Merge branch 'master' into glsl2

configs/autoconf.in

@@ -123,7 +123,7 @@ GLUT_LIB_DEPS = -L$(TOP)/$(LIB_DIR) @GLUT_MESA_DEPS@ \
 GLW_LIB_DEPS = -L$(TOP)/$(LIB_DIR) @GLW_MESA_DEPS@ \
 	$(EXTRA_LIB_PATH) @GLW_LIB_DEPS@
 APP_LIB_DEPS = $(EXTRA_LIB_PATH) @APP_LIB_DEPS@
-GLESv1_CM_LIB_DEPS = $(EXTRA_LIB_PATH) @GLESv1_LIB_DEPS@
+GLESv1_CM_LIB_DEPS = $(EXTRA_LIB_PATH) @GLESv1_CM_LIB_DEPS@
 GLESv2_LIB_DEPS = $(EXTRA_LIB_PATH) @GLESv2_LIB_DEPS@
 VG_LIB_DEPS = $(EXTRA_LIB_PATH) @VG_LIB_DEPS@
 

configure.ac

@@ -645,7 +645,7 @@ AC_SUBST([GL_PC_REQ_PRIV])
 AC_SUBST([GL_PC_LIB_PRIV])
 AC_SUBST([GL_PC_CFLAGS])
 AC_SUBST([DRI_PC_REQ_PRIV])
-AC_SUBST([GLESv1_LIB_DEPS])
+AC_SUBST([GLESv1_CM_LIB_DEPS])
 AC_SUBST([GLESv1_CM_PC_LIB_PRIV])
 AC_SUBST([GLESv2_LIB_DEPS])
 AC_SUBST([GLESv2_PC_LIB_PRIV])

src/egl/drivers/dri2/egl_dri2.c

@@ -177,9 +177,9 @@ EGLint dri2_to_egl_attribute_map[] = {
    EGL_Y_INVERTED_NOK,		/* __DRI_ATTRIB_YINVERTED */
 };
 
-static void
+static struct dri2_egl_config *
 dri2_add_config(_EGLDisplay *disp, const __DRIconfig *dri_config, int id,
-		int depth, xcb_visualtype_t *visual)
+		int depth, EGLint surface_type)
 {
    struct dri2_egl_config *conf;
    struct dri2_egl_display *dri2_dpy;
@@ -192,6 +192,10 @@ dri2_add_config(_EGLDisplay *disp, const __DRIconfig *dri_config, int id,
    _eglInitConfig(&base, disp, id);
    
    i = 0;
+   double_buffer = 0;
+   bind_to_texture_rgb = 0;
+   bind_to_texture_rgba = 0;
+
    while (dri2_dpy->core->indexConfigAttrib(dri_config, i++, &attrib, &value)) {
       switch (attrib) {
       case __DRI_ATTRIB_RENDER_TYPE:
@@ -242,35 +246,27 @@ dri2_add_config(_EGLDisplay *disp, const __DRIconfig *dri_config, int id,
     * return in the getBuffer callback to get the behaviour we want. */
 
    if (double_buffer)
-      return;
+      return NULL;
 
-   if (visual != NULL) {
+   if (depth > 0 && depth != _eglGetConfigKey(&base, EGL_BUFFER_SIZE))
-      if (depth != _eglGetConfigKey(&base, EGL_BUFFER_SIZE))
+      return NULL;
-	 return;
-
-      _eglSetConfigKey(&base, EGL_SURFACE_TYPE,
-		       EGL_WINDOW_BIT | EGL_PIXMAP_BIT | EGL_PBUFFER_BIT |
-		       EGL_SWAP_BEHAVIOR_PRESERVED_BIT);
-
-      _eglSetConfigKey(&base, EGL_NATIVE_VISUAL_ID, visual->visual_id);
-      _eglSetConfigKey(&base, EGL_NATIVE_VISUAL_TYPE, visual->_class);
-   } else {
-      _eglSetConfigKey(&base, EGL_SURFACE_TYPE,
-		       EGL_PIXMAP_BIT | EGL_PBUFFER_BIT);
-   }
 
    _eglSetConfigKey(&base, EGL_NATIVE_RENDERABLE, EGL_TRUE);
-   _eglSetConfigKey(&base, EGL_BIND_TO_TEXTURE_RGB, bind_to_texture_rgb);
+
-   if (_eglGetConfigKey(&base, EGL_ALPHA_SIZE) > 0)
+   _eglSetConfigKey(&base, EGL_SURFACE_TYPE, surface_type);
-      _eglSetConfigKey(&base,
+   if (surface_type & (EGL_PIXMAP_BIT | EGL_PBUFFER_BIT)) {
-		       EGL_BIND_TO_TEXTURE_RGBA, bind_to_texture_rgba);
+      _eglSetConfigKey(&base, EGL_BIND_TO_TEXTURE_RGB, bind_to_texture_rgb);
+      if (_eglGetConfigKey(&base, EGL_ALPHA_SIZE) > 0)
+	 _eglSetConfigKey(&base,
+			  EGL_BIND_TO_TEXTURE_RGBA, bind_to_texture_rgba);
+   }
 
    _eglSetConfigKey(&base, EGL_RENDERABLE_TYPE, disp->ClientAPIsMask);
    _eglSetConfigKey(&base, EGL_CONFORMANT, disp->ClientAPIsMask);
 
    if (!_eglValidateConfig(&base, EGL_FALSE)) {
       _eglLog(_EGL_DEBUG, "DRI2: failed to validate config %d", id);
-      return;
+      return NULL;
    }
 
    conf = malloc(sizeof *conf);
@@ -279,6 +275,8 @@ dri2_add_config(_EGLDisplay *disp, const __DRIconfig *dri_config, int id,
       conf->dri_config = dri_config;
       _eglAddConfig(disp, &conf->base);
    }
+
+   return conf;
 }
 
 /**
@@ -613,10 +611,19 @@ dri2_add_configs_for_visuals(struct dri2_egl_display *dri2_dpy,
    xcb_depth_iterator_t d;
    xcb_visualtype_t *visuals;
    int i, j, id;
+   struct dri2_egl_config *conf;
+   EGLint surface_type;
 
    s = xcb_setup_roots_iterator(xcb_get_setup(dri2_dpy->conn));
    d = xcb_screen_allowed_depths_iterator(s.data);
    id = 1;
+
+   surface_type =
+      EGL_WINDOW_BIT |
+      EGL_PIXMAP_BIT |
+      EGL_PBUFFER_BIT |
+      EGL_SWAP_BEHAVIOR_PRESERVED_BIT;
+
    while (d.rem > 0) {
       EGLBoolean class_added[6] = { 0, };
 
@@ -626,9 +633,16 @@ dri2_add_configs_for_visuals(struct dri2_egl_display *dri2_dpy,
 	    continue;
 
 	 class_added[visuals[i]._class] = EGL_TRUE;
-	 for (j = 0; dri2_dpy->driver_configs[j]; j++)
+	 for (j = 0; dri2_dpy->driver_configs[j]; j++) {
-	    dri2_add_config(disp, dri2_dpy->driver_configs[j],
+	    conf = dri2_add_config(disp, dri2_dpy->driver_configs[j],
-			    id++, d.data->depth, &visuals[i]);
+				   id++, d.data->depth, surface_type);
+	    if (conf == NULL)
+	       continue;
+	    _eglSetConfigKey(&conf->base,
+			     EGL_NATIVE_VISUAL_ID, visuals[i].visual_id);
+	    _eglSetConfigKey(&conf->base,
+			     EGL_NATIVE_VISUAL_TYPE, visuals[i]._class);
+	 }
       }
 
       xcb_depth_next(&d);      
@@ -738,6 +752,12 @@ dri2_create_screen(_EGLDisplay *disp)
    if (api_mask & (1 << __DRI_API_GLES2))
       disp->ClientAPIsMask |= EGL_OPENGL_ES2_BIT;
 
+   if (dri2_dpy->dri2->base.version >= 2) {
+      disp->Extensions.KHR_surfaceless_gles1 = EGL_TRUE;
+      disp->Extensions.KHR_surfaceless_gles2 = EGL_TRUE;
+      disp->Extensions.KHR_surfaceless_opengl = EGL_TRUE;
+   }
+
    return EGL_TRUE;
 
  cleanup_dri_screen:
@@ -948,6 +968,7 @@ dri2_initialize_drm(_EGLDriver *drv, _EGLDisplay *disp,
 		    EGLint *major, EGLint *minor)
 {
    struct dri2_egl_display *dri2_dpy;
+   int i;
 
    dri2_dpy = malloc(sizeof *dri2_dpy);
    if (!dri2_dpy)
@@ -970,10 +991,17 @@ dri2_initialize_drm(_EGLDriver *drv, _EGLDisplay *disp,
    if (!dri2_create_screen(disp))
       goto cleanup_driver;
 
+   for (i = 0; dri2_dpy->driver_configs[i]; i++)
+      dri2_add_config(disp, dri2_dpy->driver_configs[i], i + 1, 0, 0);
+
    disp->Extensions.KHR_image_base = EGL_TRUE;
    disp->Extensions.KHR_gl_renderbuffer_image = EGL_TRUE;
    disp->Extensions.KHR_gl_texture_2D_image = EGL_TRUE;
 
+   /* we're supporting EGL 1.4 */
+   *major = 1;
+   *minor = 4;
+
    return EGL_TRUE;
 
  cleanup_driver:
@@ -1041,6 +1069,7 @@ dri2_create_context(_EGLDriver *drv, _EGLDisplay *disp, _EGLConfig *conf,
    struct dri2_egl_display *dri2_dpy = dri2_egl_display(disp);
    struct dri2_egl_context *dri2_ctx_shared = dri2_egl_context(share_list);
    struct dri2_egl_config *dri2_config = dri2_egl_config(conf);
+   const __DRIconfig *dri_config;
    int api;
 
    dri2_ctx = malloc(sizeof *dri2_ctx);
@@ -1074,11 +1103,16 @@ dri2_create_context(_EGLDriver *drv, _EGLDisplay *disp, _EGLConfig *conf,
       return NULL;
    }
 
+   if (conf != NULL)
+      dri_config = dri2_config->dri_config;
+   else
+      dri_config = NULL;
+
    if (dri2_dpy->dri2->base.version >= 2) {
       dri2_ctx->dri_context =
 	 dri2_dpy->dri2->createNewContextForAPI(dri2_dpy->dri_screen,
 						api,
-						dri2_config->dri_config,
+						dri_config,
 						dri2_ctx_shared ? 
 						dri2_ctx_shared->dri_context : NULL,
 						dri2_ctx);

src/egl/main/eglapi.c

@@ -379,7 +379,11 @@ eglCreateContext(EGLDisplay dpy, EGLConfig config, EGLContext share_list,
    _EGLContext *context;
    EGLContext ret;
 
-   _EGL_CHECK_CONFIG(disp, conf, EGL_NO_CONTEXT, drv);
+   if (config)
+      _EGL_CHECK_CONFIG(disp, conf, EGL_NO_CONTEXT, drv);
+   else
+      drv = _eglCheckDisplay(disp, __FUNCTION__);
+
    if (!share && share_list != EGL_NO_CONTEXT)
       RETURN_EGL_ERROR(disp, EGL_BAD_CONTEXT, EGL_NO_CONTEXT);
 

src/egl/main/eglconfig.c

@@ -314,8 +314,10 @@ _eglValidateConfig(const _EGLConfig *conf, EGLBoolean for_matching)
                    EGL_VG_ALPHA_FORMAT_PRE_BIT |
                    EGL_MULTISAMPLE_RESOLVE_BOX_BIT |
                    EGL_SWAP_BEHAVIOR_PRESERVED_BIT;
+#ifdef EGL_MESA_screen_surface
             if (conf->Display->Extensions.MESA_screen_surface)
                mask |= EGL_SCREEN_BIT_MESA;
+#endif
             break;
          case EGL_RENDERABLE_TYPE:
          case EGL_CONFORMANT:

src/egl/main/eglcontext.c

@@ -83,7 +83,7 @@ _eglParseContextAttribList(_EGLContext *ctx, const EGLint *attrib_list)
       }
    }
 
-   if (err == EGL_SUCCESS) {
+   if (err == EGL_SUCCESS && ctx->Config) {
       EGLint renderable_type, api_bit;
 
       renderable_type = GET_CONFIG_ATTRIB(ctx->Config, EGL_RENDERABLE_TYPE);
@@ -220,45 +220,49 @@ _eglBindContextToSurfaces(_EGLContext *newCtx,
     * surface (oldDraw), the old bindings are broken first and the new one is
     * created.
     */
-   oldCtx = newDraw->CurrentContext;
+   if (newDraw) {
-   if (newCtx != oldCtx) {
+      oldCtx = newDraw->CurrentContext;
-      if (oldCtx) {
+      if (newCtx != oldCtx) {
-         assert(oldCtx->DrawSurface == newDraw);
+	 if (oldCtx) {
-         oldCtx->DrawSurface = NULL;
+	    assert(oldCtx->DrawSurface == newDraw);
+	    oldCtx->DrawSurface = NULL;
+	 }
+      
+	 newDraw->CurrentContext = newCtx;
       }
+   }
 
-      if (newCtx) {
+   if (newCtx) {
-         _EGLSurface *oldDraw = newCtx->DrawSurface;
+      _EGLSurface *oldDraw = newCtx->DrawSurface;
-         if (oldDraw)
+      if (oldDraw)
-            oldDraw->CurrentContext = NULL;
+	 oldDraw->CurrentContext = NULL;
 
-         newCtx->DrawSurface = newDraw;
+      newCtx->DrawSurface = newDraw;
-         *draw = oldDraw;
+      *draw = oldDraw;
-      }
-
-      newDraw->CurrentContext = newCtx;
    }
 
    /* likewise */
-   if (newRead != newDraw)
+   if (newRead && newRead != newDraw) {
       oldCtx = newRead->CurrentContext;
-   if (newCtx != oldCtx) {
+      if (newCtx != oldCtx) {
-      if (oldCtx) {
+	 if (oldCtx) {
-         assert(oldCtx->ReadSurface == newRead);
+	    assert(oldCtx->ReadSurface == newRead);
-         oldCtx->ReadSurface = NULL;
+	    oldCtx->ReadSurface = NULL;
+	 }
+
+	 newRead->CurrentContext = newCtx;
       }
-
-      if (newCtx) {
-         _EGLSurface *oldRead = newCtx->ReadSurface;
-         if (oldRead)
-            oldRead->CurrentContext = NULL;
-
-         newCtx->ReadSurface = newRead;
-         *read = oldRead;
-      }
-
-      newRead->CurrentContext = newCtx;
    }
+
+   if (newCtx) {
+      _EGLSurface *oldRead = newCtx->ReadSurface;
+      if (oldRead)
+	 oldRead->CurrentContext = NULL;
+
+      newCtx->ReadSurface = newRead;
+      *read = oldRead;
+   }
+
 }
 
 
@@ -297,7 +301,9 @@ static EGLBoolean
 _eglCheckMakeCurrent(_EGLContext *ctx, _EGLSurface *draw, _EGLSurface *read)
 {
    _EGLThreadInfo *t = _eglGetCurrentThread();
+   _EGLDisplay *dpy;
    EGLint conflict_api;
+   EGLBoolean surfaceless;
 
    if (_eglIsCurrentThreadDummy())
       return _eglError(EGL_BAD_ALLOC, "eglMakeCurrent");
@@ -309,8 +315,23 @@ _eglCheckMakeCurrent(_EGLContext *ctx, _EGLSurface *draw, _EGLSurface *read)
       return EGL_TRUE;
    }
 
-   /* ctx/draw/read must be all given */
+   dpy = ctx->Resource.Display;
-   if (draw == NULL || read == NULL)
+   switch (_eglGetContextAPIBit(ctx)) {
+   case EGL_OPENGL_ES_BIT:
+      surfaceless = dpy->Extensions.KHR_surfaceless_gles1;
+      break;
+   case EGL_OPENGL_ES2_BIT:
+      surfaceless = dpy->Extensions.KHR_surfaceless_gles2;
+      break;
+   case EGL_OPENGL_BIT:
+      surfaceless = dpy->Extensions.KHR_surfaceless_opengl;
+      break;
+   default:
+      surfaceless = EGL_FALSE;
+      break;
+   }
+
+   if (!surfaceless && (draw == NULL || read == NULL))
       return _eglError(EGL_BAD_MATCH, "eglMakeCurrent");
 
    /* context stealing from another thread is not allowed */
@@ -331,12 +352,13 @@ _eglCheckMakeCurrent(_EGLContext *ctx, _EGLSurface *draw, _EGLSurface *read)
     *
     * The latter is more restrictive so we can check only the latter case.
     */
-   if ((draw->CurrentContext && draw->CurrentContext != ctx) ||
+   if ((draw && draw->CurrentContext && draw->CurrentContext != ctx) ||
-       (read->CurrentContext && read->CurrentContext != ctx))
+       (read && read->CurrentContext && read->CurrentContext != ctx))
       return _eglError(EGL_BAD_ACCESS, "eglMakeCurrent");
 
    /* simply require the configs to be equal */
-   if (draw->Config != ctx->Config || read->Config != ctx->Config)
+   if ((draw && draw->Config != ctx->Config) ||
+       (read && read->Config != ctx->Config))
       return _eglError(EGL_BAD_MATCH, "eglMakeCurrent");
 
    switch (ctx->ClientAPI) {
@@ -387,7 +409,6 @@ _eglBindContext(_EGLContext **ctx, _EGLSurface **draw, _EGLSurface **read)
 
       *draw = oldCtx->DrawSurface;
       *read = oldCtx->ReadSurface;
-      assert(*draw && *read);
 
       _eglBindContextToSurfaces(NULL, draw, read);
    }

src/egl/main/eglcurrent.c

@@ -1,9 +1,9 @@
 #include <stdlib.h>
 #include <string.h>
-#include "eglglobals.h"
 #include "egllog.h"
 #include "eglmutex.h"
 #include "eglcurrent.h"
+#include "eglglobals.h"
 
 
 /* This should be kept in sync with _eglInitThreadInfo() */
@@ -300,12 +300,14 @@ _eglError(EGLint errCode, const char *msg)
       case EGL_BAD_SURFACE:
          s = "EGL_BAD_SURFACE";
          break;
+#ifdef EGL_MESA_screen_surface
       case EGL_BAD_SCREEN_MESA:
          s = "EGL_BAD_SCREEN_MESA";
          break;
       case EGL_BAD_MODE_MESA:
          s = "EGL_BAD_MODE_MESA";
          break;
+#endif
       default:
          s = "other";
       }

src/egl/main/egldisplay.h

@@ -60,6 +60,9 @@ struct _egl_extensions
    EGLBoolean KHR_gl_texture_cubemap_image;
    EGLBoolean KHR_gl_texture_3D_image;
    EGLBoolean KHR_gl_renderbuffer_image;
+   EGLBoolean KHR_surfaceless_gles1;
+   EGLBoolean KHR_surfaceless_gles2;
+   EGLBoolean KHR_surfaceless_opengl;
    EGLBoolean NOK_swap_region;
    EGLBoolean NOK_texture_from_pixmap;
 

src/egl/main/egldriver.c

@@ -14,7 +14,6 @@
 #include "egldefines.h"
 #include "egldisplay.h"
 #include "egldriver.h"
-#include "eglglobals.h"
 #include "egllog.h"
 #include "eglmisc.h"
 #include "eglmode.h"

src/egl/main/eglglobals.c

@@ -11,7 +11,6 @@ struct _egl_global _eglGlobal =
 {
    &_eglGlobalMutex,       /* Mutex */
    NULL,                   /* DisplayList */
-   1,                      /* FreeScreenHandle */
    2,                      /* NumAtExitCalls */
    {
       /* default AtExitCalls, called in reverse order */

src/egl/main/eglglobals.h

@@ -16,8 +16,6 @@ struct _egl_global
    /* the list of all displays */
    _EGLDisplay *DisplayList;
 
-   EGLScreenMESA FreeScreenHandle;
-
    EGLint NumAtExitCalls;
    void (*AtExitCalls[10])(void);
 };

src/egl/main/eglmisc.c

@@ -97,6 +97,10 @@ _eglUpdateExtensionsString(_EGLDisplay *dpy)
    _EGL_CHECK_EXTENSION(KHR_gl_texture_3D_image);
    _EGL_CHECK_EXTENSION(KHR_gl_renderbuffer_image);
 
+   _EGL_CHECK_EXTENSION(KHR_surfaceless_gles1);
+   _EGL_CHECK_EXTENSION(KHR_surfaceless_gles2);
+   _EGL_CHECK_EXTENSION(KHR_surfaceless_opengl);
+
    _EGL_CHECK_EXTENSION(NOK_swap_region);
    _EGL_CHECK_EXTENSION(NOK_texture_from_pixmap);
 #undef _EGL_CHECK_EXTENSION

src/egl/main/eglmode.c

@@ -10,6 +10,9 @@
 #include "eglstring.h"
 
 
+#ifdef EGL_MESA_screen_surface
+
+
 #define MIN2(A, B)  (((A) < (B)) ? (A) : (B))
 
 
@@ -353,3 +356,6 @@ _eglQueryModeStringMESA(_EGLDriver *drv, _EGLDisplay *dpy, _EGLMode *m)
 {
    return m->Name;
 }
+
+
+#endif /* EGL_MESA_screen_surface */

src/egl/main/eglmode.h

@@ -4,6 +4,9 @@
 #include "egltypedefs.h"
 
 
+#ifdef EGL_MESA_screen_surface
+
+
 #define EGL_NO_MODE_MESA 0
 
 
@@ -54,4 +57,7 @@ extern const char *
 _eglQueryModeStringMESA(_EGLDriver *drv, _EGLDisplay *dpy, _EGLMode *m);
 
 
+#endif /* EGL_MESA_screen_surface */
+
+
 #endif /* EGLMODE_INCLUDED */

src/egl/main/eglscreen.c

@@ -16,7 +16,6 @@
 #include <string.h>
 
 #include "egldisplay.h"
-#include "eglglobals.h"
 #include "eglcurrent.h"
 #include "eglmode.h"
 #include "eglconfig.h"
@@ -25,6 +24,14 @@
 #include "eglmutex.h"
 
 
+#ifdef EGL_MESA_screen_surface
+
+
+/* ugh, no atomic op? */
+static _EGL_DECLARE_MUTEX(_eglNextScreenHandleMutex);
+static EGLScreenMESA _eglNextScreenHandle = 1;
+
+
 /**
  * Return a new screen handle/ID.
  * NOTE: we never reuse these!
@@ -33,10 +40,10 @@ static EGLScreenMESA
 _eglAllocScreenHandle(void)
 {
    EGLScreenMESA s;
-   
+
-   _eglLockMutex(_eglGlobal.Mutex);
+   _eglLockMutex(&_eglNextScreenHandleMutex);
-   s = _eglGlobal.FreeScreenHandle++;
+   s = _eglNextScreenHandle++;
-   _eglUnlockMutex(_eglGlobal.Mutex);
+   _eglUnlockMutex(&_eglNextScreenHandleMutex);
 
    return s;
 }
@@ -263,3 +270,5 @@ _eglDestroyScreen(_EGLScreen *scrn)
    free(scrn);
 }
 
+
+#endif /* EGL_MESA_screen_surface */

src/egl/main/eglscreen.h

@@ -5,6 +5,9 @@
 #include "egltypedefs.h"
 
 
+#ifdef EGL_MESA_screen_surface
+
+
 /**
  * Per-screen information.
  * Note that an EGL screen doesn't have a size.  A screen may be set to
@@ -86,4 +89,7 @@ PUBLIC void
 _eglDestroyScreen(_EGLScreen *scrn);
 
 
+#endif /* EGL_MESA_screen_surface */
+
+
 #endif /* EGLSCREEN_INCLUDED */

src/egl/main/eglsurface.c

@@ -30,6 +30,50 @@ _eglClampSwapInterval(_EGLSurface *surf, EGLint interval)
 }
 
 
+#ifdef EGL_MESA_screen_surface
+static EGLint
+_eglParseScreenSurfaceAttribList(_EGLSurface *surf, const EGLint *attrib_list)
+{
+   EGLint i, err = EGL_SUCCESS;
+
+   if (!attrib_list)
+      return EGL_SUCCESS;
+
+   for (i = 0; attrib_list[i] != EGL_NONE; i++) {
+      EGLint attr = attrib_list[i++];
+      EGLint val = attrib_list[i];
+
+      switch (attr) {
+      case EGL_WIDTH:
+         if (val < 0) {
+            err = EGL_BAD_PARAMETER;
+            break;
+         }
+         surf->Width = val;
+         break;
+      case EGL_HEIGHT:
+         if (val < 0) {
+            err = EGL_BAD_PARAMETER;
+            break;
+         }
+         surf->Height = val;
+         break;
+      default:
+         err = EGL_BAD_ATTRIBUTE;
+         break;
+      }
+
+      if (err != EGL_SUCCESS) {
+         _eglLog(_EGL_WARNING, "bad surface attribute 0x%04x", attr);
+         break;
+      }
+   }
+
+   return err;
+}
+#endif /* EGL_MESA_screen_surface */
+
+
 /**
  * Parse the list of surface attributes and return the proper error code.
  */
@@ -44,6 +88,11 @@ _eglParseSurfaceAttribList(_EGLSurface *surf, const EGLint *attrib_list)
    if (!attrib_list)
       return EGL_SUCCESS;
 
+#ifdef EGL_MESA_screen_surface
+   if (type == EGL_SCREEN_BIT_MESA)
+      return _eglParseScreenSurfaceAttribList(surf, attrib_list);
+#endif
+
    if (dpy->Extensions.NOK_texture_from_pixmap)
       texture_type |= EGL_PIXMAP_BIT;
 
@@ -52,12 +101,8 @@ _eglParseSurfaceAttribList(_EGLSurface *surf, const EGLint *attrib_list)
       EGLint val = attrib_list[i];
 
       switch (attr) {
-      /* common (except for screen surfaces) attributes */
+      /* common attributes */
       case EGL_VG_COLORSPACE:
-         if (type == EGL_SCREEN_BIT_MESA) {
-            err = EGL_BAD_ATTRIBUTE;
-            break;
-         }
          switch (val) {
          case EGL_VG_COLORSPACE_sRGB:
          case EGL_VG_COLORSPACE_LINEAR:
@@ -71,10 +116,6 @@ _eglParseSurfaceAttribList(_EGLSurface *surf, const EGLint *attrib_list)
          surf->VGColorspace = val;
          break;
       case EGL_VG_ALPHA_FORMAT:
-         if (type == EGL_SCREEN_BIT_MESA) {
-            err = EGL_BAD_ATTRIBUTE;
-            break;
-         }
          switch (val) {
          case EGL_VG_ALPHA_FORMAT_NONPRE:
          case EGL_VG_ALPHA_FORMAT_PRE:
@@ -101,7 +142,7 @@ _eglParseSurfaceAttribList(_EGLSurface *surf, const EGLint *attrib_list)
          break;
       /* pbuffer surface attributes */
       case EGL_WIDTH:
-         if (type != EGL_PBUFFER_BIT && type != EGL_SCREEN_BIT_MESA) {
+         if (type != EGL_PBUFFER_BIT) {
             err = EGL_BAD_ATTRIBUTE;
             break;
          }
@@ -112,7 +153,7 @@ _eglParseSurfaceAttribList(_EGLSurface *surf, const EGLint *attrib_list)
          surf->Width = val;
          break;
       case EGL_HEIGHT:
-         if (type != EGL_PBUFFER_BIT && type != EGL_SCREEN_BIT_MESA) {
+         if (type != EGL_PBUFFER_BIT) {
             err = EGL_BAD_ATTRIBUTE;
             break;
          }
@@ -129,6 +170,7 @@ _eglParseSurfaceAttribList(_EGLSurface *surf, const EGLint *attrib_list)
          }
          surf->LargestPbuffer = !!val;
          break;
+      /* for eglBindTexImage */
       case EGL_TEXTURE_FORMAT:
          if (!(type & texture_type)) {
             err = EGL_BAD_ATTRIBUTE;
@@ -210,10 +252,12 @@ _eglInitSurface(_EGLSurface *surf, _EGLDisplay *dpy, EGLint type,
    case EGL_PBUFFER_BIT:
       func = "eglCreatePBufferSurface";
       break;
+#ifdef EGL_MESA_screen_surface
    case EGL_SCREEN_BIT_MESA:
       func = "eglCreateScreenSurface";
       renderBuffer = EGL_SINGLE_BUFFER; /* XXX correct? */
       break;
+#endif
    default:
       _eglLog(_EGL_WARNING, "Bad type in _eglInitSurface");
       return EGL_FALSE;

src/gallium/auxiliary/Makefile

@@ -149,6 +149,7 @@ C_SOURCES = \
 
 GALLIVM_SOURCES = \
         gallivm/lp_bld_arit.c \
+        gallivm/lp_bld_assert.c \
         gallivm/lp_bld_const.c \
         gallivm/lp_bld_conv.c \
         gallivm/lp_bld_debug.c \

src/gallium/auxiliary/SConscript

@@ -34,14 +34,14 @@ env.CodeGenerate(
     target = 'util/u_format_table.c',
     script = '#src/gallium/auxiliary/util/u_format_table.py',
     source = ['#src/gallium/auxiliary/util/u_format.csv'],
-    command = 'python $SCRIPT $SOURCE > $TARGET'
+    command = python_cmd + ' $SCRIPT $SOURCE > $TARGET'
 )
 
 env.CodeGenerate(
     target = 'util/u_half.c',
     script = 'util/u_half.py',
     source = [],
-    command = 'python $SCRIPT > $TARGET'
+    command = python_cmd + ' $SCRIPT > $TARGET'
 )
 
 env.Depends('util/u_format_table.c', [
@@ -198,6 +198,7 @@ source = [
 if env['llvm']:
     source += [
     'gallivm/lp_bld_arit.c',
+    'gallivm/lp_bld_assert.c',
     'gallivm/lp_bld_const.c',
     'gallivm/lp_bld_conv.c',
     'gallivm/lp_bld_debug.c',

src/gallium/auxiliary/draw/draw_context.c

@@ -288,12 +288,19 @@ draw_set_mapped_constant_buffer(struct draw_context *draw,
                 shader_type == PIPE_SHADER_GEOMETRY);
    debug_assert(slot < PIPE_MAX_CONSTANT_BUFFERS);
 
-   if (shader_type == PIPE_SHADER_VERTEX) {
+   switch (shader_type) {
+   case PIPE_SHADER_VERTEX:
       draw->pt.user.vs_constants[slot] = buffer;
+      draw->pt.user.vs_constants_size[slot] = size;
       draw_vs_set_constants(draw, slot, buffer, size);
-   } else if (shader_type == PIPE_SHADER_GEOMETRY) {
+      break;
+   case PIPE_SHADER_GEOMETRY:
       draw->pt.user.gs_constants[slot] = buffer;
+      draw->pt.user.gs_constants_size[slot] = size;
       draw_gs_set_constants(draw, slot, buffer, size);
+      break;
+   default:
+      assert(0 && "invalid shader type in draw_set_mapped_constant_buffer");
    }
 }
 

src/gallium/auxiliary/draw/draw_decompose_tmp.h

@@ -0,0 +1,425 @@
+/*
+ * Mesa 3-D graphics library
+ * Version:  7.9
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * Copyright (C) 2010 LunarG Inc.
+ *
+ * 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
+ * THE AUTHORS OR COPYRIGHT HOLDERS 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.
+ *
+ * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+ *    Chia-I Wu <olv@lunarg.com>
+ */
+
+/* these macros are optional */
+#ifndef LOCAL_VARS
+#define LOCAL_VARS
+#endif
+#ifndef FUNC_ENTER
+#define FUNC_ENTER do {} while (0)
+#endif
+#ifndef FUNC_EXIT
+#define FUNC_EXIT do {} while (0)
+#endif
+#ifndef LINE_ADJ
+#define LINE_ADJ(flags, a0, i0, i1, a1) LINE(flags, i0, i1)
+#endif
+#ifndef TRIANGLE_ADJ
+#define TRIANGLE_ADJ(flags, i0, a0, i1, a1, i2, a2) TRIANGLE(flags, i0, i1, i2)
+#endif
+
+static void
+FUNC(FUNC_VARS)
+{
+   unsigned idx[6], i;
+   ushort flags;
+   LOCAL_VARS
+
+   FUNC_ENTER;
+
+   /* prim, count, and last_vertex_last should have been defined */
+   if (0) {
+      debug_printf("%s: prim 0x%x, count %d, last_vertex_last %d\n",
+            __FUNCTION__, prim, count, last_vertex_last);
+   }
+
+   switch (prim) {
+   case PIPE_PRIM_POINTS:
+      for (i = 0; i < count; i++) {
+         idx[0] = GET_ELT(i);
+         POINT(idx[0]);
+      }
+      break;
+
+   case PIPE_PRIM_LINES:
+      flags = DRAW_PIPE_RESET_STIPPLE;
+      for (i = 0; i + 1 < count; i += 2) {
+         idx[0] = GET_ELT(i);
+         idx[1] = GET_ELT(i + 1);
+         LINE(flags, idx[0], idx[1]);
+      }
+      break;
+
+   case PIPE_PRIM_LINE_LOOP:
+   case PIPE_PRIM_LINE_STRIP:
+      if (count >= 2) {
+         flags = DRAW_PIPE_RESET_STIPPLE;
+         idx[1] = GET_ELT(0);
+         idx[2] = idx[1];
+
+         for (i = 1; i < count; i++, flags = 0) {
+            idx[0] = idx[1];
+            idx[1] = GET_ELT(i);
+            LINE(flags, idx[0], idx[1]);
+         }
+         /* close the loop */
+         if (prim == PIPE_PRIM_LINE_LOOP)
+            LINE(flags, idx[1], idx[2]);
+      }
+      break;
+
+   case PIPE_PRIM_TRIANGLES:
+      flags = DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL;
+      for (i = 0; i + 2 < count; i += 3) {
+         idx[0] = GET_ELT(i);
+         idx[1] = GET_ELT(i + 1);
+         idx[2] = GET_ELT(i + 2);
+         TRIANGLE(flags, idx[0], idx[1], idx[2]);
+      }
+      break;
+
+   case PIPE_PRIM_TRIANGLE_STRIP:
+      if (count >= 3) {
+         flags = DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL;
+         idx[1] = GET_ELT(0);
+         idx[2] = GET_ELT(1);
+
+         if (last_vertex_last) {
+            for (i = 0; i + 2 < count; i++) {
+               idx[0] = idx[1];
+               idx[1] = idx[2];
+               idx[2] = GET_ELT(i + 2);
+               /* always emit idx[2] last */
+               if (i & 1)
+                  TRIANGLE(flags, idx[1], idx[0], idx[2]);
+               else
+                  TRIANGLE(flags, idx[0], idx[1], idx[2]);
+            }
+         }
+         else {
+            for (i = 0; i + 2 < count; i++) {
+               idx[0] = idx[1];
+               idx[1] = idx[2];
+               idx[2] = GET_ELT(i + 2);
+               /* always emit idx[0] first */
+               if (i & 1)
+                  TRIANGLE(flags, idx[0], idx[2], idx[1]);
+               else
+                  TRIANGLE(flags, idx[0], idx[1], idx[2]);
+            }
+         }
+      }
+      break;
+
+   case PIPE_PRIM_TRIANGLE_FAN:
+      if (count >= 3) {
+         flags = DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL;
+         idx[0] = GET_ELT(0);
+         idx[2] = GET_ELT(1);
+
+         /* idx[0] is neither the first nor the last vertex */
+         if (last_vertex_last) {
+            for (i = 0; i + 2 < count; i++) {
+               idx[1] = idx[2];
+               idx[2] = GET_ELT(i + 2);
+               /* always emit idx[2] last */
+               TRIANGLE(flags, idx[0], idx[1], idx[2]);
+            }
+         }
+         else {
+            for (i = 0; i + 2 < count; i++) {
+               idx[1] = idx[2];
+               idx[2] = GET_ELT(i + 2);
+               /* always emit idx[1] first */
+               TRIANGLE(flags, idx[1], idx[2], idx[0]);
+            }
+         }
+      }
+      break;
+
+   case PIPE_PRIM_QUADS:
+      if (last_vertex_last) {
+         for (i = 0; i + 3 < count; i += 4) {
+            idx[0] = GET_ELT(i);
+            idx[1] = GET_ELT(i + 1);
+            idx[2] = GET_ELT(i + 2);
+            idx[3] = GET_ELT(i + 3);
+
+            flags = DRAW_PIPE_RESET_STIPPLE |
+                    DRAW_PIPE_EDGE_FLAG_0 |
+                    DRAW_PIPE_EDGE_FLAG_2;
+            /* always emit idx[3] last */
+            TRIANGLE(flags, idx[0], idx[1], idx[3]);
+
+            flags = DRAW_PIPE_EDGE_FLAG_0 |
+                    DRAW_PIPE_EDGE_FLAG_1;
+            TRIANGLE(flags, idx[1], idx[2], idx[3]);
+         }
+      }
+      else {
+         for (i = 0; i + 3 < count; i += 4) {
+            idx[0] = GET_ELT(i);
+            idx[1] = GET_ELT(i + 1);
+            idx[2] = GET_ELT(i + 2);
+            idx[3] = GET_ELT(i + 3);
+
+            flags = DRAW_PIPE_RESET_STIPPLE |
+                    DRAW_PIPE_EDGE_FLAG_0 |
+                    DRAW_PIPE_EDGE_FLAG_1;
+            /* XXX should always emit idx[0] first */
+            /* always emit idx[3] first */
+            TRIANGLE(flags, idx[3], idx[0], idx[1]);
+
+            flags = DRAW_PIPE_EDGE_FLAG_1 |
+                    DRAW_PIPE_EDGE_FLAG_2;
+            TRIANGLE(flags, idx[3], idx[1], idx[2]);
+         }
+      }
+      break;
+
+   case PIPE_PRIM_QUAD_STRIP:
+      if (count >= 4) {
+         idx[2] = GET_ELT(0);
+         idx[3] = GET_ELT(1);
+
+         if (last_vertex_last) {
+            for (i = 0; i + 3 < count; i += 2) {
+               idx[0] = idx[2];
+               idx[1] = idx[3];
+               idx[2] = GET_ELT(i + 2);
+               idx[3] = GET_ELT(i + 3);
+
+               /* always emit idx[3] last */
+               flags = DRAW_PIPE_RESET_STIPPLE |
+                       DRAW_PIPE_EDGE_FLAG_0 |
+                       DRAW_PIPE_EDGE_FLAG_2;
+               TRIANGLE(flags, idx[2], idx[0], idx[3]);
+
+               flags = DRAW_PIPE_EDGE_FLAG_0 |
+                       DRAW_PIPE_EDGE_FLAG_1;
+               TRIANGLE(flags, idx[0], idx[1], idx[3]);
+            }
+         }
+         else {
+            for (i = 0; i + 3 < count; i += 2) {
+               idx[0] = idx[2];
+               idx[1] = idx[3];
+               idx[2] = GET_ELT(i + 2);
+               idx[3] = GET_ELT(i + 3);
+
+               flags = DRAW_PIPE_RESET_STIPPLE |
+                       DRAW_PIPE_EDGE_FLAG_0 |
+                       DRAW_PIPE_EDGE_FLAG_1;
+               /* XXX should always emit idx[0] first */
+               /* always emit idx[3] first */
+               TRIANGLE(flags, idx[3], idx[2], idx[0]);
+
+               flags = DRAW_PIPE_EDGE_FLAG_1 |
+                       DRAW_PIPE_EDGE_FLAG_2;
+               TRIANGLE(flags, idx[3], idx[0], idx[1]);
+            }
+         }
+      }
+      break;
+
+   case PIPE_PRIM_POLYGON:
+      if (count >= 3) {
+         ushort edge_next, edge_finish;
+
+         if (last_vertex_last) {
+            flags = (DRAW_PIPE_RESET_STIPPLE |
+                     DRAW_PIPE_EDGE_FLAG_2 |
+                     DRAW_PIPE_EDGE_FLAG_0);
+            edge_next = DRAW_PIPE_EDGE_FLAG_0;
+            edge_finish = DRAW_PIPE_EDGE_FLAG_1;
+         }
+         else {
+            flags = (DRAW_PIPE_RESET_STIPPLE |
+                     DRAW_PIPE_EDGE_FLAG_0 |
+                     DRAW_PIPE_EDGE_FLAG_1);
+            edge_next = DRAW_PIPE_EDGE_FLAG_1;
+            edge_finish = DRAW_PIPE_EDGE_FLAG_2;
+         }
+
+         idx[0] = GET_ELT(0);
+         idx[2] = GET_ELT(1);
+
+         for (i = 0; i + 2 < count; i++, flags = edge_next) {
+            idx[1] = idx[2];
+            idx[2] = GET_ELT(i + 2);
+
+            if (i + 3 == count)
+               flags |= edge_finish;
+
+            /* idx[0] is both the first and the last vertex */
+            if (last_vertex_last)
+               TRIANGLE(flags, idx[1], idx[2], idx[0]);
+            else
+               TRIANGLE(flags, idx[0], idx[1], idx[2]);
+         }
+      }
+      break;
+
+   case PIPE_PRIM_LINES_ADJACENCY:
+      flags = DRAW_PIPE_RESET_STIPPLE;
+      for (i = 0; i + 3 < count; i += 4) {
+         idx[0] = GET_ELT(i);
+         idx[1] = GET_ELT(i + 1);
+         idx[2] = GET_ELT(i + 2);
+         idx[3] = GET_ELT(i + 3);
+         LINE_ADJ(flags, idx[0], idx[1], idx[2], idx[3]);
+      }
+      break;
+
+   case PIPE_PRIM_LINE_STRIP_ADJACENCY:
+      if (count >= 4) {
+         flags = DRAW_PIPE_RESET_STIPPLE;
+         idx[1] = GET_ELT(0);
+         idx[2] = GET_ELT(1);
+         idx[3] = GET_ELT(2);
+
+         for (i = 1; i + 2 < count; i++, flags = 0) {
+            idx[0] = idx[1];
+            idx[1] = idx[2];
+            idx[2] = idx[3];
+            idx[3] = GET_ELT(i + 2);
+            LINE_ADJ(flags, idx[0], idx[1], idx[2], idx[3]);
+         }
+      }
+      break;
+
+   case PIPE_PRIM_TRIANGLES_ADJACENCY:
+      flags = DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL;
+      for (i = 0; i + 5 < count; i += 6) {
+         idx[0] = GET_ELT(i);
+         idx[1] = GET_ELT(i + 1);
+         idx[2] = GET_ELT(i + 2);
+         idx[3] = GET_ELT(i + 3);
+         idx[4] = GET_ELT(i + 4);
+         idx[5] = GET_ELT(i + 5);
+         TRIANGLE_ADJ(flags, idx[0], idx[1], idx[2], idx[3], idx[4], idx[5]);
+      }
+      break;
+
+   case PIPE_PRIM_TRIANGLE_STRIP_ADJACENCY:
+      if (count >= 6) {
+         flags = DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL;
+         idx[0] = GET_ELT(1);
+         idx[2] = GET_ELT(0);
+         idx[4] = GET_ELT(2);
+         idx[3] = GET_ELT(4);
+
+         /*
+          * The vertices of the i-th triangle are stored in
+          * idx[0,2,4] = { 2*i, 2*i+2, 2*i+4 };
+          *
+          * The adjacent vertices are stored in
+          * idx[1,3,5] = { 2*i-2, 2*i+6, 2*i+3 }.
+          *
+          * However, there are two exceptions:
+          *
+          * For the first triangle, idx[1] = 1;
+          * For the  last triangle, idx[3] = 2*i+5.
+          */
+         if (last_vertex_last) {
+            for (i = 0; i + 5 < count; i += 2) {
+               idx[1] = idx[0];
+
+               idx[0] = idx[2];
+               idx[2] = idx[4];
+               idx[4] = idx[3];
+
+               idx[3] = GET_ELT(i + ((i + 7 < count) ? 6 : 5));
+               idx[5] = GET_ELT(i + 3);
+
+               /*
+                * alternate the first two vertices (idx[0] and idx[2]) and the
+                * corresponding adjacent vertices (idx[3] and idx[5]) to have
+                * the correct orientation
+                */
+               if (i & 2) {
+                  TRIANGLE_ADJ(flags,
+                        idx[2], idx[1], idx[0], idx[5], idx[4], idx[3]);
+               }
+               else {
+                  TRIANGLE_ADJ(flags,
+                        idx[0], idx[1], idx[2], idx[3], idx[4], idx[5]);
+               }
+            }
+         }
+         else {
+            for (i = 0; i + 5 < count; i += 2) {
+               idx[1] = idx[0];
+
+               idx[0] = idx[2];
+               idx[2] = idx[4];
+               idx[4] = idx[3];
+
+               idx[3] = GET_ELT(i + ((i + 7 < count) ? 6 : 5));
+               idx[5] = GET_ELT(i + 3);
+
+               /*
+                * alternate the last two vertices (idx[2] and idx[4]) and the
+                * corresponding adjacent vertices (idx[1] and idx[5]) to have
+                * the correct orientation
+                */
+               if (i & 2) {
+                  TRIANGLE_ADJ(flags,
+                        idx[0], idx[5], idx[4], idx[3], idx[2], idx[1]);
+               }
+               else {
+                  TRIANGLE_ADJ(flags,
+                        idx[0], idx[1], idx[2], idx[3], idx[4], idx[5]);
+               }
+            }
+         }
+      }
+      break;
+
+   default:
+      assert(0);
+      break;
+   }
+
+   FUNC_EXIT;
+}
+
+#undef LOCAL_VARS
+#undef FUNC_ENTER
+#undef FUNC_EXIT
+#undef LINE_ADJ
+#undef TRIANGLE_ADJ
+
+#undef FUNC
+#undef FUNC_VARS
+#undef GET_ELT
+#undef POINT
+#undef LINE
+#undef TRIANGLE

src/gallium/auxiliary/draw/draw_gs.c

@@ -1,6 +1,6 @@
 /**************************************************************************
  *
- * Copyright 2009 VMWare Inc.
+ * Copyright 2009 VMware, Inc.
  * All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
@@ -75,7 +75,10 @@ draw_gs_set_constants(struct draw_context *draw,
                       const void *constants,
                       unsigned size)
 {
-   /* noop */
+   /* noop. added here for symmetry with the VS
+    * code and in case we'll ever want to allign
+    * the constants, e.g. when we'll change to a
+    * different interpreter */
 }
 
 
@@ -370,32 +373,23 @@ static void gs_tri_adj(struct draw_geometry_shader *shader,
    gs_flush(shader, 1);
 }
 
-#define TRIANGLE(gs,i0,i1,i2) gs_tri(gs,i0,i1,i2)
+#define FUNC         gs_run
-#define TRI_ADJ(gs,i0,i1,i2,i3,i4,i5)  gs_tri_adj(gs,i0,i1,i2,i3,i4,i5)
+#define GET_ELT(idx) (idx)
-#define LINE(gs,i0,i1)        gs_line(gs,i0,i1)
-#define LINE_ADJ(gs,i0,i1,i2,i3)    gs_line_adj(gs,i0,i1,i2,i3)
-#define POINT(gs,i0)          gs_point(gs,i0)
-#define FUNC gs_run
-#define LOCAL_VARS
 #include "draw_gs_tmp.h"
 
 
-#define TRIANGLE(gs,i0,i1,i2) gs_tri(gs,elts[i0],elts[i1],elts[i2])
+#define FUNC         gs_run_elts
-#define TRI_ADJ(gs,i0,i1,i2,i3,i4,i5)           \
+#define LOCAL_VARS   const ushort *elts = input_prims->elts;
-   gs_tri_adj(gs,elts[i0],elts[i1],elts[i2],elts[i3], \
+#define GET_ELT(idx) (elts[idx] & ~DRAW_PIPE_FLAG_MASK)
-              elts[i4],elts[i5])
-#define LINE(gs,i0,i1)        gs_line(gs,elts[i0],elts[i1])
-#define LINE_ADJ(gs,i0,i1,i2,i3)  gs_line_adj(gs,elts[i0],      \
-                                              elts[i1],         \
-                                              elts[i2],elts[i3])
-#define POINT(gs,i0)          gs_point(gs,elts[i0])
-#define FUNC gs_run_elts
-#define LOCAL_VARS                         \
-   const ushort *elts = input_prims->elts;
 #include "draw_gs_tmp.h"
 
+
+/**
+ * Execute geometry shader using TGSI interpreter.
+ */
 int draw_geometry_shader_run(struct draw_geometry_shader *shader,
                              const void *constants[PIPE_MAX_CONSTANT_BUFFERS], 
+                             const unsigned constants_size[PIPE_MAX_CONSTANT_BUFFERS], 
                              const struct draw_vertex_info *input_verts,
                              const struct draw_prim_info *input_prim,
                              struct draw_vertex_info *output_verts,
@@ -405,7 +399,6 @@ int draw_geometry_shader_run(struct draw_geometry_shader *shader,
    unsigned input_stride = input_verts->vertex_size;
    unsigned vertex_size = input_verts->vertex_size;
    struct tgsi_exec_machine *machine = shader->machine;
-   unsigned int i;
    unsigned num_input_verts = input_prim->linear ?
                               input_verts->count :
                               input_prim->count;
@@ -447,9 +440,8 @@ int draw_geometry_shader_run(struct draw_geometry_shader *shader,
    }
    shader->primitive_lengths = MALLOC(max_out_prims * sizeof(unsigned));
 
-   for (i = 0; i < PIPE_MAX_CONSTANT_BUFFERS; i++) {
+   tgsi_exec_set_constant_buffers(machine, PIPE_MAX_CONSTANT_BUFFERS,
-      machine->Consts[i] = constants[i];
+                                  constants, constants_size);
-   }
 
    if (input_prim->linear)
       gs_run(shader, input_prim, input_verts,

src/gallium/auxiliary/draw/draw_gs.h

@@ -1,6 +1,6 @@
 /**************************************************************************
  * 
- * Copyright 2009 VMWare Inc.
+ * Copyright 2009 VMware, Inc.
  * All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
@@ -73,6 +73,7 @@ struct draw_geometry_shader {
  */
 int draw_geometry_shader_run(struct draw_geometry_shader *shader,
                              const void *constants[PIPE_MAX_CONSTANT_BUFFERS], 
+                             const unsigned constants_size[PIPE_MAX_CONSTANT_BUFFERS], 
                              const struct draw_vertex_info *input_verts,
                              const struct draw_prim_info *input_prim,
                              struct draw_vertex_info *output_verts,

src/gallium/auxiliary/draw/draw_gs_tmp.h

@@ -1,152 +1,34 @@
+#define FUNC_VARS struct draw_geometry_shader *gs,             \
+                  const struct draw_prim_info *input_prims,    \
+                  const struct draw_vertex_info *input_verts,  \
+                  struct draw_prim_info *output_prims,         \
+                  struct draw_vertex_info *output_verts
 
-static void FUNC( struct draw_geometry_shader *shader,
+#define FUNC_ENTER                                                \
-                  const struct draw_prim_info *input_prims,
+   /* declare more local vars */                                  \
-                  const struct draw_vertex_info *input_verts,
+   struct draw_context *draw = gs->draw;                          \
-                  struct draw_prim_info *output_prims,
+   const unsigned prim = input_prims->prim;                       \
-                  struct draw_vertex_info *output_verts)
+   const unsigned count = input_prims->count;                     \
-{
+   const boolean last_vertex_last =                               \
-   struct draw_context *draw = shader->draw;
+      !(draw->rasterizer->flatshade &&                            \
+        draw->rasterizer->flatshade_first);                       \
+   do {                                                           \
+      debug_assert(input_prims->primitive_count == 1);            \
+      switch (prim) {                                             \
+      case PIPE_PRIM_QUADS:                                       \
+      case PIPE_PRIM_QUAD_STRIP:                                  \
+      case PIPE_PRIM_POLYGON:                                     \
+         debug_assert(!"unexpected primitive type in GS");        \
+         return;                                                  \
+      default:                                                    \
+         break;                                                   \
+      }                                                           \
+   } while (0)                                                    \
 
-   boolean flatfirst = (draw->rasterizer->flatshade &&
+#define POINT(i0)                             gs_point(gs,i0)
-                        draw->rasterizer->flatshade_first);
+#define LINE(flags,i0,i1)                     gs_line(gs,i0,i1)
-   unsigned i, j;
+#define TRIANGLE(flags,i0,i1,i2)              gs_tri(gs,i0,i1,i2)
-   unsigned count = input_prims->count;
+#define LINE_ADJ(flags,i0,i1,i2,i3)           gs_line_adj(gs,i0,i1,i2,i3)
-   LOCAL_VARS
+#define TRIANGLE_ADJ(flags,i0,i1,i2,i3,i4,i5) gs_tri_adj(gs,i0,i1,i2,i3,i4,i5)
 
-   if (0) debug_printf("%s %d\n", __FUNCTION__, count);
+#include "draw_decompose_tmp.h"
-
-   debug_assert(input_prims->primitive_count == 1);
-
-   switch (input_prims->prim) {
-   case PIPE_PRIM_POINTS:
-      for (i = 0; i < count; i++) {
-	 POINT( shader, i + 0 );
-      }
-      break;
-
-   case PIPE_PRIM_LINES:
-      for (i = 0; i+1 < count; i += 2) {
-         LINE( shader , i + 0 , i + 1 );
-      }
-      break;
-
-   case PIPE_PRIM_LINE_LOOP:
-      if (count >= 2) {
-
-         for (i = 1; i < count; i++) {
-            LINE( shader, i - 1, i );
-         }
-
-	 LINE( shader, i - 1, 0 );
-      }
-      break;
-
-   case PIPE_PRIM_LINE_STRIP:
-      for (i = 1; i < count; i++) {
-         LINE( shader, i - 1, i );
-      }
-      break;
-
-   case PIPE_PRIM_TRIANGLES:
-      for (i = 0; i+2 < count; i += 3) {
-         TRIANGLE( shader, i + 0, i + 1, i + 2 );
-      }
-      break;
-
-   case PIPE_PRIM_TRIANGLE_STRIP:
-      if (flatfirst) {
-         for (i = 0; i+2 < count; i++) {
-            TRIANGLE( shader,
-                      i + 0,
-                      i + 1 + (i&1),
-                      i + 2 - (i&1) );
-         }
-      }
-      else {
-         for (i = 0; i+2 < count; i++) {
-            TRIANGLE( shader,
-                      i + 0 + (i&1),
-                      i + 1 - (i&1),
-                      i + 2 );
-         }
-      }
-      break;
-
-   case PIPE_PRIM_TRIANGLE_FAN:
-      if (count >= 3) {
-         if (flatfirst) {
-            for (i = 0; i+2 < count; i++) {
-               TRIANGLE( shader,
-                         i + 1,
-                         i + 2,
-                         0 );
-            }
-         }
-         else {
-            for (i = 0; i+2 < count; i++) {
-               TRIANGLE( shader,
-                         0,
-                         i + 1,
-                         i + 2 );
-            }
-         }
-      }
-      break;
-
-   case PIPE_PRIM_POLYGON:
-      {
-	 for (i = 0; i+2 < count; i++) {
-
-            if (flatfirst) {
-               TRIANGLE( shader, 0, i + 1, i + 2 );
-            }
-            else {
-               TRIANGLE( shader, i + 1, i + 2, 0 );
-            }
-	 }
-      }
-      break;
-
-   case PIPE_PRIM_LINES_ADJACENCY:
-      for (i = 0; i+3 < count; i += 4) {
-         LINE_ADJ( shader , i + 0 , i + 1, i + 2, i + 3 );
-      }
-      break;
-   case PIPE_PRIM_LINE_STRIP_ADJACENCY:
-      for (i = 1; i + 2 < count; i++) {
-         LINE_ADJ( shader, i - 1, i, i + 1, i + 2 );
-      }
-      break;
-
-   case PIPE_PRIM_TRIANGLES_ADJACENCY:
-      for (i = 0; i+5 < count; i += 5) {
-         TRI_ADJ( shader, i + 0, i + 1, i + 2,
-                  i + 3, i + 4, i + 5);
-      }
-      break;
-   case PIPE_PRIM_TRIANGLE_STRIP_ADJACENCY:
-      for (i = 0, j = 0; i+5 < count; i += 2, ++j) {
-         TRI_ADJ( shader,
-                  i + 0,
-                  i + 1 + 2*(j&1),
-                  i + 2 + 2*(j&1),
-                  i + 3 - 2*(j&1),
-                  i + 4 - 2*(j&1),
-                  i + 5);
-      }
-      break;
-
-   default:
-      debug_assert(!"Unsupported primitive in geometry shader");
-      break;
-   }
-}
-
-
-#undef TRIANGLE
-#undef TRI_ADJ
-#undef POINT
-#undef LINE
-#undef LINE_ADJ
-#undef FUNC
-#undef LOCAL_VARS

src/gallium/auxiliary/draw/draw_llvm.c

@@ -37,6 +37,8 @@
 #include "gallivm/lp_bld_debug.h"
 #include "gallivm/lp_bld_tgsi.h"
 #include "gallivm/lp_bld_printf.h"
+#include "gallivm/lp_bld_intr.h"
+#include "gallivm/lp_bld_init.h"
 
 #include "tgsi/tgsi_exec.h"
 #include "tgsi/tgsi_dump.h"
@@ -793,6 +795,11 @@ draw_llvm_generate(struct draw_llvm *llvm, struct draw_llvm_variant *variant)
 
    sampler->destroy(sampler);
 
+#ifdef PIPE_ARCH_X86
+   /* Avoid corrupting the FPU stack on 32bit OSes. */
+   lp_build_intrinsic(builder, "llvm.x86.mmx.emms", LLVMVoidType(), NULL, 0);
+#endif
+
    LLVMBuildRetVoid(builder);
 
    LLVMDisposeBuilder(builder);
@@ -820,6 +827,7 @@ draw_llvm_generate(struct draw_llvm *llvm, struct draw_llvm_variant *variant)
    if (gallivm_debug & GALLIVM_DEBUG_ASM) {
       lp_disassemble(code);
    }
+   lp_func_delete_body(variant->function);
 }
 
 
@@ -963,6 +971,11 @@ draw_llvm_generate_elts(struct draw_llvm *llvm, struct draw_llvm_variant *varian
 
    sampler->destroy(sampler);
 
+#ifdef PIPE_ARCH_X86
+   /* Avoid corrupting the FPU stack on 32bit OSes. */
+   lp_build_intrinsic(builder, "llvm.x86.mmx.emms", LLVMVoidType(), NULL, 0);
+#endif
+
    LLVMBuildRetVoid(builder);
 
    LLVMDisposeBuilder(builder);
@@ -990,6 +1003,7 @@ draw_llvm_generate_elts(struct draw_llvm *llvm, struct draw_llvm_variant *varian
    if (gallivm_debug & GALLIVM_DEBUG_ASM) {
       lp_disassemble(code);
    }
+   lp_func_delete_body(variant->function_elts);
 }
 
 void

src/gallium/auxiliary/draw/draw_pipe.c

@@ -169,77 +169,40 @@ static void do_triangle( struct draw_context *draw,
 /*
  * Set up macros for draw_pt_decompose.h template code.
  * This code uses vertex indexes / elements.
+ *
+ * Flags are needed by the stipple and unfilled stages.  When the two stages
+ * are active, vcache_run_extras is called and the flags are stored in the
+ * higher bits of i0.  Otherwise, flags do not matter.
  */
 
-/* emit first quad vertex as first vertex in triangles */
+#define TRIANGLE(flags,i0,i1,i2)                               \
-#define QUAD_FIRST_PV(i0,i1,i2,i3)              \
+   do_triangle( draw,                                          \
-   do_triangle( draw,                           \
+                i0,  /* flags */                               \
-                ( DRAW_PIPE_RESET_STIPPLE |     \
+                verts + stride * (i0 & ~DRAW_PIPE_FLAG_MASK),  \
-                  DRAW_PIPE_EDGE_FLAG_0 |       \
+                verts + stride * (i1),                         \
-                  DRAW_PIPE_EDGE_FLAG_1 ),      \
+                verts + stride * (i2) )
-                verts + stride * (elts[i0] & ~DRAW_PIPE_FLAG_MASK),     \
-                verts + stride * (elts[i1] & ~DRAW_PIPE_FLAG_MASK),     \
-                verts + stride * (elts[i2] & ~DRAW_PIPE_FLAG_MASK));    \
-   do_triangle( draw,                           \
-                ( DRAW_PIPE_EDGE_FLAG_1 |       \
-                  DRAW_PIPE_EDGE_FLAG_2 ),      \
-                verts + stride * (elts[i0] & ~DRAW_PIPE_FLAG_MASK),     \
-                verts + stride * (elts[i2] & ~DRAW_PIPE_FLAG_MASK),     \
-                verts + stride * (elts[i3] & ~DRAW_PIPE_FLAG_MASK))
 
-/* emit last quad vertex as last vertex in triangles */
+#define LINE(flags,i0,i1)                                      \
-#define QUAD_LAST_PV(i0,i1,i2,i3)               \
+   do_line( draw,                                              \
-   do_triangle( draw,                           \
+            i0, /* flags */                                    \
-                ( DRAW_PIPE_RESET_STIPPLE |     \
+            verts + stride * (i0 & ~DRAW_PIPE_FLAG_MASK),      \
-                  DRAW_PIPE_EDGE_FLAG_0 |       \
+            verts + stride * (i1) )
-                  DRAW_PIPE_EDGE_FLAG_2 ),      \
-                verts + stride * (elts[i0] & ~DRAW_PIPE_FLAG_MASK),     \
-                verts + stride * (elts[i1] & ~DRAW_PIPE_FLAG_MASK),     \
-                verts + stride * (elts[i3] & ~DRAW_PIPE_FLAG_MASK));    \
-   do_triangle( draw,                           \
-                ( DRAW_PIPE_EDGE_FLAG_0 |       \
-                  DRAW_PIPE_EDGE_FLAG_1 ),      \
-                verts + stride * (elts[i1] & ~DRAW_PIPE_FLAG_MASK),     \
-                verts + stride * (elts[i2] & ~DRAW_PIPE_FLAG_MASK),     \
-                verts + stride * (elts[i3] & ~DRAW_PIPE_FLAG_MASK))
-
-#define TRIANGLE(flags,i0,i1,i2)                                        \
-   do_triangle( draw,                                                   \
-                elts[i0],  /* flags */                                  \
-                verts + stride * (elts[i0] & ~DRAW_PIPE_FLAG_MASK),     \
-                verts + stride * (elts[i1] & ~DRAW_PIPE_FLAG_MASK),     \
-                verts + stride * (elts[i2] & ~DRAW_PIPE_FLAG_MASK) );
-
-#define LINE(flags,i0,i1)                                       \
-   do_line( draw,                                               \
-            elts[i0],                                           \
-            verts + stride * (elts[i0] & ~DRAW_PIPE_FLAG_MASK), \
-            verts + stride * (elts[i1] & ~DRAW_PIPE_FLAG_MASK) );
 
 #define POINT(i0)                               \
-   do_point( draw,                              \
+   do_point( draw, verts + stride * (i0) )
-             verts + stride * (elts[i0] & ~DRAW_PIPE_FLAG_MASK) )
 
-#define FUNC pipe_run
+#define GET_ELT(idx) (elts[idx])
-#define ARGS                                    \
+
+#define FUNC pipe_run_elts
+#define FUNC_VARS                               \
     struct draw_context *draw,                  \
     unsigned prim,                              \
     struct vertex_header *vertices,             \
     unsigned stride,                            \
-    const ushort *elts
+    const ushort *elts,                         \
-
+    unsigned count
-#define LOCAL_VARS                                           \
-   char *verts = (char *)vertices;                           \
-   boolean flatfirst = (draw->rasterizer->flatshade &&       \
-                        draw->rasterizer->flatshade_first);  \
-   unsigned i;                                               \
-   ushort flags
-
-#define FLUSH
 
 #include "draw_pt_decompose.h"
-#undef ARGS
-#undef LOCAL_VARS
 
 
 
@@ -269,14 +232,29 @@ void draw_pipeline_run( struct draw_context *draw,
         i < prim_info->primitive_count;
         start += prim_info->primitive_lengths[i], i++)
    {
-      unsigned count = prim_info->primitive_lengths[i];
+      const unsigned count = prim_info->primitive_lengths[i];
 
-      pipe_run(draw,
+#if DEBUG
-               prim_info->prim,
+      /* make sure none of the element indexes go outside the vertex buffer */
-               vert_info->verts,
+      {
-               vert_info->stride,
+         unsigned max_index = 0x0, i;
-               prim_info->elts + start,
+         /* find the largest element index */
-               count);
+         for (i = 0; i < count; i++) {
+            unsigned int index = (prim_info->elts[start + i]
+                                  & ~DRAW_PIPE_FLAG_MASK);
+            if (index > max_index)
+               max_index = index;
+         }
+         assert(max_index <= vert_info->count);
+      }
+#endif
+
+      pipe_run_elts(draw,
+                    prim_info->prim,
+                    vert_info->verts,
+                    vert_info->stride,
+                    prim_info->elts + start,
+                    count);
    }
 
    draw->pipeline.verts = NULL;
@@ -289,70 +267,30 @@ void draw_pipeline_run( struct draw_context *draw,
  * This code is for non-indexed (aka linear) rendering (no elts).
  */
 
-/* emit first quad vertex as first vertex in triangles */
+#define TRIANGLE(flags,i0,i1,i2)       \
-#define QUAD_FIRST_PV(i0,i1,i2,i3)                               \
+   do_triangle( draw, flags,           \
-   do_triangle( draw,                                            \
+                verts + stride * (i0), \
-                ( DRAW_PIPE_RESET_STIPPLE |                      \
+                verts + stride * (i1), \
-                  DRAW_PIPE_EDGE_FLAG_0 |                        \
+                verts + stride * (i2) )
-                  DRAW_PIPE_EDGE_FLAG_1 ),                       \
-                verts + stride * ((i0) & ~DRAW_PIPE_FLAG_MASK),  \
-                verts + stride * ((i1) & ~DRAW_PIPE_FLAG_MASK),  \
-                verts + stride * ((i2) & ~DRAW_PIPE_FLAG_MASK)); \
-   do_triangle( draw,                                            \
-                ( DRAW_PIPE_EDGE_FLAG_1 |                        \
-                  DRAW_PIPE_EDGE_FLAG_2 ),                       \
-                verts + stride * ((i0) & ~DRAW_PIPE_FLAG_MASK),  \
-                verts + stride * ((i2) & ~DRAW_PIPE_FLAG_MASK),  \
-                verts + stride * ((i3) & ~DRAW_PIPE_FLAG_MASK))
 
-/* emit last quad vertex as last vertex in triangles */
+#define LINE(flags,i0,i1)              \
-#define QUAD_LAST_PV(i0,i1,i2,i3)                                \
+   do_line( draw, flags,               \
-   do_triangle( draw,                                            \
+            verts + stride * (i0),     \
-                ( DRAW_PIPE_RESET_STIPPLE |                      \
+            verts + stride * (i1) )
-                  DRAW_PIPE_EDGE_FLAG_0 |                        \
-                  DRAW_PIPE_EDGE_FLAG_2 ),                       \
-                verts + stride * ((i0) & ~DRAW_PIPE_FLAG_MASK),  \
-                verts + stride * ((i1) & ~DRAW_PIPE_FLAG_MASK),  \
-                verts + stride * ((i3) & ~DRAW_PIPE_FLAG_MASK)); \
-   do_triangle( draw,                                            \
-                ( DRAW_PIPE_EDGE_FLAG_0 |                        \
-                  DRAW_PIPE_EDGE_FLAG_1 ),                       \
-                verts + stride * ((i1) & ~DRAW_PIPE_FLAG_MASK),  \
-                verts + stride * ((i2) & ~DRAW_PIPE_FLAG_MASK),  \
-                verts + stride * ((i3) & ~DRAW_PIPE_FLAG_MASK))
 
-#define TRIANGLE(flags,i0,i1,i2)                                 \
+#define POINT(i0)                      \
-   do_triangle( draw,                                            \
+   do_point( draw, verts + stride * (i0) )
-                flags,  /* flags */                              \
-                verts + stride * ((i0) & ~DRAW_PIPE_FLAG_MASK),  \
-                verts + stride * ((i1) & ~DRAW_PIPE_FLAG_MASK),  \
-                verts + stride * ((i2) & ~DRAW_PIPE_FLAG_MASK))
 
-#define LINE(flags,i0,i1)                                   \
-   do_line( draw,                                           \
-            flags,                                          \
-            verts + stride * ((i0) & ~DRAW_PIPE_FLAG_MASK), \
-            verts + stride * ((i1) & ~DRAW_PIPE_FLAG_MASK))
 
-#define POINT(i0)                               \
+#define GET_ELT(idx) (idx)
-   do_point( draw,                              \
-             verts + stride * ((i0) & ~DRAW_PIPE_FLAG_MASK) )
 
 #define FUNC pipe_run_linear
-#define ARGS                                    \
+#define FUNC_VARS                      \
-    struct draw_context *draw,                  \
+    struct draw_context *draw,         \
-    unsigned prim,                              \
+    unsigned prim,                     \
-    struct vertex_header *vertices,             \
+    struct vertex_header *vertices,    \
-    unsigned stride
+    unsigned stride,                   \
-
+    unsigned count
-#define LOCAL_VARS                                           \
-   char *verts = (char *)vertices;                           \
-   boolean flatfirst = (draw->rasterizer->flatshade &&       \
-                        draw->rasterizer->flatshade_first);  \
-   unsigned i;                                               \
-   ushort flags
-
-#define FLUSH
 
 #include "draw_pt_decompose.h"
 
@@ -378,6 +316,8 @@ void draw_pipeline_run_linear( struct draw_context *draw,
       draw->pipeline.vertex_stride = vert_info->stride;
       draw->pipeline.vertex_count = count;
 
+      assert(count <= vert_info->count);
+
       pipe_run_linear(draw,
                       prim_info->prim,
                       (struct vertex_header*)verts,

src/gallium/auxiliary/draw/draw_pipe_clip.c

@@ -68,8 +68,7 @@ struct clip_stage {
 };
 
 
-/* This is a bit confusing:
+/** Cast wrapper */
- */
 static INLINE struct clip_stage *clip_stage( struct draw_stage *stage )
 {
    return (struct clip_stage *)stage;
@@ -81,18 +80,22 @@ static INLINE struct clip_stage *clip_stage( struct draw_stage *stage )
 
 /* All attributes are float[4], so this is easy:
  */
-static void interp_attr( float *fdst,
+static void interp_attr( float dst[4],
 			 float t,
-			 const float *fin,
+			 const float in[4],
-			 const float *fout )
+			 const float out[4] )
 {  
-   fdst[0] = LINTERP( t, fout[0], fin[0] );
+   dst[0] = LINTERP( t, out[0], in[0] );
-   fdst[1] = LINTERP( t, fout[1], fin[1] );
+   dst[1] = LINTERP( t, out[1], in[1] );
-   fdst[2] = LINTERP( t, fout[2], fin[2] );
+   dst[2] = LINTERP( t, out[2], in[2] );
-   fdst[3] = LINTERP( t, fout[3], fin[3] );
+   dst[3] = LINTERP( t, out[3], in[3] );
 }
 
 
+/**
+ * Copy front/back, primary/secondary colors from src vertex to dst vertex.
+ * Used when flat shading.
+ */
 static void copy_colors( struct draw_stage *stage,
 			 struct vertex_header *dst,
 			 const struct vertex_header *src )
@@ -121,20 +124,17 @@ static void interp( const struct clip_stage *clip,
 
    /* Vertex header.
     */
-   {
+   dst->clipmask = 0;
-      dst->clipmask = 0;
+   dst->edgeflag = 0;        /* will get overwritten later */
-      dst->edgeflag = 0;        /* will get overwritten later */
+   dst->pad = 0;
-      dst->pad = 0;
+   dst->vertex_id = UNDEFINED_VERTEX_ID;
-      dst->vertex_id = UNDEFINED_VERTEX_ID;
-   }
 
-   /* Clip coordinates:  interpolate normally
+   /* Interpolate the clip-space coords.
     */
-   {
+   interp_attr(dst->clip, t, in->clip, out->clip);
-      interp_attr(dst->clip, t, in->clip, out->clip);
-   }
 
-   /* Do the projective divide and insert window coordinates:
+   /* Do the projective divide and viewport transformation to get
+    * new window coordinates:
     */
    {
       const float *pos = dst->clip;

src/gallium/auxiliary/draw/draw_private.h

@@ -163,9 +163,11 @@ struct draw_context
          /** vertex arrays */
          const void *vbuffer[PIPE_MAX_ATTRIBS];
          
-         /** constant buffer (for vertex/geometry shader) */
+         /** constant buffers (for vertex/geometry shader) */
          const void *vs_constants[PIPE_MAX_CONSTANT_BUFFERS];
+         unsigned vs_constants_size[PIPE_MAX_CONSTANT_BUFFERS];
          const void *gs_constants[PIPE_MAX_CONSTANT_BUFFERS];
+         unsigned gs_constants_size[PIPE_MAX_CONSTANT_BUFFERS];
       } user;
 
       boolean test_fse;         /* enable FSE even though its not correct (eg for softpipe) */
@@ -198,6 +200,7 @@ struct draw_context
    struct pipe_viewport_state viewport;
    boolean identity_viewport;
 
+   /** Vertex shader state */
    struct {
       struct draw_vertex_shader *vertex_shader;
       uint num_vs_outputs;  /**< convenience, from vertex_shader */
@@ -227,6 +230,7 @@ struct draw_context
       struct translate_cache *emit_cache;
    } vs;
 
+   /** Geometry shader state */
    struct {
       struct draw_geometry_shader *geometry_shader;
       uint num_gs_outputs;  /**< convenience, from geometry_shader */
@@ -239,6 +243,7 @@ struct draw_context
       struct tgsi_sampler **samplers;
    } gs;
 
+   /** Stream output (vertex feedback) state */
    struct {
       struct pipe_stream_output_state state;
       void *buffers[PIPE_MAX_SO_BUFFERS];

src/gallium/auxiliary/draw/draw_pt.c

@@ -259,6 +259,12 @@ draw_print_arrays(struct draw_context *draw, uint prim, int start, uint count)
       for (j = 0; j < draw->pt.nr_vertex_elements; j++) {
          uint buf = draw->pt.vertex_element[j].vertex_buffer_index;
          ubyte *ptr = (ubyte *) draw->pt.user.vbuffer[buf];
+
+         if (draw->pt.vertex_element[j].instance_divisor) {
+            ii = draw->instance_id / draw->pt.vertex_element[j].instance_divisor;
+         }
+
+         ptr += draw->pt.vertex_buffer[buf].buffer_offset;
          ptr += draw->pt.vertex_buffer[buf].stride * ii;
          ptr += draw->pt.vertex_element[j].src_offset;
 
@@ -341,19 +347,22 @@ draw_arrays_instanced(struct draw_context *draw,
    unsigned reduced_prim = u_reduced_prim(mode);
    unsigned instance;
 
+   assert(instanceCount > 0);
+
    if (reduced_prim != draw->reduced_prim) {
       draw_do_flush(draw, DRAW_FLUSH_STATE_CHANGE);
       draw->reduced_prim = reduced_prim;
    }
 
    if (0)
-      draw_print_arrays(draw, mode, start, MIN2(count, 20));
+      debug_printf("draw_arrays(mode=%u start=%u count=%u):\n",
+                   mode, start, count);
+
+   if (0)
+      tgsi_dump(draw->vs.vertex_shader->state.tokens, 0);
 
    if (0) {
       unsigned int i;
-      debug_printf("draw_arrays(mode=%u start=%u count=%u):\n",
-                   mode, start, count);
-      tgsi_dump(draw->vs.vertex_shader->state.tokens, 0);
       debug_printf("Elements:\n");
       for (i = 0; i < draw->pt.nr_vertex_elements; i++) {
          debug_printf("  %u: src_offset=%u  inst_div=%u   vbuf=%u  format=%s\n",
@@ -374,6 +383,9 @@ draw_arrays_instanced(struct draw_context *draw,
       }
    }
 
+   if (0)
+      draw_print_arrays(draw, mode, start, MIN2(count, 20));
+
    for (instance = 0; instance < instanceCount; instance++) {
       draw->instance_id = instance + startInstance;
       draw_pt_arrays(draw, mode, start, count);

src/gallium/auxiliary/draw/draw_pt_decompose.h

@@ -1,194 +1,7 @@
+#define LOCAL_VARS                           \
+   char *verts = (char *) vertices;          \
+   const boolean last_vertex_last =          \
+      !(draw->rasterizer->flatshade &&       \
+        draw->rasterizer->flatshade_first);
 
-
+#include "draw_decompose_tmp.h"
-static void FUNC( ARGS,
-                  unsigned count )
-{
-   LOCAL_VARS;
-
-   switch (prim) {
-   case PIPE_PRIM_POINTS:
-      for (i = 0; i < count; i ++) {
-	 POINT( (i + 0) );
-      }
-      break;
-
-   case PIPE_PRIM_LINES:
-      for (i = 0; i+1 < count; i += 2) {
-         LINE( DRAW_PIPE_RESET_STIPPLE,
-               (i + 0),
-               (i + 1));
-      }
-      break;
-
-   case PIPE_PRIM_LINE_LOOP:
-      if (count >= 2) {
-         flags = DRAW_PIPE_RESET_STIPPLE;
-
-         for (i = 1; i < count; i++, flags = 0) {
-            LINE( flags,
-                  (i - 1),
-                  (i ));
-         }
-
-	 LINE( flags,
-               (i - 1),
-               (0 ));
-      }
-      break;
-
-   case PIPE_PRIM_LINE_STRIP:
-      flags = DRAW_PIPE_RESET_STIPPLE;
-      for (i = 1; i < count; i++, flags = 0) {
-         LINE( flags,
-               (i - 1),
-               (i ));
-      }
-      break;
-
-   case PIPE_PRIM_TRIANGLES:
-      for (i = 0; i+2 < count; i += 3) {
-         TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL,
-                   (i + 0),
-                   (i + 1),
-                   (i + 2 ));
-      }
-      break;
-
-   case PIPE_PRIM_TRIANGLE_STRIP:
-      if (flatfirst) {
-         for (i = 0; i+2 < count; i++) {
-            /* Emit first triangle vertex as first triangle vertex */
-            TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL,
-                      (i + 0),
-                      (i + 1 + (i&1)),
-                      (i + 2 - (i&1)) );
-         }
-      }
-      else {
-         for (i = 0; i+2 < count; i++) {
-            /* Emit last triangle vertex as last triangle vertex */
-            TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL,
-                      (i + 0 + (i&1)),
-                      (i + 1 - (i&1)),
-                      (i + 2 ));
-         }
-      }
-      break;
-
-   case PIPE_PRIM_TRIANGLE_FAN:
-      if (count >= 3) {
-         if (flatfirst) {
-            for (i = 0; i+2 < count; i++) {
-               TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL,
-                         (i + 1),
-                         (i + 2),
-                         0 );
-            }
-         }
-         else {
-            for (i = 0; i+2 < count; i++) {
-               TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL,
-                         (0),
-                         (i + 1),
-                         (i + 2 ));
-            }
-         }
-      }
-      break;
-
-
-   case PIPE_PRIM_QUADS:
-      /* GL quads don't follow provoking vertex convention */
-      if (flatfirst) {
-         for (i = 0; i+3 < count; i += 4) {
-            /* emit last quad vertex as first triangle vertex */
-            QUAD_FIRST_PV( (i + 3),
-                           (i + 0),
-                           (i + 1),
-                           (i + 2) );
-         }
-      }
-      else {
-         for (i = 0; i+3 < count; i += 4) {
-            /* emit last quad vertex as last triangle vertex */
-            QUAD_LAST_PV( (i + 0),
-                          (i + 1),
-                          (i + 2),
-                          (i + 3) );
-         }
-      }
-      break;
-
-   case PIPE_PRIM_QUAD_STRIP:
-      /* GL quad strips don't follow provoking vertex convention */
-      if (flatfirst) {
-         for (i = 0; i+3 < count; i += 2) {
-            /* emit last quad vertex as first triangle vertex */
-            QUAD_FIRST_PV( (i + 3),
-                           (i + 2),
-                           (i + 0),
-                           (i + 1) );
-
-         }
-      }
-      else {
-         for (i = 0; i+3 < count; i += 2) {
-            /* emit last quad vertex as last triangle vertex */
-            QUAD_LAST_PV( (i + 2),
-                          (i + 0),
-                          (i + 1),
-                          (i + 3) );
-         }
-      }
-      break;
-
-   case PIPE_PRIM_POLYGON:
-      /* GL polygons don't follow provoking vertex convention */
-      {
-         /* These bitflags look a little odd because we submit the
-          * vertices as (1,2,0) to satisfy flatshade requirements.
-          */
-         const ushort edge_first  = DRAW_PIPE_EDGE_FLAG_2;
-         const ushort edge_middle = DRAW_PIPE_EDGE_FLAG_0;
-         const ushort edge_last   = DRAW_PIPE_EDGE_FLAG_1;
-
-         flags = DRAW_PIPE_RESET_STIPPLE | edge_first | edge_middle;
-
-	 for (i = 0; i+2 < count; i++, flags = edge_middle) {
-
-            if (i + 3 == count)
-               flags |= edge_last;
-
-            if (flatfirst) {
-               /* emit first polygon vertex as first triangle vertex */
-               TRIANGLE( flags,
-                         (0),
-                         (i + 1),
-                         (i + 2) );
-            }
-            else {
-               /* emit first polygon vertex as last triangle vertex */
-               TRIANGLE( flags,
-                         (i + 1),
-                         (i + 2),
-                         (0));
-            }
-	 }
-      }
-      break;
-
-   default:
-      assert(0);
-      break;
-   }
-
-   FLUSH;
-}
-
-
-#undef TRIANGLE
-#undef QUAD_FIRST_PV
-#undef QUAD_LAST_PV
-#undef POINT
-#undef LINE
-#undef FUNC

src/gallium/auxiliary/draw/draw_pt_emit.c

@@ -182,6 +182,7 @@ void draw_pt_emit( struct pt_emit *emit,
 			 0,
 			 ~0);
 
+   /* fetch/translate vertex attribs to fill hw_verts[] */
    translate->run( translate,
 		   0, 
 		   vertex_count,

src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c

@@ -176,6 +176,7 @@ static void emit(struct pt_emit *emit,
 
 static void draw_vertex_shader_run(struct draw_vertex_shader *vshader,
                                    const void *constants[PIPE_MAX_CONSTANT_BUFFERS], 
+                                   unsigned const_size[PIPE_MAX_CONSTANT_BUFFERS],
                                    const struct draw_vertex_info *input_verts,
                                    struct draw_vertex_info *output_verts )
 {
@@ -190,6 +191,7 @@ static void draw_vertex_shader_run(struct draw_vertex_shader *vshader,
                        (const float (*)[4])input_verts->verts->data,
                        (      float (*)[4])output_verts->verts->data,
                        constants,
+                       const_size,
                        input_verts->count,
                        input_verts->vertex_size,
                        input_verts->vertex_size);
@@ -236,6 +238,7 @@ static void fetch_pipeline_generic( struct draw_pt_middle_end *middle,
    if (fpme->opt & PT_SHADE) {
       draw_vertex_shader_run(vshader,
                              draw->pt.user.vs_constants,
+                             draw->pt.user.vs_constants_size,
                              vert_info,
                              &vs_vert_info);
 
@@ -246,6 +249,7 @@ static void fetch_pipeline_generic( struct draw_pt_middle_end *middle,
    if ((fpme->opt & PT_SHADE) && gshader) {
       draw_geometry_shader_run(gshader,
                                draw->pt.user.gs_constants,
+                               draw->pt.user.gs_constants_size,
                                vert_info,
                                prim_info,
                                &gs_vert_info,

src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline_llvm.c

@@ -1,6 +1,6 @@
 /**************************************************************************
  *
- * Copyright 2010 VMWare, Inc.
+ * Copyright 2010 VMware, Inc.
  * All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
@@ -254,6 +254,7 @@ llvm_pipeline_generic( struct draw_pt_middle_end *middle,
    if ((opt & PT_SHADE) && gshader) {
       draw_geometry_shader_run(gshader,
                                draw->pt.user.gs_constants,
+                               draw->pt.user.gs_constants_size,
                                vert_info,
                                prim_info,
                                &gs_vert_info,

src/gallium/auxiliary/draw/draw_pt_so_emit.c

@@ -218,25 +218,15 @@ static void so_tri(struct pt_so_emit *so, int i0, int i1, int i2)
 }
 
 
-#define TRIANGLE(gs,i0,i1,i2) so_tri(so,i0,i1,i2)
+#define FUNC         so_run_linear
-#define LINE(gs,i0,i1)        so_line(so,i0,i1)
+#define GET_ELT(idx) (start + (idx))
-#define POINT(gs,i0)          so_point(so,i0)
-#define FUNC so_run_linear
-#define LOCAL_VARS
 #include "draw_so_emit_tmp.h"
-#undef LOCAL_VARS
-#undef FUNC
 
 
-#define TRIANGLE(gs,i0,i1,i2) so_tri(gs,elts[i0],elts[i1],elts[i2])
+#define FUNC         so_run_elts
-#define LINE(gs,i0,i1)        so_line(gs,elts[i0],elts[i1])
+#define LOCAL_VARS   const ushort *elts = input_prims->elts;
-#define POINT(gs,i0)          so_point(gs,elts[i0])
+#define GET_ELT(idx) (elts[start + (idx)] & ~DRAW_PIPE_FLAG_MASK)
-#define FUNC so_run_elts
-#define LOCAL_VARS                         \
-   const ushort *elts = input_prims->elts;
 #include "draw_so_emit_tmp.h"
-#undef LOCAL_VARS
-#undef FUNC
 
 
 void draw_pt_so_emit( struct pt_so_emit *emit,

src/gallium/auxiliary/draw/draw_pt_vcache.c

@@ -95,7 +95,7 @@ static INLINE void
 vcache_check_flush( struct vcache_frontend *vcache )
 {
    if (vcache->draw_count + 6 >= DRAW_MAX ||
-       vcache->fetch_count + 4 >= FETCH_MAX) {
+       vcache->fetch_count + 6 >= FETCH_MAX) {
       vcache_flush( vcache );
    }
 }
@@ -180,59 +180,61 @@ vcache_point( struct vcache_frontend *vcache,
 }
 
 
-static INLINE void 
+static INLINE void
-vcache_quad( struct vcache_frontend *vcache,
+vcache_line_adj_flags( struct vcache_frontend *vcache,
-             unsigned i0,
+                       unsigned flags,
-             unsigned i1,
+                       unsigned a0, unsigned i0, unsigned i1, unsigned a1 )
-             unsigned i2,
-             unsigned i3 )
 {
-   if (vcache->draw->rasterizer->flatshade_first) {
+   vcache_elt(vcache, a0, 0);
-      /* pass last quad vertex as first triangle vertex */
+   vcache_elt(vcache, i0, flags);
-      vcache_triangle( vcache, i3, i0, i1 );
+   vcache_elt(vcache, i1, 0);
-      vcache_triangle( vcache, i3, i1, i2 );
+   vcache_elt(vcache, a1, 0);
-   }
+   vcache_check_flush(vcache);
-   else {
-      /* pass last quad vertex as last triangle vertex */
-      vcache_triangle( vcache, i0, i1, i3 );
-      vcache_triangle( vcache, i1, i2, i3 );
-   }
 }
 
 
-static INLINE void 
+static INLINE void
-vcache_ef_quad( struct vcache_frontend *vcache,
+vcache_line_adj( struct vcache_frontend *vcache,
-                unsigned i0,
+                 unsigned a0, unsigned i0, unsigned i1, unsigned a1 )
-                unsigned i1,
-                unsigned i2,
-                unsigned i3 )
 {
-   if (vcache->draw->rasterizer->flatshade_first) {
+   vcache_elt(vcache, a0, 0);
-      /* pass last quad vertex as first triangle vertex */
+   vcache_elt(vcache, i0, 0);
-      vcache_triangle_flags( vcache,
+   vcache_elt(vcache, i1, 0);
-                             ( DRAW_PIPE_RESET_STIPPLE |
+   vcache_elt(vcache, a1, 0);
-                               DRAW_PIPE_EDGE_FLAG_0 |
+   vcache_check_flush(vcache);
-                               DRAW_PIPE_EDGE_FLAG_1 ),
+}
-                             i3, i0, i1 );
 
-      vcache_triangle_flags( vcache,
-                             ( DRAW_PIPE_EDGE_FLAG_1 |
-                               DRAW_PIPE_EDGE_FLAG_2 ),
-                             i3, i1, i2 );
-   }
-   else {
-      /* pass last quad vertex as last triangle vertex */
-      vcache_triangle_flags( vcache,
-                             ( DRAW_PIPE_RESET_STIPPLE |
-                               DRAW_PIPE_EDGE_FLAG_0 |
-                               DRAW_PIPE_EDGE_FLAG_2 ),
-                             i0, i1, i3 );
 
-      vcache_triangle_flags( vcache,
+static INLINE void
-                             ( DRAW_PIPE_EDGE_FLAG_0 |
+vcache_triangle_adj_flags( struct vcache_frontend *vcache,
-                               DRAW_PIPE_EDGE_FLAG_1 ),
+                           unsigned flags,
-                             i1, i2, i3 );
+                           unsigned i0, unsigned a0,
-   }
+                           unsigned i1, unsigned a1,
+                           unsigned i2, unsigned a2 )
+{
+   vcache_elt(vcache, i0, flags);
+   vcache_elt(vcache, a0, 0);
+   vcache_elt(vcache, i1, 0);
+   vcache_elt(vcache, a1, 0);
+   vcache_elt(vcache, i2, 0);
+   vcache_elt(vcache, a2, 0);
+   vcache_check_flush(vcache);
+}
+
+
+static INLINE void
+vcache_triangle_adj( struct vcache_frontend *vcache,
+                     unsigned i0, unsigned a0,
+                     unsigned i1, unsigned a1,
+                     unsigned i2, unsigned a2 )
+{
+   vcache_elt(vcache, i0, 0);
+   vcache_elt(vcache, a0, 0);
+   vcache_elt(vcache, i1, 0);
+   vcache_elt(vcache, a1, 0);
+   vcache_elt(vcache, i2, 0);
+   vcache_elt(vcache, a2, 0);
+   vcache_check_flush(vcache);
 }
 
 
@@ -240,17 +242,23 @@ vcache_ef_quad( struct vcache_frontend *vcache,
  * this.  The two paths aren't too different though - it may be
  * possible to reunify them.
  */
-#define TRIANGLE(vc,flags,i0,i1,i2) vcache_triangle_flags(vc,flags,i0,i1,i2)
+#define TRIANGLE(flags,i0,i1,i2) vcache_triangle_flags(vcache,flags,i0,i1,i2)
-#define QUAD(vc,i0,i1,i2,i3)        vcache_ef_quad(vc,i0,i1,i2,i3)
+#define LINE(flags,i0,i1)        vcache_line_flags(vcache,flags,i0,i1)
-#define LINE(vc,flags,i0,i1)        vcache_line_flags(vc,flags,i0,i1)
+#define POINT(i0)                vcache_point(vcache,i0)
-#define POINT(vc,i0)                vcache_point(vc,i0)
+#define LINE_ADJ(flags,a0,i0,i1,a1) \
+   vcache_line_adj_flags(vcache,flags,a0,i0,i1,a1)
+#define TRIANGLE_ADJ(flags,i0,a0,i1,a1,i2,a2) \
+   vcache_triangle_adj_flags(vcache,flags,i0,a0,i1,a1,i2,a2)
 #define FUNC vcache_run_extras
 #include "draw_pt_vcache_tmp.h"
 
-#define TRIANGLE(vc,flags,i0,i1,i2) vcache_triangle(vc,i0,i1,i2)
+#define TRIANGLE(flags,i0,i1,i2) vcache_triangle(vcache,i0,i1,i2)
-#define QUAD(vc,i0,i1,i2,i3)        vcache_quad(vc,i0,i1,i2,i3)
+#define LINE(flags,i0,i1)        vcache_line(vcache,i0,i1)
-#define LINE(vc,flags,i0,i1)        vcache_line(vc,i0,i1)
+#define POINT(i0)                vcache_point(vcache,i0)
-#define POINT(vc,i0)                vcache_point(vc,i0)
+#define LINE_ADJ(flags,a0,i0,i1,a1) \
+   vcache_line_adj(vcache,a0,i0,i1,a1)
+#define TRIANGLE_ADJ(flags,i0,a0,i1,a1,i2,a2) \
+   vcache_triangle_adj(vcache,i0,a0,i1,a1,i2,a2)
 #define FUNC vcache_run
 #include "draw_pt_vcache_tmp.h"
 
@@ -339,6 +347,25 @@ format_from_get_elt( pt_elt_func get_elt )
 #endif
 
 
+/**
+ * Check if any vertex attributes use instance divisors.
+ * Note that instance divisors complicate vertex fetching so we need
+ * to take the vcache path when they're in use.
+ */
+static boolean
+any_instance_divisors(const struct draw_context *draw)
+{
+   uint i;
+
+   for (i = 0; i < draw->pt.nr_vertex_elements; i++) {
+      uint div = draw->pt.vertex_element[i].instance_divisor;
+      if (div)
+         return TRUE;
+   }
+   return FALSE;
+}
+
+
 static INLINE void 
 vcache_check_run( struct draw_pt_front_end *frontend, 
                   pt_elt_func get_elt,
@@ -382,6 +409,9 @@ vcache_check_run( struct draw_pt_front_end *frontend,
    if (max_index >= (unsigned) DRAW_PIPE_MAX_VERTICES)
       goto fail;
 
+   if (any_instance_divisors(draw))
+      goto fail;
+
    fetch_count = max_index + 1 - min_index;
 
    if (0)
@@ -518,7 +548,18 @@ vcache_prepare( struct draw_pt_front_end *frontend,
     * which is a separate issue.
     */
    vcache->input_prim = in_prim;
-   vcache->output_prim = u_reduced_prim(in_prim);
+   switch (in_prim) {
+   case PIPE_PRIM_LINES_ADJACENCY:
+   case PIPE_PRIM_LINE_STRIP_ADJACENCY:
+      vcache->output_prim = PIPE_PRIM_LINES_ADJACENCY;
+      break;
+   case PIPE_PRIM_TRIANGLES_ADJACENCY:
+   case PIPE_PRIM_TRIANGLE_STRIP_ADJACENCY:
+      vcache->output_prim = PIPE_PRIM_TRIANGLES_ADJACENCY;
+      break;
+   default:
+      vcache->output_prim = u_reduced_prim(in_prim);
+   }
 
    vcache->middle = middle;
    vcache->opt = opt;

src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h

@@ -1,198 +1,19 @@
+#define FUNC_VARS                      \
+   struct draw_pt_front_end *frontend, \
+   pt_elt_func get_elt,                \
+   const void *elts,                   \
+   int elt_bias,                       \
+   unsigned count
 
+#define LOCAL_VARS \
+   struct vcache_frontend *vcache = (struct vcache_frontend *) frontend;   \
+   struct draw_context *draw = vcache->draw;                               \
+   const unsigned prim = vcache->input_prim;                               \
+   const boolean last_vertex_last = !(draw->rasterizer->flatshade &&       \
+                                      draw->rasterizer->flatshade_first);
 
-static void FUNC( struct draw_pt_front_end *frontend, 
+#define GET_ELT(idx) (get_elt(elts, idx) + elt_bias)
-                  pt_elt_func get_elt,
-                  const void *elts,
-                  int elt_bias,
-                  unsigned count )
-{
-   struct vcache_frontend *vcache = (struct vcache_frontend *)frontend;
-   struct draw_context *draw = vcache->draw;
 
-   boolean flatfirst = (draw->rasterizer->flatshade && 
+#define FUNC_EXIT do { vcache_flush(vcache); } while (0)
-                        draw->rasterizer->flatshade_first);
-   unsigned i;
-   ushort flags;
 
-   if (0) debug_printf("%s %d\n", __FUNCTION__, count);
+#include "draw_decompose_tmp.h"
-
-
-   switch (vcache->input_prim) {
-   case PIPE_PRIM_POINTS:
-      for (i = 0; i < count; i ++) {
-	 POINT( vcache,
-                get_elt(elts, i + 0) + elt_bias );
-      }
-      break;
-
-   case PIPE_PRIM_LINES:
-      for (i = 0; i+1 < count; i += 2) {
-         LINE( vcache, 
-               DRAW_PIPE_RESET_STIPPLE,
-               get_elt(elts, i + 0) + elt_bias,
-               get_elt(elts, i + 1) + elt_bias);
-      }
-      break;
-
-   case PIPE_PRIM_LINE_LOOP:  
-      if (count >= 2) {
-         flags = DRAW_PIPE_RESET_STIPPLE;
-
-         for (i = 1; i < count; i++, flags = 0) {
-            LINE( vcache, 
-                  flags,
-                  get_elt(elts, i - 1) + elt_bias,
-                  get_elt(elts, i ) + elt_bias);
-         }
-
-	 LINE( vcache, 
-               flags,
-               get_elt(elts, i - 1) + elt_bias,
-               get_elt(elts, 0 ) + elt_bias);
-      }
-      break;
-
-   case PIPE_PRIM_LINE_STRIP:
-      flags = DRAW_PIPE_RESET_STIPPLE;
-      for (i = 1; i < count; i++, flags = 0) {
-         LINE( vcache, 
-               flags,
-               get_elt(elts, i - 1) + elt_bias,
-               get_elt(elts, i ) + elt_bias);
-      }
-      break;
-
-   case PIPE_PRIM_TRIANGLES:
-      for (i = 0; i+2 < count; i += 3) {
-         TRIANGLE( vcache,
-                   DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, 
-                   get_elt(elts, i + 0) + elt_bias,
-                   get_elt(elts, i + 1) + elt_bias,
-                   get_elt(elts, i + 2 ) + elt_bias);
-      }
-      break;
-
-   case PIPE_PRIM_TRIANGLE_STRIP:
-      if (flatfirst) {
-         for (i = 0; i+2 < count; i++) {
-            TRIANGLE( vcache,
-                      DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, 
-                      get_elt(elts, i + 0) + elt_bias,
-                      get_elt(elts, i + 1 + (i&1)) + elt_bias,
-                      get_elt(elts, i + 2 - (i&1)) + elt_bias);
-         }
-      }
-      else {
-         for (i = 0; i+2 < count; i++) {
-            TRIANGLE( vcache,
-                      DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, 
-                      get_elt(elts, i + 0 + (i&1)) + elt_bias,
-                      get_elt(elts, i + 1 - (i&1)) + elt_bias,
-                      get_elt(elts, i + 2 ) + elt_bias);
-         }
-      }
-      break;
-
-   case PIPE_PRIM_TRIANGLE_FAN:
-      if (count >= 3) {
-         if (flatfirst) {
-            for (i = 0; i+2 < count; i++) {
-               TRIANGLE( vcache,
-                         DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, 
-                         get_elt(elts, i + 1) + elt_bias,
-                         get_elt(elts, i + 2) + elt_bias,
-                         get_elt(elts, 0 ) + elt_bias);
-            }
-         }
-         else {
-            for (i = 0; i+2 < count; i++) {
-               TRIANGLE( vcache,
-                         DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, 
-                         get_elt(elts, 0) + elt_bias,
-                         get_elt(elts, i + 1) + elt_bias,
-                         get_elt(elts, i + 2 ) + elt_bias);
-            }
-         }
-      }
-      break;
-
-
-   case PIPE_PRIM_QUADS:
-      for (i = 0; i+3 < count; i += 4) {
-         QUAD( vcache,
-               get_elt(elts, i + 0) + elt_bias,
-               get_elt(elts, i + 1) + elt_bias,
-               get_elt(elts, i + 2) + elt_bias,
-               get_elt(elts, i + 3) + elt_bias );
-      }
-      break;
-
-   case PIPE_PRIM_QUAD_STRIP:
-      for (i = 0; i+3 < count; i += 2) {
-         QUAD( vcache,
-               get_elt(elts, i + 2) + elt_bias,
-               get_elt(elts, i + 0) + elt_bias,
-               get_elt(elts, i + 1) + elt_bias,
-               get_elt(elts, i + 3) + elt_bias );
-      }
-      break;
-
-   case PIPE_PRIM_POLYGON:
-      {
-         /* These bitflags look a little odd because we submit the
-          * vertices as (1,2,0) to satisfy flatshade requirements.  
-          */
-         ushort edge_next, edge_finish;
-
-         if (flatfirst) {
-            flags = (DRAW_PIPE_RESET_STIPPLE |
-                     DRAW_PIPE_EDGE_FLAG_1 |
-                     DRAW_PIPE_EDGE_FLAG_2);
-            edge_next = DRAW_PIPE_EDGE_FLAG_2;
-            edge_finish = DRAW_PIPE_EDGE_FLAG_0;
-         }
-         else {
-            flags = (DRAW_PIPE_RESET_STIPPLE |
-                     DRAW_PIPE_EDGE_FLAG_2 |
-                     DRAW_PIPE_EDGE_FLAG_0);
-            edge_next = DRAW_PIPE_EDGE_FLAG_0;
-            edge_finish = DRAW_PIPE_EDGE_FLAG_1;
-         }
-
-	 for (i = 0; i+2 < count; i++, flags = edge_next) {
-
-            if (i + 3 == count)
-               flags |= edge_finish;
-
-            if (flatfirst) {
-               TRIANGLE( vcache,
-                         flags,
-                         get_elt(elts, 0) + elt_bias,
-                         get_elt(elts, i + 1) + elt_bias,
-                         get_elt(elts, i + 2) + elt_bias );
-            }
-            else {
-               TRIANGLE( vcache,
-                         flags,
-                         get_elt(elts, i + 1) + elt_bias,
-                         get_elt(elts, i + 2) + elt_bias,
-                         get_elt(elts, 0) + elt_bias);
-            }
-	 }
-      }
-      break;
-
-   default:
-      assert(0);
-      break;
-   }
-   
-   vcache_flush( vcache );
-}
-
-
-#undef TRIANGLE
-#undef QUAD
-#undef POINT
-#undef LINE
-#undef FUNC

src/gallium/auxiliary/draw/draw_so_emit_tmp.h

@@ -1,123 +1,33 @@
+#define FUNC_VARS                               \
+   struct pt_so_emit *so,                       \
+   const struct draw_prim_info *input_prims,    \
+   const struct draw_vertex_info *input_verts,  \
+   unsigned start,                              \
+   unsigned count
 
-static void FUNC( struct pt_so_emit *so,
+#define FUNC_ENTER                                                \
-                  const struct draw_prim_info *input_prims,
+   /* declare more local vars */                                  \
-                  const struct draw_vertex_info *input_verts,
+   struct draw_context *draw = so->draw;                          \
-                  unsigned start,
+   const unsigned prim = input_prims->prim;                       \
-                  unsigned count)
+   const boolean last_vertex_last =                               \
-{
+      !(draw->rasterizer->flatshade &&                            \
-   struct draw_context *draw = so->draw;
+        draw->rasterizer->flatshade_first);                       \
+   do {                                                           \
+      debug_assert(input_prims->primitive_count == 1);            \
+      switch (prim) {                                             \
+      case PIPE_PRIM_LINES_ADJACENCY:                             \
+      case PIPE_PRIM_LINE_STRIP_ADJACENCY:                        \
+      case PIPE_PRIM_TRIANGLES_ADJACENCY:                         \
+      case PIPE_PRIM_TRIANGLE_STRIP_ADJACENCY:                    \
+         debug_assert(!"unexpected primitive type in stream output"); \
+         return;                                                  \
+      default:                                                    \
+         break;                                                   \
+      }                                                           \
+   } while (0)                                                    \
 
-   boolean flatfirst = (draw->rasterizer->flatshade &&
+#define POINT(i0)                so_point(so,i0)
-                        draw->rasterizer->flatshade_first);
+#define LINE(flags,i0,i1)        so_line(so,i0,i1)
-   unsigned i;
+#define TRIANGLE(flags,i0,i1,i2) so_tri(so,i0,i1,i2)
-   LOCAL_VARS
 
-   if (0) debug_printf("%s %d\n", __FUNCTION__, count);
+#include "draw_decompose_tmp.h"
-
-   debug_assert(input_prims->primitive_count == 1);
-
-   switch (input_prims->prim) {
-   case PIPE_PRIM_POINTS:
-      for (i = 0; i < count; i++) {
-	 POINT( so, start + i + 0 );
-      }
-      break;
-
-   case PIPE_PRIM_LINES:
-      for (i = 0; i+1 < count; i += 2) {
-         LINE( so , start + i + 0 , start + i + 1 );
-      }
-      break;
-
-   case PIPE_PRIM_LINE_LOOP:
-      if (count >= 2) {
-
-         for (i = 1; i < count; i++) {
-            LINE( so, start + i - 1, start + i );
-         }
-
-	 LINE( so, start + i - 1, start );
-      }
-      break;
-
-   case PIPE_PRIM_LINE_STRIP:
-      for (i = 1; i < count; i++) {
-         LINE( so, start + i - 1, start + i );
-      }
-      break;
-
-   case PIPE_PRIM_TRIANGLES:
-      for (i = 0; i+2 < count; i += 3) {
-         TRIANGLE( so, start + i + 0, start + i + 1, start + i + 2 );
-      }
-      break;
-
-   case PIPE_PRIM_TRIANGLE_STRIP:
-      if (flatfirst) {
-         for (i = 0; i+2 < count; i++) {
-            TRIANGLE( so,
-                      start + i + 0,
-                      start + i + 1 + (i&1),
-                      start + i + 2 - (i&1) );
-         }
-      }
-      else {
-         for (i = 0; i+2 < count; i++) {
-            TRIANGLE( so,
-                      start + i + 0 + (i&1),
-                      start + i + 1 - (i&1),
-                      start + i + 2 );
-         }
-      }
-      break;
-
-   case PIPE_PRIM_TRIANGLE_FAN:
-      if (count >= 3) {
-         if (flatfirst) {
-            for (i = 0; i+2 < count; i++) {
-               TRIANGLE( so,
-                         start + i + 1,
-                         start + i + 2,
-                         start );
-            }
-         }
-         else {
-            for (i = 0; i+2 < count; i++) {
-               TRIANGLE( so,
-                         start,
-                         start + i + 1,
-                         start + i + 2 );
-            }
-         }
-      }
-      break;
-
-   case PIPE_PRIM_POLYGON:
-      {
-         /* These bitflags look a little odd because we submit the
-          * vertices as (1,2,0) to satisfy flatshade requirements.
-          */
-
-	 for (i = 0; i+2 < count; i++) {
-
-            if (flatfirst) {
-               TRIANGLE( so, start + 0, start + i + 1, start + i + 2 );
-            }
-            else {
-               TRIANGLE( so, start + i + 1, start + i + 2, start + 0 );
-            }
-	 }
-      }
-      break;
-
-   default:
-      debug_assert(!"Unsupported primitive in stream output");
-      break;
-   }
-}
-
-
-#undef TRIANGLE
-#undef POINT
-#undef LINE
-#undef FUNC

src/gallium/auxiliary/draw/draw_vertex.h

@@ -166,7 +166,7 @@ static INLINE enum pipe_format draw_translate_vinfo_format(enum attrib_emit emit
    }
 }
 
-static INLINE enum attrib_emit draw_translate_vinfo_size(enum attrib_emit emit)
+static INLINE unsigned draw_translate_vinfo_size(enum attrib_emit emit)
 {
    switch (emit) {
    case EMIT_OMIT:

src/gallium/auxiliary/draw/draw_vs.c

@@ -48,18 +48,30 @@
 
 DEBUG_GET_ONCE_BOOL_OPTION(gallium_dump_vs, "GALLIUM_DUMP_VS", FALSE)
 
+
+/**
+ * Set a vertex shader constant buffer.
+ * \param slot  which constant buffer in [0, PIPE_MAX_CONSTANT_BUFFERS-1]
+ * \param constants  the mapped buffer
+ * \param size  size of buffer in bytes
+ */
 void
 draw_vs_set_constants(struct draw_context *draw,
                       unsigned slot,
                       const void *constants,
                       unsigned size)
 {
-   if (((uintptr_t)constants) & 0xf) {
+   const int alignment = 16;
+
+   /* check if buffer is 16-byte aligned */
+   if (((uintptr_t)constants) & (alignment - 1)) {
+      /* if not, copy the constants into a new, 16-byte aligned buffer */
       if (size > draw->vs.const_storage_size[slot]) {
          if (draw->vs.aligned_constant_storage[slot]) {
             align_free((void *)draw->vs.aligned_constant_storage[slot]);
          }
-         draw->vs.aligned_constant_storage[slot] = align_malloc(size, 16);
+         draw->vs.aligned_constant_storage[slot] =
+            align_malloc(size, alignment);
       }
       assert(constants);
       memcpy((void *)draw->vs.aligned_constant_storage[slot],

src/gallium/auxiliary/draw/draw_vs.h

@@ -133,7 +133,8 @@ struct draw_vertex_shader {
    void (*run_linear)( struct draw_vertex_shader *shader,
 		       const float (*input)[4],
 		       float (*output)[4],
-                      const void *constants[PIPE_MAX_CONSTANT_BUFFERS],
+                       const void *constants[PIPE_MAX_CONSTANT_BUFFERS],
+                       const unsigned const_size[PIPE_MAX_CONSTANT_BUFFERS],
 		       unsigned count,
 		       unsigned input_stride,
 		       unsigned output_stride );

src/gallium/auxiliary/draw/draw_vs_exec.c

@@ -85,7 +85,8 @@ static void
 vs_exec_run_linear( struct draw_vertex_shader *shader,
 		    const float (*input)[4],
 		    float (*output)[4],
-                   const void *constants[PIPE_MAX_CONSTANT_BUFFERS],
+                    const void *constants[PIPE_MAX_CONSTANT_BUFFERS],
+                    const unsigned const_size[PIPE_MAX_CONSTANT_BUFFERS],
 		    unsigned count,
 		    unsigned input_stride,
 		    unsigned output_stride )
@@ -95,9 +96,8 @@ vs_exec_run_linear( struct draw_vertex_shader *shader,
    unsigned int i, j;
    unsigned slot;
 
-   for (i = 0; i < PIPE_MAX_CONSTANT_BUFFERS; i++) {
+   tgsi_exec_set_constant_buffers(machine, PIPE_MAX_CONSTANT_BUFFERS,
-      machine->Consts[i] = constants[i];
+                                  constants, const_size);
-   }
 
    for (i = 0; i < count; i += MAX_TGSI_VERTICES) {
       unsigned int max_vertices = MIN2(MAX_TGSI_VERTICES, count - i);

src/gallium/auxiliary/draw/draw_vs_llvm.c

@@ -49,6 +49,7 @@ vs_llvm_run_linear( struct draw_vertex_shader *shader,
 		    const float (*input)[4],
 		    float (*output)[4],
                     const void *constants[PIPE_MAX_CONSTANT_BUFFERS],
+                    const unsigned constants_size[PIPE_MAX_CONSTANT_BUFFERS],
 		    unsigned count,
 		    unsigned input_stride,
 		    unsigned output_stride )

src/gallium/auxiliary/draw/draw_vs_sse.c

@@ -84,6 +84,7 @@ vs_sse_run_linear( struct draw_vertex_shader *base,
 		   const float (*input)[4],
 		   float (*output)[4],
                   const void *constants[PIPE_MAX_CONSTANT_BUFFERS],
+		  const unsigned const_size[PIPE_MAX_CONSTANT_BUFFERS],
 		   unsigned count,
 		   unsigned input_stride,
 		   unsigned output_stride )

src/gallium/auxiliary/draw/draw_vs_varient.c

@@ -149,7 +149,8 @@ static void PIPE_CDECL vsvg_run_elts( struct draw_vs_varient *varient,
    vsvg->base.vs->run_linear( vsvg->base.vs, 
                               temp_buffer,
                               temp_buffer,
-                             vsvg->base.vs->draw->pt.user.vs_constants,
+                              vsvg->base.vs->draw->pt.user.vs_constants,
+                              vsvg->base.vs->draw->pt.user.vs_constants_size,
                               count,
                               temp_vertex_stride, 
                               temp_vertex_stride);
@@ -214,7 +215,8 @@ static void PIPE_CDECL vsvg_run_linear( struct draw_vs_varient *varient,
    vsvg->base.vs->run_linear( vsvg->base.vs, 
                               temp_buffer,
                               temp_buffer,
-                             vsvg->base.vs->draw->pt.user.vs_constants,
+                              vsvg->base.vs->draw->pt.user.vs_constants,
+                              vsvg->base.vs->draw->pt.user.vs_constants_size,
                               count,
                               temp_vertex_stride, 
                               temp_vertex_stride);

src/gallium/auxiliary/gallivm/lp_bld_assert.c

@@ -0,0 +1,101 @@
+/**************************************************************************
+ *
+ * Copyright 2010 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 above copyright notice and this permission notice (including the
+ * next paragraph) 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 NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE 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.
+ *
+ **************************************************************************/
+
+#include "util/u_debug.h"
+#include "util/u_memory.h"
+#include "lp_bld_assert.h"
+#include "lp_bld_init.h"
+#include "lp_bld_printf.h"
+
+
+/**
+ * A call to lp_build_assert() will build a function call to this function.
+ */
+static void
+lp_assert(int condition, const char *msg)
+{
+   if (!condition) {
+      debug_printf("LLVM assertion '%s' failed!\n", msg);
+      assert(condition);
+   }
+}
+
+
+
+/**
+ * lp_build_assert.
+ *
+ * Build an assertion in LLVM IR by building a function call to the
+ * lp_assert() function above.
+ *
+ * \param condition should be an 'i1' or 'i32' value
+ * \param msg  a string to print if the assertion fails.
+ */
+LLVMValueRef
+lp_build_assert(LLVMBuilderRef builder, LLVMValueRef condition,
+                const char *msg)
+{
+   LLVMModuleRef module;
+   LLVMTypeRef arg_types[2];
+   LLVMValueRef msg_string, assert_func, params[2], r;
+
+   module = LLVMGetGlobalParent(LLVMGetBasicBlockParent(
+                            LLVMGetInsertBlock(builder)));
+
+   msg_string = lp_build_const_string_variable(module, msg, strlen(msg) + 1);
+
+   arg_types[0] = LLVMInt32Type();
+   arg_types[1] = LLVMPointerType(LLVMInt8Type(), 0);
+
+   /* lookup the lp_assert function */
+   assert_func = LLVMGetNamedFunction(module, "lp_assert");
+
+   /* Create the assertion function if not found */
+   if (!assert_func) {
+      LLVMTypeRef func_type =
+         LLVMFunctionType(LLVMVoidType(), arg_types, 2, 0);
+
+      assert_func = LLVMAddFunction(module, "lp_assert", func_type);
+      LLVMSetFunctionCallConv(assert_func, LLVMCCallConv);
+      LLVMSetLinkage(assert_func, LLVMExternalLinkage);
+      LLVMAddGlobalMapping(lp_build_engine, assert_func,
+                           func_to_pointer((func_pointer)lp_assert));
+   }
+   assert(assert_func);
+
+   /* build function call param list */
+   params[0] = LLVMBuildZExt(builder, condition, arg_types[0], "");
+   params[1] = LLVMBuildBitCast(builder, msg_string, arg_types[1], "");
+
+   /* check arg types */
+   assert(LLVMTypeOf(params[0]) == arg_types[0]);
+   assert(LLVMTypeOf(params[1]) == arg_types[1]);
+
+   r = LLVMBuildCall(builder, assert_func, params, 2, "");
+
+   return r;
+}

src/gallium/auxiliary/gallivm/lp_bld_assert.h

@@ -1,6 +1,6 @@
 /**************************************************************************
  *
- * Copyright 2009, VMware, Inc.
+ * Copyright 2010 VMware, Inc.
  * All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
@@ -25,13 +25,17 @@
  *
  **************************************************************************/
 
-#ifndef DRI2_H
+#ifndef LP_BLD_ASSERT_H
-#define DRI2_H
+#define LP_BLD_ASSERT_H
 
-#include "dri_drawable.h"
-#include "dri_wrapper.h"
 
-const __DRIconfig **
+#include "lp_bld.h"
-dri2_init_screen(__DRIscreen * sPriv);
+
+
+LLVMValueRef
+lp_build_assert(LLVMBuilderRef builder, LLVMValueRef condition,
+                const char *msg);
+
+
+#endif
 
-#endif /* DRI2_H */

src/gallium/auxiliary/gallivm/lp_bld_init.c

@@ -45,6 +45,8 @@ static const struct debug_named_value lp_bld_debug_flags[] = {
    { "nopt",   GALLIVM_DEBUG_NO_OPT, NULL },
    DEBUG_NAMED_VALUE_END
 };
+
+DEBUG_GET_ONCE_FLAGS_OPTION(gallivm_debug, "GALLIVM_DEBUG", lp_bld_debug_flags, 0);
 #endif
 
 
@@ -89,7 +91,7 @@ void
 lp_build_init(void)
 {
 #ifdef DEBUG
-   gallivm_debug = debug_get_flags_option("GALLIVM_DEBUG", lp_bld_debug_flags, 0 );
+   gallivm_debug = debug_get_option_gallivm_debug();
 #endif
 
    lp_set_target_options();

src/gallium/auxiliary/gallivm/lp_bld_init.h

@@ -44,5 +44,7 @@ extern LLVMPassManagerRef lp_build_pass;
 void
 lp_build_init(void);
 
+extern void
+lp_func_delete_body(LLVMValueRef func);
 
 #endif /* !LP_BLD_INIT_H */

src/gallium/auxiliary/gallivm/lp_bld_logic.c

@@ -362,10 +362,53 @@ lp_build_cmp(struct lp_build_context *bld,
 }
 
 
+/**
+ * Return (mask & a) | (~mask & b);
+ */
+LLVMValueRef
+lp_build_select_bitwise(struct lp_build_context *bld,
+                        LLVMValueRef mask,
+                        LLVMValueRef a,
+                        LLVMValueRef b)
+{
+   struct lp_type type = bld->type;
+   LLVMValueRef res;
+
+   if (a == b) {
+      return a;
+   }
+
+   if(type.floating) {
+      LLVMTypeRef int_vec_type = lp_build_int_vec_type(type);
+      a = LLVMBuildBitCast(bld->builder, a, int_vec_type, "");
+      b = LLVMBuildBitCast(bld->builder, b, int_vec_type, "");
+   }
+
+   a = LLVMBuildAnd(bld->builder, a, mask, "");
+
+   /* This often gets translated to PANDN, but sometimes the NOT is
+    * pre-computed and stored in another constant. The best strategy depends
+    * on available registers, so it is not a big deal -- hopefully LLVM does
+    * the right decision attending the rest of the program.
+    */
+   b = LLVMBuildAnd(bld->builder, b, LLVMBuildNot(bld->builder, mask, ""), "");
+
+   res = LLVMBuildOr(bld->builder, a, b, "");
+
+   if(type.floating) {
+      LLVMTypeRef vec_type = lp_build_vec_type(type);
+      res = LLVMBuildBitCast(bld->builder, res, vec_type, "");
+   }
+
+   return res;
+}
+
+
 /**
  * Return mask ? a : b;
  *
- * mask is a bitwise mask, composed of 0 or ~0 for each element.
+ * mask is a bitwise mask, composed of 0 or ~0 for each element. Any other value
+ * will yield unpredictable results.
  */
 LLVMValueRef
 lp_build_select(struct lp_build_context *bld,
@@ -424,27 +467,7 @@ lp_build_select(struct lp_build_context *bld,
       }
    }
    else {
-      if(type.floating) {
+      res = lp_build_select_bitwise(bld, mask, a, b);
-         LLVMTypeRef int_vec_type = lp_build_int_vec_type(type);
-         a = LLVMBuildBitCast(bld->builder, a, int_vec_type, "");
-         b = LLVMBuildBitCast(bld->builder, b, int_vec_type, "");
-      }
-
-      a = LLVMBuildAnd(bld->builder, a, mask, "");
-
-      /* This often gets translated to PANDN, but sometimes the NOT is
-       * pre-computed and stored in another constant. The best strategy depends
-       * on available registers, so it is not a big deal -- hopefully LLVM does
-       * the right decision attending the rest of the program.
-       */
-      b = LLVMBuildAnd(bld->builder, b, LLVMBuildNot(bld->builder, mask, ""), "");
-
-      res = LLVMBuildOr(bld->builder, a, b, "");
-
-      if(type.floating) {
-         LLVMTypeRef vec_type = lp_build_vec_type(type);
-         res = LLVMBuildBitCast(bld->builder, res, vec_type, "");
-      }
    }
 
    return res;

src/gallium/auxiliary/gallivm/lp_bld_logic.h

@@ -63,6 +63,11 @@ lp_build_cmp(struct lp_build_context *bld,
              LLVMValueRef a,
              LLVMValueRef b);
 
+LLVMValueRef
+lp_build_select_bitwise(struct lp_build_context *bld,
+                        LLVMValueRef mask,
+                        LLVMValueRef a,
+                        LLVMValueRef b);
 
 LLVMValueRef
 lp_build_select(struct lp_build_context *bld,

src/gallium/auxiliary/gallivm/lp_bld_misc.cpp

@@ -39,6 +39,7 @@
 #include <llvm/Target/TargetOptions.h>
 #include <llvm/ExecutionEngine/ExecutionEngine.h>
 #include <llvm/ExecutionEngine/JITEventListener.h>
+#include <llvm/Support/CommandLine.h>
 
 #include "pipe/p_config.h"
 #include "util/u_debug.h"
@@ -141,4 +142,35 @@ lp_set_target_options(void)
 #if 0
    llvm::UnsafeFPMath = true;
 #endif
+
+#if 0
+   /*
+    * LLVM will generate MMX instructions for vectors <= 64 bits, leading to
+    * innefficient code, and in 32bit systems, to the corruption of the FPU
+    * stack given that it expects the user to generate the EMMS instructions.
+    *
+    * See also:
+    * - http://llvm.org/bugs/show_bug.cgi?id=3287
+    * - http://l4.me.uk/post/2009/06/07/llvm-wrinkle-3-configuration-what-configuration/
+    *
+    * XXX: Unfortunately this is not working.
+    */
+   static boolean first = FALSE;
+   if (first) {
+      static const char* options[] = {
+         "prog",
+         "-disable-mmx"
+      };
+      llvm::cl::ParseCommandLineOptions(2, const_cast<char**>(options));
+      first = FALSE;
+   }
+#endif
+}
+
+
+extern "C" void
+lp_func_delete_body(LLVMValueRef FF)
+{
+   llvm::Function *func = llvm::unwrap<llvm::Function>(FF);
+   func->deleteBody();
 }

src/gallium/auxiliary/tgsi/tgsi_exec.c

@@ -557,6 +557,23 @@ print_temp(const struct tgsi_exec_machine *mach, uint index)
 #endif
 
 
+void
+tgsi_exec_set_constant_buffers(struct tgsi_exec_machine *mach,
+                               unsigned num_bufs,
+                               const void **bufs,
+                               const unsigned *buf_sizes)
+{
+   unsigned i;
+
+   for (i = 0; i < num_bufs; i++) {
+      mach->Consts[i] = bufs[i];
+      mach->ConstsSize[i] = buf_sizes[i];
+   }
+}
+
+
+
+
 /**
  * Check if there's a potential src/dst register data dependency when
  * using SOA execution.
@@ -632,6 +649,10 @@ tgsi_exec_machine_bind_shader(
 
    util_init_math();
 
+   if (numSamplers) {
+      assert(samplers);
+   }
+
    mach->Tokens = tokens;
    mach->Samplers = samplers;
 
@@ -1040,6 +1061,8 @@ fetch_src_file_channel(const struct tgsi_exec_machine *mach,
 {
    uint i;
 
+   assert(swizzle < 4);
+
    switch (file) {
    case TGSI_FILE_CONSTANT:
       for (i = 0; i < QUAD_SIZE; i++) {
@@ -1049,9 +1072,23 @@ fetch_src_file_channel(const struct tgsi_exec_machine *mach,
          if (index->i[i] < 0) {
             chan->u[i] = 0;
          } else {
-            const uint *p = (const uint *)mach->Consts[index2D->i[i]];
+            /* NOTE: copying the const value as a uint instead of float */
-
+            const uint constbuf = index2D->i[i];
-            chan->u[i] = p[index->i[i] * 4 + swizzle];
+            const uint *buf = (const uint *)mach->Consts[constbuf];
+            const int pos = index->i[i] * 4 + swizzle;
+            /* const buffer bounds check */
+            if (pos < 0 || pos >= mach->ConstsSize[constbuf]) {
+               if (0) {
+                  /* Debug: print warning */
+                  static int count = 0;
+                  if (count++ < 100)
+                     debug_printf("TGSI Exec: const buffer index %d"
+                                  " out of bounds\n", pos);
+               }
+               chan->u[i] = 0;
+            }
+            else
+               chan->u[i] = buf[pos];
          }
       }
       break;
@@ -1065,9 +1102,10 @@ fetch_src_file_channel(const struct tgsi_exec_machine *mach,
                          index2D->i[i] * TGSI_EXEC_MAX_INPUT_ATTRIBS + index->i[i],
                          index2D->i[i], index->i[i]);
                          }*/
-         chan->u[i] = mach->Inputs[index2D->i[i] *
+         int pos = index2D->i[i] * TGSI_EXEC_MAX_INPUT_ATTRIBS + index->i[i];
-                                   TGSI_EXEC_MAX_INPUT_ATTRIBS +
+         assert(pos >= 0);
-                                   index->i[i]].xyzw[swizzle].u[i];
+         assert(pos < Elements(mach->Inputs));
+         chan->u[i] = mach->Inputs[pos].xyzw[swizzle].u[i];
       }
       break;
 
@@ -1187,7 +1225,7 @@ fetch_source(const struct tgsi_exec_machine *mach,
       index2.i[1] =
       index2.i[2] =
       index2.i[3] = reg->Indirect.Index;
-
+      assert(reg->Indirect.File == TGSI_FILE_ADDRESS);
       /* get current value of address register[swizzle] */
       swizzle = tgsi_util_get_src_register_swizzle( &reg->Indirect, CHAN_X );
       fetch_src_file_channel(mach,

src/gallium/auxiliary/tgsi/tgsi_exec.h

@@ -253,7 +253,10 @@ struct tgsi_exec_machine
    struct tgsi_sampler           **Samplers;
 
    unsigned                      ImmLimit;
+
    const void *Consts[PIPE_MAX_CONSTANT_BUFFERS];
+   unsigned ConstsSize[PIPE_MAX_CONSTANT_BUFFERS];
+
    const struct tgsi_token       *Tokens;   /**< Declarations, instructions */
    unsigned                      Processor; /**< TGSI_PROCESSOR_x */
 
@@ -367,6 +370,13 @@ tgsi_set_exec_mask(struct tgsi_exec_machine *mach,
 }
 
 
+extern void
+tgsi_exec_set_constant_buffers(struct tgsi_exec_machine *mach,
+                               unsigned num_bufs,
+                               const void **bufs,
+                               const unsigned *buf_sizes);
+
+
 #if defined __cplusplus
 } /* extern "C" */
 #endif

src/gallium/auxiliary/tgsi/tgsi_sanity.c

@@ -33,6 +33,10 @@
 #include "tgsi_info.h"
 #include "tgsi_iterate.h"
 
+
+DEBUG_GET_ONCE_BOOL_OPTION(print_sanity, "TGSI_PRINT_SANITY", TRUE);
+
+
 typedef struct {
    uint file : 28;
    /* max 2 dimensions */
@@ -54,6 +58,8 @@ struct sanity_check_ctx
    uint errors;
    uint warnings;
    uint implied_array_size;
+
+   boolean print;
 };
 
 static INLINE unsigned
@@ -148,6 +154,9 @@ report_error(
 {
    va_list args;
 
+   if (!ctx->print)
+      return;
+
    debug_printf( "Error  : " );
    va_start( args, format );
    _debug_vprintf( format, args );
@@ -164,6 +173,9 @@ report_warning(
 {
    va_list args;
 
+   if (!ctx->print)
+      return;
+
    debug_printf( "Warning: " );
    va_start( args, format );
    _debug_vprintf( format, args );
@@ -539,6 +551,7 @@ tgsi_sanity_check(
    ctx.errors = 0;
    ctx.warnings = 0;
    ctx.implied_array_size = 0;
+   ctx.print = debug_get_option_print_sanity();
 
    if (!tgsi_iterate_shader( tokens, &ctx.iter ))
       return FALSE;

src/gallium/auxiliary/tgsi/tgsi_sanity.h

@@ -35,7 +35,8 @@ extern "C" {
 #endif
 
 /* Check the given token stream for errors and common mistakes.
- * Diagnostic messages are printed out to the debug output.
+ * Diagnostic messages are printed out to the debug output, and is
+ * controlled by the debug option TGSI_PRINT_SANITY (default true).
  * Returns TRUE if there are no errors, even though there could be some warnings.
  */
 boolean

src/gallium/auxiliary/translate/translate_generic.c

@@ -368,23 +368,23 @@ static void PIPE_CDECL generic_run_elts( struct translate *translate,
    /* loop over vertex attributes (vertex shader inputs)
     */
    for (i = 0; i < count; i++) {
-      unsigned elt = *elts++;
+      const unsigned elt = *elts++;
 
       for (attr = 0; attr < nr_attrs; attr++) {
 	 float data[4];
-         const uint8_t *src;
+	 char *dst = vert + tg->attrib[attr].output_offset;
-         unsigned index;
-
-	 char *dst = (vert + 
-		      tg->attrib[attr].output_offset);
 
          if (tg->attrib[attr].type == TRANSLATE_ELEMENT_NORMAL) {
+            const uint8_t *src;
+            unsigned index;
+
             if (tg->attrib[attr].instance_divisor) {
                index = instance_id / tg->attrib[attr].instance_divisor;
             } else {
                index = elt;
             }
 
+            /* clamp to void going out of bounds */
             index = MIN2(index, tg->attrib[attr].max_index);
 
             src = tg->attrib[attr].input_ptr +
@@ -392,11 +392,23 @@ static void PIPE_CDECL generic_run_elts( struct translate *translate,
 
             tg->attrib[attr].fetch( data, src, 0, 0 );
 
+            if (0)
+               debug_printf("Fetch elt attr %d  from %p  stride %d  div %u  max %u  index %d:  "
+                            " %f, %f, %f, %f \n",
+                            attr,
+                            tg->attrib[attr].input_ptr,
+                            tg->attrib[attr].input_stride,
+                            tg->attrib[attr].instance_divisor,
+                            tg->attrib[attr].max_index,
+                            index,
+                            data[0], data[1],data[2], data[3]);
          } else {
             data[0] = (float)instance_id;
          }
-         if (0) debug_printf("vert %d/%d attr %d: %f %f %f %f\n",
+
-                             i, elt, attr, data[0], data[1], data[2], data[3]);
+         if (0)
+            debug_printf("vert %d/%d attr %d: %f %f %f %f\n",
+                         i, elt, attr, data[0], data[1], data[2], data[3]);
 
 	 tg->attrib[attr].emit( data, dst );
       }
@@ -425,29 +437,42 @@ static void PIPE_CDECL generic_run( struct translate *translate,
 
       for (attr = 0; attr < nr_attrs; attr++) {
 	 float data[4];
-
+	 char *dst = vert + tg->attrib[attr].output_offset;
-	 char *dst = (vert + 
-		      tg->attrib[attr].output_offset);
 
          if (tg->attrib[attr].type == TRANSLATE_ELEMENT_NORMAL) {
             const uint8_t *src;
+            unsigned index;
 
             if (tg->attrib[attr].instance_divisor) {
-               src = tg->attrib[attr].input_ptr +
+               index = instance_id / tg->attrib[attr].instance_divisor;
-                     tg->attrib[attr].input_stride *
+            }
-                     (instance_id / tg->attrib[attr].instance_divisor);
+            else {
-            } else {
+               index = elt;
-               src = tg->attrib[attr].input_ptr +
-                     tg->attrib[attr].input_stride * elt;
             }
 
+            /* clamp to void going out of bounds */
+            index = MIN2(index, tg->attrib[attr].max_index);
+
+            src = tg->attrib[attr].input_ptr +
+                  tg->attrib[attr].input_stride * index;
+
             tg->attrib[attr].fetch( data, src, 0, 0 );
+
+            if (0)
+               debug_printf("Fetch linear attr %d  from %p  stride %d  index %d: "
+                            " %f, %f, %f, %f \n",
+                            attr,
+                            tg->attrib[attr].input_ptr,
+                            tg->attrib[attr].input_stride,
+                            index,
+                            data[0], data[1],data[2], data[3]);
          } else {
             data[0] = (float)instance_id;
          }
 
-         if (0) debug_printf("vert %d attr %d: %f %f %f %f\n",
+         if (0)
-                             i, attr, data[0], data[1], data[2], data[3]);
+            debug_printf("vert %d attr %d: %f %f %f %f\n",
+                         i, attr, data[0], data[1], data[2], data[3]);
 
 	 tg->attrib[attr].emit( data, dst );
       }

src/gallium/auxiliary/util/u_blitter.c

@@ -87,6 +87,7 @@ struct blitter_context_priv
    void *dsa_write_depth_keep_stencil;
    void *dsa_keep_depth_stencil;
    void *dsa_keep_depth_write_stencil;
+   void *dsa_flush_depth_stencil;
 
    void *velem_state;
 
@@ -156,6 +157,10 @@ struct blitter_context *util_blitter_create(struct pipe_context *pipe)
    ctx->dsa_keep_depth_stencil =
       pipe->create_depth_stencil_alpha_state(pipe, &dsa);
 
+   dsa.depth.writemask = 1;
+   ctx->dsa_flush_depth_stencil =
+      pipe->create_depth_stencil_alpha_state(pipe, &dsa);
+
    dsa.depth.enabled = 1;
    dsa.depth.writemask = 1;
    dsa.depth.func = PIPE_FUNC_ALWAYS;
@@ -940,3 +945,42 @@ void util_blitter_clear_depth_stencil(struct blitter_context *blitter,
                            UTIL_BLITTER_ATTRIB_NONE, NULL);
    blitter_restore_CSOs(ctx);
 }
+
+/* Clear a region of a depth stencil surface. */
+void util_blitter_flush_depth_stencil(struct blitter_context *blitter,
+                                      struct pipe_surface *dstsurf)
+{
+   struct blitter_context_priv *ctx = (struct blitter_context_priv*)blitter;
+   struct pipe_context *pipe = ctx->base.pipe;
+   struct pipe_framebuffer_state fb_state;
+
+   assert(dstsurf->texture);
+   if (!dstsurf->texture)
+      return;
+
+   /* check the saved state */
+   blitter_check_saved_CSOs(ctx);
+   assert(blitter->saved_fb_state.nr_cbufs != ~0);
+
+   /* bind CSOs */
+   pipe->bind_blend_state(pipe, ctx->blend_keep_color);
+   pipe->bind_depth_stencil_alpha_state(pipe, ctx->dsa_flush_depth_stencil);
+
+   pipe->bind_rasterizer_state(pipe, ctx->rs_state);
+   pipe->bind_fs_state(pipe, blitter_get_fs_col(ctx, 0));
+   pipe->bind_vs_state(pipe, ctx->vs_col);
+   pipe->bind_vertex_elements_state(pipe, ctx->velem_state);
+
+   /* set a framebuffer state */
+   fb_state.width = dstsurf->width;
+   fb_state.height = dstsurf->height;
+   fb_state.nr_cbufs = 0;
+   fb_state.cbufs[0] = 0;
+   fb_state.zsbuf = dstsurf;
+   pipe->set_framebuffer_state(pipe, &fb_state);
+
+   blitter_set_dst_dimensions(ctx, dstsurf->width, dstsurf->height);
+   blitter->draw_rectangle(blitter, 0, 0, dstsurf->width, dstsurf->height, 0,
+                           UTIL_BLITTER_ATTRIB_NONE, NULL);
+   blitter_restore_CSOs(ctx);
+}

src/gallium/auxiliary/util/u_blitter.h

@@ -200,6 +200,8 @@ void util_blitter_clear_depth_stencil(struct blitter_context *blitter,
                                       unsigned dstx, unsigned dsty,
                                       unsigned width, unsigned height);
 
+void util_blitter_flush_depth_stencil(struct blitter_context *blitter,
+                                      struct pipe_surface *dstsurf);
 /* The functions below should be used to save currently bound constant state
  * objects inside a driver. The objects are automatically restored at the end
  * of the util_blitter_{clear, copy_region, fill_region} functions and then

src/gallium/auxiliary/util/u_cpu_detect.c

@@ -73,9 +73,15 @@
 #endif
 
 
+DEBUG_GET_ONCE_BOOL_OPTION(dump_cpu, "GALLIUM_DUMP_CPU", TRUE);
+
+
 struct util_cpu_caps util_cpu_caps;
 
+#if defined(PIPE_ARCH_X86) || defined(PIPE_ARCH_X86_64)
 static int has_cpuid(void);
+#endif
+
 
 #if defined(PIPE_ARCH_X86)
 
@@ -497,23 +503,25 @@ util_cpu_detect(void)
 #endif /* PIPE_ARCH_PPC */
 
 #ifdef DEBUG
-   debug_printf("util_cpu_caps.arch = %i\n", util_cpu_caps.arch);
+   if (debug_get_option_dump_cpu()) {
-   debug_printf("util_cpu_caps.nr_cpus = %u\n", util_cpu_caps.nr_cpus);
+      debug_printf("util_cpu_caps.arch = %i\n", util_cpu_caps.arch);
+      debug_printf("util_cpu_caps.nr_cpus = %u\n", util_cpu_caps.nr_cpus);
 
-   debug_printf("util_cpu_caps.x86_cpu_type = %u\n", util_cpu_caps.x86_cpu_type);
+      debug_printf("util_cpu_caps.x86_cpu_type = %u\n", util_cpu_caps.x86_cpu_type);
-   debug_printf("util_cpu_caps.cacheline = %u\n", util_cpu_caps.cacheline);
+      debug_printf("util_cpu_caps.cacheline = %u\n", util_cpu_caps.cacheline);
 
-   debug_printf("util_cpu_caps.has_tsc = %u\n", util_cpu_caps.has_tsc);
+      debug_printf("util_cpu_caps.has_tsc = %u\n", util_cpu_caps.has_tsc);
-   debug_printf("util_cpu_caps.has_mmx = %u\n", util_cpu_caps.has_mmx);
+      debug_printf("util_cpu_caps.has_mmx = %u\n", util_cpu_caps.has_mmx);
-   debug_printf("util_cpu_caps.has_mmx2 = %u\n", util_cpu_caps.has_mmx2);
+      debug_printf("util_cpu_caps.has_mmx2 = %u\n", util_cpu_caps.has_mmx2);
-   debug_printf("util_cpu_caps.has_sse = %u\n", util_cpu_caps.has_sse);
+      debug_printf("util_cpu_caps.has_sse = %u\n", util_cpu_caps.has_sse);
-   debug_printf("util_cpu_caps.has_sse2 = %u\n", util_cpu_caps.has_sse2);
+      debug_printf("util_cpu_caps.has_sse2 = %u\n", util_cpu_caps.has_sse2);
-   debug_printf("util_cpu_caps.has_sse3 = %u\n", util_cpu_caps.has_sse3);
+      debug_printf("util_cpu_caps.has_sse3 = %u\n", util_cpu_caps.has_sse3);
-   debug_printf("util_cpu_caps.has_ssse3 = %u\n", util_cpu_caps.has_ssse3);
+      debug_printf("util_cpu_caps.has_ssse3 = %u\n", util_cpu_caps.has_ssse3);
-   debug_printf("util_cpu_caps.has_sse4_1 = %u\n", util_cpu_caps.has_sse4_1);
+      debug_printf("util_cpu_caps.has_sse4_1 = %u\n", util_cpu_caps.has_sse4_1);
-   debug_printf("util_cpu_caps.has_3dnow = %u\n", util_cpu_caps.has_3dnow);
+      debug_printf("util_cpu_caps.has_3dnow = %u\n", util_cpu_caps.has_3dnow);
-   debug_printf("util_cpu_caps.has_3dnow_ext = %u\n", util_cpu_caps.has_3dnow_ext);
+      debug_printf("util_cpu_caps.has_3dnow_ext = %u\n", util_cpu_caps.has_3dnow_ext);
-   debug_printf("util_cpu_caps.has_altivec = %u\n", util_cpu_caps.has_altivec);
+      debug_printf("util_cpu_caps.has_altivec = %u\n", util_cpu_caps.has_altivec);
+   }
 #endif
 
    util_cpu_detect_initialized = TRUE;

src/gallium/auxiliary/util/u_draw.h

@@ -0,0 +1,138 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * 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 above copyright notice and this permission notice (including the
+ * next paragraph) 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 NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS 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.
+ *
+ **************************************************************************/
+
+#ifndef U_DRAW_H
+#define U_DRAW_H
+
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_context.h"
+
+
+static INLINE void
+util_draw_init_info(struct pipe_draw_info *info)
+{
+   memset(info, 0, sizeof(*info));
+   info->instance_count = 1;
+   info->max_index = 0xffffffff;
+}
+
+
+static INLINE void
+util_draw_arrays(struct pipe_context *pipe, uint mode, uint start, uint count)
+{
+   struct pipe_draw_info info;
+
+   util_draw_init_info(&info);
+   info.mode = mode;
+   info.start = start;
+   info.count = count;
+   info.min_index = start;
+   info.max_index = start + count - 1;
+
+   pipe->draw_vbo(pipe, &info);
+}
+
+static INLINE void
+util_draw_elements(struct pipe_context *pipe, int index_bias,
+                   uint mode, uint start, uint count)
+{
+   struct pipe_draw_info info;
+
+   util_draw_init_info(&info);
+   info.indexed = TRUE;
+   info.mode = mode;
+   info.start = start;
+   info.count = count;
+   info.index_bias = index_bias;
+
+   pipe->draw_vbo(pipe, &info);
+}
+
+static INLINE void
+util_draw_arrays_instanced(struct pipe_context *pipe,
+                           uint mode, uint start, uint count,
+                           uint start_instance,
+                           uint instance_count)
+{
+   struct pipe_draw_info info;
+
+   util_draw_init_info(&info);
+   info.mode = mode;
+   info.start = start;
+   info.count = count;
+   info.start_instance = start_instance;
+   info.instance_count = instance_count;
+   info.min_index = start;
+   info.max_index = start + count - 1;
+
+   pipe->draw_vbo(pipe, &info);
+}
+
+static INLINE void
+util_draw_elements_instanced(struct pipe_context *pipe,
+                             int index_bias,
+                             uint mode, uint start, uint count,
+                             uint start_instance,
+                             uint instance_count)
+{
+   struct pipe_draw_info info;
+
+   util_draw_init_info(&info);
+   info.indexed = TRUE;
+   info.mode = mode;
+   info.start = start;
+   info.count = count;
+   info.index_bias = index_bias;
+   info.start_instance = start_instance;
+   info.instance_count = instance_count;
+
+   pipe->draw_vbo(pipe, &info);
+}
+
+static INLINE void
+util_draw_range_elements(struct pipe_context *pipe,
+                         int index_bias,
+                         uint min_index,
+                         uint max_index,
+                         uint mode, uint start, uint count)
+{
+   struct pipe_draw_info info;
+
+   util_draw_init_info(&info);
+   info.indexed = TRUE;
+   info.mode = mode;
+   info.start = start;
+   info.count = count;
+   info.index_bias = index_bias;
+   info.min_index = min_index;
+   info.max_index = max_index;
+
+   pipe->draw_vbo(pipe, &info);
+}
+
+#endif

src/gallium/auxiliary/util/u_draw_quad.c

@@ -60,7 +60,7 @@ util_draw_vertex_buffer(struct pipe_context *pipe,
    /* note: vertex elements already set by caller */
 
    /* draw */
-   pipe->draw_arrays(pipe, prim_type, 0, num_verts);
+   util_draw_arrays(pipe, prim_type, 0, num_verts);
 }
 
 

src/gallium/auxiliary/util/u_draw_quad.h

@@ -29,12 +29,18 @@
 #define U_DRAWQUAD_H
 
 
+#include "pipe/p_compiler.h"
+#include "pipe/p_context.h"
+
+
 #ifdef __cplusplus
 extern "C" {
 #endif
 
 struct pipe_resource;
 
+#include "util/u_draw.h"
+
 extern void 
 util_draw_vertex_buffer(struct pipe_context *pipe,
                         struct pipe_resource *vbuf, uint offset,

src/gallium/auxiliary/util/u_format_other.c

@@ -121,6 +121,15 @@ util_format_r1_unorm_pack_rgba_8unorm(uint8_t *dst_row, unsigned dst_stride,
  * A.k.a. D3DFMT_CxV8U8
  */
 
+static uint8_t
+r8g8bx_derive(int16_t r, int16_t g)
+{
+   /* Derive blue from red and green components.
+    * Apparently, we must always use integers to perform calculations,
+    * otherwise the results won't match D3D's CxV8U8 definition.
+    */
+   return (uint8_t)sqrtf(0x7f * 0x7f - r * r - g * g) * 0xff / 0x7f;
+}
 
 void
 util_format_r8g8bx_snorm_unpack_rgba_float(float *dst_row, unsigned dst_stride,
@@ -145,7 +154,7 @@ util_format_r8g8bx_snorm_unpack_rgba_float(float *dst_row, unsigned dst_stride,
 
          dst[0] = (float)(r * (1.0f/0x7f)); /* r */
          dst[1] = (float)(g * (1.0f/0x7f)); /* g */
-         dst[2] = sqrtf(1.0f - dst[0] * dst[0] - dst[1] * dst[1]); /* b */
+         dst[2] = r8g8bx_derive(r, g) * (1.0f/0xff); /* b */
          dst[3] = 1.0f; /* a */
          dst += 4;
       }
@@ -177,7 +186,7 @@ util_format_r8g8bx_snorm_unpack_rgba_8unorm(uint8_t *dst_row, unsigned dst_strid
 
          dst[0] = (uint8_t)(((uint16_t)MAX2(r, 0)) * 0xff / 0x7f); /* r */
          dst[1] = (uint8_t)(((uint16_t)MAX2(g, 0)) * 0xff / 0x7f); /* g */
-         dst[2] = (uint8_t)sqrtf(0x7f*0x7f - r * r - g * g) * 0xff / 0x7f; /* b */
+         dst[2] = r8g8bx_derive(r, g); /* b */
          dst[3] = 255; /* a */
          dst += 4;
       }
@@ -262,6 +271,6 @@ util_format_r8g8bx_snorm_fetch_rgba_float(float *dst, const uint8_t *src,
 
    dst[0] = r * (1.0f/0x7f); /* r */
    dst[1] = g * (1.0f/0x7f); /* g */
-   dst[2] = sqrtf(1.0f - dst[0] * dst[0] - dst[1] * dst[1]); /* b */
+   dst[2] = r8g8bx_derive(r, g) * (1.0f/0xff); /* b */
    dst[3] = 1.0f; /* a */
 }

src/gallium/auxiliary/util/u_pack_color.h

@@ -425,6 +425,53 @@ util_pack_color(const float rgba[4], enum pipe_format format, union util_color *
    }
 }
  
+/* Integer versions of util_pack_z and util_pack_z_stencil - useful for
+ * constructing clear masks.
+ */
+static INLINE uint
+util_pack_uint_z(enum pipe_format format, unsigned z)
+{
+   switch (format) {
+   case PIPE_FORMAT_Z16_UNORM:
+      return z & 0xffff;
+   case PIPE_FORMAT_Z32_UNORM:
+   case PIPE_FORMAT_Z32_FLOAT:
+      return z;
+   case PIPE_FORMAT_Z24_UNORM_S8_USCALED:
+   case PIPE_FORMAT_Z24X8_UNORM:
+      return z & 0xffffff;
+   case PIPE_FORMAT_S8_USCALED_Z24_UNORM:
+   case PIPE_FORMAT_X8Z24_UNORM:
+      return (z & 0xffffff) << 8;
+   case PIPE_FORMAT_S8_USCALED:
+      return 0;
+   default:
+      debug_print_format("gallium: unhandled format in util_pack_z()", format);
+      assert(0);
+      return 0;
+   }
+}
+
+static INLINE uint
+util_pack_uint_z_stencil(enum pipe_format format, double z, uint s)
+{
+   unsigned packed = util_pack_uint_z(format, z);
+
+   s &= 0xff;
+
+   switch (format) {
+   case PIPE_FORMAT_Z24_UNORM_S8_USCALED:
+      return packed | (s << 24);
+   case PIPE_FORMAT_S8_USCALED_Z24_UNORM:
+      return packed | s;
+   case PIPE_FORMAT_S8_USCALED:
+      return packed | s;
+   default:
+      return packed;
+   }
+}
+
+
 
 /**
  * Note: it's assumed that z is in [0,1]

src/gallium/auxiliary/util/u_prim.h

@@ -108,6 +108,20 @@ static INLINE boolean u_trim_pipe_prim( unsigned pipe_prim, unsigned *nr )
       ok = (*nr >= 4);
       *nr -= (*nr % 2);
       break;
+   case PIPE_PRIM_LINES_ADJACENCY:
+      ok = (*nr >= 4);
+      *nr -= (*nr % 4);
+      break;
+   case PIPE_PRIM_LINE_STRIP_ADJACENCY:
+      ok = (*nr >= 4);
+      break;
+   case PIPE_PRIM_TRIANGLES_ADJACENCY:
+      ok = (*nr >= 6);
+      *nr -= (*nr % 5);
+      break;
+   case PIPE_PRIM_TRIANGLE_STRIP_ADJACENCY:
+      ok = (*nr >= 4);
+      break;
    default:
       ok = 0;
       break;

src/gallium/auxiliary/util/u_sse.h

@@ -41,7 +41,6 @@
 
 #if defined(PIPE_ARCH_SSE)
 
-#include <xmmintrin.h>
 #include <emmintrin.h>
 
 

src/gallium/docs/d3d11ddi.txt

@@ -66,9 +66,6 @@ Unordered access view: view supporting random read/write access (usually from co
 clear
 	+ Gallium supports clearing both render targets and depth/stencil with a single call
 
-draw_range_elements
-	+ Gallium supports indexed draw with explicit range
-
 fence_signalled
 fence_finish
 	+ D3D10/D3D11 don't appear to support explicit fencing; queries can often substitute though, and flushing is supported
@@ -271,31 +268,27 @@ Dispatch (D3D11 only)
 DispatchIndirect (D3D11 only)
 	- Gallium does not support compute shaders
 
-Draw -> draw_arrays
+Draw -> draw_vbo
 	! D3D11 sets primitive modes separately with IaSetTopology: it's not obvious which is better
 
 DrawAuto -> draw_auto
 
-DrawIndexed -> draw_elements
+DrawIndexed -> draw_vbo
 	! D3D11 sets primitive modes separately with IaSetTopology: it's not obvious which is better
-	* may want to add a separate set_index_buffer
+	+ D3D11 lacks explicit range, which is required for OpenGL
-	- Gallium lacks base vertex for indexed draw calls
-	+ D3D11 lacks draw_range_elements functionality, which is required for OpenGL
 
-DrawIndexedInstanced -> draw_elements_instanced
+DrawIndexedInstanced -> draw_vbo
 	! D3D11 sets primitive modes separately with IaSetTopology: it's not obvious which is better
-	* may want to add a separate set_index_buffer
-	- Gallium lacks base vertex for indexed draw calls
 
-DrawIndexedInstancedIndirect (D3D11 only) -> call draw_elements_instanced multiple times in software
+DrawIndexedInstancedIndirect (D3D11 only)
-	# this allows to use an hardware buffer to specify the parameters for multiple draw_elements_instanced calls
+	# this allows to use an hardware buffer to specify the parameters for multiple draw_vbo calls
 	- Gallium does not support draw call parameter buffers and indirect draw
 
-DrawInstanced -> draw_arrays_instanced
+DrawInstanced -> draw_vbo
 	! D3D11 sets primitive modes separately with IaSetTopology: it's not obvious which is better
 
-DrawInstancedIndirect (D3D11 only) -> call draw_arrays_instanced multiple times in software
+DrawInstancedIndirect (D3D11 only)
-	# this allows to use an hardware buffer to specify the parameters for multiple draw_arrays_instanced calls
+	# this allows to use an hardware buffer to specify the parameters for multiple draw_vbo calls
 	- Gallium does not support draw call parameter buffers and indirect draws
 
 DsSetConstantBuffers (D3D11 only)
@@ -332,10 +325,9 @@ HsSetShaderResources (D3D11 only)
 HsSetShaderWithIfaces (D3D11 only)
 	- Gallium does not support hull shaders
 
-IaSetIndexBuffer
+IaSetIndexBuffer -> set_index_buffer
-	! Gallium passes this to the draw_elements or draw_elements_instanced calls
 	+ Gallium supports 8-bit indices
-	! the D3D11 interface allows index-size-unaligned byte offsets into index buffers; it's not clear whether they actually work
+	# the D3D11 interface allows index-size-unaligned byte offsets into the index buffer; most drivers will abort with an assertion
 
 IaSetInputLayout -> bind_vertex_elements_state
 

src/gallium/docs/source/conf.py

@@ -38,7 +38,7 @@ master_doc = 'index'
 
 # General information about the project.
 project = u'Gallium'
-copyright = u'2009, VMWare, X.org, Nouveau'
+copyright = u'2009, VMware, X.org, Nouveau'
 
 # The version info for the project you're documenting, acts as replacement for
 # |version| and |release|, also used in various other places throughout the
@@ -176,7 +176,7 @@ htmlhelp_basename = 'Galliumdoc'
 # (source start file, target name, title, author, documentclass [howto/manual]).
 latex_documents = [
   ('index', 'Gallium.tex', u'Gallium Documentation',
-   u'VMWare, X.org, Nouveau', 'manual'),
+   u'VMware, X.org, Nouveau', 'manual'),
 ]
 
 # The name of an image file (relative to this directory) to place at the top of

src/gallium/docs/source/context.rst

@@ -45,6 +45,7 @@ buffers, surfaces) are bound to the driver.
 
 * ``set_vertex_buffers``
 
+* ``set_index_buffer``
 
 Non-CSO State
 ^^^^^^^^^^^^^
@@ -132,50 +133,26 @@ this surface need not be bound to the framebuffer.
 Drawing
 ^^^^^^^
 
-``draw_arrays`` draws a specified primitive.
+``draw_vbo`` draws a specified primitive.  The primitive mode and other
+properties are described by ``pipe_draw_info``.
 
-This command is equivalent to calling ``draw_arrays_instanced``
+The ``mode``, ``start``, and ``count`` fields of ``pipe_draw_info`` specify the
-with ``startInstance`` set to 0 and ``instanceCount`` set to 1.
+the mode of the primitive and the vertices to be fetched, in the range between
+``start`` to ``start``+``count``-1, inclusive.
 
-``draw_elements`` draws a specified primitive using an optional
+Every instance with instanceID in the range between ``start_instance`` and
-index buffer.
+``start_instance``+``instance_count``-1, inclusive, will be drawn.
 
-This command is equivalent to calling ``draw_elements_instanced``
+All vertex indices must fall inside the range given by ``min_index`` and
-with ``startInstance`` set to 0 and ``instanceCount`` set to 1.
+``max_index``.  In case non-indexed draw, ``min_index`` should be set to
+``start`` and ``max_index`` should be set to ``start``+``count``-1.
 
-``draw_range_elements``
+``index_bias`` is a value added to every vertex index before fetching vertex
+attributes.  It does not affect ``min_index`` and ``max_index``.
 
-XXX: this is (probably) a temporary entrypoint, as the range
+If there is an index buffer bound, and ``indexed`` field is true, all vertex
-information should be available from the vertex_buffer state.
+indices will be looked up in the index buffer.  ``min_index``, ``max_index``,
-Using this to quickly evaluate a specialized path in the draw
+and ``index_bias`` apply after index lookup.
-module.
-
-``draw_arrays_instanced`` draws multiple instances of the same primitive.
-
-This command is equivalent to calling ``draw_elements_instanced``
-with ``indexBuffer`` set to NULL and ``indexSize`` set to 0.
-
-``draw_elements_instanced`` draws multiple instances of the same primitive
-using an optional index buffer.
-
-For instanceID in the range between ``startInstance``
-and ``startInstance``+``instanceCount``-1, inclusive, draw a primitive
-specified by ``mode`` and sequential numbers in the range between ``start``
-and ``start``+``count``-1, inclusive.
-
-If ``indexBuffer`` is not NULL, it specifies an index buffer with index
-byte size of ``indexSize``. The sequential numbers are used to lookup
-the index buffer and the resulting indices in turn are used to fetch
-vertex attributes.
-
-If ``indexBuffer`` is NULL, the sequential numbers are used directly
-as indices to fetch vertex attributes.
-
-``indexBias`` is a value which is added to every index read from the index 
-buffer before fetching vertex attributes.
-
-``minIndex`` and ``maxIndex`` describe minimum and maximum index contained in
-the index buffer.
 
 If a given vertex element has ``instance_divisor`` set to 0, it is said
 it contains per-vertex data and effective vertex attribute address needs

src/gallium/docs/source/cso/rasterizer.rst

@@ -126,11 +126,15 @@ sprite_coord_enable
 
 Specifies if a texture unit has its texture coordinates replaced or not. This
 is a packed bitfield containing the enable for all texcoords -- if all bits
-are zero, point sprites are effectively disabled. If any bit is set, then
+are zero, point sprites are effectively disabled. 
-point_smooth and point_quad_rasterization are ignored; point smoothing is
+
-disabled and points are always rasterized as quads. If enabled, the four
+If any bit is set, then point_smooth MUST be disabled (there are no
-vertices of the resulting quad will be assigned texture coordinates,
+round sprites) and point_quad_rasterization MUST be true (sprites are
-according to sprite_coord_mode.
+always rasterized as quads).  Any mismatch between these states should
+be considered a bug in the state-tracker.
+
+If enabled, the four vertices of the resulting quad will be assigned
+texture coordinates, according to sprite_coord_mode.
 
 sprite_coord_mode
 ^^^^^^^^^^^^^^^^^
@@ -141,20 +145,23 @@ have coordinates (0,0,0,1). For PIPE_SPRITE_COORD_UPPER_LEFT, the upper-left
 vertex will have coordinates (0,0,0,1).
 This state is used by :ref:`Draw` to generate texcoords.
 
-.. note::
-
-    When geometry shaders are available, a special geometry shader could be
-    used instead of this functionality, to convert incoming points into quads
-    with the proper texture coordinates.
-
 point_quad_rasterization
 ^^^^^^^^^^^^^^^^^^^^^^^^
 
-Determines if points should be rasterized as quads or points. Certain APIs,
+Determines if points should be rasterized according to quad or point
-like Direct3D, always use quad rasterization for points, regardless of
+rasterization rules.
-whether point sprites are enabled or not. If this state is enabled, point
+
-smoothing and antialiasing are disabled. If it is disabled, point sprite
+OpenGL actually has quite different rasterization rules for points and
-coordinates are not generated.
+point sprites - hence this indicates if points should be rasterized as
+points or according to point sprite (which decomposes them into quads,
+basically) rules.
+
+Additionally Direct3D will always use quad rasterization rules for
+points, regardless of whether point sprites are enabled or not.
+
+If this state is enabled, point smoothing and antialiasing are
+disabled. If it is disabled, point sprite coordinates are not
+generated.
 
 .. note::
 

src/gallium/docs/source/distro.rst

@@ -51,10 +51,10 @@ nVidia nv50
 
 Driver for the nVidia nv50 family of GPUs.
 
-VMWare SVGA
+VMware SVGA
 ^^^^^^^^^^^
 
-Driver for VMWare virtualized guest operating system graphics processing.
+Driver for VMware virtualized guest operating system graphics processing.
 
 ATI r300
 ^^^^^^^^

src/gallium/drivers/cell/common.h

@@ -230,7 +230,7 @@ struct cell_command_rasterizer
 {
    opcode_t opcode;    /**< CELL_CMD_STATE_RASTERIZER */
    struct pipe_rasterizer_state rasterizer;
-   uint32_t pad[1];
+   /*uint32_t pad[1];*/
 };
 
 

src/gallium/drivers/cell/ppu/cell_context.h

@@ -132,6 +132,7 @@ struct cell_context
    struct pipe_viewport_state viewport;
    struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS];
    uint num_vertex_buffers;
+   struct pipe_index_buffer index_buffer;
 
    ubyte *cbuf_map[PIPE_MAX_COLOR_BUFS];
    ubyte *zsbuf_map;
@@ -154,7 +155,7 @@ struct cell_context
    struct vertex_info vertex_info;
 
    /** Mapped constant buffers */
-   void *mapped_constants[PIPE_SHADER_TYPES];
+   const void *mapped_constants[PIPE_SHADER_TYPES];
 
    PIPE_ALIGN_VAR(16) struct cell_spu_function_info spu_functions;
 

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

@@ -56,16 +56,11 @@
  * XXX should the element buffer be specified/bound with a separate function?
  */
 static void
-cell_draw_range_elements(struct pipe_context *pipe,
+cell_draw_vbo(struct pipe_context *pipe, const struct pipe_draw_info *info)
-                         struct pipe_resource *indexBuffer,
-                         unsigned indexSize,
-                         int indexBias,
-                         unsigned min_index,
-                         unsigned max_index,
-                         unsigned mode, unsigned start, unsigned count)
 {
    struct cell_context *cell = cell_context(pipe);
    struct draw_context *draw = cell->draw;
+   void *mapped_indices = NULL;
    unsigned i;
 
    if (cell->dirty)
@@ -83,18 +78,20 @@ cell_draw_range_elements(struct pipe_context *pipe,
       draw_set_mapped_vertex_buffer(draw, i, buf);
    }
    /* Map index buffer, if present */
-   if (indexBuffer) {
+   if (info->indexed && cell->index_buffer.buffer) {
-      void *mapped_indexes = cell_resource(indexBuffer)->data;
+      mapped_indices = cell_resource(cell->index_buffer.buffer)->data;
-      draw_set_mapped_element_buffer(draw, indexSize, indexBias, mapped_indexes);
+      mapped_indices += cell->index_buffer.offset;
-   }
-   else {
-      /* no index/element buffer */
-      draw_set_mapped_element_buffer(draw, 0, 0, NULL);
    }
 
+   draw_set_mapped_element_buffer_range(draw, (mapped_indices) ?
+                                        lp->index_buffer.index_size : 0,
+                                        info->index_bias,
+                                        info->min_index,
+                                        info->max_index,
+                                        mapped_indices);
 
    /* draw! */
-   draw_arrays(draw, mode, start, count);
+   draw_arrays(draw, info->mode, info->start, info->count);
 
    /*
     * unmap vertex/index buffers - will cause draw module to flush
@@ -102,7 +99,7 @@ cell_draw_range_elements(struct pipe_context *pipe,
    for (i = 0; i < cell->num_vertex_buffers; i++) {
       draw_set_mapped_vertex_buffer(draw, i, NULL);
    }
-   if (indexBuffer) {
+   if (mapped_indices) {
       draw_set_mapped_element_buffer(draw, 0, 0, NULL);
    }
 
@@ -115,32 +112,9 @@ cell_draw_range_elements(struct pipe_context *pipe,
 }
 
 
-static void
-cell_draw_elements(struct pipe_context *pipe,
-                   struct pipe_resource *indexBuffer,
-                   unsigned indexSize, int indexBias,
-                   unsigned mode, unsigned start, unsigned count)
-{
-   cell_draw_range_elements( pipe, indexBuffer,
-                             indexSize, indexBias,
-                             0, 0xffffffff,
-                             mode, start, count );
-}
-
-
-static void
-cell_draw_arrays(struct pipe_context *pipe, unsigned mode,
-                     unsigned start, unsigned count)
-{
-   cell_draw_elements(pipe, NULL, 0, 0, mode, start, count);
-}
-
-
 void
 cell_init_draw_functions(struct cell_context *cell)
 {
-   cell->pipe.draw_arrays = cell_draw_arrays;
+   cell->pipe.draw_vbo = cell_draw_vbo;
-   cell->pipe.draw_elements = cell_draw_elements;
-   cell->pipe.draw_range_elements = cell_draw_range_elements;
 }
 

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

@@ -87,6 +87,7 @@ struct cell_buffer_node
 };
 
 
+#if 0
 static void
 cell_add_buffer_to_list(struct cell_context *cell,
                         struct cell_buffer_list *list,
@@ -100,6 +101,7 @@ cell_add_buffer_to_list(struct cell_context *cell,
       list->head = node;
    }
 }
+#endif
 
 
 /**
@@ -113,7 +115,7 @@ cell_free_fenced_buffers(struct cell_context *cell,
                          struct cell_buffer_list *list)
 {
    if (list->head) {
-      struct pipe_screen *ps = cell->pipe.screen;
+      /*struct pipe_screen *ps = cell->pipe.screen;*/
       struct cell_buffer_node *node;
 
       cell_fence_finish(cell, &list->fence);
@@ -146,7 +148,7 @@ cell_free_fenced_buffers(struct cell_context *cell,
 void
 cell_add_fenced_textures(struct cell_context *cell)
 {
-   struct cell_buffer_list *list = &cell->fenced_buffers[cell->cur_batch];
+   /*struct cell_buffer_list *list = &cell->fenced_buffers[cell->cur_batch];*/
    uint i;
 
    for (i = 0; i < cell->num_textures; i++) {

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

@@ -281,7 +281,7 @@ cell_set_fragment_sampler_views(struct pipe_context *pipe,
          struct pipe_resource *new_tex = new_view ? new_view->texture : NULL;
 
          pipe_sampler_view_reference(&cell->fragment_sampler_views[i],
-                                     views[i]);
+                                     new_view);
          pipe_resource_reference((struct pipe_resource **) &cell->texture[i],
                                 (struct pipe_resource *) new_tex);
 

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

@@ -36,7 +36,7 @@
 #include "draw/draw_context.h"
 
 
-void *
+static void *
 cell_create_vertex_elements_state(struct pipe_context *pipe,
                                   unsigned count,
                                   const struct pipe_vertex_element *attribs)
@@ -51,7 +51,7 @@ cell_create_vertex_elements_state(struct pipe_context *pipe,
    return velems;
 }
 
-void
+static void
 cell_bind_vertex_elements_state(struct pipe_context *pipe,
                                 void *velems)
 {
@@ -66,7 +66,7 @@ cell_bind_vertex_elements_state(struct pipe_context *pipe,
       draw_set_vertex_elements(cell->draw, cell_velems->count, cell_velems->velem);
 }
 
-void
+static void
 cell_delete_vertex_elements_state(struct pipe_context *pipe, void *velems)
 {
    FREE( velems );
@@ -91,10 +91,26 @@ cell_set_vertex_buffers(struct pipe_context *pipe,
 }
 
 
+static void
+cell_set_index_buffer(struct pipe_context *pipe,
+                      const struct pipe_index_buffer *ib)
+{
+   struct cell_context *cell = cell_context(pipe);
+
+   if (ib)
+      memcpy(&cell->index_buffer, ib, sizeof(cell->index_buffer));
+   else
+      memset(&cell->index_buffer, 0, sizeof(cell->index_buffer));
+
+   /* TODO make this more like a state */
+}
+
+
 void
 cell_init_vertex_functions(struct cell_context *cell)
 {
    cell->pipe.set_vertex_buffers = cell_set_vertex_buffers;
+   cell->pipe.set_index_buffer = cell_set_index_buffer;
    cell->pipe.create_vertex_elements_state = cell_create_vertex_elements_state;
    cell->pipe.bind_vertex_elements_state = cell_bind_vertex_elements_state;
    cell->pipe.delete_vertex_elements_state = cell_delete_vertex_elements_state;

src/gallium/drivers/failover/fo_context.c

@@ -50,13 +50,8 @@ void failover_fail_over( struct failover_context *failover )
 }
 
 
-static void failover_draw_elements( struct pipe_context *pipe,
+static void failover_draw_vbo( struct pipe_context *pipe,
-                                    struct pipe_resource *indexResource,
+                               const struct pipe_draw_info *info)
-                                    unsigned indexSize,
-                                    int indexBias,
-                                    unsigned prim, 
-                                    unsigned start, 
-                                    unsigned count)
 {
    struct failover_context *failover = failover_context( pipe );
 
@@ -70,13 +65,7 @@ static void failover_draw_elements( struct pipe_context *pipe,
    /* Try hardware:
     */
    if (failover->mode == FO_HW) {
-      failover->hw->draw_elements( failover->hw, 
+      failover->hw->draw_vbo( failover->hw, info );
-                                   indexResource, 
-                                   indexSize, 
-                                   indexBias,
-                                   prim, 
-                                   start, 
-                                   count );
    }
 
    /* Possibly try software:
@@ -88,13 +77,7 @@ static void failover_draw_elements( struct pipe_context *pipe,
 	 failover_state_emit( failover );
       }
 
-      failover->sw->draw_elements( failover->sw, 
+      failover->sw->draw_vbo( failover->sw, info );
-				   indexResource, 
-				   indexSize, 
-				   indexBias,
-				   prim, 
-				   start, 
-				   count );
 
       /* Be ready to switch back to hardware rendering without an
        * intervening flush.  Unlikely to be much performance impact to
@@ -104,13 +87,6 @@ static void failover_draw_elements( struct pipe_context *pipe,
    }
 }
 
-
-static void failover_draw_arrays( struct pipe_context *pipe,
-				     unsigned prim, unsigned start, unsigned count)
-{
-   failover_draw_elements(pipe, NULL, 0, 0, prim, start, count);
-}
-
 static unsigned int
 failover_is_resource_referenced( struct pipe_context *_pipe,
 				 struct pipe_resource *resource,
@@ -143,8 +119,7 @@ struct pipe_context *failover_create( struct pipe_context *hw,
    failover->pipe.get_paramf = hw->get_paramf;
 #endif
 
-   failover->pipe.draw_arrays = failover_draw_arrays;
+   failover->pipe.draw_vbo = failover_draw_vbo;
-   failover->pipe.draw_elements = failover_draw_elements;
    failover->pipe.clear = hw->clear;
    failover->pipe.clear_render_target = hw->clear_render_target;
    failover->pipe.clear_depth_stencil = hw->clear_depth_stencil;

src/gallium/drivers/failover/fo_context.h

@@ -56,6 +56,7 @@
 #define FO_NEW_VERTEX_BUFFER   0x40000
 #define FO_NEW_VERTEX_ELEMENT  0x80000
 #define FO_NEW_SAMPLE_MASK     0x100000
+#define FO_NEW_INDEX_BUFFER    0x200000
 
 
 
@@ -97,6 +98,7 @@ struct failover_context {
    struct pipe_scissor_state scissor;
    struct pipe_viewport_state viewport;
    struct pipe_vertex_buffer vertex_buffers[PIPE_MAX_ATTRIBS];
+   struct pipe_index_buffer index_buffer;
 
    uint num_vertex_buffers;
 

src/gallium/drivers/failover/fo_state.c

@@ -583,6 +583,23 @@ failover_set_vertex_buffers(struct pipe_context *pipe,
 }
 
 
+static void
+failover_set_index_buffer(struct pipe_context *pipe,
+                          const struct pipe_index_buffer *ib)
+{
+   struct failover_context *failover = failover_context(pipe);
+
+   if (ib)
+      memcpy(&failover->index_buffer, ib, sizeof(failover->index_buffer));
+   else
+      memset(&failover->index_buffer, 0, sizeof(failover->index_buffer));
+
+   failover->dirty |= FO_NEW_INDEX_BUFFER;
+   failover->sw->set_index_buffer( failover->sw, ib );
+   failover->hw->set_index_buffer( failover->hw, ib );
+}
+
+
 void
 failover_set_constant_buffer(struct pipe_context *pipe,
                              uint shader, uint index,
@@ -635,6 +652,7 @@ failover_init_state_functions( struct failover_context *failover )
    failover->pipe.set_vertex_sampler_views = failover_set_vertex_sampler_views;
    failover->pipe.set_viewport_state = failover_set_viewport_state;
    failover->pipe.set_vertex_buffers = failover_set_vertex_buffers;
+   failover->pipe.set_index_buffer = failover_set_index_buffer;
    failover->pipe.set_constant_buffer = failover_set_constant_buffer;
    failover->pipe.create_sampler_view = failover_create_sampler_view;
    failover->pipe.sampler_view_destroy = failover_sampler_view_destroy;

src/gallium/drivers/failover/fo_state_emit.c

@@ -135,5 +135,10 @@ failover_state_emit( struct failover_context *failover )
                                         failover->vertex_buffers );
    }
 
+   if (failover->dirty & FO_NEW_INDEX_BUFFER) {
+      failover->sw->set_index_buffer( failover->sw,
+                                      &failover->index_buffer );
+   }
+
    failover->dirty = 0;
 }

src/gallium/drivers/galahad/glhd_context.c

@@ -48,68 +48,13 @@ galahad_destroy(struct pipe_context *_pipe)
 }
 
 static void
-galahad_draw_arrays(struct pipe_context *_pipe,
+galahad_draw_vbo(struct pipe_context *_pipe,
-                     unsigned prim,
+                 const struct pipe_draw_info *info)
-                     unsigned start,
-                     unsigned count)
 {
    struct galahad_context *glhd_pipe = galahad_context(_pipe);
    struct pipe_context *pipe = glhd_pipe->pipe;
 
-   pipe->draw_arrays(pipe,
+   pipe->draw_vbo(pipe, info);
-                     prim,
-                     start,
-                     count);
-}
-
-static void
-galahad_draw_elements(struct pipe_context *_pipe,
-                       struct pipe_resource *_indexResource,
-                       unsigned indexSize,
-                       int indexBias,
-                       unsigned prim,
-                       unsigned start,
-                       unsigned count)
-{
-   struct galahad_context *glhd_pipe = galahad_context(_pipe);
-   struct galahad_resource *glhd_resource = galahad_resource(_indexResource);
-   struct pipe_context *pipe = glhd_pipe->pipe;
-   struct pipe_resource *indexResource = glhd_resource->resource;
-
-   pipe->draw_elements(pipe,
-                       indexResource,
-                       indexSize,
-                       indexBias,
-                       prim,
-                       start,
-                       count);
-}
-
-static void
-galahad_draw_range_elements(struct pipe_context *_pipe,
-                             struct pipe_resource *_indexResource,
-                             unsigned indexSize,
-                             int indexBias,
-                             unsigned minIndex,
-                             unsigned maxIndex,
-                             unsigned mode,
-                             unsigned start,
-                             unsigned count)
-{
-   struct galahad_context *glhd_pipe = galahad_context(_pipe);
-   struct galahad_resource *glhd_resource = galahad_resource(_indexResource);
-   struct pipe_context *pipe = glhd_pipe->pipe;
-   struct pipe_resource *indexResource = glhd_resource->resource;
-
-   pipe->draw_range_elements(pipe,
-                             indexResource,
-                             indexSize,
-                             indexBias,
-                             minIndex,
-                             maxIndex,
-                             mode,
-                             start,
-                             count);
 }
 
 static struct pipe_query *
@@ -650,6 +595,41 @@ galahad_set_vertex_buffers(struct pipe_context *_pipe,
                             num_buffers,
                             buffers);
 }
+
+static void
+galahad_set_index_buffer(struct pipe_context *_pipe,
+                         const struct pipe_index_buffer *_ib)
+{
+   struct galahad_context *glhd_pipe = galahad_context(_pipe);
+   struct pipe_context *pipe = glhd_pipe->pipe;
+   struct pipe_index_buffer unwrapped_ib, *ib = NULL;
+
+   if (_ib->buffer) {
+      switch (_ib->index_size) {
+      case 1:
+      case 2:
+      case 4:
+         break;
+      default:
+         glhd_warn("index buffer %p has unrecognized index size %d",
+               _ib->buffer, _ib->index_size);
+         break;
+      }
+   }
+   else if (_ib->offset || _ib->index_size) {
+      glhd_warn("non-indexed state with index offset %d and index size %d",
+            _ib->offset, _ib->index_size);
+   }
+
+   if (_ib) {
+      unwrapped_ib = *_ib;
+      unwrapped_ib.buffer = galahad_resource_unwrap(_ib->buffer);
+      ib = &unwrapped_ib;
+   }
+
+   pipe->set_index_buffer(pipe, ib);
+}
+
 static void
 galahad_resource_copy_region(struct pipe_context *_pipe,
                               struct pipe_resource *_dst,
@@ -934,9 +914,7 @@ galahad_context_create(struct pipe_screen *_screen, struct pipe_context *pipe)
    glhd_pipe->base.draw = NULL;
 
    glhd_pipe->base.destroy = galahad_destroy;
-   glhd_pipe->base.draw_arrays = galahad_draw_arrays;
+   glhd_pipe->base.draw_vbo = galahad_draw_vbo;
-   glhd_pipe->base.draw_elements = galahad_draw_elements;
-   glhd_pipe->base.draw_range_elements = galahad_draw_range_elements;
    glhd_pipe->base.create_query = galahad_create_query;
    glhd_pipe->base.destroy_query = galahad_destroy_query;
    glhd_pipe->base.begin_query = galahad_begin_query;
@@ -976,6 +954,7 @@ galahad_context_create(struct pipe_screen *_screen, struct pipe_context *pipe)
    glhd_pipe->base.set_fragment_sampler_views = galahad_set_fragment_sampler_views;
    glhd_pipe->base.set_vertex_sampler_views = galahad_set_vertex_sampler_views;
    glhd_pipe->base.set_vertex_buffers = galahad_set_vertex_buffers;
+   glhd_pipe->base.set_index_buffer = galahad_set_index_buffer;
    glhd_pipe->base.resource_copy_region = galahad_resource_copy_region;
    glhd_pipe->base.clear = galahad_clear;
    glhd_pipe->base.clear_render_target = galahad_clear_render_target;

src/gallium/drivers/i915/i915_context.c

@@ -45,16 +45,11 @@
 
 
 static void
-i915_draw_range_elements(struct pipe_context *pipe,
+i915_draw_vbo(struct pipe_context *pipe, const struct pipe_draw_info *info)
-                         struct pipe_resource *indexBuffer,
-                         unsigned indexSize,
-                         int indexBias,
-                         unsigned min_index,
-                         unsigned max_index,
-                         unsigned prim, unsigned start, unsigned count)
 {
    struct i915_context *i915 = i915_context(pipe);
    struct draw_context *draw = i915->draw;
+   void *mapped_indices = NULL;
    unsigned i;
 
    if (i915->dirty)
@@ -71,16 +66,18 @@ i915_draw_range_elements(struct pipe_context *pipe,
    /*
     * Map index buffer, if present
     */
-   if (indexBuffer) {
+   if (info->indexed && i915->index_buffer.buffer) {
-      void *mapped_indexes = i915_buffer(indexBuffer)->data;
+      char *indices = (char *) i915_buffer(i915->index_buffer.buffer)->data;
-      draw_set_mapped_element_buffer_range(draw, indexSize, indexBias,
+      mapped_indices = (void *) (indices + i915->index_buffer.offset);
-                                           min_index,
-                                           max_index,
-                                           mapped_indexes);
-   } else {
-      draw_set_mapped_element_buffer(draw, 0, 0, NULL);
    }
 
+   draw_set_mapped_element_buffer_range(draw, (mapped_indices) ?
+                                        i915->index_buffer.index_size : 0,
+                                        info->index_bias,
+                                        info->min_index,
+                                        info->max_index,
+                                        mapped_indices);
+
 
    draw_set_mapped_constant_buffer(draw, PIPE_SHADER_VERTEX, 0,
                                    i915->current.constants[PIPE_SHADER_VERTEX],
@@ -90,7 +87,7 @@ i915_draw_range_elements(struct pipe_context *pipe,
    /*
     * Do the drawing
     */
-   draw_arrays(i915->draw, prim, start, count);
+   draw_arrays(i915->draw, info->mode, info->start, info->count);
 
    /*
     * unmap vertex/index buffers
@@ -99,32 +96,11 @@ i915_draw_range_elements(struct pipe_context *pipe,
       draw_set_mapped_vertex_buffer(draw, i, NULL);
    }
 
-   if (indexBuffer) {
+   if (mapped_indices) {
       draw_set_mapped_element_buffer(draw, 0, 0, NULL);
    }
 }
 
-static void
-i915_draw_elements(struct pipe_context *pipe,
-                   struct pipe_resource *indexBuffer,
-                   unsigned indexSize, int indexBias,
-                   unsigned prim, unsigned start, unsigned count)
-{
-   i915_draw_range_elements(pipe, indexBuffer,
-                            indexSize, indexBias,
-                            0, 0xffffffff,
-                            prim, start, count);
-}
-
-static void
-i915_draw_arrays(struct pipe_context *pipe,
-                 unsigned prim, unsigned start, unsigned count)
-{
-   i915_draw_elements(pipe, NULL, 0, 0, prim, start, count);
-}
-
-
-
 
 /*
  * Generic context functions
@@ -168,9 +144,7 @@ i915_create_context(struct pipe_screen *screen, void *priv)
 
    i915->base.clear = i915_clear;
 
-   i915->base.draw_arrays = i915_draw_arrays;
+   i915->base.draw_vbo = i915_draw_vbo;
-   i915->base.draw_elements = i915_draw_elements;
-   i915->base.draw_range_elements = i915_draw_range_elements;
 
    /*
     * Create drawing context and plug our rendering stage into it.

src/gallium/drivers/i915/i915_context.h

@@ -221,6 +221,7 @@ struct i915_context
    struct pipe_sampler_view *fragment_sampler_views[PIPE_MAX_SAMPLERS];
    struct pipe_viewport_state viewport;
    struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS];
+   struct pipe_index_buffer index_buffer;
 
    unsigned dirty;
 

src/gallium/drivers/i915/i915_state.c

@@ -812,6 +812,19 @@ i915_delete_vertex_elements_state(struct pipe_context *pipe, void *velems)
    FREE( velems );
 }
 
+static void i915_set_index_buffer(struct pipe_context *pipe,
+                                  const struct pipe_index_buffer *ib)
+{
+   struct i915_context *i915 = i915_context(pipe);
+
+   if (ib)
+      memcpy(&i915->index_buffer, ib, sizeof(i915->index_buffer));
+   else
+      memset(&i915->index_buffer, 0, sizeof(i915->index_buffer));
+
+   /* TODO make this more like a state */
+}
+
 static void
 i915_set_sample_mask(struct pipe_context *pipe,
                      unsigned sample_mask)
@@ -860,4 +873,5 @@ i915_init_state_functions( struct i915_context *i915 )
    i915->base.sampler_view_destroy = i915_sampler_view_destroy;
    i915->base.set_viewport_state = i915_set_viewport_state;
    i915->base.set_vertex_buffers = i915_set_vertex_buffers;
+   i915->base.set_index_buffer = i915_set_index_buffer;
 }

src/gallium/drivers/i965/brw_context.h

@@ -576,6 +576,7 @@ struct brw_context
        */
       struct pipe_resource *index_buffer;
       unsigned index_size;
+      unsigned index_offset;
 
       /* Updates are signalled by PIPE_NEW_INDEX_RANGE:
        */

src/gallium/drivers/i965/brw_draw.c

@@ -142,7 +142,7 @@ static int brw_emit_prim(struct brw_context *brw,
  */
 static int
 try_draw_range_elements(struct brw_context *brw,
-			struct pipe_resource *index_buffer,
+			boolean indexed,
 			unsigned hw_prim, 
 			unsigned start, unsigned count)
 {
@@ -165,7 +165,7 @@ try_draw_range_elements(struct brw_context *brw,
    if (ret)
       return ret;
    
-   ret = brw_emit_prim(brw, start, count, index_buffer != NULL, hw_prim);
+   ret = brw_emit_prim(brw, start, count, indexed, hw_prim);
    if (ret)
       return ret;
 
@@ -177,91 +177,54 @@ try_draw_range_elements(struct brw_context *brw,
 
 
 static void
-brw_draw_range_elements(struct pipe_context *pipe,
+brw_draw_vbo(struct pipe_context *pipe, const struct pipe_draw_info *info)
-			struct pipe_resource *index_buffer,
-			unsigned index_size, int index_bias,
-			unsigned min_index,
-			unsigned max_index,
-			unsigned mode, unsigned start, unsigned count)
 {
    struct brw_context *brw = brw_context(pipe);
    int ret;
    uint32_t hw_prim;
 
-   hw_prim = brw_set_prim(brw, mode);
+   hw_prim = brw_set_prim(brw, info->mode);
 
    if (BRW_DEBUG & DEBUG_PRIMS)
       debug_printf("PRIM: %s start %d count %d index_buffer %p\n",
-                   u_prim_name(mode), start, count, (void *)index_buffer);
+                   u_prim_name(info->mode), info->start, info->count,
+                   (void *) brw->curr.index_buffer);
 
-   assert(index_bias == 0);
+   assert(info->index_bias == 0);
 
-   /* Potentially trigger upload of new index buffer.
+   /* Potentially trigger upload of new index buffer range.
-    *
+    * XXX: do we really care?
-    * XXX: do we need to go through state validation to achieve this?
-    * Could just call upload code directly.
     */
-   if (brw->curr.index_buffer != index_buffer ||
+   if (brw->curr.min_index != info->min_index ||
-       brw->curr.index_size != index_size) {
+       brw->curr.max_index != info->max_index) 
-      pipe_resource_reference( &brw->curr.index_buffer, index_buffer );
-      brw->curr.index_size = index_size;
-      brw->state.dirty.mesa |= PIPE_NEW_INDEX_BUFFER;
-   }
-
-   /* XXX: do we really care?
-    */
-   if (brw->curr.min_index != min_index ||
-       brw->curr.max_index != max_index) 
    { 
-      brw->curr.min_index = min_index;
+      brw->curr.min_index = info->min_index;
-      brw->curr.max_index = max_index;
+      brw->curr.max_index = info->max_index;
       brw->state.dirty.mesa |= PIPE_NEW_INDEX_RANGE;
    }
 
 
    /* Make a first attempt at drawing:
     */
-   ret = try_draw_range_elements(brw, index_buffer, hw_prim, start, count );
+   ret = try_draw_range_elements(brw, info->indexed,
+         hw_prim, info->start, info->count);
 
    /* Otherwise, flush and retry:
     */
    if (ret != 0) {
       brw_context_flush( brw );
-      ret = try_draw_range_elements(brw, index_buffer, hw_prim, start, count );
+      ret = try_draw_range_elements(brw, info->indexed,
+            hw_prim, info->start, info->count);
       assert(ret == 0);
    }
 }
 
-static void
-brw_draw_elements(struct pipe_context *pipe,
-		  struct pipe_resource *index_buffer,
-		  unsigned index_size, int index_bias,
-		  unsigned mode, 
-		  unsigned start, unsigned count)
-{
-   brw_draw_range_elements( pipe, index_buffer,
-                            index_size, index_bias,
-                            0, 0xffffffff,
-                            mode, 
-                            start, count );
-}
-
-static void
-brw_draw_arrays(struct pipe_context *pipe, unsigned mode,
-                     unsigned start, unsigned count)
-{
-   brw_draw_elements(pipe, NULL, 0, 0, mode, start, count);
-}
-
-
 
 boolean brw_draw_init( struct brw_context *brw )
 {
    /* Register our drawing function: 
     */
-   brw->base.draw_arrays = brw_draw_arrays;
+   brw->base.draw_vbo = brw_draw_vbo;
-   brw->base.draw_elements = brw_draw_elements;
-   brw->base.draw_range_elements = brw_draw_range_elements;
 
    /* Create helpers for uploading data in user buffers:
     */

src/gallium/drivers/i965/brw_draw_upload.c

@@ -231,7 +231,7 @@ static int brw_prepare_indices(struct brw_context *brw)
    struct pipe_resource *upload_buf = NULL;
    struct brw_winsys_buffer *bo = NULL;
    GLuint offset;
-   GLuint index_size;
+   GLuint index_size, index_offset;
    GLuint ib_size;
    int ret;
 
@@ -246,13 +246,14 @@ static int brw_prepare_indices(struct brw_context *brw)
 
    ib_size = index_buffer->width0;
    index_size = brw->curr.index_size;
+   index_offset = brw->curr.index_offset;
 
    /* Turn userbuffer into a proper hardware buffer?
     */
    if (brw_buffer_is_user_buffer(index_buffer)) {
 
       ret = u_upload_buffer( brw->vb.upload_index,
-			     0,
+			     index_offset,
 			     ib_size,
 			     index_buffer,
 			     &offset,
@@ -269,7 +270,7 @@ static int brw_prepare_indices(struct brw_context *brw)
    else {
       bo = brw_buffer(index_buffer)->bo;
       ib_size = bo->size;
-      offset = 0;
+      offset = index_offset;
    }
 
    /* Use CMD_3D_PRIM's start_vertex_offset to avoid re-uploading the

src/gallium/drivers/i965/brw_pipe_vertex.c

@@ -274,10 +274,41 @@ static void brw_set_vertex_buffers(struct pipe_context *pipe,
 }
 
 
+static void brw_set_index_buffer(struct pipe_context *pipe,
+                                 const struct pipe_index_buffer *ib)
+{
+   struct brw_context *brw = brw_context(pipe);
+
+   if (ib) {
+      if (brw->curr.index_buffer == ib->buffer &&
+          brw->curr.index_offset == ib->offset &&
+          brw->curr.index_size == ib->index_size)
+         return;
+
+      pipe_resource_reference(&brw->curr.index_buffer, ib->buffer);
+      brw->curr.index_offset = ib->offset;
+      brw->curr.index_size = ib->index_size;
+   }
+   else {
+      if (!brw->curr.index_buffer &&
+          !brw->curr.index_offset &&
+          !brw->curr.index_size)
+         return;
+
+      pipe_resource_reference(&brw->curr.index_buffer, NULL);
+      brw->curr.index_offset = 0;
+      brw->curr.index_size = 0;
+   }
+
+   brw->state.dirty.mesa |= PIPE_NEW_INDEX_BUFFER;
+}
+
+
 void 
 brw_pipe_vertex_init( struct brw_context *brw )
 {
    brw->base.set_vertex_buffers = brw_set_vertex_buffers;
+   brw->base.set_index_buffer = brw_set_index_buffer;
    brw->base.create_vertex_elements_state = brw_create_vertex_elements_state;
    brw->base.bind_vertex_elements_state = brw_bind_vertex_elements_state;
    brw->base.delete_vertex_elements_state = brw_delete_vertex_elements_state;

src/gallium/drivers/identity/id_context.c

@@ -46,68 +46,13 @@ identity_destroy(struct pipe_context *_pipe)
 }
 
 static void
-identity_draw_arrays(struct pipe_context *_pipe,
+identity_draw_vbo(struct pipe_context *_pipe,
-                     unsigned prim,
+                  const struct pipe_draw_info *info)
-                     unsigned start,
-                     unsigned count)
 {
    struct identity_context *id_pipe = identity_context(_pipe);
    struct pipe_context *pipe = id_pipe->pipe;
 
-   pipe->draw_arrays(pipe,
+   pipe->draw_vbo(pipe, info);
-                     prim,
-                     start,
-                     count);
-}
-
-static void
-identity_draw_elements(struct pipe_context *_pipe,
-                       struct pipe_resource *_indexResource,
-                       unsigned indexSize,
-                       int indexBias,
-                       unsigned prim,
-                       unsigned start,
-                       unsigned count)
-{
-   struct identity_context *id_pipe = identity_context(_pipe);
-   struct identity_resource *id_resource = identity_resource(_indexResource);
-   struct pipe_context *pipe = id_pipe->pipe;
-   struct pipe_resource *indexResource = id_resource->resource;
-
-   pipe->draw_elements(pipe,
-                       indexResource,
-                       indexSize,
-                       indexBias,
-                       prim,
-                       start,
-                       count);
-}
-
-static void
-identity_draw_range_elements(struct pipe_context *_pipe,
-                             struct pipe_resource *_indexResource,
-                             unsigned indexSize,
-                             int indexBias,
-                             unsigned minIndex,
-                             unsigned maxIndex,
-                             unsigned mode,
-                             unsigned start,
-                             unsigned count)
-{
-   struct identity_context *id_pipe = identity_context(_pipe);
-   struct identity_resource *id_resource = identity_resource(_indexResource);
-   struct pipe_context *pipe = id_pipe->pipe;
-   struct pipe_resource *indexResource = id_resource->resource;
-
-   pipe->draw_range_elements(pipe,
-                             indexResource,
-                             indexSize,
-                             indexBias,
-                             minIndex,
-                             maxIndex,
-                             mode,
-                             start,
-                             count);
 }
 
 static struct pipe_query *
@@ -611,6 +556,24 @@ identity_set_vertex_buffers(struct pipe_context *_pipe,
                             num_buffers,
                             buffers);
 }
+
+static void
+identity_set_index_buffer(struct pipe_context *_pipe,
+                          const struct pipe_index_buffer *_ib)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+   struct pipe_index_buffer unwrapped_ib, *ib = NULL;
+
+   if (_ib) {
+      unwrapped_ib = *_ib;
+      unwrapped_ib.buffer = identity_resource_unwrap(_ib->buffer);
+      ib = &unwrapped_ib;
+   }
+
+   pipe->set_index_buffer(pipe, ib);
+}
+
 static void
 identity_resource_copy_region(struct pipe_context *_pipe,
                               struct pipe_resource *_dst,
@@ -889,9 +852,7 @@ identity_context_create(struct pipe_screen *_screen, struct pipe_context *pipe)
    id_pipe->base.draw = NULL;
 
    id_pipe->base.destroy = identity_destroy;
-   id_pipe->base.draw_arrays = identity_draw_arrays;
+   id_pipe->base.draw_vbo = identity_draw_vbo;
-   id_pipe->base.draw_elements = identity_draw_elements;
-   id_pipe->base.draw_range_elements = identity_draw_range_elements;
    id_pipe->base.create_query = identity_create_query;
    id_pipe->base.destroy_query = identity_destroy_query;
    id_pipe->base.begin_query = identity_begin_query;
@@ -931,6 +892,7 @@ identity_context_create(struct pipe_screen *_screen, struct pipe_context *pipe)
    id_pipe->base.set_fragment_sampler_views = identity_set_fragment_sampler_views;
    id_pipe->base.set_vertex_sampler_views = identity_set_vertex_sampler_views;
    id_pipe->base.set_vertex_buffers = identity_set_vertex_buffers;
+   id_pipe->base.set_index_buffer = identity_set_index_buffer;
    id_pipe->base.resource_copy_region = identity_resource_copy_region;
    id_pipe->base.clear = identity_clear;
    id_pipe->base.clear_render_target = identity_clear_render_target;

src/gallium/drivers/llvmpipe/SConscript

@@ -1,3 +1,4 @@
+from sys import executable as python_cmd
 import distutils.version
 
 Import('*')
@@ -16,7 +17,7 @@ env.CodeGenerate(
 	target = 'lp_tile_soa.c',
 	script = 'lp_tile_soa.py',
 	source = ['#src/gallium/auxiliary/util/u_format.csv'],
-	command = 'python $SCRIPT $SOURCE > $TARGET'
+	command = python_cmd + ' $SCRIPT $SOURCE > $TARGET'
 )
 
 # XXX: Our dependency scanner only finds depended modules in relative dirs.

src/gallium/drivers/llvmpipe/lp_bld_depth.c

@@ -258,16 +258,16 @@ lp_build_stencil_op_single(struct lp_build_context *bld,
    }
 
    if (stencil->writemask != stencilMax) {
-      /* compute res = (res & mask) | (stencilVals & ~mask) */
+      /* mask &= stencil->writemask */
-      LLVMValueRef mask = lp_build_const_int_vec(type, stencil->writemask);
+      LLVMValueRef writemask = lp_build_const_int_vec(type, stencil->writemask);
-      LLVMValueRef cmask = LLVMBuildNot(bld->builder, mask, "notWritemask");
+      mask = LLVMBuildAnd(bld->builder, mask, writemask, "");
-      LLVMValueRef t1 = LLVMBuildAnd(bld->builder, res, mask, "t1");
+      /* res = (res & mask) | (stencilVals & ~mask) */
-      LLVMValueRef t2 = LLVMBuildAnd(bld->builder, stencilVals, cmask, "t2");
+      res = lp_build_select_bitwise(bld, writemask, res, stencilVals);
-      res = LLVMBuildOr(bld->builder, t1, t2, "t1_or_t2");
+   }
+   else {
+      /* res = mask ? res : stencilVals */
+      res = lp_build_select(bld, mask, res, stencilVals);
    }
-
-   /* only the update the vector elements enabled by 'mask' */
-   res = lp_build_select(bld, mask, res, stencilVals);
 
    return res;
 }
@@ -662,9 +662,9 @@ lp_build_depth_stencil_test(LLVMBuilderRef builder,
          }
 
          /* Mix the old and new Z buffer values.
-          * z_dst[i] = zselectmask[i] ? z_src[i] : z_dst[i]
+          * z_dst[i] = (zselectmask[i] & z_src[i]) | (~zselectmask[i] & z_dst[i])
           */
-         z_dst = lp_build_select(&bld, zselectmask, z_src, z_dst);
+         z_dst = lp_build_select_bitwise(&bld, zselectmask, z_src, z_dst);
       }
 
       if (stencil[0].enabled) {

src/gallium/drivers/llvmpipe/lp_context.c

@@ -46,6 +46,10 @@
 #include "lp_query.h"
 #include "lp_setup.h"
 
+
+DEBUG_GET_ONCE_BOOL_OPTION(lp_no_rast, "LP_NO_RAST", FALSE);
+
+
 static void llvmpipe_destroy( struct pipe_context *pipe )
 {
    struct llvmpipe_context *llvmpipe = llvmpipe_context( pipe );
@@ -130,7 +134,7 @@ llvmpipe_create_context( struct pipe_screen *screen, void *priv )
 
    /* FIXME: devise alternative to draw_texture_samplers */
 
-   if (debug_get_bool_option( "LP_NO_RAST", FALSE ))
+   if (debug_get_option_lp_no_rast())
       llvmpipe->no_rast = TRUE;
 
    llvmpipe->setup = lp_setup_create( &llvmpipe->pipe,

src/gallium/drivers/llvmpipe/lp_context.h

@@ -77,6 +77,7 @@ struct llvmpipe_context {
    struct pipe_sampler_view *vertex_sampler_views[PIPE_MAX_VERTEX_SAMPLERS];
    struct pipe_viewport_state viewport;
    struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS];
+   struct pipe_index_buffer index_buffer;
    struct {
       struct llvmpipe_resource *buffer[PIPE_MAX_SO_BUFFERS];
       int offset[PIPE_MAX_SO_BUFFERS];

src/gallium/drivers/llvmpipe/lp_draw_arrays.c

@@ -49,20 +49,11 @@
  * the drawing to the 'draw' module.
  */
 static void
-llvmpipe_draw_range_elements_instanced(struct pipe_context *pipe,
+llvmpipe_draw_vbo(struct pipe_context *pipe, const struct pipe_draw_info *info)
-                                       struct pipe_resource *indexBuffer,
-                                       unsigned indexSize,
-                                       int indexBias,
-                                       unsigned minIndex,
-                                       unsigned maxIndex,
-                                       unsigned mode,
-                                       unsigned start,
-                                       unsigned count,
-                                       unsigned startInstance,
-                                       unsigned instanceCount)
 {
    struct llvmpipe_context *lp = llvmpipe_context(pipe);
    struct draw_context *draw = lp->draw;
+   void *mapped_indices = NULL;
    unsigned i;
 
    if (lp->dirty)
@@ -77,27 +68,25 @@ llvmpipe_draw_range_elements_instanced(struct pipe_context *pipe,
    }
 
    /* Map index buffer, if present */
-   if (indexBuffer) {
+   if (info->indexed && lp->index_buffer.buffer) {
-      void *mapped_indexes = llvmpipe_resource_data(indexBuffer);
+      char *indices = (char *) llvmpipe_resource_data(lp->index_buffer.buffer);
-      draw_set_mapped_element_buffer_range(draw,
+      mapped_indices = (void *) (indices + lp->index_buffer.offset);
-                                           indexSize,
-                                           indexBias,
-                                           minIndex,
-                                           maxIndex,
-                                           mapped_indexes);
-   }
-   else {
-      /* no index/element buffer */
-      draw_set_mapped_element_buffer_range(draw, 0, 0, start,
-                                           start + count - 1, NULL);
    }
+
+   draw_set_mapped_element_buffer_range(draw, (mapped_indices) ?
+                                        lp->index_buffer.index_size : 0,
+                                        info->index_bias,
+                                        info->min_index,
+                                        info->max_index,
+                                        mapped_indices);
+
    llvmpipe_prepare_vertex_sampling(lp,
                                     lp->num_vertex_sampler_views,
                                     lp->vertex_sampler_views);
 
    /* draw! */
-   draw_arrays_instanced(draw, mode, start, count,
+   draw_arrays_instanced(draw, info->mode, info->start, info->count,
-                         startInstance, instanceCount);
+         info->start_instance, info->instance_count);
 
    /*
     * unmap vertex/index buffers
@@ -105,7 +94,7 @@ llvmpipe_draw_range_elements_instanced(struct pipe_context *pipe,
    for (i = 0; i < lp->num_vertex_buffers; i++) {
       draw_set_mapped_vertex_buffer(draw, i, NULL);
    }
-   if (indexBuffer) {
+   if (mapped_indices) {
       draw_set_mapped_element_buffer(draw, 0, 0, NULL);
    }
    llvmpipe_cleanup_vertex_sampling(lp);
@@ -119,112 +108,8 @@ llvmpipe_draw_range_elements_instanced(struct pipe_context *pipe,
 }
 
 
-static void
-llvmpipe_draw_arrays_instanced(struct pipe_context *pipe,
-                               unsigned mode,
-                               unsigned start,
-                               unsigned count,
-                               unsigned startInstance,
-                               unsigned instanceCount)
-{
-   llvmpipe_draw_range_elements_instanced(pipe,
-                                          NULL, /* no indexBuffer */
-                                          0, 0, /* indexSize, indexBias */
-                                          0, ~0, /* minIndex, maxIndex */
-                                          mode,
-                                          start,
-                                          count,
-                                          startInstance,
-                                          instanceCount);
-}
-
-
-static void
-llvmpipe_draw_elements_instanced(struct pipe_context *pipe,
-                                 struct pipe_resource *indexBuffer,
-                                 unsigned indexSize,
-                                 int indexBias,
-                                 unsigned mode,
-                                 unsigned start,
-                                 unsigned count,
-                                 unsigned startInstance,
-                                 unsigned instanceCount)
-{
-   llvmpipe_draw_range_elements_instanced(pipe,
-                                          indexBuffer,
-                                          indexSize, indexBias,
-                                          0, ~0, /* minIndex, maxIndex */
-                                          mode,
-                                          start,
-                                          count,
-                                          startInstance,
-                                          instanceCount);
-}
-
-
-static void
-llvmpipe_draw_elements(struct pipe_context *pipe,
-                       struct pipe_resource *indexBuffer,
-                       unsigned indexSize,
-                       int indexBias,
-                       unsigned mode,
-                       unsigned start,
-                       unsigned count)
-{
-   llvmpipe_draw_range_elements_instanced(pipe,
-                                          indexBuffer,
-                                          indexSize, indexBias,
-                                          0, 0xffffffff,  /* min, maxIndex */
-                                          mode, start, count,
-                                          0,  /* startInstance */
-                                          1);  /* instanceCount */
-}
-
-
-static void
-llvmpipe_draw_range_elements(struct pipe_context *pipe,
-                             struct pipe_resource *indexBuffer,
-                             unsigned indexSize,
-                             int indexBias,
-                             unsigned min_index,
-                             unsigned max_index,
-                             unsigned mode,
-                             unsigned start,
-                             unsigned count)
-{
-   llvmpipe_draw_range_elements_instanced(pipe,
-                                          indexBuffer,
-                                          indexSize, indexBias,
-                                          min_index, max_index,
-                                          mode, start, count,
-                                          0,  /* startInstance */
-                                          1);  /* instanceCount */
-}
-
-
-static void
-llvmpipe_draw_arrays(struct pipe_context *pipe,
-                     unsigned mode,
-                     unsigned start,
-                     unsigned count)
-{
-   llvmpipe_draw_range_elements_instanced(pipe,
-                                          NULL,  /* indexBuffer */
-                                          0,  /* indexSize */
-                                          0,  /* indexBias */
-                                          0, ~0,  /* min, maxIndex */
-                                          mode, start, count,
-                                          0,  /* startInstance */
-                                          1);  /* instanceCount */
-}
-
-
 void
 llvmpipe_init_draw_funcs(struct llvmpipe_context *llvmpipe)
 {
-   llvmpipe->pipe.draw_arrays = llvmpipe_draw_arrays;
+   llvmpipe->pipe.draw_vbo = llvmpipe_draw_vbo;
-   llvmpipe->pipe.draw_elements = llvmpipe_draw_elements;
-   llvmpipe->pipe.draw_range_elements = llvmpipe_draw_range_elements;
-   llvmpipe->pipe.draw_arrays_instanced = llvmpipe_draw_arrays_instanced;
-   llvmpipe->pipe.draw_elements_instanced = llvmpipe_draw_elements_instanced;
 }

src/gallium/drivers/llvmpipe/lp_rast.c

@@ -43,6 +43,12 @@
 #include "lp_scene.h"
 
 
+#ifdef DEBUG
+int jit_line = 0;
+const struct lp_rast_state *jit_state = NULL;
+#endif
+
+
 /**
  * Begin rasterizing a scene.
  * Called once per scene by one thread.
@@ -368,14 +374,15 @@ lp_rast_store_linear_color( struct lp_rasterizer_task *task,
 
    for (buf = 0; buf < rast->state.nr_cbufs; buf++) {
       struct pipe_surface *cbuf = scene->fb.cbufs[buf];
-      const unsigned face = cbuf->face, level = cbuf->level;
+      const unsigned face_slice = cbuf->face + cbuf->zslice;
+      const unsigned level = cbuf->level;
       struct llvmpipe_resource *lpt = llvmpipe_resource(cbuf->texture);
 
       if (!task->color_tiles[buf])
          continue;
 
       llvmpipe_unswizzle_cbuf_tile(lpt,
-                                   face,
+                                   face_slice,
                                    level,
                                    task->x, task->y,
                                    task->color_tiles[buf]);
@@ -418,6 +425,7 @@ lp_rast_shade_tile(struct lp_rasterizer_task *task,
          depth = lp_rast_get_depth_block_pointer(task, tile_x + x, tile_y + y);
 
          /* run shader on 4x4 block */
+         BEGIN_JIT_CALL(state);
          variant->jit_function[RAST_WHOLE]( &state->jit_context,
                                             tile_x + x, tile_y + y,
                                             inputs->facing,
@@ -428,6 +436,7 @@ lp_rast_shade_tile(struct lp_rasterizer_task *task,
                                             depth,
                                             0xffff,
                                             &task->vis_counter);
+         END_JIT_CALL();
       }
    }
 }
@@ -497,6 +506,7 @@ lp_rast_shade_quads_mask(struct lp_rasterizer_task *task,
    assert(lp_check_alignment(state->jit_context.blend_color, 16));
 
    /* run shader on 4x4 block */
+   BEGIN_JIT_CALL(state);
    variant->jit_function[RAST_EDGE_TEST](&state->jit_context,
                                          x, y,
                                          inputs->facing,
@@ -507,6 +517,7 @@ lp_rast_shade_quads_mask(struct lp_rasterizer_task *task,
                                          depth,
                                          mask,
                                          &task->vis_counter);
+   END_JIT_CALL();
 }
 
 

src/gallium/drivers/llvmpipe/lp_rast_priv.h

@@ -40,6 +40,34 @@
 #include "lp_limits.h"
 
 
+/* If we crash in a jitted function, we can examine jit_line and jit_state
+ * to get some info.  This is not thread-safe, however.
+ */
+#ifdef DEBUG
+
+extern int jit_line;
+extern const struct lp_rast_state *jit_state;
+
+#define BEGIN_JIT_CALL(state) \
+   do { \
+      jit_line = __LINE__; \
+      jit_state = state; \
+   } while (0)
+
+#define END_JIT_CALL() \
+   do { \
+      jit_line = 0; \
+      jit_state = NULL; \
+   } while (0)
+
+#else
+
+#define BEGIN_JIT_CALL(X)
+#define END_JIT_CALL()
+
+#endif
+
+
 struct lp_rasterizer;
 
 
@@ -249,6 +277,7 @@ lp_rast_shade_quads_all( struct lp_rasterizer_task *task,
    depth = lp_rast_get_depth_block_pointer(task, x, y);
 
    /* run shader on 4x4 block */
+   BEGIN_JIT_CALL(state);
    variant->jit_function[RAST_WHOLE]( &state->jit_context,
                                       x, y,
                                       inputs->facing,
@@ -259,6 +288,7 @@ lp_rast_shade_quads_all( struct lp_rasterizer_task *task,
                                       depth,
                                       0xffff,
                                       &task->vis_counter );
+   END_JIT_CALL();
 }
 
 

src/gallium/drivers/llvmpipe/lp_state_fs.c

@@ -75,6 +75,7 @@
 #include "gallivm/lp_bld_type.h"
 #include "gallivm/lp_bld_const.h"
 #include "gallivm/lp_bld_conv.h"
+#include "gallivm/lp_bld_init.h"
 #include "gallivm/lp_bld_intr.h"
 #include "gallivm/lp_bld_logic.h"
 #include "gallivm/lp_bld_tgsi.h"
@@ -676,6 +677,11 @@ generate_fragment(struct llvmpipe_context *lp,
 		     color_ptr);
    }
 
+#ifdef PIPE_ARCH_X86
+   /* Avoid corrupting the FPU stack on 32bit OSes. */
+   lp_build_intrinsic(builder, "llvm.x86.mmx.emms", LLVMVoidType(), NULL, 0);
+#endif
+
    LLVMBuildRetVoid(builder);
 
    LLVMDisposeBuilder(builder);
@@ -710,6 +716,7 @@ generate_fragment(struct llvmpipe_context *lp,
       if (gallivm_debug & GALLIVM_DEBUG_ASM) {
          lp_disassemble(f);
       }
+      lp_func_delete_body(function);
    }
 }
 

src/gallium/drivers/llvmpipe/lp_state_vertex.c

@@ -89,6 +89,19 @@ llvmpipe_set_vertex_buffers(struct pipe_context *pipe,
 }
 
 
+static void
+llvmpipe_set_index_buffer(struct pipe_context *pipe,
+                          const struct pipe_index_buffer *ib)
+{
+   struct llvmpipe_context *llvmpipe = llvmpipe_context(pipe);
+
+   if (ib)
+      memcpy(&llvmpipe->index_buffer, ib, sizeof(llvmpipe->index_buffer));
+   else
+      memset(&llvmpipe->index_buffer, 0, sizeof(llvmpipe->index_buffer));
+
+   /* TODO make this more like a state */
+}
 
 void
 llvmpipe_init_vertex_funcs(struct llvmpipe_context *llvmpipe)
@@ -98,4 +111,5 @@ llvmpipe_init_vertex_funcs(struct llvmpipe_context *llvmpipe)
    llvmpipe->pipe.delete_vertex_elements_state = llvmpipe_delete_vertex_elements_state;
 
    llvmpipe->pipe.set_vertex_buffers = llvmpipe_set_vertex_buffers;
+   llvmpipe->pipe.set_index_buffer = llvmpipe_set_index_buffer;
 }

src/gallium/drivers/llvmpipe/lp_test_blend.c

@@ -37,6 +37,7 @@
  */
 
 
+#include "gallivm/lp_bld_init.h"
 #include "gallivm/lp_bld_type.h"
 #include "gallivm/lp_bld_debug.h"
 #include "lp_bld_blend.h"
@@ -485,8 +486,7 @@ test_one(unsigned verbose,
 {
    LLVMModuleRef module = NULL;
    LLVMValueRef func = NULL;
-   LLVMExecutionEngineRef engine = NULL;
+   LLVMExecutionEngineRef engine = lp_build_engine;
-   LLVMModuleProviderRef provider = NULL;
    LLVMPassManagerRef pass = NULL;
    char *error = NULL;
    blend_test_ptr_t blend_test_ptr;
@@ -510,15 +510,6 @@ test_one(unsigned verbose,
    }
    LLVMDisposeMessage(error);
 
-   provider = LLVMCreateModuleProviderForExistingModule(module);
-   if (LLVMCreateJITCompiler(&engine, provider, 1, &error)) {
-      if(verbose < 1)
-         dump_blend_type(stderr, blend, mode, type);
-      fprintf(stderr, "%s\n", error);
-      LLVMDisposeMessage(error);
-      abort();
-   }
-
 #if 0
    pass = LLVMCreatePassManager();
    LLVMAddTargetData(LLVMGetExecutionEngineTargetData(engine), pass);
@@ -735,7 +726,6 @@ test_one(unsigned verbose,
 
    LLVMFreeMachineCodeForFunction(engine, func);
 
-   LLVMDisposeExecutionEngine(engine);
    if(pass)
       LLVMDisposePassManager(pass);