length.
+    *
+    * width*length should be a power of two greater or equal to eight.
+    *
+    * @sa LP_MAX_VECTOR_LENGTH
+    */
+   unsigned length:14;
 };
 
 
@@ -124,7 +121,7 @@ struct lp_build_context
     * This not only describes the input/output LLVM types, but also whether
     * to normalize/clamp the results.
     */
-   union lp_type type;
+   struct lp_type type;
 
    /** Same as lp_build_undef(type) */
    LLVMValueRef undef;
@@ -138,41 +135,41 @@ struct lp_build_context
 
 
 LLVMTypeRef
-lp_build_elem_type(union lp_type type);
+lp_build_elem_type(struct lp_type type);
 
 
 LLVMTypeRef
-lp_build_vec_type(union lp_type type);
+lp_build_vec_type(struct lp_type type);
 
 
 boolean
-lp_check_elem_type(union lp_type type, LLVMTypeRef elem_type);
+lp_check_elem_type(struct lp_type type, LLVMTypeRef elem_type);
 
 
 boolean
-lp_check_vec_type(union lp_type type, LLVMTypeRef vec_type);
+lp_check_vec_type(struct lp_type type, LLVMTypeRef vec_type);
 
 
 boolean
-lp_check_value(union lp_type type, LLVMValueRef val);
+lp_check_value(struct lp_type type, LLVMValueRef val);
 
 
 LLVMTypeRef
-lp_build_int_elem_type(union lp_type type);
+lp_build_int_elem_type(struct lp_type type);
 
 
 LLVMTypeRef
-lp_build_int_vec_type(union lp_type type);
+lp_build_int_vec_type(struct lp_type type);
 
 
-union lp_type
-lp_int_type(union lp_type type);
+struct lp_type
+lp_int_type(struct lp_type type);
 
 
 void
 lp_build_context_init(struct lp_build_context *bld,
                       LLVMBuilderRef builder,
-                      union lp_type type);
+                      struct lp_type type);
 
 
 #endif /* !LP_BLD_TYPE_H */

src/gallium/drivers/llvmpipe/lp_jit.c

@@ -152,7 +152,7 @@ lp_jit_screen_init(struct llvmpipe_screen *screen)
    screen->provider = LLVMCreateModuleProviderForExistingModule(screen->module);
 
    if (LLVMCreateJITCompiler(&screen->engine, screen->provider, 1, &error)) {
-      fprintf(stderr, "%s\n", error);
+      _debug_printf("%s\n", error);
       LLVMDisposeMessage(error);
       abort();
    }

src/gallium/drivers/llvmpipe/lp_screen.c

@@ -65,8 +65,6 @@ llvmpipe_get_param(struct pipe_screen *screen, int param)
       return 1;
    case PIPE_CAP_GLSL:
       return 1;
-   case PIPE_CAP_S3TC:
-      return 0;
    case PIPE_CAP_ANISOTROPIC_FILTER:
       return 0;
    case PIPE_CAP_POINT_SPRITE:

src/gallium/drivers/llvmpipe/lp_state_fs.c

@@ -133,13 +133,13 @@ generate_pos0(LLVMBuilderRef builder,
 static void
 generate_depth(LLVMBuilderRef builder,
                const struct lp_fragment_shader_variant_key *key,
-               union lp_type src_type,
+               struct lp_type src_type,
                struct lp_build_mask_context *mask,
                LLVMValueRef src,
                LLVMValueRef dst_ptr)
 {
    const struct util_format_description *format_desc;
-   union lp_type dst_type;
+   struct lp_type dst_type;
 
    if(!key->depth.enabled)
       return;
@@ -181,7 +181,7 @@ generate_fs(struct llvmpipe_context *lp,
             struct lp_fragment_shader *shader,
             const struct lp_fragment_shader_variant_key *key,
             LLVMBuilderRef builder,
-            union lp_type type,
+            struct lp_type type,
             LLVMValueRef context_ptr,
             unsigned i,
             const struct lp_build_interp_soa_context *interp,
@@ -299,7 +299,7 @@ generate_fs(struct llvmpipe_context *lp,
 static void
 generate_blend(const struct pipe_blend_state *blend,
                LLVMBuilderRef builder,
-               union lp_type type,
+               struct lp_type type,
                LLVMValueRef context_ptr,
                LLVMValueRef mask,
                LLVMValueRef *src,
@@ -364,8 +364,8 @@ generate_fragment(struct llvmpipe_context *lp,
 {
    struct llvmpipe_screen *screen = llvmpipe_screen(lp->pipe.screen);
    struct lp_fragment_shader_variant *variant;
-   union lp_type fs_type;
-   union lp_type blend_type;
+   struct lp_type fs_type;
+   struct lp_type blend_type;
    LLVMTypeRef fs_elem_type;
    LLVMTypeRef fs_vec_type;
    LLVMTypeRef fs_int_vec_type;
@@ -431,7 +431,7 @@ generate_fragment(struct llvmpipe_context *lp,
    /* TODO: actually pick these based on the fs and color buffer
     * characteristics. */
 
-   fs_type.value = 0;
+   memset(&fs_type, 0, sizeof fs_type);
    fs_type.floating = TRUE; /* floating point values */
    fs_type.sign = TRUE;     /* values are signed */
    fs_type.norm = FALSE;    /* values are not limited to [0,1] or [-1,1] */
@@ -439,7 +439,7 @@ generate_fragment(struct llvmpipe_context *lp,
    fs_type.length = 4;      /* 4 element per vector */
    num_fs = 4;
 
-   blend_type.value = 0;
+   memset(&blend_type, 0, sizeof blend_type);
    blend_type.floating = FALSE; /* values are integers */
    blend_type.sign = FALSE;     /* values are unsigned */
    blend_type.norm = TRUE;      /* values are in [0,1] or [-1,1] */

src/gallium/drivers/llvmpipe/lp_test.h

@@ -86,43 +86,43 @@ random_float(void);
 
 
 void
-dump_type(FILE *fp, union lp_type type);
+dump_type(FILE *fp, struct lp_type type);
 
 
 double
-read_elem(union lp_type type, const void *src, unsigned index);
+read_elem(struct lp_type type, const void *src, unsigned index);
 
 
 void
-write_elem(union lp_type type, void *dst, unsigned index, double src);
+write_elem(struct lp_type type, void *dst, unsigned index, double src);
 
 
 void
-random_elem(union lp_type type, void *dst, unsigned index);
+random_elem(struct lp_type type, void *dst, unsigned index);
 
 
 void
-read_vec(union lp_type type, const void *src, double *dst);
+read_vec(struct lp_type type, const void *src, double *dst);
 
 
 void
-write_vec(union lp_type type, void *dst, const double *src);
+write_vec(struct lp_type type, void *dst, const double *src);
 
 
 void
-random_vec(union lp_type type, void *dst);
+random_vec(struct lp_type type, void *dst);
 
 
 boolean
-compare_vec_with_eps(union lp_type type, const void *res, const void *ref, double eps);
+compare_vec_with_eps(struct lp_type type, const void *res, const void *ref, double eps);
 
 
 boolean
-compare_vec(union lp_type type, const void *res, const void *ref);
+compare_vec(struct lp_type type, const void *res, const void *ref);
 
 
 void
-dump_vec(FILE *fp, union lp_type type, const void *src);
+dump_vec(FILE *fp, struct lp_type type, const void *src);
 
 
 #endif /* !LP_TEST_H */

src/gallium/drivers/llvmpipe/lp_test_blend.c

@@ -80,7 +80,7 @@ static void
 write_tsv_row(FILE *fp,
               const struct pipe_blend_state *blend,
               enum vector_mode mode,
-              union lp_type type,
+              struct lp_type type,
               double cycles,
               boolean success)
 {
@@ -125,7 +125,7 @@ static void
 dump_blend_type(FILE *fp,
                 const struct pipe_blend_state *blend,
                 enum vector_mode mode,
-                union lp_type type)
+                struct lp_type type)
 {
    fprintf(fp, "%s", mode ? "soa" : "aos");
 
@@ -153,7 +153,7 @@ static LLVMValueRef
 add_blend_test(LLVMModuleRef module,
                const struct pipe_blend_state *blend,
                enum vector_mode mode,
-               union lp_type type)
+               struct lp_type type)
 {
    LLVMTypeRef ret_type;
    LLVMTypeRef vec_type;
@@ -467,7 +467,7 @@ test_one(unsigned verbose,
          FILE *fp,
          const struct pipe_blend_state *blend,
          enum vector_mode mode,
-         union lp_type type)
+         struct lp_type type)
 {
    LLVMModuleRef module = NULL;
    LLVMValueRef func = NULL;
@@ -765,10 +765,10 @@ blend_funcs[] = {
 };
 
 
-const union lp_type blend_types[] = {
+const struct lp_type blend_types[] = {
    /* float, fixed,  sign,  norm, width, len */
-   {{  TRUE, FALSE, FALSE,  TRUE,    32,   4 }}, /* f32 x 4 */
-   {{ FALSE, FALSE, FALSE,  TRUE,     8,  16 }}, /* u8n x 16 */
+   {   TRUE, FALSE, FALSE,  TRUE,    32,   4 }, /* f32 x 4 */
+   {  FALSE, FALSE, FALSE,  TRUE,     8,  16 }, /* u8n x 16 */
 };
 
 
@@ -788,7 +788,7 @@ test_all(unsigned verbose, FILE *fp)
    const unsigned *alpha_dst_factor;
    struct pipe_blend_state blend;
    enum vector_mode mode;
-   const union lp_type *type;
+   const struct lp_type *type;
    bool success = TRUE;
 
    for(rgb_func = blend_funcs; rgb_func < &blend_funcs[num_funcs]; ++rgb_func) {
@@ -841,27 +841,27 @@ test_some(unsigned verbose, FILE *fp, unsigned long n)
    const unsigned *alpha_dst_factor;
    struct pipe_blend_state blend;
    enum vector_mode mode;
-   const union lp_type *type;
+   const struct lp_type *type;
    unsigned long i;
    bool success = TRUE;
 
    for(i = 0; i < n; ++i) {
-      rgb_func = &blend_funcs[random() % num_funcs];
-      alpha_func = &blend_funcs[random() % num_funcs];
-      rgb_src_factor = &blend_factors[random() % num_factors];
-      alpha_src_factor = &blend_factors[random() % num_factors];
+      rgb_func = &blend_funcs[rand() % num_funcs];
+      alpha_func = &blend_funcs[rand() % num_funcs];
+      rgb_src_factor = &blend_factors[rand() % num_factors];
+      alpha_src_factor = &blend_factors[rand() % num_factors];
       
       do {
-         rgb_dst_factor = &blend_factors[random() % num_factors];
+         rgb_dst_factor = &blend_factors[rand() % num_factors];
       } while(*rgb_dst_factor == PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE);
 
       do {
-         alpha_dst_factor = &blend_factors[random() % num_factors];
+         alpha_dst_factor = &blend_factors[rand() % num_factors];
       } while(*alpha_dst_factor == PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE);
 
-      mode = random() & 1;
+      mode = rand() & 1;
 
-      type = &blend_types[random() % num_types];
+      type = &blend_types[rand() % num_types];
 
       memset(&blend, 0, sizeof blend);
       blend.blend_enable      = 1;

src/gallium/drivers/llvmpipe/lp_test_conv.c

@@ -59,8 +59,8 @@ write_tsv_header(FILE *fp)
 
 static void
 write_tsv_row(FILE *fp,
-              union lp_type src_type,
-              union lp_type dst_type,
+              struct lp_type src_type,
+              struct lp_type dst_type,
               double cycles,
               boolean success)
 {
@@ -80,8 +80,8 @@ write_tsv_row(FILE *fp,
 
 static void
 dump_conv_types(FILE *fp,
-               union lp_type src_type,
-               union lp_type dst_type)
+               struct lp_type src_type,
+               struct lp_type dst_type)
 {
    fprintf(fp, "src_type=");
    dump_type(fp, src_type);
@@ -96,8 +96,8 @@ dump_conv_types(FILE *fp,
 
 static LLVMValueRef
 add_conv_test(LLVMModuleRef module,
-              union lp_type src_type, unsigned num_srcs,
-              union lp_type dst_type, unsigned num_dsts)
+              struct lp_type src_type, unsigned num_srcs,
+              struct lp_type dst_type, unsigned num_dsts)
 {
    LLVMTypeRef args[2];
    LLVMValueRef func;
@@ -145,8 +145,8 @@ add_conv_test(LLVMModuleRef module,
 static boolean
 test_one(unsigned verbose,
          FILE *fp,
-         union lp_type src_type,
-         union lp_type dst_type)
+         struct lp_type src_type,
+         struct lp_type dst_type)
 {
    LLVMModuleRef module = NULL;
    LLVMValueRef func = NULL;
@@ -343,35 +343,35 @@ test_one(unsigned verbose,
 }
 
 
-const union lp_type conv_types[] = {
+const struct lp_type conv_types[] = {
    /* float, fixed,  sign,  norm, width, len */
 
-   {{  TRUE, FALSE,  TRUE,  TRUE,    32,   4 }},
-   {{  TRUE, FALSE,  TRUE, FALSE,    32,   4 }},
-   {{  TRUE, FALSE, FALSE,  TRUE,    32,   4 }},
-   {{  TRUE, FALSE, FALSE, FALSE,    32,   4 }},
+   {   TRUE, FALSE,  TRUE,  TRUE,    32,   4 },
+   {   TRUE, FALSE,  TRUE, FALSE,    32,   4 },
+   {   TRUE, FALSE, FALSE,  TRUE,    32,   4 },
+   {   TRUE, FALSE, FALSE, FALSE,    32,   4 },
 
    /* TODO: test fixed formats too */
 
-   {{ FALSE, FALSE,  TRUE,  TRUE,    16,   8 }},
-   {{ FALSE, FALSE,  TRUE, FALSE,    16,   8 }},
-   {{ FALSE, FALSE, FALSE,  TRUE,    16,   8 }},
-   {{ FALSE, FALSE, FALSE, FALSE,    16,   8 }},
+   {  FALSE, FALSE,  TRUE,  TRUE,    16,   8 },
+   {  FALSE, FALSE,  TRUE, FALSE,    16,   8 },
+   {  FALSE, FALSE, FALSE,  TRUE,    16,   8 },
+   {  FALSE, FALSE, FALSE, FALSE,    16,   8 },
 
-   {{ FALSE, FALSE,  TRUE,  TRUE,    32,   4 }},
-   {{ FALSE, FALSE,  TRUE, FALSE,    32,   4 }},
-   {{ FALSE, FALSE, FALSE,  TRUE,    32,   4 }},
-   {{ FALSE, FALSE, FALSE, FALSE,    32,   4 }},
+   {  FALSE, FALSE,  TRUE,  TRUE,    32,   4 },
+   {  FALSE, FALSE,  TRUE, FALSE,    32,   4 },
+   {  FALSE, FALSE, FALSE,  TRUE,    32,   4 },
+   {  FALSE, FALSE, FALSE, FALSE,    32,   4 },
 
-   {{ FALSE, FALSE,  TRUE,  TRUE,    16,   8 }},
-   {{ FALSE, FALSE,  TRUE, FALSE,    16,   8 }},
-   {{ FALSE, FALSE, FALSE,  TRUE,    16,   8 }},
-   {{ FALSE, FALSE, FALSE, FALSE,    16,   8 }},
+   {  FALSE, FALSE,  TRUE,  TRUE,    16,   8 },
+   {  FALSE, FALSE,  TRUE, FALSE,    16,   8 },
+   {  FALSE, FALSE, FALSE,  TRUE,    16,   8 },
+   {  FALSE, FALSE, FALSE, FALSE,    16,   8 },
 
-   {{ FALSE, FALSE,  TRUE,  TRUE,     8,  16 }},
-   {{ FALSE, FALSE,  TRUE, FALSE,     8,  16 }},
-   {{ FALSE, FALSE, FALSE,  TRUE,     8,  16 }},
-   {{ FALSE, FALSE, FALSE, FALSE,     8,  16 }},
+   {  FALSE, FALSE,  TRUE,  TRUE,     8,  16 },
+   {  FALSE, FALSE,  TRUE, FALSE,     8,  16 },
+   {  FALSE, FALSE, FALSE,  TRUE,     8,  16 },
+   {  FALSE, FALSE, FALSE, FALSE,     8,  16 },
 };
 
 
@@ -381,8 +381,8 @@ const unsigned num_types = sizeof(conv_types)/sizeof(conv_types[0]);
 boolean
 test_all(unsigned verbose, FILE *fp)
 {
-   const union lp_type *src_type;
-   const union lp_type *dst_type;
+   const struct lp_type *src_type;
+   const struct lp_type *dst_type;
    bool success = TRUE;
 
    for(src_type = conv_types; src_type < &conv_types[num_types]; ++src_type) {
@@ -407,16 +407,16 @@ test_all(unsigned verbose, FILE *fp)
 boolean
 test_some(unsigned verbose, FILE *fp, unsigned long n)
 {
-   const union lp_type *src_type;
-   const union lp_type *dst_type;
+   const struct lp_type *src_type;
+   const struct lp_type *dst_type;
    unsigned long i;
    bool success = TRUE;
 
    for(i = 0; i < n; ++i) {
-      src_type = &conv_types[random() % num_types];
+      src_type = &conv_types[rand() % num_types];
       
       do {
-         dst_type = &conv_types[random() % num_types];
+         dst_type = &conv_types[rand() % num_types];
       } while (src_type == dst_type || src_type->norm != dst_type->norm);
 
       if(!test_one(verbose, fp, *src_type, *dst_type))

src/gallium/drivers/llvmpipe/lp_test_main.c

@@ -40,7 +40,7 @@
 
 void
 dump_type(FILE *fp,
-          union lp_type type)
+          struct lp_type type)
 {
    fprintf(fp, "%s%s%u%sx%u",
            type.sign ? (type.floating || type.fixed ? "" : "s") : "u",
@@ -52,7 +52,7 @@ dump_type(FILE *fp,
 
 
 double
-read_elem(union lp_type type, const void *src, unsigned index)
+read_elem(struct lp_type type, const void *src, unsigned index)
 {
    double scale = lp_const_scale(type);
    double value;
@@ -115,7 +115,7 @@ read_elem(union lp_type type, const void *src, unsigned index)
 
 
 void
-write_elem(union lp_type type, void *dst, unsigned index, double value)
+write_elem(struct lp_type type, void *dst, unsigned index, double value)
 {
    assert(index < type.length);
    if(!type.sign && value < 0.0)
@@ -184,11 +184,11 @@ write_elem(union lp_type type, void *dst, unsigned index, double value)
 
 
 void
-random_elem(union lp_type type, void *dst, unsigned index)
+random_elem(struct lp_type type, void *dst, unsigned index)
 {
    double value;
    assert(index < type.length);
-   value = (double)random()/(double)RAND_MAX;
+   value = (double)rand()/(double)RAND_MAX;
    if(!type.norm) {
       unsigned long long mask;
       if (type.floating)
@@ -199,17 +199,17 @@ random_elem(union lp_type type, void *dst, unsigned index)
          mask = ((unsigned long long)1 << (type.width - 1)) - 1;
       else
          mask = ((unsigned long long)1 << type.width) - 1;
-      value += (double)(mask & random());
+      value += (double)(mask & rand());
    }
    if(!type.sign)
-      if(random() & 1)
+      if(rand() & 1)
          value = -value;
    write_elem(type, dst, index, value);
 }
 
 
 void
-read_vec(union lp_type type, const void *src, double *dst)
+read_vec(struct lp_type type, const void *src, double *dst)
 {
    unsigned i;
    for (i = 0; i < type.length; ++i)
@@ -218,7 +218,7 @@ read_vec(union lp_type type, const void *src, double *dst)
 
 
 void
-write_vec(union lp_type type, void *dst, const double *src)
+write_vec(struct lp_type type, void *dst, const double *src)
 {
    unsigned i;
    for (i = 0; i < type.length; ++i)
@@ -229,12 +229,12 @@ write_vec(union lp_type type, void *dst, const double *src)
 float
 random_float(void)
 {
-    return (float)((double)random()/(double)RAND_MAX);
+    return (float)((double)rand()/(double)RAND_MAX);
 }
 
 
 void
-random_vec(union lp_type type, void *dst)
+random_vec(struct lp_type type, void *dst)
 {
    unsigned i;
    for (i = 0; i < type.length; ++i)
@@ -243,7 +243,7 @@ random_vec(union lp_type type, void *dst)
 
 
 boolean
-compare_vec_with_eps(union lp_type type, const void *res, const void *ref, double eps)
+compare_vec_with_eps(struct lp_type type, const void *res, const void *ref, double eps)
 {
    unsigned i;
    for (i = 0; i < type.length; ++i) {
@@ -259,7 +259,7 @@ compare_vec_with_eps(union lp_type type, const void *res, const void *ref, doubl
 
 
 boolean
-compare_vec(union lp_type type, const void *res, const void *ref)
+compare_vec(struct lp_type type, const void *res, const void *ref)
 {
    double eps = lp_const_eps(type);
    return compare_vec_with_eps(type, res, ref, eps);
@@ -267,7 +267,7 @@ compare_vec(union lp_type type, const void *res, const void *ref)
 
 
 void
-dump_vec(FILE *fp, union lp_type type, const void *src)
+dump_vec(FILE *fp, struct lp_type type, const void *src)
 {
    unsigned i;
    for (i = 0; i < type.length; ++i) {

src/gallium/drivers/llvmpipe/lp_tex_sample_c.c

@@ -1654,7 +1654,7 @@ lp_c_sampler_soa_destroy(struct lp_build_sampler_soa *sampler)
 static void
 lp_c_sampler_soa_emit_fetch_texel(struct lp_build_sampler_soa *_sampler,
                                   LLVMBuilderRef builder,
-                                  union lp_type type,
+                                  struct lp_type type,
                                   unsigned unit,
                                   unsigned num_coords,
                                   const LLVMValueRef *coords,

src/gallium/drivers/llvmpipe/lp_tex_sample_llvm.c

@@ -149,7 +149,7 @@ lp_llvm_sampler_soa_destroy(struct lp_build_sampler_soa *sampler)
 static void
 lp_llvm_sampler_soa_emit_fetch_texel(struct lp_build_sampler_soa *base,
                                      LLVMBuilderRef builder,
-                                     union lp_type type,
+                                     struct lp_type type,
                                      unsigned unit,
                                      unsigned num_coords,
                                      const LLVMValueRef *coords,

src/gallium/drivers/nv04/nv04_screen.c

@@ -16,8 +16,6 @@ nv04_screen_get_param(struct pipe_screen *screen, int param)
 		return 0;
 	case PIPE_CAP_GLSL:
 		return 0;
-	case PIPE_CAP_S3TC:
-		return 0;
 	case PIPE_CAP_ANISOTROPIC_FILTER:
 		return 0;
 	case PIPE_CAP_POINT_SPRITE:

src/gallium/drivers/nv10/nv10_screen.c

@@ -15,8 +15,6 @@ nv10_screen_get_param(struct pipe_screen *screen, int param)
 		return 0;
 	case PIPE_CAP_GLSL:
 		return 0;
-	case PIPE_CAP_S3TC:
-		return 0;
 	case PIPE_CAP_ANISOTROPIC_FILTER:
 		return 1;
 	case PIPE_CAP_POINT_SPRITE:

src/gallium/drivers/nv20/nv20_screen.c

@@ -15,8 +15,6 @@ nv20_screen_get_param(struct pipe_screen *screen, int param)
 		return 0;
 	case PIPE_CAP_GLSL:
 		return 0;
-	case PIPE_CAP_S3TC:
-		return 0;
 	case PIPE_CAP_ANISOTROPIC_FILTER:
 		return 1;
 	case PIPE_CAP_POINT_SPRITE:

src/gallium/drivers/nv30/nv30_screen.c

@@ -22,8 +22,6 @@ nv30_screen_get_param(struct pipe_screen *pscreen, int param)
 		return 1;
 	case PIPE_CAP_GLSL:
 		return 0;
-	case PIPE_CAP_S3TC:
-		return 0;
 	case PIPE_CAP_ANISOTROPIC_FILTER:
 		return 1;
 	case PIPE_CAP_POINT_SPRITE:

src/gallium/drivers/nv40/nv40_screen.c

@@ -21,8 +21,6 @@ nv40_screen_get_param(struct pipe_screen *pscreen, int param)
 		return 1;
 	case PIPE_CAP_GLSL:
 		return 0;
-	case PIPE_CAP_S3TC:
-		return 1;
 	case PIPE_CAP_ANISOTROPIC_FILTER:
 		return 1;
 	case PIPE_CAP_POINT_SPRITE:

src/gallium/drivers/nv50/nv50_context.c

@@ -37,11 +37,12 @@ nv50_flush(struct pipe_context *pipe, unsigned flags,
 
 	/* We need this in the ddx for reliable composite, not sure what we're
 	 * actually flushing. We generate all our own flushes with flags = 0. */
-	WAIT_RING(chan, 3);
+	WAIT_RING(chan, 2);
 	BEGIN_RING(chan, eng2d, 0x0110, 1);
 	OUT_RING  (chan, 0);
 
-	FIRE_RING(chan);
+	if (flags & PIPE_FLUSH_FRAME)
+		FIRE_RING(chan);
 }
 
 static void
@@ -110,6 +111,9 @@ nv50_create(struct pipe_screen *pscreen, unsigned pctx_id)
 	nv50->pipe.is_texture_referenced = nv50_is_texture_referenced;
 	nv50->pipe.is_buffer_referenced = nv50_is_buffer_referenced;
 
+	screen->base.channel->user_private = nv50;
+	screen->base.channel->flush_notify = nv50_state_flush_notify;
+
 	nv50_init_surface_functions(nv50);
 	nv50_init_state_functions(nv50);
 	nv50_init_query_functions(nv50);

src/gallium/drivers/nv50/nv50_context.h

@@ -116,6 +116,7 @@ struct nv50_state {
 	unsigned miptree_nr;
 	struct nouveau_stateobj *vertprog;
 	struct nouveau_stateobj *fragprog;
+	struct nouveau_stateobj *programs;
 	struct nouveau_stateobj *vtxfmt;
 	struct nouveau_stateobj *vtxbuf;
 	struct nouveau_stateobj *vtxattr;
@@ -190,10 +191,12 @@ extern void nv50_clear(struct pipe_context *pipe, unsigned buffers,
 /* nv50_program.c */
 extern void nv50_vertprog_validate(struct nv50_context *nv50);
 extern void nv50_fragprog_validate(struct nv50_context *nv50);
+extern void nv50_linkage_validate(struct nv50_context *nv50);
 extern void nv50_program_destroy(struct nv50_context *nv50, struct nv50_program *p);
 
 /* nv50_state_validate.c */
 extern boolean nv50_state_validate(struct nv50_context *nv50);
+extern void nv50_state_flush_notify(struct nouveau_channel *chan);
 
 /* nv50_tex.c */
 extern void nv50_tex_validate(struct nv50_context *);

src/gallium/drivers/nv50/nv50_program.h

@@ -15,6 +15,15 @@ struct nv50_program_exec {
 	} param;
 };
 
+struct nv50_sreg4 {
+	uint8_t hw;
+	uint8_t id_vp;
+	uint8_t id_fp;
+
+	uint8_t mask;
+	boolean linear;
+};
+
 struct nv50_program {
 	struct pipe_shader_state pipe;
 	struct tgsi_shader_info info;
@@ -24,8 +33,8 @@ struct nv50_program {
 	struct nv50_program_exec *exec_head;
 	struct nv50_program_exec *exec_tail;
 	unsigned exec_size;
-	struct nouveau_resource *data[2];
-	unsigned data_start[2];
+	struct nouveau_resource *data[1];
+	unsigned data_start[1];
 
 	struct nouveau_bo *bo;
 
@@ -36,14 +45,20 @@ struct nv50_program {
 	struct {
 		unsigned high_temp;
 		unsigned high_result;
-		struct {
-			unsigned attr[2];
-		} vp;
-		struct {
-			unsigned regs[4];
-			unsigned map[5];
-			unsigned high_map;
-		} fp;
+
+		uint32_t attr[2];
+		uint32_t regs[4];
+
+		/* for VPs, io_nr doesn't count 'private' results (PSIZ etc.) */
+		unsigned io_nr;
+		struct nv50_sreg4 io[PIPE_MAX_SHADER_OUTPUTS];
+
+		/* FP colour inputs, VP/GP back colour outputs */
+		struct nv50_sreg4 two_side[2];
+
+		/* VP only */
+		uint8_t clpd, clpd_nr;
+		uint8_t psiz;
 	} cfg;
 };
 

src/gallium/drivers/nv50/nv50_screen.c

@@ -87,12 +87,10 @@ nv50_screen_get_param(struct pipe_screen *pscreen, int param)
 		return 1;
 	case PIPE_CAP_GLSL:
 		return 0;
-	case PIPE_CAP_S3TC:
-		return 1;
 	case PIPE_CAP_ANISOTROPIC_FILTER:
 		return 1;
 	case PIPE_CAP_POINT_SPRITE:
-		return 0;
+		return 1;
 	case PIPE_CAP_MAX_RENDER_TARGETS:
 		return 8;
 	case PIPE_CAP_OCCLUSION_QUERY:

src/gallium/drivers/nv50/nv50_state.c

@@ -276,6 +276,9 @@ nv50_rasterizer_state_create(struct pipe_context *pipe,
 	so_method(so, tesla, 0x1684, 1);
 	so_data  (so, cso->flatshade_first ? 0 : 1);
 
+	so_method(so, tesla, NV50TCL_VERTEX_TWO_SIDE_ENABLE, 1);
+	so_data  (so, cso->light_twoside);
+
 	so_method(so, tesla, NV50TCL_LINE_WIDTH, 1);
 	so_data  (so, fui(cso->line_width));
 	so_method(so, tesla, NV50TCL_LINE_SMOOTH_ENABLE, 1);
@@ -294,6 +297,9 @@ nv50_rasterizer_state_create(struct pipe_context *pipe,
 	so_method(so, tesla, NV50TCL_POINT_SIZE, 1);
 	so_data  (so, fui(cso->point_size));
 
+	so_method(so, tesla, NV50TCL_POINT_SPRITE_ENABLE, 1);
+	so_data  (so, cso->point_sprite);
+
 	so_method(so, tesla, NV50TCL_POLYGON_MODE_FRONT, 3);
 	if (cso->front_winding == PIPE_WINDING_CCW) {
 		so_data(so, nvgl_polygon_mode(cso->fill_ccw));

src/gallium/drivers/nv50/nv50_state_validate.c

@@ -189,6 +189,8 @@ nv50_state_emit(struct nv50_context *nv50)
 		so_emit(chan, nv50->state.vertprog);
 	if (nv50->state.dirty & NV50_NEW_FRAGPROG)
 		so_emit(chan, nv50->state.fragprog);
+	if (nv50->state.dirty & (NV50_NEW_FRAGPROG | NV50_NEW_VERTPROG))
+		so_emit(chan, nv50->state.programs);
 	if (nv50->state.dirty & NV50_NEW_RASTERIZER)
 		so_emit(chan, nv50->state.rast);
 	if (nv50->state.dirty & NV50_NEW_BLEND_COLOUR)
@@ -210,6 +212,12 @@ nv50_state_emit(struct nv50_context *nv50)
 			so_emit(chan, nv50->state.vtxattr);
 	}
 	nv50->state.dirty = 0;
+}
+
+void
+nv50_state_flush_notify(struct nouveau_channel *chan)
+{
+	struct nv50_context *nv50 = chan->user_private;
 
 	so_emit_reloc_markers(chan, nv50->state.fb);
 	so_emit_reloc_markers(chan, nv50->state.vertprog);
@@ -240,6 +248,9 @@ nv50_state_validate(struct nv50_context *nv50)
 	if (nv50->dirty & (NV50_NEW_FRAGPROG | NV50_NEW_FRAGPROG_CB))
 		nv50_fragprog_validate(nv50);
 
+	if (nv50->dirty & (NV50_NEW_FRAGPROG | NV50_NEW_VERTPROG))
+		nv50_linkage_validate(nv50);
+
 	if (nv50->dirty & NV50_NEW_RASTERIZER)
 		so_ref(nv50->rasterizer->so, &nv50->state.rast);
 

src/gallium/drivers/nv50/nv50_transfer.c

@@ -89,14 +89,14 @@ nv50_transfer_rect_m2mf(struct pipe_screen *pscreen,
 		if (src_bo->tile_flags) {
 			BEGIN_RING(chan, m2mf,
 				NV50_MEMORY_TO_MEMORY_FORMAT_TILING_POSITION_IN, 1);
-			OUT_RING  (chan, (sy << 16) | sx);
+			OUT_RING  (chan, (sy << 16) | (sx * cpp));
 		} else {
 			src_offset += (line_count * src_pitch);
 		}
 		if (dst_bo->tile_flags) {
 			BEGIN_RING(chan, m2mf,
 				NV50_MEMORY_TO_MEMORY_FORMAT_TILING_POSITION_OUT, 1);
-			OUT_RING  (chan, (dy << 16) | dx);
+			OUT_RING  (chan, (dy << 16) | (dx * cpp));
 		} else {
 			dst_offset += (line_count * dst_pitch);
 		}

src/gallium/drivers/r300/r300_screen.c

@@ -93,8 +93,6 @@ static int r300_get_param(struct pipe_screen* pscreen, int param)
             } else {
                 return 0;
             }
-        case PIPE_CAP_S3TC:
-            return 1;
         case PIPE_CAP_ANISOTROPIC_FILTER:
             return 1;
         case PIPE_CAP_POINT_SPRITE:

src/gallium/drivers/softpipe/sp_screen.c

@@ -65,8 +65,6 @@ softpipe_get_param(struct pipe_screen *screen, int param)
       return 1;
    case PIPE_CAP_GLSL:
       return 1;
-   case PIPE_CAP_S3TC:
-      return 0;
    case PIPE_CAP_ANISOTROPIC_FILTER:
       return 0;
    case PIPE_CAP_POINT_SPRITE:
@@ -141,6 +139,7 @@ softpipe_is_format_supported( struct pipe_screen *screen,
    case PIPE_FORMAT_DXT1_RGBA:
    case PIPE_FORMAT_DXT3_RGBA:
    case PIPE_FORMAT_DXT5_RGBA:
+   case PIPE_FORMAT_Z32_FLOAT:
       return FALSE;
    default:
       return TRUE;

src/gallium/include/pipe/p_defines.h

@@ -281,7 +281,7 @@ enum pipe_transfer_usage {
 #define PIPE_CAP_NPOT_TEXTURES           2
 #define PIPE_CAP_TWO_SIDED_STENCIL       3
 #define PIPE_CAP_GLSL                    4  /* XXX need something better */
-#define PIPE_CAP_S3TC                    5
+#define PIPE_CAP_S3TC                    5  /* XXX: deprecated; cap determined via supported sampler formats */
 #define PIPE_CAP_ANISOTROPIC_FILTER      6
 #define PIPE_CAP_POINT_SPRITE            7
 #define PIPE_CAP_MAX_RENDER_TARGETS      8

src/gallium/include/pipe/p_inlines.h

@@ -118,7 +118,7 @@ pipe_buffer_write(struct pipe_screen *screen,
                   unsigned offset, unsigned size,
                   const void *data)
 {
-   uint8_t *map;
+   void *map;
    
    assert(offset < buf->size);
    assert(offset + size <= buf->size);
@@ -129,7 +129,7 @@ pipe_buffer_write(struct pipe_screen *screen,
                                PIPE_BUFFER_USAGE_FLUSH_EXPLICIT);
    assert(map);
    if(map) {
-      memcpy(map + offset, data, size);
+      memcpy((uint8_t *)map + offset, data, size);
       pipe_buffer_flush_mapped_range(screen, buf, offset, size);
       pipe_buffer_unmap(screen, buf);
    }
@@ -141,7 +141,7 @@ pipe_buffer_read(struct pipe_screen *screen,
                  unsigned offset, unsigned size,
                  void *data)
 {
-   uint8_t *map;
+   void *map;
    
    assert(offset < buf->size);
    assert(offset + size <= buf->size);
@@ -150,11 +150,31 @@ pipe_buffer_read(struct pipe_screen *screen,
    map = pipe_buffer_map_range(screen, buf, offset, size, PIPE_BUFFER_USAGE_CPU_READ);
    assert(map);
    if(map) {
-      memcpy(data, map + offset, size);
+      memcpy(data, (const uint8_t *)map + offset, size);
       pipe_buffer_unmap(screen, buf);
    }
 }
 
+static INLINE void *
+pipe_transfer_map( struct pipe_transfer *transf )
+{
+   struct pipe_screen *screen = transf->texture->screen;
+   return screen->transfer_map(screen, transf);
+}
+
+static INLINE void
+pipe_transfer_unmap( struct pipe_transfer *transf )
+{
+   struct pipe_screen *screen = transf->texture->screen;
+   screen->transfer_unmap(screen, transf);
+}
+
+static INLINE void
+pipe_transfer_destroy( struct pipe_transfer *transf )
+{
+   struct pipe_screen *screen = transf->texture->screen;
+   screen->tex_transfer_destroy(transf);
+}
 
 #ifdef __cplusplus
 }

src/gallium/include/pipe/p_state.h

@@ -307,7 +307,7 @@ struct pipe_transfer
    unsigned nblocksx;            /**< allocated width in blocks */
    unsigned nblocksy;            /**< allocated height in blocks */
    unsigned stride;              /**< stride in bytes between rows of blocks */
-   unsigned usage;               /**< PIPE_TRANSFER_*  */
+   enum pipe_transfer_usage usage; /**< PIPE_TRANSFER_*  */
 
    struct pipe_texture *texture; /**< texture to transfer to/from  */
    unsigned face;

src/gallium/include/pipe/p_thread.h

@@ -39,7 +39,7 @@
 #include "util/u_debug.h" /* for assert */
 
 
-#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_HAIKU)
+#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) || defined(PIPE_OS_HAIKU)
 
 #include <pthread.h> /* POSIX threads headers */
 #include <stdio.h> /* for perror() */
@@ -213,7 +213,7 @@ typedef unsigned pipe_condvar;
  */
 
 typedef struct {
-#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_HAIKU)
+#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) || defined(PIPE_OS_HAIKU)
    pthread_key_t key;
 #elif defined(PIPE_SUBSYSTEM_WINDOWS_USER)
    DWORD key;
@@ -228,7 +228,7 @@ typedef struct {
 static INLINE void
 pipe_tsd_init(pipe_tsd *tsd)
 {
-#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_HAIKU)
+#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) || defined(PIPE_OS_HAIKU)
    if (pthread_key_create(&tsd->key, NULL/*free*/) != 0) {
       perror("pthread_key_create(): failed to allocate key for thread specific data");
       exit(-1);
@@ -245,7 +245,7 @@ pipe_tsd_get(pipe_tsd *tsd)
    if (tsd->initMagic != (int) PIPE_TSD_INIT_MAGIC) {
       pipe_tsd_init(tsd);
    }
-#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_HAIKU)
+#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) || defined(PIPE_OS_HAIKU)
    return pthread_getspecific(tsd->key);
 #elif defined(PIPE_SUBSYSTEM_WINDOWS_USER)
    assert(0);
@@ -262,7 +262,7 @@ pipe_tsd_set(pipe_tsd *tsd, void *value)
    if (tsd->initMagic != (int) PIPE_TSD_INIT_MAGIC) {
       pipe_tsd_init(tsd);
    }
-#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_HAIKU)
+#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) || defined(PIPE_OS_HAIKU)
    if (pthread_setspecific(tsd->key, value) != 0) {
       perror("pthread_set_specific() failed");
       exit(-1);

src/gallium/state_trackers/g3dvl/vl_basic_csc.c

@@ -51,6 +51,7 @@ static int vlResizeFrameBuffer
 	struct vlBasicCSC	*basic_csc;
 	struct pipe_context	*pipe;
 	struct pipe_texture	template;
+	float			clear_color[4];
 
 	assert(csc);
 
@@ -68,7 +69,12 @@ static int vlResizeFrameBuffer
 	basic_csc->viewport.translate[1] = 0;
 	basic_csc->viewport.translate[2] = 0;
 	basic_csc->viewport.translate[3] = 0;
-	
+
+	clear_color[0]	= 0.0f;
+	clear_color[1]  = 0.0f;
+	clear_color[2]  = 0.0f;
+	clear_color[3]  = 0.0f;
+
 	if (basic_csc->framebuffer_tex)
 	{
 		pipe_surface_reference(&basic_csc->framebuffer.cbufs[0], NULL);
@@ -98,7 +104,7 @@ static int vlResizeFrameBuffer
 
 	/* Clear to black, in case video doesn't fill the entire window */
 	pipe->set_framebuffer_state(pipe, &basic_csc->framebuffer);
-	pipe->clear(pipe, PIPE_CLEAR_COLOR, 0, 0.0f, 0);
+	pipe->clear(pipe, PIPE_CLEAR_COLOR, clear_color, 0.0f, 0);
 
 	return 0;
 }

src/gallium/state_trackers/xorg/xorg_composite.c

@@ -12,9 +12,9 @@ struct xorg_composite_blend {
    int op:8;
 
    unsigned rgb_src_factor:5;    /**< PIPE_BLENDFACTOR_x */
-   unsigned rgb_dst_factor:5;    /**< PIPE_BLENDFACTOR_x */
-
    unsigned alpha_src_factor:5;  /**< PIPE_BLENDFACTOR_x */
+
+   unsigned rgb_dst_factor:5;    /**< PIPE_BLENDFACTOR_x */
    unsigned alpha_dst_factor:5;  /**< PIPE_BLENDFACTOR_x */
 };
 
@@ -653,17 +653,12 @@ boolean xorg_solid_bind_state(struct exa_context *exa,
 
    exa->solid_color[3] = 1.f;
 
+#if 0
    debug_printf("Color Pixel=(%d, %d, %d, %d), RGBA=(%f, %f, %f, %f)\n",
                 (fg >> 24) & 0xff, (fg >> 16) & 0xff,
                 (fg >> 8) & 0xff,  (fg >> 0) & 0xff,
                 exa->solid_color[0], exa->solid_color[1],
                 exa->solid_color[2], exa->solid_color[3]);
-
-#if 0
-   exa->solid_color[0] = 1.f;
-   exa->solid_color[1] = 0.f;
-   exa->solid_color[2] = 0.f;
-   exa->solid_color[3] = 1.f;
 #endif
 
    vs_traits = VS_SOLID_FILL;
@@ -691,9 +686,6 @@ void xorg_solid(struct exa_context *exa,
    struct pipe_buffer *buf = 0;
    float vertices[4][2][4];
 
-   x0 = 10; y0 = 10;
-   x1 = 300; y1 = 300;
-
    /* 1st vertex */
    setup_vertex0(vertices[0], x0, y0,
                  exa->solid_color);
@@ -844,11 +836,6 @@ static void renderer_copy_texture(struct exa_context *exa,
    assert(dst->width[0] != 0);
    assert(dst->height[0] != 0);
 
-#if 0
-   debug_printf("copy texture [%f, %f, %f, %f], [%f, %f, %f, %f]\n",
-                sx1, sy1, sx2, sy2, dx1, dy1, dx2, dy2);
-#endif
-
 #if 1
    s0 = sx1 / src->width[0];
    s1 = sx2 / src->width[0];
@@ -861,8 +848,16 @@ static void renderer_copy_texture(struct exa_context *exa,
    t1 = 1;
 #endif
 
-   assert(screen->is_format_supported(screen, dst_surf->format, PIPE_TEXTURE_2D,
-                                      PIPE_TEXTURE_USAGE_RENDER_TARGET, 0));
+#if 1
+   debug_printf("copy texture src=[%f, %f, %f, %f], dst=[%f, %f, %f, %f], tex=[%f, %f, %f, %f]\n",
+                sx1, sy1, sx2, sy2, dx1, dy1, dx2, dy2,
+                s0, t0, s1, t1);
+#endif
+
+   assert(screen->is_format_supported(screen, dst_surf->format,
+                                      PIPE_TEXTURE_2D,
+                                      PIPE_TEXTURE_USAGE_RENDER_TARGET,
+                                      0));
 
    /* save state (restored below) */
    cso_save_blend(exa->cso);
@@ -926,6 +921,8 @@ static void renderer_copy_texture(struct exa_context *exa,
          fb.cbufs[i] = 0;
    }
    cso_set_framebuffer(exa->cso, &fb);
+   setup_vs_constant_buffer(exa, fb.width, fb.height);
+   setup_fs_constant_buffer(exa);
 
    /* draw quad */
    buf = setup_vertex_data_tex(exa,
@@ -955,6 +952,61 @@ static void renderer_copy_texture(struct exa_context *exa,
    pipe_surface_reference(&dst_surf, NULL);
 }
 
+
+static struct pipe_texture *
+create_sampler_texture(struct exa_context *ctx,
+                       struct pipe_texture *src)
+{
+   enum pipe_format format;
+   struct pipe_context *pipe = ctx->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_texture *pt;
+   struct pipe_texture templ;
+
+   pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
+
+   /* the coming in texture should already have that invariance */
+   debug_assert(screen->is_format_supported(screen, src->format,
+                                            PIPE_TEXTURE_2D,
+                                            PIPE_TEXTURE_USAGE_SAMPLER, 0));
+
+   format = src->format;
+
+   memset(&templ, 0, sizeof(templ));
+   templ.target = PIPE_TEXTURE_2D;
+   templ.format = format;
+   templ.last_level = 0;
+   templ.width[0] = src->width[0];
+   templ.height[0] = src->height[0];
+   templ.depth[0] = 1;
+   pf_get_block(format, &templ.block);
+   templ.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER;
+
+   pt = screen->texture_create(screen, &templ);
+
+   debug_assert(!pt || pipe_is_referenced(&pt->reference));
+
+   if (!pt)
+      return NULL;
+
+   {
+      /* copy source framebuffer surface into texture */
+      struct pipe_surface *ps_read = screen->get_tex_surface(
+         screen, src, 0, 0, 0, PIPE_BUFFER_USAGE_GPU_READ);
+      struct pipe_surface *ps_tex = screen->get_tex_surface(
+         screen, pt, 0, 0, 0, PIPE_BUFFER_USAGE_GPU_WRITE );
+      pipe->surface_copy(pipe,
+			 ps_tex, /* dest */
+			 0, 0, /* destx/y */
+			 ps_read,
+			 0, 0, src->width[0], src->height[0]);
+      pipe_surface_reference(&ps_read, NULL);
+      pipe_surface_reference(&ps_tex, NULL);
+   }
+
+   return pt;
+}
+
 void xorg_copy_pixmap(struct exa_context *ctx,
                       struct exa_pixmap_priv *dst_priv, int dx, int dy,
                       struct exa_pixmap_priv *src_priv, int sx, int sy,
@@ -966,6 +1018,8 @@ void xorg_copy_pixmap(struct exa_context *ctx,
    struct pipe_texture *dst = dst_priv->tex;
    struct pipe_texture *src = src_priv->tex;
 
+   xorg_exa_finish(ctx);
+
    dst_loc[0] = dx;
    dst_loc[1] = dy;
    dst_loc[2] = width;
@@ -1003,17 +1057,25 @@ void xorg_copy_pixmap(struct exa_context *ctx,
 
    if (src_loc[2] >= 0 && src_loc[3] >= 0 &&
        dst_loc[2] >= 0 && dst_loc[3] >= 0) {
+      struct pipe_texture *temp_src = src;
+
+      if (src == dst)
+         temp_src = create_sampler_texture(ctx, src);
+
       renderer_copy_texture(ctx,
-                            src,
+                            temp_src,
                             src_loc[0],
-                            src_loc[1] + src_loc[3],
-                            src_loc[0] + src_loc[2],
                             src_loc[1],
+                            src_loc[0] + src_loc[2],
+                            src_loc[1] + src_loc[3],
                             dst,
                             dst_loc[0],
-                            dst_loc[1] + dst_loc[3],
+                            dst_loc[1],
                             dst_loc[0] + dst_loc[2],
-                            dst_loc[1]);
+                            dst_loc[1] + dst_loc[3]);
+
+      if (src == dst)
+         pipe_texture_reference(&temp_src, NULL);
    }
 }
 

src/gallium/state_trackers/xorg/xorg_exa.c

@@ -47,6 +47,8 @@
 
 #include "util/u_rect.h"
 
+#define DEBUG_SOLID 0
+
 /*
  * Helper functions
  */
@@ -72,6 +74,9 @@ exa_get_pipe_format(int depth, enum pipe_format *format, int *bbp)
 	assert(*bbp == 16);
 	break;
     case 8:
+	*format = PIPE_FORMAT_A8_UNORM;
+	assert(*bbp == 8);
+	break;
     case 4:
     case 1:
 	*format = PIPE_FORMAT_A8R8G8B8_UNORM; /* bad bad bad */
@@ -253,7 +258,7 @@ ExaDone(PixmapPtr pPixmap)
 #if 1
     xorg_exa_flush(exa, PIPE_FLUSH_RENDER_CACHE, NULL);
 #else
-    xorg_finish(exa);
+    xorg_exa_finish(exa);
 #endif
     xorg_exa_common_done(exa);
 }
@@ -276,23 +281,35 @@ ExaPrepareSolid(PixmapPtr pPixmap, int alu, Pixel planeMask, Pixel fg)
     struct exa_pixmap_priv *priv = exaGetPixmapDriverPrivate(pPixmap);
     struct exa_context *exa = ms->exa;
 
+#if 0
     debug_printf("ExaPrepareSolid - test\n");
-    if (pPixmap->drawable.depth < 15)
-	return FALSE;
-
+#endif
     if (!EXA_PM_IS_SOLID(&pPixmap->drawable, planeMask))
 	return FALSE;
 
     if (!priv || !priv->tex)
 	return FALSE;
 
+    if (!exa->scrn->is_format_supported(exa->scrn, priv->tex->format,
+                                        priv->tex->target,
+                                        PIPE_TEXTURE_USAGE_RENDER_TARGET, 0))
+	return FALSE;
+
     if (alu != GXcopy)
 	return FALSE;
 
     if (!exa->pipe)
 	return FALSE;
 
+
+#if DEBUG_SOLID
+    fg = 0xffff0000;
+#endif
+
+#if 1
     debug_printf("  ExaPrepareSolid(0x%x)\n", fg);
+#endif
+
     return xorg_solid_bind_state(exa, priv, fg);
 }
 
@@ -310,11 +327,42 @@ ExaSolid(PixmapPtr pPixmap, int x0, int y0, int x1, int y1)
     if (x0 == 0 && y0 == 0 &&
         x1 == priv->tex->width[0] &&
         y1 == priv->tex->height[0]) {
-       exa->ctx->clear(exa->ctx, PIPE_CLEAR_COLOR | PIPE_CLEAR_DEPTHSTENCIL,
+       exa->ctx->clear(exa->pipe, PIPE_CLEAR_COLOR,
                        exa->solid_color, 1., 0);
     } else
 #endif
-       xorg_solid(exa, priv, x0, y0, x1, y1) ;
+
+#if DEBUG_SOLID
+       exa->solid_color[0] = 0.f;
+       exa->solid_color[1] = 1.f;
+       exa->solid_color[2] = 0.f;
+       exa->solid_color[3] = 1.f;
+    xorg_solid(exa, priv, 0, 0, 1024, 768);
+       exa->solid_color[0] = 1.f;
+       exa->solid_color[1] = 0.f;
+       exa->solid_color[2] = 0.f;
+       exa->solid_color[3] = 1.f;
+       xorg_solid(exa, priv, 0, 0, 300, 300);
+       xorg_solid(exa, priv, 300, 300, 350, 350);
+       xorg_solid(exa, priv, 350, 350, 500, 500);
+
+       xorg_solid(exa, priv,
+               priv->tex->width[0] - 10,
+               priv->tex->height[0] - 10,
+               priv->tex->width[0],
+               priv->tex->height[0]);
+
+    exa->solid_color[0] = 0.f;
+    exa->solid_color[1] = 0.f;
+    exa->solid_color[2] = 1.f;
+    exa->solid_color[3] = 1.f;
+
+    exa->has_solid_color = FALSE;
+    ExaPrepareCopy(pPixmap, pPixmap, 0, 0, GXcopy, 0xffffffff);
+    ExaCopy(pPixmap, 0, 0, 50, 50, 500, 500);
+#else
+    xorg_solid(exa, priv, x0, y0, x1, y1) ;
+#endif
 }
 
 static Bool
@@ -332,15 +380,20 @@ ExaPrepareCopy(PixmapPtr pSrcPixmap, PixmapPtr pDstPixmap, int xdir,
     if (alu != GXcopy)
 	return FALSE;
 
-    if (pSrcPixmap->drawable.depth < 15 || pDstPixmap->drawable.depth < 15)
-	return FALSE;
-
     if (!EXA_PM_IS_SOLID(&pSrcPixmap->drawable, planeMask))
 	return FALSE;
 
     if (!priv || !src_priv)
 	return FALSE;
 
+    if (!exa->scrn->is_format_supported(exa->scrn, priv->tex->format,
+                                        priv->tex->target,
+                                        PIPE_TEXTURE_USAGE_RENDER_TARGET, 0) ||
+        !exa->scrn->is_format_supported(exa->scrn, src_priv->tex->format,
+                                        src_priv->tex->target,
+                                        PIPE_TEXTURE_USAGE_SAMPLER, 0))
+	return FALSE;
+
     if (!priv->tex || !src_priv->tex)
 	return FALSE;
 
@@ -433,14 +486,11 @@ static void
 ExaDestroyPixmap(ScreenPtr pScreen, void *dPriv)
 {
     struct exa_pixmap_priv *priv = (struct exa_pixmap_priv *)dPriv;
-    ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum];
-    modesettingPtr ms = modesettingPTR(pScrn);
 
     if (!priv)
 	return;
 
-    if (priv->tex)
-	ms->screen->texture_destroy(priv->tex);
+    pipe_texture_reference(&priv->tex, NULL);
 
     xfree(priv);
 }
@@ -583,8 +633,8 @@ ExaModifyPixmapHeader(PixmapPtr pPixmap, int width, int height,
 		struct pipe_surface *dst_surf;
                 struct pipe_surface *src_surf;
 
-		dst_surf = exa->scrn->get_tex_surface(exa->scrn, texture, 0, 0, 0,
-						      PIPE_BUFFER_USAGE_GPU_WRITE);
+		dst_surf = exa->scrn->get_tex_surface(
+                   exa->scrn, texture, 0, 0, 0, PIPE_BUFFER_USAGE_GPU_WRITE);
 		src_surf = exa_gpu_surface(exa, priv);
 		exa->pipe->surface_copy(exa->pipe, dst_surf, 0, 0, src_surf,
                                         0, 0, min(width, texture->width[0]),

src/gallium/state_trackers/xorg/xorg_exa_tgsi.c

@@ -52,8 +52,7 @@ struct xorg_shaders {
 
 static const char over_op[] =
    "SUB TEMP[3], CONST[0].wwww, TEMP[1].wwww\n"
-   "MUL TEMP[3], TEMP[0], TEMP[3]\n"
-   "ADD TEMP[0], TEMP[3], TEMP[0]\n";
+   "MAD TEMP[3], TEMP[0], TEMP[3], TEMP[0]\n";
 
 
 static INLINE void
@@ -79,8 +78,7 @@ vs_normalize_coords(struct ureg_program *ureg, struct ureg_src coords,
 {
    struct ureg_dst tmp = ureg_DECL_temporary(ureg);
    struct ureg_src ret;
-   ureg_MUL(ureg, tmp, coords, const0);
-   ureg_ADD(ureg, tmp, ureg_src(tmp), const1);
+   ureg_MAD(ureg, tmp, coords, const0, const1);
    ret = ureg_src(tmp);
    ureg_release_temporary(ureg, tmp);
    return ret;
@@ -261,7 +259,7 @@ create_vs(struct pipe_context *pipe,
 
    if (is_composite) {
       src = ureg_DECL_vs_input(ureg, input_slot++);
-      dst = ureg_DECL_output(ureg, TGSI_SEMANTIC_GENERIC, 1);
+      dst = ureg_DECL_output(ureg, TGSI_SEMANTIC_GENERIC, 0);
       ureg_MOV(ureg, dst, src);
    }
 
@@ -312,7 +310,7 @@ create_fs(struct pipe_context *pipe,
    if (is_composite) {
       src_sampler = ureg_DECL_sampler(ureg, 0);
       src_input = ureg_DECL_fs_input(ureg,
-                                     TGSI_SEMANTIC_POSITION,
+                                     TGSI_SEMANTIC_GENERIC,
                                      0,
                                      TGSI_INTERPOLATE_PERSPECTIVE);
    } else {
@@ -367,11 +365,11 @@ create_fs(struct pipe_context *pipe,
          else
             src = out;
 
-         coords = ureg_DECL_constant(ureg);
-         const0124 = ureg_DECL_constant(ureg);
-         matrow0 = ureg_DECL_constant(ureg);
-         matrow1 = ureg_DECL_constant(ureg);
-         matrow2 = ureg_DECL_constant(ureg);
+         coords = ureg_DECL_constant(ureg, 0);
+         const0124 = ureg_DECL_constant(ureg, 1);
+         matrow0 = ureg_DECL_constant(ureg, 2);
+         matrow1 = ureg_DECL_constant(ureg, 3);
+         matrow2 = ureg_DECL_constant(ureg, 4);
 
          if (is_lingrad) {
             linear_gradient(ureg, src,

src/gallium/winsys/g3dvl/xsp_winsys.c

@@ -105,6 +105,7 @@ static struct pipe_buffer* xsp_surface_buffer_create
 	unsigned height,
 	enum pipe_format format,
 	unsigned usage,
+	unsigned tex_usage,
 	unsigned *stride
 )
 {

src/gallium/winsys/gdi/SConscript

@@ -5,35 +5,39 @@ Import('*')
 
 if env['platform'] == 'windows':
 
-	env = env.Clone()
+    env = env.Clone()
 
-	env.Append(CPPPATH = [
-		'#src/gallium/state_trackers/wgl',
-	])
+    env.Append(CPPPATH = [
+        '#src/gallium/state_trackers/wgl',
+    ])
 
-	env.Append(LIBS = [
-		'gdi32',
-		'user32',
-		'kernel32',
-		'ws2_32',
-	])
+    env.Append(LIBS = [
+        'gdi32',
+        'user32',
+        'kernel32',
+        'ws2_32',
+    ])
 
-	sources = [
-		'gdi_softpipe_winsys.c',
-	]
-	
-	if env['gcc']:
-		sources += ['#src/gallium/state_trackers/wgl/opengl32.mingw.def']
-	else:
-		sources += ['#src/gallium/state_trackers/wgl/opengl32.def']
-		
-	drivers = [
-		trace,
-		softpipe,
-	]
+    if 'llvmpipe' in env['drivers']:
+        sources = ['gdi_llvmpipe_winsys.c']
+        drivers = [llvmpipe]
+        env.Tool('llvm')
+    elif 'softpipe' in env['drivers']:
+        sources = ['gdi_softpipe_winsys.c']
+        drivers = [softpipe]
+    else:
+        print 'warning: softpipe or llvmpipe not selected, gdi winsys disabled'
+        Return()
+    
+    if env['gcc']:
+        sources += ['#src/gallium/state_trackers/wgl/opengl32.mingw.def']
+    else:
+        sources += ['#src/gallium/state_trackers/wgl/opengl32.def']
+        
+    drivers += [trace]
 
-	env.SharedLibrary(
-		target ='opengl32',
-		source = sources,
-		LIBS = wgl + glapi + mesa + drivers + auxiliaries + glsl + env['LIBS'],
-	)
+    env.SharedLibrary(
+        target ='opengl32',
+        source = sources,
+        LIBS = wgl + glapi + mesa + drivers + auxiliaries + glsl + env['LIBS'],
+    )

src/gallium/winsys/gdi/gdi_llvmpipe_winsys.c

@@ -0,0 +1,284 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * 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, sub license, 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 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 NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS 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.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * LLVMpipe support.
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+
+#include <windows.h>
+
+#include "pipe/p_format.h"
+#include "pipe/p_context.h"
+#include "pipe/p_inlines.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "llvmpipe/lp_winsys.h"
+#include "llvmpipe/lp_texture.h"
+#include "shared/stw_winsys.h"
+
+
+struct gdi_llvmpipe_displaytarget
+{
+   enum pipe_format format;
+   struct pipe_format_block block;
+   unsigned width;
+   unsigned height;
+   unsigned stride;
+
+   unsigned size;
+
+   void *data;
+
+   BITMAPINFO bmi;
+};
+
+
+/** Cast wrapper */
+static INLINE struct gdi_llvmpipe_displaytarget *
+gdi_llvmpipe_displaytarget( struct llvmpipe_displaytarget *buf )
+{
+   return (struct gdi_llvmpipe_displaytarget *)buf;
+}
+
+
+static boolean
+gdi_llvmpipe_is_displaytarget_format_supported( struct llvmpipe_winsys *ws,
+                                                enum pipe_format format )
+{
+   switch(format) {
+   case PIPE_FORMAT_X8R8G8B8_UNORM:
+   case PIPE_FORMAT_A8R8G8B8_UNORM:
+      return TRUE;
+
+   /* TODO: Support other formats possible with BMPs, as described in 
+    * http://msdn.microsoft.com/en-us/library/dd183376(VS.85).aspx */
+      
+   default:
+      return FALSE;
+   }
+}
+
+
+static void *
+gdi_llvmpipe_displaytarget_map(struct llvmpipe_winsys *ws,
+                               struct llvmpipe_displaytarget *dt,
+                               unsigned flags )
+{
+   struct gdi_llvmpipe_displaytarget *gdt = gdi_llvmpipe_displaytarget(dt);
+
+   return gdt->data;
+}
+
+
+static void
+gdi_llvmpipe_displaytarget_unmap(struct llvmpipe_winsys *ws,
+                                 struct llvmpipe_displaytarget *dt )
+{
+
+}
+
+
+static void
+gdi_llvmpipe_displaytarget_destroy(struct llvmpipe_winsys *winsys,
+                                   struct llvmpipe_displaytarget *dt)
+{
+   struct gdi_llvmpipe_displaytarget *gdt = gdi_llvmpipe_displaytarget(dt);
+
+   align_free(gdt->data);
+   FREE(gdt);
+}
+
+
+/**
+ * Round n up to next multiple.
+ */
+static INLINE unsigned
+round_up(unsigned n, unsigned multiple)
+{
+   return (n + multiple - 1) & ~(multiple - 1);
+}
+
+
+static struct llvmpipe_displaytarget *
+gdi_llvmpipe_displaytarget_create(struct llvmpipe_winsys *winsys,
+                                  enum pipe_format format,
+                                  unsigned width, unsigned height,
+                                  unsigned alignment,
+                                  unsigned *stride)
+{
+   struct gdi_llvmpipe_displaytarget *gdt;
+   unsigned cpp;
+   unsigned bpp;
+   
+   gdt = CALLOC_STRUCT(gdi_llvmpipe_displaytarget);
+   if(!gdt)
+      goto no_gdt;
+
+   gdt->format = format;
+   gdt->width = width;
+   gdt->height = height;
+
+   bpp = pf_get_bits(format);
+   cpp = pf_get_size(format);
+   
+   gdt->stride = round_up(width * cpp, alignment);
+   gdt->size = gdt->stride * height;
+   
+   gdt->data = align_malloc(gdt->size, alignment);
+   if(!gdt->data)
+      goto no_data;
+
+   gdt->bmi.bmiHeader.biSize = sizeof(BITMAPINFOHEADER);
+   gdt->bmi.bmiHeader.biWidth = gdt->stride / cpp;
+   gdt->bmi.bmiHeader.biHeight= -(long)height;
+   gdt->bmi.bmiHeader.biPlanes = 1;
+   gdt->bmi.bmiHeader.biBitCount = bpp;
+   gdt->bmi.bmiHeader.biCompression = BI_RGB;
+   gdt->bmi.bmiHeader.biSizeImage = 0;
+   gdt->bmi.bmiHeader.biXPelsPerMeter = 0;
+   gdt->bmi.bmiHeader.biYPelsPerMeter = 0;
+   gdt->bmi.bmiHeader.biClrUsed = 0;
+   gdt->bmi.bmiHeader.biClrImportant = 0;
+
+   *stride = gdt->stride;
+   return (struct llvmpipe_displaytarget *)gdt;
+
+no_data:
+   FREE(gdt);
+no_gdt:
+   return NULL;
+}
+
+
+static void
+gdi_llvmpipe_displaytarget_display(struct llvmpipe_winsys *winsys, 
+                                   struct llvmpipe_displaytarget *dt,
+                                   void *context_private)
+{
+   assert(0);
+}
+
+
+static void
+gdi_llvmpipe_destroy(struct llvmpipe_winsys *winsys)
+{
+   FREE(winsys);
+}
+
+
+static struct pipe_screen *
+gdi_llvmpipe_screen_create(void)
+{
+   static struct llvmpipe_winsys *winsys;
+   struct pipe_screen *screen;
+
+   winsys = CALLOC_STRUCT(llvmpipe_winsys);
+   if(!winsys)
+      goto no_winsys;
+
+   winsys->destroy = gdi_llvmpipe_destroy;
+   winsys->is_displaytarget_format_supported = gdi_llvmpipe_is_displaytarget_format_supported;
+   winsys->displaytarget_create = gdi_llvmpipe_displaytarget_create;
+   winsys->displaytarget_map = gdi_llvmpipe_displaytarget_map;
+   winsys->displaytarget_unmap = gdi_llvmpipe_displaytarget_unmap;
+   winsys->displaytarget_display = gdi_llvmpipe_displaytarget_display;
+   winsys->displaytarget_destroy = gdi_llvmpipe_displaytarget_destroy;
+
+   screen = llvmpipe_create_screen(winsys);
+   if(!screen)
+      goto no_screen;
+
+   return screen;
+   
+no_screen:
+   FREE(winsys);
+no_winsys:
+   return NULL;
+}
+
+
+static struct pipe_context *
+gdi_llvmpipe_context_create(struct pipe_screen *screen)
+{
+   return llvmpipe_create(screen);
+}
+
+
+static void
+gdi_llvmpipe_flush_frontbuffer(struct pipe_screen *screen,
+                               struct pipe_surface *surface,
+                               HDC hDC)
+{
+    struct llvmpipe_texture *texture;
+    struct gdi_llvmpipe_displaytarget *gdt;
+
+    texture = llvmpipe_texture(surface->texture);
+    gdt = gdi_llvmpipe_displaytarget(texture->dt);
+
+    StretchDIBits(hDC,
+                  0, 0, gdt->width, gdt->height,
+                  0, 0, gdt->width, gdt->height,
+                  gdt->data, &gdt->bmi, 0, SRCCOPY);
+}
+
+
+static const struct stw_winsys stw_winsys = {
+   &gdi_llvmpipe_screen_create,
+   &gdi_llvmpipe_context_create,
+   &gdi_llvmpipe_flush_frontbuffer
+};
+
+
+BOOL WINAPI
+DllMain(HINSTANCE hinstDLL, DWORD fdwReason, LPVOID lpReserved)
+{
+   switch (fdwReason) {
+   case DLL_PROCESS_ATTACH:
+      if (!stw_init(&stw_winsys)) {
+         return FALSE;
+      }
+      return stw_init_thread();
+
+   case DLL_THREAD_ATTACH:
+      return stw_init_thread();
+
+   case DLL_THREAD_DETACH:
+      stw_cleanup_thread();
+      break;
+
+   case DLL_PROCESS_DETACH:
+      stw_cleanup_thread();
+      stw_cleanup();
+      break;
+   }
+   return TRUE;
+}

src/gallium/winsys/xlib/Makefile

@@ -34,6 +34,7 @@ XLIB_WINSYS_SOURCES = \
 	xlib_brw_aub.c \
 	xlib_brw_context.c \
 	xlib_brw_screen.c \
+	xlib_llvmpipe.c \
 	xlib_softpipe.c \
 	xlib_trace.c 
 

src/gallium/winsys/xlib/xlib_llvmpipe.c

@@ -33,6 +33,8 @@
  */
 
 
+#if defined(GALLIUM_LLVMPIPE)
+
 #include "xm_api.h"
 
 #undef ASSERT
@@ -459,3 +461,4 @@ struct xm_driver xlib_llvmpipe_driver =
 
 
 
+#endif /* GALLIUM_LLVMPIPE */

src/glx/x11/glx_pbuffer.c

@@ -40,6 +40,23 @@
 #include "glcontextmodes.h"
 
 
+/**
+ * Emit a warning when clients use GLX 1.3 functions on pre-1.3 systems.
+ */
+static void
+warn_GLX_1_3(Display *dpy, const char *function_name)
+{
+   __GLXdisplayPrivate *priv = __glXInitialize(dpy);
+
+   if (priv->minorVersion < 3) {
+      fprintf(stderr, 
+	      "WARNING: Application calling GLX 1.3 function \"%s\" "
+	      "when GLX 1.3 is not supported!  This is an application bug!\n",
+	      function_name);
+   }
+}
+
+
 /**
  * Change a drawable's attribute.
  *
@@ -559,6 +576,8 @@ glXCreatePbuffer(Display * dpy, GLXFBConfig config, const int *attrib_list)
    width = 0;
    height = 0;
 
+   warn_GLX_1_3(dpy, __func__);
+
    for (i = 0; attrib_list[i * 2]; i++) {
       switch (attrib_list[i * 2]) {
       case GLX_PBUFFER_WIDTH:
@@ -592,6 +611,7 @@ PUBLIC void
 glXQueryDrawable(Display * dpy, GLXDrawable drawable,
                  int attribute, unsigned int *value)
 {
+   warn_GLX_1_3(dpy, __func__);
    GetDrawableAttribute(dpy, drawable, attribute, value);
 }
 
@@ -645,6 +665,8 @@ PUBLIC GLXPixmap
 glXCreatePixmap(Display * dpy, GLXFBConfig config, Pixmap pixmap,
                 const int *attrib_list)
 {
+   warn_GLX_1_3(dpy, __func__);
+
    return CreateDrawable(dpy, (__GLcontextModes *) config,
                          (Drawable) pixmap, attrib_list, X_GLXCreatePixmap);
 }
@@ -654,6 +676,8 @@ PUBLIC GLXWindow
 glXCreateWindow(Display * dpy, GLXFBConfig config, Window win,
                 const int *attrib_list)
 {
+   warn_GLX_1_3(dpy, __func__);
+
    return CreateDrawable(dpy, (__GLcontextModes *) config,
                          (Drawable) win, attrib_list, X_GLXCreateWindow);
 }
@@ -662,6 +686,8 @@ glXCreateWindow(Display * dpy, GLXFBConfig config, Window win,
 PUBLIC void
 glXDestroyPixmap(Display * dpy, GLXPixmap pixmap)
 {
+   warn_GLX_1_3(dpy, __func__);
+
    DestroyDrawable(dpy, (GLXDrawable) pixmap, X_GLXDestroyPixmap);
 }
 
@@ -669,6 +695,8 @@ glXDestroyPixmap(Display * dpy, GLXPixmap pixmap)
 PUBLIC void
 glXDestroyWindow(Display * dpy, GLXWindow win)
 {
+   warn_GLX_1_3(dpy, __func__);
+
    DestroyDrawable(dpy, (GLXDrawable) win, X_GLXDestroyWindow);
 }
 

src/mesa/drivers/common/meta.c

@@ -67,6 +67,29 @@
 #define OFFSET(FIELD) ((void *) offsetof(struct vertex, FIELD))
 
 
+/**
+ * Flags passed to _mesa_meta_begin().
+ */
+/*@{*/
+#define META_ALL              ~0x0
+#define META_ALPHA_TEST        0x1
+#define META_BLEND             0x2  /**< includes logicop */
+#define META_COLOR_MASK        0x4
+#define META_DEPTH_TEST        0x8
+#define META_FOG              0x10
+#define META_PIXEL_STORE      0x20
+#define META_PIXEL_TRANSFER   0x40
+#define META_RASTERIZATION    0x80
+#define META_SCISSOR         0x100
+#define META_SHADER          0x200
+#define META_STENCIL_TEST    0x400
+#define META_TRANSFORM       0x800 /**< modelview, projection, clip planes */
+#define META_TEXTURE        0x1000
+#define META_VERTEX         0x2000
+#define META_VIEWPORT       0x4000
+/*@}*/
+
+
 /**
  * State which we may save/restore across meta ops.
  * XXX this may be incomplete...
@@ -94,6 +117,17 @@ struct save_state
    /** META_PIXEL_STORE */
    struct gl_pixelstore_attrib Pack, Unpack;
 
+   /** META_PIXEL_TRANSFER */
+   GLfloat RedBias, RedScale;
+   GLfloat GreenBias, GreenScale;
+   GLfloat BlueBias, BlueScale;
+   GLfloat AlphaBias, AlphaScale;
+   GLfloat DepthBias, DepthScale;
+   GLboolean MapColorFlag;
+   GLboolean Convolution1DEnabled;
+   GLboolean Convolution2DEnabled;
+   GLboolean Separable2DEnabled;
+
    /** META_RASTERIZATION */
    GLenum FrontPolygonMode, BackPolygonMode;
    GLboolean PolygonOffset;
@@ -363,6 +397,35 @@ _mesa_meta_begin(GLcontext *ctx, GLbitfield state)
       ctx->Unpack = ctx->DefaultPacking;
    }
 
+   if (state & META_PIXEL_TRANSFER) {
+      save->RedScale = ctx->Pixel.RedScale;
+      save->RedBias = ctx->Pixel.RedBias;
+      save->GreenScale = ctx->Pixel.GreenScale;
+      save->GreenBias = ctx->Pixel.GreenBias;
+      save->BlueScale = ctx->Pixel.BlueScale;
+      save->BlueBias = ctx->Pixel.BlueBias;
+      save->AlphaScale = ctx->Pixel.AlphaScale;
+      save->AlphaBias = ctx->Pixel.AlphaBias;
+      save->MapColorFlag = ctx->Pixel.MapColorFlag;
+      save->Convolution1DEnabled = ctx->Pixel.Convolution1DEnabled;
+      save->Convolution2DEnabled = ctx->Pixel.Convolution2DEnabled;
+      save->Separable2DEnabled = ctx->Pixel.Separable2DEnabled;
+      ctx->Pixel.RedScale = 1.0F;
+      ctx->Pixel.RedBias = 0.0F;
+      ctx->Pixel.GreenScale = 1.0F;
+      ctx->Pixel.GreenBias = 0.0F;
+      ctx->Pixel.BlueScale = 1.0F;
+      ctx->Pixel.BlueBias = 0.0F;
+      ctx->Pixel.AlphaScale = 1.0F;
+      ctx->Pixel.AlphaBias = 0.0F;
+      ctx->Pixel.MapColorFlag = GL_FALSE;
+      ctx->Pixel.Convolution1DEnabled = GL_FALSE;
+      ctx->Pixel.Convolution2DEnabled = GL_FALSE;
+      ctx->Pixel.Separable2DEnabled = GL_FALSE;
+      /* XXX more state */
+      ctx->NewState |=_NEW_PIXEL;
+   }
+
    if (state & META_RASTERIZATION) {
       save->FrontPolygonMode = ctx->Polygon.FrontMode;
       save->BackPolygonMode = ctx->Polygon.BackMode;
@@ -415,11 +478,6 @@ _mesa_meta_begin(GLcontext *ctx, GLbitfield state)
       save->ClientActiveUnit = ctx->Array.ActiveTexture;
       save->EnvMode = ctx->Texture.Unit[0].EnvMode;
 
-      if (ctx->Texture._EnabledUnits |
-          ctx->Texture._EnabledCoordUnits |
-          ctx->Texture._TexGenEnabled |
-          ctx->Texture._TexMatEnabled) {
-
       /* Disable all texture units */
       for (u = 0; u < ctx->Const.MaxTextureUnits; u++) {
          save->TexEnabled[u] = ctx->Texture.Unit[u].Enabled;
@@ -438,7 +496,6 @@ _mesa_meta_begin(GLcontext *ctx, GLbitfield state)
             _mesa_set_enable(ctx, GL_TEXTURE_GEN_Q, GL_FALSE);
          }
       }
-      }
 
       /* save current texture objects for unit[0] only */
       for (tgt = 0; tgt < NUM_TEXTURE_TARGETS; tgt++) {
@@ -564,6 +621,23 @@ _mesa_meta_end(GLcontext *ctx)
       ctx->Unpack = save->Unpack;
    }
 
+   if (state & META_PIXEL_TRANSFER) {
+      ctx->Pixel.RedScale = save->RedScale;
+      ctx->Pixel.RedBias = save->RedBias;
+      ctx->Pixel.GreenScale = save->GreenScale;
+      ctx->Pixel.GreenBias = save->GreenBias;
+      ctx->Pixel.BlueScale = save->BlueScale;
+      ctx->Pixel.BlueBias = save->BlueBias;
+      ctx->Pixel.AlphaScale = save->AlphaScale;
+      ctx->Pixel.AlphaBias = save->AlphaBias;
+      ctx->Pixel.MapColorFlag = save->MapColorFlag;
+      ctx->Pixel.Convolution1DEnabled = save->Convolution1DEnabled;
+      ctx->Pixel.Convolution2DEnabled = save->Convolution2DEnabled;
+      ctx->Pixel.Separable2DEnabled = save->Separable2DEnabled;
+      /* XXX more state */
+      ctx->NewState |=_NEW_PIXEL;
+   }
+
    if (state & META_RASTERIZATION) {
       _mesa_PolygonMode(GL_FRONT, save->FrontPolygonMode);
       _mesa_PolygonMode(GL_BACK, save->BackPolygonMode);
@@ -2037,3 +2111,231 @@ _mesa_meta_generate_mipmap(GLcontext *ctx, GLenum target,
    /* restore (XXX add to meta_begin/end()? */
    _mesa_BindFramebufferEXT(GL_FRAMEBUFFER_EXT, fboSave);
 }
+
+
+/**
+ * Determine the GL data type to use for the temporary image read with
+ * ReadPixels() and passed to Tex[Sub]Image().
+ */
+static GLenum
+get_temp_image_type(GLcontext *ctx, GLenum baseFormat)
+{
+   switch (baseFormat) {
+   case GL_RGBA:
+   case GL_RGB:
+   case GL_ALPHA:
+   case GL_LUMINANCE:
+   case GL_LUMINANCE_ALPHA:
+   case GL_INTENSITY:
+      if (ctx->DrawBuffer->Visual.redBits <= 8)
+         return GL_UNSIGNED_BYTE;
+      else if (ctx->DrawBuffer->Visual.redBits <= 8)
+         return GL_UNSIGNED_SHORT;
+      else
+         return GL_FLOAT;
+   case GL_DEPTH_COMPONENT:
+      return GL_UNSIGNED_INT;
+   case GL_DEPTH_STENCIL:
+      return GL_UNSIGNED_INT_24_8;
+   default:
+      _mesa_problem(ctx, "Unexpected format in get_temp_image_type()");
+      return 0;
+   }
+}
+
+
+/**
+ * Helper for _mesa_meta_CopyTexImage1/2D() functions.
+ * Have to be careful with locking and meta state for pixel transfer.
+ */
+static void
+copy_tex_image(GLcontext *ctx, GLuint dims, GLenum target, GLint level,
+               GLenum internalFormat, GLint x, GLint y,
+               GLsizei width, GLsizei height, GLint border)
+{
+   struct gl_texture_unit *texUnit;
+   struct gl_texture_object *texObj;
+   struct gl_texture_image *texImage;
+   GLenum format, type;
+   GLint bpp;
+   void *buf;
+
+   texUnit = _mesa_get_current_tex_unit(ctx);
+   texObj = _mesa_select_tex_object(ctx, texUnit, target);
+   texImage = _mesa_get_tex_image(ctx, texObj, target, level);
+
+   format = _mesa_base_tex_format(ctx, internalFormat);
+   type = get_temp_image_type(ctx, format);
+   bpp = _mesa_bytes_per_pixel(format, type);
+   if (bpp <= 0) {
+      _mesa_problem(ctx, "Bad bpp in meta copy_tex_image()");
+      return;
+   }
+
+   /*
+    * Alloc image buffer (XXX could use a PBO)
+    */
+   buf = _mesa_malloc(width * height * bpp);
+   if (!buf) {
+      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glCopyTexImage%uD", dims);
+      return;
+   }
+
+   _mesa_unlock_texture(ctx, texObj); /* need to unlock first */
+
+   /*
+    * Read image from framebuffer (disable pixel transfer ops)
+    */
+   _mesa_meta_begin(ctx, META_PIXEL_STORE | META_PIXEL_TRANSFER);
+   ctx->Driver.ReadPixels(ctx, x, y, width, height,
+			  format, type, &ctx->Pack, buf);
+   _mesa_meta_end(ctx);
+
+   /*
+    * Store texture data (with pixel transfer ops)
+    */
+   _mesa_meta_begin(ctx, META_PIXEL_STORE);
+   if (target == GL_TEXTURE_1D) {
+      ctx->Driver.TexImage1D(ctx, target, level, internalFormat,
+                             width, border, format, type,
+                             buf, &ctx->Unpack, texObj, texImage);
+   }
+   else {
+      ctx->Driver.TexImage2D(ctx, target, level, internalFormat,
+                             width, height, border, format, type,
+                             buf, &ctx->Unpack, texObj, texImage);
+   }
+   _mesa_meta_end(ctx);
+
+   _mesa_lock_texture(ctx, texObj); /* re-lock */
+
+   _mesa_free(buf);
+}
+
+
+void
+_mesa_meta_CopyTexImage1D(GLcontext *ctx, GLenum target, GLint level,
+                          GLenum internalFormat, GLint x, GLint y,
+                          GLsizei width, GLint border)
+{
+   copy_tex_image(ctx, 1, target, level, internalFormat, x, y,
+                  width, 1, border);
+}
+
+
+void
+_mesa_meta_CopyTexImage2D(GLcontext *ctx, GLenum target, GLint level,
+                          GLenum internalFormat, GLint x, GLint y,
+                          GLsizei width, GLsizei height, GLint border)
+{
+   copy_tex_image(ctx, 2, target, level, internalFormat, x, y,
+                  width, height, border);
+}
+
+
+
+/**
+ * Helper for _mesa_meta_CopyTexSubImage1/2/3D() functions.
+ * Have to be careful with locking and meta state for pixel transfer.
+ */
+static void
+copy_tex_sub_image(GLcontext *ctx, GLuint dims, GLenum target, GLint level,
+                   GLint xoffset, GLint yoffset, GLint zoffset,
+                   GLint x, GLint y,
+                   GLsizei width, GLsizei height)
+{
+   struct gl_texture_unit *texUnit;
+   struct gl_texture_object *texObj;
+   struct gl_texture_image *texImage;
+   GLenum format, type;
+   GLint bpp;
+   void *buf;
+
+   texUnit = _mesa_get_current_tex_unit(ctx);
+   texObj = _mesa_select_tex_object(ctx, texUnit, target);
+   texImage = _mesa_select_tex_image(ctx, texObj, target, level);
+
+   format = texImage->TexFormat->BaseFormat;
+   type = get_temp_image_type(ctx, format);
+   bpp = _mesa_bytes_per_pixel(format, type);
+   if (bpp <= 0) {
+      _mesa_problem(ctx, "Bad bpp in meta copy_tex_sub_image()");
+      return;
+   }
+
+   /*
+    * Alloc image buffer (XXX could use a PBO)
+    */
+   buf = _mesa_malloc(width * height * bpp);
+   if (!buf) {
+      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glCopyTexSubImage%uD", dims);
+      return;
+   }
+
+   _mesa_unlock_texture(ctx, texObj); /* need to unlock first */
+
+   /*
+    * Read image from framebuffer (disable pixel transfer ops)
+    */
+   _mesa_meta_begin(ctx, META_PIXEL_STORE | META_PIXEL_TRANSFER);
+   ctx->Driver.ReadPixels(ctx, x, y, width, height,
+			  format, type, &ctx->Pack, buf);
+   _mesa_meta_end(ctx);
+
+   /*
+    * Store texture data (with pixel transfer ops)
+    */
+   _mesa_meta_begin(ctx, META_PIXEL_STORE);
+   if (target == GL_TEXTURE_1D) {
+      ctx->Driver.TexSubImage1D(ctx, target, level, xoffset,
+                                width, format, type, buf,
+                                &ctx->Unpack, texObj, texImage);
+   }
+   else if (target == GL_TEXTURE_3D) {
+      ctx->Driver.TexSubImage3D(ctx, target, level, xoffset, yoffset, zoffset,
+                                width, height, 1, format, type, buf,
+                                &ctx->Unpack, texObj, texImage);
+   }
+   else {
+      ctx->Driver.TexSubImage2D(ctx, target, level, xoffset, yoffset,
+                                width, height, format, type, buf,
+                                &ctx->Unpack, texObj, texImage);
+   }
+   _mesa_meta_end(ctx);
+
+   _mesa_lock_texture(ctx, texObj); /* re-lock */
+
+   _mesa_free(buf);
+}
+
+
+void
+_mesa_meta_CopyTexSubImage1D(GLcontext *ctx, GLenum target, GLint level,
+                             GLint xoffset,
+                             GLint x, GLint y, GLsizei width)
+{
+   copy_tex_sub_image(ctx, 1, target, level, xoffset, 0, 0,
+                      x, y, width, 1);
+}
+
+
+void
+_mesa_meta_CopyTexSubImage2D(GLcontext *ctx, GLenum target, GLint level,
+                             GLint xoffset, GLint yoffset,
+                             GLint x, GLint y,
+                             GLsizei width, GLsizei height)
+{
+   copy_tex_sub_image(ctx, 2, target, level, xoffset, yoffset, 0,
+                      x, y, width, height);
+}
+
+
+void
+_mesa_meta_CopyTexSubImage3D(GLcontext *ctx, GLenum target, GLint level,
+                             GLint xoffset, GLint yoffset, GLint zoffset,
+                             GLint x, GLint y,
+                             GLsizei width, GLsizei height)
+{
+   copy_tex_sub_image(ctx, 3, target, level, xoffset, yoffset, zoffset,
+                      x, y, width, height);
+}

src/mesa/drivers/common/meta.h

@@ -27,29 +27,6 @@
 #define META_H
 
 
-/**
- * Flags passed to _mesa_meta_begin().
- * XXX these flags may evolve...
- */
-/*@{*/
-#define META_ALPHA_TEST      0x1
-#define META_BLEND           0x2  /**< includes logicop */
-#define META_COLOR_MASK      0x4
-#define META_DEPTH_TEST      0x8
-#define META_FOG            0x10
-#define META_RASTERIZATION  0x20
-#define META_SCISSOR        0x40
-#define META_SHADER         0x80
-#define META_STENCIL_TEST  0x100
-#define META_TRANSFORM     0x200 /**< modelview, projection */
-#define META_TEXTURE       0x400
-#define META_VERTEX        0x800
-#define META_VIEWPORT     0x1000
-#define META_PIXEL_STORE  0x2000
-#define META_ALL            ~0x0
-/*@}*/
-
-
 extern void
 _mesa_meta_init(GLcontext *ctx);
 
@@ -87,5 +64,32 @@ extern void
 _mesa_meta_generate_mipmap(GLcontext *ctx, GLenum target,
                            struct gl_texture_object *texObj);
 
+extern void
+_mesa_meta_CopyTexImage1D(GLcontext *ctx, GLenum target, GLint level,
+                          GLenum internalFormat, GLint x, GLint y,
+                          GLsizei width, GLint border);
+
+extern void
+_mesa_meta_CopyTexImage2D(GLcontext *ctx, GLenum target, GLint level,
+                          GLenum internalFormat, GLint x, GLint y,
+                          GLsizei width, GLsizei height, GLint border);
+
+extern void
+_mesa_meta_CopyTexSubImage1D(GLcontext *ctx, GLenum target, GLint level,
+                             GLint xoffset,
+                             GLint x, GLint y, GLsizei width);
+
+extern void
+_mesa_meta_CopyTexSubImage2D(GLcontext *ctx, GLenum target, GLint level,
+                             GLint xoffset, GLint yoffset,
+                             GLint x, GLint y,
+                             GLsizei width, GLsizei height);
+
+extern void
+_mesa_meta_CopyTexSubImage3D(GLcontext *ctx, GLenum target, GLint level,
+                             GLint xoffset, GLint yoffset, GLint zoffset,
+                             GLint x, GLint y,
+                             GLsizei width, GLsizei height);
+
 
 #endif /* META_H */

src/mesa/drivers/dri/ffb/ffb_tex.c

@@ -30,24 +30,6 @@
 #include "ffb_tex.h"
 
 /* No texture unit, all software. */
-/* XXX this function isn't needed since _mesa_init_driver_functions()
- * will make all these assignments.
- */
 void ffbDDInitTexFuncs(GLcontext *ctx)
 {
-        /*
-	ctx->Driver.ChooseTextureFormat = _mesa_choose_tex_format;
-	ctx->Driver.TexImage1D = _mesa_store_teximage1d;
-	ctx->Driver.TexImage2D = _mesa_store_teximage2d;
-	ctx->Driver.TexImage3D = _mesa_store_teximage3d;
-	ctx->Driver.TexSubImage1D = _mesa_store_texsubimage1d;
-	ctx->Driver.TexSubImage2D = _mesa_store_texsubimage2d;
-	ctx->Driver.TexSubImage3D = _mesa_store_texsubimage3d;
-	ctx->Driver.CopyTexImage1D = _swrast_copy_teximage1d;
-	ctx->Driver.CopyTexImage2D = _swrast_copy_teximage2d;
-	ctx->Driver.CopyTexSubImage1D = _swrast_copy_texsubimage1d;
-	ctx->Driver.CopyTexSubImage2D = _swrast_copy_texsubimage2d;
-	ctx->Driver.CopyTexSubImage3D = _swrast_copy_texsubimage3d;
-	ctx->Driver.TestProxyTexImage = _mesa_test_proxy_teximage;
-	*/
 }