Merge commit 'origin/master' into gallium-sampler-view

Conflicts:
	src/gallium/drivers/nv30/nv30_context.h
	src/gallium/drivers/nv30/nv30_state.c
	src/gallium/drivers/nv40/nv40_context.h
	src/gallium/drivers/nv40/nv40_state.c
	src/gallium/drivers/r300/r300_emit.c

150 files changed, 6852 insertions, 10110 deletions

diff --git a/configs/autoconf.in b/configs/autoconf.in
index bf34f3bffa..f50fb7dd09 100644
--- a/configs/autoconf.in
+++ b/configs/autoconf.in
@@ -24,6 +24,8 @@ RADEON_CFLAGS = @RADEON_CFLAGS@
 RADEON_LDFLAGS = @RADEON_LDFLAGS@
 INTEL_LIBS = @INTEL_LIBS@
 INTEL_CFLAGS = @INTEL_CFLAGS@
+X_LIBS = @X_LIBS@
+X_CFLAGS = @X_CFLAGS@
 
 # Assembler
 MESA_ASM_SOURCES = @MESA_ASM_SOURCES@
diff --git a/configure.ac b/configure.ac
index ed47f428c9..0f51097ef6 100644
--- a/configure.ac
+++ b/configure.ac
@@ -547,7 +547,9 @@ else
     x11_pkgconfig=no
 fi
 dnl Use the autoconf macro if no pkg-config files
-if test "$x11_pkgconfig" = no; then
+if test "$x11_pkgconfig" = yes; then
+    PKG_CHECK_MODULES([X], [x11])
+else
     AC_PATH_XTRA
 fi
 
@@ -1358,7 +1360,7 @@ AC_ARG_ENABLE([gallium-nouveau],
     [enable_gallium_nouveau=no])
 if test "x$enable_gallium_nouveau" = xyes; then
     GALLIUM_WINSYS_DRM_DIRS="$GALLIUM_WINSYS_DRM_DIRS nouveau"
-    GALLIUM_DRIVERS_DIRS="$GALLIUM_DRIVERS_DIRS nouveau nv30 nv40 nv50"
+    GALLIUM_DRIVERS_DIRS="$GALLIUM_DRIVERS_DIRS nouveau nvfx nv50"
 fi
 
 dnl
diff --git a/progs/egl/Makefile b/progs/egl/Makefile
index 25de6e1f70..5f51104fed 100644
--- a/progs/egl/Makefile
+++ b/progs/egl/Makefile
@@ -57,13 +57,13 @@ peglgears: peglgears.o $(HEADERS) $(LIB_DEP)
 	$(CC) $(CFLAGS) $(LDFLAGS) -o $@ $< $(LIBS) $(LIBDRM_LIB) -lm
 
 xeglgears: xeglgears.o $(HEADERS) $(LIB_DEP)
-	$(CC) $(CFLAGS) $(LDFLAGS) -o $@ $< $(LIBS) -lm -L$(libdir) -lX11
+	$(CC) $(CFLAGS) $(LDFLAGS) -o $@ $< $(LIBS) -lm $(X_LIBS)
 
 xeglthreads: xeglthreads.o $(HEADERS) $(LIB_DEP)
-	$(CC) $(CFLAGS) $(LDFLAGS) -o $@ $< $(LIBS) -lm -L$(libdir) -lX11
+	$(CC) $(CFLAGS) $(LDFLAGS) -o $@ $< $(LIBS) -lm $(X_LIBS)
 
 xegl_tri: xegl_tri.o $(HEADERS) $(LIB_DEP)
-	$(CC) $(CFLAGS) $(LDFLAGS) -o $@ $< $(LIBS) -lm -L$(libdir) -lX11
+	$(CC) $(CFLAGS) $(LDFLAGS) -o $@ $< $(LIBS) -lm $(X_LIBS)
 
 clean:
 	-rm -f *.o *~
diff --git a/progs/xdemos/Makefile b/progs/xdemos/Makefile
index e87d55d011..f81aafe00f 100644
--- a/progs/xdemos/Makefile
+++ b/progs/xdemos/Makefile
@@ -9,9 +9,9 @@ INCDIR = $(TOP)/include
 LIB_DEP = $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME)
 
 # Add X11 and pthread libs to satisfy GNU gold.
-APP_LIB_DEPS += -lX11 -lpthread
+APP_LIB_DEPS += $(X_LIBS) -lpthread
 
-LIBS = -L$(TOP)/$(LIB_DIR) -l$(GL_LIB) -L$(libdir) $(APP_LIB_DEPS)
+LIBS = -L$(TOP)/$(LIB_DIR) -l$(GL_LIB) $(APP_LIB_DEPS)
 
 PROGS = \
 	corender \
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_sample.c b/src/gallium/auxiliary/gallivm/lp_bld_sample.c
index 543fd5fea3..2f74aa5e00 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_sample.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_sample.c
@@ -84,8 +84,12 @@ lp_sampler_static_state(struct lp_sampler_static_state *state,
    state->wrap_t            = sampler->wrap_t;
    state->wrap_r            = sampler->wrap_r;
    state->min_img_filter    = sampler->min_img_filter;
-   state->min_mip_filter    = sampler->min_mip_filter;
    state->mag_img_filter    = sampler->mag_img_filter;
+   if (texture->last_level) {
+      state->min_mip_filter = sampler->min_mip_filter;
+   } else {
+      state->min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
+   }
 
    state->compare_mode      = sampler->compare_mode;
    if (sampler->compare_mode != PIPE_TEX_COMPARE_NONE) {
diff --git a/src/gallium/drivers/llvmpipe/lp_flush.c b/src/gallium/drivers/llvmpipe/lp_flush.c
index 1b4e889935..636d72a9bb 100644
--- a/src/gallium/drivers/llvmpipe/lp_flush.c
+++ b/src/gallium/drivers/llvmpipe/lp_flush.c
@@ -92,3 +92,68 @@ llvmpipe_flush( struct pipe_context *pipe,
 #endif
 }
 
+
+/**
+ * Flush context if necessary.
+ *
+ * TODO: move this logic to an auxiliary library?
+ *
+ * FIXME: We must implement DISCARD/DONTBLOCK/UNSYNCHRONIZED/etc for
+ * textures to avoid blocking.
+ */
+boolean
+llvmpipe_flush_texture(struct pipe_context *pipe,
+                       struct pipe_texture *texture,
+                       unsigned face,
+                       unsigned level,
+                       unsigned flush_flags,
+                       boolean read_only,
+                       boolean cpu_access,
+                       boolean do_not_flush)
+{
+   struct pipe_fence_handle *last_fence = NULL;
+   unsigned referenced;
+
+   referenced = pipe->is_texture_referenced(pipe, texture, face, level);
+
+   if ((referenced & PIPE_REFERENCED_FOR_WRITE) ||
+       ((referenced & PIPE_REFERENCED_FOR_READ) && !read_only)) {
+
+      if (do_not_flush)
+         return FALSE;
+
+      /*
+       * TODO: The semantics of these flush flags are too obtuse. They should
+       * disappear and the pipe driver should just ensure that all visible
+       * side-effects happen when they need to happen.
+       */
+      if (referenced & PIPE_REFERENCED_FOR_WRITE)
+         flush_flags |= PIPE_FLUSH_RENDER_CACHE;
+
+      if (referenced & PIPE_REFERENCED_FOR_READ)
+         flush_flags |= PIPE_FLUSH_TEXTURE_CACHE;
+
+      if (cpu_access) {
+         /*
+          * Flush and wait.
+          */
+
+         struct pipe_fence_handle *fence = NULL;
+
+         pipe->flush(pipe, flush_flags, &fence);
+
+         if (last_fence) {
+            pipe->screen->fence_finish(pipe->screen, fence, 0);
+            pipe->screen->fence_reference(pipe->screen, &fence, NULL);
+         }
+      } else {
+         /*
+          * Just flush.
+          */
+
+         pipe->flush(pipe, flush_flags, NULL);
+      }
+   }
+
+   return TRUE;
+}
diff --git a/src/gallium/drivers/llvmpipe/lp_flush.h b/src/gallium/drivers/llvmpipe/lp_flush.h
index 10b2b52583..e13f57ccec 100644
--- a/src/gallium/drivers/llvmpipe/lp_flush.h
+++ b/src/gallium/drivers/llvmpipe/lp_flush.h
@@ -28,10 +28,22 @@
 #ifndef LP_FLUSH_H
 #define LP_FLUSH_H
 
+#include "pipe/p_compiler.h"
+
 struct pipe_context;
 struct pipe_fence_handle;
 
 void llvmpipe_flush(struct pipe_context *pipe, unsigned flags,
                     struct pipe_fence_handle **fence);
 
+boolean
+llvmpipe_flush_texture(struct pipe_context *pipe,
+                       struct pipe_texture *texture,
+                       unsigned face,
+                       unsigned level,
+                       unsigned flush_flags,
+                       boolean read_only,
+                       boolean cpu_access,
+                       boolean do_not_flush);
+
 #endif
diff --git a/src/gallium/drivers/llvmpipe/lp_rast.c b/src/gallium/drivers/llvmpipe/lp_rast.c
index dd9a8e8856..81ea11a16b 100644
--- a/src/gallium/drivers/llvmpipe/lp_rast.c
+++ b/src/gallium/drivers/llvmpipe/lp_rast.c
@@ -62,18 +62,20 @@ lp_rast_begin( struct lp_rasterizer *rast,
    rast->state.write_color = write_color;
    
    for (i = 0; i < rast->state.nr_cbufs; i++) {
+      struct pipe_surface *cbuf = scene->fb.cbufs[i];
       rast->cbuf[i].map = scene->cbuf_map[i];
-      rast->cbuf[i].format = scene->cbuf_transfer[i]->texture->format;
-      rast->cbuf[i].width = scene->cbuf_transfer[i]->width;
-      rast->cbuf[i].height = scene->cbuf_transfer[i]->height;
-      rast->cbuf[i].stride = scene->cbuf_transfer[i]->stride;
+      rast->cbuf[i].format = cbuf->texture->format;
+      rast->cbuf[i].width = cbuf->width;
+      rast->cbuf[i].height = cbuf->height;
+      rast->cbuf[i].stride = llvmpipe_texture_stride(cbuf->texture, cbuf->level);
    }
 
    if (write_zstencil) {
+      struct pipe_surface *zsbuf = scene->fb.zsbuf;
       rast->zsbuf.map = scene->zsbuf_map;
-      rast->zsbuf.stride = scene->zsbuf_transfer->stride;
+      rast->zsbuf.stride = llvmpipe_texture_stride(zsbuf->texture, zsbuf->level);
       rast->zsbuf.blocksize = 
-         util_format_get_blocksize(scene->zsbuf_transfer->texture->format);
+         util_format_get_blocksize(zsbuf->texture->format);
    }
 
    lp_scene_bin_iter_begin( scene );
diff --git a/src/gallium/drivers/llvmpipe/lp_scene.c b/src/gallium/drivers/llvmpipe/lp_scene.c
index 505cb21503..681ce674d4 100644
--- a/src/gallium/drivers/llvmpipe/lp_scene.c
+++ b/src/gallium/drivers/llvmpipe/lp_scene.c
@@ -397,7 +397,6 @@ end:
 static boolean
 lp_scene_map_buffers( struct lp_scene *scene )
 {
-   struct pipe_context *pipe = scene->pipe;
    struct pipe_surface *cbuf, *zsbuf;
    int i;
 
@@ -409,20 +408,10 @@ lp_scene_map_buffers( struct lp_scene *scene )
    for (i = 0; i < scene->fb.nr_cbufs; i++) {
       cbuf = scene->fb.cbufs[i];
       if (cbuf) {
-	 scene->cbuf_transfer[i] = pipe->get_tex_transfer(pipe,
-                                                          cbuf->texture,
-                                                          cbuf->face,
-                                                          cbuf->level,
-                                                          cbuf->zslice,
-                                                          PIPE_TRANSFER_READ_WRITE,
-                                                          0, 0,
-                                                          cbuf->width, 
-                                                          cbuf->height);
-	 if (!scene->cbuf_transfer[i])
-	    goto fail;
-
-	 scene->cbuf_map[i] = pipe->transfer_map(pipe, 
-                                                 scene->cbuf_transfer[i]);
+	 scene->cbuf_map[i] = llvmpipe_texture_map(cbuf->texture,
+	                                           cbuf->face,
+                                                   cbuf->level,
+                                                   cbuf->zslice);
 	 if (!scene->cbuf_map[i])
 	    goto fail;
       }
@@ -432,20 +421,10 @@ lp_scene_map_buffers( struct lp_scene *scene )
     */
    zsbuf = scene->fb.zsbuf;
    if (zsbuf) {
-      scene->zsbuf_transfer = pipe->get_tex_transfer(pipe,
-                                                       zsbuf->texture,
-                                                       zsbuf->face,
-                                                       zsbuf->level,
-                                                       zsbuf->zslice,
-                                                       PIPE_TRANSFER_READ_WRITE,
-                                                       0, 0,
-                                                       zsbuf->width,
-                                                       zsbuf->height);
-      if (!scene->zsbuf_transfer)
-         goto fail;
-
-      scene->zsbuf_map = pipe->transfer_map(pipe, 
-                                              scene->zsbuf_transfer);
+      scene->zsbuf_map = llvmpipe_texture_map(zsbuf->texture,
+                                              zsbuf->face,
+                                              zsbuf->level,
+                                              zsbuf->zslice);
       if (!scene->zsbuf_map)
 	 goto fail;
    }
@@ -469,28 +448,27 @@ fail:
 static void
 lp_scene_unmap_buffers( struct lp_scene *scene )
 {
-   struct pipe_context *pipe = scene->pipe;
    unsigned i;
 
    for (i = 0; i < scene->fb.nr_cbufs; i++) {
-      if (scene->cbuf_map[i]) 
-	 pipe->transfer_unmap(pipe, scene->cbuf_transfer[i]);
-
-      if (scene->cbuf_transfer[i])
-	 pipe->tex_transfer_destroy(pipe, scene->cbuf_transfer[i]);
-
-      scene->cbuf_transfer[i] = NULL;
-      scene->cbuf_map[i] = NULL;
+      if (scene->cbuf_map[i]) {
+         struct pipe_surface *cbuf = scene->fb.cbufs[i];
+         llvmpipe_texture_unmap(cbuf->texture,
+                                cbuf->face,
+                                cbuf->level,
+                                cbuf->zslice);
+         scene->cbuf_map[i] = NULL;
+      }
    }
 
-   if (scene->zsbuf_map) 
-      pipe->transfer_unmap(pipe, scene->zsbuf_transfer);
-
-   if (scene->zsbuf_transfer)
-      pipe->tex_transfer_destroy(pipe, scene->zsbuf_transfer);
-
-   scene->zsbuf_transfer = NULL;
-   scene->zsbuf_map = NULL;
+   if (scene->zsbuf_map) {
+      struct pipe_surface *zsbuf = scene->fb.zsbuf;
+      llvmpipe_texture_unmap(zsbuf->texture,
+                             zsbuf->face,
+                             zsbuf->level,
+                             zsbuf->zslice);
+      scene->zsbuf_map = NULL;
+   }
 
    util_unreference_framebuffer_state( &scene->fb );
 }
diff --git a/src/gallium/drivers/llvmpipe/lp_scene.h b/src/gallium/drivers/llvmpipe/lp_scene.h
index 739ac22908..b602b1e8a0 100644
--- a/src/gallium/drivers/llvmpipe/lp_scene.h
+++ b/src/gallium/drivers/llvmpipe/lp_scene.h
@@ -114,8 +114,6 @@ struct texture_ref {
  */
 struct lp_scene {
    struct pipe_context *pipe;
-   struct pipe_transfer *cbuf_transfer[PIPE_MAX_COLOR_BUFS];
-   struct pipe_transfer *zsbuf_transfer;
 
    /* Scene's buffers are mapped at the time the scene is enqueued:
     */
diff --git a/src/gallium/drivers/llvmpipe/lp_surface.c b/src/gallium/drivers/llvmpipe/lp_surface.c
index 6110b0a193..ca3d62c361 100644
--- a/src/gallium/drivers/llvmpipe/lp_surface.c
+++ b/src/gallium/drivers/llvmpipe/lp_surface.c
@@ -27,6 +27,7 @@
 
 #include "util/u_rect.h"
 #include "lp_context.h"
+#include "lp_flush.h"
 #include "lp_surface.h"
 
 
@@ -36,6 +37,20 @@ lp_surface_copy(struct pipe_context *pipe,
                 struct pipe_surface *src, unsigned srcx, unsigned srcy,
                 unsigned width, unsigned height)
 {
+   llvmpipe_flush_texture(pipe,
+                          dest->texture, dest->face, dest->level,
+                          0, /* flush_flags */
+                          FALSE, /* read_only */
+                          FALSE, /* cpu_access */
+                          FALSE); /* do_not_flush */
+
+   llvmpipe_flush_texture(pipe,
+                          src->texture, src->face, src->level,
+                          0, /* flush_flags */
+                          TRUE, /* read_only */
+                          FALSE, /* cpu_access */
+                          FALSE); /* do_not_flush */
+
    util_surface_copy(pipe, FALSE,
                      dest, destx, desty,
                      src, srcx, srcy,
diff --git a/src/gallium/drivers/llvmpipe/lp_texture.c b/src/gallium/drivers/llvmpipe/lp_texture.c
index f2c6dbd088..9a85a42897 100644
--- a/src/gallium/drivers/llvmpipe/lp_texture.c
+++ b/src/gallium/drivers/llvmpipe/lp_texture.c
@@ -40,6 +40,7 @@
 
 #include "lp_context.h"
 #include "lp_screen.h"
+#include "lp_flush.h"
 #include "lp_texture.h"
 #include "lp_tile_size.h"
 #include "state_tracker/sw_winsys.h"
@@ -163,6 +164,92 @@ llvmpipe_texture_destroy(struct pipe_texture *pt)
 }
 
 
+/**
+ * Map a texture. Without any synchronization.
+ */
+void *
+llvmpipe_texture_map(struct pipe_texture *texture,
+                     unsigned face,
+                     unsigned level,
+                     unsigned zslice)
+{
+   struct llvmpipe_texture *lpt = llvmpipe_texture(texture);
+   uint8_t *map;
+
+   if (lpt->dt) {
+      /* display target */
+      struct llvmpipe_screen *screen = llvmpipe_screen(texture->screen);
+      struct sw_winsys *winsys = screen->winsys;
+      const unsigned usage = PIPE_BUFFER_USAGE_CPU_READ_WRITE;
+
+      assert(face == 0);
+      assert(level == 0);
+      assert(zslice == 0);
+
+      /* FIXME: keep map count? */
+      map = winsys->displaytarget_map(winsys, lpt->dt, usage);
+   }
+   else {
+      /* regular texture */
+      unsigned offset;
+      unsigned stride;
+
+      map = lpt->data;
+
+      assert(level < LP_MAX_TEXTURE_2D_LEVELS);
+
+      offset = lpt->level_offset[level];
+      stride = lpt->stride[level];
+
+      /* XXX shouldn't that rather be
+         tex_height = align(u_minify(texture->height0, level), 2)
+         to account for alignment done in llvmpipe_texture_layout ?
+      */
+      if (texture->target == PIPE_TEXTURE_CUBE) {
+         unsigned tex_height = u_minify(texture->height0, level);
+         offset += face *  util_format_get_nblocksy(texture->format, tex_height) * stride;
+      }
+      else if (texture->target == PIPE_TEXTURE_3D) {
+         unsigned tex_height = u_minify(texture->height0, level);
+         offset += zslice * util_format_get_nblocksy(texture->format, tex_height) * stride;
+      }
+      else {
+         assert(face == 0);
+         assert(zslice == 0);
+      }
+
+      map += offset;
+   }
+
+   return map;
+}
+
+
+/**
+ * Unmap a texture. Without any synchronization.
+ */
+void
+llvmpipe_texture_unmap(struct pipe_texture *texture,
+                       unsigned face,
+                       unsigned level,
+                       unsigned zslice)
+{
+   struct llvmpipe_texture *lpt = llvmpipe_texture(texture);
+
+   if (lpt->dt) {
+      /* display target */
+      struct llvmpipe_screen *lp_screen = llvmpipe_screen(texture->screen);
+      struct sw_winsys *winsys = lp_screen->winsys;
+
+      assert(face == 0);
+      assert(level == 0);
+      assert(zslice == 0);
+
+      winsys->displaytarget_unmap(winsys, lpt->dt);
+   }
+}
+
+
 static struct pipe_surface *
 llvmpipe_get_tex_surface(struct pipe_screen *screen,
                          struct pipe_texture *pt,
@@ -181,7 +268,6 @@ llvmpipe_get_tex_surface(struct pipe_screen *screen,
       ps->format = pt->format;
       ps->width = u_minify(pt->width0, level);
       ps->height = u_minify(pt->height0, level);
-      ps->offset = lpt->level_offset[level];
       ps->usage = usage;
 
       /* Because we are llvmpipe, anything that the state tracker
@@ -207,23 +293,6 @@ llvmpipe_get_tex_surface(struct pipe_screen *screen,
       ps->face = face;
       ps->level = level;
       ps->zslice = zslice;
-
-      /* XXX shouldn't that rather be
-         tex_height = align(ps->height, 2);
-         to account for alignment done in llvmpipe_texture_layout ?
-      */
-      if (pt->target == PIPE_TEXTURE_CUBE) {
-         unsigned tex_height = ps->height;
-         ps->offset += face * util_format_get_nblocksy(pt->format, tex_height) * lpt->stride[level];
-      }
-      else if (pt->target == PIPE_TEXTURE_3D) {
-         unsigned tex_height = ps->height;
-         ps->offset += zslice * util_format_get_nblocksy(pt->format, tex_height) * lpt->stride[level];
-      }
-      else {
-         assert(face == 0);
-         assert(zslice == 0);
-      }
    }
    return ps;
 }
@@ -269,24 +338,6 @@ llvmpipe_get_tex_transfer(struct pipe_context *pipe,
       pt->level = level;
       pt->zslice = zslice;
 
-      lpt->offset = lptex->level_offset[level];
-
-      /* XXX shouldn't that rather be
-         tex_height = align(u_minify(texture->height0, level), 2)
-         to account for alignment done in llvmpipe_texture_layout ?
-      */
-      if (texture->target == PIPE_TEXTURE_CUBE) {
-         unsigned tex_height = u_minify(texture->height0, level);
-         lpt->offset += face *  util_format_get_nblocksy(texture->format, tex_height) * pt->stride;
-      }
-      else if (texture->target == PIPE_TEXTURE_3D) {
-         unsigned tex_height = u_minify(texture->height0, level);
-         lpt->offset += zslice * util_format_get_nblocksy(texture->format, tex_height) * pt->stride;
-      }
-      else {
-         assert(face == 0);
-         assert(zslice == 0);
-      }
       return pt;
    }
    return NULL;
@@ -312,7 +363,7 @@ llvmpipe_transfer_map( struct pipe_context *pipe,
                        struct pipe_transfer *transfer )
 {
    struct llvmpipe_screen *screen = llvmpipe_screen(pipe->screen);
-   ubyte *map, *xfer_map;
+   ubyte *map;
    struct llvmpipe_texture *lpt;
    enum pipe_format format;
 
@@ -320,34 +371,34 @@ llvmpipe_transfer_map( struct pipe_context *pipe,
    lpt = llvmpipe_texture(transfer->texture);
    format = lpt->base.format;
 
-   if (lpt->dt) {
-      /* display target */
-      struct sw_winsys *winsys = screen->winsys;
+   /*
+    * Transfers, like other pipe operations, must happen in order, so flush the
+    * context if necessary.
+    */
+   llvmpipe_flush_texture(pipe,
+                          transfer->texture, transfer->face, transfer->level,
+                          0, /* flush_flags */
+                          !(transfer->usage & PIPE_TRANSFER_WRITE), /* read_only */
+                          TRUE, /* cpu_access */
+                          FALSE); /* do_not_flush */
 
-      map = winsys->displaytarget_map(winsys, lpt->dt,
-                                      pipe_transfer_buffer_flags(transfer));
-      if (map == NULL)
-         return NULL;
-   }
-   else {
-      /* regular texture */
-      map = lpt->data;
-   }
+   map = llvmpipe_texture_map(transfer->texture,
+                              transfer->face, transfer->level, transfer->zslice);
 
    /* May want to different things here depending on read/write nature
     * of the map:
     */
-   if (transfer->texture && (transfer->usage & PIPE_TRANSFER_WRITE)) {
+   if (transfer->usage & PIPE_TRANSFER_WRITE) {
       /* Do something to notify sharing contexts of a texture change.
        */
       screen->timestamp++;
    }
    
-   xfer_map = map + llvmpipe_transfer(transfer)->offset +
+   map +=
       transfer->y / util_format_get_blockheight(format) * transfer->stride +
       transfer->x / util_format_get_blockwidth(format) * util_format_get_blocksize(format);
-   /*printf("map = %p  xfer map = %p\n", map, xfer_map);*/
-   return xfer_map;
+
+   return map;
 }
 
 
@@ -355,17 +406,10 @@ static void
 llvmpipe_transfer_unmap(struct pipe_context *pipe,
                         struct pipe_transfer *transfer)
 {
-   struct llvmpipe_screen *lp_screen = llvmpipe_screen(pipe->screen);
-   struct llvmpipe_texture *lpt;
-
    assert(transfer->texture);
-   lpt = llvmpipe_texture(transfer->texture);
 
-   if (lpt->dt) {
-      /* display target */
-      struct sw_winsys *winsys = lp_screen->winsys;
-      winsys->displaytarget_unmap(winsys, lpt->dt);
-   }
+   llvmpipe_texture_unmap(transfer->texture,
+                          transfer->face, transfer->level, transfer->zslice);
 }
 
 
diff --git a/src/gallium/drivers/llvmpipe/lp_texture.h b/src/gallium/drivers/llvmpipe/lp_texture.h
index 94b667abf3..2350c26e4f 100644
--- a/src/gallium/drivers/llvmpipe/lp_texture.h
+++ b/src/gallium/drivers/llvmpipe/lp_texture.h
@@ -95,6 +95,28 @@ llvmpipe_transfer(struct pipe_transfer *pt)
 }
 
 
+static INLINE unsigned
+llvmpipe_texture_stride(struct pipe_texture *texture,
+                        unsigned level)
+{
+   struct llvmpipe_texture *lpt = llvmpipe_texture(texture);
+   assert(level < LP_MAX_TEXTURE_2D_LEVELS);
+   return lpt->stride[level];
+}
+
+
+void *
+llvmpipe_texture_map(struct pipe_texture *texture,
+                     unsigned face,
+                     unsigned level,
+                     unsigned zslice);
+
+void
+llvmpipe_texture_unmap(struct pipe_texture *texture,
+                       unsigned face,
+                       unsigned level,
+                       unsigned zslice);
+
 extern void
 llvmpipe_init_screen_texture_funcs(struct pipe_screen *screen);
 
diff --git a/src/gallium/drivers/nouveau/Makefile b/src/gallium/drivers/nouveau/Makefile
index 0e02680bc6..0cb66041d5 100644
--- a/src/gallium/drivers/nouveau/Makefile
+++ b/src/gallium/drivers/nouveau/Makefile
@@ -4,7 +4,6 @@ include $(TOP)/configs/current
 LIBNAME = nouveau
 
 C_SOURCES = nouveau_screen.c \
-	    nouveau_context.c \
-	    nv04_surface_2d.c
+	    nouveau_context.c
 
 include ../../Makefile.template
diff --git a/src/gallium/drivers/nouveau/nouveau_util.h b/src/gallium/drivers/nouveau/nouveau_util.h
index 7f16e31c3f..ab7761a31d 100644
--- a/src/gallium/drivers/nouveau/nouveau_util.h
+++ b/src/gallium/drivers/nouveau/nouveau_util.h
@@ -33,7 +33,7 @@ nouveau_vbuf_split(unsigned remaining, unsigned overhead, unsigned vpp,
 		max = max - (max % 3);
 		break;
 	case PIPE_PRIM_QUADS:
-		max = max & 3;
+		max = max & ~3;
 		break;
 	case PIPE_PRIM_LINE_LOOP:
 	case PIPE_PRIM_LINE_STRIP:
diff --git a/src/gallium/drivers/nouveau/nouveau_winsys.h b/src/gallium/drivers/nouveau/nouveau_winsys.h
index af9ddd558c..bed014b9ce 100644
--- a/src/gallium/drivers/nouveau/nouveau_winsys.h
+++ b/src/gallium/drivers/nouveau/nouveau_winsys.h
@@ -27,10 +27,7 @@
 #define NOUVEAU_BUFFER_USAGE_NO_RENDER (1 << 19)
 
 extern struct pipe_screen *
-nv30_screen_create(struct pipe_winsys *ws, struct nouveau_device *);
-
-extern struct pipe_screen *
-nv40_screen_create(struct pipe_winsys *ws, struct nouveau_device *);
+nvfx_screen_create(struct pipe_winsys *ws, struct nouveau_device *);
 
 extern struct pipe_screen *
 nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *);
diff --git a/src/gallium/drivers/nv30/Makefile b/src/gallium/drivers/nv30/Makefile
deleted file mode 100644
index 364c80d8f3..0000000000
--- a/src/gallium/drivers/nv30/Makefile
+++ /dev/null
@@ -1,29 +0,0 @@
-TOP = ../../../..
-include $(TOP)/configs/current
-
-LIBNAME = nv30
-
-C_SOURCES = \
-	nv30_clear.c \
-	nv30_context.c \
-	nv30_draw.c \
-	nv30_fragprog.c \
-	nv30_fragtex.c \
-	nv30_miptree.c \
-	nv30_query.c \
-	nv30_screen.c \
-	nv30_state.c \
-	nv30_state_blend.c \
-	nv30_state_emit.c \
-	nv30_state_fb.c \
-	nv30_state_rasterizer.c \
-	nv30_state_scissor.c \
-	nv30_state_stipple.c \
-	nv30_state_viewport.c \
-	nv30_state_zsa.c \
-	nv30_surface.c \
-	nv30_transfer.c \
-	nv30_vbo.c \
-	nv30_vertprog.c
-
-include ../../Makefile.template
diff --git a/src/gallium/drivers/nv30/nv30_clear.c b/src/gallium/drivers/nv30/nv30_clear.c
deleted file mode 100644
index c4ba926664..0000000000
--- a/src/gallium/drivers/nv30/nv30_clear.c
+++ /dev/null
@@ -1,14 +0,0 @@
-#include "pipe/p_context.h"
-#include "pipe/p_defines.h"
-#include "pipe/p_state.h"
-#include "util/u_clear.h"
-
-#include "nv30_context.h"
-
-void
-nv30_clear(struct pipe_context *pipe, unsigned buffers,
-           const float *rgba, double depth, unsigned stencil)
-{
-	util_clear(pipe, &nv30_context(pipe)->framebuffer, buffers, rgba, depth,
-		   stencil);
-}
diff --git a/src/gallium/drivers/nv30/nv30_context.c b/src/gallium/drivers/nv30/nv30_context.c
deleted file mode 100644
index 825c167b01..0000000000
--- a/src/gallium/drivers/nv30/nv30_context.c
+++ /dev/null
@@ -1,88 +0,0 @@
-#include "draw/draw_context.h"
-#include "pipe/p_defines.h"
-
-#include "nv30_context.h"
-#include "nv30_screen.h"
-
-static void
-nv30_flush(struct pipe_context *pipe, unsigned flags,
-	   struct pipe_fence_handle **fence)
-{
-	struct nv30_context *nv30 = nv30_context(pipe);
-	struct nv30_screen *screen = nv30->screen;
-	struct nouveau_channel *chan = screen->base.channel;
-	struct nouveau_grobj *rankine = screen->rankine;
-
-	if (flags & PIPE_FLUSH_TEXTURE_CACHE) {
-		BEGIN_RING(chan, rankine, 0x1fd8, 1);
-		OUT_RING  (chan, 2);
-		BEGIN_RING(chan, rankine, 0x1fd8, 1);
-		OUT_RING  (chan, 1);
-	}
-
-	FIRE_RING(chan);
-	if (fence)
-		*fence = NULL;
-}
-
-static void
-nv30_destroy(struct pipe_context *pipe)
-{
-	struct nv30_context *nv30 = nv30_context(pipe);
-	unsigned i;
-
-	for (i = 0; i < NV30_STATE_MAX; i++) {
-		if (nv30->state.hw[i])
-			so_ref(NULL, &nv30->state.hw[i]);
-	}
-
-	if (nv30->draw)
-		draw_destroy(nv30->draw);
-	FREE(nv30);
-}
-
-struct pipe_context *
-nv30_create(struct pipe_screen *pscreen, void *priv)
-{
-	struct nv30_screen *screen = nv30_screen(pscreen);
-	struct pipe_winsys *ws = pscreen->winsys;
-	struct nv30_context *nv30;
-	struct nouveau_winsys *nvws = screen->nvws;
-
-	nv30 = CALLOC(1, sizeof(struct nv30_context));
-	if (!nv30)
-		return NULL;
-	nv30->screen = screen;
-
-	nv30->nvws = nvws;
-
-	nv30->pipe.winsys = ws;
-	nv30->pipe.screen = pscreen;
-	nv30->pipe.priv = priv;
-	nv30->pipe.destroy = nv30_destroy;
-	nv30->pipe.draw_arrays = nv30_draw_arrays;
-	nv30->pipe.draw_elements = nv30_draw_elements;
-	nv30->pipe.clear = nv30_clear;
-	nv30->pipe.flush = nv30_flush;
-
-	nv30->pipe.is_texture_referenced = nouveau_is_texture_referenced;
-	nv30->pipe.is_buffer_referenced = nouveau_is_buffer_referenced;
-
-	screen->base.channel->user_private = nv30;
-	screen->base.channel->flush_notify = nv30_state_flush_notify;
-
-	nv30_init_query_functions(nv30);
-	nv30_init_surface_functions(nv30);
-	nv30_init_state_functions(nv30);
-	nv30_init_transfer_functions(nv30);
-
-	/* Create, configure, and install fallback swtnl path */
-	nv30->draw = draw_create();
-	draw_wide_point_threshold(nv30->draw, 9999999.0);
-	draw_wide_line_threshold(nv30->draw, 9999999.0);
-	draw_enable_line_stipple(nv30->draw, FALSE);
-	draw_enable_point_sprites(nv30->draw, FALSE);
-	draw_set_rasterize_stage(nv30->draw, nv30_draw_render_stage(nv30));
-
-	return &nv30->pipe;
-}
diff --git a/src/gallium/drivers/nv30/nv30_context.h b/src/gallium/drivers/nv30/nv30_context.h
deleted file mode 100644
index 38a17a4455..0000000000
--- a/src/gallium/drivers/nv30/nv30_context.h
+++ /dev/null
@@ -1,224 +0,0 @@
-#ifndef __NV30_CONTEXT_H__
-#define __NV30_CONTEXT_H__
-
-#include <stdio.h>
-
-#include "pipe/p_context.h"
-#include "pipe/p_defines.h"
-#include "pipe/p_state.h"
-#include "pipe/p_compiler.h"
-
-#include "util/u_memory.h"
-#include "util/u_math.h"
-#include "util/u_inlines.h"
-
-#include "draw/draw_vertex.h"
-
-#include "nouveau/nouveau_winsys.h"
-#include "nouveau/nouveau_gldefs.h"
-#include "nouveau/nouveau_context.h"
-#include "nouveau/nouveau_stateobj.h"
-
-#include "nv30_state.h"
-
-#define NOUVEAU_ERR(fmt, args...) \
-	fprintf(stderr, "%s:%d -  "fmt, __func__, __LINE__, ##args);
-#define NOUVEAU_MSG(fmt, args...) \
-	fprintf(stderr, "nouveau: "fmt, ##args);
-
-enum nv30_state_index {
-	NV30_STATE_FB = 0,
-	NV30_STATE_VIEWPORT = 1,
-	NV30_STATE_BLEND = 2,
-	NV30_STATE_RAST = 3,
-	NV30_STATE_ZSA = 4,
-	NV30_STATE_BCOL = 5,
-	NV30_STATE_CLIP = 6,
-	NV30_STATE_SCISSOR = 7,
-	NV30_STATE_STIPPLE = 8,
-	NV30_STATE_FRAGPROG = 9,
-	NV30_STATE_VERTPROG = 10,
-	NV30_STATE_FRAGTEX0 = 11,
-	NV30_STATE_FRAGTEX1 = 12,
-	NV30_STATE_FRAGTEX2 = 13,
-	NV30_STATE_FRAGTEX3 = 14,
-	NV30_STATE_FRAGTEX4 = 15,
-	NV30_STATE_FRAGTEX5 = 16,
-	NV30_STATE_FRAGTEX6 = 17,
-	NV30_STATE_FRAGTEX7 = 18,
-	NV30_STATE_FRAGTEX8 = 19,
-	NV30_STATE_FRAGTEX9 = 20,
-	NV30_STATE_FRAGTEX10 = 21,
-	NV30_STATE_FRAGTEX11 = 22,
-	NV30_STATE_FRAGTEX12 = 23,
-	NV30_STATE_FRAGTEX13 = 24,
-	NV30_STATE_FRAGTEX14 = 25,
-	NV30_STATE_FRAGTEX15 = 26,
-	NV30_STATE_VERTTEX0 = 27,
-	NV30_STATE_VERTTEX1 = 28,
-	NV30_STATE_VERTTEX2 = 29,
-	NV30_STATE_VERTTEX3 = 30,
-	NV30_STATE_VTXBUF = 31,
-	NV30_STATE_VTXFMT = 32,
-	NV30_STATE_VTXATTR = 33,
-	NV30_STATE_SR = 34,
-	NV30_STATE_MAX = 35
-};
-
-#include "nv30_screen.h"
-
-#define NV30_NEW_BLEND		(1 <<  0)
-#define NV30_NEW_RAST		(1 <<  1)
-#define NV30_NEW_ZSA		(1 <<  2)
-#define NV30_NEW_SAMPLER	(1 <<  3)
-#define NV30_NEW_FB		(1 <<  4)
-#define NV30_NEW_STIPPLE	(1 <<  5)
-#define NV30_NEW_SCISSOR	(1 <<  6)
-#define NV30_NEW_VIEWPORT	(1 <<  7)
-#define NV30_NEW_BCOL		(1 <<  8)
-#define NV30_NEW_VERTPROG	(1 <<  9)
-#define NV30_NEW_FRAGPROG	(1 << 10)
-#define NV30_NEW_ARRAYS		(1 << 11)
-#define NV30_NEW_UCP		(1 << 12)
-#define NV30_NEW_SR		(1 << 13)
-
-struct nv30_rasterizer_state {
-	struct pipe_rasterizer_state pipe;
-	struct nouveau_stateobj *so;
-};
-
-struct nv30_zsa_state {
-	struct pipe_depth_stencil_alpha_state pipe;
-	struct nouveau_stateobj *so;
-};
-
-struct nv30_blend_state {
-	struct pipe_blend_state pipe;
-	struct nouveau_stateobj *so;
-};
-
-
-struct nv30_state {
-	unsigned scissor_enabled;
-	unsigned stipple_enabled;
-	unsigned fp_samplers;
-
-	uint64_t dirty;
-	struct nouveau_stateobj *hw[NV30_STATE_MAX];
-};
-
-struct nv30_vtxelt_state {
-	struct pipe_vertex_element pipe[16];
-	unsigned num_elements;
-};
-
-struct nv30_context {
-	struct pipe_context pipe;
-
-	struct nouveau_winsys *nvws;
-	struct nv30_screen *screen;
-
-	struct draw_context *draw;
-
-	/* HW state derived from pipe states */
-	struct nv30_state state;
-
-	/* Context state */
-	unsigned dirty;
-	struct pipe_scissor_state scissor;
-	unsigned stipple[32];
-	struct nv30_vertex_program *vertprog;
-	struct nv30_fragment_program *fragprog;
-	struct pipe_buffer *constbuf[PIPE_SHADER_TYPES];
-	unsigned constbuf_nr[PIPE_SHADER_TYPES];
-	struct nv30_rasterizer_state *rasterizer;
-	struct nv30_zsa_state *zsa;
-	struct nv30_blend_state *blend;
-	struct pipe_blend_color blend_colour;
-	struct pipe_stencil_ref stencil_ref;
-	struct pipe_viewport_state viewport;
-	struct pipe_framebuffer_state framebuffer;
-	struct pipe_buffer *idxbuf;
-	unsigned idxbuf_format;
-	struct nv30_sampler_state *tex_sampler[PIPE_MAX_SAMPLERS];
-	struct nv30_miptree *tex_miptree[PIPE_MAX_SAMPLERS];
-	struct pipe_sampler_view *fragment_sampler_views[PIPE_MAX_SAMPLERS];
-	unsigned nr_samplers;
-	unsigned nr_textures;
-	unsigned dirty_samplers;
-	struct pipe_vertex_buffer vtxbuf[PIPE_MAX_ATTRIBS];
-	unsigned vtxbuf_nr;
-	struct nv30_vtxelt_state *vtxelt;
-};
-
-static INLINE struct nv30_context *
-nv30_context(struct pipe_context *pipe)
-{
-	return (struct nv30_context *)pipe;
-}
-
-struct nv30_state_entry {
-	boolean (*validate)(struct nv30_context *nv30);
-	struct {
-		unsigned pipe;
-		unsigned hw;
-	} dirty;
-};
-
-extern void nv30_init_state_functions(struct nv30_context *nv30);
-extern void nv30_init_surface_functions(struct nv30_context *nv30);
-extern void nv30_init_query_functions(struct nv30_context *nv30);
-extern void nv30_init_transfer_functions(struct nv30_context *nv30);
-
-extern void nv30_screen_init_miptree_functions(struct pipe_screen *pscreen);
-
-/* nv30_draw.c */
-extern struct draw_stage *nv30_draw_render_stage(struct nv30_context *nv30);
-
-/* nv30_vertprog.c */
-extern void nv30_vertprog_destroy(struct nv30_context *,
-				  struct nv30_vertex_program *);
-
-/* nv30_fragprog.c */
-extern void nv30_fragprog_destroy(struct nv30_context *,
-				  struct nv30_fragment_program *);
-
-/* nv30_fragtex.c */
-extern void nv30_fragtex_bind(struct nv30_context *);
-
-/* nv30_state.c and friends */
-extern boolean nv30_state_validate(struct nv30_context *nv30);
-extern void nv30_state_emit(struct nv30_context *nv30);
-extern void nv30_state_flush_notify(struct nouveau_channel *chan);
-extern struct nv30_state_entry nv30_state_rasterizer;
-extern struct nv30_state_entry nv30_state_scissor;
-extern struct nv30_state_entry nv30_state_stipple;
-extern struct nv30_state_entry nv30_state_fragprog;
-extern struct nv30_state_entry nv30_state_vertprog;
-extern struct nv30_state_entry nv30_state_blend;
-extern struct nv30_state_entry nv30_state_blend_colour;
-extern struct nv30_state_entry nv30_state_zsa;
-extern struct nv30_state_entry nv30_state_viewport;
-extern struct nv30_state_entry nv30_state_framebuffer;
-extern struct nv30_state_entry nv30_state_fragtex;
-extern struct nv30_state_entry nv30_state_vbo;
-extern struct nv30_state_entry nv30_state_sr;
-
-/* nv30_vbo.c */
-extern void nv30_draw_arrays(struct pipe_context *, unsigned mode,
-				unsigned start, unsigned count);
-extern void nv30_draw_elements(struct pipe_context *pipe,
-				  struct pipe_buffer *indexBuffer,
-				  unsigned indexSize,
-				  unsigned mode, unsigned start,
-				  unsigned count);
-
-/* nv30_clear.c */
-extern void nv30_clear(struct pipe_context *pipe, unsigned buffers,
-		       const float *rgba, double depth, unsigned stencil);
-
-/* nv30_context.c */
-struct pipe_context *
-nv30_create(struct pipe_screen *pscreen, void *priv);
-
-#endif
diff --git a/src/gallium/drivers/nv30/nv30_draw.c b/src/gallium/drivers/nv30/nv30_draw.c
deleted file mode 100644
index 74fc138c05..0000000000
--- a/src/gallium/drivers/nv30/nv30_draw.c
+++ /dev/null
@@ -1,61 +0,0 @@
-#include "draw/draw_pipe.h"
-
-#include "nv30_context.h"
-
-struct nv30_draw_stage {
-	struct draw_stage draw;
-	struct nv30_context *nv30;
-};
-
-static void
-nv30_draw_point(struct draw_stage *draw, struct prim_header *prim)
-{
-	NOUVEAU_ERR("\n");
-}
-
-static void
-nv30_draw_line(struct draw_stage *draw, struct prim_header *prim)
-{
-	NOUVEAU_ERR("\n");
-}
-
-static void
-nv30_draw_tri(struct draw_stage *draw, struct prim_header *prim)
-{
-	NOUVEAU_ERR("\n");
-}
-
-static void
-nv30_draw_flush(struct draw_stage *draw, unsigned flags)
-{
-}
-
-static void
-nv30_draw_reset_stipple_counter(struct draw_stage *draw)
-{
-	NOUVEAU_ERR("\n");
-}
-
-static void
-nv30_draw_destroy(struct draw_stage *draw)
-{
-	FREE(draw);
-}
-
-struct draw_stage *
-nv30_draw_render_stage(struct nv30_context *nv30)
-{
-	struct nv30_draw_stage *nv30draw = CALLOC_STRUCT(nv30_draw_stage);
-
-	nv30draw->nv30 = nv30;
-	nv30draw->draw.draw = nv30->draw;
-	nv30draw->draw.point = nv30_draw_point;
-	nv30draw->draw.line = nv30_draw_line;
-	nv30draw->draw.tri = nv30_draw_tri;
-	nv30draw->draw.flush = nv30_draw_flush;
-	nv30draw->draw.reset_stipple_counter = nv30_draw_reset_stipple_counter;
-	nv30draw->draw.destroy = nv30_draw_destroy;
-
-	return &nv30draw->draw;
-}
-
diff --git a/src/gallium/drivers/nv30/nv30_fragprog.c b/src/gallium/drivers/nv30/nv30_fragprog.c
deleted file mode 100644
index 2c432c6dfa..0000000000
--- a/src/gallium/drivers/nv30/nv30_fragprog.c
+++ /dev/null
@@ -1,905 +0,0 @@
-#include "pipe/p_context.h"
-#include "pipe/p_defines.h"
-#include "pipe/p_state.h"
-#include "util/u_inlines.h"
-
-#include "pipe/p_shader_tokens.h"
-#include "tgsi/tgsi_dump.h"
-#include "tgsi/tgsi_parse.h"
-#include "tgsi/tgsi_util.h"
-
-#include "nv30_context.h"
-
-#define SWZ_X 0
-#define SWZ_Y 1
-#define SWZ_Z 2
-#define SWZ_W 3
-#define MASK_X 1
-#define MASK_Y 2
-#define MASK_Z 4
-#define MASK_W 8
-#define MASK_ALL (MASK_X|MASK_Y|MASK_Z|MASK_W)
-#define DEF_SCALE NV30_FP_OP_DST_SCALE_1X
-#define DEF_CTEST NV30_FP_OP_COND_TR
-#include "nv30_shader.h"
-
-#define swz(s,x,y,z,w) nv30_sr_swz((s), SWZ_##x, SWZ_##y, SWZ_##z, SWZ_##w)
-#define neg(s) nv30_sr_neg((s))
-#define abs(s) nv30_sr_abs((s))
-#define scale(s,v) nv30_sr_scale((s), NV30_FP_OP_DST_SCALE_##v)
-
-#define MAX_CONSTS 128
-#define MAX_IMM 32
-struct nv30_fpc {
-	struct nv30_fragment_program *fp;
-
-	uint attrib_map[PIPE_MAX_SHADER_INPUTS];
-
-	int high_temp;
-	int temp_temp_count;
-	int num_regs;
-
-	uint depth_id;
-	uint colour_id;
-
-	unsigned inst_offset;
-
-	struct {
-		int pipe;
-		float vals[4];
-	} consts[MAX_CONSTS];
-	int nr_consts;
-
-	struct nv30_sreg imm[MAX_IMM];
-	unsigned nr_imm;
-};
-
-static INLINE struct nv30_sreg
-temp(struct nv30_fpc *fpc)
-{
-	int idx;
-
-	idx  = fpc->temp_temp_count++;
-	idx += fpc->high_temp + 1;
-	return nv30_sr(NV30SR_TEMP, idx);
-}
-
-static INLINE struct nv30_sreg
-constant(struct nv30_fpc *fpc, int pipe, float vals[4])
-{
-	int idx;
-
-	if (fpc->nr_consts == MAX_CONSTS)
-		assert(0);
-	idx = fpc->nr_consts++;
-
-	fpc->consts[idx].pipe = pipe;
-	if (pipe == -1)
-		memcpy(fpc->consts[idx].vals, vals, 4 * sizeof(float));
-	return nv30_sr(NV30SR_CONST, idx);
-}
-
-#define arith(cc,s,o,d,m,s0,s1,s2) \
-	nv30_fp_arith((cc), (s), NV30_FP_OP_OPCODE_##o, \
-			(d), (m), (s0), (s1), (s2))
-#define tex(cc,s,o,u,d,m,s0,s1,s2) \
-	nv30_fp_tex((cc), (s), NV30_FP_OP_OPCODE_##o, (u), \
-		    (d), (m), (s0), none, none)
-
-static void
-grow_insns(struct nv30_fpc *fpc, int size)
-{
-	struct nv30_fragment_program *fp = fpc->fp;
-
-	fp->insn_len += size;
-	fp->insn = realloc(fp->insn, sizeof(uint32_t) * fp->insn_len);
-}
-
-static void
-emit_src(struct nv30_fpc *fpc, int pos, struct nv30_sreg src)
-{
-	struct nv30_fragment_program *fp = fpc->fp;
-	uint32_t *hw = &fp->insn[fpc->inst_offset];
-	uint32_t sr = 0;
-
-	switch (src.type) {
-	case NV30SR_INPUT:
-		sr |= (NV30_FP_REG_TYPE_INPUT << NV30_FP_REG_TYPE_SHIFT);
-		hw[0] |= (src.index << NV30_FP_OP_INPUT_SRC_SHIFT);
-		break;
-	case NV30SR_OUTPUT:
-		sr |= NV30_FP_REG_SRC_HALF;
-		/* fall-through */
-	case NV30SR_TEMP:
-		sr |= (NV30_FP_REG_TYPE_TEMP << NV30_FP_REG_TYPE_SHIFT);
-		sr |= (src.index << NV30_FP_REG_SRC_SHIFT);
-		break;
-	case NV30SR_CONST:
-		grow_insns(fpc, 4);
-		hw = &fp->insn[fpc->inst_offset];
-		if (fpc->consts[src.index].pipe >= 0) {
-			struct nv30_fragment_program_data *fpd;
-
-			fp->consts = realloc(fp->consts, ++fp->nr_consts *
-					     sizeof(*fpd));
-			fpd = &fp->consts[fp->nr_consts - 1];
-			fpd->offset = fpc->inst_offset + 4;
-			fpd->index = fpc->consts[src.index].pipe;
-			memset(&fp->insn[fpd->offset], 0, sizeof(uint32_t) * 4);
-		} else {
-			memcpy(&fp->insn[fpc->inst_offset + 4],
-				fpc->consts[src.index].vals,
-				sizeof(uint32_t) * 4);
-		}
-
-		sr |= (NV30_FP_REG_TYPE_CONST << NV30_FP_REG_TYPE_SHIFT);
-		break;
-	case NV30SR_NONE:
-		sr |= (NV30_FP_REG_TYPE_INPUT << NV30_FP_REG_TYPE_SHIFT);
-		break;
-	default:
-		assert(0);
-	}
-
-	if (src.negate)
-		sr |= NV30_FP_REG_NEGATE;
-
-	if (src.abs)
-		hw[1] |= (1 << (29 + pos));
-
-	sr |= ((src.swz[0] << NV30_FP_REG_SWZ_X_SHIFT) |
-	       (src.swz[1] << NV30_FP_REG_SWZ_Y_SHIFT) |
-	       (src.swz[2] << NV30_FP_REG_SWZ_Z_SHIFT) |
-	       (src.swz[3] << NV30_FP_REG_SWZ_W_SHIFT));
-
-	hw[pos + 1] |= sr;
-}
-
-static void
-emit_dst(struct nv30_fpc *fpc, struct nv30_sreg dst)
-{
-	struct nv30_fragment_program *fp = fpc->fp;
-	uint32_t *hw = &fp->insn[fpc->inst_offset];
-
-	switch (dst.type) {
-	case NV30SR_TEMP:
-		if (fpc->num_regs < (dst.index + 1))
-			fpc->num_regs = dst.index + 1;
-		break;
-	case NV30SR_OUTPUT:
-		if (dst.index == 1) {
-			fp->fp_control |= 0xe;
-		} else {
-			hw[0] |= NV30_FP_OP_OUT_REG_HALF;
-		}
-		break;
-	case NV30SR_NONE:
-		hw[0] |= (1 << 30);
-		break;
-	default:
-		assert(0);
-	}
-
-	hw[0] |= (dst.index << NV30_FP_OP_OUT_REG_SHIFT);
-}
-
-static void
-nv30_fp_arith(struct nv30_fpc *fpc, int sat, int op,
-	      struct nv30_sreg dst, int mask,
-	      struct nv30_sreg s0, struct nv30_sreg s1, struct nv30_sreg s2)
-{
-	struct nv30_fragment_program *fp = fpc->fp;
-	uint32_t *hw;
-
-	fpc->inst_offset = fp->insn_len;
-	grow_insns(fpc, 4);
-	hw = &fp->insn[fpc->inst_offset];
-	memset(hw, 0, sizeof(uint32_t) * 4);
-
-	if (op == NV30_FP_OP_OPCODE_KIL)
-		fp->fp_control |= NV34TCL_FP_CONTROL_USES_KIL;
-	hw[0] |= (op << NV30_FP_OP_OPCODE_SHIFT);
-	hw[0] |= (mask << NV30_FP_OP_OUTMASK_SHIFT);
-	hw[2] |= (dst.dst_scale << NV30_FP_OP_DST_SCALE_SHIFT);
-
-	if (sat)
-		hw[0] |= NV30_FP_OP_OUT_SAT;
-
-	if (dst.cc_update)
-		hw[0] |= NV30_FP_OP_COND_WRITE_ENABLE;
-	hw[1] |= (dst.cc_test << NV30_FP_OP_COND_SHIFT);
-	hw[1] |= ((dst.cc_swz[0] << NV30_FP_OP_COND_SWZ_X_SHIFT) |
-		  (dst.cc_swz[1] << NV30_FP_OP_COND_SWZ_Y_SHIFT) |
-		  (dst.cc_swz[2] << NV30_FP_OP_COND_SWZ_Z_SHIFT) |
-		  (dst.cc_swz[3] << NV30_FP_OP_COND_SWZ_W_SHIFT));
-
-	emit_dst(fpc, dst);
-	emit_src(fpc, 0, s0);
-	emit_src(fpc, 1, s1);
-	emit_src(fpc, 2, s2);
-}
-
-static void
-nv30_fp_tex(struct nv30_fpc *fpc, int sat, int op, int unit,
-	    struct nv30_sreg dst, int mask,
-	    struct nv30_sreg s0, struct nv30_sreg s1, struct nv30_sreg s2)
-{
-	struct nv30_fragment_program *fp = fpc->fp;
-
-	nv30_fp_arith(fpc, sat, op, dst, mask, s0, s1, s2);
-
-	fp->insn[fpc->inst_offset] |= (unit << NV30_FP_OP_TEX_UNIT_SHIFT);
-	fp->samplers |= (1 << unit);
-}
-
-static INLINE struct nv30_sreg
-tgsi_src(struct nv30_fpc *fpc, const struct tgsi_full_src_register *fsrc)
-{
-	struct nv30_sreg src;
-
-	switch (fsrc->Register.File) {
-	case TGSI_FILE_INPUT:
-		src = nv30_sr(NV30SR_INPUT,
-			      fpc->attrib_map[fsrc->Register.Index]);
-		break;
-	case TGSI_FILE_CONSTANT:
-		src = constant(fpc, fsrc->Register.Index, NULL);
-		break;
-	case TGSI_FILE_IMMEDIATE:
-		assert(fsrc->Register.Index < fpc->nr_imm);
-		src = fpc->imm[fsrc->Register.Index];
-		break;
-	case TGSI_FILE_TEMPORARY:
-		src = nv30_sr(NV30SR_TEMP, fsrc->Register.Index + 1);
-		if (fpc->high_temp < src.index)
-			fpc->high_temp = src.index;
-		break;
-	/* This is clearly insane, but gallium hands us shaders like this.
-	 * Luckily fragprog results are just temp regs..
-	 */
-	case TGSI_FILE_OUTPUT:
-		if (fsrc->Register.Index == fpc->colour_id)
-			return nv30_sr(NV30SR_OUTPUT, 0);
-		else
-			return nv30_sr(NV30SR_OUTPUT, 1);
-		break;
-	default:
-		NOUVEAU_ERR("bad src file\n");
-		break;
-	}
-
-	src.abs = fsrc->Register.Absolute;
-	src.negate = fsrc->Register.Negate;
-	src.swz[0] = fsrc->Register.SwizzleX;
-	src.swz[1] = fsrc->Register.SwizzleY;
-	src.swz[2] = fsrc->Register.SwizzleZ;
-	src.swz[3] = fsrc->Register.SwizzleW;
-	return src;
-}
-
-static INLINE struct nv30_sreg
-tgsi_dst(struct nv30_fpc *fpc, const struct tgsi_full_dst_register *fdst) {
-	int idx;
-
-	switch (fdst->Register.File) {
-	case TGSI_FILE_OUTPUT:
-		if (fdst->Register.Index == fpc->colour_id)
-			return nv30_sr(NV30SR_OUTPUT, 0);
-		else
-			return nv30_sr(NV30SR_OUTPUT, 1);
-		break;
-	case TGSI_FILE_TEMPORARY:
-		idx = fdst->Register.Index + 1;
-		if (fpc->high_temp < idx)
-			fpc->high_temp = idx;
-		return nv30_sr(NV30SR_TEMP, idx);
-	case TGSI_FILE_NULL:
-		return nv30_sr(NV30SR_NONE, 0);
-	default:
-		NOUVEAU_ERR("bad dst file %d\n", fdst->Register.File);
-		return nv30_sr(NV30SR_NONE, 0);
-	}
-}
-
-static INLINE int
-tgsi_mask(uint tgsi)
-{
-	int mask = 0;
-
-	if (tgsi & TGSI_WRITEMASK_X) mask |= MASK_X;
-	if (tgsi & TGSI_WRITEMASK_Y) mask |= MASK_Y;
-	if (tgsi & TGSI_WRITEMASK_Z) mask |= MASK_Z;
-	if (tgsi & TGSI_WRITEMASK_W) mask |= MASK_W;
-	return mask;
-}
-
-static boolean
-src_native_swz(struct nv30_fpc *fpc, const struct tgsi_full_src_register *fsrc,
-	       struct nv30_sreg *src)
-{
-	const struct nv30_sreg none = nv30_sr(NV30SR_NONE, 0);
-	struct nv30_sreg tgsi = tgsi_src(fpc, fsrc);
-	uint mask = 0;
-	uint c;
-
-	for (c = 0; c < 4; c++) {
-		switch (tgsi_util_get_full_src_register_swizzle(fsrc, c)) {
-		case TGSI_SWIZZLE_X:
-		case TGSI_SWIZZLE_Y:
-		case TGSI_SWIZZLE_Z:
-		case TGSI_SWIZZLE_W:
-			mask |= (1 << c);
-			break;
-		default:
-			assert(0);
-		}
-	}
-
-	if (mask == MASK_ALL)
-		return TRUE;
-
-	*src = temp(fpc);
-
-	if (mask)
-		arith(fpc, 0, MOV, *src, mask, tgsi, none, none);
-
-	return FALSE;
-}
-
-static boolean
-nv30_fragprog_parse_instruction(struct nv30_fpc *fpc,
-				const struct tgsi_full_instruction *finst)
-{
-	const struct nv30_sreg none = nv30_sr(NV30SR_NONE, 0);
-	struct nv30_sreg src[3], dst, tmp;
-	int mask, sat, unit = 0;
-	int ai = -1, ci = -1;
-	int i;
-
-	if (finst->Instruction.Opcode == TGSI_OPCODE_END)
-		return TRUE;
-
-	fpc->temp_temp_count = 0;
-	for (i = 0; i < finst->Instruction.NumSrcRegs; i++) {
-		const struct tgsi_full_src_register *fsrc;
-
-		fsrc = &finst->Src[i];
-		if (fsrc->Register.File == TGSI_FILE_TEMPORARY) {
-			src[i] = tgsi_src(fpc, fsrc);
-		}
-	}
-
-	for (i = 0; i < finst->Instruction.NumSrcRegs; i++) {
-		const struct tgsi_full_src_register *fsrc;
-
-		fsrc = &finst->Src[i];
-
-		switch (fsrc->Register.File) {
-		case TGSI_FILE_INPUT:
-		case TGSI_FILE_CONSTANT:
-		case TGSI_FILE_TEMPORARY:
-			if (!src_native_swz(fpc, fsrc, &src[i]))
-				continue;
-			break;
-		default:
-			break;
-		}
-
-		switch (fsrc->Register.File) {
-		case TGSI_FILE_INPUT:
-			if (ai == -1 || ai == fsrc->Register.Index) {
-				ai = fsrc->Register.Index;
-				src[i] = tgsi_src(fpc, fsrc);
-			} else {
-				NOUVEAU_MSG("extra src attr %d\n",
-					 fsrc->Register.Index);
-				src[i] = temp(fpc);
-				arith(fpc, 0, MOV, src[i], MASK_ALL,
-				      tgsi_src(fpc, fsrc), none, none);
-			}
-			break;
-		case TGSI_FILE_CONSTANT:
-		case TGSI_FILE_IMMEDIATE:
-			if (ci == -1 || ci == fsrc->Register.Index) {
-				ci = fsrc->Register.Index;
-				src[i] = tgsi_src(fpc, fsrc);
-			} else {
-				src[i] = temp(fpc);
-				arith(fpc, 0, MOV, src[i], MASK_ALL,
-				      tgsi_src(fpc, fsrc), none, none);
-			}
-			break;
-		case TGSI_FILE_TEMPORARY:
-			/* handled above */
-			break;
-		case TGSI_FILE_SAMPLER:
-			unit = fsrc->Register.Index;
-			break;
-		case TGSI_FILE_OUTPUT:
-			break;
-		default:
-			NOUVEAU_ERR("bad src file\n");
-			return FALSE;
-		}
-	}
-
-	dst  = tgsi_dst(fpc, &finst->Dst[0]);
-	mask = tgsi_mask(finst->Dst[0].Register.WriteMask);
-	sat  = (finst->Instruction.Saturate == TGSI_SAT_ZERO_ONE);
-
-	switch (finst->Instruction.Opcode) {
-	case TGSI_OPCODE_ABS:
-		arith(fpc, sat, MOV, dst, mask, abs(src[0]), none, none);
-		break;
-	case TGSI_OPCODE_ADD:
-		arith(fpc, sat, ADD, dst, mask, src[0], src[1], none);
-		break;
-	case TGSI_OPCODE_CMP:
-		tmp = nv30_sr(NV30SR_NONE, 0);
-		tmp.cc_update = 1;
-		arith(fpc, 0, MOV, tmp, 0xf, src[0], none, none);
-		dst.cc_test = NV30_VP_INST_COND_GE;
-		arith(fpc, sat, MOV, dst, mask, src[2], none, none);
-		dst.cc_test = NV30_VP_INST_COND_LT;
-		arith(fpc, sat, MOV, dst, mask, src[1], none, none);
-		break;
-	case TGSI_OPCODE_COS:
-		arith(fpc, sat, COS, dst, mask, src[0], none, none);
-		break;
-	case TGSI_OPCODE_DP3:
-		arith(fpc, sat, DP3, dst, mask, src[0], src[1], none);
-		break;
-	case TGSI_OPCODE_DP4:
-		arith(fpc, sat, DP4, dst, mask, src[0], src[1], none);
-		break;
-	case TGSI_OPCODE_DPH:
-		tmp = temp(fpc);
-		arith(fpc, 0, DP3, tmp, MASK_X, src[0], src[1], none);
-		arith(fpc, sat, ADD, dst, mask, swz(tmp, X, X, X, X),
-		      swz(src[1], W, W, W, W), none);
-		break;
-	case TGSI_OPCODE_DST:
-		arith(fpc, sat, DST, dst, mask, src[0], src[1], none);
-		break;
-	case TGSI_OPCODE_EX2:
-		arith(fpc, sat, EX2, dst, mask, src[0], none, none);
-		break;
-	case TGSI_OPCODE_FLR:
-		arith(fpc, sat, FLR, dst, mask, src[0], none, none);
-		break;
-	case TGSI_OPCODE_FRC:
-		arith(fpc, sat, FRC, dst, mask, src[0], none, none);
-		break;
-	case TGSI_OPCODE_KILP:
-		arith(fpc, 0, KIL, none, 0, none, none, none);
-		break;
-	case TGSI_OPCODE_KIL:
-		dst = nv30_sr(NV30SR_NONE, 0);
-		dst.cc_update = 1;
-		arith(fpc, 0, MOV, dst, MASK_ALL, src[0], none, none);
-		dst.cc_update = 0; dst.cc_test = NV30_FP_OP_COND_LT;
-		arith(fpc, 0, KIL, dst, 0, none, none, none);
-		break;
-	case TGSI_OPCODE_LG2:
-		arith(fpc, sat, LG2, dst, mask, src[0], none, none);
-		break;
-//	case TGSI_OPCODE_LIT:
-	case TGSI_OPCODE_LRP:
-		arith(fpc, sat, LRP, dst, mask, src[0], src[1], src[2]);
-		break;
-	case TGSI_OPCODE_MAD:
-		arith(fpc, sat, MAD, dst, mask, src[0], src[1], src[2]);
-		break;
-	case TGSI_OPCODE_MAX:
-		arith(fpc, sat, MAX, dst, mask, src[0], src[1], none);
-		break;
-	case TGSI_OPCODE_MIN:
-		arith(fpc, sat, MIN, dst, mask, src[0], src[1], none);
-		break;
-	case TGSI_OPCODE_MOV:
-		arith(fpc, sat, MOV, dst, mask, src[0], none, none);
-		break;
-	case TGSI_OPCODE_MUL:
-		arith(fpc, sat, MUL, dst, mask, src[0], src[1], none);
-		break;
-	case TGSI_OPCODE_POW:
-		arith(fpc, sat, POW, dst, mask, src[0], src[1], none);
-		break;
-	case TGSI_OPCODE_RCP:
-		arith(fpc, sat, RCP, dst, mask, src[0], none, none);
-		break;
-	case TGSI_OPCODE_RET:
-		assert(0);
-		break;
-	case TGSI_OPCODE_RFL:
-		arith(fpc, 0, RFL, dst, mask, src[0], src[1], none);
-		break;
-	case TGSI_OPCODE_RSQ:
-		arith(fpc, sat, RSQ, dst, mask, abs(swz(src[0], X, X, X, X)), none, none);
-		break;
-	case TGSI_OPCODE_SCS:
-		/* avoid overwriting the source */
-		if(src[0].swz[SWZ_X] != SWZ_X)
-		{
-			if (mask & MASK_X) {
-				arith(fpc, sat, COS, dst, MASK_X,
-				      swz(src[0], X, X, X, X), none, none);
-			}
-			if (mask & MASK_Y) {
-				arith(fpc, sat, SIN, dst, MASK_Y,
-				      swz(src[0], X, X, X, X), none, none);
-			}
-		}
-		else
-		{
-			if (mask & MASK_Y) {
-				arith(fpc, sat, SIN, dst, MASK_Y,
-				      swz(src[0], X, X, X, X), none, none);
-			}
-			if (mask & MASK_X) {
-				arith(fpc, sat, COS, dst, MASK_X,
-				      swz(src[0], X, X, X, X), none, none);
-			}
-		}
-		break;
-	case TGSI_OPCODE_SIN:
-		arith(fpc, sat, SIN, dst, mask, src[0], none, none);
-		break;
-	case TGSI_OPCODE_SGE:
-		arith(fpc, sat, SGE, dst, mask, src[0], src[1], none);
-		break;
-	case TGSI_OPCODE_SGT:
-		arith(fpc, sat, SGT, dst, mask, src[0], src[1], none);
-		break;
-	case TGSI_OPCODE_SLT:
-		arith(fpc, sat, SLT, dst, mask, src[0], src[1], none);
-		break;
-	case TGSI_OPCODE_SUB:
-		arith(fpc, sat, ADD, dst, mask, src[0], neg(src[1]), none);
-		break;
-	case TGSI_OPCODE_TEX:
-		tex(fpc, sat, TEX, unit, dst, mask, src[0], none, none);
-		break;
-	case TGSI_OPCODE_TXB:
-		tex(fpc, sat, TXB, unit, dst, mask, src[0], none, none);
-		break;
-	case TGSI_OPCODE_TXP:
-		tex(fpc, sat, TXP, unit, dst, mask, src[0], none, none);
-		break;
-	case TGSI_OPCODE_XPD:
-		tmp = temp(fpc);
-		arith(fpc, 0, MUL, tmp, mask,
-		      swz(src[0], Z, X, Y, Y), swz(src[1], Y, Z, X, X), none);
-		arith(fpc, sat, MAD, dst, (mask & ~MASK_W),
-		      swz(src[0], Y, Z, X, X), swz(src[1], Z, X, Y, Y),
-		      neg(tmp));
-		break;
-	default:
-		NOUVEAU_ERR("invalid opcode %d\n", finst->Instruction.Opcode);
-		return FALSE;
-	}
-
-	return TRUE;
-}
-
-static boolean
-nv30_fragprog_parse_decl_attrib(struct nv30_fpc *fpc,
-				const struct tgsi_full_declaration *fdec)
-{
-	int hw;
-
-	switch (fdec->Semantic.Name) {
-	case TGSI_SEMANTIC_POSITION:
-		hw = NV30_FP_OP_INPUT_SRC_POSITION;
-		break;
-	case TGSI_SEMANTIC_COLOR:
-		if (fdec->Semantic.Index == 0) {
-			hw = NV30_FP_OP_INPUT_SRC_COL0;
-		} else
-		if (fdec->Semantic.Index == 1) {
-			hw = NV30_FP_OP_INPUT_SRC_COL1;
-		} else {
-			NOUVEAU_ERR("bad colour semantic index\n");
-			return FALSE;
-		}
-		break;
-	case TGSI_SEMANTIC_FOG:
-		hw = NV30_FP_OP_INPUT_SRC_FOGC;
-		break;
-	case TGSI_SEMANTIC_GENERIC:
-		if (fdec->Semantic.Index <= 7) {
-			hw = NV30_FP_OP_INPUT_SRC_TC(fdec->Semantic.
-						     Index);
-		} else {
-			NOUVEAU_ERR("bad generic semantic index\n");
-			return FALSE;
-		}
-		break;
-	default:
-		NOUVEAU_ERR("bad input semantic\n");
-		return FALSE;
-	}
-
-	fpc->attrib_map[fdec->Range.First] = hw;
-	return TRUE;
-}
-
-static boolean
-nv30_fragprog_parse_decl_output(struct nv30_fpc *fpc,
-				const struct tgsi_full_declaration *fdec)
-{
-	switch (fdec->Semantic.Name) {
-	case TGSI_SEMANTIC_POSITION:
-		fpc->depth_id = fdec->Range.First;
-		break;
-	case TGSI_SEMANTIC_COLOR:
-		fpc->colour_id = fdec->Range.First;
-		break;
-	default:
-		NOUVEAU_ERR("bad output semantic\n");
-		return FALSE;
-	}
-
-	return TRUE;
-}
-
-static boolean
-nv30_fragprog_prepare(struct nv30_fpc *fpc)
-{
-	struct tgsi_parse_context p;
-	/*int high_temp = -1, i;*/
-
-	tgsi_parse_init(&p, fpc->fp->pipe.tokens);
-	while (!tgsi_parse_end_of_tokens(&p)) {
-		const union tgsi_full_token *tok = &p.FullToken;
-
-		tgsi_parse_token(&p);
-		switch(tok->Token.Type) {
-		case TGSI_TOKEN_TYPE_DECLARATION:
-		{
-			const struct tgsi_full_declaration *fdec;
-			fdec = &p.FullToken.FullDeclaration;
-			switch (fdec->Declaration.File) {
-			case TGSI_FILE_INPUT:
-				if (!nv30_fragprog_parse_decl_attrib(fpc, fdec))
-					goto out_err;
-				break;
-			case TGSI_FILE_OUTPUT:
-				if (!nv30_fragprog_parse_decl_output(fpc, fdec))
-					goto out_err;
-				break;
-			/*case TGSI_FILE_TEMPORARY:
-				if (fdec->Range.Last > high_temp) {
-					high_temp =
-						fdec->Range.Last;
-				}
-				break;*/
-			default:
-				break;
-			}
-		}
-			break;
-		case TGSI_TOKEN_TYPE_IMMEDIATE:
-		{
-			struct tgsi_full_immediate *imm;
-			float vals[4];
-
-			imm = &p.FullToken.FullImmediate;
-			assert(imm->Immediate.DataType == TGSI_IMM_FLOAT32);
-			assert(fpc->nr_imm < MAX_IMM);
-
-			vals[0] = imm->u[0].Float;
-			vals[1] = imm->u[1].Float;
-			vals[2] = imm->u[2].Float;
-			vals[3] = imm->u[3].Float;
-			fpc->imm[fpc->nr_imm++] = constant(fpc, -1, vals);
-		}
-			break;
-		default:
-			break;
-		}
-	}
-	tgsi_parse_free(&p);
-
-	/*if (++high_temp) {
-		fpc->r_temp = CALLOC(high_temp, sizeof(struct nv30_sreg));
-		for (i = 0; i < high_temp; i++)
-			fpc->r_temp[i] = temp(fpc);
-		fpc->r_temps_discard = 0;
-	}*/
-
-	return TRUE;
-
-out_err:
-	/*if (fpc->r_temp)
-		FREE(fpc->r_temp);*/
-	tgsi_parse_free(&p);
-	return FALSE;
-}
-
-static void
-nv30_fragprog_translate(struct nv30_context *nv30,
-			struct nv30_fragment_program *fp)
-{
-	struct tgsi_parse_context parse;
-	struct nv30_fpc *fpc = NULL;
-
-	tgsi_dump(fp->pipe.tokens,0);
-
-	fpc = CALLOC(1, sizeof(struct nv30_fpc));
-	if (!fpc)
-		return;
-	fpc->fp = fp;
-	fpc->high_temp = -1;
-	fpc->num_regs = 2;
-
-	if (!nv30_fragprog_prepare(fpc)) {
-		FREE(fpc);
-		return;
-	}
-
-	tgsi_parse_init(&parse, fp->pipe.tokens);
-
-	while (!tgsi_parse_end_of_tokens(&parse)) {
-		tgsi_parse_token(&parse);
-
-		switch (parse.FullToken.Token.Type) {
-		case TGSI_TOKEN_TYPE_INSTRUCTION:
-		{
-			const struct tgsi_full_instruction *finst;
-
-			finst = &parse.FullToken.FullInstruction;
-			if (!nv30_fragprog_parse_instruction(fpc, finst))
-				goto out_err;
-		}
-			break;
-		default:
-			break;
-		}
-	}
-
-	fp->fp_control |= (fpc->num_regs-1)/2;
-	fp->fp_reg_control = (1<<16)|0x4;
-
-	/* Terminate final instruction */
-	fp->insn[fpc->inst_offset] |= 0x00000001;
-
-	/* Append NOP + END instruction, may or may not be necessary. */
-	fpc->inst_offset = fp->insn_len;
-	grow_insns(fpc, 4);
-	fp->insn[fpc->inst_offset + 0] = 0x00000001;
-	fp->insn[fpc->inst_offset + 1] = 0x00000000;
-	fp->insn[fpc->inst_offset + 2] = 0x00000000;
-	fp->insn[fpc->inst_offset + 3] = 0x00000000;
-
-	fp->translated = TRUE;
-	fp->on_hw = FALSE;
-out_err:
-	tgsi_parse_free(&parse);
-	FREE(fpc);
-}
-
-static void
-nv30_fragprog_upload(struct nv30_context *nv30,
-		     struct nv30_fragment_program *fp)
-{
-	struct pipe_screen *pscreen = nv30->pipe.screen;
-	const uint32_t le = 1;
-	uint32_t *map;
-	int i;
-
-	map = pipe_buffer_map(pscreen, fp->buffer, PIPE_BUFFER_USAGE_CPU_WRITE);
-
-#if 0
-	for (i = 0; i < fp->insn_len; i++) {
-		fflush(stdout); fflush(stderr);
-		NOUVEAU_ERR("%d 0x%08x\n", i, fp->insn[i]);
-		fflush(stdout); fflush(stderr);
-	}
-#endif
-
-	if ((*(const uint8_t *)&le)) {
-		for (i = 0; i < fp->insn_len; i++) {
-			map[i] = fp->insn[i];
-		}
-	} else {
-		/* Weird swapping for big-endian chips */
-		for (i = 0; i < fp->insn_len; i++) {
-			map[i] = ((fp->insn[i] & 0xffff) << 16) |
-				  ((fp->insn[i] >> 16) & 0xffff);
-		}
-	}
-
-	pipe_buffer_unmap(pscreen, fp->buffer);
-}
-
-static boolean
-nv30_fragprog_validate(struct nv30_context *nv30)
-{
-	struct nv30_fragment_program *fp = nv30->fragprog;
-	struct pipe_buffer *constbuf =
-		nv30->constbuf[PIPE_SHADER_FRAGMENT];
-	struct pipe_screen *pscreen = nv30->pipe.screen;
-	struct nouveau_stateobj *so;
-	boolean new_consts = FALSE;
-	int i;
-
-	if (fp->translated)
-		goto update_constants;
-
-	/*nv30->fallback_swrast &= ~NV30_NEW_FRAGPROG;*/
-	nv30_fragprog_translate(nv30, fp);
-	if (!fp->translated) {
-		/*nv30->fallback_swrast |= NV30_NEW_FRAGPROG;*/
-		return FALSE;
-	}
-
-	fp->buffer = pscreen->buffer_create(pscreen, 0x100, 0, fp->insn_len * 4);
-	nv30_fragprog_upload(nv30, fp);
-
-	so = so_new(4, 4, 1);
-	so_method(so, nv30->screen->rankine, NV34TCL_FP_ACTIVE_PROGRAM, 1);
-	so_reloc (so, nouveau_bo(fp->buffer), 0, NOUVEAU_BO_VRAM |
-		      NOUVEAU_BO_GART | NOUVEAU_BO_RD | NOUVEAU_BO_LOW |
-		      NOUVEAU_BO_OR, NV34TCL_FP_ACTIVE_PROGRAM_DMA0,
-		      NV34TCL_FP_ACTIVE_PROGRAM_DMA1);
-	so_method(so, nv30->screen->rankine, NV34TCL_FP_CONTROL, 1);
-	so_data  (so, fp->fp_control);
-	so_method(so, nv30->screen->rankine, NV34TCL_FP_REG_CONTROL, 1);
-	so_data  (so, fp->fp_reg_control);
-	so_method(so, nv30->screen->rankine, NV34TCL_TX_UNITS_ENABLE, 1);
-	so_data  (so, fp->samplers);
-	so_ref(so, &fp->so);
-	so_ref(NULL, &so);
-
-update_constants:
-	if (fp->nr_consts) {
-		float *map;
-
-		map = pipe_buffer_map(pscreen, constbuf,
-				      PIPE_BUFFER_USAGE_CPU_READ);
-		for (i = 0; i < fp->nr_consts; i++) {
-			struct nv30_fragment_program_data *fpd = &fp->consts[i];
-			uint32_t *p = &fp->insn[fpd->offset];
-			uint32_t *cb = (uint32_t *)&map[fpd->index * 4];
-
-			if (!memcmp(p, cb, 4 * sizeof(float)))
-				continue;
-			memcpy(p, cb, 4 * sizeof(float));
-			new_consts = TRUE;
-		}
-		pipe_buffer_unmap(pscreen, constbuf);
-
-		if (new_consts)
-			nv30_fragprog_upload(nv30, fp);
-	}
-
-	if (new_consts || fp->so != nv30->state.hw[NV30_STATE_FRAGPROG]) {
-		so_ref(fp->so, &nv30->state.hw[NV30_STATE_FRAGPROG]);
-		return TRUE;
-	}
-
-	return FALSE;
-}
-
-void
-nv30_fragprog_destroy(struct nv30_context *nv30,
-		      struct nv30_fragment_program *fp)
-{
-	if (fp->buffer)
-		pipe_buffer_reference(&fp->buffer, NULL);
-
-	if (fp->so)
-		so_ref(NULL, &fp->so);
-
-	if (fp->insn_len)
-		FREE(fp->insn);
-}
-
-struct nv30_state_entry nv30_state_fragprog = {
-	.validate = nv30_fragprog_validate,
-	.dirty = {
-		.pipe = NV30_NEW_FRAGPROG,
-		.hw = NV30_STATE_FRAGPROG
-	}
-};
diff --git a/src/gallium/drivers/nv30/nv30_miptree.c b/src/gallium/drivers/nv30/nv30_miptree.c
deleted file mode 100644
index bfa27b632f..0000000000
--- a/src/gallium/drivers/nv30/nv30_miptree.c
+++ /dev/null
@@ -1,240 +0,0 @@
-#include "pipe/p_state.h"
-#include "pipe/p_defines.h"
-#include "util/u_inlines.h"
-#include "util/u_format.h"
-#include "util/u_math.h"
-
-#include "nv30_context.h"
-#include "../nouveau/nv04_surface_2d.h"
-
-static void
-nv30_miptree_layout(struct nv30_miptree *nv30mt)
-{
-	struct pipe_texture *pt = &nv30mt->base;
-	uint width = pt->width0;
-	uint offset = 0;
-	int nr_faces, l, f;
-	uint wide_pitch = pt->tex_usage & (PIPE_TEXTURE_USAGE_SAMPLER |
-		                           PIPE_TEXTURE_USAGE_DEPTH_STENCIL |
-		                           PIPE_TEXTURE_USAGE_RENDER_TARGET |
-		                           PIPE_TEXTURE_USAGE_DISPLAY_TARGET |
-		                           PIPE_TEXTURE_USAGE_SCANOUT);
-
-	if (pt->target == PIPE_TEXTURE_CUBE) {
-		nr_faces = 6;
-	} else
-	if (pt->target == PIPE_TEXTURE_3D) {
-		nr_faces = pt->depth0;
-	} else {
-		nr_faces = 1;
-	}
-
-	for (l = 0; l <= pt->last_level; l++) {
-		if (wide_pitch && (pt->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR))
-			nv30mt->level[l].pitch = align(util_format_get_stride(pt->format, pt->width0), 64);
-		else
-			nv30mt->level[l].pitch = util_format_get_stride(pt->format, width);
-
-		nv30mt->level[l].image_offset =
-			CALLOC(nr_faces, sizeof(unsigned));
-
-		width  = u_minify(width, 1);
-	}
-
-	for (f = 0; f < nr_faces; f++) {
-		for (l = 0; l < pt->last_level; l++) {
-			nv30mt->level[l].image_offset[f] = offset;
-
-			if (!(pt->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR) &&
-			    u_minify(pt->width0, l + 1) > 1 && u_minify(pt->height0, l + 1) > 1)
-				offset += align(nv30mt->level[l].pitch * u_minify(pt->height0, l), 64);
-			else
-				offset += nv30mt->level[l].pitch * u_minify(pt->height0, l);
-		}
-
-		nv30mt->level[l].image_offset[f] = offset;
-		offset += nv30mt->level[l].pitch * u_minify(pt->height0, l);
-	}
-
-	nv30mt->total_size = offset;
-}
-
-static struct pipe_texture *
-nv30_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *pt)
-{
-	struct nv30_miptree *mt;
-	unsigned buf_usage = PIPE_BUFFER_USAGE_PIXEL |
-	                     NOUVEAU_BUFFER_USAGE_TEXTURE;
-
-	mt = MALLOC(sizeof(struct nv30_miptree));
-	if (!mt)
-		return NULL;
-	mt->base = *pt;
-	pipe_reference_init(&mt->base.reference, 1);
-	mt->base.screen = pscreen;
-
-	/* Swizzled textures must be POT */
-	if (pt->width0 & (pt->width0 - 1) ||
-	    pt->height0 & (pt->height0 - 1))
-		mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR;
-	else
-	if (pt->tex_usage & (PIPE_TEXTURE_USAGE_SCANOUT |
-	                     PIPE_TEXTURE_USAGE_DISPLAY_TARGET |
-	                     PIPE_TEXTURE_USAGE_DEPTH_STENCIL))
-		mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR;
-	else
-	if (pt->tex_usage & PIPE_TEXTURE_USAGE_DYNAMIC)
-		mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR;
-	else {
-		switch (pt->format) {
-		/* TODO: Figure out which formats can be swizzled */
-		case PIPE_FORMAT_B8G8R8A8_UNORM:
-		case PIPE_FORMAT_B8G8R8X8_UNORM:
-		case PIPE_FORMAT_R16_SNORM:
-		case PIPE_FORMAT_B5G6R5_UNORM:
-		case PIPE_FORMAT_L8A8_UNORM:
-		case PIPE_FORMAT_A8_UNORM:
-		case PIPE_FORMAT_L8_UNORM:
-		case PIPE_FORMAT_I8_UNORM:
-		{
-			if (debug_get_bool_option("NOUVEAU_NO_SWIZZLE", FALSE))
-				mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR;
- 			break;
-		}
-		default:
-			mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR;
-		}
-	}
-
-	if (pt->tex_usage & PIPE_TEXTURE_USAGE_DYNAMIC)
-		buf_usage |= PIPE_BUFFER_USAGE_CPU_READ_WRITE;
-
-	/* apparently we can't render to swizzled surfaces smaller than 64 bytes, so make them linear.
-	 * If the user did not ask for a render target, they can still render to it, but it will cost them an extra copy.
-	 * This also happens for small mipmaps of large textures. */
-	if (pt->tex_usage & PIPE_TEXTURE_USAGE_RENDER_TARGET && util_format_get_stride(pt->format, pt->width0) < 64)
-		mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR;
-
-	nv30_miptree_layout(mt);
-
-	mt->buffer = pscreen->buffer_create(pscreen, 256, buf_usage,
-				       mt->total_size);
-	if (!mt->buffer) {
-		FREE(mt);
-		return NULL;
-	}
-	mt->bo = nouveau_bo(mt->buffer);
-
-	return &mt->base;
-}
-
-static struct pipe_texture *
-nv30_miptree_blanket(struct pipe_screen *pscreen, const struct pipe_texture *pt,
-		     const unsigned *stride, struct pipe_buffer *pb)
-{
-	struct nv30_miptree *mt;
-
-	/* Only supports 2D, non-mipmapped textures for the moment */
-	if (pt->target != PIPE_TEXTURE_2D || pt->last_level != 0 ||
-	    pt->depth0 != 1)
-		return NULL;
-
-	mt = CALLOC_STRUCT(nv30_miptree);
-	if (!mt)
-		return NULL;
-
-	mt->base = *pt;
-	pipe_reference_init(&mt->base.reference, 1);
-	mt->base.screen = pscreen;
-	mt->level[0].pitch = stride[0];
-	mt->level[0].image_offset = CALLOC(1, sizeof(unsigned));
-
-	/* Assume whoever created this buffer expects it to be linear for now */
-	mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR;
-
-	pipe_buffer_reference(&mt->buffer, pb);
-	mt->bo = nouveau_bo(mt->buffer);
-	return &mt->base;
-}
-
-static void
-nv30_miptree_destroy(struct pipe_texture *pt)
-{
-	struct nv30_miptree *mt = (struct nv30_miptree *)pt;
-	int l;
-
-	pipe_buffer_reference(&mt->buffer, NULL);
-	for (l = 0; l <= pt->last_level; l++) {
-		if (mt->level[l].image_offset)
-			FREE(mt->level[l].image_offset);
-	}
-
-	FREE(mt);
-}
-
-static struct pipe_surface *
-nv30_miptree_surface_new(struct pipe_screen *pscreen, struct pipe_texture *pt,
-			 unsigned face, unsigned level, unsigned zslice,
-			 unsigned flags)
-{
-	struct nv30_miptree *nv30mt = (struct nv30_miptree *)pt;
-	struct nv04_surface *ns;
-
-	ns = CALLOC_STRUCT(nv04_surface);
-	if (!ns)
-		return NULL;
-	pipe_texture_reference(&ns->base.texture, pt);
-	ns->base.format = pt->format;
-	ns->base.width = u_minify(pt->width0, level);
-	ns->base.height = u_minify(pt->height0, level);
-	ns->base.usage = flags;
-	pipe_reference_init(&ns->base.reference, 1);
-	ns->base.face = face;
-	ns->base.level = level;
-	ns->base.zslice = zslice;
-	ns->pitch = nv30mt->level[level].pitch;
-
-	if (pt->target == PIPE_TEXTURE_CUBE) {
-		ns->base.offset = nv30mt->level[level].image_offset[face];
-	} else
-	if (pt->target == PIPE_TEXTURE_3D) {
-		ns->base.offset = nv30mt->level[level].image_offset[zslice];
-	} else {
-		ns->base.offset = nv30mt->level[level].image_offset[0];
-	}
-
-	/* create a linear temporary that we can render into if necessary.
-	 * Note that ns->pitch is always a multiple of 64 for linear surfaces and swizzled surfaces are POT, so
-	 * ns->pitch & 63 is equivalent to (ns->pitch < 64 && swizzled)*/
-	if((ns->pitch & 63) && (ns->base.usage & (PIPE_BUFFER_USAGE_GPU_WRITE | NOUVEAU_BUFFER_USAGE_NO_RENDER)) == PIPE_BUFFER_USAGE_GPU_WRITE)
-		return &nv04_surface_wrap_for_render(pscreen, ((struct nv30_screen*)pscreen)->eng2d, ns)->base;
-
-	return &ns->base;
-}
-
-static void
-nv30_miptree_surface_del(struct pipe_surface *ps)
-{
-	struct nv04_surface* ns = (struct nv04_surface*)ps;
-	if(ns->backing)
-	{
-		struct nv30_screen* screen = (struct nv30_screen*)ps->texture->screen;
-		if(ns->backing->base.usage & PIPE_BUFFER_USAGE_GPU_WRITE)
-			screen->eng2d->copy(screen->eng2d, &ns->backing->base, 0, 0, ps, 0, 0, ns->base.width, ns->base.height);
-		nv30_miptree_surface_del(&ns->backing->base);
-	}
-
-	pipe_texture_reference(&ps->texture, NULL);
-	FREE(ps);
-}
-
-void
-nv30_screen_init_miptree_functions(struct pipe_screen *pscreen)
-{
-	pscreen->texture_create = nv30_miptree_create;
-	pscreen->texture_destroy = nv30_miptree_destroy;
-	pscreen->get_tex_surface = nv30_miptree_surface_new;
-	pscreen->tex_surface_destroy = nv30_miptree_surface_del;
-
-	nouveau_screen(pscreen)->texture_blanket = nv30_miptree_blanket;
-}
diff --git a/src/gallium/drivers/nv30/nv30_query.c b/src/gallium/drivers/nv30/nv30_query.c
deleted file mode 100644
index e27e9ccbf6..0000000000
--- a/src/gallium/drivers/nv30/nv30_query.c
+++ /dev/null
@@ -1,127 +0,0 @@
-#include "pipe/p_context.h"
-
-#include "nv30_context.h"
-
-struct nv30_query {
-	struct nouveau_resource *object;
-	unsigned type;
-	boolean ready;
-	uint64_t result;
-};
-
-static INLINE struct nv30_query *
-nv30_query(struct pipe_query *pipe)
-{
-	return (struct nv30_query *)pipe;
-}
-
-static struct pipe_query *
-nv30_query_create(struct pipe_context *pipe, unsigned query_type)
-{
-	struct nv30_query *q;
-
-	q = CALLOC(1, sizeof(struct nv30_query));
-	q->type = query_type;
-
-	return (struct pipe_query *)q;
-}
-
-static void
-nv30_query_destroy(struct pipe_context *pipe, struct pipe_query *pq)
-{
-	struct nv30_query *q = nv30_query(pq);
-
-	if (q->object)
-		nouveau_resource_free(&q->object);
-	FREE(q);
-}
-
-static void
-nv30_query_begin(struct pipe_context *pipe, struct pipe_query *pq)
-{
-	struct nv30_context *nv30 = nv30_context(pipe);
-	struct nv30_query *q = nv30_query(pq);
-	struct nv30_screen *screen = nv30->screen;
-	struct nouveau_channel *chan = screen->base.channel;
-	struct nouveau_grobj *rankine = screen->rankine;
-
-	assert(q->type == PIPE_QUERY_OCCLUSION_COUNTER);
-
-	/* Happens when end_query() is called, then another begin_query()
-	 * without querying the result in-between.  For now we'll wait for
-	 * the existing query to notify completion, but it could be better.
-	 */
-	if (q->object) {
-		uint64_t tmp;
-		pipe->get_query_result(pipe, pq, 1, &tmp);
-	}
-
-	if (nouveau_resource_alloc(nv30->screen->query_heap, 1, NULL, &q->object))
-		assert(0);
-	nouveau_notifier_reset(nv30->screen->query, q->object->start);
-
-	BEGIN_RING(chan, rankine, NV34TCL_QUERY_RESET, 1);
-	OUT_RING  (chan, 1);
-	BEGIN_RING(chan, rankine, NV34TCL_QUERY_UNK17CC, 1);
-	OUT_RING  (chan, 1);
-
-	q->ready = FALSE;
-}
-
-static void
-nv30_query_end(struct pipe_context *pipe, struct pipe_query *pq)
-{
-	struct nv30_context *nv30 = nv30_context(pipe);
-	struct nv30_screen *screen = nv30->screen;
-	struct nouveau_channel *chan = screen->base.channel;
-	struct nouveau_grobj *rankine = screen->rankine;
-	struct nv30_query *q = nv30_query(pq);
-
-	BEGIN_RING(chan, rankine, NV34TCL_QUERY_GET, 1);
-	OUT_RING  (chan, (0x01 << NV34TCL_QUERY_GET_UNK24_SHIFT) |
-		   ((q->object->start * 32) << NV34TCL_QUERY_GET_OFFSET_SHIFT));
-	FIRE_RING(chan);
-}
-
-static boolean
-nv30_query_result(struct pipe_context *pipe, struct pipe_query *pq,
-		  boolean wait, uint64_t *result)
-{
-	struct nv30_context *nv30 = nv30_context(pipe);
-	struct nv30_query *q = nv30_query(pq);
-
-	assert(q->object && q->type == PIPE_QUERY_OCCLUSION_COUNTER);
-
-	if (!q->ready) {
-		unsigned status;
-
-		status = nouveau_notifier_status(nv30->screen->query,
-						 q->object->start);
-		if (status != NV_NOTIFY_STATE_STATUS_COMPLETED) {
-			if (wait == FALSE)
-				return FALSE;
-
-			nouveau_notifier_wait_status(nv30->screen->query,
-					q->object->start,
-					NV_NOTIFY_STATE_STATUS_COMPLETED, 0);
-		}
-
-		q->result = nouveau_notifier_return_val(nv30->screen->query,
-							q->object->start);
-		q->ready = TRUE;
-		nouveau_resource_free(&q->object);
-	}
-
-	*result = q->result;
-	return TRUE;
-}
-
-void
-nv30_init_query_functions(struct nv30_context *nv30)
-{
-	nv30->pipe.create_query = nv30_query_create;
-	nv30->pipe.destroy_query = nv30_query_destroy;
-	nv30->pipe.begin_query = nv30_query_begin;
-	nv30->pipe.end_query = nv30_query_end;
-	nv30->pipe.get_query_result = nv30_query_result;
-}
diff --git a/src/gallium/drivers/nv30/nv30_shader.h b/src/gallium/drivers/nv30/nv30_shader.h
deleted file mode 100644
index dd3a36f78f..0000000000
--- a/src/gallium/drivers/nv30/nv30_shader.h
+++ /dev/null
@@ -1,490 +0,0 @@
-#ifndef __NV30_SHADER_H__
-#define __NV30_SHADER_H__
-
-/* Vertex programs instruction set
- *
- * 128bit opcodes, split into 4 32-bit ones for ease of use.
- *
- * Non-native instructions
- *   ABS - MOV + NV40_VP_INST0_DEST_ABS
- *   POW - EX2 + MUL + LG2
- *   SUB - ADD, second source negated
- *   SWZ - MOV
- *   XPD -  
- *
- * Register access
- *   - Only one INPUT can be accessed per-instruction (move extras into TEMPs)
- *   - Only one CONST can be accessed per-instruction (move extras into TEMPs)
- *
- * Relative Addressing
- *   According to the value returned for
- *   MAX_PROGRAM_NATIVE_ADDRESS_REGISTERS_ARB
- *
- *   there are only two address registers available.  The destination in the
- *   ARL instruction is set to TEMP <n> (The temp isn't actually written).
- *
- *   When using vanilla ARB_v_p, the proprietary driver will squish both the
- *   available ADDRESS regs into the first hardware reg in the X and Y
- *   components.
- *
- *   To use an address reg as an index into consts, the CONST_SRC is set to
- *   (const_base + offset) and INDEX_CONST is set.
- *
- *   To access the second address reg use ADDR_REG_SELECT_1. A particular
- *   component of the address regs is selected with ADDR_SWZ.
- *
- *   Only one address register can be accessed per instruction.
- *
- * Conditional execution (see NV_vertex_program{2,3} for details) Conditional
- * execution of an instruction is enabled by setting COND_TEST_ENABLE, and
- * selecting the condition which will allow the test to pass with
- * COND_{FL,LT,...}.  It is possible to swizzle the values in the condition
- * register, which allows for testing against an individual component.
- *
- * Branching:
- *
- *   The BRA/CAL instructions seem to follow a slightly different opcode
- *   layout.  The destination instruction ID (IADDR) overlaps a source field.
- *   Instruction ID's seem to be numbered based on the UPLOAD_FROM_ID FIFO
- *   command, and is incremented automatically on each UPLOAD_INST FIFO
- *   command.
- *
- *   Conditional branching is achieved by using the condition tests described
- *   above.  There doesn't appear to be dedicated looping instructions, but
- *   this can be done using a temp reg + conditional branching.
- *
- *   Subroutines may be uploaded before the main program itself, but the first
- *   executed instruction is determined by the PROGRAM_START_ID FIFO command.
- *
- */
-
-/* DWORD 0 */
-
-#define NV30_VP_INST_ADDR_REG_SELECT_1        (1 << 24)
-#define NV30_VP_INST_SRC2_ABS           (1 << 23) /* guess */
-#define NV30_VP_INST_SRC1_ABS           (1 << 22) /* guess */
-#define NV30_VP_INST_SRC0_ABS           (1 << 21) /* guess */
-#define NV30_VP_INST_VEC_RESULT         (1 << 20)
-#define NV30_VP_INST_DEST_TEMP_ID_SHIFT        16
-#define NV30_VP_INST_DEST_TEMP_ID_MASK        (0x0F << 16)
-#define NV30_VP_INST_COND_UPDATE_ENABLE        (1<<15)
-#define NV30_VP_INST_VEC_DEST_TEMP_MASK      (0xF << 16)
-#define NV30_VP_INST_COND_TEST_ENABLE        (1<<14)
-#define NV30_VP_INST_COND_SHIFT          11
-#define NV30_VP_INST_COND_MASK          (0x07 << 11)
-#  define NV30_VP_INST_COND_FL  0 /* guess */  
-#  define NV30_VP_INST_COND_LT  1  
-#  define NV30_VP_INST_COND_EQ  2
-#  define NV30_VP_INST_COND_LE  3
-#  define NV30_VP_INST_COND_GT  4
-#  define NV30_VP_INST_COND_NE  5
-#  define NV30_VP_INST_COND_GE  6
-#  define NV30_VP_INST_COND_TR  7 /* guess */
-#define NV30_VP_INST_COND_SWZ_X_SHIFT        9
-#define NV30_VP_INST_COND_SWZ_X_MASK        (0x03 <<  9)
-#define NV30_VP_INST_COND_SWZ_Y_SHIFT        7
-#define NV30_VP_INST_COND_SWZ_Y_MASK        (0x03 <<  7)
-#define NV30_VP_INST_COND_SWZ_Z_SHIFT        5
-#define NV30_VP_INST_COND_SWZ_Z_MASK        (0x03 <<  5)
-#define NV30_VP_INST_COND_SWZ_W_SHIFT        3
-#define NV30_VP_INST_COND_SWZ_W_MASK        (0x03 <<  3)
-#define NV30_VP_INST_COND_SWZ_ALL_SHIFT        3
-#define NV30_VP_INST_COND_SWZ_ALL_MASK        (0xFF <<  3)
-#define NV30_VP_INST_ADDR_SWZ_SHIFT        1
-#define NV30_VP_INST_ADDR_SWZ_MASK        (0x03 <<  1)
-#define NV30_VP_INST_SCA_OPCODEH_SHIFT        0
-#define NV30_VP_INST_SCA_OPCODEH_MASK        (0x01 <<  0)
-
-/* DWORD 1 */
-#define NV30_VP_INST_SCA_OPCODEL_SHIFT        28
-#define NV30_VP_INST_SCA_OPCODEL_MASK        (0x0F << 28)
-#  define NV30_VP_INST_OP_NOP  0x00
-#  define NV30_VP_INST_OP_RCP  0x02
-#  define NV30_VP_INST_OP_RCC  0x03
-#  define NV30_VP_INST_OP_RSQ  0x04
-#  define NV30_VP_INST_OP_EXP  0x05
-#  define NV30_VP_INST_OP_LOG  0x06
-#  define NV30_VP_INST_OP_LIT  0x07
-#  define NV30_VP_INST_OP_BRA  0x09
-#  define NV30_VP_INST_OP_CAL  0x0B
-#  define NV30_VP_INST_OP_RET  0x0C
-#  define NV30_VP_INST_OP_LG2  0x0D
-#  define NV30_VP_INST_OP_EX2  0x0E
-#  define NV30_VP_INST_OP_SIN  0x0F
-#  define NV30_VP_INST_OP_COS  0x10
-#define NV30_VP_INST_VEC_OPCODE_SHIFT        23
-#define NV30_VP_INST_VEC_OPCODE_MASK        (0x1F << 23)
-#  define NV30_VP_INST_OP_NOPV  0x00
-#  define NV30_VP_INST_OP_MOV  0x01
-#  define NV30_VP_INST_OP_MUL  0x02
-#  define NV30_VP_INST_OP_ADD  0x03
-#  define NV30_VP_INST_OP_MAD  0x04
-#  define NV30_VP_INST_OP_DP3  0x05
-#  define NV30_VP_INST_OP_DP4  0x07
-#  define NV30_VP_INST_OP_DPH  0x06
-#  define NV30_VP_INST_OP_DST  0x08
-#  define NV30_VP_INST_OP_MIN  0x09
-#  define NV30_VP_INST_OP_MAX  0x0A
-#  define NV30_VP_INST_OP_SLT  0x0B
-#  define NV30_VP_INST_OP_SGE  0x0C
-#  define NV30_VP_INST_OP_ARL  0x0D
-#  define NV30_VP_INST_OP_FRC  0x0E
-#  define NV30_VP_INST_OP_FLR  0x0F
-#  define NV30_VP_INST_OP_SEQ  0x10
-#  define NV30_VP_INST_OP_SFL  0x11
-#  define NV30_VP_INST_OP_SGT  0x12
-#  define NV30_VP_INST_OP_SLE  0x13
-#  define NV30_VP_INST_OP_SNE  0x14
-#  define NV30_VP_INST_OP_STR  0x15
-#  define NV30_VP_INST_OP_SSG  0x16
-#  define NV30_VP_INST_OP_ARR  0x17
-#  define NV30_VP_INST_OP_ARA  0x18
-#define NV30_VP_INST_CONST_SRC_SHIFT        14
-#define NV30_VP_INST_CONST_SRC_MASK        (0xFF << 14)
-#define NV30_VP_INST_INPUT_SRC_SHIFT        9    /*NV20*/
-#define NV30_VP_INST_INPUT_SRC_MASK        (0x0F <<  9)  /*NV20*/
-#  define NV30_VP_INST_IN_POS  0    /* These seem to match the bindings specified in */
-#  define NV30_VP_INST_IN_WEIGHT  1    /* the ARB_v_p spec (2.14.3.1) */
-#  define NV30_VP_INST_IN_NORMAL  2    
-#  define NV30_VP_INST_IN_COL0  3    /* Should probably confirm them all though */
-#  define NV30_VP_INST_IN_COL1  4
-#  define NV30_VP_INST_IN_FOGC  5
-#  define NV30_VP_INST_IN_TC0  8
-#  define NV30_VP_INST_IN_TC(n)  (8+n)
-#define NV30_VP_INST_SRC0H_SHIFT        0    /*NV20*/
-#define NV30_VP_INST_SRC0H_MASK          (0x1FF << 0)  /*NV20*/
-
-/* Please note: the IADDR fields overlap other fields because they are used
- * only for branch instructions.  See Branching: label above
- *
- * DWORD 2
- */
-#define NV30_VP_INST_SRC0L_SHIFT        26    /*NV20*/
-#define NV30_VP_INST_SRC0L_MASK         (0x3F  <<26)  /* NV30_VP_SRC0_LOW_MASK << 26 */
-#define NV30_VP_INST_SRC1_SHIFT         11    /*NV20*/
-#define NV30_VP_INST_SRC1_MASK          (0x7FFF<<11)  /*NV20*/
-#define NV30_VP_INST_SRC2H_SHIFT        0    /*NV20*/
-#define NV30_VP_INST_SRC2H_MASK          (0x7FF << 0)  /* NV30_VP_SRC2_HIGH_MASK >> 4*/
-#define NV30_VP_INST_IADDR_SHIFT        2
-#define NV30_VP_INST_IADDR_MASK          (0xF <<  28)   /* NV30_VP_SRC2_LOW_MASK << 28 */
-
-/* DWORD 3 */
-#define NV30_VP_INST_SRC2L_SHIFT        28    /*NV20*/
-#define NV30_VP_INST_SRC2L_MASK          (0x0F  <<28)  /*NV20*/
-#define NV30_VP_INST_STEMP_WRITEMASK_SHIFT      24
-#define NV30_VP_INST_STEMP_WRITEMASK_MASK      (0x0F << 24)
-#define NV30_VP_INST_VTEMP_WRITEMASK_SHIFT      20
-#define NV30_VP_INST_VTEMP_WRITEMASK_MASK      (0x0F << 20)
-#define NV30_VP_INST_SDEST_WRITEMASK_SHIFT      16
-#define NV30_VP_INST_SDEST_WRITEMASK_MASK      (0x0F << 16)
-#define NV30_VP_INST_VDEST_WRITEMASK_SHIFT      12    /*NV20*/
-#define NV30_VP_INST_VDEST_WRITEMASK_MASK      (0x0F << 12)  /*NV20*/
-#define NV30_VP_INST_DEST_SHIFT        2
-#define NV30_VP_INST_DEST_MASK        (0x0F <<  2)
-#  define NV30_VP_INST_DEST_POS  0
-#  define NV30_VP_INST_DEST_BFC0  1
-#  define NV30_VP_INST_DEST_BFC1  2
-#  define NV30_VP_INST_DEST_COL0  3
-#  define NV30_VP_INST_DEST_COL1  4
-#  define NV30_VP_INST_DEST_FOGC  5
-#  define NV30_VP_INST_DEST_PSZ   6
-#  define NV30_VP_INST_DEST_TC(n)  (8+n)
-
-#define NV30_VP_INST_LAST                           (1 << 0)
-
-/* Useful to split the source selection regs into their pieces */
-#define NV30_VP_SRC0_HIGH_SHIFT                                                6
-#define NV30_VP_SRC0_HIGH_MASK                                        0x00007FC0
-#define NV30_VP_SRC0_LOW_MASK                                         0x0000003F
-#define NV30_VP_SRC2_HIGH_SHIFT                                                4
-#define NV30_VP_SRC2_HIGH_MASK                                        0x00007FF0
-#define NV30_VP_SRC2_LOW_MASK                                         0x0000000F
-
-
-/* Source-register definition - matches NV20 exactly */
-#define NV30_VP_SRC_NEGATE          (1<<14)
-#define NV30_VP_SRC_SWZ_X_SHIFT        12
-#define NV30_VP_SRC_REG_SWZ_X_MASK        (0x03  <<12)
-#define NV30_VP_SRC_SWZ_Y_SHIFT        10
-#define NV30_VP_SRC_REG_SWZ_Y_MASK        (0x03  <<10)
-#define NV30_VP_SRC_SWZ_Z_SHIFT        8
-#define NV30_VP_SRC_REG_SWZ_Z_MASK        (0x03  << 8)
-#define NV30_VP_SRC_SWZ_W_SHIFT        6
-#define NV30_VP_SRC_REG_SWZ_W_MASK        (0x03  << 6)
-#define NV30_VP_SRC_REG_SWZ_ALL_SHIFT        6
-#define NV30_VP_SRC_REG_SWZ_ALL_MASK        (0xFF  << 6)
-#define NV30_VP_SRC_TEMP_SRC_SHIFT        2
-#define NV30_VP_SRC_REG_TEMP_ID_MASK        (0x0F  << 0)
-#define NV30_VP_SRC_REG_TYPE_SHIFT        0
-#define NV30_VP_SRC_REG_TYPE_MASK        (0x03  << 0)
-#define NV30_VP_SRC_REG_TYPE_TEMP  1
-#define NV30_VP_SRC_REG_TYPE_INPUT  2
-#define NV30_VP_SRC_REG_TYPE_CONST  3 /* guess */
-
-/*
- * Each fragment program opcode appears to be comprised of 4 32-bit values.
- *
- *   0 - Opcode, output reg/mask, ATTRIB source
- *   1 - Source 0
- *   2 - Source 1
- *   3 - Source 2
- *
- * There appears to be no special difference between result regs and temp regs.
- *     result.color == R0.xyzw
- *     result.depth == R1.z
- * When the fragprog contains instructions to write depth, NV30_TCL_PRIMITIVE_3D_UNK1D78=0
- * otherwise it is set to 1.
- *
- * Constants are inserted directly after the instruction that uses them.
- * 
- * It appears that it's not possible to use two input registers in one
- * instruction as the input sourcing is done in the instruction dword
- * and not the source selection dwords.  As such instructions such as:
- * 
- *     ADD result.color, fragment.color, fragment.texcoord[0];
- *
- * must be split into two MOV's and then an ADD (nvidia does this) but
- * I'm not sure why it's not just one MOV and then source the second input
- * in the ADD instruction..
- *
- * Negation of the full source is done with NV30_FP_REG_NEGATE, arbitrary
- * negation requires multiplication with a const.
- *
- * Arbitrary swizzling is supported with the exception of SWIZZLE_ZERO/SWIZZLE_ONE
- * The temp/result regs appear to be initialised to (0.0, 0.0, 0.0, 0.0) as SWIZZLE_ZERO
- * is implemented simply by not writing to the relevant components of the destination.
- *
- * Conditional execution
- *   TODO
- * 
- * Non-native instructions:
- *   LIT
- *   LRP - MAD+MAD
- *   SUB - ADD, negate second source
- *   RSQ - LG2 + EX2
- *   POW - LG2 + MUL + EX2
- *   SCS - COS + SIN
- *   XPD
- */
-
-//== Opcode / Destination selection ==
-#define NV30_FP_OP_PROGRAM_END          (1 << 0)
-#define NV30_FP_OP_OUT_REG_SHIFT        1
-#define NV30_FP_OP_OUT_REG_MASK          (31 << 1)  /* uncertain */
-/* Needs to be set when writing outputs to get expected result.. */
-#define NV30_FP_OP_OUT_REG_HALF          (1 << 7)
-#define NV30_FP_OP_COND_WRITE_ENABLE        (1 << 8)
-#define NV30_FP_OP_OUTMASK_SHIFT        9
-#define NV30_FP_OP_OUTMASK_MASK          (0xF << 9)
-#  define NV30_FP_OP_OUT_X  (1<<9)
-#  define NV30_FP_OP_OUT_Y  (1<<10)
-#  define NV30_FP_OP_OUT_Z  (1<<11)
-#  define NV30_FP_OP_OUT_W  (1<<12)
-/* Uncertain about these, especially the input_src values.. it's possible that
- * they can be dynamically changed.
- */
-#define NV30_FP_OP_INPUT_SRC_SHIFT        13
-#define NV30_FP_OP_INPUT_SRC_MASK        (15 << 13)
-#  define NV30_FP_OP_INPUT_SRC_POSITION  0x0
-#  define NV30_FP_OP_INPUT_SRC_COL0  0x1
-#  define NV30_FP_OP_INPUT_SRC_COL1  0x2
-#  define NV30_FP_OP_INPUT_SRC_FOGC  0x3
-#  define NV30_FP_OP_INPUT_SRC_TC0    0x4
-#  define NV30_FP_OP_INPUT_SRC_TC(n)  (0x4 + n)
-#define NV30_FP_OP_TEX_UNIT_SHIFT        17
-#define NV30_FP_OP_TEX_UNIT_MASK        (0xF << 17) /* guess */
-#define NV30_FP_OP_PRECISION_SHIFT        22
-#define NV30_FP_OP_PRECISION_MASK        (3 << 22)
-#   define NV30_FP_PRECISION_FP32  0
-#   define NV30_FP_PRECISION_FP16  1
-#   define NV30_FP_PRECISION_FX12  2
-#define NV30_FP_OP_OPCODE_SHIFT          24
-#define NV30_FP_OP_OPCODE_MASK          (0x3F << 24)
-#  define NV30_FP_OP_OPCODE_NOP  0x00
-#  define NV30_FP_OP_OPCODE_MOV  0x01
-#  define NV30_FP_OP_OPCODE_MUL  0x02
-#  define NV30_FP_OP_OPCODE_ADD  0x03
-#  define NV30_FP_OP_OPCODE_MAD  0x04
-#  define NV30_FP_OP_OPCODE_DP3  0x05
-#  define NV30_FP_OP_OPCODE_DP4  0x06
-#  define NV30_FP_OP_OPCODE_DST  0x07
-#  define NV30_FP_OP_OPCODE_MIN  0x08
-#  define NV30_FP_OP_OPCODE_MAX  0x09
-#  define NV30_FP_OP_OPCODE_SLT  0x0A
-#  define NV30_FP_OP_OPCODE_SGE  0x0B
-#  define NV30_FP_OP_OPCODE_SLE  0x0C
-#  define NV30_FP_OP_OPCODE_SGT  0x0D
-#  define NV30_FP_OP_OPCODE_SNE  0x0E
-#  define NV30_FP_OP_OPCODE_SEQ  0x0F
-#  define NV30_FP_OP_OPCODE_FRC  0x10
-#  define NV30_FP_OP_OPCODE_FLR  0x11
-#  define NV30_FP_OP_OPCODE_KIL  0x12
-#  define NV30_FP_OP_OPCODE_PK4B   0x13
-#  define NV30_FP_OP_OPCODE_UP4B   0x14
-#  define NV30_FP_OP_OPCODE_DDX  0x15 /* can only write XY */
-#  define NV30_FP_OP_OPCODE_DDY  0x16 /* can only write XY */
-#  define NV30_FP_OP_OPCODE_TEX  0x17
-#  define NV30_FP_OP_OPCODE_TXP  0x18
-#  define NV30_FP_OP_OPCODE_TXD  0x19
-#  define NV30_FP_OP_OPCODE_RCP  0x1A
-#  define NV30_FP_OP_OPCODE_RSQ  0x1B
-#  define NV30_FP_OP_OPCODE_EX2  0x1C
-#  define NV30_FP_OP_OPCODE_LG2  0x1D
-#  define NV30_FP_OP_OPCODE_LIT  0x1E
-#  define NV30_FP_OP_OPCODE_LRP  0x1F
-#  define NV30_FP_OP_OPCODE_STR  0x20 
-#  define NV30_FP_OP_OPCODE_SFL  0x21
-#  define NV30_FP_OP_OPCODE_COS  0x22
-#  define NV30_FP_OP_OPCODE_SIN  0x23
-#  define NV30_FP_OP_OPCODE_PK2H   0x24
-#  define NV30_FP_OP_OPCODE_UP2H   0x25
-#  define NV30_FP_OP_OPCODE_POW  0x26
-#  define NV30_FP_OP_OPCODE_PK4UB  0x27
-#  define NV30_FP_OP_OPCODE_UP4UB  0x28
-#  define NV30_FP_OP_OPCODE_PK2US  0x29
-#  define NV30_FP_OP_OPCODE_UP2US  0x2A
-#  define NV30_FP_OP_OPCODE_DP2A   0x2E
-#  define NV30_FP_OP_OPCODE_TXB  0x31
-#  define NV30_FP_OP_OPCODE_RFL  0x36
-#  define NV30_FP_OP_OPCODE_DIV  0x3A
-#define NV30_FP_OP_OUT_SAT          (1 << 31)
-
-/* high order bits of SRC0 */
-#define NV30_FP_OP_OUT_ABS          (1 << 29)
-#define NV30_FP_OP_COND_SWZ_W_SHIFT        27
-#define NV30_FP_OP_COND_SWZ_W_MASK        (3 << 27)
-#define NV30_FP_OP_COND_SWZ_Z_SHIFT        25
-#define NV30_FP_OP_COND_SWZ_Z_MASK        (3 << 25)
-#define NV30_FP_OP_COND_SWZ_Y_SHIFT        23
-#define NV30_FP_OP_COND_SWZ_Y_MASK        (3 << 23)
-#define NV30_FP_OP_COND_SWZ_X_SHIFT        21
-#define NV30_FP_OP_COND_SWZ_X_MASK        (3 << 21)
-#define NV30_FP_OP_COND_SWZ_ALL_SHIFT        21
-#define NV30_FP_OP_COND_SWZ_ALL_MASK        (0xFF << 21)
-#define NV30_FP_OP_COND_SHIFT          18
-#define NV30_FP_OP_COND_MASK          (0x07 << 18)
-#  define NV30_FP_OP_COND_FL  0
-#  define NV30_FP_OP_COND_LT  1
-#  define NV30_FP_OP_COND_EQ  2
-#  define NV30_FP_OP_COND_LE  3
-#  define NV30_FP_OP_COND_GT  4
-#  define NV30_FP_OP_COND_NE  5
-#  define NV30_FP_OP_COND_GE  6
-#  define NV30_FP_OP_COND_TR  7
-
-/* high order bits of SRC1 */
-#define NV30_FP_OP_DST_SCALE_SHIFT        28
-#define NV30_FP_OP_DST_SCALE_MASK        (3 << 28)
-#define NV30_FP_OP_DST_SCALE_1X                                                0
-#define NV30_FP_OP_DST_SCALE_2X                                                1
-#define NV30_FP_OP_DST_SCALE_4X                                                2
-#define NV30_FP_OP_DST_SCALE_8X                                                3
-#define NV30_FP_OP_DST_SCALE_INV_2X                                            5
-#define NV30_FP_OP_DST_SCALE_INV_4X                                            6
-#define NV30_FP_OP_DST_SCALE_INV_8X                                            7
-
-
-/* high order bits of SRC2 */
-#define NV30_FP_OP_INDEX_INPUT          (1 << 30)
-
-//== Register selection ==
-#define NV30_FP_REG_TYPE_SHIFT          0
-#define NV30_FP_REG_TYPE_MASK          (3 << 0)
-#  define NV30_FP_REG_TYPE_TEMP  0
-#  define NV30_FP_REG_TYPE_INPUT  1
-#  define NV30_FP_REG_TYPE_CONST  2
-#define NV30_FP_REG_SRC_SHIFT          2 /* uncertain */
-#define NV30_FP_REG_SRC_MASK          (31 << 2)
-#define NV30_FP_REG_SRC_HALF          (1 << 8)
-#define NV30_FP_REG_SWZ_ALL_SHIFT        9
-#define NV30_FP_REG_SWZ_ALL_MASK        (255 << 9)
-#define NV30_FP_REG_SWZ_X_SHIFT          9
-#define NV30_FP_REG_SWZ_X_MASK          (3 << 9)
-#define NV30_FP_REG_SWZ_Y_SHIFT          11
-#define NV30_FP_REG_SWZ_Y_MASK          (3 << 11)
-#define NV30_FP_REG_SWZ_Z_SHIFT          13
-#define NV30_FP_REG_SWZ_Z_MASK          (3 << 13)
-#define NV30_FP_REG_SWZ_W_SHIFT          15
-#define NV30_FP_REG_SWZ_W_MASK          (3 << 15)
-#  define NV30_FP_SWIZZLE_X  0
-#  define NV30_FP_SWIZZLE_Y  1
-#  define NV30_FP_SWIZZLE_Z  2
-#  define NV30_FP_SWIZZLE_W  3
-#define NV30_FP_REG_NEGATE          (1 << 17)
-
-#define NV30SR_NONE	0
-#define NV30SR_OUTPUT	1
-#define NV30SR_INPUT	2
-#define NV30SR_TEMP	3
-#define NV30SR_CONST	4
-
-struct nv30_sreg {
-	int type;
-	int index;
-
-	int dst_scale;
-
-	int negate;
-	int abs;
-	int swz[4];
-
-	int cc_update;
-	int cc_update_reg;
-	int cc_test;
-	int cc_test_reg;
-	int cc_swz[4];
-};
-
-static INLINE struct nv30_sreg
-nv30_sr(int type, int index)
-{
-	struct nv30_sreg temp = {
-		.type = type,
-		.index = index,
-		.dst_scale = DEF_SCALE,
-		.abs = 0,
-		.negate = 0,
-		.swz = { 0, 1, 2, 3 },
-		.cc_update = 0,
-		.cc_update_reg = 0,
-		.cc_test = DEF_CTEST,
-		.cc_test_reg = 0,
-		.cc_swz = { 0, 1, 2, 3 },
-	};
-	return temp;
-}
-
-static INLINE struct nv30_sreg
-nv30_sr_swz(struct nv30_sreg src, int x, int y, int z, int w)
-{
-	struct nv30_sreg dst = src;
-
-	dst.swz[SWZ_X] = src.swz[x];
-	dst.swz[SWZ_Y] = src.swz[y];
-	dst.swz[SWZ_Z] = src.swz[z];
-	dst.swz[SWZ_W] = src.swz[w];
-	return dst;
-}
-
-static INLINE struct nv30_sreg
-nv30_sr_neg(struct nv30_sreg src)
-{
-	src.negate = !src.negate;
-	return src;
-}
-
-static INLINE struct nv30_sreg
-nv30_sr_abs(struct nv30_sreg src)
-{
-	src.abs = 1;
-	return src;
-}
-
-static INLINE struct nv30_sreg
-nv30_sr_scale(struct nv30_sreg src, int scale)
-{
-	src.dst_scale = scale;
-	return src;
-}
-
-#endif
diff --git a/src/gallium/drivers/nv30/nv30_state.c b/src/gallium/drivers/nv30/nv30_state.c
deleted file mode 100644
index fb3075f4c8..0000000000
--- a/src/gallium/drivers/nv30/nv30_state.c
+++ /dev/null
@@ -1,782 +0,0 @@
-#include "pipe/p_state.h"
-#include "pipe/p_defines.h"
-#include "util/u_inlines.h"
-
-#include "tgsi/tgsi_parse.h"
-
-#include "nv30_context.h"
-#include "nv30_state.h"
-
-static void *
-nv30_blend_state_create(struct pipe_context *pipe,
-			const struct pipe_blend_state *cso)
-{
-	struct nv30_context *nv30 = nv30_context(pipe);
-	struct nouveau_grobj *rankine = nv30->screen->rankine;
-	struct nv30_blend_state *bso = CALLOC(1, sizeof(*bso));
-	struct nouveau_stateobj *so = so_new(5, 8, 0);
-
-	if (cso->rt[0].blend_enable) {
-		so_method(so, rankine, NV34TCL_BLEND_FUNC_ENABLE, 3);
-		so_data  (so, 1);
-		so_data  (so, (nvgl_blend_func(cso->rt[0].alpha_src_factor) << 16) |
-			       nvgl_blend_func(cso->rt[0].rgb_src_factor));
-		so_data  (so, nvgl_blend_func(cso->rt[0].alpha_dst_factor) << 16 |
-			      nvgl_blend_func(cso->rt[0].rgb_dst_factor));
-		/* FIXME: Gallium assumes GL_EXT_blend_func_separate.
-		   It is not the case for NV30 */
-		so_method(so, rankine, NV34TCL_BLEND_EQUATION, 1);
-		so_data  (so, nvgl_blend_eqn(cso->rt[0].rgb_func));
-	} else {
-		so_method(so, rankine, NV34TCL_BLEND_FUNC_ENABLE, 1);
-		so_data  (so, 0);
-	}
-
-	so_method(so, rankine, NV34TCL_COLOR_MASK, 1);
-	so_data  (so, (((cso->rt[0].colormask & PIPE_MASK_A) ? (0x01 << 24) : 0) |
-		       ((cso->rt[0].colormask & PIPE_MASK_R) ? (0x01 << 16) : 0) |
-		       ((cso->rt[0].colormask & PIPE_MASK_G) ? (0x01 <<  8) : 0) |
-		       ((cso->rt[0].colormask & PIPE_MASK_B) ? (0x01 <<  0) : 0)));
-
-	if (cso->logicop_enable) {
-		so_method(so, rankine, NV34TCL_COLOR_LOGIC_OP_ENABLE, 2);
-		so_data  (so, 1);
-		so_data  (so, nvgl_logicop_func(cso->logicop_func));
-	} else {
-		so_method(so, rankine, NV34TCL_COLOR_LOGIC_OP_ENABLE, 1);
-		so_data  (so, 0);
-	}
-
-	so_method(so, rankine, NV34TCL_DITHER_ENABLE, 1);
-	so_data  (so, cso->dither ? 1 : 0);
-
-	so_ref(so, &bso->so);
-	so_ref(NULL, &so);
-	bso->pipe = *cso;
-	return (void *)bso;
-}
-
-static void
-nv30_blend_state_bind(struct pipe_context *pipe, void *hwcso)
-{
-	struct nv30_context *nv30 = nv30_context(pipe);
-
-	nv30->blend = hwcso;
-	nv30->dirty |= NV30_NEW_BLEND;
-}
-
-static void
-nv30_blend_state_delete(struct pipe_context *pipe, void *hwcso)
-{
-	struct nv30_blend_state *bso = hwcso;
-
-	so_ref(NULL, &bso->so);
-	FREE(bso);
-}
-
-
-static INLINE unsigned
-wrap_mode(unsigned wrap) {
-	unsigned ret;
-
-	switch (wrap) {
-	case PIPE_TEX_WRAP_REPEAT:
-		ret = NV34TCL_TX_WRAP_S_REPEAT;
-		break;
-	case PIPE_TEX_WRAP_MIRROR_REPEAT:
-		ret = NV34TCL_TX_WRAP_S_MIRRORED_REPEAT;
-		break;
-	case PIPE_TEX_WRAP_CLAMP_TO_EDGE:
-		ret = NV34TCL_TX_WRAP_S_CLAMP_TO_EDGE;
-		break;
-	case PIPE_TEX_WRAP_CLAMP_TO_BORDER:
-		ret = NV34TCL_TX_WRAP_S_CLAMP_TO_BORDER;
-		break;
-	case PIPE_TEX_WRAP_CLAMP:
-		ret = NV34TCL_TX_WRAP_S_CLAMP;
-		break;
-/*	case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE:
-		ret = NV34TCL_TX_WRAP_S_MIRROR_CLAMP_TO_EDGE;
-		break;
-	case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER:
-		ret = NV34TCL_TX_WRAP_S_MIRROR_CLAMP_TO_BORDER;
-		break;
-	case PIPE_TEX_WRAP_MIRROR_CLAMP:
-		ret = NV34TCL_TX_WRAP_S_MIRROR_CLAMP;
-		break;*/
-	default:
-		NOUVEAU_ERR("unknown wrap mode: %d\n", wrap);
-		ret = NV34TCL_TX_WRAP_S_REPEAT;
-		break;
-	}
-
-	return ret >> NV34TCL_TX_WRAP_S_SHIFT;
-}
-
-static void *
-nv30_sampler_state_create(struct pipe_context *pipe,
-			  const struct pipe_sampler_state *cso)
-{
-	struct nv30_sampler_state *ps;
-	uint32_t filter = 0;
-
-	ps = MALLOC(sizeof(struct nv30_sampler_state));
-
-	ps->fmt = 0;
-	/* TODO: Not all RECTs formats have this bit set, bits 15-8 of format
-	   are the tx format to use. We should store normalized coord flag
-	   in sampler state structure, and set appropriate format in
-	   nvxx_fragtex_build()
-	 */
-	/*NV34TCL_TX_FORMAT_RECT*/
-	/*if (!cso->normalized_coords) {
-		ps->fmt |= (1<<14) ;
-	}*/
-
-	ps->wrap = ((wrap_mode(cso->wrap_s) << NV34TCL_TX_WRAP_S_SHIFT) |
-		    (wrap_mode(cso->wrap_t) << NV34TCL_TX_WRAP_T_SHIFT) |
-		    (wrap_mode(cso->wrap_r) << NV34TCL_TX_WRAP_R_SHIFT));
-
-	ps->en = 0;
-
-	if (cso->max_anisotropy >= 8) {
-		ps->en |= NV34TCL_TX_ENABLE_ANISO_8X;
-	} else
-	if (cso->max_anisotropy >= 4) {
-		ps->en |= NV34TCL_TX_ENABLE_ANISO_4X;
-	} else
-	if (cso->max_anisotropy >= 2) {
-		ps->en |= NV34TCL_TX_ENABLE_ANISO_2X;
-	}
-
-	switch (cso->mag_img_filter) {
-	case PIPE_TEX_FILTER_LINEAR:
-		filter |= NV34TCL_TX_FILTER_MAGNIFY_LINEAR;
-		break;
-	case PIPE_TEX_FILTER_NEAREST:
-	default:
-		filter |= NV34TCL_TX_FILTER_MAGNIFY_NEAREST;
-		break;
-	}
-
-	switch (cso->min_img_filter) {
-	case PIPE_TEX_FILTER_LINEAR:
-		switch (cso->min_mip_filter) {
-		case PIPE_TEX_MIPFILTER_NEAREST:
-			filter |= NV34TCL_TX_FILTER_MINIFY_LINEAR_MIPMAP_NEAREST;
-			break;
-		case PIPE_TEX_MIPFILTER_LINEAR:
-			filter |= NV34TCL_TX_FILTER_MINIFY_LINEAR_MIPMAP_LINEAR;
-			break;
-		case PIPE_TEX_MIPFILTER_NONE:
-		default:
-			filter |= NV34TCL_TX_FILTER_MINIFY_LINEAR;
-			break;
-		}
-		break;
-	case PIPE_TEX_FILTER_NEAREST:
-	default:
-		switch (cso->min_mip_filter) {
-		case PIPE_TEX_MIPFILTER_NEAREST:
-			filter |= NV34TCL_TX_FILTER_MINIFY_NEAREST_MIPMAP_NEAREST;
-		break;
-		case PIPE_TEX_MIPFILTER_LINEAR:
-			filter |= NV34TCL_TX_FILTER_MINIFY_NEAREST_MIPMAP_LINEAR;
-			break;
-		case PIPE_TEX_MIPFILTER_NONE:
-		default:
-			filter |= NV34TCL_TX_FILTER_MINIFY_NEAREST;
-			break;
-		}
-		break;
-	}
-
-	ps->filt = filter;
-
-	{
-		float limit;
-
-		limit = CLAMP(cso->lod_bias, -16.0, 15.0);
-		ps->filt |= (int)(cso->lod_bias * 256.0) & 0x1fff;
-
-		limit = CLAMP(cso->max_lod, 0.0, 15.0);
-		ps->en |= (int)(limit) << 14 /*NV34TCL_TX_ENABLE_MIPMAP_MAX_LOD_SHIFT*/;
-
-		limit = CLAMP(cso->min_lod, 0.0, 15.0);
-		ps->en |= (int)(limit) << 26 /*NV34TCL_TX_ENABLE_MIPMAP_MIN_LOD_SHIFT*/;
-	}
-
-	if (cso->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) {
-		switch (cso->compare_func) {
-		case PIPE_FUNC_NEVER:
-			ps->wrap |= NV34TCL_TX_WRAP_RCOMP_NEVER;
-			break;
-		case PIPE_FUNC_GREATER:
-			ps->wrap |= NV34TCL_TX_WRAP_RCOMP_GREATER;
-			break;
-		case PIPE_FUNC_EQUAL:
-			ps->wrap |= NV34TCL_TX_WRAP_RCOMP_EQUAL;
-			break;
-		case PIPE_FUNC_GEQUAL:
-			ps->wrap |= NV34TCL_TX_WRAP_RCOMP_GEQUAL;
-			break;
-		case PIPE_FUNC_LESS:
-			ps->wrap |= NV34TCL_TX_WRAP_RCOMP_LESS;
-			break;
-		case PIPE_FUNC_NOTEQUAL:
-			ps->wrap |= NV34TCL_TX_WRAP_RCOMP_NOTEQUAL;
-			break;
-		case PIPE_FUNC_LEQUAL:
-			ps->wrap |= NV34TCL_TX_WRAP_RCOMP_LEQUAL;
-			break;
-		case PIPE_FUNC_ALWAYS:
-			ps->wrap |= NV34TCL_TX_WRAP_RCOMP_ALWAYS;
-			break;
-		default:
-			break;
-		}
-	}
-
-	ps->bcol = ((float_to_ubyte(cso->border_color[3]) << 24) |
-		    (float_to_ubyte(cso->border_color[0]) << 16) |
-		    (float_to_ubyte(cso->border_color[1]) <<  8) |
-		    (float_to_ubyte(cso->border_color[2]) <<  0));
-
-	return (void *)ps;
-}
-
-static void
-nv30_sampler_state_bind(struct pipe_context *pipe, unsigned nr, void **sampler)
-{
-	struct nv30_context *nv30 = nv30_context(pipe);
-	unsigned unit;
-
-	for (unit = 0; unit < nr; unit++) {
-		nv30->tex_sampler[unit] = sampler[unit];
-		nv30->dirty_samplers |= (1 << unit);
-	}
-
-	for (unit = nr; unit < nv30->nr_samplers; unit++) {
-		nv30->tex_sampler[unit] = NULL;
-		nv30->dirty_samplers |= (1 << unit);
-	}
-
-	nv30->nr_samplers = nr;
-	nv30->dirty |= NV30_NEW_SAMPLER;
-}
-
-static void
-nv30_sampler_state_delete(struct pipe_context *pipe, void *hwcso)
-{
-	FREE(hwcso);
-}
-
-static void
-nv30_set_fragment_sampler_views(struct pipe_context *pipe,
-				unsigned nr,
-				struct pipe_sampler_view **views)
-{
-	struct nv30_context *nv30 = nv30_context(pipe);
-	unsigned unit;
-
-	for (unit = 0; unit < nr; unit++) {
-		pipe_sampler_view_reference(&nv30->fragment_sampler_views[unit], views[unit]);
-		pipe_texture_reference((struct pipe_texture **)
-				       &nv30->tex_miptree[unit], views[unit]->texture);
-		nv30->dirty_samplers |= (1 << unit);
-	}
-
-	for (unit = nr; unit < nv30->nr_textures; unit++) {
-		pipe_sampler_view_reference(&nv30->fragment_sampler_views[unit], NULL);
-		pipe_texture_reference((struct pipe_texture **)
-				       &nv30->tex_miptree[unit], NULL);
-		nv30->dirty_samplers |= (1 << unit);
-	}
-
-	nv30->nr_textures = nr;
-	nv30->dirty |= NV30_NEW_SAMPLER;
-}
-
-static struct pipe_sampler_view *
-nv30_create_sampler_view(struct pipe_context *pipe,
-			 struct pipe_texture *texture,
-			 const struct pipe_sampler_view *templ)
-{
-	struct pipe_sampler_view *view = CALLOC_STRUCT(pipe_sampler_view);
-
-	if (view) {
-		*view = *templ;
-		view->reference.count = 1;
-		view->texture = NULL;
-		pipe_texture_reference(&view->texture, texture);
-		view->context = pipe;
-	}
-
-	return view;
-}
-
-
-static void
-nv30_sampler_view_destroy(struct pipe_context *pipe,
-			  struct pipe_sampler_view *view)
-{
-	pipe_texture_reference(&view->texture, NULL);
-	FREE(view);
-}
-
-static void *
-nv30_rasterizer_state_create(struct pipe_context *pipe,
-			     const struct pipe_rasterizer_state *cso)
-{
-	struct nv30_context *nv30 = nv30_context(pipe);
-	struct nv30_rasterizer_state *rsso = CALLOC(1, sizeof(*rsso));
-	struct nouveau_stateobj *so = so_new(9, 19, 0);
-	struct nouveau_grobj *rankine = nv30->screen->rankine;
-
-	/*XXX: ignored:
-	 * 	light_twoside
-	 * 	point_smooth -nohw
-	 * 	multisample
-	 */
-
-	so_method(so, rankine, NV34TCL_SHADE_MODEL, 1);
-	so_data  (so, cso->flatshade ? NV34TCL_SHADE_MODEL_FLAT :
-				       NV34TCL_SHADE_MODEL_SMOOTH);
-
-	so_method(so, rankine, NV34TCL_LINE_WIDTH, 2);
-	so_data  (so, (unsigned char)(cso->line_width * 8.0) & 0xff);
-	so_data  (so, cso->line_smooth ? 1 : 0);
-	so_method(so, rankine, NV34TCL_LINE_STIPPLE_ENABLE, 2);
-	so_data  (so, cso->line_stipple_enable ? 1 : 0);
-	so_data  (so, (cso->line_stipple_pattern << 16) |
-		       cso->line_stipple_factor);
-
-	so_method(so, rankine, NV34TCL_POINT_SIZE, 1);
-	so_data  (so, fui(cso->point_size));
-
-	so_method(so, rankine, NV34TCL_POLYGON_MODE_FRONT, 6);
-	if (cso->front_winding == PIPE_WINDING_CCW) {
-		so_data(so, nvgl_polygon_mode(cso->fill_ccw));
-		so_data(so, nvgl_polygon_mode(cso->fill_cw));
-		switch (cso->cull_mode) {
-		case PIPE_WINDING_CCW:
-			so_data(so, NV34TCL_CULL_FACE_FRONT);
-			break;
-		case PIPE_WINDING_CW:
-			so_data(so, NV34TCL_CULL_FACE_BACK);
-			break;
-		case PIPE_WINDING_BOTH:
-			so_data(so, NV34TCL_CULL_FACE_FRONT_AND_BACK);
-			break;
-		default:
-			so_data(so, NV34TCL_CULL_FACE_BACK);
-			break;
-		}
-		so_data(so, NV34TCL_FRONT_FACE_CCW);
-	} else {
-		so_data(so, nvgl_polygon_mode(cso->fill_cw));
-		so_data(so, nvgl_polygon_mode(cso->fill_ccw));
-		switch (cso->cull_mode) {
-		case PIPE_WINDING_CCW:
-			so_data(so, NV34TCL_CULL_FACE_BACK);
-			break;
-		case PIPE_WINDING_CW:
-			so_data(so, NV34TCL_CULL_FACE_FRONT);
-			break;
-		case PIPE_WINDING_BOTH:
-			so_data(so, NV34TCL_CULL_FACE_FRONT_AND_BACK);
-			break;
-		default:
-			so_data(so, NV34TCL_CULL_FACE_BACK);
-			break;
-		}
-		so_data(so, NV34TCL_FRONT_FACE_CW);
-	}
-	so_data(so, cso->poly_smooth ? 1 : 0);
-	so_data(so, (cso->cull_mode != PIPE_WINDING_NONE) ? 1 : 0);
-
-	so_method(so, rankine, NV34TCL_POLYGON_STIPPLE_ENABLE, 1);
-	so_data  (so, cso->poly_stipple_enable ? 1 : 0);
-
-	so_method(so, rankine, NV34TCL_POLYGON_OFFSET_POINT_ENABLE, 3);
-	if ((cso->offset_cw && cso->fill_cw == PIPE_POLYGON_MODE_POINT) ||
-	    (cso->offset_ccw && cso->fill_ccw == PIPE_POLYGON_MODE_POINT))
-		so_data(so, 1);
-	else
-		so_data(so, 0);
-	if ((cso->offset_cw && cso->fill_cw == PIPE_POLYGON_MODE_LINE) ||
-	    (cso->offset_ccw && cso->fill_ccw == PIPE_POLYGON_MODE_LINE))
-		so_data(so, 1);
-	else
-		so_data(so, 0);
-	if ((cso->offset_cw && cso->fill_cw == PIPE_POLYGON_MODE_FILL) ||
-	    (cso->offset_ccw && cso->fill_ccw == PIPE_POLYGON_MODE_FILL))
-		so_data(so, 1);
-	else
-		so_data(so, 0);
-	if (cso->offset_cw || cso->offset_ccw) {
-		so_method(so, rankine, NV34TCL_POLYGON_OFFSET_FACTOR, 2);
-		so_data  (so, fui(cso->offset_scale));
-		so_data  (so, fui(cso->offset_units * 2));
-	}
-
-	so_method(so, rankine, NV34TCL_POINT_SPRITE, 1);
-	if (cso->point_quad_rasterization) {
-		unsigned psctl = (1 << 0), i;
-
-		for (i = 0; i < 8; i++) {
-			if ((cso->sprite_coord_enable >> i) & 1)
-				psctl |= (1 << (8 + i));
-		}
-
-		so_data(so, psctl);
-	} else {
-		so_data(so, 0);
-	}
-
-	so_ref(so, &rsso->so);
-	so_ref(NULL, &so);
-	rsso->pipe = *cso;
-	return (void *)rsso;
-}
-
-static void
-nv30_rasterizer_state_bind(struct pipe_context *pipe, void *hwcso)
-{
-	struct nv30_context *nv30 = nv30_context(pipe);
-
-	nv30->rasterizer = hwcso;
-	nv30->dirty |= NV30_NEW_RAST;
-	/*nv30->draw_dirty |= NV30_NEW_RAST;*/
-}
-
-static void
-nv30_rasterizer_state_delete(struct pipe_context *pipe, void *hwcso)
-{
-	struct nv30_rasterizer_state *rsso = hwcso;
-
-	so_ref(NULL, &rsso->so);
-	FREE(rsso);
-}
-
-static void *
-nv30_depth_stencil_alpha_state_create(struct pipe_context *pipe,
-			const struct pipe_depth_stencil_alpha_state *cso)
-{
-	struct nv30_context *nv30 = nv30_context(pipe);
-	struct nv30_zsa_state *zsaso = CALLOC(1, sizeof(*zsaso));
-	struct nouveau_stateobj *so = so_new(6, 20, 0);
-	struct nouveau_grobj *rankine = nv30->screen->rankine;
-
-	so_method(so, rankine, NV34TCL_DEPTH_FUNC, 3);
-	so_data  (so, nvgl_comparison_op(cso->depth.func));
-	so_data  (so, cso->depth.writemask ? 1 : 0);
-	so_data  (so, cso->depth.enabled ? 1 : 0);
-
-	so_method(so, rankine, NV34TCL_ALPHA_FUNC_ENABLE, 3);
-	so_data  (so, cso->alpha.enabled ? 1 : 0);
-	so_data  (so, nvgl_comparison_op(cso->alpha.func));
-	so_data  (so, float_to_ubyte(cso->alpha.ref_value));
-
-	if (cso->stencil[0].enabled) {
-		so_method(so, rankine, NV34TCL_STENCIL_FRONT_ENABLE, 3);
-		so_data  (so, cso->stencil[0].enabled ? 1 : 0);
-		so_data  (so, cso->stencil[0].writemask);
-		so_data  (so, nvgl_comparison_op(cso->stencil[0].func));
-		so_method(so, rankine, NV34TCL_STENCIL_FRONT_FUNC_MASK, 4);
-		so_data  (so, cso->stencil[0].valuemask);
-		so_data  (so, nvgl_stencil_op(cso->stencil[0].fail_op));
-		so_data  (so, nvgl_stencil_op(cso->stencil[0].zfail_op));
-		so_data  (so, nvgl_stencil_op(cso->stencil[0].zpass_op));
-	} else {
-		so_method(so, rankine, NV34TCL_STENCIL_FRONT_ENABLE, 1);
-		so_data  (so, 0);
-	}
-
-	if (cso->stencil[1].enabled) {
-		so_method(so, rankine, NV34TCL_STENCIL_BACK_ENABLE, 3);
-		so_data  (so, cso->stencil[1].enabled ? 1 : 0);
-		so_data  (so, cso->stencil[1].writemask);
-		so_data  (so, nvgl_comparison_op(cso->stencil[1].func));
-		so_method(so, rankine, NV34TCL_STENCIL_BACK_FUNC_MASK, 4);
-		so_data  (so, cso->stencil[1].valuemask);
-		so_data  (so, nvgl_stencil_op(cso->stencil[1].fail_op));
-		so_data  (so, nvgl_stencil_op(cso->stencil[1].zfail_op));
-		so_data  (so, nvgl_stencil_op(cso->stencil[1].zpass_op));
-	} else {
-		so_method(so, rankine, NV34TCL_STENCIL_BACK_ENABLE, 1);
-		so_data  (so, 0);
-	}
-
-	so_ref(so, &zsaso->so);
-	so_ref(NULL, &so);
-	zsaso->pipe = *cso;
-	return (void *)zsaso;
-}
-
-static void
-nv30_depth_stencil_alpha_state_bind(struct pipe_context *pipe, void *hwcso)
-{
-	struct nv30_context *nv30 = nv30_context(pipe);
-
-	nv30->zsa = hwcso;
-	nv30->dirty |= NV30_NEW_ZSA;
-}
-
-static void
-nv30_depth_stencil_alpha_state_delete(struct pipe_context *pipe, void *hwcso)
-{
-	struct nv30_zsa_state *zsaso = hwcso;
-
-	so_ref(NULL, &zsaso->so);
-	FREE(zsaso);
-}
-
-static void *
-nv30_vp_state_create(struct pipe_context *pipe,
-		     const struct pipe_shader_state *cso)
-{
-	/*struct nv30_context *nv30 = nv30_context(pipe);*/
-	struct nv30_vertex_program *vp;
-
-	vp = CALLOC(1, sizeof(struct nv30_vertex_program));
-	vp->pipe.tokens = tgsi_dup_tokens(cso->tokens);
-	/*vp->draw = draw_create_vertex_shader(nv30->draw, &vp->pipe);*/
-
-	return (void *)vp;
-}
-
-static void
-nv30_vp_state_bind(struct pipe_context *pipe, void *hwcso)
-{
-	struct nv30_context *nv30 = nv30_context(pipe);
-
-	nv30->vertprog = hwcso;
-	nv30->dirty |= NV30_NEW_VERTPROG;
-	/*nv30->draw_dirty |= NV30_NEW_VERTPROG;*/
-}
-
-static void
-nv30_vp_state_delete(struct pipe_context *pipe, void *hwcso)
-{
-	struct nv30_context *nv30 = nv30_context(pipe);
-	struct nv30_vertex_program *vp = hwcso;
-
-	/*draw_delete_vertex_shader(nv30->draw, vp->draw);*/
-	nv30_vertprog_destroy(nv30, vp);
-	FREE((void*)vp->pipe.tokens);
-	FREE(vp);
-}
-
-static void *
-nv30_fp_state_create(struct pipe_context *pipe,
-		     const struct pipe_shader_state *cso)
-{
-	struct nv30_fragment_program *fp;
-
-	fp = CALLOC(1, sizeof(struct nv30_fragment_program));
-	fp->pipe.tokens = tgsi_dup_tokens(cso->tokens);
-
-	tgsi_scan_shader(fp->pipe.tokens, &fp->info);
-
-	return (void *)fp;
-}
-
-static void
-nv30_fp_state_bind(struct pipe_context *pipe, void *hwcso)
-{
-	struct nv30_context *nv30 = nv30_context(pipe);
-
-	nv30->fragprog = hwcso;
-	nv30->dirty |= NV30_NEW_FRAGPROG;
-}
-
-static void
-nv30_fp_state_delete(struct pipe_context *pipe, void *hwcso)
-{
-	struct nv30_context *nv30 = nv30_context(pipe);
-	struct nv30_fragment_program *fp = hwcso;
-
-	nv30_fragprog_destroy(nv30, fp);
-	FREE((void*)fp->pipe.tokens);
-	FREE(fp);
-}
-
-static void
-nv30_set_blend_color(struct pipe_context *pipe,
-		     const struct pipe_blend_color *bcol)
-{
-	struct nv30_context *nv30 = nv30_context(pipe);
-
-	nv30->blend_colour = *bcol;
-	nv30->dirty |= NV30_NEW_BCOL;
-}
-
-static void
-nv30_set_stencil_ref(struct pipe_context *pipe,
-		     const struct pipe_stencil_ref *sr)
-{
-	struct nv30_context *nv30 = nv30_context(pipe);
-
-	nv30->stencil_ref = *sr;
-	nv30->dirty |= NV30_NEW_SR;
-}
-
-static void
-nv30_set_clip_state(struct pipe_context *pipe,
-		    const struct pipe_clip_state *clip)
-{
-}
-
-static void
-nv30_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index,
-			 struct pipe_buffer *buf )
-{
-	struct nv30_context *nv30 = nv30_context(pipe);
-
-	nv30->constbuf[shader] = buf;
-	nv30->constbuf_nr[shader] = buf->size / (4 * sizeof(float));
-
-	if (shader == PIPE_SHADER_VERTEX) {
-		nv30->dirty |= NV30_NEW_VERTPROG;
-	} else
-	if (shader == PIPE_SHADER_FRAGMENT) {
-		nv30->dirty |= NV30_NEW_FRAGPROG;
-	}
-}
-
-static void
-nv30_set_framebuffer_state(struct pipe_context *pipe,
-			   const struct pipe_framebuffer_state *fb)
-{
-	struct nv30_context *nv30 = nv30_context(pipe);
-
-	nv30->framebuffer = *fb;
-	nv30->dirty |= NV30_NEW_FB;
-}
-
-static void
-nv30_set_polygon_stipple(struct pipe_context *pipe,
-			 const struct pipe_poly_stipple *stipple)
-{
-	struct nv30_context *nv30 = nv30_context(pipe);
-
-	memcpy(nv30->stipple, stipple->stipple, 4 * 32);
-	nv30->dirty |= NV30_NEW_STIPPLE;
-}
-
-static void
-nv30_set_scissor_state(struct pipe_context *pipe,
-		       const struct pipe_scissor_state *s)
-{
-	struct nv30_context *nv30 = nv30_context(pipe);
-
-	nv30->scissor = *s;
-	nv30->dirty |= NV30_NEW_SCISSOR;
-}
-
-static void
-nv30_set_viewport_state(struct pipe_context *pipe,
-			const struct pipe_viewport_state *vpt)
-{
-	struct nv30_context *nv30 = nv30_context(pipe);
-
-	nv30->viewport = *vpt;
-	nv30->dirty |= NV30_NEW_VIEWPORT;
-	/*nv30->draw_dirty |= NV30_NEW_VIEWPORT;*/
-}
-
-static void
-nv30_set_vertex_buffers(struct pipe_context *pipe, unsigned count,
-			const struct pipe_vertex_buffer *vb)
-{
-	struct nv30_context *nv30 = nv30_context(pipe);
-
-	memcpy(nv30->vtxbuf, vb, sizeof(*vb) * count);
-	nv30->vtxbuf_nr = count;
-
-	nv30->dirty |= NV30_NEW_ARRAYS;
-	/*nv30->draw_dirty |= NV30_NEW_ARRAYS;*/
-}
-
-static void *
-nv30_vtxelts_state_create(struct pipe_context *pipe,
-			  unsigned num_elements,
-			  const struct pipe_vertex_element *elements)
-{
-	struct nv30_vtxelt_state *cso = CALLOC_STRUCT(nv30_vtxelt_state);
-
-	assert(num_elements < 16); /* not doing fallbacks yet */
-	cso->num_elements = num_elements;
-	memcpy(cso->pipe, elements, num_elements * sizeof(*elements));
-
-/*	nv30_vtxelt_construct(cso);*/
-
-	return (void *)cso;
-}
-
-static void
-nv30_vtxelts_state_delete(struct pipe_context *pipe, void *hwcso)
-{
-	FREE(hwcso);
-}
-
-static void
-nv30_vtxelts_state_bind(struct pipe_context *pipe, void *hwcso)
-{
-	struct nv30_context *nv30 = nv30_context(pipe);
-
-	nv30->vtxelt = hwcso;
-	nv30->dirty |= NV30_NEW_ARRAYS;
-	/*nv30->draw_dirty |= NV30_NEW_ARRAYS;*/
-}
-
-void
-nv30_init_state_functions(struct nv30_context *nv30)
-{
-	nv30->pipe.create_blend_state = nv30_blend_state_create;
-	nv30->pipe.bind_blend_state = nv30_blend_state_bind;
-	nv30->pipe.delete_blend_state = nv30_blend_state_delete;
-
-	nv30->pipe.create_sampler_state = nv30_sampler_state_create;
-	nv30->pipe.bind_fragment_sampler_states = nv30_sampler_state_bind;
-	nv30->pipe.delete_sampler_state = nv30_sampler_state_delete;
-	nv30->pipe.set_fragment_sampler_views = nv30_set_fragment_sampler_views;
-	nv30->pipe.create_sampler_view = nv30_create_sampler_view;
-	nv30->pipe.sampler_view_destroy = nv30_sampler_view_destroy;
-
-	nv30->pipe.create_rasterizer_state = nv30_rasterizer_state_create;
-	nv30->pipe.bind_rasterizer_state = nv30_rasterizer_state_bind;
-	nv30->pipe.delete_rasterizer_state = nv30_rasterizer_state_delete;
-
-	nv30->pipe.create_depth_stencil_alpha_state =
-		nv30_depth_stencil_alpha_state_create;
-	nv30->pipe.bind_depth_stencil_alpha_state =
-		nv30_depth_stencil_alpha_state_bind;
-	nv30->pipe.delete_depth_stencil_alpha_state =
-		nv30_depth_stencil_alpha_state_delete;
-
-	nv30->pipe.create_vs_state = nv30_vp_state_create;
-	nv30->pipe.bind_vs_state = nv30_vp_state_bind;
-	nv30->pipe.delete_vs_state = nv30_vp_state_delete;
-
-	nv30->pipe.create_fs_state = nv30_fp_state_create;
-	nv30->pipe.bind_fs_state = nv30_fp_state_bind;
-	nv30->pipe.delete_fs_state = nv30_fp_state_delete;
-
-	nv30->pipe.set_blend_color = nv30_set_blend_color;
-        nv30->pipe.set_stencil_ref = nv30_set_stencil_ref;
-	nv30->pipe.set_clip_state = nv30_set_clip_state;
-	nv30->pipe.set_constant_buffer = nv30_set_constant_buffer;
-	nv30->pipe.set_framebuffer_state = nv30_set_framebuffer_state;
-	nv30->pipe.set_polygon_stipple = nv30_set_polygon_stipple;
-	nv30->pipe.set_scissor_state = nv30_set_scissor_state;
-	nv30->pipe.set_viewport_state = nv30_set_viewport_state;
-
-	nv30->pipe.create_vertex_elements_state = nv30_vtxelts_state_create;
-	nv30->pipe.delete_vertex_elements_state = nv30_vtxelts_state_delete;
-	nv30->pipe.bind_vertex_elements_state = nv30_vtxelts_state_bind;
-
-	nv30->pipe.set_vertex_buffers = nv30_set_vertex_buffers;
-}
-
diff --git a/src/gallium/drivers/nv30/nv30_state.h b/src/gallium/drivers/nv30/nv30_state.h
deleted file mode 100644
index 66c26360cb..0000000000
--- a/src/gallium/drivers/nv30/nv30_state.h
+++ /dev/null
@@ -1,88 +0,0 @@
-#ifndef __NV30_STATE_H__
-#define __NV30_STATE_H__
-
-#include "pipe/p_state.h"
-#include "tgsi/tgsi_scan.h"
-
-struct nv30_sampler_state {
-	uint32_t fmt;
-	uint32_t wrap;
-	uint32_t en;
-	uint32_t filt;
-	uint32_t bcol;
-};
-
-struct nv30_vertex_program_exec {
-	uint32_t data[4];
-	boolean has_branch_offset;
-	int const_index;
-};
-
-struct nv30_vertex_program_data {
-	int index; /* immediates == -1 */
-	float value[4];
-};
-
-struct nv30_vertex_program {
-	struct pipe_shader_state pipe;
-
-	boolean translated;
-
-	struct nv30_vertex_program_exec *insns;
-	unsigned nr_insns;
-	struct nv30_vertex_program_data *consts;
-	unsigned nr_consts;
-
-	struct nouveau_resource *exec;
-	unsigned exec_start;
-	struct nouveau_resource *data;
-	unsigned data_start;
-	unsigned data_start_min;
-
-	uint32_t ir;
-	uint32_t or;
-	struct nouveau_stateobj *so;
-};
-
-struct nv30_fragment_program_data {
-	unsigned offset;
-	unsigned index;
-};
-
-struct nv30_fragment_program {
-	struct pipe_shader_state pipe;
-	struct tgsi_shader_info info;
-
-	boolean translated;
-	boolean on_hw;
-	unsigned samplers;
-
-	uint32_t *insn;
-	int       insn_len;
-
-	struct nv30_fragment_program_data *consts;
-	unsigned nr_consts;
-
-	struct pipe_buffer *buffer;
-
-	uint32_t fp_control;
-	uint32_t fp_reg_control;
-	struct nouveau_stateobj *so;
-};
-
-#define NV30_MAX_TEXTURE_LEVELS  16
-
-struct nv30_miptree {
-	struct pipe_texture base;
-	struct nouveau_bo *bo;
-
-	struct pipe_buffer *buffer;
-	uint total_size;
-
-	struct {
-		uint pitch;
-		uint *image_offset;
-	} level[NV30_MAX_TEXTURE_LEVELS];
-};
-
-#endif
diff --git a/src/gallium/drivers/nv30/nv30_state_blend.c b/src/gallium/drivers/nv30/nv30_state_blend.c
deleted file mode 100644
index c36d58c040..0000000000
--- a/src/gallium/drivers/nv30/nv30_state_blend.c
+++ /dev/null
@@ -1,41 +0,0 @@
-#include "nv30_context.h"
-
-static boolean
-nv30_state_blend_validate(struct nv30_context *nv30)
-{
-	so_ref(nv30->blend->so, &nv30->state.hw[NV30_STATE_BLEND]);
-	return TRUE;
-}
-
-struct nv30_state_entry nv30_state_blend = {
-	.validate = nv30_state_blend_validate,
-	.dirty = {
-		.pipe = NV30_NEW_BLEND,
-		.hw = NV30_STATE_BLEND
-	}
-};
-
-static boolean
-nv30_state_blend_colour_validate(struct nv30_context *nv30)
-{
-	struct nouveau_stateobj *so = so_new(1, 1, 0);
-	struct pipe_blend_color *bcol = &nv30->blend_colour;
-
-	so_method(so, nv30->screen->rankine, NV34TCL_BLEND_COLOR, 1);
-	so_data  (so, ((float_to_ubyte(bcol->color[3]) << 24) |
-		       (float_to_ubyte(bcol->color[0]) << 16) |
-		       (float_to_ubyte(bcol->color[1]) <<  8) |
-		       (float_to_ubyte(bcol->color[2]) <<  0)));
-
-	so_ref(so, &nv30->state.hw[NV30_STATE_BCOL]);
-	so_ref(NULL, &so);
-	return TRUE;
-}
-
-struct nv30_state_entry nv30_state_blend_colour = {
-	.validate = nv30_state_blend_colour_validate,
-	.dirty = {
-		.pipe = NV30_NEW_BCOL,
-		.hw = NV30_STATE_BCOL
-	}
-};
diff --git a/src/gallium/drivers/nv30/nv30_state_emit.c b/src/gallium/drivers/nv30/nv30_state_emit.c
deleted file mode 100644
index deefe7fd8d..0000000000
--- a/src/gallium/drivers/nv30/nv30_state_emit.c
+++ /dev/null
@@ -1,122 +0,0 @@
-#include "nv30_context.h"
-#include "nv30_state.h"
-
-static struct nv30_state_entry *render_states[] = {
-	&nv30_state_framebuffer,
-	&nv30_state_rasterizer,
-	&nv30_state_scissor,
-	&nv30_state_stipple,
-	&nv30_state_fragprog,
-	&nv30_state_fragtex,
-	&nv30_state_vertprog,
-	&nv30_state_blend,
-	&nv30_state_blend_colour,
-	&nv30_state_zsa,
-	&nv30_state_sr,
-	&nv30_state_viewport,
-	&nv30_state_vbo,
-	NULL
-};
-
-static void
-nv30_state_do_validate(struct nv30_context *nv30,
-		       struct nv30_state_entry **states)
-{
-	while (*states) {
-		struct nv30_state_entry *e = *states;
-
-		if (nv30->dirty & e->dirty.pipe) {
-			if (e->validate(nv30)) {
-				nv30->state.dirty |= (1ULL << e->dirty.hw);
-			}
-		}
-
-		states++;
-	}
-	nv30->dirty = 0;
-}
-
-void
-nv30_state_emit(struct nv30_context *nv30)
-{
-	struct nouveau_channel *chan = nv30->screen->base.channel;
-	struct nv30_state *state = &nv30->state;
-	struct nv30_screen *screen = nv30->screen;
-	unsigned i;
-	uint64_t states;
-
-	/* XXX: racy!
-	 */
-	if (nv30 != screen->cur_ctx) {
-		for (i = 0; i < NV30_STATE_MAX; i++) {
-			if (state->hw[i] && screen->state[i] != state->hw[i])
-				state->dirty |= (1ULL << i);
-		}
-
-		screen->cur_ctx = nv30;
-	}
-
-	for (i = 0, states = state->dirty; states; i++) {
-		if (!(states & (1ULL << i)))
-			continue;
-		so_ref (state->hw[i], &nv30->screen->state[i]);
-		if (state->hw[i])
-			so_emit(chan, nv30->screen->state[i]);
-		states &= ~(1ULL << i);
-	}
-
-	state->dirty = 0;
-}
-
-void
-nv30_state_flush_notify(struct nouveau_channel *chan)
-{
-	struct nv30_context *nv30 = chan->user_private;
-	struct nv30_state *state = &nv30->state;
-	unsigned i, samplers;
-
-	so_emit_reloc_markers(chan, state->hw[NV30_STATE_FB]);
-	for (i = 0, samplers = state->fp_samplers; i < 16 && samplers; i++) {
-		if (!(samplers & (1 << i)))
-			continue;
-		so_emit_reloc_markers(chan,
-				      state->hw[NV30_STATE_FRAGTEX0+i]);
-		samplers &= ~(1ULL << i);
-	}
-	so_emit_reloc_markers(chan, state->hw[NV30_STATE_FRAGPROG]);
-	if (state->hw[NV30_STATE_VTXBUF] /*&& nv30->render_mode == HW*/)
-		so_emit_reloc_markers(chan, state->hw[NV30_STATE_VTXBUF]);
-}
-
-boolean
-nv30_state_validate(struct nv30_context *nv30)
-{
-#if 0
-	boolean was_sw = nv30->fallback_swtnl ? TRUE : FALSE;
-
-	if (nv30->render_mode != HW) {
-		/* Don't even bother trying to go back to hw if none
-		 * of the states that caused swtnl previously have changed.
-		 */
-		if ((nv30->fallback_swtnl & nv30->dirty)
-				!= nv30->fallback_swtnl)
-			return FALSE;
-
-		/* Attempt to go to hwtnl again */
-		nv30->pipe.flush(&nv30->pipe, 0, NULL);
-		nv30->dirty |= (NV30_NEW_VIEWPORT |
-				NV30_NEW_VERTPROG |
-				NV30_NEW_ARRAYS);
-		nv30->render_mode = HW;
-	}
-#endif
-	nv30_state_do_validate(nv30, render_states);
-#if 0
-	if (nv30->fallback_swtnl || nv30->fallback_swrast)
-		return FALSE;
-	
-	if (was_sw)
-		NOUVEAU_ERR("swtnl->hw\n");
-#endif
-	return TRUE;
-}
diff --git a/src/gallium/drivers/nv30/nv30_state_fb.c b/src/gallium/drivers/nv30/nv30_state_fb.c
deleted file mode 100644
index f7fe9833c7..0000000000
--- a/src/gallium/drivers/nv30/nv30_state_fb.c
+++ /dev/null
@@ -1,173 +0,0 @@
-#include "nv30_context.h"
-#include "nouveau/nouveau_util.h"
-
-static boolean
-nv30_state_framebuffer_validate(struct nv30_context *nv30)
-{
-	struct pipe_framebuffer_state *fb = &nv30->framebuffer;
-	struct nouveau_channel *chan = nv30->screen->base.channel;
-	struct nouveau_grobj *rankine = nv30->screen->rankine;
-	struct nv04_surface *rt[2], *zeta = NULL;
-	uint32_t rt_enable = 0, rt_format = 0;
-	int i, colour_format = 0, zeta_format = 0, depth_only = 0;
-	struct nouveau_stateobj *so = so_new(12, 18, 10);
-	unsigned rt_flags = NOUVEAU_BO_RDWR | NOUVEAU_BO_VRAM;
-	unsigned w = fb->width;
-	unsigned h = fb->height;
-	struct nv30_miptree *nv30mt;
-	int colour_bits = 32, zeta_bits = 32;
-
-	for (i = 0; i < fb->nr_cbufs; i++) {
-		if (colour_format) {
-			assert(colour_format == fb->cbufs[i]->format);
-		} else {
-			colour_format = fb->cbufs[i]->format;
-			rt_enable |= (NV34TCL_RT_ENABLE_COLOR0 << i);
-			rt[i] = (struct nv04_surface *)fb->cbufs[i];
-		}
-	}
-
-	if (rt_enable & NV34TCL_RT_ENABLE_COLOR1)
-		rt_enable |= NV34TCL_RT_ENABLE_MRT;
-
-	if (fb->zsbuf) {
-		zeta_format = fb->zsbuf->format;
-		zeta = (struct nv04_surface *)fb->zsbuf;
-	}
-
-	if (rt_enable & (NV34TCL_RT_ENABLE_COLOR0|NV34TCL_RT_ENABLE_COLOR1)) {
-		/* Render to at least a colour buffer */
-		if (!(rt[0]->base.texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) {
-			assert(!(fb->width & (fb->width - 1)) && !(fb->height & (fb->height - 1)));
-			for (i = 1; i < fb->nr_cbufs; i++)
-				assert(!(rt[i]->base.texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR));
-
-			rt_format = NV34TCL_RT_FORMAT_TYPE_SWIZZLED |
-				(log2i(rt[0]->base.width) << NV34TCL_RT_FORMAT_LOG2_WIDTH_SHIFT) |
-				(log2i(rt[0]->base.height) << NV34TCL_RT_FORMAT_LOG2_HEIGHT_SHIFT);
-		}
-		else
-			rt_format = NV34TCL_RT_FORMAT_TYPE_LINEAR;
-	} else if (fb->zsbuf) {
-		depth_only = 1;
-
-		/* Render to depth buffer only */
-		if (!(zeta->base.texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) {
-			assert(!(fb->width & (fb->width - 1)) && !(fb->height & (fb->height - 1)));
-
-			rt_format = NV34TCL_RT_FORMAT_TYPE_SWIZZLED |
-				(log2i(zeta->base.width) << NV34TCL_RT_FORMAT_LOG2_WIDTH_SHIFT) |
-				(log2i(zeta->base.height) << NV34TCL_RT_FORMAT_LOG2_HEIGHT_SHIFT);
-		}
-		else
-			rt_format = NV34TCL_RT_FORMAT_TYPE_LINEAR;
-	} else {
-		return FALSE;
-	}
-
-	switch (colour_format) {
-	case PIPE_FORMAT_B8G8R8X8_UNORM:
-		rt_format |= NV34TCL_RT_FORMAT_COLOR_X8R8G8B8;
-		break;
-	case PIPE_FORMAT_B8G8R8A8_UNORM:
-	case 0:
-		rt_format |= NV34TCL_RT_FORMAT_COLOR_A8R8G8B8;
-		break;
-	case PIPE_FORMAT_B5G6R5_UNORM:
-		rt_format |= NV34TCL_RT_FORMAT_COLOR_R5G6B5;
-		colour_bits = 16;
-		break;
-	default:
-		assert(0);
-	}
-
-	switch (zeta_format) {
-	case PIPE_FORMAT_Z16_UNORM:
-		rt_format |= NV34TCL_RT_FORMAT_ZETA_Z16;
-		zeta_bits = 16;
-		break;
-	case PIPE_FORMAT_S8Z24_UNORM:
-	case PIPE_FORMAT_X8Z24_UNORM:
-	case 0:
-		rt_format |= NV34TCL_RT_FORMAT_ZETA_Z24S8;
-		break;
-	default:
-		assert(0);
-	}
-
-	if (colour_bits > zeta_bits) {
-		return FALSE;
-	}
-
-	if (depth_only || (rt_enable & NV34TCL_RT_ENABLE_COLOR0)) {
-		struct nv04_surface *rt0 = (depth_only ? zeta : rt[0]);
-		uint32_t pitch = rt0->pitch;
-
-		if (zeta) {
-			pitch |= (zeta->pitch << 16);
-		} else {
-			pitch |= (pitch << 16);
-		}
-
-		nv30mt = (struct nv30_miptree *) rt0->base.texture;
-		so_method(so, rankine, NV34TCL_DMA_COLOR0, 1);
-		so_reloc (so, nouveau_bo(nv30mt->buffer), 0, rt_flags | NOUVEAU_BO_OR,
-			      chan->vram->handle, chan->gart->handle);
-		so_method(so, rankine, NV34TCL_COLOR0_PITCH, 2);
-		so_data  (so, pitch);
-		so_reloc (so, nouveau_bo(nv30mt->buffer), rt0->base.offset,
-			      rt_flags | NOUVEAU_BO_LOW, 0, 0);
-	}
-
-	if (rt_enable & NV34TCL_RT_ENABLE_COLOR1) {
-		nv30mt = (struct nv30_miptree *)rt[1]->base.texture;
-		so_method(so, rankine, NV34TCL_DMA_COLOR1, 1);
-		so_reloc (so, nouveau_bo(nv30mt->buffer), 0, rt_flags | NOUVEAU_BO_OR,
-			      chan->vram->handle, chan->gart->handle);
-		so_method(so, rankine, NV34TCL_COLOR1_OFFSET, 2);
-		so_reloc (so, nouveau_bo(nv30mt->buffer), rt[1]->base.offset,
-			      rt_flags | NOUVEAU_BO_LOW, 0, 0);
-		so_data  (so, rt[1]->pitch);
-	}
-
-	if (zeta_format) {
-		nv30mt = (struct nv30_miptree *)zeta->base.texture;
-		so_method(so, rankine, NV34TCL_DMA_ZETA, 1);
-		so_reloc (so, nouveau_bo(nv30mt->buffer), 0, rt_flags | NOUVEAU_BO_OR,
-			      chan->vram->handle, chan->gart->handle);
-		so_method(so, rankine, NV34TCL_ZETA_OFFSET, 1);
-		so_reloc (so, nouveau_bo(nv30mt->buffer), zeta->base.offset,
-			      rt_flags | NOUVEAU_BO_LOW, 0, 0);
-		/* TODO: allocate LMA depth buffer */
-	}
-
-	so_method(so, rankine, NV34TCL_RT_ENABLE, 1);
-	so_data  (so, rt_enable);
-	so_method(so, rankine, NV34TCL_RT_HORIZ, 3);
-	so_data  (so, (w << 16) | 0);
-	so_data  (so, (h << 16) | 0);
-	so_data  (so, rt_format);
-	so_method(so, rankine, NV34TCL_VIEWPORT_HORIZ, 2);
-	so_data  (so, (w << 16) | 0);
-	so_data  (so, (h << 16) | 0);
-	so_method(so, rankine, NV34TCL_VIEWPORT_CLIP_HORIZ(0), 2);
-	so_data  (so, ((w - 1) << 16) | 0);
-	so_data  (so, ((h - 1) << 16) | 0);
-	so_method(so, rankine, 0x1d88, 1);
-	so_data  (so, (1 << 12) | h);
-	/* Wonder why this is needed, context should all be set to zero on init */
-	so_method(so, rankine, NV34TCL_VIEWPORT_TX_ORIGIN, 1);
-	so_data  (so, 0);
-
-	so_ref(so, &nv30->state.hw[NV30_STATE_FB]);
-	so_ref(NULL, &so);
-	return TRUE;
-}
-
-struct nv30_state_entry nv30_state_framebuffer = {
-	.validate = nv30_state_framebuffer_validate,
-	.dirty = {
-		.pipe = NV30_NEW_FB,
-		.hw = NV30_STATE_FB
-	}
-};
diff --git a/src/gallium/drivers/nv30/nv30_state_rasterizer.c b/src/gallium/drivers/nv30/nv30_state_rasterizer.c
deleted file mode 100644
index 6d1b60e043..0000000000
--- a/src/gallium/drivers/nv30/nv30_state_rasterizer.c
+++ /dev/null
@@ -1,17 +0,0 @@
-#include "nv30_context.h"
-
-static boolean
-nv30_state_rasterizer_validate(struct nv30_context *nv30)
-{
-	so_ref(nv30->rasterizer->so,
-	       &nv30->state.hw[NV30_STATE_RAST]);
-	return TRUE;
-}
-
-struct nv30_state_entry nv30_state_rasterizer = {
-	.validate = nv30_state_rasterizer_validate,
-	.dirty = {
-		.pipe = NV30_NEW_RAST,
-		.hw = NV30_STATE_RAST
-	}
-};
diff --git a/src/gallium/drivers/nv30/nv30_state_scissor.c b/src/gallium/drivers/nv30/nv30_state_scissor.c
deleted file mode 100644
index ba61a9e24a..0000000000
--- a/src/gallium/drivers/nv30/nv30_state_scissor.c
+++ /dev/null
@@ -1,36 +0,0 @@
-#include "nv30_context.h"
-
-static boolean
-nv30_state_scissor_validate(struct nv30_context *nv30)
-{
-	struct pipe_rasterizer_state *rast = &nv30->rasterizer->pipe;
-	struct pipe_scissor_state *s = &nv30->scissor;
-	struct nouveau_stateobj *so;
-
-	if (nv30->state.hw[NV30_STATE_SCISSOR] &&
-	    (rast->scissor == 0 && nv30->state.scissor_enabled == 0))
-		return FALSE;
-	nv30->state.scissor_enabled = rast->scissor;
-
-	so = so_new(1, 2, 0);
-	so_method(so, nv30->screen->rankine, NV34TCL_SCISSOR_HORIZ, 2);
-	if (nv30->state.scissor_enabled) {
-		so_data  (so, ((s->maxx - s->minx) << 16) | s->minx);
-		so_data  (so, ((s->maxy - s->miny) << 16) | s->miny);
-	} else {
-		so_data  (so, 4096 << 16);
-		so_data  (so, 4096 << 16);
-	}
-
-	so_ref(so, &nv30->state.hw[NV30_STATE_SCISSOR]);
-	so_ref(NULL, &so);
-	return TRUE;
-}
-
-struct nv30_state_entry nv30_state_scissor = {
-	.validate = nv30_state_scissor_validate,
-	.dirty = {
-		.pipe = NV30_NEW_SCISSOR | NV30_NEW_RAST,
-		.hw = NV30_STATE_SCISSOR
-	}
-};
diff --git a/src/gallium/drivers/nv30/nv30_state_stipple.c b/src/gallium/drivers/nv30/nv30_state_stipple.c
deleted file mode 100644
index ed520a4f43..0000000000
--- a/src/gallium/drivers/nv30/nv30_state_stipple.c
+++ /dev/null
@@ -1,40 +0,0 @@
-#include "nv30_context.h"
-
-static boolean
-nv30_state_stipple_validate(struct nv30_context *nv30)
-{
-	struct pipe_rasterizer_state *rast = &nv30->rasterizer->pipe;
-	struct nouveau_grobj *rankine = nv30->screen->rankine;
-	struct nouveau_stateobj *so;
-
-	if (nv30->state.hw[NV30_STATE_STIPPLE] &&
-	   (rast->poly_stipple_enable == 0 && nv30->state.stipple_enabled == 0))
-		return FALSE;
-
-	if (rast->poly_stipple_enable) {
-		unsigned i;
-
-		so = so_new(2, 33, 0);
-		so_method(so, rankine, NV34TCL_POLYGON_STIPPLE_ENABLE, 1);
-		so_data  (so, 1);
-		so_method(so, rankine, NV34TCL_POLYGON_STIPPLE_PATTERN(0), 32);
-		for (i = 0; i < 32; i++)
-			so_data(so, nv30->stipple[i]);
-	} else {
-		so = so_new(1, 1, 0);
-		so_method(so, rankine, NV34TCL_POLYGON_STIPPLE_ENABLE, 1);
-		so_data  (so, 0);
-	}
-
-	so_ref(so, &nv30->state.hw[NV30_STATE_STIPPLE]);
-	so_ref(NULL, &so);
-	return TRUE;
-}
-
-struct nv30_state_entry nv30_state_stipple = {
-	.validate = nv30_state_stipple_validate,
-	.dirty = {
-		.pipe = NV30_NEW_STIPPLE | NV30_NEW_RAST,
-		.hw = NV30_STATE_STIPPLE,
-	}
-};
diff --git a/src/gallium/drivers/nv30/nv30_state_viewport.c b/src/gallium/drivers/nv30/nv30_state_viewport.c
deleted file mode 100644
index 6fccd6b60e..0000000000
--- a/src/gallium/drivers/nv30/nv30_state_viewport.c
+++ /dev/null
@@ -1,42 +0,0 @@
-#include "nv30_context.h"
-
-static boolean
-nv30_state_viewport_validate(struct nv30_context *nv30)
-{
-	struct pipe_viewport_state *vpt = &nv30->viewport;
-	struct nouveau_stateobj *so;
-
-	if (nv30->state.hw[NV30_STATE_VIEWPORT] &&
-	    !(nv30->dirty & NV30_NEW_VIEWPORT))
-		return FALSE;
-
-	so = so_new(3, 10, 0);
-	so_method(so, nv30->screen->rankine,
-		  NV34TCL_VIEWPORT_TRANSLATE_X, 8);
-	so_data  (so, fui(vpt->translate[0]));
-	so_data  (so, fui(vpt->translate[1]));
-	so_data  (so, fui(vpt->translate[2]));
-	so_data  (so, fui(vpt->translate[3]));
-	so_data  (so, fui(vpt->scale[0]));
-	so_data  (so, fui(vpt->scale[1]));
-	so_data  (so, fui(vpt->scale[2]));
-	so_data  (so, fui(vpt->scale[3]));
-/*	so_method(so, nv30->screen->rankine, 0x1d78, 1);
-	so_data  (so, 1);
-*/
-	/* TODO/FIXME: never saw value 0x0110 in renouveau dumps, only 0x0001 */
-	so_method(so, nv30->screen->rankine, 0x1d78, 1);
-	so_data  (so, 1);
-
-	so_ref(so, &nv30->state.hw[NV30_STATE_VIEWPORT]);
-	so_ref(NULL, &so);
-	return TRUE;
-}
-
-struct nv30_state_entry nv30_state_viewport = {
-	.validate = nv30_state_viewport_validate,
-	.dirty = {
-		.pipe = NV30_NEW_VIEWPORT | NV30_NEW_RAST,
-		.hw = NV30_STATE_VIEWPORT
-	}
-};
diff --git a/src/gallium/drivers/nv30/nv30_state_zsa.c b/src/gallium/drivers/nv30/nv30_state_zsa.c
deleted file mode 100644
index 88cd74f180..0000000000
--- a/src/gallium/drivers/nv30/nv30_state_zsa.c
+++ /dev/null
@@ -1,41 +0,0 @@
-#include "nv30_context.h"
-
-static boolean
-nv30_state_zsa_validate(struct nv30_context *nv30)
-{
-	so_ref(nv30->zsa->so,
-	       &nv30->state.hw[NV30_STATE_ZSA]);
-	return TRUE;
-}
-
-struct nv30_state_entry nv30_state_zsa = {
-	.validate = nv30_state_zsa_validate,
-	.dirty = {
-		.pipe = NV30_NEW_ZSA,
-		.hw = NV30_STATE_ZSA
-	}
-};
-
-static boolean
-nv30_state_sr_validate(struct nv30_context *nv30)
-{
-	struct nouveau_stateobj *so = so_new(2, 2, 0);
-	struct pipe_stencil_ref *sr = &nv30->stencil_ref;
-
-	so_method(so, nv30->screen->rankine, NV34TCL_STENCIL_FRONT_FUNC_REF, 1);
-	so_data  (so, sr->ref_value[0]);
-	so_method(so, nv30->screen->rankine, NV34TCL_STENCIL_BACK_FUNC_REF, 1);
-	so_data  (so, sr->ref_value[1]);
-
-	so_ref(so, &nv30->state.hw[NV30_STATE_SR]);
-	so_ref(NULL, &so);
-	return TRUE;
-}
-
-struct nv30_state_entry nv30_state_sr = {
-	.validate = nv30_state_sr_validate,
-	.dirty = {
-		.pipe = NV30_NEW_SR,
-		.hw = NV30_STATE_SR
-	}
-};
diff --git a/src/gallium/drivers/nv30/nv30_vertprog.c b/src/gallium/drivers/nv30/nv30_vertprog.c
deleted file mode 100644
index 809be3712d..0000000000
--- a/src/gallium/drivers/nv30/nv30_vertprog.c
+++ /dev/null
@@ -1,842 +0,0 @@
-#include "pipe/p_context.h"
-#include "pipe/p_defines.h"
-#include "pipe/p_state.h"
-#include "util/u_inlines.h"
-
-#include "pipe/p_shader_tokens.h"
-#include "tgsi/tgsi_parse.h"
-#include "tgsi/tgsi_dump.h"
-
-#include "nv30_context.h"
-#include "nv30_state.h"
-
-/* TODO (at least...):
- *  1. Indexed consts  + ARL
- *  2. Arb. swz/negation
- *  3. NV_vp11, NV_vp2, NV_vp3 features
- *       - extra arith opcodes
- *       - branching
- *       - texture sampling
- *       - indexed attribs
- *       - indexed results
- *  4. bugs
- */
-
-#define SWZ_X 0
-#define SWZ_Y 1
-#define SWZ_Z 2
-#define SWZ_W 3
-#define MASK_X 8
-#define MASK_Y 4
-#define MASK_Z 2
-#define MASK_W 1
-#define MASK_ALL (MASK_X|MASK_Y|MASK_Z|MASK_W)
-#define DEF_SCALE 0
-#define DEF_CTEST 0
-#include "nv30_shader.h"
-
-#define swz(s,x,y,z,w) nv30_sr_swz((s), SWZ_##x, SWZ_##y, SWZ_##z, SWZ_##w)
-#define neg(s) nv30_sr_neg((s))
-#define abs(s) nv30_sr_abs((s))
-
-struct nv30_vpc {
-	struct nv30_vertex_program *vp;
-
-	struct nv30_vertex_program_exec *vpi;
-
-	unsigned output_map[PIPE_MAX_SHADER_OUTPUTS];
-
-	int high_temp;
-	int temp_temp_count;
-
-	struct nv30_sreg *imm;
-	unsigned nr_imm;
-};
-
-static struct nv30_sreg
-temp(struct nv30_vpc *vpc)
-{
-	int idx;
-
-	idx  = vpc->temp_temp_count++;
-	idx += vpc->high_temp + 1;
-	return nv30_sr(NV30SR_TEMP, idx);
-}
-
-static struct nv30_sreg
-constant(struct nv30_vpc *vpc, int pipe, float x, float y, float z, float w)
-{
-	struct nv30_vertex_program *vp = vpc->vp;
-	struct nv30_vertex_program_data *vpd;
-	int idx;
-
-	if (pipe >= 0) {
-		for (idx = 0; idx < vp->nr_consts; idx++) {
-			if (vp->consts[idx].index == pipe)
-				return nv30_sr(NV30SR_CONST, idx);
-		}
-	}
-
-	idx = vp->nr_consts++;
-	vp->consts = realloc(vp->consts, sizeof(*vpd) * vp->nr_consts);
-	vpd = &vp->consts[idx];
-
-	vpd->index = pipe;
-	vpd->value[0] = x;
-	vpd->value[1] = y;
-	vpd->value[2] = z;
-	vpd->value[3] = w;
-	return nv30_sr(NV30SR_CONST, idx);
-}
-
-#define arith(cc,s,o,d,m,s0,s1,s2) \
-	nv30_vp_arith((cc), (s), NV30_VP_INST_##o, (d), (m), (s0), (s1), (s2))
-
-static void
-emit_src(struct nv30_vpc *vpc, uint32_t *hw, int pos, struct nv30_sreg src)
-{
-	struct nv30_vertex_program *vp = vpc->vp;
-	uint32_t sr = 0;
-
-	switch (src.type) {
-	case NV30SR_TEMP:
-		sr |= (NV30_VP_SRC_REG_TYPE_TEMP << NV30_VP_SRC_REG_TYPE_SHIFT);
-		sr |= (src.index << NV30_VP_SRC_TEMP_SRC_SHIFT);
-		break;
-	case NV30SR_INPUT:
-		sr |= (NV30_VP_SRC_REG_TYPE_INPUT <<
-		       NV30_VP_SRC_REG_TYPE_SHIFT);
-		vp->ir |= (1 << src.index);
-		hw[1] |= (src.index << NV30_VP_INST_INPUT_SRC_SHIFT);
-		break;
-	case NV30SR_CONST:
-		sr |= (NV30_VP_SRC_REG_TYPE_CONST <<
-		       NV30_VP_SRC_REG_TYPE_SHIFT);
-		assert(vpc->vpi->const_index == -1 ||
-		       vpc->vpi->const_index == src.index);
-		vpc->vpi->const_index = src.index;
-		break;
-	case NV30SR_NONE:
-		sr |= (NV30_VP_SRC_REG_TYPE_INPUT <<
-		       NV30_VP_SRC_REG_TYPE_SHIFT);
-		break;
-	default:
-		assert(0);
-	}
-
-	if (src.negate)
-		sr |= NV30_VP_SRC_NEGATE;
-
-	if (src.abs)
-		hw[0] |= (1 << (21 + pos));
-
-	sr |= ((src.swz[0] << NV30_VP_SRC_SWZ_X_SHIFT) |
-	       (src.swz[1] << NV30_VP_SRC_SWZ_Y_SHIFT) |
-	       (src.swz[2] << NV30_VP_SRC_SWZ_Z_SHIFT) |
-	       (src.swz[3] << NV30_VP_SRC_SWZ_W_SHIFT));
-
-/*
- * |VVV|
- * d�.�b
- *  \u/
- *
- */
-
-	switch (pos) {
-	case 0:
-		hw[1] |= ((sr & NV30_VP_SRC0_HIGH_MASK) >>
-			  NV30_VP_SRC0_HIGH_SHIFT) << NV30_VP_INST_SRC0H_SHIFT;
-		hw[2] |= (sr & NV30_VP_SRC0_LOW_MASK) <<
-			  NV30_VP_INST_SRC0L_SHIFT;
-		break;
-	case 1:
-		hw[2] |= sr << NV30_VP_INST_SRC1_SHIFT;
-		break;
-	case 2:
-		hw[2] |= ((sr & NV30_VP_SRC2_HIGH_MASK) >>
-			  NV30_VP_SRC2_HIGH_SHIFT) << NV30_VP_INST_SRC2H_SHIFT;
-		hw[3] |= (sr & NV30_VP_SRC2_LOW_MASK) <<
-			  NV30_VP_INST_SRC2L_SHIFT;
-		break;
-	default:
-		assert(0);
-	}
-}
-
-static void
-emit_dst(struct nv30_vpc *vpc, uint32_t *hw, int slot, struct nv30_sreg dst)
-{
-	struct nv30_vertex_program *vp = vpc->vp;
-
-	switch (dst.type) {
-	case NV30SR_TEMP:
-		hw[0] |= (dst.index << NV30_VP_INST_DEST_TEMP_ID_SHIFT);
-		break;
-	case NV30SR_OUTPUT:
-		switch (dst.index) {
-		case NV30_VP_INST_DEST_COL0 : vp->or |= (1 << 0); break;
-		case NV30_VP_INST_DEST_COL1 : vp->or |= (1 << 1); break;
-		case NV30_VP_INST_DEST_BFC0 : vp->or |= (1 << 2); break;
-		case NV30_VP_INST_DEST_BFC1 : vp->or |= (1 << 3); break;
-		case NV30_VP_INST_DEST_FOGC : vp->or |= (1 << 4); break;
-		case NV30_VP_INST_DEST_PSZ  : vp->or |= (1 << 5); break;
-		case NV30_VP_INST_DEST_TC(0): vp->or |= (1 << 14); break;
-		case NV30_VP_INST_DEST_TC(1): vp->or |= (1 << 15); break;
-		case NV30_VP_INST_DEST_TC(2): vp->or |= (1 << 16); break;
-		case NV30_VP_INST_DEST_TC(3): vp->or |= (1 << 17); break;
-		case NV30_VP_INST_DEST_TC(4): vp->or |= (1 << 18); break;
-		case NV30_VP_INST_DEST_TC(5): vp->or |= (1 << 19); break;
-		case NV30_VP_INST_DEST_TC(6): vp->or |= (1 << 20); break;
-		case NV30_VP_INST_DEST_TC(7): vp->or |= (1 << 21); break;
-		default:
-			break;
-		}
-
-		hw[3] |= (dst.index << NV30_VP_INST_DEST_SHIFT);
-		hw[0] |= NV30_VP_INST_VEC_DEST_TEMP_MASK | (1<<20);
-
-		/*XXX: no way this is entirely correct, someone needs to
-		 *     figure out what exactly it is.
-		 */
-		hw[3] |= 0x800;
-		break;
-	default:
-		assert(0);
-	}
-}
-
-static void
-nv30_vp_arith(struct nv30_vpc *vpc, int slot, int op,
-	      struct nv30_sreg dst, int mask,
-	      struct nv30_sreg s0, struct nv30_sreg s1,
-	      struct nv30_sreg s2)
-{
-	struct nv30_vertex_program *vp = vpc->vp;
-	uint32_t *hw;
-
-	vp->insns = realloc(vp->insns, ++vp->nr_insns * sizeof(*vpc->vpi));
-	vpc->vpi = &vp->insns[vp->nr_insns - 1];
-	memset(vpc->vpi, 0, sizeof(*vpc->vpi));
-	vpc->vpi->const_index = -1;
-
-	hw = vpc->vpi->data;
-
-	hw[0] |= (NV30_VP_INST_COND_TR << NV30_VP_INST_COND_SHIFT);
-	hw[0] |= ((0 << NV30_VP_INST_COND_SWZ_X_SHIFT) |
-		  (1 << NV30_VP_INST_COND_SWZ_Y_SHIFT) |
-		  (2 << NV30_VP_INST_COND_SWZ_Z_SHIFT) |
-		  (3 << NV30_VP_INST_COND_SWZ_W_SHIFT));
-
-	hw[1] |= (op << NV30_VP_INST_VEC_OPCODE_SHIFT);
-//	hw[3] |= NV30_VP_INST_SCA_DEST_TEMP_MASK;
-//	hw[3] |= (mask << NV30_VP_INST_VEC_WRITEMASK_SHIFT);
-
-	if (dst.type == NV30SR_OUTPUT) {
-		if (slot)
-			hw[3] |= (mask << NV30_VP_INST_SDEST_WRITEMASK_SHIFT);
-		else
-			hw[3] |= (mask << NV30_VP_INST_VDEST_WRITEMASK_SHIFT);
-	} else {
-		if (slot)
-			hw[3] |= (mask << NV30_VP_INST_STEMP_WRITEMASK_SHIFT);
-		else
-			hw[3] |= (mask << NV30_VP_INST_VTEMP_WRITEMASK_SHIFT);
-	}
-
-	emit_dst(vpc, hw, slot, dst);
-	emit_src(vpc, hw, 0, s0);
-	emit_src(vpc, hw, 1, s1);
-	emit_src(vpc, hw, 2, s2);
-}
-
-static INLINE struct nv30_sreg
-tgsi_src(struct nv30_vpc *vpc, const struct tgsi_full_src_register *fsrc) {
-	struct nv30_sreg src;
-
-	switch (fsrc->Register.File) {
-	case TGSI_FILE_INPUT:
-		src = nv30_sr(NV30SR_INPUT, fsrc->Register.Index);
-		break;
-	case TGSI_FILE_CONSTANT:
-		src = constant(vpc, fsrc->Register.Index, 0, 0, 0, 0);
-		break;
-	case TGSI_FILE_IMMEDIATE:
-		src = vpc->imm[fsrc->Register.Index];
-		break;
-	case TGSI_FILE_TEMPORARY:
-		if (vpc->high_temp < fsrc->Register.Index)
-			vpc->high_temp = fsrc->Register.Index;
-		src = nv30_sr(NV30SR_TEMP, fsrc->Register.Index);
-		break;
-	default:
-		NOUVEAU_ERR("bad src file\n");
-		break;
-	}
-
-	src.abs = fsrc->Register.Absolute;
-	src.negate = fsrc->Register.Negate;
-	src.swz[0] = fsrc->Register.SwizzleX;
-	src.swz[1] = fsrc->Register.SwizzleY;
-	src.swz[2] = fsrc->Register.SwizzleZ;
-	src.swz[3] = fsrc->Register.SwizzleW;
-	return src;
-}
-
-static INLINE struct nv30_sreg
-tgsi_dst(struct nv30_vpc *vpc, const struct tgsi_full_dst_register *fdst) {
-	struct nv30_sreg dst;
-
-	switch (fdst->Register.File) {
-	case TGSI_FILE_OUTPUT:
-		dst = nv30_sr(NV30SR_OUTPUT,
-			      vpc->output_map[fdst->Register.Index]);
-
-		break;
-	case TGSI_FILE_TEMPORARY:
-		dst = nv30_sr(NV30SR_TEMP, fdst->Register.Index);
-		if (vpc->high_temp < dst.index)
-			vpc->high_temp = dst.index;
-		break;
-	default:
-		NOUVEAU_ERR("bad dst file\n");
-		break;
-	}
-
-	return dst;
-}
-
-static INLINE int
-tgsi_mask(uint tgsi)
-{
-	int mask = 0;
-
-	if (tgsi & TGSI_WRITEMASK_X) mask |= MASK_X;
-	if (tgsi & TGSI_WRITEMASK_Y) mask |= MASK_Y;
-	if (tgsi & TGSI_WRITEMASK_Z) mask |= MASK_Z;
-	if (tgsi & TGSI_WRITEMASK_W) mask |= MASK_W;
-	return mask;
-}
-
-static boolean
-nv30_vertprog_parse_instruction(struct nv30_vpc *vpc,
-				const struct tgsi_full_instruction *finst)
-{
-	struct nv30_sreg src[3], dst, tmp;
-	struct nv30_sreg none = nv30_sr(NV30SR_NONE, 0);
-	int mask;
-	int ai = -1, ci = -1;
-	int i;
-
-	if (finst->Instruction.Opcode == TGSI_OPCODE_END)
-		return TRUE;
-
-	vpc->temp_temp_count = 0;
-	for (i = 0; i < finst->Instruction.NumSrcRegs; i++) {
-		const struct tgsi_full_src_register *fsrc;
-
-		fsrc = &finst->Src[i];
-		if (fsrc->Register.File == TGSI_FILE_TEMPORARY) {
-			src[i] = tgsi_src(vpc, fsrc);
-		}
-	}
-
-	for (i = 0; i < finst->Instruction.NumSrcRegs; i++) {
-		const struct tgsi_full_src_register *fsrc;
-
-		fsrc = &finst->Src[i];
-		switch (fsrc->Register.File) {
-		case TGSI_FILE_INPUT:
-			if (ai == -1 || ai == fsrc->Register.Index) {
-				ai = fsrc->Register.Index;
-				src[i] = tgsi_src(vpc, fsrc);
-			} else {
-				src[i] = temp(vpc);
-				arith(vpc, 0, OP_MOV, src[i], MASK_ALL,
-				      tgsi_src(vpc, fsrc), none, none);
-			}
-			break;
-		/*XXX: index comparison is broken now that consts come from
-		 *     two different register files.
-		 */
-		case TGSI_FILE_CONSTANT:
-		case TGSI_FILE_IMMEDIATE:
-			if (ci == -1 || ci == fsrc->Register.Index) {
-				ci = fsrc->Register.Index;
-				src[i] = tgsi_src(vpc, fsrc);
-			} else {
-				src[i] = temp(vpc);
-				arith(vpc, 0, OP_MOV, src[i], MASK_ALL,
-				      tgsi_src(vpc, fsrc), none, none);
-			}
-			break;
-		case TGSI_FILE_TEMPORARY:
-			/* handled above */
-			break;
-		default:
-			NOUVEAU_ERR("bad src file\n");
-			return FALSE;
-		}
-	}
-
-	dst  = tgsi_dst(vpc, &finst->Dst[0]);
-	mask = tgsi_mask(finst->Dst[0].Register.WriteMask);
-
-	switch (finst->Instruction.Opcode) {
-	case TGSI_OPCODE_ABS:
-		arith(vpc, 0, OP_MOV, dst, mask, abs(src[0]), none, none);
-		break;
-	case TGSI_OPCODE_ADD:
-		arith(vpc, 0, OP_ADD, dst, mask, src[0], none, src[1]);
-		break;
-	case TGSI_OPCODE_ARL:
-		arith(vpc, 0, OP_ARL, dst, mask, src[0], none, none);
-		break;
-	case TGSI_OPCODE_DP3:
-		arith(vpc, 0, OP_DP3, dst, mask, src[0], src[1], none);
-		break;
-	case TGSI_OPCODE_DP4:
-		arith(vpc, 0, OP_DP4, dst, mask, src[0], src[1], none);
-		break;
-	case TGSI_OPCODE_DPH:
-		arith(vpc, 0, OP_DPH, dst, mask, src[0], src[1], none);
-		break;
-	case TGSI_OPCODE_DST:
-		arith(vpc, 0, OP_DST, dst, mask, src[0], src[1], none);
-		break;
-	case TGSI_OPCODE_EX2:
-		arith(vpc, 1, OP_EX2, dst, mask, none, none, src[0]);
-		break;
-	case TGSI_OPCODE_EXP:
-		arith(vpc, 1, OP_EXP, dst, mask, none, none, src[0]);
-		break;
-	case TGSI_OPCODE_FLR:
-		arith(vpc, 0, OP_FLR, dst, mask, src[0], none, none);
-		break;
-	case TGSI_OPCODE_FRC:
-		arith(vpc, 0, OP_FRC, dst, mask, src[0], none, none);
-		break;
-	case TGSI_OPCODE_LG2:
-		arith(vpc, 1, OP_LG2, dst, mask, none, none, src[0]);
-		break;
-	case TGSI_OPCODE_LIT:
-		arith(vpc, 1, OP_LIT, dst, mask, none, none, src[0]);
-		break;
-	case TGSI_OPCODE_LOG:
-		arith(vpc, 1, OP_LOG, dst, mask, none, none, src[0]);
-		break;
-	case TGSI_OPCODE_MAD:
-		arith(vpc, 0, OP_MAD, dst, mask, src[0], src[1], src[2]);
-		break;
-	case TGSI_OPCODE_MAX:
-		arith(vpc, 0, OP_MAX, dst, mask, src[0], src[1], none);
-		break;
-	case TGSI_OPCODE_MIN:
-		arith(vpc, 0, OP_MIN, dst, mask, src[0], src[1], none);
-		break;
-	case TGSI_OPCODE_MOV:
-		arith(vpc, 0, OP_MOV, dst, mask, src[0], none, none);
-		break;
-	case TGSI_OPCODE_MUL:
-		arith(vpc, 0, OP_MUL, dst, mask, src[0], src[1], none);
-		break;
-	case TGSI_OPCODE_POW:
-		tmp = temp(vpc);
-		arith(vpc, 1, OP_LG2, tmp, MASK_X, none, none,
-		      swz(src[0], X, X, X, X));
-		arith(vpc, 0, OP_MUL, tmp, MASK_X, swz(tmp, X, X, X, X),
-		      swz(src[1], X, X, X, X), none);
-		arith(vpc, 1, OP_EX2, dst, mask, none, none,
-		      swz(tmp, X, X, X, X));
-		break;
-	case TGSI_OPCODE_RCP:
-		arith(vpc, 1, OP_RCP, dst, mask, none, none, src[0]);
-		break;
-	case TGSI_OPCODE_RET:
-		break;
-	case TGSI_OPCODE_RSQ:
-		arith(vpc, 1, OP_RSQ, dst, mask, none, none, src[0]);
-		break;
-	case TGSI_OPCODE_SGE:
-		arith(vpc, 0, OP_SGE, dst, mask, src[0], src[1], none);
-		break;
-	case TGSI_OPCODE_SGT:
-		arith(vpc, 0, OP_SGT, dst, mask, src[0], src[1], none);
-		break;
-	case TGSI_OPCODE_SLT:
-		arith(vpc, 0, OP_SLT, dst, mask, src[0], src[1], none);
-		break;
-	case TGSI_OPCODE_SUB:
-		arith(vpc, 0, OP_ADD, dst, mask, src[0], none, neg(src[1]));
-		break;
-	case TGSI_OPCODE_XPD:
-		tmp = temp(vpc);
-		arith(vpc, 0, OP_MUL, tmp, mask,
-		      swz(src[0], Z, X, Y, Y), swz(src[1], Y, Z, X, X), none);
-		arith(vpc, 0, OP_MAD, dst, (mask & ~MASK_W),
-		      swz(src[0], Y, Z, X, X), swz(src[1], Z, X, Y, Y),
-		      neg(tmp));
-		break;
-	default:
-		NOUVEAU_ERR("invalid opcode %d\n", finst->Instruction.Opcode);
-		return FALSE;
-	}
-
-	return TRUE;
-}
-
-static boolean
-nv30_vertprog_parse_decl_output(struct nv30_vpc *vpc,
-				const struct tgsi_full_declaration *fdec)
-{
-	int hw;
-
-	switch (fdec->Semantic.Name) {
-	case TGSI_SEMANTIC_POSITION:
-		hw = NV30_VP_INST_DEST_POS;
-		break;
-	case TGSI_SEMANTIC_COLOR:
-		if (fdec->Semantic.Index == 0) {
-			hw = NV30_VP_INST_DEST_COL0;
-		} else
-		if (fdec->Semantic.Index == 1) {
-			hw = NV30_VP_INST_DEST_COL1;
-		} else {
-			NOUVEAU_ERR("bad colour semantic index\n");
-			return FALSE;
-		}
-		break;
-	case TGSI_SEMANTIC_BCOLOR:
-		if (fdec->Semantic.Index == 0) {
-			hw = NV30_VP_INST_DEST_BFC0;
-		} else
-		if (fdec->Semantic.Index == 1) {
-			hw = NV30_VP_INST_DEST_BFC1;
-		} else {
-			NOUVEAU_ERR("bad bcolour semantic index\n");
-			return FALSE;
-		}
-		break;
-	case TGSI_SEMANTIC_FOG:
-		hw = NV30_VP_INST_DEST_FOGC;
-		break;
-	case TGSI_SEMANTIC_PSIZE:
-		hw = NV30_VP_INST_DEST_PSZ;
-		break;
-	case TGSI_SEMANTIC_GENERIC:
-		if (fdec->Semantic.Index <= 7) {
-			hw = NV30_VP_INST_DEST_TC(fdec->Semantic.Index);
-		} else {
-			NOUVEAU_ERR("bad generic semantic index\n");
-			return FALSE;
-		}
-		break;
-	case TGSI_SEMANTIC_EDGEFLAG:
-		NOUVEAU_ERR("cannot handle edgeflag output\n");
-		return FALSE;
-	default:
-		NOUVEAU_ERR("bad output semantic\n");
-		return FALSE;
-	}
-
-	vpc->output_map[fdec->Range.First] = hw;
-	return TRUE;
-}
-
-static boolean
-nv30_vertprog_prepare(struct nv30_vpc *vpc)
-{
-	struct tgsi_parse_context p;
-	int nr_imm = 0;
-
-	tgsi_parse_init(&p, vpc->vp->pipe.tokens);
-	while (!tgsi_parse_end_of_tokens(&p)) {
-		const union tgsi_full_token *tok = &p.FullToken;
-
-		tgsi_parse_token(&p);
-		switch(tok->Token.Type) {
-		case TGSI_TOKEN_TYPE_IMMEDIATE:
-			nr_imm++;
-			break;
-		default:
-			break;
-		}
-	}
-	tgsi_parse_free(&p);
-
-	if (nr_imm) {
-		vpc->imm = CALLOC(nr_imm, sizeof(struct nv30_sreg));
-		assert(vpc->imm);
-	}
-
-	return TRUE;
-}
-
-static void
-nv30_vertprog_translate(struct nv30_context *nv30,
-			struct nv30_vertex_program *vp)
-{
-	struct tgsi_parse_context parse;
-	struct nv30_vpc *vpc = NULL;
-
-	tgsi_dump(vp->pipe.tokens,0);
-
-	vpc = CALLOC(1, sizeof(struct nv30_vpc));
-	if (!vpc)
-		return;
-	vpc->vp = vp;
-	vpc->high_temp = -1;
-
-	if (!nv30_vertprog_prepare(vpc)) {
-		FREE(vpc);
-		return;
-	}
-
-	tgsi_parse_init(&parse, vp->pipe.tokens);
-
-	while (!tgsi_parse_end_of_tokens(&parse)) {
-		tgsi_parse_token(&parse);
-
-		switch (parse.FullToken.Token.Type) {
-		case TGSI_TOKEN_TYPE_DECLARATION:
-		{
-			const struct tgsi_full_declaration *fdec;
-			fdec = &parse.FullToken.FullDeclaration;
-			switch (fdec->Declaration.File) {
-			case TGSI_FILE_OUTPUT:
-				if (!nv30_vertprog_parse_decl_output(vpc, fdec))
-					goto out_err;
-				break;
-			default:
-				break;
-			}
-		}
-			break;
-		case TGSI_TOKEN_TYPE_IMMEDIATE:
-		{
-			const struct tgsi_full_immediate *imm;
-
-			imm = &parse.FullToken.FullImmediate;
-			assert(imm->Immediate.DataType == TGSI_IMM_FLOAT32);
-			assert(imm->Immediate.NrTokens == 4 + 1);
-			vpc->imm[vpc->nr_imm++] =
-				constant(vpc, -1,
-					 imm->u[0].Float,
-					 imm->u[1].Float,
-					 imm->u[2].Float,
-					 imm->u[3].Float);
-		}
-			break;
-		case TGSI_TOKEN_TYPE_INSTRUCTION:
-		{
-			const struct tgsi_full_instruction *finst;
-			finst = &parse.FullToken.FullInstruction;
-			if (!nv30_vertprog_parse_instruction(vpc, finst))
-				goto out_err;
-		}
-			break;
-		default:
-			break;
-		}
-	}
-
-	vp->insns[vp->nr_insns - 1].data[3] |= NV30_VP_INST_LAST;
-	vp->translated = TRUE;
-out_err:
-	tgsi_parse_free(&parse);
-	FREE(vpc);
-}
-
-static boolean
-nv30_vertprog_validate(struct nv30_context *nv30)
-{ 
-	struct pipe_screen *pscreen = nv30->pipe.screen;
-	struct nv30_screen *screen = nv30->screen;
-	struct nouveau_channel *chan = screen->base.channel;
-	struct nouveau_grobj *rankine = screen->rankine;
-	struct nv30_vertex_program *vp;
-	struct pipe_buffer *constbuf;
-	boolean upload_code = FALSE, upload_data = FALSE;
-	int i;
-
-	vp = nv30->vertprog;
-	constbuf = nv30->constbuf[PIPE_SHADER_VERTEX];
-
-	/* Translate TGSI shader into hw bytecode */
-	if (!vp->translated) {
-		nv30_vertprog_translate(nv30, vp);
-		if (!vp->translated)
-			return FALSE;
-	}
-
-	/* Allocate hw vtxprog exec slots */
-	if (!vp->exec) {
-		struct nouveau_resource *heap = nv30->screen->vp_exec_heap;
-		struct nouveau_stateobj *so;
-		uint vplen = vp->nr_insns;
-
-		if (nouveau_resource_alloc(heap, vplen, vp, &vp->exec)) {
-			while (heap->next && heap->size < vplen) {
-				struct nv30_vertex_program *evict;
-				
-				evict = heap->next->priv;
-				nouveau_resource_free(&evict->exec);
-			}
-
-			if (nouveau_resource_alloc(heap, vplen, vp, &vp->exec))
-				assert(0);
-		}
-
-		so = so_new(1, 1, 0);
-		so_method(so, rankine, NV34TCL_VP_START_FROM_ID, 1);
-		so_data  (so, vp->exec->start);
-		so_ref(so, &vp->so);
-		so_ref(NULL, &so);
-
-		upload_code = TRUE;
-	}
-
-	/* Allocate hw vtxprog const slots */
-	if (vp->nr_consts && !vp->data) {
-		struct nouveau_resource *heap = nv30->screen->vp_data_heap;
-
-		if (nouveau_resource_alloc(heap, vp->nr_consts, vp, &vp->data)) {
-			while (heap->next && heap->size < vp->nr_consts) {
-				struct nv30_vertex_program *evict;
-				
-				evict = heap->next->priv;
-				nouveau_resource_free(&evict->data);
-			}
-
-			if (nouveau_resource_alloc(heap, vp->nr_consts, vp,
-						   &vp->data))
-				assert(0);
-		}
-
-		/*XXX: handle this some day */
-		assert(vp->data->start >= vp->data_start_min);
-
-		upload_data = TRUE;
-		if (vp->data_start != vp->data->start)
-			upload_code = TRUE;
-	}
-
-	/* If exec or data segments moved we need to patch the program to
-	 * fixup offsets and register IDs.
-	 */
-	if (vp->exec_start != vp->exec->start) {
-		for (i = 0; i < vp->nr_insns; i++) {
-			struct nv30_vertex_program_exec *vpi = &vp->insns[i];
-
-			if (vpi->has_branch_offset) {
-				assert(0);
-			}
-		}
-
-		vp->exec_start = vp->exec->start;
-	}
-
-	if (vp->nr_consts && vp->data_start != vp->data->start) {
-		for (i = 0; i < vp->nr_insns; i++) {
-			struct nv30_vertex_program_exec *vpi = &vp->insns[i];
-
-			if (vpi->const_index >= 0) {
-				vpi->data[1] &= ~NV30_VP_INST_CONST_SRC_MASK;
-				vpi->data[1] |=
-					(vpi->const_index + vp->data->start) <<
-					NV30_VP_INST_CONST_SRC_SHIFT;
-
-			}
-		}
-
-		vp->data_start = vp->data->start;
-	}
-
-	/* Update + Upload constant values */
-	if (vp->nr_consts) {
-		float *map = NULL;
-
-		if (constbuf) {
-			map = pipe_buffer_map(pscreen, constbuf,
-					      PIPE_BUFFER_USAGE_CPU_READ);
-		}
-
-		for (i = 0; i < vp->nr_consts; i++) {
-			struct nv30_vertex_program_data *vpd = &vp->consts[i];
-
-			if (vpd->index >= 0) {
-				if (!upload_data &&
-				    !memcmp(vpd->value, &map[vpd->index * 4],
-					    4 * sizeof(float)))
-					continue;
-				memcpy(vpd->value, &map[vpd->index * 4],
-				       4 * sizeof(float));
-			}
-
-			BEGIN_RING(chan, rankine, NV34TCL_VP_UPLOAD_CONST_ID, 5);
-			OUT_RING  (chan, i + vp->data->start);
-			OUT_RINGp (chan, (uint32_t *)vpd->value, 4);
-		}
-
-		if (constbuf)
-			pipe_buffer_unmap(pscreen, constbuf);
-	}
-
-	/* Upload vtxprog */
-	if (upload_code) {
-#if 0
-		for (i = 0; i < vp->nr_insns; i++) {
-			NOUVEAU_MSG("VP inst %d: 0x%08x 0x%08x 0x%08x 0x%08x\n",
-				i, vp->insns[i].data[0], vp->insns[i].data[1],
-				vp->insns[i].data[2], vp->insns[i].data[3]);
-		}
-#endif
-		BEGIN_RING(chan, rankine, NV34TCL_VP_UPLOAD_FROM_ID, 1);
-		OUT_RING  (chan, vp->exec->start);
-		for (i = 0; i < vp->nr_insns; i++) {
-			BEGIN_RING(chan, rankine, NV34TCL_VP_UPLOAD_INST(0), 4);
-			OUT_RINGp (chan, vp->insns[i].data, 4);
-		}
-	}
-
-	if (vp->so != nv30->state.hw[NV30_STATE_VERTPROG]) {
-		so_ref(vp->so, &nv30->state.hw[NV30_STATE_VERTPROG]);
-		return TRUE;
-	}
-
-	return FALSE;
-}
-
-void
-nv30_vertprog_destroy(struct nv30_context *nv30, struct nv30_vertex_program *vp)
-{
-	vp->translated = FALSE;
-
-	if (vp->nr_insns) {
-		FREE(vp->insns);
-		vp->insns = NULL;
-		vp->nr_insns = 0;
-	}
-
-	if (vp->nr_consts) {
-		FREE(vp->consts);
-		vp->consts = NULL;
-		vp->nr_consts = 0;
-	}
-
-	nouveau_resource_free(&vp->exec);
-	vp->exec_start = 0;
-	nouveau_resource_free(&vp->data);
-	vp->data_start = 0;
-	vp->data_start_min = 0;
-
-	vp->ir = vp->or = 0;
-	so_ref(NULL, &vp->so);
-}
-
-struct nv30_state_entry nv30_state_vertprog = {
-	.validate = nv30_vertprog_validate,
-	.dirty = {
-		.pipe = NV30_NEW_VERTPROG /*| NV30_NEW_UCP*/,
-		.hw = NV30_STATE_VERTPROG,
-	}
-};
diff --git a/src/gallium/drivers/nv40/Makefile b/src/gallium/drivers/nv40/Makefile
deleted file mode 100644
index 0ecae2b491..0000000000
--- a/src/gallium/drivers/nv40/Makefile
+++ /dev/null
@@ -1,29 +0,0 @@
-TOP = ../../../..
-include $(TOP)/configs/current
-
-LIBNAME = nv40
-
-C_SOURCES = \
-	nv40_clear.c \
-	nv40_context.c \
-	nv40_draw.c \
-	nv40_fragprog.c \
-	nv40_fragtex.c \
-	nv40_miptree.c \
-	nv40_query.c \
-	nv40_screen.c \
-	nv40_state.c \
-	nv40_state_blend.c \
-	nv40_state_emit.c \
-	nv40_state_fb.c \
-	nv40_state_rasterizer.c \
-	nv40_state_scissor.c \
-	nv40_state_stipple.c \
-	nv40_state_viewport.c \
-	nv40_state_zsa.c \
-	nv40_surface.c \
-	nv40_transfer.c \
-	nv40_vbo.c \
-	nv40_vertprog.c
-
-include ../../Makefile.template
diff --git a/src/gallium/drivers/nv40/nv40_context.c b/src/gallium/drivers/nv40/nv40_context.c
deleted file mode 100644
index e828f17643..0000000000
--- a/src/gallium/drivers/nv40/nv40_context.c
+++ /dev/null
@@ -1,88 +0,0 @@
-#include "draw/draw_context.h"
-#include "pipe/p_defines.h"
-
-#include "nv40_context.h"
-#include "nv40_screen.h"
-
-static void
-nv40_flush(struct pipe_context *pipe, unsigned flags,
-	   struct pipe_fence_handle **fence)
-{
-	struct nv40_context *nv40 = nv40_context(pipe);
-	struct nv40_screen *screen = nv40->screen;
-	struct nouveau_channel *chan = screen->base.channel;
-	struct nouveau_grobj *curie = screen->curie;
-
-	if (flags & PIPE_FLUSH_TEXTURE_CACHE) {
-		BEGIN_RING(chan, curie, 0x1fd8, 1);
-		OUT_RING  (chan, 2);
-		BEGIN_RING(chan, curie, 0x1fd8, 1);
-		OUT_RING  (chan, 1);
-	}
-
-	FIRE_RING(chan);
-	if (fence)
-		*fence = NULL;
-}
-
-static void
-nv40_destroy(struct pipe_context *pipe)
-{
-	struct nv40_context *nv40 = nv40_context(pipe);
-	unsigned i;
-
-	for (i = 0; i < NV40_STATE_MAX; i++) {
-		if (nv40->state.hw[i])
-			so_ref(NULL, &nv40->state.hw[i]);
-	}
-
-	if (nv40->draw)
-		draw_destroy(nv40->draw);
-	FREE(nv40);
-}
-
-struct pipe_context *
-nv40_create(struct pipe_screen *pscreen, void *priv)
-{
-	struct nv40_screen *screen = nv40_screen(pscreen);
-	struct pipe_winsys *ws = pscreen->winsys;
-	struct nv40_context *nv40;
-	struct nouveau_winsys *nvws = screen->nvws;
-
-	nv40 = CALLOC(1, sizeof(struct nv40_context));
-	if (!nv40)
-		return NULL;
-	nv40->screen = screen;
-
-	nv40->nvws = nvws;
-
-	nv40->pipe.winsys = ws;
-	nv40->pipe.priv = priv;
-	nv40->pipe.screen = pscreen;
-	nv40->pipe.destroy = nv40_destroy;
-	nv40->pipe.draw_arrays = nv40_draw_arrays;
-	nv40->pipe.draw_elements = nv40_draw_elements;
-	nv40->pipe.clear = nv40_clear;
-	nv40->pipe.flush = nv40_flush;
-
-	nv40->pipe.is_texture_referenced = nouveau_is_texture_referenced;
-	nv40->pipe.is_buffer_referenced = nouveau_is_buffer_referenced;
-
-	screen->base.channel->user_private = nv40;
-	screen->base.channel->flush_notify = nv40_state_flush_notify;
-
-	nv40_init_query_functions(nv40);
-	nv40_init_surface_functions(nv40);
-	nv40_init_state_functions(nv40);
-	nv40_init_transfer_functions(nv40);
-
-	/* Create, configure, and install fallback swtnl path */
-	nv40->draw = draw_create();
-	draw_wide_point_threshold(nv40->draw, 9999999.0);
-	draw_wide_line_threshold(nv40->draw, 9999999.0);
-	draw_enable_line_stipple(nv40->draw, FALSE);
-	draw_enable_point_sprites(nv40->draw, FALSE);
-	draw_set_rasterize_stage(nv40->draw, nv40_draw_render_stage(nv40));
-
-	return &nv40->pipe;
-}
diff --git a/src/gallium/drivers/nv40/nv40_context.h b/src/gallium/drivers/nv40/nv40_context.h
deleted file mode 100644
index 1e1d64ee49..0000000000
--- a/src/gallium/drivers/nv40/nv40_context.h
+++ /dev/null
@@ -1,246 +0,0 @@
-#ifndef __NV40_CONTEXT_H__
-#define __NV40_CONTEXT_H__
-
-#include <stdio.h>
-
-#include "pipe/p_context.h"
-#include "pipe/p_defines.h"
-#include "pipe/p_state.h"
-#include "pipe/p_compiler.h"
-
-#include "util/u_memory.h"
-#include "util/u_math.h"
-#include "util/u_inlines.h"
-
-#include "draw/draw_vertex.h"
-
-#include "nouveau/nouveau_winsys.h"
-#include "nouveau/nouveau_gldefs.h"
-#include "nouveau/nouveau_context.h"
-#include "nouveau/nouveau_stateobj.h"
-
-#include "nv40_state.h"
-
-#define NOUVEAU_ERR(fmt, args...) \
-	fprintf(stderr, "%s:%d -  "fmt, __func__, __LINE__, ##args);
-#define NOUVEAU_MSG(fmt, args...) \
-	fprintf(stderr, "nouveau: "fmt, ##args);
-
-enum nv40_state_index {
-	NV40_STATE_FB = 0,
-	NV40_STATE_VIEWPORT = 1,
-	NV40_STATE_BLEND = 2,
-	NV40_STATE_RAST = 3,
-	NV40_STATE_ZSA = 4,
-	NV40_STATE_BCOL = 5,
-	NV40_STATE_CLIP = 6,
-	NV40_STATE_SCISSOR = 7,
-	NV40_STATE_STIPPLE = 8,
-	NV40_STATE_FRAGPROG = 9,
-	NV40_STATE_VERTPROG = 10,
-	NV40_STATE_FRAGTEX0 = 11,
-	NV40_STATE_FRAGTEX1 = 12,
-	NV40_STATE_FRAGTEX2 = 13,
-	NV40_STATE_FRAGTEX3 = 14,
-	NV40_STATE_FRAGTEX4 = 15,
-	NV40_STATE_FRAGTEX5 = 16,
-	NV40_STATE_FRAGTEX6 = 17,
-	NV40_STATE_FRAGTEX7 = 18,
-	NV40_STATE_FRAGTEX8 = 19,
-	NV40_STATE_FRAGTEX9 = 20,
-	NV40_STATE_FRAGTEX10 = 21,
-	NV40_STATE_FRAGTEX11 = 22,
-	NV40_STATE_FRAGTEX12 = 23,
-	NV40_STATE_FRAGTEX13 = 24,
-	NV40_STATE_FRAGTEX14 = 25,
-	NV40_STATE_FRAGTEX15 = 26,
-	NV40_STATE_VERTTEX0 = 27,
-	NV40_STATE_VERTTEX1 = 28,
-	NV40_STATE_VERTTEX2 = 29,
-	NV40_STATE_VERTTEX3 = 30,
-	NV40_STATE_VTXBUF = 31,
-	NV40_STATE_VTXFMT = 32,
-	NV40_STATE_VTXATTR = 33,
-	NV40_STATE_SR = 34,
-	NV40_STATE_MAX = 35
-};
-
-#include "nv40_screen.h"
-
-#define NV40_NEW_BLEND		(1 <<  0)
-#define NV40_NEW_RAST		(1 <<  1)
-#define NV40_NEW_ZSA		(1 <<  2)
-#define NV40_NEW_SAMPLER	(1 <<  3)
-#define NV40_NEW_FB		(1 <<  4)
-#define NV40_NEW_STIPPLE	(1 <<  5)
-#define NV40_NEW_SCISSOR	(1 <<  6)
-#define NV40_NEW_VIEWPORT	(1 <<  7)
-#define NV40_NEW_BCOL		(1 <<  8)
-#define NV40_NEW_VERTPROG	(1 <<  9)
-#define NV40_NEW_FRAGPROG	(1 << 10)
-#define NV40_NEW_ARRAYS		(1 << 11)
-#define NV40_NEW_UCP		(1 << 12)
-#define NV40_NEW_SR		(1 << 13)
-
-struct nv40_rasterizer_state {
-	struct pipe_rasterizer_state pipe;
-	struct nouveau_stateobj *so;
-};
-
-struct nv40_zsa_state {
-	struct pipe_depth_stencil_alpha_state pipe;
-	struct nouveau_stateobj *so;
-};
-
-struct nv40_blend_state {
-	struct pipe_blend_state pipe;
-	struct nouveau_stateobj *so;
-};
-
-
-struct nv40_state {
-	unsigned scissor_enabled;
-	unsigned stipple_enabled;
-	unsigned fp_samplers;
-
-	uint64_t dirty;
-	struct nouveau_stateobj *hw[NV40_STATE_MAX];
-};
-
-
-struct nv40_vtxelt_state {
-	struct pipe_vertex_element pipe[16];
-	unsigned num_elements;
-};
-
-struct nv40_context {
-	struct pipe_context pipe;
-
-	struct nouveau_winsys *nvws;
-	struct nv40_screen *screen;
-
-	struct draw_context *draw;
-
-	/* HW state derived from pipe states */
-	struct nv40_state state;
-	struct {
-		struct nv40_vertex_program *vertprog;
-
-		unsigned nr_attribs;
-		unsigned hw[PIPE_MAX_SHADER_INPUTS];
-		unsigned draw[PIPE_MAX_SHADER_INPUTS];
-		unsigned emit[PIPE_MAX_SHADER_INPUTS];
-	} swtnl;
-
-	enum {
-		HW, SWTNL, SWRAST
-	} render_mode;
-	unsigned fallback_swtnl;
-	unsigned fallback_swrast;
-
-	/* Context state */
-	unsigned dirty, draw_dirty;
-	struct pipe_scissor_state scissor;
-	unsigned stipple[32];
-	struct pipe_clip_state clip;
-	struct nv40_vertex_program *vertprog;
-	struct nv40_fragment_program *fragprog;
-	struct pipe_buffer *constbuf[PIPE_SHADER_TYPES];
-	unsigned constbuf_nr[PIPE_SHADER_TYPES];
-	struct nv40_rasterizer_state *rasterizer;
-	struct nv40_zsa_state *zsa;
-	struct nv40_blend_state *blend;
-	struct pipe_blend_color blend_colour;
-	struct pipe_stencil_ref stencil_ref;
-	struct pipe_viewport_state viewport;
-	struct pipe_framebuffer_state framebuffer;
-	struct pipe_buffer *idxbuf;
-	unsigned idxbuf_format;
-	struct nv40_sampler_state *tex_sampler[PIPE_MAX_SAMPLERS];
-	struct nv40_miptree *tex_miptree[PIPE_MAX_SAMPLERS];
-	struct pipe_sampler_view *fragment_sampler_views[PIPE_MAX_SAMPLERS];
-	unsigned nr_samplers;
-	unsigned nr_textures;
-	unsigned dirty_samplers;
-	struct pipe_vertex_buffer vtxbuf[PIPE_MAX_ATTRIBS];
-	unsigned vtxbuf_nr;
-	struct nv40_vtxelt_state *vtxelt;
-};
-
-static INLINE struct nv40_context *
-nv40_context(struct pipe_context *pipe)
-{
-	return (struct nv40_context *)pipe;
-}
-
-struct nv40_state_entry {
-	boolean (*validate)(struct nv40_context *nv40);
-	struct {
-		unsigned pipe;
-		unsigned hw;
-	} dirty;
-};
-
-extern void nv40_init_state_functions(struct nv40_context *nv40);
-extern void nv40_init_surface_functions(struct nv40_context *nv40);
-extern void nv40_init_query_functions(struct nv40_context *nv40);
-extern void nv40_init_transfer_functions(struct nv40_context *nv40);
-
-extern void nv40_screen_init_miptree_functions(struct pipe_screen *pscreen);
-
-/* nv40_draw.c */
-extern struct draw_stage *nv40_draw_render_stage(struct nv40_context *nv40);
-extern void nv40_draw_elements_swtnl(struct pipe_context *pipe,
-					struct pipe_buffer *idxbuf,
-					unsigned ib_size, unsigned mode,
-					unsigned start, unsigned count);
-
-/* nv40_vertprog.c */
-extern void nv40_vertprog_destroy(struct nv40_context *,
-				  struct nv40_vertex_program *);
-
-/* nv40_fragprog.c */
-extern void nv40_fragprog_destroy(struct nv40_context *,
-				  struct nv40_fragment_program *);
-
-/* nv40_fragtex.c */
-extern void nv40_fragtex_bind(struct nv40_context *);
-
-/* nv40_state.c and friends */
-extern boolean nv40_state_validate(struct nv40_context *nv40);
-extern boolean nv40_state_validate_swtnl(struct nv40_context *nv40);
-extern void nv40_state_emit(struct nv40_context *nv40);
-extern void nv40_state_flush_notify(struct nouveau_channel *chan);
-extern struct nv40_state_entry nv40_state_rasterizer;
-extern struct nv40_state_entry nv40_state_scissor;
-extern struct nv40_state_entry nv40_state_stipple;
-extern struct nv40_state_entry nv40_state_fragprog;
-extern struct nv40_state_entry nv40_state_vertprog;
-extern struct nv40_state_entry nv40_state_blend;
-extern struct nv40_state_entry nv40_state_blend_colour;
-extern struct nv40_state_entry nv40_state_zsa;
-extern struct nv40_state_entry nv40_state_viewport;
-extern struct nv40_state_entry nv40_state_framebuffer;
-extern struct nv40_state_entry nv40_state_fragtex;
-extern struct nv40_state_entry nv40_state_vbo;
-extern struct nv40_state_entry nv40_state_vtxfmt;
-extern struct nv40_state_entry nv40_state_sr;
-
-/* nv40_vbo.c */
-extern void nv40_draw_arrays(struct pipe_context *, unsigned mode,
-				unsigned start, unsigned count);
-extern void nv40_draw_elements(struct pipe_context *pipe,
-				  struct pipe_buffer *indexBuffer,
-				  unsigned indexSize,
-				  unsigned mode, unsigned start,
-				  unsigned count);
-
-/* nv40_clear.c */
-extern void nv40_clear(struct pipe_context *pipe, unsigned buffers,
-		       const float *rgba, double depth, unsigned stencil);
-
-/* nv40_context.c */
-struct pipe_context *
-nv40_create(struct pipe_screen *pscreen, void *priv);
-
-#endif
diff --git a/src/gallium/drivers/nv40/nv40_draw.c b/src/gallium/drivers/nv40/nv40_draw.c
deleted file mode 100644
index 48bd84d16c..0000000000
--- a/src/gallium/drivers/nv40/nv40_draw.c
+++ /dev/null
@@ -1,360 +0,0 @@
-#include "pipe/p_shader_tokens.h"
-#include "util/u_inlines.h"
-
-#include "util/u_pack_color.h"
-
-#include "draw/draw_context.h"
-#include "draw/draw_vertex.h"
-#include "draw/draw_pipe.h"
-
-#include "nv40_context.h"
-#define NV40_SHADER_NO_FUCKEDNESS
-#include "nv40_shader.h"
-
-/* Simple, but crappy, swtnl path, hopefully we wont need to hit this very
- * often at all.  Uses "quadro style" vertex submission + a fixed vertex
- * layout to avoid the need to generate a vertex program or vtxfmt.
- */
-
-struct nv40_render_stage {
-	struct draw_stage stage;
-	struct nv40_context *nv40;
-	unsigned prim;
-};
-
-static INLINE struct nv40_render_stage *
-nv40_render_stage(struct draw_stage *stage)
-{
-	return (struct nv40_render_stage *)stage;
-}
-
-static INLINE void
-nv40_render_vertex(struct nv40_context *nv40, const struct vertex_header *v)
-{
-	struct nv40_screen *screen = nv40->screen;
-	struct nouveau_channel *chan = screen->base.channel;
-	struct nouveau_grobj *curie = screen->curie;
-	unsigned i;
-
-	for (i = 0; i < nv40->swtnl.nr_attribs; i++) {
-		unsigned idx = nv40->swtnl.draw[i];
-		unsigned hw = nv40->swtnl.hw[i];
-
-		switch (nv40->swtnl.emit[i]) {
-		case EMIT_OMIT:
-			break;
-		case EMIT_1F:
-			BEGIN_RING(chan, curie, NV40TCL_VTX_ATTR_1F(hw), 1);
-			OUT_RING  (chan, fui(v->data[idx][0]));
-			break;
-		case EMIT_2F:
-			BEGIN_RING(chan, curie, NV40TCL_VTX_ATTR_2F_X(hw), 2);
-			OUT_RING  (chan, fui(v->data[idx][0]));
-			OUT_RING  (chan, fui(v->data[idx][1]));
-			break;
-		case EMIT_3F:
-			BEGIN_RING(chan, curie, NV40TCL_VTX_ATTR_3F_X(hw), 3);
-			OUT_RING  (chan, fui(v->data[idx][0]));
-			OUT_RING  (chan, fui(v->data[idx][1]));
-			OUT_RING  (chan, fui(v->data[idx][2]));
-			break;
-		case EMIT_4F:
-			BEGIN_RING(chan, curie, NV40TCL_VTX_ATTR_4F_X(hw), 4);
-			OUT_RING  (chan, fui(v->data[idx][0]));
-			OUT_RING  (chan, fui(v->data[idx][1]));
-			OUT_RING  (chan, fui(v->data[idx][2]));
-			OUT_RING  (chan, fui(v->data[idx][3]));
-			break;
-		case EMIT_4UB:
-			BEGIN_RING(chan, curie, NV40TCL_VTX_ATTR_4UB(hw), 1);
-			OUT_RING  (chan, pack_ub4(float_to_ubyte(v->data[idx][0]),
-					    float_to_ubyte(v->data[idx][1]),
-					    float_to_ubyte(v->data[idx][2]),
-					    float_to_ubyte(v->data[idx][3])));
-			break;
-		default:
-			assert(0);
-			break;
-		}
-	}
-}
-
-static INLINE void
-nv40_render_prim(struct draw_stage *stage, struct prim_header *prim,
-	       unsigned mode, unsigned count)
-{
-	struct nv40_render_stage *rs = nv40_render_stage(stage);
-	struct nv40_context *nv40 = rs->nv40;
-
-	struct nv40_screen *screen = nv40->screen;
-	struct nouveau_channel *chan = screen->base.channel;
-	struct nouveau_grobj *curie = screen->curie;
-	unsigned i;
-
-	/* Ensure there's room for 4xfloat32 + potentially 3 begin/end */
-	if (AVAIL_RING(chan) < ((count * 20) + 6)) {
-		if (rs->prim != NV40TCL_BEGIN_END_STOP) {
-			NOUVEAU_ERR("AIII, missed flush\n");
-			assert(0);
-		}
-		FIRE_RING(chan);
-		nv40_state_emit(nv40);
-	}
-
-	/* Switch primitive modes if necessary */
-	if (rs->prim != mode) {
-		if (rs->prim != NV40TCL_BEGIN_END_STOP) {
-			BEGIN_RING(chan, curie, NV40TCL_BEGIN_END, 1);
-			OUT_RING  (chan, NV40TCL_BEGIN_END_STOP);
-		}
-
-		BEGIN_RING(chan, curie, NV40TCL_BEGIN_END, 1);
-		OUT_RING  (chan, mode);
-		rs->prim = mode;
-	}
-
-	/* Emit vertex data */
-	for (i = 0; i < count; i++)
-		nv40_render_vertex(nv40, prim->v[i]);
-
-	/* If it's likely we'll need to empty the push buffer soon, finish
-	 * off the primitive now.
-	 */
-	if (AVAIL_RING(chan) < ((count * 20) + 6)) {
-		BEGIN_RING(chan, curie, NV40TCL_BEGIN_END, 1);
-		OUT_RING  (chan, NV40TCL_BEGIN_END_STOP);
-		rs->prim = NV40TCL_BEGIN_END_STOP;
-	}
-}
-
-static void
-nv40_render_point(struct draw_stage *draw, struct prim_header *prim)
-{
-	nv40_render_prim(draw, prim, NV40TCL_BEGIN_END_POINTS, 1);
-}
-
-static void
-nv40_render_line(struct draw_stage *draw, struct prim_header *prim)
-{
-	nv40_render_prim(draw, prim, NV40TCL_BEGIN_END_LINES, 2);
-}
-
-static void
-nv40_render_tri(struct draw_stage *draw, struct prim_header *prim)
-{
-	nv40_render_prim(draw, prim, NV40TCL_BEGIN_END_TRIANGLES, 3);
-}
-
-static void
-nv40_render_flush(struct draw_stage *draw, unsigned flags)
-{
-	struct nv40_render_stage *rs = nv40_render_stage(draw);
-	struct nv40_context *nv40 = rs->nv40;
-	struct nv40_screen *screen = nv40->screen;
-	struct nouveau_channel *chan = screen->base.channel;
-	struct nouveau_grobj *curie = screen->curie;
-
-	if (rs->prim != NV40TCL_BEGIN_END_STOP) {
-		BEGIN_RING(chan, curie, NV40TCL_BEGIN_END, 1);
-		OUT_RING  (chan, NV40TCL_BEGIN_END_STOP);
-		rs->prim = NV40TCL_BEGIN_END_STOP;
-	}
-}
-
-static void
-nv40_render_reset_stipple_counter(struct draw_stage *draw)
-{
-}
-
-static void
-nv40_render_destroy(struct draw_stage *draw)
-{
-	FREE(draw);
-}
-
-static INLINE void
-emit_mov(struct nv40_vertex_program *vp,
-	 unsigned dst, unsigned src, unsigned vor, unsigned mask)
-{
-	struct nv40_vertex_program_exec *inst;
-
-	vp->insns = realloc(vp->insns,
-			    sizeof(struct nv40_vertex_program_exec) *
-			    ++vp->nr_insns);
-	inst = &vp->insns[vp->nr_insns - 1];
-
-	inst->data[0] = 0x401f9c6c;
-	inst->data[1] = 0x0040000d | (src << 8);
-	inst->data[2] = 0x8106c083;
-	inst->data[3] = 0x6041ff80 | (dst << 2) | (mask << 13);
-	inst->const_index = -1;
-	inst->has_branch_offset = FALSE;
-
-	vp->ir |= (1 << src);
-	if (vor != ~0)
-		vp->or |= (1 << vor);
-}
-
-static struct nv40_vertex_program *
-create_drawvp(struct nv40_context *nv40)
-{
-	struct nv40_vertex_program *vp = CALLOC_STRUCT(nv40_vertex_program);
-	unsigned i;
-
-	emit_mov(vp, NV40_VP_INST_DEST_POS, 0, ~0, 0xf);
-	emit_mov(vp, NV40_VP_INST_DEST_COL0, 3, 0, 0xf);
-	emit_mov(vp, NV40_VP_INST_DEST_COL1, 4, 1, 0xf);
-	emit_mov(vp, NV40_VP_INST_DEST_BFC0, 3, 2, 0xf);
-	emit_mov(vp, NV40_VP_INST_DEST_BFC1, 4, 3, 0xf);
-	emit_mov(vp, NV40_VP_INST_DEST_FOGC, 5, 4, 0x8);
-	for (i = 0; i < 8; i++)
-		emit_mov(vp, NV40_VP_INST_DEST_TC(i), 8 + i, 14 + i, 0xf);
-
-	vp->insns[vp->nr_insns - 1].data[3] |= 1;
-	vp->translated = TRUE;
-	return vp;
-}
-
-struct draw_stage *
-nv40_draw_render_stage(struct nv40_context *nv40)
-{
-	struct nv40_render_stage *render = CALLOC_STRUCT(nv40_render_stage);
-
-	if (!nv40->swtnl.vertprog)
-		nv40->swtnl.vertprog = create_drawvp(nv40);
-
-	render->nv40 = nv40;
-	render->stage.draw = nv40->draw;
-	render->stage.point = nv40_render_point;
-	render->stage.line = nv40_render_line;
-	render->stage.tri = nv40_render_tri;
-	render->stage.flush = nv40_render_flush;
-	render->stage.reset_stipple_counter = nv40_render_reset_stipple_counter;
-	render->stage.destroy = nv40_render_destroy;
-
-	return &render->stage;
-}
-
-void
-nv40_draw_elements_swtnl(struct pipe_context *pipe,
-			 struct pipe_buffer *idxbuf, unsigned idxbuf_size,
-			 unsigned mode, unsigned start, unsigned count)
-{
-	struct nv40_context *nv40 = nv40_context(pipe);
-	struct pipe_screen *pscreen = pipe->screen;
-	unsigned i;
-	void *map;
-
-	if (!nv40_state_validate_swtnl(nv40))
-		return;
-	nv40->state.dirty &= ~(1ULL << NV40_STATE_VTXBUF);
-	nv40_state_emit(nv40);
-
-	for (i = 0; i < nv40->vtxbuf_nr; i++) {
-		map = pipe_buffer_map(pscreen, nv40->vtxbuf[i].buffer,
-                                      PIPE_BUFFER_USAGE_CPU_READ);
-		draw_set_mapped_vertex_buffer(nv40->draw, i, map);
-	}
-
-	if (idxbuf) {
-		map = pipe_buffer_map(pscreen, idxbuf,
-				      PIPE_BUFFER_USAGE_CPU_READ);
-		draw_set_mapped_element_buffer(nv40->draw, idxbuf_size, map);
-	} else {
-		draw_set_mapped_element_buffer(nv40->draw, 0, NULL);
-	}
-
-	if (nv40->constbuf[PIPE_SHADER_VERTEX]) {
-		const unsigned nr = nv40->constbuf_nr[PIPE_SHADER_VERTEX];
-
-		map = pipe_buffer_map(pscreen,
-				      nv40->constbuf[PIPE_SHADER_VERTEX],
-				      PIPE_BUFFER_USAGE_CPU_READ);
-		draw_set_mapped_constant_buffer(nv40->draw, PIPE_SHADER_VERTEX, 0,
-                                                map, nr);
-	}
-
-	draw_arrays(nv40->draw, mode, start, count);
-
-	for (i = 0; i < nv40->vtxbuf_nr; i++)
-		pipe_buffer_unmap(pscreen, nv40->vtxbuf[i].buffer);
-
-	if (idxbuf)
-		pipe_buffer_unmap(pscreen, idxbuf);
-
-	if (nv40->constbuf[PIPE_SHADER_VERTEX])
-		pipe_buffer_unmap(pscreen, nv40->constbuf[PIPE_SHADER_VERTEX]);
-
-	draw_flush(nv40->draw);
-	pipe->flush(pipe, 0, NULL);
-}
-
-static INLINE void
-emit_attrib(struct nv40_context *nv40, unsigned hw, unsigned emit,
-	    unsigned semantic, unsigned index)
-{
-	unsigned draw_out = draw_find_shader_output(nv40->draw, semantic, index);
-	unsigned a = nv40->swtnl.nr_attribs++;
-
-	nv40->swtnl.hw[a] = hw;
-	nv40->swtnl.emit[a] = emit;
-	nv40->swtnl.draw[a] = draw_out;
-}
-
-static boolean
-nv40_state_vtxfmt_validate(struct nv40_context *nv40)
-{
-	struct nv40_fragment_program *fp = nv40->fragprog;
-	unsigned colour = 0, texcoords = 0, fog = 0, i;
-
-	/* Determine needed fragprog inputs */
-	for (i = 0; i < fp->info.num_inputs; i++) {
-		switch (fp->info.input_semantic_name[i]) {
-		case TGSI_SEMANTIC_POSITION:
-			break;
-		case TGSI_SEMANTIC_COLOR:
-			colour |= (1 << fp->info.input_semantic_index[i]);
-			break;
-		case TGSI_SEMANTIC_GENERIC:
-			texcoords |= (1 << fp->info.input_semantic_index[i]);
-			break;
-		case TGSI_SEMANTIC_FOG:
-			fog = 1;
-			break;
-		default:
-			assert(0);
-		}
-	}
-
-	nv40->swtnl.nr_attribs = 0;
-
-	/* Map draw vtxprog output to hw attribute IDs */
-	for (i = 0; i < 2; i++) {
-		if (!(colour & (1 << i)))
-			continue;
-		emit_attrib(nv40, 3 + i, EMIT_4UB, TGSI_SEMANTIC_COLOR, i);
-	}
-
-	for (i = 0; i < 8; i++) {
-		if (!(texcoords & (1 << i)))
-			continue;
-		emit_attrib(nv40, 8 + i, EMIT_4F, TGSI_SEMANTIC_GENERIC, i);
-	}
-
-	if (fog) {
-		emit_attrib(nv40, 5, EMIT_1F, TGSI_SEMANTIC_FOG, 0);
-	}
-
-	emit_attrib(nv40, 0, EMIT_3F, TGSI_SEMANTIC_POSITION, 0);
-
-	return FALSE;
-}
-
-struct nv40_state_entry nv40_state_vtxfmt = {
-	.validate = nv40_state_vtxfmt_validate,
-	.dirty = {
-		.pipe = NV40_NEW_ARRAYS | NV40_NEW_FRAGPROG,
-		.hw = 0
-	}
-};
-
diff --git a/src/gallium/drivers/nv40/nv40_query.c b/src/gallium/drivers/nv40/nv40_query.c
deleted file mode 100644
index 8ed4a67dd0..0000000000
--- a/src/gallium/drivers/nv40/nv40_query.c
+++ /dev/null
@@ -1,127 +0,0 @@
-#include "pipe/p_context.h"
-
-#include "nv40_context.h"
-
-struct nv40_query {
-	struct nouveau_resource *object;
-	unsigned type;
-	boolean ready;
-	uint64_t result;
-};
-
-static INLINE struct nv40_query *
-nv40_query(struct pipe_query *pipe)
-{
-	return (struct nv40_query *)pipe;
-}
-
-static struct pipe_query *
-nv40_query_create(struct pipe_context *pipe, unsigned query_type)
-{
-	struct nv40_query *q;
-
-	q = CALLOC(1, sizeof(struct nv40_query));
-	q->type = query_type;
-
-	return (struct pipe_query *)q;
-}
-
-static void
-nv40_query_destroy(struct pipe_context *pipe, struct pipe_query *pq)
-{
-	struct nv40_query *q = nv40_query(pq);
-
-	if (q->object)
-		nouveau_resource_free(&q->object);
-	FREE(q);
-}
-
-static void
-nv40_query_begin(struct pipe_context *pipe, struct pipe_query *pq)
-{
-	struct nv40_context *nv40 = nv40_context(pipe);
-	struct nv40_query *q = nv40_query(pq);
-	struct nv40_screen *screen = nv40->screen;
-	struct nouveau_channel *chan = screen->base.channel;
-	struct nouveau_grobj *curie = screen->curie;
-
-	assert(q->type == PIPE_QUERY_OCCLUSION_COUNTER);
-
-	/* Happens when end_query() is called, then another begin_query()
-	 * without querying the result in-between.  For now we'll wait for
-	 * the existing query to notify completion, but it could be better.
-	 */
-	if (q->object) {
-		uint64_t tmp;
-		pipe->get_query_result(pipe, pq, 1, &tmp);
-	}
-
-	if (nouveau_resource_alloc(nv40->screen->query_heap, 1, NULL, &q->object))
-		assert(0);
-	nouveau_notifier_reset(nv40->screen->query, q->object->start);
-
-	BEGIN_RING(chan, curie, NV40TCL_QUERY_RESET, 1);
-	OUT_RING  (chan, 1);
-	BEGIN_RING(chan, curie, NV40TCL_QUERY_UNK17CC, 1);
-	OUT_RING  (chan, 1);
-
-	q->ready = FALSE;
-}
-
-static void
-nv40_query_end(struct pipe_context *pipe, struct pipe_query *pq)
-{
-	struct nv40_context *nv40 = nv40_context(pipe);
-	struct nv40_query *q = nv40_query(pq);
-	struct nv40_screen *screen = nv40->screen;
-	struct nouveau_channel *chan = screen->base.channel;
-	struct nouveau_grobj *curie = screen->curie;
-
-	BEGIN_RING(chan, curie, NV40TCL_QUERY_GET, 1);
-	OUT_RING  (chan, (0x01 << NV40TCL_QUERY_GET_UNK24_SHIFT) |
-		   ((q->object->start * 32) << NV40TCL_QUERY_GET_OFFSET_SHIFT));
-	FIRE_RING(chan);
-}
-
-static boolean
-nv40_query_result(struct pipe_context *pipe, struct pipe_query *pq,
-		  boolean wait, uint64_t *result)
-{
-	struct nv40_context *nv40 = nv40_context(pipe);
-	struct nv40_query *q = nv40_query(pq);
-
-	assert(q->object && q->type == PIPE_QUERY_OCCLUSION_COUNTER);
-
-	if (!q->ready) {
-		unsigned status;
-
-		status = nouveau_notifier_status(nv40->screen->query,
-						 q->object->start);
-		if (status != NV_NOTIFY_STATE_STATUS_COMPLETED) {
-			if (wait == FALSE)
-				return FALSE;
-			nouveau_notifier_wait_status(nv40->screen->query,
-					      q->object->start,
-					      NV_NOTIFY_STATE_STATUS_COMPLETED,
-					      0);
-		}
-
-		q->result = nouveau_notifier_return_val(nv40->screen->query,
-							q->object->start);
-		q->ready = TRUE;
-		nouveau_resource_free(&q->object);
-	}
-
-	*result = q->result;
-	return TRUE;
-}
-
-void
-nv40_init_query_functions(struct nv40_context *nv40)
-{
-	nv40->pipe.create_query = nv40_query_create;
-	nv40->pipe.destroy_query = nv40_query_destroy;
-	nv40->pipe.begin_query = nv40_query_begin;
-	nv40->pipe.end_query = nv40_query_end;
-	nv40->pipe.get_query_result = nv40_query_result;
-}
diff --git a/src/gallium/drivers/nv40/nv40_screen.c b/src/gallium/drivers/nv40/nv40_screen.c
deleted file mode 100644
index dbcc33d8d9..0000000000
--- a/src/gallium/drivers/nv40/nv40_screen.c
+++ /dev/null
@@ -1,319 +0,0 @@
-#include "pipe/p_screen.h"
-
-#include "nv40_context.h"
-#include "nv40_screen.h"
-
-#define NV4X_GRCLASS4097_CHIPSETS 0x00000baf
-#define NV4X_GRCLASS4497_CHIPSETS 0x00005450
-#define NV6X_GRCLASS4497_CHIPSETS 0x00000088
-
-static int
-nv40_screen_get_param(struct pipe_screen *pscreen, int param)
-{
-	struct nv40_screen *screen = nv40_screen(pscreen);
-
-	switch (param) {
-	case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS:
-		return 16;
-	case PIPE_CAP_NPOT_TEXTURES:
-		return 1;
-	case PIPE_CAP_TWO_SIDED_STENCIL:
-		return 1;
-	case PIPE_CAP_GLSL:
-		return 0;
-	case PIPE_CAP_ANISOTROPIC_FILTER:
-		return 1;
-	case PIPE_CAP_POINT_SPRITE:
-		return 1;
-	case PIPE_CAP_MAX_RENDER_TARGETS:
-		return 4;
-	case PIPE_CAP_OCCLUSION_QUERY:
-		return 1;
-	case PIPE_CAP_TEXTURE_SHADOW_MAP:
-		return 1;
-	case PIPE_CAP_MAX_TEXTURE_2D_LEVELS:
-		return 13;
-	case PIPE_CAP_MAX_TEXTURE_3D_LEVELS:
-		return 10;
-	case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS:
-		return 13;
-	case PIPE_CAP_TEXTURE_MIRROR_CLAMP:
-	case PIPE_CAP_TEXTURE_MIRROR_REPEAT:
-		return 1;
-	case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS:
-		return 0; /* We have 4 - but unsupported currently */
-	case PIPE_CAP_TGSI_CONT_SUPPORTED:
-		return 0;
-	case PIPE_CAP_BLEND_EQUATION_SEPARATE:
-		return 1;
-	case NOUVEAU_CAP_HW_VTXBUF:
-		return 1;
-	case NOUVEAU_CAP_HW_IDXBUF:
-		if (screen->curie->grclass == NV40TCL)
-			return 1;
-		return 0;
-	case PIPE_CAP_INDEP_BLEND_ENABLE:
-		return 0;
-	case PIPE_CAP_INDEP_BLEND_FUNC:
-		return 0;
-	case PIPE_CAP_TGSI_FS_COORD_ORIGIN_LOWER_LEFT:
-	case PIPE_CAP_TGSI_FS_COORD_PIXEL_CENTER_HALF_INTEGER:
-		return 1;
-	case PIPE_CAP_TGSI_FS_COORD_ORIGIN_UPPER_LEFT:
-	case PIPE_CAP_TGSI_FS_COORD_PIXEL_CENTER_INTEGER:
-		return 0;
-	case PIPE_CAP_MAX_COMBINED_SAMPLERS:
-		return 16;
-	default:
-		NOUVEAU_ERR("Unknown PIPE_CAP %d\n", param);
-		return 0;
-	}
-}
-
-static float
-nv40_screen_get_paramf(struct pipe_screen *pscreen, int param)
-{
-	switch (param) {
-	case PIPE_CAP_MAX_LINE_WIDTH:
-	case PIPE_CAP_MAX_LINE_WIDTH_AA:
-		return 10.0;
-	case PIPE_CAP_MAX_POINT_WIDTH:
-	case PIPE_CAP_MAX_POINT_WIDTH_AA:
-		return 64.0;
-	case PIPE_CAP_MAX_TEXTURE_ANISOTROPY:
-		return 16.0;
-	case PIPE_CAP_MAX_TEXTURE_LOD_BIAS:
-		return 16.0;
-	default:
-		NOUVEAU_ERR("Unknown PIPE_CAP %d\n", param);
-		return 0.0;
-	}
-}
-
-static boolean
-nv40_screen_surface_format_supported(struct pipe_screen *pscreen,
-				     enum pipe_format format,
-				     enum pipe_texture_target target,
-				     unsigned tex_usage, unsigned geom_flags)
-{
-	if (tex_usage & PIPE_TEXTURE_USAGE_RENDER_TARGET) {
-		switch (format) {
-		case PIPE_FORMAT_B8G8R8A8_UNORM:
-		case PIPE_FORMAT_B5G6R5_UNORM: 
-			return TRUE;
-		default:
-			break;
-		}
-	} else
-	if (tex_usage & PIPE_TEXTURE_USAGE_DEPTH_STENCIL) {
-		switch (format) {
-		case PIPE_FORMAT_S8Z24_UNORM:
-		case PIPE_FORMAT_X8Z24_UNORM:
-		case PIPE_FORMAT_Z16_UNORM:
-			return TRUE;
-		default:
-			break;
-		}
-	} else {
-		switch (format) {
-		case PIPE_FORMAT_B8G8R8A8_UNORM:
-		case PIPE_FORMAT_B5G5R5A1_UNORM:
-		case PIPE_FORMAT_B4G4R4A4_UNORM:
-		case PIPE_FORMAT_B5G6R5_UNORM:
-		case PIPE_FORMAT_R16_SNORM:
-		case PIPE_FORMAT_L8_UNORM:
-		case PIPE_FORMAT_A8_UNORM:
-		case PIPE_FORMAT_I8_UNORM:
-		case PIPE_FORMAT_L8A8_UNORM:
-		case PIPE_FORMAT_Z16_UNORM:
-		case PIPE_FORMAT_S8Z24_UNORM:
-		case PIPE_FORMAT_DXT1_RGB:
-		case PIPE_FORMAT_DXT1_RGBA:
-		case PIPE_FORMAT_DXT3_RGBA:
-		case PIPE_FORMAT_DXT5_RGBA:
-			return TRUE;
-		default:
-			break;
-		}
-	}
-
-	return FALSE;
-}
-
-static struct pipe_buffer *
-nv40_surface_buffer(struct pipe_surface *surf)
-{
-	struct nv40_miptree *mt = (struct nv40_miptree *)surf->texture;
-
-	return mt->buffer;
-}
-
-static void
-nv40_screen_destroy(struct pipe_screen *pscreen)
-{
-	struct nv40_screen *screen = nv40_screen(pscreen);
-	unsigned i;
-
-	for (i = 0; i < NV40_STATE_MAX; i++) {
-		if (screen->state[i])
-			so_ref(NULL, &screen->state[i]);
-	}
-
-	nouveau_resource_destroy(&screen->vp_exec_heap);
-	nouveau_resource_destroy(&screen->vp_data_heap);
-	nouveau_resource_destroy(&screen->query_heap);
-	nouveau_notifier_free(&screen->query);
-	nouveau_notifier_free(&screen->sync);
-	nouveau_grobj_free(&screen->curie);
-	nv04_surface_2d_takedown(&screen->eng2d);
-
-	nouveau_screen_fini(&screen->base);
-
-	FREE(pscreen);
-}
-
-struct pipe_screen *
-nv40_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev)
-{
-	struct nv40_screen *screen = CALLOC_STRUCT(nv40_screen);
-	struct nouveau_channel *chan;
-	struct pipe_screen *pscreen;
-	struct nouveau_stateobj *so;
-	unsigned curie_class = 0;
-	int ret;
-
-	if (!screen)
-		return NULL;
-	pscreen = &screen->base.base;
-
-	ret = nouveau_screen_init(&screen->base, dev);
-	if (ret) {
-		nv40_screen_destroy(pscreen);
-		return NULL;
-	}
-	chan = screen->base.channel;
-
-	pscreen->winsys = ws;
-	pscreen->destroy = nv40_screen_destroy;
-	pscreen->get_param = nv40_screen_get_param;
-	pscreen->get_paramf = nv40_screen_get_paramf;
-	pscreen->is_format_supported = nv40_screen_surface_format_supported;
-	pscreen->context_create = nv40_create;
-
-	nv40_screen_init_miptree_functions(pscreen);
-
-	/* 3D object */
-	switch (dev->chipset & 0xf0) {
-	case 0x40:
-		if (NV4X_GRCLASS4097_CHIPSETS & (1 << (dev->chipset & 0x0f)))
-			curie_class = NV40TCL;
-		else
-		if (NV4X_GRCLASS4497_CHIPSETS & (1 << (dev->chipset & 0x0f)))
-			curie_class = NV44TCL;
-		break;
-	case 0x60:
-		if (NV6X_GRCLASS4497_CHIPSETS & (1 << (dev->chipset & 0x0f)))
-			curie_class = NV44TCL;
-		break;
-	}
-
-	if (!curie_class) {
-		NOUVEAU_ERR("Unknown nv4x chipset: nv%02x\n", dev->chipset);
-		return NULL;
-	}
-
-	ret = nouveau_grobj_alloc(chan, 0xbeef3097, curie_class, &screen->curie);
-	if (ret) {
-		NOUVEAU_ERR("Error creating 3D object: %d\n", ret);
-		return FALSE;
-	}
-
-	/* 2D engine setup */
-	screen->eng2d = nv04_surface_2d_init(&screen->base);
-	screen->eng2d->buf = nv40_surface_buffer;
-
-	/* Notifier for sync purposes */
-	ret = nouveau_notifier_alloc(chan, 0xbeef0301, 1, &screen->sync);
-	if (ret) {
-		NOUVEAU_ERR("Error creating notifier object: %d\n", ret);
-		nv40_screen_destroy(pscreen);
-		return NULL;
-	}
-
-	/* Query objects */
-	ret = nouveau_notifier_alloc(chan, 0xbeef0302, 32, &screen->query);
-	if (ret) {
-		NOUVEAU_ERR("Error initialising query objects: %d\n", ret);
-		nv40_screen_destroy(pscreen);
-		return NULL;
-	}
-
-	nouveau_resource_init(&screen->query_heap, 0, 32);
-	if (ret) {
-		NOUVEAU_ERR("Error initialising query object heap: %d\n", ret);
-		nv40_screen_destroy(pscreen);
-		return NULL;
-	}
-
-	/* Vtxprog resources */
-	if (nouveau_resource_init(&screen->vp_exec_heap, 0, 512) ||
-	    nouveau_resource_init(&screen->vp_data_heap, 0, 256)) {
-		nv40_screen_destroy(pscreen);
-		return NULL;
-	}
-
-	/* Static curie initialisation */
-	so = so_new(16, 25, 0);
-	so_method(so, screen->curie, NV40TCL_DMA_NOTIFY, 1);
-	so_data  (so, screen->sync->handle);
-	so_method(so, screen->curie, NV40TCL_DMA_TEXTURE0, 2);
-	so_data  (so, chan->vram->handle);
-	so_data  (so, chan->gart->handle);
-	so_method(so, screen->curie, NV40TCL_DMA_COLOR1, 1);
-	so_data  (so, chan->vram->handle);
-	so_method(so, screen->curie, NV40TCL_DMA_COLOR0, 2);
-	so_data  (so, chan->vram->handle);
-	so_data  (so, chan->vram->handle);
-	so_method(so, screen->curie, NV40TCL_DMA_VTXBUF0, 2);
-	so_data  (so, chan->vram->handle);
-	so_data  (so, chan->gart->handle);
-	so_method(so, screen->curie, NV40TCL_DMA_FENCE, 2);
-	so_data  (so, 0);
-	so_data  (so, screen->query->handle);
-	so_method(so, screen->curie, NV40TCL_DMA_UNK01AC, 2);
-	so_data  (so, chan->vram->handle);
-	so_data  (so, chan->vram->handle);
-	so_method(so, screen->curie, NV40TCL_DMA_COLOR2, 2);
-	so_data  (so, chan->vram->handle);
-	so_data  (so, chan->vram->handle);
-
-	so_method(so, screen->curie, 0x1ea4, 3);
-	so_data  (so, 0x00000010);
-	so_data  (so, 0x01000100);
-	so_data  (so, 0xff800006);
-
-	/* vtxprog output routing */
-	so_method(so, screen->curie, 0x1fc4, 1);
-	so_data  (so, 0x06144321);
-	so_method(so, screen->curie, 0x1fc8, 2);
-	so_data  (so, 0xedcba987);
-	so_data  (so, 0x00000021);
-	so_method(so, screen->curie, 0x1fd0, 1);
-	so_data  (so, 0x00171615);
-	so_method(so, screen->curie, 0x1fd4, 1);
-	so_data  (so, 0x001b1a19);
-
-	so_method(so, screen->curie, 0x1ef8, 1);
-	so_data  (so, 0x0020ffff);
-	so_method(so, screen->curie, 0x1d64, 1);
-	so_data  (so, 0x00d30000);
-	so_method(so, screen->curie, 0x1e94, 1);
-	so_data  (so, 0x00000001);
-
-	so_emit(chan, so);
-	so_ref(NULL, &so);
-	nouveau_pushbuf_flush(chan, 0);
-
-	return pscreen;
-}
-
diff --git a/src/gallium/drivers/nv40/nv40_screen.h b/src/gallium/drivers/nv40/nv40_screen.h
deleted file mode 100644
index 2765ab764a..0000000000
--- a/src/gallium/drivers/nv40/nv40_screen.h
+++ /dev/null
@@ -1,37 +0,0 @@
-#ifndef __NV40_SCREEN_H__
-#define __NV40_SCREEN_H__
-
-#include "nouveau/nouveau_screen.h"
-#include "nouveau/nv04_surface_2d.h"
-
-struct nv40_screen {
-	struct nouveau_screen base;
-
-	struct nouveau_winsys *nvws;
-
-	struct nv40_context *cur_ctx;
-
-	/* HW graphics objects */
-	struct nv04_surface_2d *eng2d;
-	struct nouveau_grobj *curie;
-	struct nouveau_notifier *sync;
-
-	/* Query object resources */
-	struct nouveau_notifier *query;
-	struct nouveau_resource *query_heap;
-
-	/* Vtxprog resources */
-	struct nouveau_resource *vp_exec_heap;
-	struct nouveau_resource *vp_data_heap;
-
-	/* Current 3D state of channel */
-	struct nouveau_stateobj *state[NV40_STATE_MAX];
-};
-
-static INLINE struct nv40_screen *
-nv40_screen(struct pipe_screen *screen)
-{
-	return (struct nv40_screen *)screen;
-}
-
-#endif
diff --git a/src/gallium/drivers/nv40/nv40_shader.h b/src/gallium/drivers/nv40/nv40_shader.h
deleted file mode 100644
index 854dccf548..0000000000
--- a/src/gallium/drivers/nv40/nv40_shader.h
+++ /dev/null
@@ -1,556 +0,0 @@
-#ifndef __NV40_SHADER_H__
-#define __NV40_SHADER_H__
-
-/* Vertex programs instruction set
- *
- * The NV40 instruction set is very similar to NV30.  Most fields are in
- * a slightly different position in the instruction however.
- *
- * Merged instructions
- *     In some cases it is possible to put two instructions into one opcode
- *     slot.  The rules for when this is OK is not entirely clear to me yet.
- *
- *     There are separate writemasks and dest temp register fields for each
- *     grouping of instructions.  There is however only one field with the
- *     ID of a result register.  Writing to temp/result regs is selected by
- *     setting VEC_RESULT/SCA_RESULT.
- *
- * Temporary registers
- *     The source/dest temp register fields have been extended by 1 bit, to
- *     give a total of 32 temporary registers.
- *
- * Relative Addressing
- *     NV40 can use an address register to index into vertex attribute regs.
- *     This is done by putting the offset value into INPUT_SRC and setting
- *     the INDEX_INPUT flag.
- *
- * Conditional execution (see NV_vertex_program{2,3} for details)
- *     There is a second condition code register on NV40, it's use is enabled
- *     by setting the COND_REG_SELECT_1 flag.
- *
- * Texture lookup
- *     TODO
- */
-
-/* ---- OPCODE BITS 127:96 / data DWORD 0 --- */
-#define NV40_VP_INST_VEC_RESULT                                        (1 << 30)
-/* uncertain.. */
-#define NV40_VP_INST_COND_UPDATE_ENABLE                        ((1 << 14)|1<<29)
-/* use address reg as index into attribs */
-#define NV40_VP_INST_INDEX_INPUT                                       (1 << 27)
-#define NV40_VP_INST_COND_REG_SELECT_1                                 (1 << 25)
-#define NV40_VP_INST_ADDR_REG_SELECT_1                                 (1 << 24)
-#define NV40_VP_INST_SRC2_ABS                                          (1 << 23)
-#define NV40_VP_INST_SRC1_ABS                                          (1 << 22)
-#define NV40_VP_INST_SRC0_ABS                                          (1 << 21)
-#define NV40_VP_INST_VEC_DEST_TEMP_SHIFT                                      15
-#define NV40_VP_INST_VEC_DEST_TEMP_MASK                             (0x1F << 15)
-#define NV40_VP_INST_COND_TEST_ENABLE                                  (1 << 13)
-#define NV40_VP_INST_COND_SHIFT                                               10
-#define NV40_VP_INST_COND_MASK                                       (0x7 << 10)
-#    define NV40_VP_INST_COND_FL                                               0
-#    define NV40_VP_INST_COND_LT                                               1
-#    define NV40_VP_INST_COND_EQ                                               2
-#    define NV40_VP_INST_COND_LE                                               3
-#    define NV40_VP_INST_COND_GT                                               4
-#    define NV40_VP_INST_COND_NE                                               5
-#    define NV40_VP_INST_COND_GE                                               6
-#    define NV40_VP_INST_COND_TR                                               7
-#define NV40_VP_INST_COND_SWZ_X_SHIFT                                          8
-#define NV40_VP_INST_COND_SWZ_X_MASK                                    (3 << 8)
-#define NV40_VP_INST_COND_SWZ_Y_SHIFT                                          6
-#define NV40_VP_INST_COND_SWZ_Y_MASK                                    (3 << 6)
-#define NV40_VP_INST_COND_SWZ_Z_SHIFT                                          4
-#define NV40_VP_INST_COND_SWZ_Z_MASK                                    (3 << 4)
-#define NV40_VP_INST_COND_SWZ_W_SHIFT                                          2
-#define NV40_VP_INST_COND_SWZ_W_MASK                                    (3 << 2)
-#define NV40_VP_INST_COND_SWZ_ALL_SHIFT                                        2
-#define NV40_VP_INST_COND_SWZ_ALL_MASK                               (0xFF << 2)
-#define NV40_VP_INST_ADDR_SWZ_SHIFT                                            0
-#define NV40_VP_INST_ADDR_SWZ_MASK                                   (0x03 << 0)
-#define NV40_VP_INST0_KNOWN ( \
-                NV40_VP_INST_INDEX_INPUT | \
-                NV40_VP_INST_COND_REG_SELECT_1 | \
-                NV40_VP_INST_ADDR_REG_SELECT_1 | \
-                NV40_VP_INST_SRC2_ABS | \
-                NV40_VP_INST_SRC1_ABS | \
-                NV40_VP_INST_SRC0_ABS | \
-                NV40_VP_INST_VEC_DEST_TEMP_MASK | \
-                NV40_VP_INST_COND_TEST_ENABLE | \
-                NV40_VP_INST_COND_MASK | \
-                NV40_VP_INST_COND_SWZ_ALL_MASK | \
-                NV40_VP_INST_ADDR_SWZ_MASK)
-
-/* ---- OPCODE BITS 95:64 / data DWORD 1 --- */
-#define NV40_VP_INST_VEC_OPCODE_SHIFT                                         22
-#define NV40_VP_INST_VEC_OPCODE_MASK                                (0x1F << 22)
-#    define NV40_VP_INST_OP_NOP                                             0x00
-#    define NV40_VP_INST_OP_MOV                                             0x01
-#    define NV40_VP_INST_OP_MUL                                             0x02
-#    define NV40_VP_INST_OP_ADD                                             0x03
-#    define NV40_VP_INST_OP_MAD                                             0x04
-#    define NV40_VP_INST_OP_DP3                                             0x05
-#    define NV40_VP_INST_OP_DPH                                             0x06
-#    define NV40_VP_INST_OP_DP4                                             0x07
-#    define NV40_VP_INST_OP_DST                                             0x08
-#    define NV40_VP_INST_OP_MIN                                             0x09
-#    define NV40_VP_INST_OP_MAX                                             0x0A
-#    define NV40_VP_INST_OP_SLT                                             0x0B
-#    define NV40_VP_INST_OP_SGE                                             0x0C
-#    define NV40_VP_INST_OP_ARL                                             0x0D
-#    define NV40_VP_INST_OP_FRC                                             0x0E
-#    define NV40_VP_INST_OP_FLR                                             0x0F
-#    define NV40_VP_INST_OP_SEQ                                             0x10
-#    define NV40_VP_INST_OP_SFL                                             0x11
-#    define NV40_VP_INST_OP_SGT                                             0x12
-#    define NV40_VP_INST_OP_SLE                                             0x13
-#    define NV40_VP_INST_OP_SNE                                             0x14
-#    define NV40_VP_INST_OP_STR                                             0x15
-#    define NV40_VP_INST_OP_SSG                                             0x16
-#    define NV40_VP_INST_OP_ARR                                             0x17
-#    define NV40_VP_INST_OP_ARA                                             0x18
-#    define NV40_VP_INST_OP_TXL                                             0x19
-#define NV40_VP_INST_SCA_OPCODE_SHIFT                                         27
-#define NV40_VP_INST_SCA_OPCODE_MASK                                (0x1F << 27)
-#    define NV40_VP_INST_OP_NOP                                             0x00
-#    define NV40_VP_INST_OP_MOV                                             0x01
-#    define NV40_VP_INST_OP_RCP                                             0x02
-#    define NV40_VP_INST_OP_RCC                                             0x03
-#    define NV40_VP_INST_OP_RSQ                                             0x04
-#    define NV40_VP_INST_OP_EXP                                             0x05
-#    define NV40_VP_INST_OP_LOG                                             0x06
-#    define NV40_VP_INST_OP_LIT                                             0x07
-#    define NV40_VP_INST_OP_BRA                                             0x09
-#    define NV40_VP_INST_OP_CAL                                             0x0B
-#    define NV40_VP_INST_OP_RET                                             0x0C
-#    define NV40_VP_INST_OP_LG2                                             0x0D
-#    define NV40_VP_INST_OP_EX2                                             0x0E
-#    define NV40_VP_INST_OP_SIN                                             0x0F
-#    define NV40_VP_INST_OP_COS                                             0x10
-#    define NV40_VP_INST_OP_PUSHA                                           0x13
-#    define NV40_VP_INST_OP_POPA                                            0x14
-#define NV40_VP_INST_CONST_SRC_SHIFT                                          12
-#define NV40_VP_INST_CONST_SRC_MASK                                 (0xFF << 12)
-#define NV40_VP_INST_INPUT_SRC_SHIFT                                           8
-#define NV40_VP_INST_INPUT_SRC_MASK                                  (0x0F << 8)
-#    define NV40_VP_INST_IN_POS                                                0
-#    define NV40_VP_INST_IN_WEIGHT                                             1
-#    define NV40_VP_INST_IN_NORMAL                                             2
-#    define NV40_VP_INST_IN_COL0                                               3
-#    define NV40_VP_INST_IN_COL1                                               4
-#    define NV40_VP_INST_IN_FOGC                                               5
-#    define NV40_VP_INST_IN_TC0                                                8
-#    define NV40_VP_INST_IN_TC(n)                                          (8+n)
-#define NV40_VP_INST_SRC0H_SHIFT                                               0
-#define NV40_VP_INST_SRC0H_MASK                                      (0xFF << 0)
-#define NV40_VP_INST1_KNOWN ( \
-                NV40_VP_INST_VEC_OPCODE_MASK | \
-                NV40_VP_INST_SCA_OPCODE_MASK | \
-                NV40_VP_INST_CONST_SRC_MASK  | \
-                NV40_VP_INST_INPUT_SRC_MASK  | \
-                NV40_VP_INST_SRC0H_MASK \
-                )
-
-/* ---- OPCODE BITS 63:32 / data DWORD 2 --- */
-#define NV40_VP_INST_SRC0L_SHIFT                                              23
-#define NV40_VP_INST_SRC0L_MASK                                    (0x1FF << 23)
-#define NV40_VP_INST_SRC1_SHIFT                                                6
-#define NV40_VP_INST_SRC1_MASK                                    (0x1FFFF << 6)
-#define NV40_VP_INST_SRC2H_SHIFT                                               0
-#define NV40_VP_INST_SRC2H_MASK                                      (0x3F << 0)
-#define NV40_VP_INST_IADDRH_SHIFT                                              0
-#define NV40_VP_INST_IADDRH_MASK                                     (0x1F << 0)
-
-/* ---- OPCODE BITS 31:0 / data DWORD 3 --- */
-#define NV40_VP_INST_IADDRL_SHIFT                                             29
-#define NV40_VP_INST_IADDRL_MASK                                       (7 << 29)
-#define NV40_VP_INST_SRC2L_SHIFT                                              21
-#define NV40_VP_INST_SRC2L_MASK                                    (0x7FF << 21)
-#define NV40_VP_INST_SCA_WRITEMASK_SHIFT                                      17
-#define NV40_VP_INST_SCA_WRITEMASK_MASK                              (0xF << 17)
-#    define NV40_VP_INST_SCA_WRITEMASK_X                               (1 << 20)
-#    define NV40_VP_INST_SCA_WRITEMASK_Y                               (1 << 19)
-#    define NV40_VP_INST_SCA_WRITEMASK_Z                               (1 << 18)
-#    define NV40_VP_INST_SCA_WRITEMASK_W                               (1 << 17)
-#define NV40_VP_INST_VEC_WRITEMASK_SHIFT                                      13
-#define NV40_VP_INST_VEC_WRITEMASK_MASK                              (0xF << 13)
-#    define NV40_VP_INST_VEC_WRITEMASK_X                               (1 << 16)
-#    define NV40_VP_INST_VEC_WRITEMASK_Y                               (1 << 15)
-#    define NV40_VP_INST_VEC_WRITEMASK_Z                               (1 << 14)
-#    define NV40_VP_INST_VEC_WRITEMASK_W                               (1 << 13)
-#define NV40_VP_INST_SCA_RESULT                                        (1 << 12)
-#define NV40_VP_INST_SCA_DEST_TEMP_SHIFT                                       7
-#define NV40_VP_INST_SCA_DEST_TEMP_MASK                              (0x1F << 7)
-#define NV40_VP_INST_DEST_SHIFT                                                2
-#define NV40_VP_INST_DEST_MASK                                         (31 << 2)
-#    define NV40_VP_INST_DEST_POS                                              0
-#    define NV40_VP_INST_DEST_COL0                                             1
-#    define NV40_VP_INST_DEST_COL1                                             2
-#    define NV40_VP_INST_DEST_BFC0                                             3
-#    define NV40_VP_INST_DEST_BFC1                                             4
-#    define NV40_VP_INST_DEST_FOGC                                             5
-#    define NV40_VP_INST_DEST_PSZ                                              6
-#    define NV40_VP_INST_DEST_TC0                                              7
-#    define NV40_VP_INST_DEST_TC(n)                                        (7+n)
-#    define NV40_VP_INST_DEST_TEMP                                          0x1F
-#define NV40_VP_INST_INDEX_CONST                                        (1 << 1)
-#define NV40_VP_INST_LAST                                               (1 << 0)
-#define NV40_VP_INST3_KNOWN ( \
-                NV40_VP_INST_SRC2L_MASK |\
-                NV40_VP_INST_SCA_WRITEMASK_MASK |\
-                NV40_VP_INST_VEC_WRITEMASK_MASK |\
-                NV40_VP_INST_SCA_DEST_TEMP_MASK |\
-                NV40_VP_INST_DEST_MASK |\
-                NV40_VP_INST_INDEX_CONST)
-
-/* Useful to split the source selection regs into their pieces */
-#define NV40_VP_SRC0_HIGH_SHIFT                                                9
-#define NV40_VP_SRC0_HIGH_MASK                                        0x0001FE00
-#define NV40_VP_SRC0_LOW_MASK                                         0x000001FF
-#define NV40_VP_SRC2_HIGH_SHIFT                                               11
-#define NV40_VP_SRC2_HIGH_MASK                                        0x0001F800
-#define NV40_VP_SRC2_LOW_MASK                                         0x000007FF
-
-/* Source selection - these are the bits you fill NV40_VP_INST_SRCn with */
-#define NV40_VP_SRC_NEGATE                                             (1 << 16)
-#define NV40_VP_SRC_SWZ_X_SHIFT                                               14
-#define NV40_VP_SRC_SWZ_X_MASK                                         (3 << 14)
-#define NV40_VP_SRC_SWZ_Y_SHIFT                                               12
-#define NV40_VP_SRC_SWZ_Y_MASK                                         (3 << 12)
-#define NV40_VP_SRC_SWZ_Z_SHIFT                                               10
-#define NV40_VP_SRC_SWZ_Z_MASK                                         (3 << 10)
-#define NV40_VP_SRC_SWZ_W_SHIFT                                                8
-#define NV40_VP_SRC_SWZ_W_MASK                                          (3 << 8)
-#define NV40_VP_SRC_SWZ_ALL_SHIFT                                              8
-#define NV40_VP_SRC_SWZ_ALL_MASK                                     (0xFF << 8)
-#define NV40_VP_SRC_TEMP_SRC_SHIFT                                             2
-#define NV40_VP_SRC_TEMP_SRC_MASK                                    (0x1F << 2)
-#define NV40_VP_SRC_REG_TYPE_SHIFT                                             0
-#define NV40_VP_SRC_REG_TYPE_MASK                                       (3 << 0)
-#    define NV40_VP_SRC_REG_TYPE_UNK0                                          0
-#    define NV40_VP_SRC_REG_TYPE_TEMP                                          1
-#    define NV40_VP_SRC_REG_TYPE_INPUT                                         2
-#    define NV40_VP_SRC_REG_TYPE_CONST                                         3
-
-
-/*
- * Each fragment program opcode appears to be comprised of 4 32-bit values.
- *
- *         0 - Opcode, output reg/mask, ATTRIB source
- *         1 - Source 0
- *         2 - Source 1
- *         3 - Source 2
- *
- * There appears to be no special difference between result regs and temp regs.
- *                 result.color == R0.xyzw
- *                 result.depth == R1.z
- * When the fragprog contains instructions to write depth,
- * NV30_TCL_PRIMITIVE_3D_UNK1D78=0 otherwise it is set to 1.
- *
- * Constants are inserted directly after the instruction that uses them.
- * 
- * It appears that it's not possible to use two input registers in one
- * instruction as the input sourcing is done in the instruction dword
- * and not the source selection dwords.  As such instructions such as:
- * 
- *                 ADD result.color, fragment.color, fragment.texcoord[0];
- *
- * must be split into two MOV's and then an ADD (nvidia does this) but
- * I'm not sure why it's not just one MOV and then source the second input
- * in the ADD instruction..
- *
- * Negation of the full source is done with NV30_FP_REG_NEGATE, arbitrary
- * negation requires multiplication with a const.
- *
- * Arbitrary swizzling is supported with the exception of SWIZZLE_ZERO and
- * SWIZZLE_ONE.
- *
- * The temp/result regs appear to be initialised to (0.0, 0.0, 0.0, 0.0) as
- * SWIZZLE_ZERO is implemented simply by not writing to the relevant components
- * of the destination.
- *
- * Looping
- *   Loops appear to be fairly expensive on NV40 at least, the proprietary
- *   driver goes to a lot of effort to avoid using the native looping
- *   instructions.  If the total number of *executed* instructions between
- *   REP/ENDREP or LOOP/ENDLOOP is <=500, the driver will unroll the loop.
- *   The maximum loop count is 255.
- *
- * Conditional execution
- *   TODO
- * 
- * Non-native instructions:
- *         LIT
- *         LRP - MAD+MAD
- *         SUB - ADD, negate second source
- *         RSQ - LG2 + EX2
- *         POW - LG2 + MUL + EX2
- *         SCS - COS + SIN
- *         XPD
- *         DP2 - MUL + ADD
- *         NRM
- */
-
-//== Opcode / Destination selection ==
-#define NV40_FP_OP_PROGRAM_END                                          (1 << 0)
-#define NV40_FP_OP_OUT_REG_SHIFT                                               1
-#define NV40_FP_OP_OUT_REG_MASK                                        (63 << 1)
-/* Needs to be set when writing outputs to get expected result.. */
-#define NV40_FP_OP_OUT_REG_HALF                                         (1 << 7)
-#define NV40_FP_OP_COND_WRITE_ENABLE                                    (1 << 8)
-#define NV40_FP_OP_OUTMASK_SHIFT                                               9
-#define NV40_FP_OP_OUTMASK_MASK                                       (0xF << 9)
-#    define NV40_FP_OP_OUT_X                                            (1 << 9)
-#    define NV40_FP_OP_OUT_Y                                            (1 <<10)
-#    define NV40_FP_OP_OUT_Z                                            (1 <<11)
-#    define NV40_FP_OP_OUT_W                                            (1 <<12)
-/* Uncertain about these, especially the input_src values.. it's possible that
- * they can be dynamically changed.
- */
-#define NV40_FP_OP_INPUT_SRC_SHIFT                                            13
-#define NV40_FP_OP_INPUT_SRC_MASK                                     (15 << 13)
-#    define NV40_FP_OP_INPUT_SRC_POSITION                                    0x0
-#    define NV40_FP_OP_INPUT_SRC_COL0                                        0x1
-#    define NV40_FP_OP_INPUT_SRC_COL1                                        0x2
-#    define NV40_FP_OP_INPUT_SRC_FOGC                                        0x3
-#    define NV40_FP_OP_INPUT_SRC_TC0                                         0x4
-#    define NV40_FP_OP_INPUT_SRC_TC(n)                                 (0x4 + n)
-#    define NV40_FP_OP_INPUT_SRC_FACING                                      0xE
-#define NV40_FP_OP_TEX_UNIT_SHIFT                                             17
-#define NV40_FP_OP_TEX_UNIT_MASK                                     (0xF << 17)
-#define NV40_FP_OP_PRECISION_SHIFT                                            22
-#define NV40_FP_OP_PRECISION_MASK                                      (3 << 22)
-#   define NV40_FP_PRECISION_FP32                                              0
-#   define NV40_FP_PRECISION_FP16                                              1
-#   define NV40_FP_PRECISION_FX12                                              2
-#define NV40_FP_OP_OPCODE_SHIFT                                               24
-#define NV40_FP_OP_OPCODE_MASK                                      (0x3F << 24)
-#        define NV40_FP_OP_OPCODE_NOP                                       0x00
-#        define NV40_FP_OP_OPCODE_MOV                                       0x01
-#        define NV40_FP_OP_OPCODE_MUL                                       0x02
-#        define NV40_FP_OP_OPCODE_ADD                                       0x03
-#        define NV40_FP_OP_OPCODE_MAD                                       0x04
-#        define NV40_FP_OP_OPCODE_DP3                                       0x05
-#        define NV40_FP_OP_OPCODE_DP4                                       0x06
-#        define NV40_FP_OP_OPCODE_DST                                       0x07
-#        define NV40_FP_OP_OPCODE_MIN                                       0x08
-#        define NV40_FP_OP_OPCODE_MAX                                       0x09
-#        define NV40_FP_OP_OPCODE_SLT                                       0x0A
-#        define NV40_FP_OP_OPCODE_SGE                                       0x0B
-#        define NV40_FP_OP_OPCODE_SLE                                       0x0C
-#        define NV40_FP_OP_OPCODE_SGT                                       0x0D
-#        define NV40_FP_OP_OPCODE_SNE                                       0x0E
-#        define NV40_FP_OP_OPCODE_SEQ                                       0x0F
-#        define NV40_FP_OP_OPCODE_FRC                                       0x10
-#        define NV40_FP_OP_OPCODE_FLR                                       0x11
-#        define NV40_FP_OP_OPCODE_KIL                                       0x12
-#        define NV40_FP_OP_OPCODE_PK4B                                      0x13
-#        define NV40_FP_OP_OPCODE_UP4B                                      0x14
-/* DDX/DDY can only write to XY */
-#        define NV40_FP_OP_OPCODE_DDX                                       0x15
-#        define NV40_FP_OP_OPCODE_DDY                                       0x16
-#        define NV40_FP_OP_OPCODE_TEX                                       0x17
-#        define NV40_FP_OP_OPCODE_TXP                                       0x18
-#        define NV40_FP_OP_OPCODE_TXD                                       0x19
-#        define NV40_FP_OP_OPCODE_RCP                                       0x1A
-#        define NV40_FP_OP_OPCODE_EX2                                       0x1C
-#        define NV40_FP_OP_OPCODE_LG2                                       0x1D
-#        define NV40_FP_OP_OPCODE_STR                                       0x20
-#        define NV40_FP_OP_OPCODE_SFL                                       0x21
-#        define NV40_FP_OP_OPCODE_COS                                       0x22
-#        define NV40_FP_OP_OPCODE_SIN                                       0x23
-#        define NV40_FP_OP_OPCODE_PK2H                                      0x24
-#        define NV40_FP_OP_OPCODE_UP2H                                      0x25
-#        define NV40_FP_OP_OPCODE_PK4UB                                     0x27
-#        define NV40_FP_OP_OPCODE_UP4UB                                     0x28
-#        define NV40_FP_OP_OPCODE_PK2US                                     0x29
-#        define NV40_FP_OP_OPCODE_UP2US                                     0x2A
-#        define NV40_FP_OP_OPCODE_DP2A                                      0x2E
-#        define NV40_FP_OP_OPCODE_TXL                                       0x2F
-#        define NV40_FP_OP_OPCODE_TXB                                       0x31
-#        define NV40_FP_OP_OPCODE_DIV                                       0x3A
-#        define NV40_FP_OP_OPCODE_UNK_LIT                                   0x3C
-/* The use of these instructions appears to be indicated by bit 31 of DWORD 2.*/
-#        define NV40_FP_OP_BRA_OPCODE_BRK                                    0x0
-#        define NV40_FP_OP_BRA_OPCODE_CAL                                    0x1
-#        define NV40_FP_OP_BRA_OPCODE_IF                                     0x2
-#        define NV40_FP_OP_BRA_OPCODE_LOOP                                   0x3
-#        define NV40_FP_OP_BRA_OPCODE_REP                                    0x4
-#        define NV40_FP_OP_BRA_OPCODE_RET                                    0x5
-#define NV40_FP_OP_OUT_SAT                                             (1 << 31)
-
-/* high order bits of SRC0 */
-#define NV40_FP_OP_OUT_ABS                                             (1 << 29)
-#define NV40_FP_OP_COND_SWZ_W_SHIFT                                           27
-#define NV40_FP_OP_COND_SWZ_W_MASK                                     (3 << 27)
-#define NV40_FP_OP_COND_SWZ_Z_SHIFT                                           25
-#define NV40_FP_OP_COND_SWZ_Z_MASK                                     (3 << 25)
-#define NV40_FP_OP_COND_SWZ_Y_SHIFT                                           23
-#define NV40_FP_OP_COND_SWZ_Y_MASK                                     (3 << 23)
-#define NV40_FP_OP_COND_SWZ_X_SHIFT                                           21
-#define NV40_FP_OP_COND_SWZ_X_MASK                                     (3 << 21)
-#define NV40_FP_OP_COND_SWZ_ALL_SHIFT                                         21
-#define NV40_FP_OP_COND_SWZ_ALL_MASK                                (0xFF << 21)
-#define NV40_FP_OP_COND_SHIFT                                                 18
-#define NV40_FP_OP_COND_MASK                                        (0x07 << 18)
-#        define NV40_FP_OP_COND_FL                                             0
-#        define NV40_FP_OP_COND_LT                                             1
-#        define NV40_FP_OP_COND_EQ                                             2
-#        define NV40_FP_OP_COND_LE                                             3
-#        define NV40_FP_OP_COND_GT                                             4
-#        define NV40_FP_OP_COND_NE                                             5
-#        define NV40_FP_OP_COND_GE                                             6
-#        define NV40_FP_OP_COND_TR                                             7
-
-/* high order bits of SRC1 */
-#define NV40_FP_OP_OPCODE_IS_BRANCH                                      (1<<31)
-#define NV40_FP_OP_DST_SCALE_SHIFT                                            28
-#define NV40_FP_OP_DST_SCALE_MASK                                      (3 << 28)
-#define NV40_FP_OP_DST_SCALE_1X                                                0
-#define NV40_FP_OP_DST_SCALE_2X                                                1
-#define NV40_FP_OP_DST_SCALE_4X                                                2
-#define NV40_FP_OP_DST_SCALE_8X                                                3
-#define NV40_FP_OP_DST_SCALE_INV_2X                                            5
-#define NV40_FP_OP_DST_SCALE_INV_4X                                            6
-#define NV40_FP_OP_DST_SCALE_INV_8X                                            7
-
-/* SRC1 LOOP */
-#define NV40_FP_OP_LOOP_INCR_SHIFT                                            19
-#define NV40_FP_OP_LOOP_INCR_MASK                                   (0xFF << 19)
-#define NV40_FP_OP_LOOP_INDEX_SHIFT                                           10
-#define NV40_FP_OP_LOOP_INDEX_MASK                                  (0xFF << 10)
-#define NV40_FP_OP_LOOP_COUNT_SHIFT                                            2
-#define NV40_FP_OP_LOOP_COUNT_MASK                                   (0xFF << 2)
-
-/* SRC1 IF */
-#define NV40_FP_OP_ELSE_ID_SHIFT                                               2
-#define NV40_FP_OP_ELSE_ID_MASK                                      (0xFF << 2)
-
-/* SRC1 CAL */
-#define NV40_FP_OP_IADDR_SHIFT                                                 2
-#define NV40_FP_OP_IADDR_MASK                                        (0xFF << 2)
-
-/* SRC1 REP
- *   I have no idea why there are 3 count values here..  but they
- *   have always been filled with the same value in my tests so
- *   far..
- */
-#define NV40_FP_OP_REP_COUNT1_SHIFT                                            2
-#define NV40_FP_OP_REP_COUNT1_MASK                                   (0xFF << 2)
-#define NV40_FP_OP_REP_COUNT2_SHIFT                                           10
-#define NV40_FP_OP_REP_COUNT2_MASK                                  (0xFF << 10)
-#define NV40_FP_OP_REP_COUNT3_SHIFT                                           19
-#define NV40_FP_OP_REP_COUNT3_MASK                                  (0xFF << 19)
-
-/* SRC2 REP/IF */
-#define NV40_FP_OP_END_ID_SHIFT                                                2
-#define NV40_FP_OP_END_ID_MASK                                       (0xFF << 2)
-
-// SRC2 high-order
-#define NV40_FP_OP_INDEX_INPUT                                         (1 << 30)
-#define NV40_FP_OP_ADDR_INDEX_SHIFT                                           19
-#define NV40_FP_OP_ADDR_INDEX_MASK                                   (0xF << 19)
-
-//== Register selection ==
-#define NV40_FP_REG_TYPE_SHIFT                                                 0
-#define NV40_FP_REG_TYPE_MASK                                           (3 << 0)
-#        define NV40_FP_REG_TYPE_TEMP                                          0
-#        define NV40_FP_REG_TYPE_INPUT                                         1
-#        define NV40_FP_REG_TYPE_CONST                                         2
-#define NV40_FP_REG_SRC_SHIFT                                                  2
-#define NV40_FP_REG_SRC_MASK                                           (63 << 2)
-#define NV40_FP_REG_SRC_HALF                                            (1 << 8)
-#define NV40_FP_REG_SWZ_ALL_SHIFT                                              9
-#define NV40_FP_REG_SWZ_ALL_MASK                                      (255 << 9)
-#define NV40_FP_REG_SWZ_X_SHIFT                                                9
-#define NV40_FP_REG_SWZ_X_MASK                                          (3 << 9)
-#define NV40_FP_REG_SWZ_Y_SHIFT                                               11
-#define NV40_FP_REG_SWZ_Y_MASK                                         (3 << 11)
-#define NV40_FP_REG_SWZ_Z_SHIFT                                               13
-#define NV40_FP_REG_SWZ_Z_MASK                                         (3 << 13)
-#define NV40_FP_REG_SWZ_W_SHIFT                                               15
-#define NV40_FP_REG_SWZ_W_MASK                                         (3 << 15)
-#        define NV40_FP_SWIZZLE_X                                              0
-#        define NV40_FP_SWIZZLE_Y                                              1
-#        define NV40_FP_SWIZZLE_Z                                              2
-#        define NV40_FP_SWIZZLE_W                                              3
-#define NV40_FP_REG_NEGATE                                             (1 << 17)
-
-#ifndef NV40_SHADER_NO_FUCKEDNESS
-#define NV40SR_NONE	0
-#define NV40SR_OUTPUT	1
-#define NV40SR_INPUT	2
-#define NV40SR_TEMP	3
-#define NV40SR_CONST	4
-
-struct nv40_sreg {
-	int type;
-	int index;
-
-	int dst_scale;
-
-	int negate;
-	int abs;
-	int swz[4];
-
-	int cc_update;
-	int cc_update_reg;
-	int cc_test;
-	int cc_test_reg;
-	int cc_swz[4];
-};
-
-static INLINE struct nv40_sreg
-nv40_sr(int type, int index)
-{
-	struct nv40_sreg temp = {
-		.type = type,
-		.index = index,
-		.dst_scale = DEF_SCALE,
-		.abs = 0,
-		.negate = 0,
-		.swz = { 0, 1, 2, 3 },
-		.cc_update = 0,
-		.cc_update_reg = 0,
-		.cc_test = DEF_CTEST,
-		.cc_test_reg = 0,
-		.cc_swz = { 0, 1, 2, 3 },
-	};
-	return temp;
-}
-
-static INLINE struct nv40_sreg
-nv40_sr_swz(struct nv40_sreg src, int x, int y, int z, int w)
-{
-	struct nv40_sreg dst = src;
-
-	dst.swz[SWZ_X] = src.swz[x];
-	dst.swz[SWZ_Y] = src.swz[y];
-	dst.swz[SWZ_Z] = src.swz[z];
-	dst.swz[SWZ_W] = src.swz[w];
-	return dst;
-}
-
-static INLINE struct nv40_sreg
-nv40_sr_neg(struct nv40_sreg src)
-{
-	src.negate = !src.negate;
-	return src;
-}
-
-static INLINE struct nv40_sreg
-nv40_sr_abs(struct nv40_sreg src)
-{
-	src.abs = 1;
-	return src;
-}
-
-static INLINE struct nv40_sreg
-nv40_sr_scale(struct nv40_sreg src, int scale)
-{
-	src.dst_scale = scale;
-	return src;
-}
-#endif
-
-#endif
diff --git a/src/gallium/drivers/nv40/nv40_state.c b/src/gallium/drivers/nv40/nv40_state.c
deleted file mode 100644
index 28a48a63f1..0000000000
--- a/src/gallium/drivers/nv40/nv40_state.c
+++ /dev/null
@@ -1,797 +0,0 @@
-#include "pipe/p_state.h"
-#include "pipe/p_defines.h"
-#include "util/u_inlines.h"
-
-#include "draw/draw_context.h"
-
-#include "tgsi/tgsi_parse.h"
-
-#include "nv40_context.h"
-#include "nv40_state.h"
-
-static void *
-nv40_blend_state_create(struct pipe_context *pipe,
-			const struct pipe_blend_state *cso)
-{
-	struct nv40_context *nv40 = nv40_context(pipe);
-	struct nouveau_grobj *curie = nv40->screen->curie;
-	struct nv40_blend_state *bso = CALLOC(1, sizeof(*bso));
-	struct nouveau_stateobj *so = so_new(5, 8, 0);
-
-	if (cso->rt[0].blend_enable) {
-		so_method(so, curie, NV40TCL_BLEND_ENABLE, 3);
-		so_data  (so, 1);
-		so_data  (so, (nvgl_blend_func(cso->rt[0].alpha_src_factor) << 16) |
-			       nvgl_blend_func(cso->rt[0].rgb_src_factor));
-		so_data  (so, nvgl_blend_func(cso->rt[0].alpha_dst_factor) << 16 |
-			      nvgl_blend_func(cso->rt[0].rgb_dst_factor));
-		so_method(so, curie, NV40TCL_BLEND_EQUATION, 1);
-		so_data  (so, nvgl_blend_eqn(cso->rt[0].alpha_func) << 16 |
-			      nvgl_blend_eqn(cso->rt[0].rgb_func));
-	} else {
-		so_method(so, curie, NV40TCL_BLEND_ENABLE, 1);
-		so_data  (so, 0);
-	}
-
-	so_method(so, curie, NV40TCL_COLOR_MASK, 1);
-	so_data  (so, (((cso->rt[0].colormask & PIPE_MASK_A) ? (0x01 << 24) : 0) |
-		       ((cso->rt[0].colormask & PIPE_MASK_R) ? (0x01 << 16) : 0) |
-		       ((cso->rt[0].colormask & PIPE_MASK_G) ? (0x01 <<  8) : 0) |
-		       ((cso->rt[0].colormask & PIPE_MASK_B) ? (0x01 <<  0) : 0)));
-
-	if (cso->logicop_enable) {
-		so_method(so, curie, NV40TCL_COLOR_LOGIC_OP_ENABLE, 2);
-		so_data  (so, 1);
-		so_data  (so, nvgl_logicop_func(cso->logicop_func));
-	} else {
-		so_method(so, curie, NV40TCL_COLOR_LOGIC_OP_ENABLE, 1);
-		so_data  (so, 0);
-	}
-
-	so_method(so, curie, NV40TCL_DITHER_ENABLE, 1);
-	so_data  (so, cso->dither ? 1 : 0);
-
-	so_ref(so, &bso->so);
-	so_ref(NULL, &so);
-	bso->pipe = *cso;
-	return (void *)bso;
-}
-
-static void
-nv40_blend_state_bind(struct pipe_context *pipe, void *hwcso)
-{
-	struct nv40_context *nv40 = nv40_context(pipe);
-
-	nv40->blend = hwcso;
-	nv40->dirty |= NV40_NEW_BLEND;
-}
-
-static void
-nv40_blend_state_delete(struct pipe_context *pipe, void *hwcso)
-{
-	struct nv40_blend_state *bso = hwcso;
-
-	so_ref(NULL, &bso->so);
-	FREE(bso);
-}
-
-
-static INLINE unsigned
-wrap_mode(unsigned wrap) {
-	unsigned ret;
-
-	switch (wrap) {
-	case PIPE_TEX_WRAP_REPEAT:
-		ret = NV40TCL_TEX_WRAP_S_REPEAT;
-		break;
-	case PIPE_TEX_WRAP_MIRROR_REPEAT:
-		ret = NV40TCL_TEX_WRAP_S_MIRRORED_REPEAT;
-		break;
-	case PIPE_TEX_WRAP_CLAMP_TO_EDGE:
-		ret = NV40TCL_TEX_WRAP_S_CLAMP_TO_EDGE;
-		break;
-	case PIPE_TEX_WRAP_CLAMP_TO_BORDER:
-		ret = NV40TCL_TEX_WRAP_S_CLAMP_TO_BORDER;
-		break;
-	case PIPE_TEX_WRAP_CLAMP:
-		ret = NV40TCL_TEX_WRAP_S_CLAMP;
-		break;
-	case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE:
-		ret = NV40TCL_TEX_WRAP_S_MIRROR_CLAMP_TO_EDGE;
-		break;
-	case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER:
-		ret = NV40TCL_TEX_WRAP_S_MIRROR_CLAMP_TO_BORDER;
-		break;
-	case PIPE_TEX_WRAP_MIRROR_CLAMP:
-		ret = NV40TCL_TEX_WRAP_S_MIRROR_CLAMP;
-		break;
-	default:
-		NOUVEAU_ERR("unknown wrap mode: %d\n", wrap);
-		ret = NV40TCL_TEX_WRAP_S_REPEAT;
-		break;
-	}
-
-	return ret >> NV40TCL_TEX_WRAP_S_SHIFT;
-}
-
-static void *
-nv40_sampler_state_create(struct pipe_context *pipe,
-			  const struct pipe_sampler_state *cso)
-{
-	struct nv40_sampler_state *ps;
-	uint32_t filter = 0;
-
-	ps = MALLOC(sizeof(struct nv40_sampler_state));
-
-	ps->fmt = 0;
-	if (!cso->normalized_coords)
-		ps->fmt |= NV40TCL_TEX_FORMAT_RECT;
-
-	ps->wrap = ((wrap_mode(cso->wrap_s) << NV40TCL_TEX_WRAP_S_SHIFT) |
-		    (wrap_mode(cso->wrap_t) << NV40TCL_TEX_WRAP_T_SHIFT) |
-		    (wrap_mode(cso->wrap_r) << NV40TCL_TEX_WRAP_R_SHIFT));
-
-	ps->en = 0;
-	if (cso->max_anisotropy >= 2) {
-		/* no idea, binary driver sets it, works without it.. meh.. */
-		ps->wrap |= (1 << 5);
-
-		if (cso->max_anisotropy >= 16) {
-			ps->en |= NV40TCL_TEX_ENABLE_ANISO_16X;
-		} else
-		if (cso->max_anisotropy >= 12) {
-			ps->en |= NV40TCL_TEX_ENABLE_ANISO_12X;
-		} else
-		if (cso->max_anisotropy >= 10) {
-			ps->en |= NV40TCL_TEX_ENABLE_ANISO_10X;
-		} else
-		if (cso->max_anisotropy >= 8) {
-			ps->en |= NV40TCL_TEX_ENABLE_ANISO_8X;
-		} else
-		if (cso->max_anisotropy >= 6) {
-			ps->en |= NV40TCL_TEX_ENABLE_ANISO_6X;
-		} else
-		if (cso->max_anisotropy >= 4) {
-			ps->en |= NV40TCL_TEX_ENABLE_ANISO_4X;
-		} else {
-			ps->en |= NV40TCL_TEX_ENABLE_ANISO_2X;
-		}
-	}
-
-	switch (cso->mag_img_filter) {
-	case PIPE_TEX_FILTER_LINEAR:
-		filter |= NV40TCL_TEX_FILTER_MAG_LINEAR;
-		break;
-	case PIPE_TEX_FILTER_NEAREST:
-	default:
-		filter |= NV40TCL_TEX_FILTER_MAG_NEAREST;
-		break;
-	}
-
-	switch (cso->min_img_filter) {
-	case PIPE_TEX_FILTER_LINEAR:
-		switch (cso->min_mip_filter) {
-		case PIPE_TEX_MIPFILTER_NEAREST:
-			filter |= NV40TCL_TEX_FILTER_MIN_LINEAR_MIPMAP_NEAREST;
-			break;
-		case PIPE_TEX_MIPFILTER_LINEAR:
-			filter |= NV40TCL_TEX_FILTER_MIN_LINEAR_MIPMAP_LINEAR;
-			break;
-		case PIPE_TEX_MIPFILTER_NONE:
-		default:
-			filter |= NV40TCL_TEX_FILTER_MIN_LINEAR;
-			break;
-		}
-		break;
-	case PIPE_TEX_FILTER_NEAREST:
-	default:
-		switch (cso->min_mip_filter) {
-		case PIPE_TEX_MIPFILTER_NEAREST:
-			filter |= NV40TCL_TEX_FILTER_MIN_NEAREST_MIPMAP_NEAREST;
-		break;
-		case PIPE_TEX_MIPFILTER_LINEAR:
-			filter |= NV40TCL_TEX_FILTER_MIN_NEAREST_MIPMAP_LINEAR;
-			break;
-		case PIPE_TEX_MIPFILTER_NONE:
-		default:
-			filter |= NV40TCL_TEX_FILTER_MIN_NEAREST;
-			break;
-		}
-		break;
-	}
-
-	ps->filt = filter;
-
-	{
-		float limit;
-
-		limit = CLAMP(cso->lod_bias, -16.0, 15.0);
-		ps->filt |= (int)(cso->lod_bias * 256.0) & 0x1fff;
-
-		limit = CLAMP(cso->max_lod, 0.0, 15.0);
-		ps->en |= (int)(limit * 256.0) << 7;
-
-		limit = CLAMP(cso->min_lod, 0.0, 15.0);
-		ps->en |= (int)(limit * 256.0) << 19;
-	}
-
-
-	if (cso->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) {
-		switch (cso->compare_func) {
-		case PIPE_FUNC_NEVER:
-			ps->wrap |= NV40TCL_TEX_WRAP_RCOMP_NEVER;
-			break;
-		case PIPE_FUNC_GREATER:
-			ps->wrap |= NV40TCL_TEX_WRAP_RCOMP_GREATER;
-			break;
-		case PIPE_FUNC_EQUAL:
-			ps->wrap |= NV40TCL_TEX_WRAP_RCOMP_EQUAL;
-			break;
-		case PIPE_FUNC_GEQUAL:
-			ps->wrap |= NV40TCL_TEX_WRAP_RCOMP_GEQUAL;
-			break;
-		case PIPE_FUNC_LESS:
-			ps->wrap |= NV40TCL_TEX_WRAP_RCOMP_LESS;
-			break;
-		case PIPE_FUNC_NOTEQUAL:
-			ps->wrap |= NV40TCL_TEX_WRAP_RCOMP_NOTEQUAL;
-			break;
-		case PIPE_FUNC_LEQUAL:
-			ps->wrap |= NV40TCL_TEX_WRAP_RCOMP_LEQUAL;
-			break;
-		case PIPE_FUNC_ALWAYS:
-			ps->wrap |= NV40TCL_TEX_WRAP_RCOMP_ALWAYS;
-			break;
-		default:
-			break;
-		}
-	}
-
-	ps->bcol = ((float_to_ubyte(cso->border_color[3]) << 24) |
-		    (float_to_ubyte(cso->border_color[0]) << 16) |
-		    (float_to_ubyte(cso->border_color[1]) <<  8) |
-		    (float_to_ubyte(cso->border_color[2]) <<  0));
-
-	return (void *)ps;
-}
-
-static void
-nv40_sampler_state_bind(struct pipe_context *pipe, unsigned nr, void **sampler)
-{
-	struct nv40_context *nv40 = nv40_context(pipe);
-	unsigned unit;
-
-	for (unit = 0; unit < nr; unit++) {
-		nv40->tex_sampler[unit] = sampler[unit];
-		nv40->dirty_samplers |= (1 << unit);
-	}
-
-	for (unit = nr; unit < nv40->nr_samplers; unit++) {
-		nv40->tex_sampler[unit] = NULL;
-		nv40->dirty_samplers |= (1 << unit);
-	}
-
-	nv40->nr_samplers = nr;
-	nv40->dirty |= NV40_NEW_SAMPLER;
-}
-
-static void
-nv40_sampler_state_delete(struct pipe_context *pipe, void *hwcso)
-{
-	FREE(hwcso);
-}
-
-static void
-nv40_set_fragment_sampler_views(struct pipe_context *pipe,
-				unsigned nr,
-				struct pipe_sampler_view **views)
-{
-	struct nv40_context *nv40 = nv40_context(pipe);
-	unsigned unit;
-
-	for (unit = 0; unit < nr; unit++) {
-		pipe_sampler_view_reference(&nv40->fragment_sampler_views[unit], views[unit]);
-		pipe_texture_reference((struct pipe_texture **)
-				       &nv40->tex_miptree[unit], views[unit]->texture);
-		nv40->dirty_samplers |= (1 << unit);
-	}
-
-	for (unit = nr; unit < nv40->nr_textures; unit++) {
-		pipe_sampler_view_reference(&nv40->fragment_sampler_views[unit], NULL);
-		pipe_texture_reference((struct pipe_texture **)
-				       &nv40->tex_miptree[unit], NULL);
-		nv40->dirty_samplers |= (1 << unit);
-	}
-
-	nv40->nr_textures = nr;
-	nv40->dirty |= NV40_NEW_SAMPLER;
-}
-
-static struct pipe_sampler_view *
-nv40_create_sampler_view(struct pipe_context *pipe,
-			 struct pipe_texture *texture,
-			 const struct pipe_sampler_view *templ)
-{
-	struct pipe_sampler_view *view = CALLOC_STRUCT(pipe_sampler_view);
-
-	if (view) {
-		*view = *templ;
-		view->reference.count = 1;
-		view->texture = NULL;
-		pipe_texture_reference(&view->texture, texture);
-		view->context = pipe;
-	}
-
-	return view;
-}
-
-
-static void
-nv40_sampler_view_destroy(struct pipe_context *pipe,
-			  struct pipe_sampler_view *view)
-{
-	pipe_texture_reference(&view->texture, NULL);
-	FREE(view);
-}
-
-static void *
-nv40_rasterizer_state_create(struct pipe_context *pipe,
-			     const struct pipe_rasterizer_state *cso)
-{
-	struct nv40_context *nv40 = nv40_context(pipe);
-	struct nv40_rasterizer_state *rsso = CALLOC(1, sizeof(*rsso));
-	struct nouveau_stateobj *so = so_new(9, 19, 0);
-	struct nouveau_grobj *curie = nv40->screen->curie;
-
-	/*XXX: ignored:
-	 * 	light_twoside
-	 * 	point_smooth -nohw
-	 * 	multisample
-	 */
-
-	so_method(so, curie, NV40TCL_SHADE_MODEL, 1);
-	so_data  (so, cso->flatshade ? NV40TCL_SHADE_MODEL_FLAT :
-				       NV40TCL_SHADE_MODEL_SMOOTH);
-
-	so_method(so, curie, NV40TCL_LINE_WIDTH, 2);
-	so_data  (so, (unsigned char)(cso->line_width * 8.0) & 0xff);
-	so_data  (so, cso->line_smooth ? 1 : 0);
-	so_method(so, curie, NV40TCL_LINE_STIPPLE_ENABLE, 2);
-	so_data  (so, cso->line_stipple_enable ? 1 : 0);
-	so_data  (so, (cso->line_stipple_pattern << 16) |
-		       cso->line_stipple_factor);
-
-	so_method(so, curie, NV40TCL_POINT_SIZE, 1);
-	so_data  (so, fui(cso->point_size));
-
-	so_method(so, curie, NV40TCL_POLYGON_MODE_FRONT, 6);
-	if (cso->front_winding == PIPE_WINDING_CCW) {
-		so_data(so, nvgl_polygon_mode(cso->fill_ccw));
-		so_data(so, nvgl_polygon_mode(cso->fill_cw));
-		switch (cso->cull_mode) {
-		case PIPE_WINDING_CCW:
-			so_data(so, NV40TCL_CULL_FACE_FRONT);
-			break;
-		case PIPE_WINDING_CW:
-			so_data(so, NV40TCL_CULL_FACE_BACK);
-			break;
-		case PIPE_WINDING_BOTH:
-			so_data(so, NV40TCL_CULL_FACE_FRONT_AND_BACK);
-			break;
-		default:
-			so_data(so, NV40TCL_CULL_FACE_BACK);
-			break;
-		}
-		so_data(so, NV40TCL_FRONT_FACE_CCW);
-	} else {
-		so_data(so, nvgl_polygon_mode(cso->fill_cw));
-		so_data(so, nvgl_polygon_mode(cso->fill_ccw));
-		switch (cso->cull_mode) {
-		case PIPE_WINDING_CCW:
-			so_data(so, NV40TCL_CULL_FACE_BACK);
-			break;
-		case PIPE_WINDING_CW:
-			so_data(so, NV40TCL_CULL_FACE_FRONT);
-			break;
-		case PIPE_WINDING_BOTH:
-			so_data(so, NV40TCL_CULL_FACE_FRONT_AND_BACK);
-			break;
-		default:
-			so_data(so, NV40TCL_CULL_FACE_BACK);
-			break;
-		}
-		so_data(so, NV40TCL_FRONT_FACE_CW);
-	}
-	so_data(so, cso->poly_smooth ? 1 : 0);
-	so_data(so, (cso->cull_mode != PIPE_WINDING_NONE) ? 1 : 0);
-
-	so_method(so, curie, NV40TCL_POLYGON_STIPPLE_ENABLE, 1);
-	so_data  (so, cso->poly_stipple_enable ? 1 : 0);
-
-	so_method(so, curie, NV40TCL_POLYGON_OFFSET_POINT_ENABLE, 3);
-	if ((cso->offset_cw && cso->fill_cw == PIPE_POLYGON_MODE_POINT) ||
-	    (cso->offset_ccw && cso->fill_ccw == PIPE_POLYGON_MODE_POINT))
-		so_data(so, 1);
-	else
-		so_data(so, 0);
-	if ((cso->offset_cw && cso->fill_cw == PIPE_POLYGON_MODE_LINE) ||
-	    (cso->offset_ccw && cso->fill_ccw == PIPE_POLYGON_MODE_LINE))
-		so_data(so, 1);
-	else
-		so_data(so, 0);
-	if ((cso->offset_cw && cso->fill_cw == PIPE_POLYGON_MODE_FILL) ||
-	    (cso->offset_ccw && cso->fill_ccw == PIPE_POLYGON_MODE_FILL))
-		so_data(so, 1);
-	else
-		so_data(so, 0);
-	if (cso->offset_cw || cso->offset_ccw) {
-		so_method(so, curie, NV40TCL_POLYGON_OFFSET_FACTOR, 2);
-		so_data  (so, fui(cso->offset_scale));
-		so_data  (so, fui(cso->offset_units * 2));
-	}
-
-	so_method(so, curie, NV40TCL_POINT_SPRITE, 1);
-	if (cso->point_quad_rasterization) {
-		unsigned psctl = (1 << 0), i;
-
-		for (i = 0; i < 8; i++) {
-			if ((cso->sprite_coord_enable >> i) & 1)
-				psctl |= (1 << (8 + i));
-		}
-
-		so_data(so, psctl);
-	} else {
-		so_data(so, 0);
-	}
-
-	so_ref(so, &rsso->so);
-	so_ref(NULL, &so);
-	rsso->pipe = *cso;
-	return (void *)rsso;
-}
-
-static void
-nv40_rasterizer_state_bind(struct pipe_context *pipe, void *hwcso)
-{
-	struct nv40_context *nv40 = nv40_context(pipe);
-
-	nv40->rasterizer = hwcso;
-	nv40->dirty |= NV40_NEW_RAST;
-	nv40->draw_dirty |= NV40_NEW_RAST;
-}
-
-static void
-nv40_rasterizer_state_delete(struct pipe_context *pipe, void *hwcso)
-{
-	struct nv40_rasterizer_state *rsso = hwcso;
-
-	so_ref(NULL, &rsso->so);
-	FREE(rsso);
-}
-
-static void *
-nv40_depth_stencil_alpha_state_create(struct pipe_context *pipe,
-			const struct pipe_depth_stencil_alpha_state *cso)
-{
-	struct nv40_context *nv40 = nv40_context(pipe);
-	struct nv40_zsa_state *zsaso = CALLOC(1, sizeof(*zsaso));
-	struct nouveau_stateobj *so = so_new(6, 20, 0);
-	struct nouveau_grobj *curie = nv40->screen->curie;
-
-	so_method(so, curie, NV40TCL_DEPTH_FUNC, 3);
-	so_data  (so, nvgl_comparison_op(cso->depth.func));
-	so_data  (so, cso->depth.writemask ? 1 : 0);
-	so_data  (so, cso->depth.enabled ? 1 : 0);
-
-	so_method(so, curie, NV40TCL_ALPHA_TEST_ENABLE, 3);
-	so_data  (so, cso->alpha.enabled ? 1 : 0);
-	so_data  (so, nvgl_comparison_op(cso->alpha.func));
-	so_data  (so, float_to_ubyte(cso->alpha.ref_value));
-
-	if (cso->stencil[0].enabled) {
-		so_method(so, curie, NV40TCL_STENCIL_FRONT_ENABLE, 3);
-		so_data  (so, cso->stencil[0].enabled ? 1 : 0);
-		so_data  (so, cso->stencil[0].writemask);
-		so_data  (so, nvgl_comparison_op(cso->stencil[0].func));
-		so_method(so, curie, NV40TCL_STENCIL_FRONT_FUNC_MASK, 4);
-		so_data  (so, cso->stencil[0].valuemask);
-		so_data  (so, nvgl_stencil_op(cso->stencil[0].fail_op));
-		so_data  (so, nvgl_stencil_op(cso->stencil[0].zfail_op));
-		so_data  (so, nvgl_stencil_op(cso->stencil[0].zpass_op));
-	} else {
-		so_method(so, curie, NV40TCL_STENCIL_FRONT_ENABLE, 1);
-		so_data  (so, 0);
-	}
-
-	if (cso->stencil[1].enabled) {
-		so_method(so, curie, NV40TCL_STENCIL_BACK_ENABLE, 3);
-		so_data  (so, cso->stencil[1].enabled ? 1 : 0);
-		so_data  (so, cso->stencil[1].writemask);
-		so_data  (so, nvgl_comparison_op(cso->stencil[1].func));
-		so_method(so, curie, NV40TCL_STENCIL_BACK_FUNC_MASK, 4);
-		so_data  (so, cso->stencil[1].valuemask);
-		so_data  (so, nvgl_stencil_op(cso->stencil[1].fail_op));
-		so_data  (so, nvgl_stencil_op(cso->stencil[1].zfail_op));
-		so_data  (so, nvgl_stencil_op(cso->stencil[1].zpass_op));
-	} else {
-		so_method(so, curie, NV40TCL_STENCIL_BACK_ENABLE, 1);
-		so_data  (so, 0);
-	}
-
-	so_ref(so, &zsaso->so);
-	so_ref(NULL, &so);
-	zsaso->pipe = *cso;
-	return (void *)zsaso;
-}
-
-static void
-nv40_depth_stencil_alpha_state_bind(struct pipe_context *pipe, void *hwcso)
-{
-	struct nv40_context *nv40 = nv40_context(pipe);
-
-	nv40->zsa = hwcso;
-	nv40->dirty |= NV40_NEW_ZSA;
-}
-
-static void
-nv40_depth_stencil_alpha_state_delete(struct pipe_context *pipe, void *hwcso)
-{
-	struct nv40_zsa_state *zsaso = hwcso;
-
-	so_ref(NULL, &zsaso->so);
-	FREE(zsaso);
-}
-
-static void *
-nv40_vp_state_create(struct pipe_context *pipe,
-		     const struct pipe_shader_state *cso)
-{
-	struct nv40_context *nv40 = nv40_context(pipe);
-	struct nv40_vertex_program *vp;
-
-	vp = CALLOC(1, sizeof(struct nv40_vertex_program));
-	vp->pipe.tokens = tgsi_dup_tokens(cso->tokens);
-	vp->draw = draw_create_vertex_shader(nv40->draw, &vp->pipe);
-
-	return (void *)vp;
-}
-
-static void
-nv40_vp_state_bind(struct pipe_context *pipe, void *hwcso)
-{
-	struct nv40_context *nv40 = nv40_context(pipe);
-
-	nv40->vertprog = hwcso;
-	nv40->dirty |= NV40_NEW_VERTPROG;
-	nv40->draw_dirty |= NV40_NEW_VERTPROG;
-}
-
-static void
-nv40_vp_state_delete(struct pipe_context *pipe, void *hwcso)
-{
-	struct nv40_context *nv40 = nv40_context(pipe);
-	struct nv40_vertex_program *vp = hwcso;
-
-	draw_delete_vertex_shader(nv40->draw, vp->draw);
-	nv40_vertprog_destroy(nv40, vp);
-	FREE((void*)vp->pipe.tokens);
-	FREE(vp);
-}
-
-static void *
-nv40_fp_state_create(struct pipe_context *pipe,
-		     const struct pipe_shader_state *cso)
-{
-	struct nv40_fragment_program *fp;
-
-	fp = CALLOC(1, sizeof(struct nv40_fragment_program));
-	fp->pipe.tokens = tgsi_dup_tokens(cso->tokens);
-
-	tgsi_scan_shader(fp->pipe.tokens, &fp->info);
-
-	return (void *)fp;
-}
-
-static void
-nv40_fp_state_bind(struct pipe_context *pipe, void *hwcso)
-{
-	struct nv40_context *nv40 = nv40_context(pipe);
-
-	nv40->fragprog = hwcso;
-	nv40->dirty |= NV40_NEW_FRAGPROG;
-}
-
-static void
-nv40_fp_state_delete(struct pipe_context *pipe, void *hwcso)
-{
-	struct nv40_context *nv40 = nv40_context(pipe);
-	struct nv40_fragment_program *fp = hwcso;
-
-	nv40_fragprog_destroy(nv40, fp);
-	FREE((void*)fp->pipe.tokens);
-	FREE(fp);
-}
-
-static void
-nv40_set_blend_color(struct pipe_context *pipe,
-		     const struct pipe_blend_color *bcol)
-{
-	struct nv40_context *nv40 = nv40_context(pipe);
-
-	nv40->blend_colour = *bcol;
-	nv40->dirty |= NV40_NEW_BCOL;
-}
-
- static void
-nv40_set_stencil_ref(struct pipe_context *pipe,
-		     const struct pipe_stencil_ref *sr)
-{
-	struct nv40_context *nv40 = nv40_context(pipe);
-
-	nv40->stencil_ref = *sr;
-	nv40->dirty |= NV40_NEW_SR;
-}
-
-static void
-nv40_set_clip_state(struct pipe_context *pipe,
-		    const struct pipe_clip_state *clip)
-{
-	struct nv40_context *nv40 = nv40_context(pipe);
-
-	nv40->clip = *clip;
-	nv40->dirty |= NV40_NEW_UCP;
-	nv40->draw_dirty |= NV40_NEW_UCP;
-}
-
-static void
-nv40_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index,
-			 struct pipe_buffer *buf )
-{
-	struct nv40_context *nv40 = nv40_context(pipe);
-
-	nv40->constbuf[shader] = buf;
-	nv40->constbuf_nr[shader] = buf->size / (4 * sizeof(float));
-
-	if (shader == PIPE_SHADER_VERTEX) {
-		nv40->dirty |= NV40_NEW_VERTPROG;
-	} else
-	if (shader == PIPE_SHADER_FRAGMENT) {
-		nv40->dirty |= NV40_NEW_FRAGPROG;
-	}
-}
-
-static void
-nv40_set_framebuffer_state(struct pipe_context *pipe,
-			   const struct pipe_framebuffer_state *fb)
-{
-	struct nv40_context *nv40 = nv40_context(pipe);
-
-	nv40->framebuffer = *fb;
-	nv40->dirty |= NV40_NEW_FB;
-}
-
-static void
-nv40_set_polygon_stipple(struct pipe_context *pipe,
-			 const struct pipe_poly_stipple *stipple)
-{
-	struct nv40_context *nv40 = nv40_context(pipe);
-
-	memcpy(nv40->stipple, stipple->stipple, 4 * 32);
-	nv40->dirty |= NV40_NEW_STIPPLE;
-}
-
-static void
-nv40_set_scissor_state(struct pipe_context *pipe,
-		       const struct pipe_scissor_state *s)
-{
-	struct nv40_context *nv40 = nv40_context(pipe);
-
-	nv40->scissor = *s;
-	nv40->dirty |= NV40_NEW_SCISSOR;
-}
-
-static void
-nv40_set_viewport_state(struct pipe_context *pipe,
-			const struct pipe_viewport_state *vpt)
-{
-	struct nv40_context *nv40 = nv40_context(pipe);
-
-	nv40->viewport = *vpt;
-	nv40->dirty |= NV40_NEW_VIEWPORT;
-	nv40->draw_dirty |= NV40_NEW_VIEWPORT;
-}
-
-static void
-nv40_set_vertex_buffers(struct pipe_context *pipe, unsigned count,
-			const struct pipe_vertex_buffer *vb)
-{
-	struct nv40_context *nv40 = nv40_context(pipe);
-
-	memcpy(nv40->vtxbuf, vb, sizeof(*vb) * count);
-	nv40->vtxbuf_nr = count;
-
-	nv40->dirty |= NV40_NEW_ARRAYS;
-	nv40->draw_dirty |= NV40_NEW_ARRAYS;
-}
-
-static void *
-nv40_vtxelts_state_create(struct pipe_context *pipe,
-			  unsigned num_elements,
-			  const struct pipe_vertex_element *elements)
-{
-	struct nv40_vtxelt_state *cso = CALLOC_STRUCT(nv40_vtxelt_state);
-
-	assert(num_elements < 16); /* not doing fallbacks yet */
-	cso->num_elements = num_elements;
-	memcpy(cso->pipe, elements, num_elements * sizeof(*elements));
-
-/*	nv40_vtxelt_construct(cso);*/
-
-	return (void *)cso;
-}
-
-static void
-nv40_vtxelts_state_delete(struct pipe_context *pipe, void *hwcso)
-{
-	FREE(hwcso);
-}
-
-static void
-nv40_vtxelts_state_bind(struct pipe_context *pipe, void *hwcso)
-{
-	struct nv40_context *nv40 = nv40_context(pipe);
-
-	nv40->vtxelt = hwcso;
-	nv40->dirty |= NV40_NEW_ARRAYS;
-	nv40->draw_dirty |= NV40_NEW_ARRAYS;
-}
-
-void
-nv40_init_state_functions(struct nv40_context *nv40)
-{
-	nv40->pipe.create_blend_state = nv40_blend_state_create;
-	nv40->pipe.bind_blend_state = nv40_blend_state_bind;
-	nv40->pipe.delete_blend_state = nv40_blend_state_delete;
-
-	nv40->pipe.create_sampler_state = nv40_sampler_state_create;
-	nv40->pipe.bind_fragment_sampler_states = nv40_sampler_state_bind;
-	nv40->pipe.delete_sampler_state = nv40_sampler_state_delete;
-	nv40->pipe.set_fragment_sampler_views = nv40_set_fragment_sampler_views;
-	nv40->pipe.create_sampler_view = nv40_create_sampler_view;
-	nv40->pipe.sampler_view_destroy = nv40_sampler_view_destroy;
-
-	nv40->pipe.create_rasterizer_state = nv40_rasterizer_state_create;
-	nv40->pipe.bind_rasterizer_state = nv40_rasterizer_state_bind;
-	nv40->pipe.delete_rasterizer_state = nv40_rasterizer_state_delete;
-
-	nv40->pipe.create_depth_stencil_alpha_state =
-		nv40_depth_stencil_alpha_state_create;
-	nv40->pipe.bind_depth_stencil_alpha_state =
-		nv40_depth_stencil_alpha_state_bind;
-	nv40->pipe.delete_depth_stencil_alpha_state =
-		nv40_depth_stencil_alpha_state_delete;
-
-	nv40->pipe.create_vs_state = nv40_vp_state_create;
-	nv40->pipe.bind_vs_state = nv40_vp_state_bind;
-	nv40->pipe.delete_vs_state = nv40_vp_state_delete;
-
-	nv40->pipe.create_fs_state = nv40_fp_state_create;
-	nv40->pipe.bind_fs_state = nv40_fp_state_bind;
-	nv40->pipe.delete_fs_state = nv40_fp_state_delete;
-
-	nv40->pipe.set_blend_color = nv40_set_blend_color;
-        nv40->pipe.set_stencil_ref = nv40_set_stencil_ref;
-	nv40->pipe.set_clip_state = nv40_set_clip_state;
-	nv40->pipe.set_constant_buffer = nv40_set_constant_buffer;
-	nv40->pipe.set_framebuffer_state = nv40_set_framebuffer_state;
-	nv40->pipe.set_polygon_stipple = nv40_set_polygon_stipple;
-	nv40->pipe.set_scissor_state = nv40_set_scissor_state;
-	nv40->pipe.set_viewport_state = nv40_set_viewport_state;
-
-	nv40->pipe.create_vertex_elements_state = nv40_vtxelts_state_create;
-	nv40->pipe.delete_vertex_elements_state = nv40_vtxelts_state_delete;
-	nv40->pipe.bind_vertex_elements_state = nv40_vtxelts_state_bind;
-
-	nv40->pipe.set_vertex_buffers = nv40_set_vertex_buffers;
-}
-
diff --git a/src/gallium/drivers/nv40/nv40_state_blend.c b/src/gallium/drivers/nv40/nv40_state_blend.c
deleted file mode 100644
index 3ff00a37f6..0000000000
--- a/src/gallium/drivers/nv40/nv40_state_blend.c
+++ /dev/null
@@ -1,41 +0,0 @@
-#include "nv40_context.h"
-
-static boolean
-nv40_state_blend_validate(struct nv40_context *nv40)
-{
-	so_ref(nv40->blend->so, &nv40->state.hw[NV40_STATE_BLEND]);
-	return TRUE;
-}
-
-struct nv40_state_entry nv40_state_blend = {
-	.validate = nv40_state_blend_validate,
-	.dirty = {
-		.pipe = NV40_NEW_BLEND,
-		.hw = NV40_STATE_BLEND
-	}
-};
-
-static boolean
-nv40_state_blend_colour_validate(struct nv40_context *nv40)
-{
-	struct nouveau_stateobj *so = so_new(1, 1, 0);
-	struct pipe_blend_color *bcol = &nv40->blend_colour;
-
-	so_method(so, nv40->screen->curie, NV40TCL_BLEND_COLOR, 1);
-	so_data  (so, ((float_to_ubyte(bcol->color[3]) << 24) |
-		       (float_to_ubyte(bcol->color[0]) << 16) |
-		       (float_to_ubyte(bcol->color[1]) <<  8) |
-		       (float_to_ubyte(bcol->color[2]) <<  0)));
-
-	so_ref(so, &nv40->state.hw[NV40_STATE_BCOL]);
-	so_ref(NULL, &so);
-	return TRUE;
-}
-
-struct nv40_state_entry nv40_state_blend_colour = {
-	.validate = nv40_state_blend_colour_validate,
-	.dirty = {
-		.pipe = NV40_NEW_BCOL,
-		.hw = NV40_STATE_BCOL
-	}
-};
diff --git a/src/gallium/drivers/nv40/nv40_state_emit.c b/src/gallium/drivers/nv40/nv40_state_emit.c
deleted file mode 100644
index 297d71f4fa..0000000000
--- a/src/gallium/drivers/nv40/nv40_state_emit.c
+++ /dev/null
@@ -1,189 +0,0 @@
-#include "nv40_context.h"
-#include "nv40_state.h"
-#include "draw/draw_context.h"
-
-static struct nv40_state_entry *render_states[] = {
-	&nv40_state_framebuffer,
-	&nv40_state_rasterizer,
-	&nv40_state_scissor,
-	&nv40_state_stipple,
-	&nv40_state_fragprog,
-	&nv40_state_fragtex,
-	&nv40_state_vertprog,
-	&nv40_state_blend,
-	&nv40_state_blend_colour,
-	&nv40_state_zsa,
-	&nv40_state_sr,
-	&nv40_state_viewport,
-	&nv40_state_vbo,
-	NULL
-};
-
-static struct nv40_state_entry *swtnl_states[] = {
-	&nv40_state_framebuffer,
-	&nv40_state_rasterizer,
-	&nv40_state_scissor,
-	&nv40_state_stipple,
-	&nv40_state_fragprog,
-	&nv40_state_fragtex,
-	&nv40_state_vertprog,
-	&nv40_state_blend,
-	&nv40_state_blend_colour,
-	&nv40_state_zsa,
-	&nv40_state_sr,
-	&nv40_state_viewport,
-	&nv40_state_vtxfmt,
-	NULL
-};
-
-static void
-nv40_state_do_validate(struct nv40_context *nv40,
-		       struct nv40_state_entry **states)
-{
-	while (*states) {
-		struct nv40_state_entry *e = *states;
-
-		if (nv40->dirty & e->dirty.pipe) {
-			if (e->validate(nv40))
-				nv40->state.dirty |= (1ULL << e->dirty.hw);
-		}
-
-		states++;
-	}
-	nv40->dirty = 0;
-}
-
-void
-nv40_state_emit(struct nv40_context *nv40)
-{
-	struct nv40_state *state = &nv40->state;
-	struct nv40_screen *screen = nv40->screen;
-	struct nouveau_channel *chan = screen->base.channel;
-	struct nouveau_grobj *curie = screen->curie;
-	unsigned i;
-	uint64_t states;
-
-	/* XXX: race conditions
-	 */
-	if (nv40 != screen->cur_ctx) {
-		for (i = 0; i < NV40_STATE_MAX; i++) {
-			if (state->hw[i] && screen->state[i] != state->hw[i])
-				state->dirty |= (1ULL << i);
-		}
-
-		screen->cur_ctx = nv40;
-	}
-
-	for (i = 0, states = state->dirty; states; i++) {
-		if (!(states & (1ULL << i)))
-			continue;
-		so_ref (state->hw[i], &nv40->screen->state[i]);
-		if (state->hw[i])
-			so_emit(chan, nv40->screen->state[i]);
-		states &= ~(1ULL << i);
-	}
-
-	if (state->dirty & ((1ULL << NV40_STATE_FRAGPROG) |
-			    (1ULL << NV40_STATE_FRAGTEX0))) {
-		BEGIN_RING(chan, curie, NV40TCL_TEX_CACHE_CTL, 1);
-		OUT_RING  (chan, 2);
-		BEGIN_RING(chan, curie, NV40TCL_TEX_CACHE_CTL, 1);
-		OUT_RING  (chan, 1);
-	}
-
-	state->dirty = 0;
-}
-
-void
-nv40_state_flush_notify(struct nouveau_channel *chan)
-{
-	struct nv40_context *nv40 = chan->user_private;
-	struct nv40_state *state = &nv40->state;
-	unsigned i, samplers;
-
-	so_emit_reloc_markers(chan, state->hw[NV40_STATE_FB]);
-	for (i = 0, samplers = state->fp_samplers; i < 16 && samplers; i++) {
-		if (!(samplers & (1 << i)))
-			continue;
-		so_emit_reloc_markers(chan,
-				      state->hw[NV40_STATE_FRAGTEX0+i]);
-		samplers &= ~(1ULL << i);
-	}
-	so_emit_reloc_markers(chan, state->hw[NV40_STATE_FRAGPROG]);
-	if (state->hw[NV40_STATE_VTXBUF] && nv40->render_mode == HW)
-		so_emit_reloc_markers(chan, state->hw[NV40_STATE_VTXBUF]);
-}
-
-boolean
-nv40_state_validate(struct nv40_context *nv40)
-{
-	boolean was_sw = nv40->fallback_swtnl ? TRUE : FALSE;
-
-	if (nv40->render_mode != HW) {
-		/* Don't even bother trying to go back to hw if none
-		 * of the states that caused swtnl previously have changed.
-		 */
-		if ((nv40->fallback_swtnl & nv40->dirty)
-				!= nv40->fallback_swtnl)
-			return FALSE;
-
-		/* Attempt to go to hwtnl again */
-		nv40->pipe.flush(&nv40->pipe, 0, NULL);
-		nv40->dirty |= (NV40_NEW_VIEWPORT |
-				NV40_NEW_VERTPROG |
-				NV40_NEW_ARRAYS);
-		nv40->render_mode = HW;
-	}
-
-	nv40_state_do_validate(nv40, render_states);
-	if (nv40->fallback_swtnl || nv40->fallback_swrast)
-		return FALSE;
-	
-	if (was_sw)
-		NOUVEAU_ERR("swtnl->hw\n");
-
-	return TRUE;
-}
-
-boolean
-nv40_state_validate_swtnl(struct nv40_context *nv40)
-{
-	struct draw_context *draw = nv40->draw;
-
-	/* Setup for swtnl */
-	if (nv40->render_mode == HW) {
-		NOUVEAU_ERR("hw->swtnl 0x%08x\n", nv40->fallback_swtnl);
-		nv40->pipe.flush(&nv40->pipe, 0, NULL);
-		nv40->dirty |= (NV40_NEW_VIEWPORT |
-				NV40_NEW_VERTPROG |
-				NV40_NEW_ARRAYS);
-		nv40->render_mode = SWTNL;
-	}
-
-	if (nv40->draw_dirty & NV40_NEW_VERTPROG)
-		draw_bind_vertex_shader(draw, nv40->vertprog->draw);
-
-	if (nv40->draw_dirty & NV40_NEW_RAST)
-		draw_set_rasterizer_state(draw, &nv40->rasterizer->pipe);
-
-	if (nv40->draw_dirty & NV40_NEW_UCP)
-		draw_set_clip_state(draw, &nv40->clip);
-
-	if (nv40->draw_dirty & NV40_NEW_VIEWPORT)
-		draw_set_viewport_state(draw, &nv40->viewport);
-
-	if (nv40->draw_dirty & NV40_NEW_ARRAYS) {
-		draw_set_vertex_buffers(draw, nv40->vtxbuf_nr, nv40->vtxbuf);
-		draw_set_vertex_elements(draw, nv40->vtxelt->num_elements, nv40->vtxelt->pipe);	
-	}
-
-	nv40_state_do_validate(nv40, swtnl_states);
-	if (nv40->fallback_swrast) {
-		NOUVEAU_ERR("swtnl->swrast 0x%08x\n", nv40->fallback_swrast);
-		return FALSE;
-	}
-
-	nv40->draw_dirty = 0;
-	return TRUE;
-}
-
diff --git a/src/gallium/drivers/nv40/nv40_state_fb.c b/src/gallium/drivers/nv40/nv40_state_fb.c
deleted file mode 100644
index fd3fdfddc0..0000000000
--- a/src/gallium/drivers/nv40/nv40_state_fb.c
+++ /dev/null
@@ -1,175 +0,0 @@
-#include "nv40_context.h"
-#include "nouveau/nouveau_util.h"
-
-static struct pipe_buffer *
-nv40_do_surface_buffer(struct pipe_surface *surface)
-{
-	struct nv40_miptree *mt = (struct nv40_miptree *)surface->texture;
-	return mt->buffer;
-}
-
-#define nv40_surface_buffer(ps) nouveau_bo(nv40_do_surface_buffer(ps))
-
-static boolean
-nv40_state_framebuffer_validate(struct nv40_context *nv40)
-{
-	struct nouveau_channel *chan = nv40->screen->base.channel;
-	struct nouveau_grobj *curie = nv40->screen->curie;
-	struct pipe_framebuffer_state *fb = &nv40->framebuffer;
-	struct nv04_surface *rt[4], *zeta;
-	uint32_t rt_enable, rt_format;
-	int i, colour_format = 0, zeta_format = 0;
-	struct nouveau_stateobj *so = so_new(18, 24, 10);
-	unsigned rt_flags = NOUVEAU_BO_RDWR | NOUVEAU_BO_VRAM;
-	unsigned w = fb->width;
-	unsigned h = fb->height;
-
-	rt_enable = 0;
-	for (i = 0; i < fb->nr_cbufs; i++) {
-		if (colour_format) {
-			assert(colour_format == fb->cbufs[i]->format);
-		} else {
-			colour_format = fb->cbufs[i]->format;
-			rt_enable |= (NV40TCL_RT_ENABLE_COLOR0 << i);
-			rt[i] = (struct nv04_surface *)fb->cbufs[i];
-		}
-	}
-
-	if (rt_enable & (NV40TCL_RT_ENABLE_COLOR1 | NV40TCL_RT_ENABLE_COLOR2 |
-			 NV40TCL_RT_ENABLE_COLOR3))
-		rt_enable |= NV40TCL_RT_ENABLE_MRT;
-
-	if (fb->zsbuf) {
-		zeta_format = fb->zsbuf->format;
-		zeta = (struct nv04_surface *)fb->zsbuf;
-	}
-
-	if (!(rt[0]->base.texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) {
-		assert(!(fb->width & (fb->width - 1)) && !(fb->height & (fb->height - 1)));
-		for (i = 1; i < fb->nr_cbufs; i++)
-			assert(!(rt[i]->base.texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR));
-
-		rt_format = NV40TCL_RT_FORMAT_TYPE_SWIZZLED |
-		            log2i(fb->width) << NV40TCL_RT_FORMAT_LOG2_WIDTH_SHIFT |
-		            log2i(fb->height) << NV40TCL_RT_FORMAT_LOG2_HEIGHT_SHIFT;
-	}
-	else
-		rt_format = NV40TCL_RT_FORMAT_TYPE_LINEAR;
-
-	switch (colour_format) {
-	case PIPE_FORMAT_B8G8R8X8_UNORM:
-		rt_format |= NV40TCL_RT_FORMAT_COLOR_X8R8G8B8;
-		break;
-	case PIPE_FORMAT_B8G8R8A8_UNORM:
-	case 0:
-		rt_format |= NV40TCL_RT_FORMAT_COLOR_A8R8G8B8;
-		break;
-	case PIPE_FORMAT_B5G6R5_UNORM:
-		rt_format |= NV40TCL_RT_FORMAT_COLOR_R5G6B5;
-		break;
-	default:
-		assert(0);
-	}
-
-	switch (zeta_format) {
-	case PIPE_FORMAT_Z16_UNORM:
-		rt_format |= NV40TCL_RT_FORMAT_ZETA_Z16;
-		break;
-	case PIPE_FORMAT_S8Z24_UNORM:
-	case PIPE_FORMAT_X8Z24_UNORM:
-	case 0:
-		rt_format |= NV40TCL_RT_FORMAT_ZETA_Z24S8;
-		break;
-	default:
-		assert(0);
-	}
-
-	if (rt_enable & NV40TCL_RT_ENABLE_COLOR0) {
-		so_method(so, curie, NV40TCL_DMA_COLOR0, 1);
-		so_reloc (so, nv40_surface_buffer(&rt[0]->base), 0,
-			      rt_flags | NOUVEAU_BO_OR,
-			      chan->vram->handle, chan->gart->handle);
-		so_method(so, curie, NV40TCL_COLOR0_PITCH, 2);
-		so_data  (so, rt[0]->pitch);
-		so_reloc (so, nv40_surface_buffer(&rt[0]->base),
-			      rt[0]->base.offset, rt_flags | NOUVEAU_BO_LOW,
-			      0, 0);
-	}
-
-	if (rt_enable & NV40TCL_RT_ENABLE_COLOR1) {
-		so_method(so, curie, NV40TCL_DMA_COLOR1, 1);
-		so_reloc (so, nv40_surface_buffer(&rt[1]->base), 0,
-			      rt_flags | NOUVEAU_BO_OR,
-			      chan->vram->handle, chan->gart->handle);
-		so_method(so, curie, NV40TCL_COLOR1_OFFSET, 2);
-		so_reloc (so, nv40_surface_buffer(&rt[1]->base),
-			      rt[1]->base.offset, rt_flags | NOUVEAU_BO_LOW,
-			      0, 0);
-		so_data  (so, rt[1]->pitch);
-	}
-
-	if (rt_enable & NV40TCL_RT_ENABLE_COLOR2) {
-		so_method(so, curie, NV40TCL_DMA_COLOR2, 1);
-		so_reloc (so, nv40_surface_buffer(&rt[2]->base), 0,
-			      rt_flags | NOUVEAU_BO_OR,
-			      chan->vram->handle, chan->gart->handle);
-		so_method(so, curie, NV40TCL_COLOR2_OFFSET, 1);
-		so_reloc (so, nv40_surface_buffer(&rt[2]->base),
-			      rt[2]->base.offset, rt_flags | NOUVEAU_BO_LOW,
-			      0, 0);
-		so_method(so, curie, NV40TCL_COLOR2_PITCH, 1);
-		so_data  (so, rt[2]->pitch);
-	}
-
-	if (rt_enable & NV40TCL_RT_ENABLE_COLOR3) {
-		so_method(so, curie, NV40TCL_DMA_COLOR3, 1);
-		so_reloc (so, nv40_surface_buffer(&rt[3]->base), 0,
-			      rt_flags | NOUVEAU_BO_OR,
-			      chan->vram->handle, chan->gart->handle);
-		so_method(so, curie, NV40TCL_COLOR3_OFFSET, 1);
-		so_reloc (so, nv40_surface_buffer(&rt[3]->base),
-			      rt[3]->base.offset, rt_flags | NOUVEAU_BO_LOW,
-			      0, 0);
-		so_method(so, curie, NV40TCL_COLOR3_PITCH, 1);
-		so_data  (so, rt[3]->pitch);
-	}
-
-	if (zeta_format) {
-		so_method(so, curie, NV40TCL_DMA_ZETA, 1);
-		so_reloc (so, nv40_surface_buffer(&zeta->base), 0,
-			      rt_flags | NOUVEAU_BO_OR,
-			      chan->vram->handle, chan->gart->handle);
-		so_method(so, curie, NV40TCL_ZETA_OFFSET, 1);
-		so_reloc (so, nv40_surface_buffer(&zeta->base),
-			      zeta->base.offset, rt_flags | NOUVEAU_BO_LOW, 0, 0);
-		so_method(so, curie, NV40TCL_ZETA_PITCH, 1);
-		so_data  (so, zeta->pitch);
-	}
-
-	so_method(so, curie, NV40TCL_RT_ENABLE, 1);
-	so_data  (so, rt_enable);
-	so_method(so, curie, NV40TCL_RT_HORIZ, 3);
-	so_data  (so, (w << 16) | 0);
-	so_data  (so, (h << 16) | 0);
-	so_data  (so, rt_format);
-	so_method(so, curie, NV40TCL_VIEWPORT_HORIZ, 2);
-	so_data  (so, (w << 16) | 0);
-	so_data  (so, (h << 16) | 0);
-	so_method(so, curie, NV40TCL_VIEWPORT_CLIP_HORIZ(0), 2);
-	so_data  (so, ((w - 1) << 16) | 0);
-	so_data  (so, ((h - 1) << 16) | 0);
-	so_method(so, curie, 0x1d88, 1);
-	so_data  (so, (1 << 12) | h);
-
-	so_ref(so, &nv40->state.hw[NV40_STATE_FB]);
-	so_ref(NULL, &so);
-	return TRUE;
-}
-
-struct nv40_state_entry nv40_state_framebuffer = {
-	.validate = nv40_state_framebuffer_validate,
-	.dirty = {
-		.pipe = NV40_NEW_FB,
-		.hw = NV40_STATE_FB
-	}
-};
diff --git a/src/gallium/drivers/nv40/nv40_state_rasterizer.c b/src/gallium/drivers/nv40/nv40_state_rasterizer.c
deleted file mode 100644
index 9ecda5990f..0000000000
--- a/src/gallium/drivers/nv40/nv40_state_rasterizer.c
+++ /dev/null
@@ -1,17 +0,0 @@
-#include "nv40_context.h"
-
-static boolean
-nv40_state_rasterizer_validate(struct nv40_context *nv40)
-{
-	so_ref(nv40->rasterizer->so,
-	       &nv40->state.hw[NV40_STATE_RAST]);
-	return TRUE;
-}
-
-struct nv40_state_entry nv40_state_rasterizer = {
-	.validate = nv40_state_rasterizer_validate,
-	.dirty = {
-		.pipe = NV40_NEW_RAST,
-		.hw = NV40_STATE_RAST
-	}
-};
diff --git a/src/gallium/drivers/nv40/nv40_state_scissor.c b/src/gallium/drivers/nv40/nv40_state_scissor.c
deleted file mode 100644
index 753a505e93..0000000000
--- a/src/gallium/drivers/nv40/nv40_state_scissor.c
+++ /dev/null
@@ -1,36 +0,0 @@
-#include "nv40_context.h"
-
-static boolean
-nv40_state_scissor_validate(struct nv40_context *nv40)
-{
-	struct pipe_rasterizer_state *rast = &nv40->rasterizer->pipe;
-	struct pipe_scissor_state *s = &nv40->scissor;
-	struct nouveau_stateobj *so;
-
-	if (nv40->state.hw[NV40_STATE_SCISSOR] &&
-	    (rast->scissor == 0 && nv40->state.scissor_enabled == 0))
-		return FALSE;
-	nv40->state.scissor_enabled = rast->scissor;
-
-	so = so_new(1, 2, 0);
-	so_method(so, nv40->screen->curie, NV40TCL_SCISSOR_HORIZ, 2);
-	if (nv40->state.scissor_enabled) {
-		so_data  (so, ((s->maxx - s->minx) << 16) | s->minx);
-		so_data  (so, ((s->maxy - s->miny) << 16) | s->miny);
-	} else {
-		so_data  (so, 4096 << 16);
-		so_data  (so, 4096 << 16);
-	}
-
-	so_ref(so, &nv40->state.hw[NV40_STATE_SCISSOR]);
-	so_ref(NULL, &so);
-	return TRUE;
-}
-
-struct nv40_state_entry nv40_state_scissor = {
-	.validate = nv40_state_scissor_validate,
-	.dirty = {
-		.pipe = NV40_NEW_SCISSOR | NV40_NEW_RAST,
-		.hw = NV40_STATE_SCISSOR
-	}
-};
diff --git a/src/gallium/drivers/nv40/nv40_state_stipple.c b/src/gallium/drivers/nv40/nv40_state_stipple.c
deleted file mode 100644
index 2b371ebfec..0000000000
--- a/src/gallium/drivers/nv40/nv40_state_stipple.c
+++ /dev/null
@@ -1,39 +0,0 @@
-#include "nv40_context.h"
-
-static boolean
-nv40_state_stipple_validate(struct nv40_context *nv40)
-{
-	struct pipe_rasterizer_state *rast = &nv40->rasterizer->pipe;
-	struct nouveau_grobj *curie = nv40->screen->curie;
-	struct nouveau_stateobj *so;
-
-	if (nv40->state.hw[NV40_STATE_STIPPLE] &&
-	   (rast->poly_stipple_enable == 0 && nv40->state.stipple_enabled == 0))
-		return FALSE;
-
-	if (rast->poly_stipple_enable) {
-		unsigned i;
-
-		so = so_new(2, 33, 0);
-		so_method(so, curie, NV40TCL_POLYGON_STIPPLE_ENABLE, 1);
-		so_data  (so, 1);
-		so_method(so, curie, NV40TCL_POLYGON_STIPPLE_PATTERN(0), 32);
-		for (i = 0; i < 32; i++)
-			so_data(so, nv40->stipple[i]);
-	} else {
-		so = so_new(1, 1, 0);
-		so_method(so, curie, NV40TCL_POLYGON_STIPPLE_ENABLE, 1);
-		so_data  (so, 0);
-	}
-
-	so_ref(so, &nv40->state.hw[NV40_STATE_STIPPLE]);
-	return TRUE;
-}
-
-struct nv40_state_entry nv40_state_stipple = {
-	.validate = nv40_state_stipple_validate,
-	.dirty = {
-		.pipe = NV40_NEW_STIPPLE | NV40_NEW_RAST,
-		.hw = NV40_STATE_STIPPLE,
-	}
-};
diff --git a/src/gallium/drivers/nv40/nv40_state_viewport.c b/src/gallium/drivers/nv40/nv40_state_viewport.c
deleted file mode 100644
index 3aacb00f99..0000000000
--- a/src/gallium/drivers/nv40/nv40_state_viewport.c
+++ /dev/null
@@ -1,38 +0,0 @@
-#include "nv40_context.h"
-
-static boolean
-nv40_state_viewport_validate(struct nv40_context *nv40)
-{
-	struct pipe_viewport_state *vpt = &nv40->viewport;
-	struct nouveau_stateobj *so;
-
-	if (nv40->state.hw[NV40_STATE_VIEWPORT] &&
-	    !(nv40->dirty & NV40_NEW_VIEWPORT))
-		return FALSE;
-
-	so = so_new(2, 9, 0);
-	so_method(so, nv40->screen->curie,
-		  NV40TCL_VIEWPORT_TRANSLATE_X, 8);
-	so_data  (so, fui(vpt->translate[0]));
-	so_data  (so, fui(vpt->translate[1]));
-	so_data  (so, fui(vpt->translate[2]));
-	so_data  (so, fui(vpt->translate[3]));
-	so_data  (so, fui(vpt->scale[0]));
-	so_data  (so, fui(vpt->scale[1]));
-	so_data  (so, fui(vpt->scale[2]));
-	so_data  (so, fui(vpt->scale[3]));
-	so_method(so, nv40->screen->curie, 0x1d78, 1);
-	so_data  (so, 1);
-
-	so_ref(so, &nv40->state.hw[NV40_STATE_VIEWPORT]);
-	so_ref(NULL, &so);
-	return TRUE;
-}
-
-struct nv40_state_entry nv40_state_viewport = {
-	.validate = nv40_state_viewport_validate,
-	.dirty = {
-		.pipe = NV40_NEW_VIEWPORT | NV40_NEW_RAST,
-		.hw = NV40_STATE_VIEWPORT
-	}
-};
diff --git a/src/gallium/drivers/nv40/nv40_state_zsa.c b/src/gallium/drivers/nv40/nv40_state_zsa.c
deleted file mode 100644
index 9cbe7da6db..0000000000
--- a/src/gallium/drivers/nv40/nv40_state_zsa.c
+++ /dev/null
@@ -1,41 +0,0 @@
-#include "nv40_context.h"
-
-static boolean
-nv40_state_zsa_validate(struct nv40_context *nv40)
-{
-	so_ref(nv40->zsa->so,
-	       &nv40->state.hw[NV40_STATE_ZSA]);
-	return TRUE;
-}
-
-struct nv40_state_entry nv40_state_zsa = {
-	.validate = nv40_state_zsa_validate,
-	.dirty = {
-		.pipe = NV40_NEW_ZSA,
-		.hw = NV40_STATE_ZSA
-	}
-};
-
-static boolean
-nv40_state_sr_validate(struct nv40_context *nv40)
-{
-	struct nouveau_stateobj *so = so_new(2, 2, 0);
-	struct pipe_stencil_ref *sr = &nv40->stencil_ref;
-
-	so_method(so, nv40->screen->curie, NV40TCL_STENCIL_FRONT_FUNC_REF, 1);
-	so_data  (so, sr->ref_value[0]);
-	so_method(so, nv40->screen->curie, NV40TCL_STENCIL_BACK_FUNC_REF, 1);
-	so_data  (so, sr->ref_value[1]);
-
-	so_ref(so, &nv40->state.hw[NV40_STATE_SR]);
-	so_ref(NULL, &so);
-	return TRUE;
-}
-
-struct nv40_state_entry nv40_state_sr = {
-	.validate = nv40_state_sr_validate,
-	.dirty = {
-		.pipe = NV40_NEW_SR,
-		.hw = NV40_STATE_SR
-	}
-};
diff --git a/src/gallium/drivers/nv40/nv40_surface.c b/src/gallium/drivers/nv40/nv40_surface.c
deleted file mode 100644
index 02ecfd7bbb..0000000000
--- a/src/gallium/drivers/nv40/nv40_surface.c
+++ /dev/null
@@ -1,64 +0,0 @@
-
-/**************************************************************************
- * 
- * Copyright 2003 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.
- * 
- **************************************************************************/
-
-#include "pipe/p_defines.h"
-#include "util/u_inlines.h"
-
-#include "util/u_tile.h"
-
-#include "nv40_context.h"
-
-static void
-nv40_surface_copy(struct pipe_context *pipe,
-		  struct pipe_surface *dest, unsigned destx, unsigned desty,
-		  struct pipe_surface *src, unsigned srcx, unsigned srcy,
-		  unsigned width, unsigned height)
-{
-	struct nv40_context *nv40 = nv40_context(pipe);
-	struct nv04_surface_2d *eng2d = nv40->screen->eng2d;
-
-	eng2d->copy(eng2d, dest, destx, desty, src, srcx, srcy, width, height);
-}
-
-static void
-nv40_surface_fill(struct pipe_context *pipe, struct pipe_surface *dest,
-		  unsigned destx, unsigned desty, unsigned width,
-		  unsigned height, unsigned value)
-{
-	struct nv40_context *nv40 = nv40_context(pipe);
-	struct nv04_surface_2d *eng2d = nv40->screen->eng2d;
-
-	eng2d->fill(eng2d, dest, destx, desty, width, height, value);
-}
-
-void
-nv40_init_surface_functions(struct nv40_context *nv40)
-{
-	nv40->pipe.surface_copy = nv40_surface_copy;
-	nv40->pipe.surface_fill = nv40_surface_fill;
-}
diff --git a/src/gallium/drivers/nv40/nv40_transfer.c b/src/gallium/drivers/nv40/nv40_transfer.c
deleted file mode 100644
index c552a68113..0000000000
--- a/src/gallium/drivers/nv40/nv40_transfer.c
+++ /dev/null
@@ -1,181 +0,0 @@
-#include "pipe/p_state.h"
-#include "pipe/p_defines.h"
-#include "util/u_inlines.h"
-#include "util/u_format.h"
-#include "util/u_memory.h"
-#include "util/u_math.h"
-#include "nouveau/nouveau_winsys.h"
-#include "nv40_context.h"
-#include "nv40_screen.h"
-#include "nv40_state.h"
-
-struct nv40_transfer {
-	struct pipe_transfer base;
-	struct pipe_surface *surface;
-	boolean direct;
-};
-
-static void
-nv40_compatible_transfer_tex(struct pipe_texture *pt, unsigned width, unsigned height,
-                             struct pipe_texture *template)
-{
-	memset(template, 0, sizeof(struct pipe_texture));
-	template->target = pt->target;
-	template->format = pt->format;
-	template->width0 = width;
-	template->height0 = height;
-	template->depth0 = 1;
-	template->last_level = 0;
-	template->nr_samples = pt->nr_samples;
-
-	template->tex_usage = PIPE_TEXTURE_USAGE_DYNAMIC |
-	                      NOUVEAU_TEXTURE_USAGE_LINEAR;
-}
-
-static struct pipe_transfer *
-nv40_transfer_new(struct pipe_context *pcontext, struct pipe_texture *pt,
-		  unsigned face, unsigned level, unsigned zslice,
-		  enum pipe_transfer_usage usage,
-		  unsigned x, unsigned y, unsigned w, unsigned h)
-{
-        struct pipe_screen *pscreen = pcontext->screen;
-	struct nv40_miptree *mt = (struct nv40_miptree *)pt;
-	struct nv40_transfer *tx;
-	struct pipe_texture tx_tex_template, *tx_tex;
-
-	tx = CALLOC_STRUCT(nv40_transfer);
-	if (!tx)
-		return NULL;
-
-	pipe_texture_reference(&tx->base.texture, pt);
-	tx->base.x = x;
-	tx->base.y = y;
-	tx->base.width = w;
-	tx->base.height = h;
-	tx->base.stride = mt->level[level].pitch;
-	tx->base.usage = usage;
-	tx->base.face = face;
-	tx->base.level = level;
-	tx->base.zslice = zslice;
-
-	/* Direct access to texture */
-	if ((pt->tex_usage & PIPE_TEXTURE_USAGE_DYNAMIC ||
-	     debug_get_bool_option("NOUVEAU_NO_TRANSFER", TRUE/*XXX:FALSE*/)) &&
-	    pt->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)
-	{
-		tx->direct = true;
-		tx->surface = pscreen->get_tex_surface(pscreen, pt,
-	                                               face, level, zslice,
-	                                               pipe_transfer_buffer_flags(&tx->base));
-		return &tx->base;
-	}
-
-	tx->direct = false;
-
-	nv40_compatible_transfer_tex(pt, w, h, &tx_tex_template);
-
-	tx_tex = pscreen->texture_create(pscreen, &tx_tex_template);
-	if (!tx_tex)
-	{
-		FREE(tx);
-		return NULL;
-	}
-
-	tx->base.stride = ((struct nv40_miptree*)tx_tex)->level[0].pitch;
-
-	tx->surface = pscreen->get_tex_surface(pscreen, tx_tex,
-	                                       0, 0, 0,
-	                                       pipe_transfer_buffer_flags(&tx->base));
-
-	pipe_texture_reference(&tx_tex, NULL);
-
-	if (!tx->surface)
-	{
-		pipe_surface_reference(&tx->surface, NULL);
-		FREE(tx);
-		return NULL;
-	}
-
-	if (usage & PIPE_TRANSFER_READ) {
-		struct nv40_screen *nvscreen = nv40_screen(pscreen);
-		struct pipe_surface *src;
-
-		src = pscreen->get_tex_surface(pscreen, pt,
-	                                       face, level, zslice,
-	                                       PIPE_BUFFER_USAGE_GPU_READ);
-
-		/* TODO: Check if SIFM can deal with x,y,w,h when swizzling */
-		/* TODO: Check if SIFM can un-swizzle */
-		nvscreen->eng2d->copy(nvscreen->eng2d,
-		                      tx->surface, 0, 0,
-		                      src, x, y,
-		                      w, h);
-
-		pipe_surface_reference(&src, NULL);
-	}
-
-	return &tx->base;
-}
-
-static void
-nv40_transfer_del(struct pipe_context *pcontext, struct pipe_transfer *ptx)
-{
-	struct nv40_transfer *tx = (struct nv40_transfer *)ptx;
-
-	if (!tx->direct && (ptx->usage & PIPE_TRANSFER_WRITE)) {
-		struct pipe_screen *pscreen = pcontext->screen;
-		struct nv40_screen *nvscreen = nv40_screen(pscreen);
-		struct pipe_surface *dst;
-
-		dst = pscreen->get_tex_surface(pscreen, ptx->texture,
-	                                       ptx->face, ptx->level, ptx->zslice,
-	                                       PIPE_BUFFER_USAGE_GPU_WRITE | NOUVEAU_BUFFER_USAGE_NO_RENDER);
-
-		/* TODO: Check if SIFM can deal with x,y,w,h when swizzling */
-		nvscreen->eng2d->copy(nvscreen->eng2d,
-		                      dst, tx->base.x, tx->base.y,
-		                      tx->surface, 0, 0,
-		                      tx->base.width, tx->base.height);
-
-		pipe_surface_reference(&dst, NULL);
-	}
-
-	pipe_surface_reference(&tx->surface, NULL);
-	pipe_texture_reference(&ptx->texture, NULL);
-	FREE(ptx);
-}
-
-static void *
-nv40_transfer_map(struct pipe_context *pcontext, struct pipe_transfer *ptx)
-{
-        struct pipe_screen *pscreen = pcontext->screen;
-	struct nv40_transfer *tx = (struct nv40_transfer *)ptx;
-	struct nv04_surface *ns = (struct nv04_surface *)tx->surface;
-	struct nv40_miptree *mt = (struct nv40_miptree *)tx->surface->texture;
-	void *map = pipe_buffer_map(pscreen, mt->buffer,
-	                            pipe_transfer_buffer_flags(ptx));
-
-	if(!tx->direct)
-		return map + ns->base.offset;
-	else
-		return map + ns->base.offset + ptx->y * ns->pitch + ptx->x * util_format_get_blocksize(ptx->texture->format);
-}
-
-static void
-nv40_transfer_unmap(struct pipe_context *pcontext, struct pipe_transfer *ptx)
-{
-        struct pipe_screen *pscreen = pcontext->screen;
-	struct nv40_transfer *tx = (struct nv40_transfer *)ptx;
-	struct nv40_miptree *mt = (struct nv40_miptree *)tx->surface->texture;
-
-	pipe_buffer_unmap(pscreen, mt->buffer);
-}
-
-void
-nv40_init_transfer_functions(struct nv40_context *nv40)
-{
-	nv40->pipe.get_tex_transfer = nv40_transfer_new;
-	nv40->pipe.tex_transfer_destroy = nv40_transfer_del;
-	nv40->pipe.transfer_map = nv40_transfer_map;
-	nv40->pipe.transfer_unmap = nv40_transfer_unmap;
-}
diff --git a/src/gallium/drivers/nv40/nv40_vbo.c b/src/gallium/drivers/nv40/nv40_vbo.c
deleted file mode 100644
index fabdf4bf23..0000000000
--- a/src/gallium/drivers/nv40/nv40_vbo.c
+++ /dev/null
@@ -1,565 +0,0 @@
-#include "pipe/p_context.h"
-#include "pipe/p_state.h"
-#include "util/u_inlines.h"
-#include "util/u_format.h"
-
-#include "nv40_context.h"
-#include "nv40_state.h"
-
-#include "nouveau/nouveau_channel.h"
-#include "nouveau/nouveau_pushbuf.h"
-#include "nouveau/nouveau_util.h"
-
-#define FORCE_SWTNL 0
-
-static INLINE int
-nv40_vbo_format_to_hw(enum pipe_format pipe, unsigned *fmt, unsigned *ncomp)
-{
-	switch (pipe) {
-	case PIPE_FORMAT_R32_FLOAT:
-	case PIPE_FORMAT_R32G32_FLOAT:
-	case PIPE_FORMAT_R32G32B32_FLOAT:
-	case PIPE_FORMAT_R32G32B32A32_FLOAT:
-		*fmt = NV40TCL_VTXFMT_TYPE_FLOAT;
-		break;
-	case PIPE_FORMAT_R8_UNORM:
-	case PIPE_FORMAT_R8G8_UNORM:
-	case PIPE_FORMAT_R8G8B8_UNORM:
-	case PIPE_FORMAT_R8G8B8A8_UNORM:
-		*fmt = NV40TCL_VTXFMT_TYPE_UBYTE;
-		break;
-	case PIPE_FORMAT_R16_SSCALED:
-	case PIPE_FORMAT_R16G16_SSCALED:
-	case PIPE_FORMAT_R16G16B16_SSCALED:
-	case PIPE_FORMAT_R16G16B16A16_SSCALED:
-		*fmt = NV40TCL_VTXFMT_TYPE_USHORT;
-		break;
-	default:
-		NOUVEAU_ERR("Unknown format %s\n", util_format_name(pipe));
-		return 1;
-	}
-
-	switch (pipe) {
-	case PIPE_FORMAT_R8_UNORM:
-	case PIPE_FORMAT_R32_FLOAT:
-	case PIPE_FORMAT_R16_SSCALED:
-		*ncomp = 1;
-		break;
-	case PIPE_FORMAT_R8G8_UNORM:
-	case PIPE_FORMAT_R32G32_FLOAT:
-	case PIPE_FORMAT_R16G16_SSCALED:
-		*ncomp = 2;
-		break;
-	case PIPE_FORMAT_R8G8B8_UNORM:
-	case PIPE_FORMAT_R32G32B32_FLOAT:
-	case PIPE_FORMAT_R16G16B16_SSCALED:
-		*ncomp = 3;
-		break;
-	case PIPE_FORMAT_R8G8B8A8_UNORM:
-	case PIPE_FORMAT_R32G32B32A32_FLOAT:
-	case PIPE_FORMAT_R16G16B16A16_SSCALED:
-		*ncomp = 4;
-		break;
-	default:
-		NOUVEAU_ERR("Unknown format %s\n", util_format_name(pipe));
-		return 1;
-	}
-
-	return 0;
-}
-
-static boolean
-nv40_vbo_set_idxbuf(struct nv40_context *nv40, struct pipe_buffer *ib,
-		    unsigned ib_size)
-{
-	struct pipe_screen *pscreen = &nv40->screen->base.base;
-	unsigned type;
-
-	if (!ib) {
-		nv40->idxbuf = NULL;
-		nv40->idxbuf_format = 0xdeadbeef;
-		return FALSE;
-	}
-
-	if (!pscreen->get_param(pscreen, NOUVEAU_CAP_HW_IDXBUF) || ib_size == 1)
-		return FALSE;
-
-	switch (ib_size) {
-	case 2:
-		type = NV40TCL_IDXBUF_FORMAT_TYPE_U16;
-		break;
-	case 4:
-		type = NV40TCL_IDXBUF_FORMAT_TYPE_U32;
-		break;
-	default:
-		return FALSE;
-	}
-
-	if (ib != nv40->idxbuf ||
-	    type != nv40->idxbuf_format) {
-		nv40->dirty |= NV40_NEW_ARRAYS;
-		nv40->idxbuf = ib;
-		nv40->idxbuf_format = type;
-	}
-
-	return TRUE;
-}
-
-static boolean
-nv40_vbo_static_attrib(struct nv40_context *nv40, struct nouveau_stateobj *so,
-		       int attrib, struct pipe_vertex_element *ve,
-		       struct pipe_vertex_buffer *vb)
-{
-	struct pipe_screen *pscreen = nv40->pipe.screen;
-	struct nouveau_grobj *curie = nv40->screen->curie;
-	unsigned type, ncomp;
-	void *map;
-
-	if (nv40_vbo_format_to_hw(ve->src_format, &type, &ncomp))
-		return FALSE;
-
-	map  = pipe_buffer_map(pscreen, vb->buffer, PIPE_BUFFER_USAGE_CPU_READ);
-	map += vb->buffer_offset + ve->src_offset;
-
-	switch (type) {
-	case NV40TCL_VTXFMT_TYPE_FLOAT:
-	{
-		float *v = map;
-
-		switch (ncomp) {
-		case 4:
-			so_method(so, curie, NV40TCL_VTX_ATTR_4F_X(attrib), 4);
-			so_data  (so, fui(v[0]));
-			so_data  (so, fui(v[1]));
-			so_data  (so, fui(v[2]));
-			so_data  (so, fui(v[3]));
-			break;
-		case 3:
-			so_method(so, curie, NV40TCL_VTX_ATTR_3F_X(attrib), 3);
-			so_data  (so, fui(v[0]));
-			so_data  (so, fui(v[1]));
-			so_data  (so, fui(v[2]));
-			break;
-		case 2:
-			so_method(so, curie, NV40TCL_VTX_ATTR_2F_X(attrib), 2);
-			so_data  (so, fui(v[0]));
-			so_data  (so, fui(v[1]));
-			break;
-		case 1:
-			so_method(so, curie, NV40TCL_VTX_ATTR_1F(attrib), 1);
-			so_data  (so, fui(v[0]));
-			break;
-		default:
-			pipe_buffer_unmap(pscreen, vb->buffer);
-			return FALSE;
-		}
-	}
-		break;
-	default:
-		pipe_buffer_unmap(pscreen, vb->buffer);
-		return FALSE;
-	}
-
-	pipe_buffer_unmap(pscreen, vb->buffer);
-
-	return TRUE;
-}
-
-void
-nv40_draw_arrays(struct pipe_context *pipe,
-		 unsigned mode, unsigned start, unsigned count)
-{
-	struct nv40_context *nv40 = nv40_context(pipe);
-	struct nv40_screen *screen = nv40->screen;
-	struct nouveau_channel *chan = screen->base.channel;
-	struct nouveau_grobj *curie = screen->curie;
-	unsigned restart;
-
-	nv40_vbo_set_idxbuf(nv40, NULL, 0);
-	if (FORCE_SWTNL || !nv40_state_validate(nv40)) {
-		nv40_draw_elements_swtnl(pipe, NULL, 0,
-                                         mode, start, count);
-                return;
-	}
-
-	while (count) {
-		unsigned vc, nr;
-
-		nv40_state_emit(nv40);
-
-		vc = nouveau_vbuf_split(AVAIL_RING(chan), 6, 256,
-					mode, start, count, &restart);
-		if (!vc) {
-			FIRE_RING(chan);
-			continue;
-		}
-
-		BEGIN_RING(chan, curie, NV40TCL_BEGIN_END, 1);
-		OUT_RING  (chan, nvgl_primitive(mode));
-
-		nr = (vc & 0xff);
-		if (nr) {
-			BEGIN_RING(chan, curie, NV40TCL_VB_VERTEX_BATCH, 1);
-			OUT_RING  (chan, ((nr - 1) << 24) | start);
-			start += nr;
-		}
-
-		nr = vc >> 8;
-		while (nr) {
-			unsigned push = nr > 2047 ? 2047 : nr;
-
-			nr -= push;
-
-			BEGIN_RING_NI(chan, curie, NV40TCL_VB_VERTEX_BATCH, push);
-			while (push--) {
-				OUT_RING(chan, ((0x100 - 1) << 24) | start);
-				start += 0x100;
-			}
-		}
-
-		BEGIN_RING(chan, curie, NV40TCL_BEGIN_END, 1);
-		OUT_RING  (chan, 0);
-
-		count -= vc;
-		start = restart;
-	}
-
-	pipe->flush(pipe, 0, NULL);
-}
-
-static INLINE void
-nv40_draw_elements_u08(struct nv40_context *nv40, void *ib,
-		       unsigned mode, unsigned start, unsigned count)
-{
-	struct nv40_screen *screen = nv40->screen;
-	struct nouveau_channel *chan = screen->base.channel;
-	struct nouveau_grobj *curie = screen->curie;
-
-	while (count) {
-		uint8_t *elts = (uint8_t *)ib + start;
-		unsigned vc, push, restart;
-
-		nv40_state_emit(nv40);
-
-		vc = nouveau_vbuf_split(AVAIL_RING(chan), 6, 2,
-					mode, start, count, &restart);
-		if (vc == 0) {
-			FIRE_RING(chan);
-			continue;
-		}
-		count -= vc;
-
-		BEGIN_RING(chan, curie, NV40TCL_BEGIN_END, 1);
-		OUT_RING  (chan, nvgl_primitive(mode));
-
-		if (vc & 1) {
-			BEGIN_RING(chan, curie, NV40TCL_VB_ELEMENT_U32, 1);
-			OUT_RING  (chan, elts[0]);
-			elts++; vc--;
-		}
-
-		while (vc) {
-			unsigned i;
-
-			push = MIN2(vc, 2047 * 2);
-
-			BEGIN_RING_NI(chan, curie, NV40TCL_VB_ELEMENT_U16, push >> 1);
-			for (i = 0; i < push; i+=2)
-				OUT_RING(chan, (elts[i+1] << 16) | elts[i]);
-
-			vc -= push;
-			elts += push;
-		}
-
-		BEGIN_RING(chan, curie, NV40TCL_BEGIN_END, 1);
-		OUT_RING  (chan, 0);
-
-		start = restart;
-	}
-}
-
-static INLINE void
-nv40_draw_elements_u16(struct nv40_context *nv40, void *ib,
-		       unsigned mode, unsigned start, unsigned count)
-{
-	struct nv40_screen *screen = nv40->screen;
-	struct nouveau_channel *chan = screen->base.channel;
-	struct nouveau_grobj *curie = screen->curie;
-
-	while (count) {
-		uint16_t *elts = (uint16_t *)ib + start;
-		unsigned vc, push, restart;
-
-		nv40_state_emit(nv40);
-
-		vc = nouveau_vbuf_split(AVAIL_RING(chan), 6, 2,
-					mode, start, count, &restart);
-		if (vc == 0) {
-			FIRE_RING(chan);
-			continue;
-		}
-		count -= vc;
-
-		BEGIN_RING(chan, curie, NV40TCL_BEGIN_END, 1);
-		OUT_RING  (chan, nvgl_primitive(mode));
-
-		if (vc & 1) {
-			BEGIN_RING(chan, curie, NV40TCL_VB_ELEMENT_U32, 1);
-			OUT_RING  (chan, elts[0]);
-			elts++; vc--;
-		}
-
-		while (vc) {
-			unsigned i;
-
-			push = MIN2(vc, 2047 * 2);
-
-			BEGIN_RING_NI(chan, curie, NV40TCL_VB_ELEMENT_U16, push >> 1);
-			for (i = 0; i < push; i+=2)
-				OUT_RING(chan, (elts[i+1] << 16) | elts[i]);
-
-			vc -= push;
-			elts += push;
-		}
-
-		BEGIN_RING(chan, curie, NV40TCL_BEGIN_END, 1);
-		OUT_RING  (chan, 0);
-
-		start = restart;
-	}
-}
-
-static INLINE void
-nv40_draw_elements_u32(struct nv40_context *nv40, void *ib,
-		       unsigned mode, unsigned start, unsigned count)
-{
-	struct nv40_screen *screen = nv40->screen;
-	struct nouveau_channel *chan = screen->base.channel;
-	struct nouveau_grobj *curie = screen->curie;
-
-	while (count) {
-		uint32_t *elts = (uint32_t *)ib + start;
-		unsigned vc, push, restart;
-
-		nv40_state_emit(nv40);
-
-		vc = nouveau_vbuf_split(AVAIL_RING(chan), 5, 1,
-					mode, start, count, &restart);
-		if (vc == 0) {
-			FIRE_RING(chan);
-			continue;
-		}
-		count -= vc;
-
-		BEGIN_RING(chan, curie, NV40TCL_BEGIN_END, 1);
-		OUT_RING  (chan, nvgl_primitive(mode));
-
-		while (vc) {
-			push = MIN2(vc, 2047);
-
-			BEGIN_RING_NI(chan, curie, NV40TCL_VB_ELEMENT_U32, push);
-			OUT_RINGp    (chan, elts, push);
-
-			vc -= push;
-			elts += push;
-		}
-
-		BEGIN_RING(chan, curie, NV40TCL_BEGIN_END, 1);
-		OUT_RING  (chan, 0);
-
-		start = restart;
-	}
-}
-
-static void
-nv40_draw_elements_inline(struct pipe_context *pipe,
-			  struct pipe_buffer *ib, unsigned ib_size,
-			  unsigned mode, unsigned start, unsigned count)
-{
-	struct nv40_context *nv40 = nv40_context(pipe);
-	struct pipe_screen *pscreen = pipe->screen;
-	void *map;
-
-	map = pipe_buffer_map(pscreen, ib, PIPE_BUFFER_USAGE_CPU_READ);
-	if (!ib) {
-		NOUVEAU_ERR("failed mapping ib\n");
-		return;
-	}
-
-	switch (ib_size) {
-	case 1:
-		nv40_draw_elements_u08(nv40, map, mode, start, count);
-		break;
-	case 2:
-		nv40_draw_elements_u16(nv40, map, mode, start, count);
-		break;
-	case 4:
-		nv40_draw_elements_u32(nv40, map, mode, start, count);
-		break;
-	default:
-		NOUVEAU_ERR("invalid idxbuf fmt %d\n", ib_size);
-		break;
-	}
-
-	pipe_buffer_unmap(pscreen, ib);
-}
-
-static void
-nv40_draw_elements_vbo(struct pipe_context *pipe,
-		       unsigned mode, unsigned start, unsigned count)
-{
-	struct nv40_context *nv40 = nv40_context(pipe);
-	struct nv40_screen *screen = nv40->screen;
-	struct nouveau_channel *chan = screen->base.channel;
-	struct nouveau_grobj *curie = screen->curie;
-	unsigned restart;
-
-	while (count) {
-		unsigned nr, vc;
-
-		nv40_state_emit(nv40);
-
-		vc = nouveau_vbuf_split(AVAIL_RING(chan), 6, 256,
-					mode, start, count, &restart);
-		if (!vc) {
-			FIRE_RING(chan);
-			continue;
-		}
-
-		BEGIN_RING(chan, curie, NV40TCL_BEGIN_END, 1);
-		OUT_RING  (chan, nvgl_primitive(mode));
-
-		nr = (vc & 0xff);
-		if (nr) {
-			BEGIN_RING(chan, curie, NV40TCL_VB_INDEX_BATCH, 1);
-			OUT_RING  (chan, ((nr - 1) << 24) | start);
-			start += nr;
-		}
-
-		nr = vc >> 8;
-		while (nr) {
-			unsigned push = nr > 2047 ? 2047 : nr;
-
-			nr -= push;
-
-			BEGIN_RING_NI(chan, curie, NV40TCL_VB_INDEX_BATCH, push);
-			while (push--) {
-				OUT_RING(chan, ((0x100 - 1) << 24) | start);
-				start += 0x100;
-			}
-		}
-
-		BEGIN_RING(chan, curie, NV40TCL_BEGIN_END, 1);
-		OUT_RING  (chan, 0);
-
-		count -= vc;
-		start = restart;
-	}
-}
-
-void
-nv40_draw_elements(struct pipe_context *pipe,
-		   struct pipe_buffer *indexBuffer, unsigned indexSize,
-		   unsigned mode, unsigned start, unsigned count)
-{
-	struct nv40_context *nv40 = nv40_context(pipe);
-	boolean idxbuf;
-
-	idxbuf = nv40_vbo_set_idxbuf(nv40, indexBuffer, indexSize);
-	if (FORCE_SWTNL || !nv40_state_validate(nv40)) {
-		nv40_draw_elements_swtnl(pipe, NULL, 0,
-                                         mode, start, count);
-                return;
-	}
-
-	if (idxbuf) {
-		nv40_draw_elements_vbo(pipe, mode, start, count);
-	} else {
-		nv40_draw_elements_inline(pipe, indexBuffer, indexSize,
-					  mode, start, count);
-	}
-
-	pipe->flush(pipe, 0, NULL);
-}
-
-static boolean
-nv40_vbo_validate(struct nv40_context *nv40)
-{
-	struct nouveau_stateobj *vtxbuf, *vtxfmt, *sattr = NULL;
-	struct nouveau_grobj *curie = nv40->screen->curie;
-	struct pipe_buffer *ib = nv40->idxbuf;
-	unsigned ib_format = nv40->idxbuf_format;
-	unsigned vb_flags = NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD;
-	int hw;
-
-	vtxbuf = so_new(3, 17, 18);
-	so_method(vtxbuf, curie, NV40TCL_VTXBUF_ADDRESS(0), nv40->vtxelt->num_elements);
-	vtxfmt = so_new(1, 16, 0);
-	so_method(vtxfmt, curie, NV40TCL_VTXFMT(0), nv40->vtxelt->num_elements);
-
-	for (hw = 0; hw < nv40->vtxelt->num_elements; hw++) {
-		struct pipe_vertex_element *ve;
-		struct pipe_vertex_buffer *vb;
-		unsigned type, ncomp;
-
-		ve = &nv40->vtxelt->pipe[hw];
-		vb = &nv40->vtxbuf[ve->vertex_buffer_index];
-
-		if (!vb->stride) {
-			if (!sattr)
-				sattr = so_new(16, 16 * 4, 0);
-
-			if (nv40_vbo_static_attrib(nv40, sattr, hw, ve, vb)) {
-				so_data(vtxbuf, 0);
-				so_data(vtxfmt, NV40TCL_VTXFMT_TYPE_FLOAT);
-				continue;
-			}
-		}
-
-		if (nv40_vbo_format_to_hw(ve->src_format, &type, &ncomp)) {
-			nv40->fallback_swtnl |= NV40_NEW_ARRAYS;
-			so_ref(NULL, &vtxbuf);
-			so_ref(NULL, &vtxfmt);
-			return FALSE;
-		}
-
-		so_reloc(vtxbuf, nouveau_bo(vb->buffer),
-				 vb->buffer_offset + ve->src_offset,
-				 vb_flags | NOUVEAU_BO_LOW | NOUVEAU_BO_OR,
-				 0, NV40TCL_VTXBUF_ADDRESS_DMA1);
-		so_data (vtxfmt, ((vb->stride << NV40TCL_VTXFMT_STRIDE_SHIFT) |
-				  (ncomp << NV40TCL_VTXFMT_SIZE_SHIFT) | type));
-	}
-
-	if (ib) {
-		struct nouveau_bo *bo = nouveau_bo(ib);
-
-		so_method(vtxbuf, curie, NV40TCL_IDXBUF_ADDRESS, 2);
-		so_reloc (vtxbuf, bo, 0, vb_flags | NOUVEAU_BO_LOW, 0, 0);
-		so_reloc (vtxbuf, bo, ib_format, vb_flags | NOUVEAU_BO_OR,
-			  0, NV40TCL_IDXBUF_FORMAT_DMA1);
-	}
-
-	so_method(vtxbuf, curie, 0x1710, 1);
-	so_data  (vtxbuf, 0);
-
-	so_ref(vtxbuf, &nv40->state.hw[NV40_STATE_VTXBUF]);
-	so_ref(NULL, &vtxbuf);
-	nv40->state.dirty |= (1ULL << NV40_STATE_VTXBUF);
-	so_ref(vtxfmt, &nv40->state.hw[NV40_STATE_VTXFMT]);
-	so_ref(NULL, &vtxfmt);
-	nv40->state.dirty |= (1ULL << NV40_STATE_VTXFMT);
-	so_ref(sattr, &nv40->state.hw[NV40_STATE_VTXATTR]);
-	so_ref(NULL, &sattr);
-	nv40->state.dirty |= (1ULL << NV40_STATE_VTXATTR);
-	return FALSE;
-}
-
-struct nv40_state_entry nv40_state_vbo = {
-	.validate = nv40_vbo_validate,
-	.dirty = {
-		.pipe = NV40_NEW_ARRAYS,
-		.hw = 0,
-	}
-};
-
diff --git a/src/gallium/drivers/nv50/nv50_vbo.c b/src/gallium/drivers/nv50/nv50_vbo.c
index 6b9c1ee231..5047286806 100644
--- a/src/gallium/drivers/nv50/nv50_vbo.c
+++ b/src/gallium/drivers/nv50/nv50_vbo.c
@@ -549,6 +549,7 @@ nv50_vbo_validate(struct nv50_context *nv50)
 	if (nv50->vtxbuf_nr == 0)
 		return NULL;
 
+	nv50->vbo_fifo = 0;
 	if (nv50->screen->force_push ||
 	    nv50->vertprog->cfg.edgeflag_in < 16)
 		nv50->vbo_fifo = 0xffff;
diff --git a/src/gallium/drivers/nvfx/Makefile b/src/gallium/drivers/nvfx/Makefile
new file mode 100644
index 0000000000..dfe97e6ed5
--- /dev/null
+++ b/src/gallium/drivers/nvfx/Makefile
@@ -0,0 +1,32 @@
+TOP = ../../../..
+include $(TOP)/configs/current
+
+LIBNAME = nvfx
+
+C_SOURCES = \
+	nv04_surface_2d.c \
+	nvfx_context.c \
+	nvfx_clear.c \
+	nvfx_draw.c \
+	nvfx_fragprog.c \
+	nvfx_fragtex.c \
+	nv30_fragtex.c \
+	nv40_fragtex.c \
+	nvfx_miptree.c \
+	nvfx_query.c \
+	nvfx_screen.c \
+	nvfx_state.c \
+	nvfx_state_blend.c \
+        nvfx_state_emit.c \
+	nvfx_state_fb.c \
+	nvfx_state_rasterizer.c \
+	nvfx_state_scissor.c \
+        nvfx_state_stipple.c \
+	nvfx_state_viewport.c \
+	nvfx_state_zsa.c \
+	nvfx_surface.c \
+	nvfx_transfer.c \
+	nvfx_vbo.c \
+	nvfx_vertprog.c
+
+include ../../Makefile.template
diff --git a/src/gallium/drivers/nouveau/nv04_surface_2d.c b/src/gallium/drivers/nvfx/nv04_surface_2d.c
index 93114465d5..ed18c9f24d 100644
--- a/src/gallium/drivers/nouveau/nv04_surface_2d.c
+++ b/src/gallium/drivers/nvfx/nv04_surface_2d.c
@@ -543,4 +543,3 @@ nv04_surface_wrap_for_render(struct pipe_screen *pscreen, struct nv04_surface_2d
 
 	return temp_ns;
 }
-
diff --git a/src/gallium/drivers/nouveau/nv04_surface_2d.h b/src/gallium/drivers/nvfx/nv04_surface_2d.h
index ce696a11a3..ce696a11a3 100644
--- a/src/gallium/drivers/nouveau/nv04_surface_2d.h
+++ b/src/gallium/drivers/nvfx/nv04_surface_2d.h
diff --git a/src/gallium/drivers/nv30/nv30_fragtex.c b/src/gallium/drivers/nvfx/nv30_fragtex.c
index f7d98f3f20..2b56f45492 100644
--- a/src/gallium/drivers/nv30/nv30_fragtex.c
+++ b/src/gallium/drivers/nvfx/nv30_fragtex.c
@@ -1,7 +1,37 @@
 #include "util/u_format.h"
 
-#include "nv30_context.h"
+#include "nvfx_context.h"
 #include "nouveau/nouveau_util.h"
+#include "nvfx_tex.h"
+
+void
+nv30_sampler_state_init(struct pipe_context *pipe,
+			  struct nvfx_sampler_state *ps,
+			  const struct pipe_sampler_state *cso)
+{
+	if (cso->max_anisotropy >= 8) {
+		ps->en |= NV34TCL_TX_ENABLE_ANISO_8X;
+	} else
+	if (cso->max_anisotropy >= 4) {
+		ps->en |= NV34TCL_TX_ENABLE_ANISO_4X;
+	} else
+	if (cso->max_anisotropy >= 2) {
+		ps->en |= NV34TCL_TX_ENABLE_ANISO_2X;
+	}
+
+	{
+		float limit;
+
+		limit = CLAMP(cso->lod_bias, -16.0, 15.0);
+		ps->filt |= (int)(cso->lod_bias * 256.0) & 0x1fff;
+
+		limit = CLAMP(cso->max_lod, 0.0, 15.0);
+		ps->en |= (int)(limit) << 14 /*NV34TCL_TX_ENABLE_MIPMAP_MAX_LOD_SHIFT*/;
+
+		limit = CLAMP(cso->min_lod, 0.0, 15.0);
+		ps->en |= (int)(limit) << 26 /*NV34TCL_TX_ENABLE_MIPMAP_MIN_LOD_SHIFT*/;
+	}
+}
 
 #define _(m,tf,ts0x,ts0y,ts0z,ts0w,ts1x,ts1y,ts1z,ts1w)                        \
 {                                                                              \
@@ -57,11 +87,11 @@ nv30_fragtex_format(uint pipe_format)
 }
 
 
-static struct nouveau_stateobj *
-nv30_fragtex_build(struct nv30_context *nv30, int unit)
+struct nouveau_stateobj *
+nv30_fragtex_build(struct nvfx_context *nvfx, int unit)
 {
-	struct nv30_sampler_state *ps = nv30->tex_sampler[unit];
-	struct nv30_miptree *nv30mt = nv30->tex_miptree[unit];
+	struct nvfx_sampler_state *ps = nvfx->tex_sampler[unit];
+	struct nvfx_miptree *nv30mt = nvfx->tex_miptree[unit];
 	struct pipe_texture *pt = &nv30mt->base;
 	struct nouveau_bo *bo = nouveau_bo(nv30mt->buffer);
 	struct nv30_texture_format *tf;
@@ -101,7 +131,7 @@ nv30_fragtex_build(struct nv30_context *nv30, int unit)
 	txs = tf->swizzle;
 
 	so = so_new(1, 8, 2);
-	so_method(so, nv30->screen->rankine, NV34TCL_TX_OFFSET(unit), 8);
+	so_method(so, nvfx->screen->eng3d, NV34TCL_TX_OFFSET(unit), 8);
 	so_reloc (so, bo, 0, tex_flags | NOUVEAU_BO_LOW, 0, 0);
 	so_reloc (so, bo, txf, tex_flags | NOUVEAU_BO_OR,
 		      NV34TCL_TX_FORMAT_DMA0, NV34TCL_TX_FORMAT_DMA1);
@@ -115,47 +145,3 @@ nv30_fragtex_build(struct nv30_context *nv30, int unit)
 
 	return so;
 }
-
-static boolean
-nv30_fragtex_validate(struct nv30_context *nv30)
-{
-	struct nv30_fragment_program *fp = nv30->fragprog;
-	struct nv30_state *state = &nv30->state;
-	struct nouveau_stateobj *so;
-	unsigned samplers, unit;
-
-	samplers = state->fp_samplers & ~fp->samplers;
-	while (samplers) {
-		unit = ffs(samplers) - 1;
-		samplers &= ~(1 << unit);
-
-		so = so_new(1, 1, 0);
-		so_method(so, nv30->screen->rankine, NV34TCL_TX_ENABLE(unit), 1);
-		so_data  (so, 0);
-		so_ref(so, &nv30->state.hw[NV30_STATE_FRAGTEX0 + unit]);
-		so_ref(NULL, &so);
-		state->dirty |= (1ULL << (NV30_STATE_FRAGTEX0 + unit));
-	}
-
-	samplers = nv30->dirty_samplers & fp->samplers;
-	while (samplers) {
-		unit = ffs(samplers) - 1;
-		samplers &= ~(1 << unit);
-
-		so = nv30_fragtex_build(nv30, unit);
-		so_ref(so, &nv30->state.hw[NV30_STATE_FRAGTEX0 + unit]);
-		so_ref(NULL, &so);
-		state->dirty |= (1ULL << (NV30_STATE_FRAGTEX0 + unit));
-	}
-
-	nv30->state.fp_samplers = fp->samplers;
-	return FALSE;
-}
-
-struct nv30_state_entry nv30_state_fragtex = {
-	.validate = nv30_fragtex_validate,
-	.dirty = {
-		.pipe = NV30_NEW_SAMPLER | NV30_NEW_FRAGPROG,
-		.hw = 0
-	}
-};
diff --git a/src/gallium/drivers/nvfx/nv30_vertprog.h b/src/gallium/drivers/nvfx/nv30_vertprog.h
new file mode 100644
index 0000000000..ec0444c07f
--- /dev/null
+++ b/src/gallium/drivers/nvfx/nv30_vertprog.h
@@ -0,0 +1,169 @@
+#ifndef __NV30_SHADER_H__
+#define __NV30_SHADER_H__
+
+/* Vertex programs instruction set
+ *
+ * 128bit opcodes, split into 4 32-bit ones for ease of use.
+ *
+ * Non-native instructions
+ *   ABS - MOV + NV40_VP_INST0_DEST_ABS
+ *   POW - EX2 + MUL + LG2
+ *   SUB - ADD, second source negated
+ *   SWZ - MOV
+ *   XPD -
+ *
+ * Register access
+ *   - Only one INPUT can be accessed per-instruction (move extras into TEMPs)
+ *   - Only one CONST can be accessed per-instruction (move extras into TEMPs)
+ *
+ * Relative Addressing
+ *   According to the value returned for
+ *   MAX_PROGRAM_NATIVE_ADDRESS_REGISTERS_ARB
+ *
+ *   there are only two address registers available.  The destination in the
+ *   ARL instruction is set to TEMP <n> (The temp isn't actually written).
+ *
+ *   When using vanilla ARB_v_p, the proprietary driver will squish both the
+ *   available ADDRESS regs into the first hardware reg in the X and Y
+ *   components.
+ *
+ *   To use an address reg as an index into consts, the CONST_SRC is set to
+ *   (const_base + offset) and INDEX_CONST is set.
+ *
+ *   To access the second address reg use ADDR_REG_SELECT_1. A particular
+ *   component of the address regs is selected with ADDR_SWZ.
+ *
+ *   Only one address register can be accessed per instruction.
+ *
+ * Conditional execution (see NV_vertex_program{2,3} for details) Conditional
+ * execution of an instruction is enabled by setting COND_TEST_ENABLE, and
+ * selecting the condition which will allow the test to pass with
+ * COND_{FL,LT,...}.  It is possible to swizzle the values in the condition
+ * register, which allows for testing against an individual component.
+ *
+ * Branching:
+ *
+ *   The BRA/CAL instructions seem to follow a slightly different opcode
+ *   layout.  The destination instruction ID (IADDR) overlaps a source field.
+ *   Instruction ID's seem to be numbered based on the UPLOAD_FROM_ID FIFO
+ *   command, and is incremented automatically on each UPLOAD_INST FIFO
+ *   command.
+ *
+ *   Conditional branching is achieved by using the condition tests described
+ *   above.  There doesn't appear to be dedicated looping instructions, but
+ *   this can be done using a temp reg + conditional branching.
+ *
+ *   Subroutines may be uploaded before the main program itself, but the first
+ *   executed instruction is determined by the PROGRAM_START_ID FIFO command.
+ *
+ */
+
+/* DWORD 0 */
+
+#define NV30_VP_INST_ADDR_REG_SELECT_1        (1 << 24)
+#define NV30_VP_INST_SRC2_ABS           (1 << 23) /* guess */
+#define NV30_VP_INST_SRC1_ABS           (1 << 22) /* guess */
+#define NV30_VP_INST_SRC0_ABS           (1 << 21) /* guess */
+#define NV30_VP_INST_VEC_RESULT         (1 << 20)
+#define NV30_VP_INST_DEST_TEMP_ID_SHIFT        16
+#define NV30_VP_INST_DEST_TEMP_ID_MASK        (0x0F << 16)
+#define NV30_VP_INST_COND_UPDATE_ENABLE        (1<<15)
+#define NV30_VP_INST_VEC_DEST_TEMP_MASK      (0xF << 16)
+#define NV30_VP_INST_COND_TEST_ENABLE        (1<<14)
+#define NV30_VP_INST_COND_SHIFT          11
+#define NV30_VP_INST_COND_MASK          (0x07 << 11)
+#define NV30_VP_INST_COND_SWZ_X_SHIFT        9
+#define NV30_VP_INST_COND_SWZ_X_MASK        (0x03 <<  9)
+#define NV30_VP_INST_COND_SWZ_Y_SHIFT        7
+#define NV30_VP_INST_COND_SWZ_Y_MASK        (0x03 <<  7)
+#define NV30_VP_INST_COND_SWZ_Z_SHIFT        5
+#define NV30_VP_INST_COND_SWZ_Z_MASK        (0x03 <<  5)
+#define NV30_VP_INST_COND_SWZ_W_SHIFT        3
+#define NV30_VP_INST_COND_SWZ_W_MASK        (0x03 <<  3)
+#define NV30_VP_INST_COND_SWZ_ALL_SHIFT        3
+#define NV30_VP_INST_COND_SWZ_ALL_MASK        (0xFF <<  3)
+#define NV30_VP_INST_ADDR_SWZ_SHIFT        1
+#define NV30_VP_INST_ADDR_SWZ_MASK        (0x03 <<  1)
+#define NV30_VP_INST_SCA_OPCODEH_SHIFT        0
+#define NV30_VP_INST_SCA_OPCODEH_MASK        (0x01 <<  0)
+
+/* DWORD 1 */
+#define NV30_VP_INST_SCA_OPCODEL_SHIFT        28
+#define NV30_VP_INST_SCA_OPCODEL_MASK        (0x0F << 28)
+#define NV30_VP_INST_VEC_OPCODE_SHIFT        23
+#define NV30_VP_INST_VEC_OPCODE_MASK        (0x1F << 23)
+#define NV30_VP_INST_CONST_SRC_SHIFT        14
+#define NV30_VP_INST_CONST_SRC_MASK        (0xFF << 14)
+#define NV30_VP_INST_INPUT_SRC_SHIFT        9    /*NV20*/
+#define NV30_VP_INST_INPUT_SRC_MASK        (0x0F <<  9)  /*NV20*/
+#define NV30_VP_INST_SRC0H_SHIFT        0    /*NV20*/
+#define NV30_VP_INST_SRC0H_MASK          (0x1FF << 0)  /*NV20*/
+
+/* Please note: the IADDR fields overlap other fields because they are used
+ * only for branch instructions.  See Branching: label above
+ *
+ * DWORD 2
+ */
+#define NV30_VP_INST_SRC0L_SHIFT        26    /*NV20*/
+#define NV30_VP_INST_SRC0L_MASK         (0x3F  <<26)  /* NV30_VP_SRC0_LOW_MASK << 26 */
+#define NV30_VP_INST_SRC1_SHIFT         11    /*NV20*/
+#define NV30_VP_INST_SRC1_MASK          (0x7FFF<<11)  /*NV20*/
+#define NV30_VP_INST_SRC2H_SHIFT        0    /*NV20*/
+#define NV30_VP_INST_SRC2H_MASK          (0x7FF << 0)  /* NV30_VP_SRC2_HIGH_MASK >> 4*/
+#define NV30_VP_INST_IADDR_SHIFT        2
+#define NV30_VP_INST_IADDR_MASK          (0xF <<  28)   /* NV30_VP_SRC2_LOW_MASK << 28 */
+
+/* DWORD 3 */
+#define NV30_VP_INST_SRC2L_SHIFT        28    /*NV20*/
+#define NV30_VP_INST_SRC2L_MASK          (0x0F  <<28)  /*NV20*/
+#define NV30_VP_INST_STEMP_WRITEMASK_SHIFT      24
+#define NV30_VP_INST_STEMP_WRITEMASK_MASK      (0x0F << 24)
+#define NV30_VP_INST_VTEMP_WRITEMASK_SHIFT      20
+#define NV30_VP_INST_VTEMP_WRITEMASK_MASK      (0x0F << 20)
+#define NV30_VP_INST_SDEST_WRITEMASK_SHIFT      16
+#define NV30_VP_INST_SDEST_WRITEMASK_MASK      (0x0F << 16)
+#define NV30_VP_INST_VDEST_WRITEMASK_SHIFT      12    /*NV20*/
+#define NV30_VP_INST_VDEST_WRITEMASK_MASK      (0x0F << 12)  /*NV20*/
+#define NV30_VP_INST_DEST_SHIFT        2
+#define NV30_VP_INST_DEST_MASK        (0x0F <<  2)
+#  define NV30_VP_INST_DEST_POS  0
+#  define NV30_VP_INST_DEST_BFC0  1
+#  define NV30_VP_INST_DEST_BFC1  2
+#  define NV30_VP_INST_DEST_COL0  3
+#  define NV30_VP_INST_DEST_COL1  4
+#  define NV30_VP_INST_DEST_FOGC  5
+#  define NV30_VP_INST_DEST_PSZ   6
+#  define NV30_VP_INST_DEST_TC(n)  (8+n)
+
+/* Useful to split the source selection regs into their pieces */
+#define NV30_VP_SRC0_HIGH_SHIFT                                                6
+#define NV30_VP_SRC0_HIGH_MASK                                        0x00007FC0
+#define NV30_VP_SRC0_LOW_MASK                                         0x0000003F
+#define NV30_VP_SRC2_HIGH_SHIFT                                                4
+#define NV30_VP_SRC2_HIGH_MASK                                        0x00007FF0
+#define NV30_VP_SRC2_LOW_MASK                                         0x0000000F
+
+
+/* Source-register definition - matches NV20 exactly */
+#define NV30_VP_SRC_NEGATE          (1<<14)
+#define NV30_VP_SRC_SWZ_X_SHIFT        12
+#define NV30_VP_SRC_REG_SWZ_X_MASK        (0x03  <<12)
+#define NV30_VP_SRC_SWZ_Y_SHIFT        10
+#define NV30_VP_SRC_REG_SWZ_Y_MASK        (0x03  <<10)
+#define NV30_VP_SRC_SWZ_Z_SHIFT        8
+#define NV30_VP_SRC_REG_SWZ_Z_MASK        (0x03  << 8)
+#define NV30_VP_SRC_SWZ_W_SHIFT        6
+#define NV30_VP_SRC_REG_SWZ_W_MASK        (0x03  << 6)
+#define NV30_VP_SRC_REG_SWZ_ALL_SHIFT        6
+#define NV30_VP_SRC_REG_SWZ_ALL_MASK        (0xFF  << 6)
+#define NV30_VP_SRC_TEMP_SRC_SHIFT        2
+#define NV30_VP_SRC_REG_TEMP_ID_MASK        (0x0F  << 0)
+#define NV30_VP_SRC_REG_TYPE_SHIFT        0
+#define NV30_VP_SRC_REG_TYPE_MASK        (0x03  << 0)
+#define NV30_VP_SRC_REG_TYPE_TEMP  1
+#define NV30_VP_SRC_REG_TYPE_INPUT  2
+#define NV30_VP_SRC_REG_TYPE_CONST  3 /* guess */
+
+#include "nvfx_shader.h"
+
+#endif
diff --git a/src/gallium/drivers/nv40/nv40_fragtex.c b/src/gallium/drivers/nvfx/nv40_fragtex.c
index b60118922a..5889b5e40d 100644
--- a/src/gallium/drivers/nv40/nv40_fragtex.c
+++ b/src/gallium/drivers/nvfx/nv40_fragtex.c
@@ -1,18 +1,63 @@
 #include "util/u_format.h"
+#include "nvfx_context.h"
+#include "nvfx_tex.h"
 
-#include "nv40_context.h"
+void
+nv40_sampler_state_init(struct pipe_context *pipe,
+			  struct nvfx_sampler_state *ps,
+			  const struct pipe_sampler_state *cso)
+{
+	if (cso->max_anisotropy >= 2) {
+		/* no idea, binary driver sets it, works without it.. meh.. */
+		ps->wrap |= (1 << 5);
+
+		if (cso->max_anisotropy >= 16) {
+			ps->en |= NV40TCL_TEX_ENABLE_ANISO_16X;
+		} else
+		if (cso->max_anisotropy >= 12) {
+			ps->en |= NV40TCL_TEX_ENABLE_ANISO_12X;
+		} else
+		if (cso->max_anisotropy >= 10) {
+			ps->en |= NV40TCL_TEX_ENABLE_ANISO_10X;
+		} else
+		if (cso->max_anisotropy >= 8) {
+			ps->en |= NV40TCL_TEX_ENABLE_ANISO_8X;
+		} else
+		if (cso->max_anisotropy >= 6) {
+			ps->en |= NV40TCL_TEX_ENABLE_ANISO_6X;
+		} else
+		if (cso->max_anisotropy >= 4) {
+			ps->en |= NV40TCL_TEX_ENABLE_ANISO_4X;
+		} else {
+			ps->en |= NV40TCL_TEX_ENABLE_ANISO_2X;
+		}
+	}
+
+	{
+		float limit;
+
+		limit = CLAMP(cso->lod_bias, -16.0, 15.0);
+		ps->filt |= (int)(cso->lod_bias * 256.0) & 0x1fff;
+
+		limit = CLAMP(cso->max_lod, 0.0, 15.0);
+		ps->en |= (int)(limit * 256.0) << 7;
+
+		limit = CLAMP(cso->min_lod, 0.0, 15.0);
+		ps->en |= (int)(limit * 256.0) << 19;
+	}
+}
 
 #define _(m,tf,ts0x,ts0y,ts0z,ts0w,ts1x,ts1y,ts1z,ts1w,sx,sy,sz,sw)            \
 {                                                                              \
   TRUE,                                                                        \
   PIPE_FORMAT_##m,                                                             \
   NV40TCL_TEX_FORMAT_FORMAT_##tf,                                              \
-  (NV40TCL_TEX_SWIZZLE_S0_X_##ts0x | NV40TCL_TEX_SWIZZLE_S0_Y_##ts0y |         \
-   NV40TCL_TEX_SWIZZLE_S0_Z_##ts0z | NV40TCL_TEX_SWIZZLE_S0_W_##ts0w |         \
-   NV40TCL_TEX_SWIZZLE_S1_X_##ts1x | NV40TCL_TEX_SWIZZLE_S1_Y_##ts1y |         \
-   NV40TCL_TEX_SWIZZLE_S1_Z_##ts1z | NV40TCL_TEX_SWIZZLE_S1_W_##ts1w),         \
-  ((NV40TCL_TEX_FILTER_SIGNED_RED*sx) | (NV40TCL_TEX_FILTER_SIGNED_GREEN*sy) |       \
-   (NV40TCL_TEX_FILTER_SIGNED_BLUE*sz) | (NV40TCL_TEX_FILTER_SIGNED_ALPHA*sw))       \
+  (NV34TCL_TX_SWIZZLE_S0_X_##ts0x | NV34TCL_TX_SWIZZLE_S0_Y_##ts0y |         \
+   NV34TCL_TX_SWIZZLE_S0_Z_##ts0z | NV34TCL_TX_SWIZZLE_S0_W_##ts0w |         \
+   NV34TCL_TX_SWIZZLE_S1_X_##ts1x | NV34TCL_TX_SWIZZLE_S1_Y_##ts1y |         \
+   NV34TCL_TX_SWIZZLE_S1_Z_##ts1z | NV34TCL_TX_SWIZZLE_S1_W_##ts1w),         \
+  ((NV34TCL_TX_FILTER_SIGNED_RED*sx) | (NV34TCL_TX_FILTER_SIGNED_GREEN*sy) |       \
+   (NV34TCL_TX_FILTER_SIGNED_BLUE*sz) | (NV34TCL_TX_FILTER_SIGNED_ALPHA*sw))       \
 }
 
 struct nv40_texture_format {
@@ -60,11 +105,11 @@ nv40_fragtex_format(uint pipe_format)
 }
 
 
-static struct nouveau_stateobj *
-nv40_fragtex_build(struct nv40_context *nv40, int unit)
+struct nouveau_stateobj *
+nv40_fragtex_build(struct nvfx_context *nvfx, int unit)
 {
-	struct nv40_sampler_state *ps = nv40->tex_sampler[unit];
-	struct nv40_miptree *nv40mt = nv40->tex_miptree[unit];
+	struct nvfx_sampler_state *ps = nvfx->tex_sampler[unit];
+	struct nvfx_miptree *nv40mt = nvfx->tex_miptree[unit];
 	struct nouveau_bo *bo = nouveau_bo(nv40mt->buffer);
 	struct pipe_texture *pt = &nv40mt->base;
 	struct nv40_texture_format *tf;
@@ -81,20 +126,20 @@ nv40_fragtex_build(struct nv40_context *nv40, int unit)
 	txf |= ((pt->last_level + 1) << NV40TCL_TEX_FORMAT_MIPMAP_COUNT_SHIFT);
 
 	if (1) /* XXX */
-		txf |= NV40TCL_TEX_FORMAT_NO_BORDER;
+		txf |= NV34TCL_TX_FORMAT_NO_BORDER;
 
 	switch (pt->target) {
 	case PIPE_TEXTURE_CUBE:
-		txf |= NV40TCL_TEX_FORMAT_CUBIC;
+		txf |= NV34TCL_TX_FORMAT_CUBIC;
 		/* fall-through */
 	case PIPE_TEXTURE_2D:
-		txf |= NV40TCL_TEX_FORMAT_DIMS_2D;
+		txf |= NV34TCL_TX_FORMAT_DIMS_2D;
 		break;
 	case PIPE_TEXTURE_3D:
-		txf |= NV40TCL_TEX_FORMAT_DIMS_3D;
+		txf |= NV34TCL_TX_FORMAT_DIMS_3D;
 		break;
 	case PIPE_TEXTURE_1D:
-		txf |= NV40TCL_TEX_FORMAT_DIMS_1D;
+		txf |= NV34TCL_TX_FORMAT_DIMS_1D;
 		break;
 	default:
 		NOUVEAU_ERR("Unknown target %d\n", pt->target);
@@ -111,63 +156,19 @@ nv40_fragtex_build(struct nv40_context *nv40, int unit)
 	txs = tf->swizzle;
 
 	so = so_new(2, 9, 2);
-	so_method(so, nv40->screen->curie, NV40TCL_TEX_OFFSET(unit), 8);
+	so_method(so, nvfx->screen->eng3d, NV34TCL_TX_OFFSET(unit), 8);
 	so_reloc (so, bo, 0, tex_flags | NOUVEAU_BO_LOW, 0, 0);
 	so_reloc (so, bo, txf, tex_flags | NOUVEAU_BO_OR,
-		      NV40TCL_TEX_FORMAT_DMA0, NV40TCL_TEX_FORMAT_DMA1);
+		      NV34TCL_TX_FORMAT_DMA0, NV34TCL_TX_FORMAT_DMA1);
 	so_data  (so, ps->wrap);
 	so_data  (so, NV40TCL_TEX_ENABLE_ENABLE | ps->en);
 	so_data  (so, txs);
 	so_data  (so, ps->filt | tf->sign | 0x2000 /*voodoo*/);
-	so_data  (so, (pt->width0 << NV40TCL_TEX_SIZE0_W_SHIFT) |
+	so_data  (so, (pt->width0 << NV34TCL_TX_NPOT_SIZE_W_SHIFT) |
 		       pt->height0);
 	so_data  (so, ps->bcol);
-	so_method(so, nv40->screen->curie, NV40TCL_TEX_SIZE1(unit), 1);
+	so_method(so, nvfx->screen->eng3d, NV40TCL_TEX_SIZE1(unit), 1);
 	so_data  (so, (pt->depth0 << NV40TCL_TEX_SIZE1_DEPTH_SHIFT) | txp);
 
 	return so;
 }
-
-static boolean
-nv40_fragtex_validate(struct nv40_context *nv40)
-{
-	struct nv40_fragment_program *fp = nv40->fragprog;
-	struct nv40_state *state = &nv40->state;
-	struct nouveau_stateobj *so;
-	unsigned samplers, unit;
-
-	samplers = state->fp_samplers & ~fp->samplers;
-	while (samplers) {
-		unit = ffs(samplers) - 1;
-		samplers &= ~(1 << unit);
-
-		so = so_new(1, 1, 0);
-		so_method(so, nv40->screen->curie, NV40TCL_TEX_ENABLE(unit), 1);
-		so_data  (so, 0);
-		so_ref(so, &nv40->state.hw[NV40_STATE_FRAGTEX0 + unit]);
-		state->dirty |= (1ULL << (NV40_STATE_FRAGTEX0 + unit));
-	}
-
-	samplers = nv40->dirty_samplers & fp->samplers;
-	while (samplers) {
-		unit = ffs(samplers) - 1;
-		samplers &= ~(1 << unit);
-
-		so = nv40_fragtex_build(nv40, unit);
-		so_ref(so, &nv40->state.hw[NV40_STATE_FRAGTEX0 + unit]);
-		so_ref(NULL, &so);
-		state->dirty |= (1ULL << (NV40_STATE_FRAGTEX0 + unit));
-	}
-
-	nv40->state.fp_samplers = fp->samplers;
-	return FALSE;
-}
-
-struct nv40_state_entry nv40_state_fragtex = {
-	.validate = nv40_fragtex_validate,
-	.dirty = {
-		.pipe = NV40_NEW_SAMPLER | NV40_NEW_FRAGPROG,
-		.hw = 0
-	}
-};
-
diff --git a/src/gallium/drivers/nvfx/nv40_vertprog.h b/src/gallium/drivers/nvfx/nv40_vertprog.h
new file mode 100644
index 0000000000..7337293bab
--- /dev/null
+++ b/src/gallium/drivers/nvfx/nv40_vertprog.h
@@ -0,0 +1,177 @@
+#ifndef __NV40_SHADER_H__
+#define __NV40_SHADER_H__
+
+/* Vertex programs instruction set
+ *
+ * The NV40 instruction set is very similar to NV30.  Most fields are in
+ * a slightly different position in the instruction however.
+ *
+ * Merged instructions
+ *     In some cases it is possible to put two instructions into one opcode
+ *     slot.  The rules for when this is OK is not entirely clear to me yet.
+ *
+ *     There are separate writemasks and dest temp register fields for each
+ *     grouping of instructions.  There is however only one field with the
+ *     ID of a result register.  Writing to temp/result regs is selected by
+ *     setting VEC_RESULT/SCA_RESULT.
+ *
+ * Temporary registers
+ *     The source/dest temp register fields have been extended by 1 bit, to
+ *     give a total of 32 temporary registers.
+ *
+ * Relative Addressing
+ *     NV40 can use an address register to index into vertex attribute regs.
+ *     This is done by putting the offset value into INPUT_SRC and setting
+ *     the INDEX_INPUT flag.
+ *
+ * Conditional execution (see NV_vertex_program{2,3} for details)
+ *     There is a second condition code register on NV40, it's use is enabled
+ *     by setting the COND_REG_SELECT_1 flag.
+ *
+ * Texture lookup
+ *     TODO
+ */
+
+/* ---- OPCODE BITS 127:96 / data DWORD 0 --- */
+#define NV40_VP_INST_VEC_RESULT                                        (1 << 30)
+/* uncertain.. */
+#define NV40_VP_INST_COND_UPDATE_ENABLE                        ((1 << 14)|1<<29)
+/* use address reg as index into attribs */
+#define NV40_VP_INST_INDEX_INPUT                                       (1 << 27)
+#define NV40_VP_INST_COND_REG_SELECT_1                                 (1 << 25)
+#define NV40_VP_INST_ADDR_REG_SELECT_1                                 (1 << 24)
+#define NV40_VP_INST_SRC2_ABS                                          (1 << 23)
+#define NV40_VP_INST_SRC1_ABS                                          (1 << 22)
+#define NV40_VP_INST_SRC0_ABS                                          (1 << 21)
+#define NV40_VP_INST_VEC_DEST_TEMP_SHIFT                                      15
+#define NV40_VP_INST_VEC_DEST_TEMP_MASK                             (0x1F << 15)
+#define NV40_VP_INST_COND_TEST_ENABLE                                  (1 << 13)
+#define NV40_VP_INST_COND_SHIFT                                               10
+#define NV40_VP_INST_COND_MASK                                       (0x7 << 10)
+#define NV40_VP_INST_COND_SWZ_X_SHIFT                                          8
+#define NV40_VP_INST_COND_SWZ_X_MASK                                    (3 << 8)
+#define NV40_VP_INST_COND_SWZ_Y_SHIFT                                          6
+#define NV40_VP_INST_COND_SWZ_Y_MASK                                    (3 << 6)
+#define NV40_VP_INST_COND_SWZ_Z_SHIFT                                          4
+#define NV40_VP_INST_COND_SWZ_Z_MASK                                    (3 << 4)
+#define NV40_VP_INST_COND_SWZ_W_SHIFT                                          2
+#define NV40_VP_INST_COND_SWZ_W_MASK                                    (3 << 2)
+#define NV40_VP_INST_COND_SWZ_ALL_SHIFT                                        2
+#define NV40_VP_INST_COND_SWZ_ALL_MASK                               (0xFF << 2)
+#define NV40_VP_INST_ADDR_SWZ_SHIFT                                            0
+#define NV40_VP_INST_ADDR_SWZ_MASK                                   (0x03 << 0)
+#define NV40_VP_INST0_KNOWN ( \
+                NV40_VP_INST_INDEX_INPUT | \
+                NV40_VP_INST_COND_REG_SELECT_1 | \
+                NV40_VP_INST_ADDR_REG_SELECT_1 | \
+                NV40_VP_INST_SRC2_ABS | \
+                NV40_VP_INST_SRC1_ABS | \
+                NV40_VP_INST_SRC0_ABS | \
+                NV40_VP_INST_VEC_DEST_TEMP_MASK | \
+                NV40_VP_INST_COND_TEST_ENABLE | \
+                NV40_VP_INST_COND_MASK | \
+                NV40_VP_INST_COND_SWZ_ALL_MASK | \
+                NV40_VP_INST_ADDR_SWZ_MASK)
+
+/* ---- OPCODE BITS 95:64 / data DWORD 1 --- */
+#define NV40_VP_INST_VEC_OPCODE_SHIFT                                         22
+#define NV40_VP_INST_VEC_OPCODE_MASK                                (0x1F << 22)
+#define NV40_VP_INST_SCA_OPCODE_SHIFT                                         27
+#define NV40_VP_INST_SCA_OPCODE_MASK                                (0x1F << 27)
+#define NV40_VP_INST_CONST_SRC_SHIFT                                          12
+#define NV40_VP_INST_CONST_SRC_MASK                                 (0xFF << 12)
+#define NV40_VP_INST_INPUT_SRC_SHIFT                                           8
+#define NV40_VP_INST_INPUT_SRC_MASK                                  (0x0F << 8)
+#define NV40_VP_INST_SRC0H_SHIFT                                               0
+#define NV40_VP_INST_SRC0H_MASK                                      (0xFF << 0)
+#define NV40_VP_INST1_KNOWN ( \
+                NV40_VP_INST_VEC_OPCODE_MASK | \
+                NV40_VP_INST_SCA_OPCODE_MASK | \
+                NV40_VP_INST_CONST_SRC_MASK  | \
+                NV40_VP_INST_INPUT_SRC_MASK  | \
+                NV40_VP_INST_SRC0H_MASK \
+                )
+
+/* ---- OPCODE BITS 63:32 / data DWORD 2 --- */
+#define NV40_VP_INST_SRC0L_SHIFT                                              23
+#define NV40_VP_INST_SRC0L_MASK                                    (0x1FF << 23)
+#define NV40_VP_INST_SRC1_SHIFT                                                6
+#define NV40_VP_INST_SRC1_MASK                                    (0x1FFFF << 6)
+#define NV40_VP_INST_SRC2H_SHIFT                                               0
+#define NV40_VP_INST_SRC2H_MASK                                      (0x3F << 0)
+#define NV40_VP_INST_IADDRH_SHIFT                                              0
+#define NV40_VP_INST_IADDRH_MASK                                     (0x1F << 0)
+
+/* ---- OPCODE BITS 31:0 / data DWORD 3 --- */
+#define NV40_VP_INST_IADDRL_SHIFT                                             29
+#define NV40_VP_INST_IADDRL_MASK                                       (7 << 29)
+#define NV40_VP_INST_SRC2L_SHIFT                                              21
+#define NV40_VP_INST_SRC2L_MASK                                    (0x7FF << 21)
+#define NV40_VP_INST_SCA_WRITEMASK_SHIFT                                      17
+#define NV40_VP_INST_SCA_WRITEMASK_MASK                              (0xF << 17)
+#    define NV40_VP_INST_SCA_WRITEMASK_X                               (1 << 20)
+#    define NV40_VP_INST_SCA_WRITEMASK_Y                               (1 << 19)
+#    define NV40_VP_INST_SCA_WRITEMASK_Z                               (1 << 18)
+#    define NV40_VP_INST_SCA_WRITEMASK_W                               (1 << 17)
+#define NV40_VP_INST_VEC_WRITEMASK_SHIFT                                      13
+#define NV40_VP_INST_VEC_WRITEMASK_MASK                              (0xF << 13)
+#    define NV40_VP_INST_VEC_WRITEMASK_X                               (1 << 16)
+#    define NV40_VP_INST_VEC_WRITEMASK_Y                               (1 << 15)
+#    define NV40_VP_INST_VEC_WRITEMASK_Z                               (1 << 14)
+#    define NV40_VP_INST_VEC_WRITEMASK_W                               (1 << 13)
+#define NV40_VP_INST_SCA_RESULT                                        (1 << 12)
+#define NV40_VP_INST_SCA_DEST_TEMP_SHIFT                                       7
+#define NV40_VP_INST_SCA_DEST_TEMP_MASK                              (0x1F << 7)
+#define NV40_VP_INST_DEST_SHIFT                                                2
+#define NV40_VP_INST_DEST_MASK                                         (31 << 2)
+#    define NV40_VP_INST_DEST_POS                                              0
+#    define NV40_VP_INST_DEST_COL0                                             1
+#    define NV40_VP_INST_DEST_COL1                                             2
+#    define NV40_VP_INST_DEST_BFC0                                             3
+#    define NV40_VP_INST_DEST_BFC1                                             4
+#    define NV40_VP_INST_DEST_FOGC                                             5
+#    define NV40_VP_INST_DEST_PSZ                                              6
+#    define NV40_VP_INST_DEST_TC0                                              7
+#    define NV40_VP_INST_DEST_TC(n)                                        (7+n)
+#    define NV40_VP_INST_DEST_TEMP                                          0x1F
+#define NV40_VP_INST_INDEX_CONST                                        (1 << 1)
+#define NV40_VP_INST3_KNOWN ( \
+                NV40_VP_INST_SRC2L_MASK |\
+                NV40_VP_INST_SCA_WRITEMASK_MASK |\
+                NV40_VP_INST_VEC_WRITEMASK_MASK |\
+                NV40_VP_INST_SCA_DEST_TEMP_MASK |\
+                NV40_VP_INST_DEST_MASK |\
+                NV40_VP_INST_INDEX_CONST)
+
+/* Useful to split the source selection regs into their pieces */
+#define NV40_VP_SRC0_HIGH_SHIFT                                                9
+#define NV40_VP_SRC0_HIGH_MASK                                        0x0001FE00
+#define NV40_VP_SRC0_LOW_MASK                                         0x000001FF
+#define NV40_VP_SRC2_HIGH_SHIFT                                               11
+#define NV40_VP_SRC2_HIGH_MASK                                        0x0001F800
+#define NV40_VP_SRC2_LOW_MASK                                         0x000007FF
+
+/* Source selection - these are the bits you fill NV40_VP_INST_SRCn with */
+#define NV40_VP_SRC_NEGATE                                             (1 << 16)
+#define NV40_VP_SRC_SWZ_X_SHIFT                                               14
+#define NV40_VP_SRC_SWZ_X_MASK                                         (3 << 14)
+#define NV40_VP_SRC_SWZ_Y_SHIFT                                               12
+#define NV40_VP_SRC_SWZ_Y_MASK                                         (3 << 12)
+#define NV40_VP_SRC_SWZ_Z_SHIFT                                               10
+#define NV40_VP_SRC_SWZ_Z_MASK                                         (3 << 10)
+#define NV40_VP_SRC_SWZ_W_SHIFT                                                8
+#define NV40_VP_SRC_SWZ_W_MASK                                          (3 << 8)
+#define NV40_VP_SRC_SWZ_ALL_SHIFT                                              8
+#define NV40_VP_SRC_SWZ_ALL_MASK                                     (0xFF << 8)
+#define NV40_VP_SRC_TEMP_SRC_SHIFT                                             2
+#define NV40_VP_SRC_TEMP_SRC_MASK                                    (0x1F << 2)
+#define NV40_VP_SRC_REG_TYPE_SHIFT                                             0
+#define NV40_VP_SRC_REG_TYPE_MASK                                       (3 << 0)
+#    define NV40_VP_SRC_REG_TYPE_UNK0                                          0
+#    define NV40_VP_SRC_REG_TYPE_TEMP                                          1
+#    define NV40_VP_SRC_REG_TYPE_INPUT                                         2
+#    define NV40_VP_SRC_REG_TYPE_CONST                                         3
+
+#include "nvfx_shader.h"
+
+#endif
diff --git a/src/gallium/drivers/nv40/nv40_clear.c b/src/gallium/drivers/nvfx/nvfx_clear.c
index ddf13addf3..2be70fcee4 100644
--- a/src/gallium/drivers/nv40/nv40_clear.c
+++ b/src/gallium/drivers/nvfx/nvfx_clear.c
@@ -3,12 +3,12 @@
 #include "pipe/p_state.h"
 #include "util/u_clear.h"
 
-#include "nv40_context.h"
+#include "nvfx_context.h"
 
 void
-nv40_clear(struct pipe_context *pipe, unsigned buffers,
+nvfx_clear(struct pipe_context *pipe, unsigned buffers,
            const float *rgba, double depth, unsigned stencil)
 {
-	util_clear(pipe, &nv40_context(pipe)->framebuffer, buffers, rgba, depth,
+	util_clear(pipe, &nvfx_context(pipe)->framebuffer, buffers, rgba, depth,
 		   stencil);
 }
diff --git a/src/gallium/drivers/nvfx/nvfx_context.c b/src/gallium/drivers/nvfx/nvfx_context.c
new file mode 100644
index 0000000000..fc3cbdb558
--- /dev/null
+++ b/src/gallium/drivers/nvfx/nvfx_context.c
@@ -0,0 +1,90 @@
+#include "draw/draw_context.h"
+#include "pipe/p_defines.h"
+
+#include "nvfx_context.h"
+#include "nvfx_screen.h"
+
+static void
+nvfx_flush(struct pipe_context *pipe, unsigned flags,
+	   struct pipe_fence_handle **fence)
+{
+	struct nvfx_context *nvfx = nvfx_context(pipe);
+	struct nvfx_screen *screen = nvfx->screen;
+	struct nouveau_channel *chan = screen->base.channel;
+	struct nouveau_grobj *eng3d = screen->eng3d;
+
+	if (flags & PIPE_FLUSH_TEXTURE_CACHE) {
+		BEGIN_RING(chan, eng3d, 0x1fd8, 1);
+		OUT_RING  (chan, 2);
+		BEGIN_RING(chan, eng3d, 0x1fd8, 1);
+		OUT_RING  (chan, 1);
+	}
+
+	FIRE_RING(chan);
+	if (fence)
+		*fence = NULL;
+}
+
+static void
+nvfx_destroy(struct pipe_context *pipe)
+{
+	struct nvfx_context *nvfx = nvfx_context(pipe);
+	unsigned i;
+
+	for (i = 0; i < NVFX_STATE_MAX; i++) {
+		if (nvfx->state.hw[i])
+			so_ref(NULL, &nvfx->state.hw[i]);
+	}
+
+	if (nvfx->draw)
+		draw_destroy(nvfx->draw);
+	FREE(nvfx);
+}
+
+struct pipe_context *
+nvfx_create(struct pipe_screen *pscreen, void *priv)
+{
+	struct nvfx_screen *screen = nvfx_screen(pscreen);
+	struct pipe_winsys *ws = pscreen->winsys;
+	struct nvfx_context *nvfx;
+	struct nouveau_winsys *nvws = screen->nvws;
+
+	nvfx = CALLOC(1, sizeof(struct nvfx_context));
+	if (!nvfx)
+		return NULL;
+	nvfx->screen = screen;
+
+	nvfx->nvws = nvws;
+
+	nvfx->pipe.winsys = ws;
+	nvfx->pipe.screen = pscreen;
+	nvfx->pipe.priv = priv;
+	nvfx->pipe.destroy = nvfx_destroy;
+	nvfx->pipe.draw_arrays = nvfx_draw_arrays;
+	nvfx->pipe.draw_elements = nvfx_draw_elements;
+	nvfx->pipe.clear = nvfx_clear;
+	nvfx->pipe.flush = nvfx_flush;
+
+	nvfx->pipe.is_texture_referenced = nouveau_is_texture_referenced;
+	nvfx->pipe.is_buffer_referenced = nouveau_is_buffer_referenced;
+
+	screen->base.channel->user_private = nvfx;
+	screen->base.channel->flush_notify = nvfx_state_flush_notify;
+
+	nvfx->is_nv4x = screen->is_nv4x;
+
+	nvfx_init_query_functions(nvfx);
+	nvfx_init_surface_functions(nvfx);
+	nvfx_init_state_functions(nvfx);
+	nvfx_init_transfer_functions(nvfx);
+
+	/* Create, configure, and install fallback swtnl path */
+	nvfx->draw = draw_create();
+	draw_wide_point_threshold(nvfx->draw, 9999999.0);
+	draw_wide_line_threshold(nvfx->draw, 9999999.0);
+	draw_enable_line_stipple(nvfx->draw, FALSE);
+	draw_enable_point_sprites(nvfx->draw, FALSE);
+	draw_set_rasterize_stage(nvfx->draw, nvfx_draw_render_stage(nvfx));
+
+	return &nvfx->pipe;
+}
diff --git a/src/gallium/drivers/nvfx/nvfx_context.h b/src/gallium/drivers/nvfx/nvfx_context.h
new file mode 100644
index 0000000000..001b19eedf
--- /dev/null
+++ b/src/gallium/drivers/nvfx/nvfx_context.h
@@ -0,0 +1,265 @@
+#ifndef __NVFX_CONTEXT_H__
+#define __NVFX_CONTEXT_H__
+
+#include <stdio.h>
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_state.h"
+#include "pipe/p_compiler.h"
+
+#include "util/u_memory.h"
+#include "util/u_math.h"
+#include "util/u_inlines.h"
+
+#include "draw/draw_vertex.h"
+
+#include "nouveau/nouveau_winsys.h"
+#include "nouveau/nouveau_gldefs.h"
+#include "nouveau/nouveau_context.h"
+#include "nouveau/nouveau_stateobj.h"
+
+#include "nvfx_state.h"
+
+#define NOUVEAU_ERR(fmt, args...) \
+	fprintf(stderr, "%s:%d -  "fmt, __func__, __LINE__, ##args);
+#define NOUVEAU_MSG(fmt, args...) \
+	fprintf(stderr, "nouveau: "fmt, ##args);
+
+enum nvfx_state_index {
+	NVFX_STATE_FB = 0,
+	NVFX_STATE_VIEWPORT = 1,
+	NVFX_STATE_BLEND = 2,
+	NVFX_STATE_RAST = 3,
+	NVFX_STATE_ZSA = 4,
+	NVFX_STATE_BCOL = 5,
+	NVFX_STATE_CLIP = 6,
+	NVFX_STATE_SCISSOR = 7,
+	NVFX_STATE_STIPPLE = 8,
+	NVFX_STATE_FRAGPROG = 9,
+	NVFX_STATE_VERTPROG = 10,
+	NVFX_STATE_FRAGTEX0 = 11,
+	NVFX_STATE_FRAGTEX1 = 12,
+	NVFX_STATE_FRAGTEX2 = 13,
+	NVFX_STATE_FRAGTEX3 = 14,
+	NVFX_STATE_FRAGTEX4 = 15,
+	NVFX_STATE_FRAGTEX5 = 16,
+	NVFX_STATE_FRAGTEX6 = 17,
+	NVFX_STATE_FRAGTEX7 = 18,
+	NVFX_STATE_FRAGTEX8 = 19,
+	NVFX_STATE_FRAGTEX9 = 20,
+	NVFX_STATE_FRAGTEX10 = 21,
+	NVFX_STATE_FRAGTEX11 = 22,
+	NVFX_STATE_FRAGTEX12 = 23,
+	NVFX_STATE_FRAGTEX13 = 24,
+	NVFX_STATE_FRAGTEX14 = 25,
+	NVFX_STATE_FRAGTEX15 = 26,
+	NVFX_STATE_VERTTEX0 = 27,
+	NVFX_STATE_VERTTEX1 = 28,
+	NVFX_STATE_VERTTEX2 = 29,
+	NVFX_STATE_VERTTEX3 = 30,
+	NVFX_STATE_VTXBUF = 31,
+	NVFX_STATE_VTXFMT = 32,
+	NVFX_STATE_VTXATTR = 33,
+	NVFX_STATE_SR = 34,
+	NVFX_STATE_MAX = 35
+};
+
+#include "nvfx_screen.h"
+
+#define NVFX_NEW_BLEND		(1 <<  0)
+#define NVFX_NEW_RAST		(1 <<  1)
+#define NVFX_NEW_ZSA		(1 <<  2)
+#define NVFX_NEW_SAMPLER	(1 <<  3)
+#define NVFX_NEW_FB		(1 <<  4)
+#define NVFX_NEW_STIPPLE	(1 <<  5)
+#define NVFX_NEW_SCISSOR	(1 <<  6)
+#define NVFX_NEW_VIEWPORT	(1 <<  7)
+#define NVFX_NEW_BCOL		(1 <<  8)
+#define NVFX_NEW_VERTPROG	(1 <<  9)
+#define NVFX_NEW_FRAGPROG	(1 << 10)
+#define NVFX_NEW_ARRAYS		(1 << 11)
+#define NVFX_NEW_UCP		(1 << 12)
+#define NVFX_NEW_SR		(1 << 13)
+
+struct nvfx_rasterizer_state {
+	struct pipe_rasterizer_state pipe;
+	struct nouveau_stateobj *so;
+};
+
+struct nvfx_zsa_state {
+	struct pipe_depth_stencil_alpha_state pipe;
+	struct nouveau_stateobj *so;
+};
+
+struct nvfx_blend_state {
+	struct pipe_blend_state pipe;
+	struct nouveau_stateobj *so;
+};
+
+
+struct nvfx_state {
+	unsigned scissor_enabled;
+	unsigned stipple_enabled;
+	unsigned fp_samplers;
+
+	uint64_t dirty;
+	struct nouveau_stateobj *hw[NVFX_STATE_MAX];
+};
+
+struct nvfx_vtxelt_state {
+	struct pipe_vertex_element pipe[16];
+	unsigned num_elements;
+};
+
+struct nvfx_context {
+	struct pipe_context pipe;
+
+	struct nouveau_winsys *nvws;
+	struct nvfx_screen *screen;
+
+	unsigned is_nv4x; /* either 0 or ~0 */
+
+	struct draw_context *draw;
+
+	/* HW state derived from pipe states */
+	struct nvfx_state state;
+	struct {
+		struct nvfx_vertex_program *vertprog;
+
+		unsigned nr_attribs;
+		unsigned hw[PIPE_MAX_SHADER_INPUTS];
+		unsigned draw[PIPE_MAX_SHADER_INPUTS];
+		unsigned emit[PIPE_MAX_SHADER_INPUTS];
+	} swtnl;
+
+	enum {
+		HW, SWTNL, SWRAST
+	} render_mode;
+	unsigned fallback_swtnl;
+	unsigned fallback_swrast;
+
+	/* Context state */
+	unsigned dirty, draw_dirty;
+	struct pipe_scissor_state scissor;
+	unsigned stipple[32];
+	struct pipe_clip_state clip;
+	struct nvfx_vertex_program *vertprog;
+	struct nvfx_fragment_program *fragprog;
+	struct pipe_buffer *constbuf[PIPE_SHADER_TYPES];
+	unsigned constbuf_nr[PIPE_SHADER_TYPES];
+	struct nvfx_rasterizer_state *rasterizer;
+	struct nvfx_zsa_state *zsa;
+	struct nvfx_blend_state *blend;
+	struct pipe_blend_color blend_colour;
+	struct pipe_stencil_ref stencil_ref;
+	struct pipe_viewport_state viewport;
+	struct pipe_framebuffer_state framebuffer;
+	struct pipe_buffer *idxbuf;
+	unsigned idxbuf_format;
+	struct nvfx_sampler_state *tex_sampler[PIPE_MAX_SAMPLERS];
+	struct nvfx_miptree *tex_miptree[PIPE_MAX_SAMPLERS];
+	struct pipe_sampler_view *fragment_sampler_views[PIPE_MAX_SAMPLERS];
+	unsigned nr_samplers;
+	unsigned nr_textures;
+	unsigned dirty_samplers;
+	struct pipe_vertex_buffer vtxbuf[PIPE_MAX_ATTRIBS];
+	unsigned vtxbuf_nr;
+	struct nvfx_vtxelt_state *vtxelt;
+};
+
+static INLINE struct nvfx_context *
+nvfx_context(struct pipe_context *pipe)
+{
+	return (struct nvfx_context *)pipe;
+}
+
+struct nvfx_state_entry {
+	boolean (*validate)(struct nvfx_context *nvfx);
+	struct {
+		unsigned pipe;
+		unsigned hw;
+	} dirty;
+};
+
+extern struct nvfx_state_entry nvfx_state_blend;
+extern struct nvfx_state_entry nvfx_state_blend_colour;
+extern struct nvfx_state_entry nvfx_state_fragprog;
+extern struct nvfx_state_entry nvfx_state_fragtex;
+extern struct nvfx_state_entry nvfx_state_framebuffer;
+extern struct nvfx_state_entry nvfx_state_rasterizer;
+extern struct nvfx_state_entry nvfx_state_scissor;
+extern struct nvfx_state_entry nvfx_state_sr;
+extern struct nvfx_state_entry nvfx_state_stipple;
+extern struct nvfx_state_entry nvfx_state_vbo;
+extern struct nvfx_state_entry nvfx_state_vertprog;
+extern struct nvfx_state_entry nvfx_state_viewport;
+extern struct nvfx_state_entry nvfx_state_vtxfmt;
+extern struct nvfx_state_entry nvfx_state_zsa;
+
+extern void nvfx_init_query_functions(struct nvfx_context *nvfx);
+extern void nvfx_init_surface_functions(struct nvfx_context *nvfx);
+
+/* nvfx_context.c */
+struct pipe_context *
+nvfx_create(struct pipe_screen *pscreen, void *priv);
+
+/* nvfx_clear.c */
+extern void nvfx_clear(struct pipe_context *pipe, unsigned buffers,
+		       const float *rgba, double depth, unsigned stencil);
+
+/* nvfx_draw.c */
+extern struct draw_stage *nvfx_draw_render_stage(struct nvfx_context *nvfx);
+extern void nvfx_draw_elements_swtnl(struct pipe_context *pipe,
+					struct pipe_buffer *idxbuf,
+					unsigned ib_size, unsigned mode,
+					unsigned start, unsigned count);
+
+/* nvfx_fragprog.c */
+extern void nvfx_fragprog_destroy(struct nvfx_context *,
+				    struct nvfx_fragment_program *);
+
+/* nv30_fragtex.c */
+extern void
+nv30_sampler_state_init(struct pipe_context *pipe,
+			  struct nvfx_sampler_state *ps,
+			  const struct pipe_sampler_state *cso);
+extern void nv30_fragtex_bind(struct nvfx_context *);
+extern struct nouveau_stateobj *
+nv30_fragtex_build(struct nvfx_context *nvfx, int unit);
+
+/* nv40_fragtex.c */
+extern void
+nv40_sampler_state_init(struct pipe_context *pipe,
+			  struct nvfx_sampler_state *ps,
+			  const struct pipe_sampler_state *cso);
+extern void nv40_fragtex_bind(struct nvfx_context *);
+extern struct nouveau_stateobj *
+nv40_fragtex_build(struct nvfx_context *nvfx, int unit);
+
+/* nvfx_state.c */
+extern void nvfx_init_state_functions(struct nvfx_context *nvfx);
+
+/* nvfx_state_emit.c */
+extern void nvfx_state_flush_notify(struct nouveau_channel *chan);
+extern boolean nvfx_state_validate(struct nvfx_context *nvfx);
+extern boolean nvfx_state_validate_swtnl(struct nvfx_context *nvfx);
+extern void nvfx_state_emit(struct nvfx_context *nvfx);
+
+/* nvfx_transfer.c */
+extern void nvfx_init_transfer_functions(struct nvfx_context *nvfx);
+
+/* nvfx_vbo.c */
+extern void nvfx_draw_arrays(struct pipe_context *, unsigned mode,
+				unsigned start, unsigned count);
+extern void nvfx_draw_elements(struct pipe_context *pipe,
+				  struct pipe_buffer *indexBuffer,
+				  unsigned indexSize,
+				  unsigned mode, unsigned start,
+				  unsigned count);
+
+/* nvfx_vertprog.c */
+extern void nvfx_vertprog_destroy(struct nvfx_context *,
+				  struct nvfx_vertex_program *);
+
+#endif
diff --git a/src/gallium/drivers/nvfx/nvfx_draw.c b/src/gallium/drivers/nvfx/nvfx_draw.c
new file mode 100644
index 0000000000..5379b29efd
--- /dev/null
+++ b/src/gallium/drivers/nvfx/nvfx_draw.c
@@ -0,0 +1,350 @@
+#include "pipe/p_shader_tokens.h"
+#include "util/u_inlines.h"
+#include "tgsi/tgsi_ureg.h"
+
+#include "util/u_pack_color.h"
+
+#include "draw/draw_context.h"
+#include "draw/draw_vertex.h"
+#include "draw/draw_pipe.h"
+
+#include "nvfx_context.h"
+#include "nv30_vertprog.h"
+#include "nv40_vertprog.h"
+
+/* Simple, but crappy, swtnl path, hopefully we wont need to hit this very
+ * often at all.  Uses "quadro style" vertex submission + a fixed vertex
+ * layout to avoid the need to generate a vertex program or vtxfmt.
+ */
+
+struct nvfx_render_stage {
+	struct draw_stage stage;
+	struct nvfx_context *nvfx;
+	unsigned prim;
+};
+
+static INLINE struct nvfx_render_stage *
+nvfx_render_stage(struct draw_stage *stage)
+{
+	return (struct nvfx_render_stage *)stage;
+}
+
+static INLINE void
+nvfx_render_vertex(struct nvfx_context *nvfx, const struct vertex_header *v)
+{
+	struct nvfx_screen *screen = nvfx->screen;
+	struct nouveau_channel *chan = screen->base.channel;
+	struct nouveau_grobj *eng3d = screen->eng3d;
+	unsigned i;
+
+	for (i = 0; i < nvfx->swtnl.nr_attribs; i++) {
+		unsigned idx = nvfx->swtnl.draw[i];
+		unsigned hw = nvfx->swtnl.hw[i];
+
+		switch (nvfx->swtnl.emit[i]) {
+		case EMIT_OMIT:
+			break;
+		case EMIT_1F:
+			BEGIN_RING(chan, eng3d, NV34TCL_VTX_ATTR_1F(hw), 1);
+			OUT_RING  (chan, fui(v->data[idx][0]));
+			break;
+		case EMIT_2F:
+			BEGIN_RING(chan, eng3d, NV34TCL_VTX_ATTR_2F_X(hw), 2);
+			OUT_RING  (chan, fui(v->data[idx][0]));
+			OUT_RING  (chan, fui(v->data[idx][1]));
+			break;
+		case EMIT_3F:
+			BEGIN_RING(chan, eng3d, NV34TCL_VTX_ATTR_3F_X(hw), 3);
+			OUT_RING  (chan, fui(v->data[idx][0]));
+			OUT_RING  (chan, fui(v->data[idx][1]));
+			OUT_RING  (chan, fui(v->data[idx][2]));
+			break;
+		case EMIT_4F:
+			BEGIN_RING(chan, eng3d, NV34TCL_VTX_ATTR_4F_X(hw), 4);
+			OUT_RING  (chan, fui(v->data[idx][0]));
+			OUT_RING  (chan, fui(v->data[idx][1]));
+			OUT_RING  (chan, fui(v->data[idx][2]));
+			OUT_RING  (chan, fui(v->data[idx][3]));
+			break;
+		case 0xff:
+			BEGIN_RING(chan, eng3d, NV34TCL_VTX_ATTR_4F_X(hw), 4);
+			OUT_RING  (chan, fui(v->data[idx][0] / v->data[idx][3]));
+			OUT_RING  (chan, fui(v->data[idx][1] / v->data[idx][3]));
+			OUT_RING  (chan, fui(v->data[idx][2] / v->data[idx][3]));
+			OUT_RING  (chan, fui(1.0f / v->data[idx][3]));
+			break;
+		case EMIT_4UB:
+			BEGIN_RING(chan, eng3d, NV34TCL_VTX_ATTR_4UB(hw), 1);
+			OUT_RING  (chan, pack_ub4(float_to_ubyte(v->data[idx][0]),
+					    float_to_ubyte(v->data[idx][1]),
+					    float_to_ubyte(v->data[idx][2]),
+					    float_to_ubyte(v->data[idx][3])));
+			break;
+		default:
+			assert(0);
+			break;
+		}
+	}
+}
+
+static INLINE void
+nvfx_render_prim(struct draw_stage *stage, struct prim_header *prim,
+	       unsigned mode, unsigned count)
+{
+	struct nvfx_render_stage *rs = nvfx_render_stage(stage);
+	struct nvfx_context *nvfx = rs->nvfx;
+
+	struct nvfx_screen *screen = nvfx->screen;
+	struct nouveau_channel *chan = screen->base.channel;
+	struct nouveau_grobj *eng3d = screen->eng3d;
+	unsigned i;
+
+	/* Ensure there's room for 4xfloat32 + potentially 3 begin/end */
+	if (AVAIL_RING(chan) < ((count * 20) + 6)) {
+		if (rs->prim != NV34TCL_VERTEX_BEGIN_END_STOP) {
+			NOUVEAU_ERR("AIII, missed flush\n");
+			assert(0);
+		}
+		FIRE_RING(chan);
+		nvfx_state_emit(nvfx);
+	}
+
+	/* Switch primitive modes if necessary */
+	if (rs->prim != mode) {
+		if (rs->prim != NV34TCL_VERTEX_BEGIN_END_STOP) {
+			BEGIN_RING(chan, eng3d, NV34TCL_VERTEX_BEGIN_END, 1);
+			OUT_RING  (chan, NV34TCL_VERTEX_BEGIN_END_STOP);
+		}
+
+		BEGIN_RING(chan, eng3d, NV34TCL_VERTEX_BEGIN_END, 1);
+		OUT_RING  (chan, mode);
+		rs->prim = mode;
+	}
+
+	/* Emit vertex data */
+	for (i = 0; i < count; i++)
+		nvfx_render_vertex(nvfx, prim->v[i]);
+
+	/* If it's likely we'll need to empty the push buffer soon, finish
+	 * off the primitive now.
+	 */
+	if (AVAIL_RING(chan) < ((count * 20) + 6)) {
+		BEGIN_RING(chan, eng3d, NV34TCL_VERTEX_BEGIN_END, 1);
+		OUT_RING  (chan, NV34TCL_VERTEX_BEGIN_END_STOP);
+		rs->prim = NV34TCL_VERTEX_BEGIN_END_STOP;
+	}
+}
+
+static void
+nvfx_render_point(struct draw_stage *draw, struct prim_header *prim)
+{
+	nvfx_render_prim(draw, prim, NV34TCL_VERTEX_BEGIN_END_POINTS, 1);
+}
+
+static void
+nvfx_render_line(struct draw_stage *draw, struct prim_header *prim)
+{
+	nvfx_render_prim(draw, prim, NV34TCL_VERTEX_BEGIN_END_LINES, 2);
+}
+
+static void
+nvfx_render_tri(struct draw_stage *draw, struct prim_header *prim)
+{
+	nvfx_render_prim(draw, prim, NV34TCL_VERTEX_BEGIN_END_TRIANGLES, 3);
+}
+
+static void
+nvfx_render_flush(struct draw_stage *draw, unsigned flags)
+{
+	struct nvfx_render_stage *rs = nvfx_render_stage(draw);
+	struct nvfx_context *nvfx = rs->nvfx;
+	struct nvfx_screen *screen = nvfx->screen;
+	struct nouveau_channel *chan = screen->base.channel;
+	struct nouveau_grobj *eng3d = screen->eng3d;
+
+	if (rs->prim != NV34TCL_VERTEX_BEGIN_END_STOP) {
+		BEGIN_RING(chan, eng3d, NV34TCL_VERTEX_BEGIN_END, 1);
+		OUT_RING  (chan, NV34TCL_VERTEX_BEGIN_END_STOP);
+		rs->prim = NV34TCL_VERTEX_BEGIN_END_STOP;
+	}
+}
+
+static void
+nvfx_render_reset_stipple_counter(struct draw_stage *draw)
+{
+}
+
+static void
+nvfx_render_destroy(struct draw_stage *draw)
+{
+	FREE(draw);
+}
+
+static struct nvfx_vertex_program *
+nvfx_create_drawvp(struct nvfx_context *nvfx)
+{
+	struct ureg_program *ureg;
+	uint i;
+
+	ureg = ureg_create( TGSI_PROCESSOR_VERTEX );
+	if (ureg == NULL)
+		return NULL;
+
+	ureg_MOV(ureg, ureg_DECL_output(ureg, TGSI_SEMANTIC_POSITION, 0), ureg_DECL_vs_input(ureg, 0));
+	ureg_MOV(ureg, ureg_DECL_output(ureg, TGSI_SEMANTIC_COLOR, 0), ureg_DECL_vs_input(ureg, 3));
+	ureg_MOV(ureg, ureg_DECL_output(ureg, TGSI_SEMANTIC_COLOR, 1), ureg_DECL_vs_input(ureg, 4));
+	ureg_MOV(ureg, ureg_DECL_output(ureg, TGSI_SEMANTIC_BCOLOR, 0), ureg_DECL_vs_input(ureg, 3));
+	ureg_MOV(ureg, ureg_DECL_output(ureg, TGSI_SEMANTIC_BCOLOR, 1), ureg_DECL_vs_input(ureg, 4));
+	ureg_MOV(ureg,
+		   ureg_writemask(ureg_DECL_output(ureg, TGSI_SEMANTIC_FOG, 1), TGSI_WRITEMASK_X),
+		   ureg_DECL_vs_input(ureg, 5));
+	for (i = 0; i < 8; ++i)
+		ureg_MOV(ureg, ureg_DECL_output(ureg, TGSI_SEMANTIC_GENERIC, i), ureg_DECL_vs_input(ureg, 8 + i));
+
+	ureg_END( ureg );
+
+	return ureg_create_shader_and_destroy( ureg, &nvfx->pipe );
+}
+
+struct draw_stage *
+nvfx_draw_render_stage(struct nvfx_context *nvfx)
+{
+	struct nvfx_render_stage *render = CALLOC_STRUCT(nvfx_render_stage);
+
+	if (!nvfx->swtnl.vertprog)
+		nvfx->swtnl.vertprog = nvfx_create_drawvp(nvfx);
+
+	render->nvfx = nvfx;
+	render->stage.draw = nvfx->draw;
+	render->stage.point = nvfx_render_point;
+	render->stage.line = nvfx_render_line;
+	render->stage.tri = nvfx_render_tri;
+	render->stage.flush = nvfx_render_flush;
+	render->stage.reset_stipple_counter = nvfx_render_reset_stipple_counter;
+	render->stage.destroy = nvfx_render_destroy;
+
+	return &render->stage;
+}
+
+void
+nvfx_draw_elements_swtnl(struct pipe_context *pipe,
+			 struct pipe_buffer *idxbuf, unsigned idxbuf_size,
+			 unsigned mode, unsigned start, unsigned count)
+{
+	struct nvfx_context *nvfx = nvfx_context(pipe);
+	struct pipe_screen *pscreen = pipe->screen;
+	unsigned i;
+	void *map;
+
+	if (!nvfx_state_validate_swtnl(nvfx))
+		return;
+	nvfx->state.dirty &= ~(1ULL << NVFX_STATE_VTXBUF);
+	nvfx_state_emit(nvfx);
+
+	for (i = 0; i < nvfx->vtxbuf_nr; i++) {
+		map = pipe_buffer_map(pscreen, nvfx->vtxbuf[i].buffer,
+                                      PIPE_BUFFER_USAGE_CPU_READ);
+		draw_set_mapped_vertex_buffer(nvfx->draw, i, map);
+	}
+
+	if (idxbuf) {
+		map = pipe_buffer_map(pscreen, idxbuf,
+				      PIPE_BUFFER_USAGE_CPU_READ);
+		draw_set_mapped_element_buffer(nvfx->draw, idxbuf_size, map);
+	} else {
+		draw_set_mapped_element_buffer(nvfx->draw, 0, NULL);
+	}
+
+	if (nvfx->constbuf[PIPE_SHADER_VERTEX]) {
+		const unsigned nr = nvfx->constbuf_nr[PIPE_SHADER_VERTEX];
+
+		map = pipe_buffer_map(pscreen,
+				      nvfx->constbuf[PIPE_SHADER_VERTEX],
+				      PIPE_BUFFER_USAGE_CPU_READ);
+		draw_set_mapped_constant_buffer(nvfx->draw, PIPE_SHADER_VERTEX, 0,
+                                                map, nr);
+	}
+
+	draw_arrays(nvfx->draw, mode, start, count);
+
+	for (i = 0; i < nvfx->vtxbuf_nr; i++)
+		pipe_buffer_unmap(pscreen, nvfx->vtxbuf[i].buffer);
+
+	if (idxbuf)
+		pipe_buffer_unmap(pscreen, idxbuf);
+
+	if (nvfx->constbuf[PIPE_SHADER_VERTEX])
+		pipe_buffer_unmap(pscreen, nvfx->constbuf[PIPE_SHADER_VERTEX]);
+
+	draw_flush(nvfx->draw);
+	pipe->flush(pipe, 0, NULL);
+}
+
+static INLINE void
+emit_attrib(struct nvfx_context *nvfx, unsigned hw, unsigned emit,
+	    unsigned semantic, unsigned index)
+{
+	unsigned draw_out = draw_find_shader_output(nvfx->draw, semantic, index);
+	unsigned a = nvfx->swtnl.nr_attribs++;
+
+	nvfx->swtnl.hw[a] = hw;
+	nvfx->swtnl.emit[a] = emit;
+	nvfx->swtnl.draw[a] = draw_out;
+}
+
+static boolean
+nvfx_state_vtxfmt_validate(struct nvfx_context *nvfx)
+{
+	struct nvfx_fragment_program *fp = nvfx->fragprog;
+	unsigned colour = 0, texcoords = 0, fog = 0, i;
+
+	/* Determine needed fragprog inputs */
+	for (i = 0; i < fp->info.num_inputs; i++) {
+		switch (fp->info.input_semantic_name[i]) {
+		case TGSI_SEMANTIC_POSITION:
+			break;
+		case TGSI_SEMANTIC_COLOR:
+			colour |= (1 << fp->info.input_semantic_index[i]);
+			break;
+		case TGSI_SEMANTIC_GENERIC:
+			texcoords |= (1 << fp->info.input_semantic_index[i]);
+			break;
+		case TGSI_SEMANTIC_FOG:
+			fog = 1;
+			break;
+		default:
+			assert(0);
+		}
+	}
+
+	nvfx->swtnl.nr_attribs = 0;
+
+	/* Map draw vtxprog output to hw attribute IDs */
+	for (i = 0; i < 2; i++) {
+		if (!(colour & (1 << i)))
+			continue;
+		emit_attrib(nvfx, 3 + i, EMIT_4F, TGSI_SEMANTIC_COLOR, i);
+	}
+
+	for (i = 0; i < 8; i++) {
+		if (!(texcoords & (1 << i)))
+			continue;
+		emit_attrib(nvfx, 8 + i, EMIT_4F, TGSI_SEMANTIC_GENERIC, i);
+	}
+
+	if (fog) {
+		emit_attrib(nvfx, 5, EMIT_1F, TGSI_SEMANTIC_FOG, 0);
+	}
+
+	emit_attrib(nvfx, 0, 0xff, TGSI_SEMANTIC_POSITION, 0);
+
+	return FALSE;
+}
+
+struct nvfx_state_entry nvfx_state_vtxfmt = {
+	.validate = nvfx_state_vtxfmt_validate,
+	.dirty = {
+		.pipe = NVFX_NEW_ARRAYS | NVFX_NEW_FRAGPROG,
+		.hw = 0
+	}
+};
diff --git a/src/gallium/drivers/nv40/nv40_fragprog.c b/src/gallium/drivers/nvfx/nvfx_fragprog.c
index dc24f9b08a..76351430f4 100644
--- a/src/gallium/drivers/nv40/nv40_fragprog.c
+++ b/src/gallium/drivers/nvfx/nvfx_fragprog.c
@@ -7,37 +7,20 @@
 #include "tgsi/tgsi_parse.h"
 #include "tgsi/tgsi_util.h"
 
-#include "nv40_context.h"
-
-#define SWZ_X 0
-#define SWZ_Y 1
-#define SWZ_Z 2
-#define SWZ_W 3
-#define MASK_X 1
-#define MASK_Y 2
-#define MASK_Z 4
-#define MASK_W 8
-#define MASK_ALL (MASK_X|MASK_Y|MASK_Z|MASK_W)
-#define DEF_SCALE NV40_FP_OP_DST_SCALE_1X
-#define DEF_CTEST NV40_FP_OP_COND_TR
-#include "nv40_shader.h"
-
-#define swz(s,x,y,z,w) nv40_sr_swz((s), SWZ_##x, SWZ_##y, SWZ_##z, SWZ_##w)
-#define neg(s) nv40_sr_neg((s))
-#define abs(s) nv40_sr_abs((s))
-#define scale(s,v) nv40_sr_scale((s), NV40_FP_OP_DST_SCALE_##v)
+#include "nvfx_context.h"
+#include "nvfx_shader.h"
 
 #define MAX_CONSTS 128
 #define MAX_IMM 32
-struct nv40_fpc {
-	struct nv40_fragment_program *fp;
+struct nvfx_fpc {
+	struct nvfx_fragment_program *fp;
 
 	uint attrib_map[PIPE_MAX_SHADER_INPUTS];
 
 	unsigned r_temps;
 	unsigned r_temps_discard;
-	struct nv40_sreg r_result[PIPE_MAX_SHADER_OUTPUTS];
-	struct nv40_sreg *r_temp;
+	struct nvfx_sreg r_result[PIPE_MAX_SHADER_OUTPUTS];
+	struct nvfx_sreg *r_temp;
 
 	int num_regs;
 
@@ -50,35 +33,35 @@ struct nv40_fpc {
 	} consts[MAX_CONSTS];
 	int nr_consts;
 
-	struct nv40_sreg imm[MAX_IMM];
+	struct nvfx_sreg imm[MAX_IMM];
 	unsigned nr_imm;
 };
 
-static INLINE struct nv40_sreg
-temp(struct nv40_fpc *fpc)
+static INLINE struct nvfx_sreg
+temp(struct nvfx_fpc *fpc)
 {
 	int idx = ffs(~fpc->r_temps) - 1;
 
 	if (idx < 0) {
 		NOUVEAU_ERR("out of temps!!\n");
 		assert(0);
-		return nv40_sr(NV40SR_TEMP, 0);
+		return nvfx_sr(NVFXSR_TEMP, 0);
 	}
 
 	fpc->r_temps |= (1 << idx);
 	fpc->r_temps_discard |= (1 << idx);
-	return nv40_sr(NV40SR_TEMP, idx);
+	return nvfx_sr(NVFXSR_TEMP, idx);
 }
 
 static INLINE void
-release_temps(struct nv40_fpc *fpc)
+release_temps(struct nvfx_fpc *fpc)
 {
 	fpc->r_temps &= ~fpc->r_temps_discard;
 	fpc->r_temps_discard = 0;
 }
 
-static INLINE struct nv40_sreg
-constant(struct nv40_fpc *fpc, int pipe, float vals[4])
+static INLINE struct nvfx_sreg
+constant(struct nvfx_fpc *fpc, int pipe, float vals[4])
 {
 	int idx;
 
@@ -89,45 +72,45 @@ constant(struct nv40_fpc *fpc, int pipe, float vals[4])
 	fpc->consts[idx].pipe = pipe;
 	if (pipe == -1)
 		memcpy(fpc->consts[idx].vals, vals, 4 * sizeof(float));
-	return nv40_sr(NV40SR_CONST, idx);
+	return nvfx_sr(NVFXSR_CONST, idx);
 }
 
 #define arith(cc,s,o,d,m,s0,s1,s2) \
-	nv40_fp_arith((cc), (s), NV40_FP_OP_OPCODE_##o, \
+	nvfx_fp_arith((cc), (s), NVFX_FP_OP_OPCODE_##o, \
 			(d), (m), (s0), (s1), (s2))
 #define tex(cc,s,o,u,d,m,s0,s1,s2) \
-	nv40_fp_tex((cc), (s), NV40_FP_OP_OPCODE_##o, (u), \
+	nvfx_fp_tex((cc), (s), NVFX_FP_OP_OPCODE_##o, (u), \
 		    (d), (m), (s0), none, none)
 
 static void
-grow_insns(struct nv40_fpc *fpc, int size)
+grow_insns(struct nvfx_fpc *fpc, int size)
 {
-	struct nv40_fragment_program *fp = fpc->fp;
+	struct nvfx_fragment_program *fp = fpc->fp;
 
 	fp->insn_len += size;
 	fp->insn = realloc(fp->insn, sizeof(uint32_t) * fp->insn_len);
 }
 
 static void
-emit_src(struct nv40_fpc *fpc, int pos, struct nv40_sreg src)
+emit_src(struct nvfx_fpc *fpc, int pos, struct nvfx_sreg src)
 {
-	struct nv40_fragment_program *fp = fpc->fp;
+	struct nvfx_fragment_program *fp = fpc->fp;
 	uint32_t *hw = &fp->insn[fpc->inst_offset];
 	uint32_t sr = 0;
 
 	switch (src.type) {
-	case NV40SR_INPUT:
-		sr |= (NV40_FP_REG_TYPE_INPUT << NV40_FP_REG_TYPE_SHIFT);
-		hw[0] |= (src.index << NV40_FP_OP_INPUT_SRC_SHIFT);
+	case NVFXSR_INPUT:
+		sr |= (NVFX_FP_REG_TYPE_INPUT << NVFX_FP_REG_TYPE_SHIFT);
+		hw[0] |= (src.index << NVFX_FP_OP_INPUT_SRC_SHIFT);
 		break;
-	case NV40SR_OUTPUT:
-		sr |= NV40_FP_REG_SRC_HALF;
+	case NVFXSR_OUTPUT:
+		sr |= NVFX_FP_REG_SRC_HALF;
 		/* fall-through */
-	case NV40SR_TEMP:
-		sr |= (NV40_FP_REG_TYPE_TEMP << NV40_FP_REG_TYPE_SHIFT);
-		sr |= (src.index << NV40_FP_REG_SRC_SHIFT);
+	case NVFXSR_TEMP:
+		sr |= (NVFX_FP_REG_TYPE_TEMP << NVFX_FP_REG_TYPE_SHIFT);
+		sr |= (src.index << NVFX_FP_REG_SRC_SHIFT);
 		break;
-	case NV40SR_CONST:
+	case NVFXSR_CONST:
 		if (!fpc->have_const) {
 			grow_insns(fpc, 4);
 			fpc->have_const = 1;
@@ -135,7 +118,7 @@ emit_src(struct nv40_fpc *fpc, int pos, struct nv40_sreg src)
 
 		hw = &fp->insn[fpc->inst_offset];
 		if (fpc->consts[src.index].pipe >= 0) {
-			struct nv40_fragment_program_data *fpd;
+			struct nvfx_fragment_program_data *fpd;
 
 			fp->consts = realloc(fp->consts, ++fp->nr_consts *
 					     sizeof(*fpd));
@@ -149,63 +132,63 @@ emit_src(struct nv40_fpc *fpc, int pos, struct nv40_sreg src)
 				sizeof(uint32_t) * 4);
 		}
 
-		sr |= (NV40_FP_REG_TYPE_CONST << NV40_FP_REG_TYPE_SHIFT);
+		sr |= (NVFX_FP_REG_TYPE_CONST << NVFX_FP_REG_TYPE_SHIFT);
 		break;
-	case NV40SR_NONE:
-		sr |= (NV40_FP_REG_TYPE_INPUT << NV40_FP_REG_TYPE_SHIFT);
+	case NVFXSR_NONE:
+		sr |= (NVFX_FP_REG_TYPE_INPUT << NVFX_FP_REG_TYPE_SHIFT);
 		break;
 	default:
 		assert(0);
 	}
 
 	if (src.negate)
-		sr |= NV40_FP_REG_NEGATE;
+		sr |= NVFX_FP_REG_NEGATE;
 
 	if (src.abs)
 		hw[1] |= (1 << (29 + pos));
 
-	sr |= ((src.swz[0] << NV40_FP_REG_SWZ_X_SHIFT) |
-	       (src.swz[1] << NV40_FP_REG_SWZ_Y_SHIFT) |
-	       (src.swz[2] << NV40_FP_REG_SWZ_Z_SHIFT) |
-	       (src.swz[3] << NV40_FP_REG_SWZ_W_SHIFT));
+	sr |= ((src.swz[0] << NVFX_FP_REG_SWZ_X_SHIFT) |
+	       (src.swz[1] << NVFX_FP_REG_SWZ_Y_SHIFT) |
+	       (src.swz[2] << NVFX_FP_REG_SWZ_Z_SHIFT) |
+	       (src.swz[3] << NVFX_FP_REG_SWZ_W_SHIFT));
 
 	hw[pos + 1] |= sr;
 }
 
 static void
-emit_dst(struct nv40_fpc *fpc, struct nv40_sreg dst)
+emit_dst(struct nvfx_fpc *fpc, struct nvfx_sreg dst)
 {
-	struct nv40_fragment_program *fp = fpc->fp;
+	struct nvfx_fragment_program *fp = fpc->fp;
 	uint32_t *hw = &fp->insn[fpc->inst_offset];
 
 	switch (dst.type) {
-	case NV40SR_TEMP:
+	case NVFXSR_TEMP:
 		if (fpc->num_regs < (dst.index + 1))
 			fpc->num_regs = dst.index + 1;
 		break;
-	case NV40SR_OUTPUT:
+	case NVFXSR_OUTPUT:
 		if (dst.index == 1) {
 			fp->fp_control |= 0xe;
 		} else {
-			hw[0] |= NV40_FP_OP_OUT_REG_HALF;
+			hw[0] |= NVFX_FP_OP_OUT_REG_HALF;
 		}
 		break;
-	case NV40SR_NONE:
+	case NVFXSR_NONE:
 		hw[0] |= (1 << 30);
 		break;
 	default:
 		assert(0);
 	}
 
-	hw[0] |= (dst.index << NV40_FP_OP_OUT_REG_SHIFT);
+	hw[0] |= (dst.index << NVFX_FP_OP_OUT_REG_SHIFT);
 }
 
 static void
-nv40_fp_arith(struct nv40_fpc *fpc, int sat, int op,
-	      struct nv40_sreg dst, int mask,
-	      struct nv40_sreg s0, struct nv40_sreg s1, struct nv40_sreg s2)
+nvfx_fp_arith(struct nvfx_fpc *fpc, int sat, int op,
+	      struct nvfx_sreg dst, int mask,
+	      struct nvfx_sreg s0, struct nvfx_sreg s1, struct nvfx_sreg s2)
 {
-	struct nv40_fragment_program *fp = fpc->fp;
+	struct nvfx_fragment_program *fp = fpc->fp;
 	uint32_t *hw;
 
 	fpc->inst_offset = fp->insn_len;
@@ -214,22 +197,22 @@ nv40_fp_arith(struct nv40_fpc *fpc, int sat, int op,
 	hw = &fp->insn[fpc->inst_offset];
 	memset(hw, 0, sizeof(uint32_t) * 4);
 
-	if (op == NV40_FP_OP_OPCODE_KIL)
-		fp->fp_control |= NV40TCL_FP_CONTROL_KIL;
-	hw[0] |= (op << NV40_FP_OP_OPCODE_SHIFT);
-	hw[0] |= (mask << NV40_FP_OP_OUTMASK_SHIFT);
-	hw[2] |= (dst.dst_scale << NV40_FP_OP_DST_SCALE_SHIFT);
+	if (op == NVFX_FP_OP_OPCODE_KIL)
+		fp->fp_control |= NV34TCL_FP_CONTROL_USES_KIL;
+	hw[0] |= (op << NVFX_FP_OP_OPCODE_SHIFT);
+	hw[0] |= (mask << NVFX_FP_OP_OUTMASK_SHIFT);
+	hw[2] |= (dst.dst_scale << NVFX_FP_OP_DST_SCALE_SHIFT);
 
 	if (sat)
-		hw[0] |= NV40_FP_OP_OUT_SAT;
+		hw[0] |= NVFX_FP_OP_OUT_SAT;
 
 	if (dst.cc_update)
-		hw[0] |= NV40_FP_OP_COND_WRITE_ENABLE;
-	hw[1] |= (dst.cc_test << NV40_FP_OP_COND_SHIFT);
-	hw[1] |= ((dst.cc_swz[0] << NV40_FP_OP_COND_SWZ_X_SHIFT) |
-		  (dst.cc_swz[1] << NV40_FP_OP_COND_SWZ_Y_SHIFT) |
-		  (dst.cc_swz[2] << NV40_FP_OP_COND_SWZ_Z_SHIFT) |
-		  (dst.cc_swz[3] << NV40_FP_OP_COND_SWZ_W_SHIFT));
+		hw[0] |= NVFX_FP_OP_COND_WRITE_ENABLE;
+	hw[1] |= (dst.cc_test << NVFX_FP_OP_COND_SHIFT);
+	hw[1] |= ((dst.cc_swz[0] << NVFX_FP_OP_COND_SWZ_X_SHIFT) |
+		  (dst.cc_swz[1] << NVFX_FP_OP_COND_SWZ_Y_SHIFT) |
+		  (dst.cc_swz[2] << NVFX_FP_OP_COND_SWZ_Z_SHIFT) |
+		  (dst.cc_swz[3] << NVFX_FP_OP_COND_SWZ_W_SHIFT));
 
 	emit_dst(fpc, dst);
 	emit_src(fpc, 0, s0);
@@ -238,26 +221,26 @@ nv40_fp_arith(struct nv40_fpc *fpc, int sat, int op,
 }
 
 static void
-nv40_fp_tex(struct nv40_fpc *fpc, int sat, int op, int unit,
-	    struct nv40_sreg dst, int mask,
-	    struct nv40_sreg s0, struct nv40_sreg s1, struct nv40_sreg s2)
+nvfx_fp_tex(struct nvfx_fpc *fpc, int sat, int op, int unit,
+	    struct nvfx_sreg dst, int mask,
+	    struct nvfx_sreg s0, struct nvfx_sreg s1, struct nvfx_sreg s2)
 {
-	struct nv40_fragment_program *fp = fpc->fp;
+	struct nvfx_fragment_program *fp = fpc->fp;
 
-	nv40_fp_arith(fpc, sat, op, dst, mask, s0, s1, s2);
+	nvfx_fp_arith(fpc, sat, op, dst, mask, s0, s1, s2);
 
-	fp->insn[fpc->inst_offset] |= (unit << NV40_FP_OP_TEX_UNIT_SHIFT);
+	fp->insn[fpc->inst_offset] |= (unit << NVFX_FP_OP_TEX_UNIT_SHIFT);
 	fp->samplers |= (1 << unit);
 }
 
-static INLINE struct nv40_sreg
-tgsi_src(struct nv40_fpc *fpc, const struct tgsi_full_src_register *fsrc)
+static INLINE struct nvfx_sreg
+tgsi_src(struct nvfx_fpc *fpc, const struct tgsi_full_src_register *fsrc)
 {
-	struct nv40_sreg src;
+	struct nvfx_sreg src;
 
 	switch (fsrc->Register.File) {
 	case TGSI_FILE_INPUT:
-		src = nv40_sr(NV40SR_INPUT,
+		src = nvfx_sr(NVFXSR_INPUT,
 			      fpc->attrib_map[fsrc->Register.Index]);
 		break;
 	case TGSI_FILE_CONSTANT:
@@ -288,18 +271,18 @@ tgsi_src(struct nv40_fpc *fpc, const struct tgsi_full_src_register *fsrc)
 	return src;
 }
 
-static INLINE struct nv40_sreg
-tgsi_dst(struct nv40_fpc *fpc, const struct tgsi_full_dst_register *fdst) {
+static INLINE struct nvfx_sreg
+tgsi_dst(struct nvfx_fpc *fpc, const struct tgsi_full_dst_register *fdst) {
 	switch (fdst->Register.File) {
 	case TGSI_FILE_OUTPUT:
 		return fpc->r_result[fdst->Register.Index];
 	case TGSI_FILE_TEMPORARY:
 		return fpc->r_temp[fdst->Register.Index];
 	case TGSI_FILE_NULL:
-		return nv40_sr(NV40SR_NONE, 0);
+		return nvfx_sr(NVFXSR_NONE, 0);
 	default:
 		NOUVEAU_ERR("bad dst file %d\n", fdst->Register.File);
-		return nv40_sr(NV40SR_NONE, 0);
+		return nvfx_sr(NVFXSR_NONE, 0);
 	}
 }
 
@@ -308,52 +291,19 @@ tgsi_mask(uint tgsi)
 {
 	int mask = 0;
 
-	if (tgsi & TGSI_WRITEMASK_X) mask |= MASK_X;
-	if (tgsi & TGSI_WRITEMASK_Y) mask |= MASK_Y;
-	if (tgsi & TGSI_WRITEMASK_Z) mask |= MASK_Z;
-	if (tgsi & TGSI_WRITEMASK_W) mask |= MASK_W;
+	if (tgsi & TGSI_WRITEMASK_X) mask |= NVFX_FP_MASK_X;
+	if (tgsi & TGSI_WRITEMASK_Y) mask |= NVFX_FP_MASK_Y;
+	if (tgsi & TGSI_WRITEMASK_Z) mask |= NVFX_FP_MASK_Z;
+	if (tgsi & TGSI_WRITEMASK_W) mask |= NVFX_FP_MASK_W;
 	return mask;
 }
 
 static boolean
-src_native_swz(struct nv40_fpc *fpc, const struct tgsi_full_src_register *fsrc,
-	       struct nv40_sreg *src)
-{
-	const struct nv40_sreg none = nv40_sr(NV40SR_NONE, 0);
-	struct nv40_sreg tgsi = tgsi_src(fpc, fsrc);
-	uint mask = 0;
-	uint c;
-
-	for (c = 0; c < 4; c++) {
-		switch (tgsi_util_get_full_src_register_swizzle(fsrc, c)) {
-		case TGSI_SWIZZLE_X:
-		case TGSI_SWIZZLE_Y:
-		case TGSI_SWIZZLE_Z:
-		case TGSI_SWIZZLE_W:
-			mask |= (1 << c);
-			break;
-		default:
-			assert(0);
-		}
-	}
-
-	if (mask == MASK_ALL)
-		return TRUE;
-
-	*src = temp(fpc);
-
-	if (mask)
-		arith(fpc, 0, MOV, *src, mask, tgsi, none, none);
-
-	return FALSE;
-}
-
-static boolean
-nv40_fragprog_parse_instruction(struct nv40_fpc *fpc,
+nvfx_fragprog_parse_instruction(struct nvfx_context* nvfx, struct nvfx_fpc *fpc,
 				const struct tgsi_full_instruction *finst)
 {
-	const struct nv40_sreg none = nv40_sr(NV40SR_NONE, 0);
-	struct nv40_sreg src[3], dst, tmp;
+	const struct nvfx_sreg none = nvfx_sr(NVFXSR_NONE, 0);
+	struct nvfx_sreg src[3], dst, tmp;
 	int mask, sat, unit;
 	int ai = -1, ci = -1, ii = -1;
 	int i;
@@ -377,23 +327,12 @@ nv40_fragprog_parse_instruction(struct nv40_fpc *fpc,
 
 		switch (fsrc->Register.File) {
 		case TGSI_FILE_INPUT:
-		case TGSI_FILE_CONSTANT:
-		case TGSI_FILE_TEMPORARY:
-			if (!src_native_swz(fpc, fsrc, &src[i]))
-				continue;
-			break;
-		default:
-			break;
-		}
-
-		switch (fsrc->Register.File) {
-		case TGSI_FILE_INPUT:
 			if (ai == -1 || ai == fsrc->Register.Index) {
 				ai = fsrc->Register.Index;
 				src[i] = tgsi_src(fpc, fsrc);
 			} else {
 				src[i] = temp(fpc);
-				arith(fpc, 0, MOV, src[i], MASK_ALL,
+				arith(fpc, 0, MOV, src[i], NVFX_FP_MASK_ALL,
 				      tgsi_src(fpc, fsrc), none, none);
 			}
 			break;
@@ -404,7 +343,7 @@ nv40_fragprog_parse_instruction(struct nv40_fpc *fpc,
 				src[i] = tgsi_src(fpc, fsrc);
 			} else {
 				src[i] = temp(fpc);
-				arith(fpc, 0, MOV, src[i], MASK_ALL,
+				arith(fpc, 0, MOV, src[i], NVFX_FP_MASK_ALL,
 				      tgsi_src(fpc, fsrc), none, none);
 			}
 			break;
@@ -415,7 +354,7 @@ nv40_fragprog_parse_instruction(struct nv40_fpc *fpc,
 				src[i] = tgsi_src(fpc, fsrc);
 			} else {
 				src[i] = temp(fpc);
-				arith(fpc, 0, MOV, src[i], MASK_ALL,
+				arith(fpc, 0, MOV, src[i], NVFX_FP_MASK_ALL,
 				      tgsi_src(fpc, fsrc), none, none);
 			}
 			break;
@@ -445,25 +384,25 @@ nv40_fragprog_parse_instruction(struct nv40_fpc *fpc,
 		arith(fpc, sat, ADD, dst, mask, src[0], src[1], none);
 		break;
 	case TGSI_OPCODE_CMP:
-		tmp = nv40_sr(NV40SR_NONE, 0);
+		tmp = nvfx_sr(NVFXSR_NONE, 0);
 		tmp.cc_update = 1;
 		arith(fpc, 0, MOV, tmp, 0xf, src[0], none, none);
-		dst.cc_test = NV40_VP_INST_COND_GE;
+		dst.cc_test = NVFX_COND_GE;
 		arith(fpc, sat, MOV, dst, mask, src[2], none, none);
-		dst.cc_test = NV40_VP_INST_COND_LT;
+		dst.cc_test = NVFX_COND_LT;
 		arith(fpc, sat, MOV, dst, mask, src[1], none, none);
 		break;
 	case TGSI_OPCODE_COS:
 		arith(fpc, sat, COS, dst, mask, src[0], none, none);
 		break;
 	case TGSI_OPCODE_DDX:
-		if (mask & (MASK_Z | MASK_W)) {
+		if (mask & (NVFX_FP_MASK_Z | NVFX_FP_MASK_W)) {
 			tmp = temp(fpc);
-			arith(fpc, sat, DDX, tmp, MASK_X | MASK_Y,
+			arith(fpc, sat, DDX, tmp, NVFX_FP_MASK_X | NVFX_FP_MASK_Y,
 			      swz(src[0], Z, W, Z, W), none, none);
-			arith(fpc, 0, MOV, tmp, MASK_Z | MASK_W,
+			arith(fpc, 0, MOV, tmp, NVFX_FP_MASK_Z | NVFX_FP_MASK_W,
 			      swz(tmp, X, Y, X, Y), none, none);
-			arith(fpc, sat, DDX, tmp, MASK_X | MASK_Y, src[0],
+			arith(fpc, sat, DDX, tmp, NVFX_FP_MASK_X | NVFX_FP_MASK_Y, src[0],
 			      none, none);
 			arith(fpc, 0, MOV, dst, mask, tmp, none, none);
 		} else {
@@ -471,13 +410,13 @@ nv40_fragprog_parse_instruction(struct nv40_fpc *fpc,
 		}
 		break;
 	case TGSI_OPCODE_DDY:
-		if (mask & (MASK_Z | MASK_W)) {
+		if (mask & (NVFX_FP_MASK_Z | NVFX_FP_MASK_W)) {
 			tmp = temp(fpc);
-			arith(fpc, sat, DDY, tmp, MASK_X | MASK_Y,
+			arith(fpc, sat, DDY, tmp, NVFX_FP_MASK_X | NVFX_FP_MASK_Y,
 			      swz(src[0], Z, W, Z, W), none, none);
-			arith(fpc, 0, MOV, tmp, MASK_Z | MASK_W,
+			arith(fpc, 0, MOV, tmp, NVFX_FP_MASK_Z | NVFX_FP_MASK_W,
 			      swz(tmp, X, Y, X, Y), none, none);
-			arith(fpc, sat, DDY, tmp, MASK_X | MASK_Y, src[0],
+			arith(fpc, sat, DDY, tmp, NVFX_FP_MASK_X | NVFX_FP_MASK_Y, src[0],
 			      none, none);
 			arith(fpc, 0, MOV, dst, mask, tmp, none, none);
 		} else {
@@ -492,7 +431,7 @@ nv40_fragprog_parse_instruction(struct nv40_fpc *fpc,
 		break;
 	case TGSI_OPCODE_DPH:
 		tmp = temp(fpc);
-		arith(fpc, 0, DP3, tmp, MASK_X, src[0], src[1], none);
+		arith(fpc, 0, DP3, tmp, NVFX_FP_MASK_X, src[0], src[1], none);
 		arith(fpc, sat, ADD, dst, mask, swz(tmp, X, X, X, X),
 		      swz(src[1], W, W, W, W), none);
 		break;
@@ -512,10 +451,10 @@ nv40_fragprog_parse_instruction(struct nv40_fpc *fpc,
 		arith(fpc, 0, KIL, none, 0, none, none, none);
 		break;
 	case TGSI_OPCODE_KIL:
-		dst = nv40_sr(NV40SR_NONE, 0);
+		dst = nvfx_sr(NVFXSR_NONE, 0);
 		dst.cc_update = 1;
-		arith(fpc, 0, MOV, dst, MASK_ALL, src[0], none, none);
-		dst.cc_update = 0; dst.cc_test = NV40_FP_OP_COND_LT;
+		arith(fpc, 0, MOV, dst, NVFX_FP_MASK_ALL, src[0], none, none);
+		dst.cc_update = 0; dst.cc_test = NVFX_COND_LT;
 		arith(fpc, 0, KIL, dst, 0, none, none, none);
 		break;
 	case TGSI_OPCODE_LG2:
@@ -523,9 +462,13 @@ nv40_fragprog_parse_instruction(struct nv40_fpc *fpc,
 		break;
 //	case TGSI_OPCODE_LIT:
 	case TGSI_OPCODE_LRP:
-		tmp = temp(fpc);
-		arith(fpc, 0, MAD, tmp, mask, neg(src[0]), src[2], src[2]);
-		arith(fpc, sat, MAD, dst, mask, src[0], src[1], tmp);
+		if(!nvfx->is_nv4x)
+			arith(fpc, sat, LRP_NV30, dst, mask, src[0], src[1], src[2]);
+		else {
+			tmp = temp(fpc);
+			arith(fpc, 0, MAD, tmp, mask, neg(src[0]), src[2], src[2]);
+			arith(fpc, sat, MAD, dst, mask, src[0], src[1], tmp);
+		}
 		break;
 	case TGSI_OPCODE_MAD:
 		arith(fpc, sat, MAD, dst, mask, src[0], src[1], src[2]);
@@ -543,13 +486,17 @@ nv40_fragprog_parse_instruction(struct nv40_fpc *fpc,
 		arith(fpc, sat, MUL, dst, mask, src[0], src[1], none);
 		break;
 	case TGSI_OPCODE_POW:
-		tmp = temp(fpc);
-		arith(fpc, 0, LG2, tmp, MASK_X,
-		      swz(src[0], X, X, X, X), none, none);
-		arith(fpc, 0, MUL, tmp, MASK_X, swz(tmp, X, X, X, X),
-		      swz(src[1], X, X, X, X), none);
-		arith(fpc, sat, EX2, dst, mask,
-		      swz(tmp, X, X, X, X), none, none);
+		if(!nvfx->is_nv4x)
+			arith(fpc, sat, POW_NV30, dst, mask, src[0], src[1], none);
+		else {
+			tmp = temp(fpc);
+			arith(fpc, 0, LG2, tmp, NVFX_FP_MASK_X,
+			      swz(src[0], X, X, X, X), none, none);
+			arith(fpc, 0, MUL, tmp, NVFX_FP_MASK_X, swz(tmp, X, X, X, X),
+			      swz(src[1], X, X, X, X), none);
+			arith(fpc, sat, EX2, dst, mask,
+			      swz(tmp, X, X, X, X), none, none);
+		}
 		break;
 	case TGSI_OPCODE_RCP:
 		arith(fpc, sat, RCP, dst, mask, src[0], none, none);
@@ -558,42 +505,50 @@ nv40_fragprog_parse_instruction(struct nv40_fpc *fpc,
 		assert(0);
 		break;
 	case TGSI_OPCODE_RFL:
-		tmp = temp(fpc);
-		arith(fpc, 0, DP3, tmp, MASK_X, src[0], src[0], none);
-		arith(fpc, 0, DP3, tmp, MASK_Y, src[0], src[1], none);
-		arith(fpc, 0, DIV, scale(tmp, 2X), MASK_Z,
-		      swz(tmp, Y, Y, Y, Y), swz(tmp, X, X, X, X), none);
-		arith(fpc, sat, MAD, dst, mask,
-		      swz(tmp, Z, Z, Z, Z), src[0], neg(src[1]));
+		if(!nvfx->is_nv4x)
+			arith(fpc, 0, RFL_NV30, dst, mask, src[0], src[1], none);
+		else {
+			tmp = temp(fpc);
+			arith(fpc, 0, DP3, tmp, NVFX_FP_MASK_X, src[0], src[0], none);
+			arith(fpc, 0, DP3, tmp, NVFX_FP_MASK_Y, src[0], src[1], none);
+			arith(fpc, 0, DIV, scale(tmp, 2X), NVFX_FP_MASK_Z,
+			      swz(tmp, Y, Y, Y, Y), swz(tmp, X, X, X, X), none);
+			arith(fpc, sat, MAD, dst, mask,
+			      swz(tmp, Z, Z, Z, Z), src[0], neg(src[1]));
+		}
 		break;
 	case TGSI_OPCODE_RSQ:
-		tmp = temp(fpc);
-		arith(fpc, 0, LG2, scale(tmp, INV_2X), MASK_X,
-		      abs(swz(src[0], X, X, X, X)), none, none);
-		arith(fpc, sat, EX2, dst, mask,
-		      neg(swz(tmp, X, X, X, X)), none, none);
+		if(!nvfx->is_nv4x)
+			arith(fpc, sat, RSQ_NV30, dst, mask, abs(swz(src[0], X, X, X, X)), none, none);
+		else {
+			tmp = temp(fpc);
+			arith(fpc, 0, LG2, scale(tmp, INV_2X), NVFX_FP_MASK_X,
+			      abs(swz(src[0], X, X, X, X)), none, none);
+			arith(fpc, sat, EX2, dst, mask,
+			      neg(swz(tmp, X, X, X, X)), none, none);
+		}
 		break;
 	case TGSI_OPCODE_SCS:
 		/* avoid overwriting the source */
-		if(src[0].swz[SWZ_X] != SWZ_X)
+		if(src[0].swz[NVFX_SWZ_X] != NVFX_SWZ_X)
 		{
-			if (mask & MASK_X) {
-				arith(fpc, sat, COS, dst, MASK_X,
+			if (mask & NVFX_FP_MASK_X) {
+				arith(fpc, sat, COS, dst, NVFX_FP_MASK_X,
 				      swz(src[0], X, X, X, X), none, none);
 			}
-			if (mask & MASK_Y) {
-				arith(fpc, sat, SIN, dst, MASK_Y,
+			if (mask & NVFX_FP_MASK_Y) {
+				arith(fpc, sat, SIN, dst, NVFX_FP_MASK_Y,
 				      swz(src[0], X, X, X, X), none, none);
 			}
 		}
 		else
 		{
-			if (mask & MASK_Y) {
-				arith(fpc, sat, SIN, dst, MASK_Y,
+			if (mask & NVFX_FP_MASK_Y) {
+				arith(fpc, sat, SIN, dst, NVFX_FP_MASK_Y,
 				      swz(src[0], X, X, X, X), none, none);
 			}
-			if (mask & MASK_X) {
-				arith(fpc, sat, COS, dst, MASK_X,
+			if (mask & NVFX_FP_MASK_X) {
+				arith(fpc, sat, COS, dst, NVFX_FP_MASK_X,
 				      swz(src[0], X, X, X, X), none, none);
 			}
 		}
@@ -641,7 +596,7 @@ nv40_fragprog_parse_instruction(struct nv40_fpc *fpc,
 		tmp = temp(fpc);
 		arith(fpc, 0, MUL, tmp, mask,
 		      swz(src[0], Z, X, Y, Y), swz(src[1], Y, Z, X, X), none);
-		arith(fpc, sat, MAD, dst, (mask & ~MASK_W),
+		arith(fpc, sat, MAD, dst, (mask & ~NVFX_FP_MASK_W),
 		      swz(src[0], Y, Z, X, X), swz(src[1], Z, X, Y, Y),
 		      neg(tmp));
 		break;
@@ -655,32 +610,32 @@ nv40_fragprog_parse_instruction(struct nv40_fpc *fpc,
 }
 
 static boolean
-nv40_fragprog_parse_decl_attrib(struct nv40_fpc *fpc,
+nvfx_fragprog_parse_decl_attrib(struct nvfx_context* nvfx, struct nvfx_fpc *fpc,
 				const struct tgsi_full_declaration *fdec)
 {
 	int hw;
 
 	switch (fdec->Semantic.Name) {
 	case TGSI_SEMANTIC_POSITION:
-		hw = NV40_FP_OP_INPUT_SRC_POSITION;
+		hw = NVFX_FP_OP_INPUT_SRC_POSITION;
 		break;
 	case TGSI_SEMANTIC_COLOR:
 		if (fdec->Semantic.Index == 0) {
-			hw = NV40_FP_OP_INPUT_SRC_COL0;
+			hw = NVFX_FP_OP_INPUT_SRC_COL0;
 		} else
 		if (fdec->Semantic.Index == 1) {
-			hw = NV40_FP_OP_INPUT_SRC_COL1;
+			hw = NVFX_FP_OP_INPUT_SRC_COL1;
 		} else {
 			NOUVEAU_ERR("bad colour semantic index\n");
 			return FALSE;
 		}
 		break;
 	case TGSI_SEMANTIC_FOG:
-		hw = NV40_FP_OP_INPUT_SRC_FOGC;
+		hw = NVFX_FP_OP_INPUT_SRC_FOGC;
 		break;
 	case TGSI_SEMANTIC_GENERIC:
 		if (fdec->Semantic.Index <= 7) {
-			hw = NV40_FP_OP_INPUT_SRC_TC(fdec->Semantic.
+			hw = NVFX_FP_OP_INPUT_SRC_TC(fdec->Semantic.
 						     Index);
 		} else {
 			NOUVEAU_ERR("bad generic semantic index\n");
@@ -697,7 +652,7 @@ nv40_fragprog_parse_decl_attrib(struct nv40_fpc *fpc,
 }
 
 static boolean
-nv40_fragprog_parse_decl_output(struct nv40_fpc *fpc,
+nvfx_fragprog_parse_decl_output(struct nvfx_context* nvfx, struct nvfx_fpc *fpc,
 				const struct tgsi_full_declaration *fdec)
 {
 	unsigned idx = fdec->Range.First;
@@ -708,12 +663,14 @@ nv40_fragprog_parse_decl_output(struct nv40_fpc *fpc,
 		hw = 1;
 		break;
 	case TGSI_SEMANTIC_COLOR:
+		hw = ~0;
 		switch (fdec->Semantic.Index) {
 		case 0: hw = 0; break;
 		case 1: hw = 2; break;
 		case 2: hw = 3; break;
 		case 3: hw = 4; break;
-		default:
+		}
+		if(hw > ((nvfx->is_nv4x) ? 4 : 2)) {
 			NOUVEAU_ERR("bad rcol index\n");
 			return FALSE;
 		}
@@ -723,13 +680,13 @@ nv40_fragprog_parse_decl_output(struct nv40_fpc *fpc,
 		return FALSE;
 	}
 
-	fpc->r_result[idx] = nv40_sr(NV40SR_OUTPUT, hw);
+	fpc->r_result[idx] = nvfx_sr(NVFXSR_OUTPUT, hw);
 	fpc->r_temps |= (1 << hw);
 	return TRUE;
 }
 
 static boolean
-nv40_fragprog_prepare(struct nv40_fpc *fpc)
+nvfx_fragprog_prepare(struct nvfx_context* nvfx, struct nvfx_fpc *fpc)
 {
 	struct tgsi_parse_context p;
 	int high_temp = -1, i;
@@ -746,11 +703,11 @@ nv40_fragprog_prepare(struct nv40_fpc *fpc)
 			fdec = &p.FullToken.FullDeclaration;
 			switch (fdec->Declaration.File) {
 			case TGSI_FILE_INPUT:
-				if (!nv40_fragprog_parse_decl_attrib(fpc, fdec))
+				if (!nvfx_fragprog_parse_decl_attrib(nvfx, fpc, fdec))
 					goto out_err;
 				break;
 			case TGSI_FILE_OUTPUT:
-				if (!nv40_fragprog_parse_decl_output(fpc, fdec))
+				if (!nvfx_fragprog_parse_decl_output(nvfx, fpc, fdec))
 					goto out_err;
 				break;
 			case TGSI_FILE_TEMPORARY:
@@ -787,7 +744,7 @@ nv40_fragprog_prepare(struct nv40_fpc *fpc)
 	tgsi_parse_free(&p);
 
 	if (++high_temp) {
-		fpc->r_temp = CALLOC(high_temp, sizeof(struct nv40_sreg));
+		fpc->r_temp = CALLOC(high_temp, sizeof(struct nvfx_sreg));
 		for (i = 0; i < high_temp; i++)
 			fpc->r_temp[i] = temp(fpc);
 		fpc->r_temps_discard = 0;
@@ -803,19 +760,19 @@ out_err:
 }
 
 static void
-nv40_fragprog_translate(struct nv40_context *nv40,
-			struct nv40_fragment_program *fp)
+nvfx_fragprog_translate(struct nvfx_context *nvfx,
+			struct nvfx_fragment_program *fp)
 {
 	struct tgsi_parse_context parse;
-	struct nv40_fpc *fpc = NULL;
+	struct nvfx_fpc *fpc = NULL;
 
-	fpc = CALLOC(1, sizeof(struct nv40_fpc));
+	fpc = CALLOC(1, sizeof(struct nvfx_fpc));
 	if (!fpc)
 		return;
 	fpc->fp = fp;
 	fpc->num_regs = 2;
 
-	if (!nv40_fragprog_prepare(fpc)) {
+	if (!nvfx_fragprog_prepare(nvfx, fpc)) {
 		FREE(fpc);
 		return;
 	}
@@ -831,7 +788,7 @@ nv40_fragprog_translate(struct nv40_context *nv40,
 			const struct tgsi_full_instruction *finst;
 
 			finst = &parse.FullToken.FullInstruction;
-			if (!nv40_fragprog_parse_instruction(fpc, finst))
+			if (!nvfx_fragprog_parse_instruction(nvfx, fpc, finst))
 				goto out_err;
 		}
 			break;
@@ -840,7 +797,10 @@ nv40_fragprog_translate(struct nv40_context *nv40,
 		}
 	}
 
-	fp->fp_control |= fpc->num_regs << NV40TCL_FP_CONTROL_TEMP_COUNT_SHIFT;
+	if(!nvfx->is_nv4x)
+		fp->fp_control |= (fpc->num_regs-1)/2;
+	else
+		fp->fp_control |= fpc->num_regs << NV40TCL_FP_CONTROL_TEMP_COUNT_SHIFT;
 
 	/* Terminate final instruction */
 	fp->insn[fpc->inst_offset] |= 0x00000001;
@@ -862,10 +822,10 @@ out_err:
 }
 
 static void
-nv40_fragprog_upload(struct nv40_context *nv40,
-		     struct nv40_fragment_program *fp)
+nvfx_fragprog_upload(struct nvfx_context *nvfx,
+		     struct nvfx_fragment_program *fp)
 {
-	struct pipe_screen *pscreen = nv40->pipe.screen;
+	struct pipe_screen *pscreen = nvfx->pipe.screen;
 	const uint32_t le = 1;
 	uint32_t *map;
 	int i;
@@ -896,12 +856,12 @@ nv40_fragprog_upload(struct nv40_context *nv40,
 }
 
 static boolean
-nv40_fragprog_validate(struct nv40_context *nv40)
+nvfx_fragprog_validate(struct nvfx_context *nvfx)
 {
-	struct nv40_fragment_program *fp = nv40->fragprog;
+	struct nvfx_fragment_program *fp = nvfx->fragprog;
 	struct pipe_buffer *constbuf =
-		nv40->constbuf[PIPE_SHADER_FRAGMENT];
-	struct pipe_screen *pscreen = nv40->pipe.screen;
+		nvfx->constbuf[PIPE_SHADER_FRAGMENT];
+	struct pipe_screen *pscreen = nvfx->pipe.screen;
 	struct nouveau_stateobj *so;
 	boolean new_consts = FALSE;
 	int i;
@@ -909,24 +869,31 @@ nv40_fragprog_validate(struct nv40_context *nv40)
 	if (fp->translated)
 		goto update_constants;
 
-	nv40->fallback_swrast &= ~NV40_NEW_FRAGPROG;
-	nv40_fragprog_translate(nv40, fp);
+	nvfx->fallback_swrast &= ~NVFX_NEW_FRAGPROG;
+	nvfx_fragprog_translate(nvfx, fp);
 	if (!fp->translated) {
-		nv40->fallback_swrast |= NV40_NEW_FRAGPROG;
+		nvfx->fallback_swrast |= NVFX_NEW_FRAGPROG;
 		return FALSE;
 	}
 
 	fp->buffer = pscreen->buffer_create(pscreen, 0x100, 0, fp->insn_len * 4);
-	nv40_fragprog_upload(nv40, fp);
+	nvfx_fragprog_upload(nvfx, fp);
 
-	so = so_new(2, 2, 1);
-	so_method(so, nv40->screen->curie, NV40TCL_FP_ADDRESS, 1);
+	so = so_new(4, 4, 1);
+	so_method(so, nvfx->screen->eng3d, NV34TCL_FP_ACTIVE_PROGRAM, 1);
 	so_reloc (so, nouveau_bo(fp->buffer), 0, NOUVEAU_BO_VRAM |
 		      NOUVEAU_BO_GART | NOUVEAU_BO_RD | NOUVEAU_BO_LOW |
-		      NOUVEAU_BO_OR, NV40TCL_FP_ADDRESS_DMA0,
-		      NV40TCL_FP_ADDRESS_DMA1);
-	so_method(so, nv40->screen->curie, NV40TCL_FP_CONTROL, 1);
+		      NOUVEAU_BO_OR, NV34TCL_FP_ACTIVE_PROGRAM_DMA0,
+		      NV34TCL_FP_ACTIVE_PROGRAM_DMA1);
+	so_method(so, nvfx->screen->eng3d, NV34TCL_FP_CONTROL, 1);
 	so_data  (so, fp->fp_control);
+	if(!nvfx->is_nv4x) {
+		so_method(so, nvfx->screen->eng3d, NV34TCL_FP_REG_CONTROL, 1);
+		so_data  (so, (1<<16)|0x4);
+		so_method(so, nvfx->screen->eng3d, NV34TCL_TX_UNITS_ENABLE, 1);
+		so_data  (so, fp->samplers);
+	}
+
 	so_ref(so, &fp->so);
 	so_ref(NULL, &so);
 
@@ -937,7 +904,7 @@ update_constants:
 		map = pipe_buffer_map(pscreen, constbuf,
 				      PIPE_BUFFER_USAGE_CPU_READ);
 		for (i = 0; i < fp->nr_consts; i++) {
-			struct nv40_fragment_program_data *fpd = &fp->consts[i];
+			struct nvfx_fragment_program_data *fpd = &fp->consts[i];
 			uint32_t *p = &fp->insn[fpd->offset];
 			uint32_t *cb = (uint32_t *)&map[fpd->index * 4];
 
@@ -949,11 +916,11 @@ update_constants:
 		pipe_buffer_unmap(pscreen, constbuf);
 
 		if (new_consts)
-			nv40_fragprog_upload(nv40, fp);
+			nvfx_fragprog_upload(nvfx, fp);
 	}
 
-	if (new_consts || fp->so != nv40->state.hw[NV40_STATE_FRAGPROG]) {
-		so_ref(fp->so, &nv40->state.hw[NV40_STATE_FRAGPROG]);
+	if (new_consts || fp->so != nvfx->state.hw[NVFX_STATE_FRAGPROG]) {
+		so_ref(fp->so, &nvfx->state.hw[NVFX_STATE_FRAGPROG]);
 		return TRUE;
 	}
 
@@ -961,8 +928,8 @@ update_constants:
 }
 
 void
-nv40_fragprog_destroy(struct nv40_context *nv40,
-		      struct nv40_fragment_program *fp)
+nvfx_fragprog_destroy(struct nvfx_context *nvfx,
+		      struct nvfx_fragment_program *fp)
 {
 	if (fp->buffer)
 		pipe_buffer_reference(&fp->buffer, NULL);
@@ -974,11 +941,10 @@ nv40_fragprog_destroy(struct nv40_context *nv40,
 		FREE(fp->insn);
 }
 
-struct nv40_state_entry nv40_state_fragprog = {
-	.validate = nv40_fragprog_validate,
+struct nvfx_state_entry nvfx_state_fragprog = {
+	.validate = nvfx_fragprog_validate,
 	.dirty = {
-		.pipe = NV40_NEW_FRAGPROG,
-		.hw = NV40_STATE_FRAGPROG
+		.pipe = NVFX_NEW_FRAGPROG,
+		.hw = NVFX_STATE_FRAGPROG
 	}
 };
-
diff --git a/src/gallium/drivers/nvfx/nvfx_fragtex.c b/src/gallium/drivers/nvfx/nvfx_fragtex.c
new file mode 100644
index 0000000000..84e4eb1004
--- /dev/null
+++ b/src/gallium/drivers/nvfx/nvfx_fragtex.c
@@ -0,0 +1,49 @@
+#include "nvfx_context.h"
+
+static boolean
+nvfx_fragtex_validate(struct nvfx_context *nvfx)
+{
+	struct nvfx_fragment_program *fp = nvfx->fragprog;
+	struct nvfx_state *state = &nvfx->state;
+	struct nouveau_stateobj *so;
+	unsigned samplers, unit;
+
+	samplers = state->fp_samplers & ~fp->samplers;
+	while (samplers) {
+		unit = ffs(samplers) - 1;
+		samplers &= ~(1 << unit);
+
+		so = so_new(1, 1, 0);
+		so_method(so, nvfx->screen->eng3d, NV34TCL_TX_ENABLE(unit), 1);
+		so_data  (so, 0);
+		so_ref(so, &nvfx->state.hw[NVFX_STATE_FRAGTEX0 + unit]);
+		so_ref(NULL, &so);
+		state->dirty |= (1ULL << (NVFX_STATE_FRAGTEX0 + unit));
+	}
+
+	samplers = nvfx->dirty_samplers & fp->samplers;
+	while (samplers) {
+		unit = ffs(samplers) - 1;
+		samplers &= ~(1 << unit);
+
+		if(!nvfx->is_nv4x)
+			so = nv30_fragtex_build(nvfx, unit);
+		else
+			so = nv40_fragtex_build(nvfx, unit);
+
+		so_ref(so, &nvfx->state.hw[NVFX_STATE_FRAGTEX0 + unit]);
+		so_ref(NULL, &so);
+		state->dirty |= (1ULL << (NVFX_STATE_FRAGTEX0 + unit));
+	}
+
+	nvfx->state.fp_samplers = fp->samplers;
+	return FALSE;
+}
+
+struct nvfx_state_entry nvfx_state_fragtex = {
+	.validate = nvfx_fragtex_validate,
+	.dirty = {
+		.pipe = NVFX_NEW_SAMPLER | NVFX_NEW_FRAGPROG,
+		.hw = 0
+	}
+};
diff --git a/src/gallium/drivers/nv40/nv40_miptree.c b/src/gallium/drivers/nvfx/nvfx_miptree.c
index 62e97bcea4..0f5ed61aab 100644
--- a/src/gallium/drivers/nv40/nv40_miptree.c
+++ b/src/gallium/drivers/nvfx/nvfx_miptree.c
@@ -4,13 +4,13 @@
 #include "util/u_format.h"
 #include "util/u_math.h"
 
-#include "nv40_context.h"
-#include "../nouveau/nv04_surface_2d.h"
+#include "nvfx_context.h"
+#include "nv04_surface_2d.h"
 
 
 
 static void
-nv40_miptree_layout(struct nv40_miptree *mt)
+nvfx_miptree_layout(struct nvfx_miptree *mt)
 {
 	struct pipe_texture *pt = &mt->base;
 	uint width = pt->width0;
@@ -62,13 +62,13 @@ nv40_miptree_layout(struct nv40_miptree *mt)
 }
 
 static struct pipe_texture *
-nv40_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *pt)
+nvfx_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *pt)
 {
-	struct nv40_miptree *mt;
+	struct nvfx_miptree *mt;
 	unsigned buf_usage = PIPE_BUFFER_USAGE_PIXEL |
 	                     NOUVEAU_BUFFER_USAGE_TEXTURE;
 
-	mt = MALLOC(sizeof(struct nv40_miptree));
+	mt = MALLOC(sizeof(struct nvfx_miptree));
 	if (!mt)
 		return NULL;
 	mt->base = *pt;
@@ -89,6 +89,18 @@ nv40_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *pt)
 		mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR;
 	else {
 		switch (pt->format) {
+		case PIPE_FORMAT_B5G6R5_UNORM:
+		case PIPE_FORMAT_L8A8_UNORM:
+		case PIPE_FORMAT_A8_UNORM:
+		case PIPE_FORMAT_L8_UNORM:
+		case PIPE_FORMAT_I8_UNORM:
+			/* TODO: we can actually swizzle these formats on nv40, we
+				are just preserving the pre-unification behavior.
+				The whole 2D code is going to be rewritten anyway. */
+			if(nvfx_screen(pscreen)->is_nv4x) {
+				mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR;
+				break;
+			}
 		/* TODO: Figure out which formats can be swizzled */
 		case PIPE_FORMAT_B8G8R8A8_UNORM:
 		case PIPE_FORMAT_B8G8R8X8_UNORM:
@@ -112,7 +124,7 @@ nv40_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *pt)
 	if (pt->tex_usage & PIPE_TEXTURE_USAGE_RENDER_TARGET && util_format_get_stride(pt->format, pt->width0) < 64)
 		mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR;
 
-	nv40_miptree_layout(mt);
+	nvfx_miptree_layout(mt);
 
 	mt->buffer = pscreen->buffer_create(pscreen, 256, buf_usage, mt->total_size);
 	if (!mt->buffer) {
@@ -124,17 +136,17 @@ nv40_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *pt)
 }
 
 static struct pipe_texture *
-nv40_miptree_blanket(struct pipe_screen *pscreen, const struct pipe_texture *pt,
+nvfx_miptree_blanket(struct pipe_screen *pscreen, const struct pipe_texture *pt,
 		     const unsigned *stride, struct pipe_buffer *pb)
 {
-	struct nv40_miptree *mt;
+	struct nvfx_miptree *mt;
 
 	/* Only supports 2D, non-mipmapped textures for the moment */
 	if (pt->target != PIPE_TEXTURE_2D || pt->last_level != 0 ||
 	    pt->depth0 != 1)
 		return NULL;
 
-	mt = CALLOC_STRUCT(nv40_miptree);
+	mt = CALLOC_STRUCT(nvfx_miptree);
 	if (!mt)
 		return NULL;
 
@@ -153,9 +165,9 @@ nv40_miptree_blanket(struct pipe_screen *pscreen, const struct pipe_texture *pt,
 }
 
 static void
-nv40_miptree_destroy(struct pipe_texture *pt)
+nvfx_miptree_destroy(struct pipe_texture *pt)
 {
-	struct nv40_miptree *mt = (struct nv40_miptree *)pt;
+	struct nvfx_miptree *mt = (struct nvfx_miptree *)pt;
 	int l;
 
 	pipe_buffer_reference(&mt->buffer, NULL);
@@ -168,11 +180,11 @@ nv40_miptree_destroy(struct pipe_texture *pt)
 }
 
 static struct pipe_surface *
-nv40_miptree_surface_new(struct pipe_screen *pscreen, struct pipe_texture *pt,
+nvfx_miptree_surface_new(struct pipe_screen *pscreen, struct pipe_texture *pt,
 			 unsigned face, unsigned level, unsigned zslice,
 			 unsigned flags)
 {
-	struct nv40_miptree *mt = (struct nv40_miptree *)pt;
+	struct nvfx_miptree *mt = (struct nvfx_miptree *)pt;
 	struct nv04_surface *ns;
 
 	ns = CALLOC_STRUCT(nv04_surface);
@@ -202,21 +214,21 @@ nv40_miptree_surface_new(struct pipe_screen *pscreen, struct pipe_texture *pt,
 	 * Note that ns->pitch is always a multiple of 64 for linear surfaces and swizzled surfaces are POT, so
 	 * ns->pitch & 63 is equivalent to (ns->pitch < 64 && swizzled)*/
 	if((ns->pitch & 63) && (ns->base.usage & (PIPE_BUFFER_USAGE_GPU_WRITE | NOUVEAU_BUFFER_USAGE_NO_RENDER)) == PIPE_BUFFER_USAGE_GPU_WRITE)
-		return &nv04_surface_wrap_for_render(pscreen, ((struct nv40_screen*)pscreen)->eng2d, ns)->base;
+		return &nv04_surface_wrap_for_render(pscreen, ((struct nvfx_screen*)pscreen)->eng2d, ns)->base;
 
 	return &ns->base;
 }
 
 static void
-nv40_miptree_surface_del(struct pipe_surface *ps)
+nvfx_miptree_surface_del(struct pipe_surface *ps)
 {
 	struct nv04_surface* ns = (struct nv04_surface*)ps;
 	if(ns->backing)
 	{
-		struct nv40_screen* screen = (struct nv40_screen*)ps->texture->screen;
+		struct nvfx_screen* screen = (struct nvfx_screen*)ps->texture->screen;
 		if(ns->backing->base.usage & PIPE_BUFFER_USAGE_GPU_WRITE)
 			screen->eng2d->copy(screen->eng2d, &ns->backing->base, 0, 0, ps, 0, 0, ns->base.width, ns->base.height);
-		nv40_miptree_surface_del(&ns->backing->base);
+		nvfx_miptree_surface_del(&ns->backing->base);
 	}
 
 	pipe_texture_reference(&ps->texture, NULL);
@@ -224,13 +236,12 @@ nv40_miptree_surface_del(struct pipe_surface *ps)
 }
 
 void
-nv40_screen_init_miptree_functions(struct pipe_screen *pscreen)
+nvfx_screen_init_miptree_functions(struct pipe_screen *pscreen)
 {
-	pscreen->texture_create = nv40_miptree_create;
-	pscreen->texture_destroy = nv40_miptree_destroy;
-	pscreen->get_tex_surface = nv40_miptree_surface_new;
-	pscreen->tex_surface_destroy = nv40_miptree_surface_del;
+	pscreen->texture_create = nvfx_miptree_create;
+	pscreen->texture_destroy = nvfx_miptree_destroy;
+	pscreen->get_tex_surface = nvfx_miptree_surface_new;
+	pscreen->tex_surface_destroy = nvfx_miptree_surface_del;
 
-	nouveau_screen(pscreen)->texture_blanket = nv40_miptree_blanket;
+	nouveau_screen(pscreen)->texture_blanket = nvfx_miptree_blanket;
 }
-
diff --git a/src/gallium/drivers/nvfx/nvfx_query.c b/src/gallium/drivers/nvfx/nvfx_query.c
new file mode 100644
index 0000000000..acbaf75a23
--- /dev/null
+++ b/src/gallium/drivers/nvfx/nvfx_query.c
@@ -0,0 +1,127 @@
+#include "pipe/p_context.h"
+
+#include "nvfx_context.h"
+
+struct nvfx_query {
+	struct nouveau_resource *object;
+	unsigned type;
+	boolean ready;
+	uint64_t result;
+};
+
+static INLINE struct nvfx_query *
+nvfx_query(struct pipe_query *pipe)
+{
+	return (struct nvfx_query *)pipe;
+}
+
+static struct pipe_query *
+nvfx_query_create(struct pipe_context *pipe, unsigned query_type)
+{
+	struct nvfx_query *q;
+
+	q = CALLOC(1, sizeof(struct nvfx_query));
+	q->type = query_type;
+
+	return (struct pipe_query *)q;
+}
+
+static void
+nvfx_query_destroy(struct pipe_context *pipe, struct pipe_query *pq)
+{
+	struct nvfx_query *q = nvfx_query(pq);
+
+	if (q->object)
+		nouveau_resource_free(&q->object);
+	FREE(q);
+}
+
+static void
+nvfx_query_begin(struct pipe_context *pipe, struct pipe_query *pq)
+{
+	struct nvfx_context *nvfx = nvfx_context(pipe);
+	struct nvfx_query *q = nvfx_query(pq);
+	struct nvfx_screen *screen = nvfx->screen;
+	struct nouveau_channel *chan = screen->base.channel;
+	struct nouveau_grobj *eng3d = screen->eng3d;
+
+	assert(q->type == PIPE_QUERY_OCCLUSION_COUNTER);
+
+	/* Happens when end_query() is called, then another begin_query()
+	 * without querying the result in-between.  For now we'll wait for
+	 * the existing query to notify completion, but it could be better.
+	 */
+	if (q->object) {
+		uint64_t tmp;
+		pipe->get_query_result(pipe, pq, 1, &tmp);
+	}
+
+	if (nouveau_resource_alloc(nvfx->screen->query_heap, 1, NULL, &q->object))
+		assert(0);
+	nouveau_notifier_reset(nvfx->screen->query, q->object->start);
+
+	BEGIN_RING(chan, eng3d, NV34TCL_QUERY_RESET, 1);
+	OUT_RING  (chan, 1);
+	BEGIN_RING(chan, eng3d, NV34TCL_QUERY_UNK17CC, 1);
+	OUT_RING  (chan, 1);
+
+	q->ready = FALSE;
+}
+
+static void
+nvfx_query_end(struct pipe_context *pipe, struct pipe_query *pq)
+{
+	struct nvfx_context *nvfx = nvfx_context(pipe);
+	struct nvfx_screen *screen = nvfx->screen;
+	struct nouveau_channel *chan = screen->base.channel;
+	struct nouveau_grobj *eng3d = screen->eng3d;
+	struct nvfx_query *q = nvfx_query(pq);
+
+	BEGIN_RING(chan, eng3d, NV34TCL_QUERY_GET, 1);
+	OUT_RING  (chan, (0x01 << NV34TCL_QUERY_GET_UNK24_SHIFT) |
+		   ((q->object->start * 32) << NV34TCL_QUERY_GET_OFFSET_SHIFT));
+	FIRE_RING(chan);
+}
+
+static boolean
+nvfx_query_result(struct pipe_context *pipe, struct pipe_query *pq,
+		  boolean wait, uint64_t *result)
+{
+	struct nvfx_context *nvfx = nvfx_context(pipe);
+	struct nvfx_query *q = nvfx_query(pq);
+
+	assert(q->object && q->type == PIPE_QUERY_OCCLUSION_COUNTER);
+
+	if (!q->ready) {
+		unsigned status;
+
+		status = nouveau_notifier_status(nvfx->screen->query,
+						 q->object->start);
+		if (status != NV_NOTIFY_STATE_STATUS_COMPLETED) {
+			if (wait == FALSE)
+				return FALSE;
+
+			nouveau_notifier_wait_status(nvfx->screen->query,
+					q->object->start,
+					NV_NOTIFY_STATE_STATUS_COMPLETED, 0);
+		}
+
+		q->result = nouveau_notifier_return_val(nvfx->screen->query,
+							q->object->start);
+		q->ready = TRUE;
+		nouveau_resource_free(&q->object);
+	}
+
+	*result = q->result;
+	return TRUE;
+}
+
+void
+nvfx_init_query_functions(struct nvfx_context *nvfx)
+{
+	nvfx->pipe.create_query = nvfx_query_create;
+	nvfx->pipe.destroy_query = nvfx_query_destroy;
+	nvfx->pipe.begin_query = nvfx_query_begin;
+	nvfx->pipe.end_query = nvfx_query_end;
+	nvfx->pipe.get_query_result = nvfx_query_result;
+}
diff --git a/src/gallium/drivers/nv30/nv30_screen.c b/src/gallium/drivers/nvfx/nvfx_screen.c
index db24335b7c..8138715cc7 100644
--- a/src/gallium/drivers/nv30/nv30_screen.c
+++ b/src/gallium/drivers/nvfx/nvfx_screen.c
@@ -3,18 +3,18 @@
 
 #include "nouveau/nouveau_screen.h"
 
-#include "nv30_context.h"
-#include "nv30_screen.h"
+#include "nvfx_context.h"
+#include "nvfx_screen.h"
 
 #define NV30TCL_CHIPSET_3X_MASK 0x00000003
 #define NV34TCL_CHIPSET_3X_MASK 0x00000010
 #define NV35TCL_CHIPSET_3X_MASK 0x000001e0
 
 /* FIXME: It seems I should not include directly ../../winsys/drm/nouveau/drm/nouveau_drm_api.h
- * to get the pointer to the context front buffer, so I copied nouveau_winsys here.
- * nv30_screen_surface_format_supported() can then use it to enforce creating fbo
- * with same number of bits everywhere.
- */
+* to get the pointer to the context front buffer, so I copied nouveau_winsys here.
+* nv30_screen_surface_format_supported() can then use it to enforce creating fbo
+* with same number of bits everywhere.
+*/
 struct nouveau_winsys {
 	struct pipe_winsys base;
 
@@ -22,15 +22,21 @@ struct nouveau_winsys {
 
 	struct pipe_surface *front;
 };
+#define NV4X_GRCLASS4097_CHIPSETS 0x00000baf
+#define NV4X_GRCLASS4497_CHIPSETS 0x00005450
+#define NV6X_GRCLASS4497_CHIPSETS 0x00000088
 
 static int
-nv30_screen_get_param(struct pipe_screen *pscreen, int param)
+nvfx_screen_get_param(struct pipe_screen *pscreen, int param)
 {
+	struct nvfx_screen *screen = nvfx_screen(pscreen);
+
 	switch (param) {
 	case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS:
-		return 8;
+		/* TODO: check this */
+		return screen->is_nv4x ? 16 : 8;
 	case PIPE_CAP_NPOT_TEXTURES:
-		return 0;
+		return !!screen->is_nv4x;
 	case PIPE_CAP_TWO_SIDED_STENCIL:
 		return 1;
 	case PIPE_CAP_GLSL:
@@ -40,7 +46,7 @@ nv30_screen_get_param(struct pipe_screen *pscreen, int param)
 	case PIPE_CAP_POINT_SPRITE:
 		return 1;
 	case PIPE_CAP_MAX_RENDER_TARGETS:
-		return 2;
+		return screen->is_nv4x ? 4 : 2;
 	case PIPE_CAP_OCCLUSION_QUERY:
 		return 1;
 	case PIPE_CAP_TEXTURE_SHADOW_MAP:
@@ -52,21 +58,26 @@ nv30_screen_get_param(struct pipe_screen *pscreen, int param)
 	case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS:
 		return 13;
 	case PIPE_CAP_TEXTURE_MIRROR_CLAMP:
-		return 0;
+		return !!screen->is_nv4x;
 	case PIPE_CAP_TEXTURE_MIRROR_REPEAT:
 		return 1;
 	case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS:
-		return 0;
+		return 0; /* We have 4 on nv40 - but unsupported currently */
 	case PIPE_CAP_TGSI_CONT_SUPPORTED:
 		return 0;
 	case PIPE_CAP_BLEND_EQUATION_SEPARATE:
-		return 0;
+		return !!screen->is_nv4x;
 	case NOUVEAU_CAP_HW_VTXBUF:
+		/* TODO: this is almost surely wrong */
+		return !!screen->is_nv4x;
 	case NOUVEAU_CAP_HW_IDXBUF:
-		return 1;
+		/* TODO: this is also almost surely wrong */
+		return screen->is_nv4x && screen->eng3d->grclass == NV40TCL;
 	case PIPE_CAP_MAX_COMBINED_SAMPLERS:
 		return 16;
 	case PIPE_CAP_INDEP_BLEND_ENABLE:
+		/* TODO: on nv40 we have separate color masks */
+		/* TODO: nv40 mrt blending is probably broken */
 		return 0;
 	case PIPE_CAP_INDEP_BLEND_FUNC:
 		return 0;
@@ -83,8 +94,10 @@ nv30_screen_get_param(struct pipe_screen *pscreen, int param)
 }
 
 static float
-nv30_screen_get_paramf(struct pipe_screen *pscreen, int param)
+nvfx_screen_get_paramf(struct pipe_screen *pscreen, int param)
 {
+	struct nvfx_screen *screen = nvfx_screen(pscreen);
+
 	switch (param) {
 	case PIPE_CAP_MAX_LINE_WIDTH:
 	case PIPE_CAP_MAX_LINE_WIDTH_AA:
@@ -93,9 +106,9 @@ nv30_screen_get_paramf(struct pipe_screen *pscreen, int param)
 	case PIPE_CAP_MAX_POINT_WIDTH_AA:
 		return 64.0;
 	case PIPE_CAP_MAX_TEXTURE_ANISOTROPY:
-		return 8.0;
+		return screen->is_nv4x ? 16.0 : 8.0;
 	case PIPE_CAP_MAX_TEXTURE_LOD_BIAS:
-		return 4.0;
+		return screen->is_nv4x ? 16.0 : 4.0;
 	default:
 		NOUVEAU_ERR("Unknown PIPE_CAP %d\n", param);
 		return 0.0;
@@ -103,11 +116,12 @@ nv30_screen_get_paramf(struct pipe_screen *pscreen, int param)
 }
 
 static boolean
-nv30_screen_surface_format_supported(struct pipe_screen *pscreen,
+nvfx_screen_surface_format_supported(struct pipe_screen *pscreen,
 				     enum pipe_format format,
 				     enum pipe_texture_target target,
 				     unsigned tex_usage, unsigned geom_flags)
 {
+	struct nvfx_screen *screen = nvfx_screen(pscreen);
 	struct pipe_surface *front = ((struct nouveau_winsys *) pscreen->winsys)->front;
 
 	if (tex_usage & PIPE_TEXTURE_USAGE_RENDER_TARGET) {
@@ -125,9 +139,9 @@ nv30_screen_surface_format_supported(struct pipe_screen *pscreen,
 		case PIPE_FORMAT_X8Z24_UNORM:
 			return TRUE;
 		case PIPE_FORMAT_Z16_UNORM:
-			if (front) {
+			/* TODO: this nv30 limitation probably does not exist */
+			if (!screen->is_nv4x && front)
 				return (front->format == PIPE_FORMAT_B5G6R5_UNORM);
-			}
 			return TRUE;
 		default:
 			break;
@@ -144,7 +158,14 @@ nv30_screen_surface_format_supported(struct pipe_screen *pscreen,
 		case PIPE_FORMAT_L8A8_UNORM:
 		case PIPE_FORMAT_Z16_UNORM:
 		case PIPE_FORMAT_S8Z24_UNORM:
+		case PIPE_FORMAT_DXT1_RGB:
+		case PIPE_FORMAT_DXT1_RGBA:
+		case PIPE_FORMAT_DXT3_RGBA:
+		case PIPE_FORMAT_DXT5_RGBA:
 			return TRUE;
+		/* TODO: does nv30 support this? */
+		case PIPE_FORMAT_R16_SNORM:
+			return !!screen->is_nv4x;
 		default:
 			break;
 		}
@@ -154,20 +175,20 @@ nv30_screen_surface_format_supported(struct pipe_screen *pscreen,
 }
 
 static struct pipe_buffer *
-nv30_surface_buffer(struct pipe_surface *surf)
+nvfx_surface_buffer(struct pipe_surface *surf)
 {
-	struct nv30_miptree *mt = (struct nv30_miptree *)surf->texture;
+	struct nvfx_miptree *mt = (struct nvfx_miptree *)surf->texture;
 
 	return mt->buffer;
 }
 
 static void
-nv30_screen_destroy(struct pipe_screen *pscreen)
+nvfx_screen_destroy(struct pipe_screen *pscreen)
 {
-	struct nv30_screen *screen = nv30_screen(pscreen);
+	struct nvfx_screen *screen = nvfx_screen(pscreen);
 	unsigned i;
 
-	for (i = 0; i < NV30_STATE_MAX; i++) {
+	for (i = 0; i < NVFX_STATE_MAX; i++) {
 		if (screen->state[i])
 			so_ref(NULL, &screen->state[i]);
 	}
@@ -177,7 +198,7 @@ nv30_screen_destroy(struct pipe_screen *pscreen)
 	nouveau_resource_destroy(&screen->query_heap);
 	nouveau_notifier_free(&screen->query);
 	nouveau_notifier_free(&screen->sync);
-	nouveau_grobj_free(&screen->rankine);
+	nouveau_grobj_free(&screen->eng3d);
 	nv04_surface_2d_takedown(&screen->eng2d);
 
 	nouveau_screen_fini(&screen->base);
@@ -185,59 +206,155 @@ nv30_screen_destroy(struct pipe_screen *pscreen)
 	FREE(pscreen);
 }
 
+static void nv30_screen_init(struct nvfx_screen *screen, struct nouveau_stateobj* so)
+{
+	int i;
+
+	/* TODO: perhaps we should do some of this on nv40 too? */
+	for (i=1; i<8; i++) {
+		so_method(so, screen->eng3d, NV34TCL_VIEWPORT_CLIP_HORIZ(i), 1);
+		so_data  (so, 0);
+		so_method(so, screen->eng3d, NV34TCL_VIEWPORT_CLIP_VERT(i), 1);
+		so_data  (so, 0);
+	}
+
+	so_method(so, screen->eng3d, 0x220, 1);
+	so_data  (so, 1);
+
+	so_method(so, screen->eng3d, 0x03b0, 1);
+	so_data  (so, 0x00100000);
+	so_method(so, screen->eng3d, 0x1454, 1);
+	so_data  (so, 0);
+	so_method(so, screen->eng3d, 0x1d80, 1);
+	so_data  (so, 3);
+	so_method(so, screen->eng3d, 0x1450, 1);
+	so_data  (so, 0x00030004);
+
+	/* NEW */
+	so_method(so, screen->eng3d, 0x1e98, 1);
+	so_data  (so, 0);
+	so_method(so, screen->eng3d, 0x17e0, 3);
+	so_data  (so, fui(0.0));
+	so_data  (so, fui(0.0));
+	so_data  (so, fui(1.0));
+	so_method(so, screen->eng3d, 0x1f80, 16);
+	for (i=0; i<16; i++) {
+		so_data  (so, (i==8) ? 0x0000ffff : 0);
+	}
+
+	so_method(so, screen->eng3d, 0x120, 3);
+	so_data  (so, 0);
+	so_data  (so, 1);
+	so_data  (so, 2);
+
+	so_method(so, screen->eng3d, 0x1d88, 1);
+	so_data  (so, 0x00001200);
+
+	so_method(so, screen->eng3d, NV34TCL_RC_ENABLE, 1);
+	so_data  (so, 0);
+
+	so_method(so, screen->eng3d, NV34TCL_DEPTH_RANGE_NEAR, 2);
+	so_data  (so, fui(0.0));
+	so_data  (so, fui(1.0));
+
+	so_method(so, screen->eng3d, NV34TCL_MULTISAMPLE_CONTROL, 1);
+	so_data  (so, 0xffff0000);
+
+	/* enables use of vp rather than fixed-function somehow */
+	so_method(so, screen->eng3d, 0x1e94, 1);
+	so_data  (so, 0x13);
+}
+
+static void nv40_screen_init(struct nvfx_screen *screen, struct nouveau_stateobj* so)
+{
+	so_method(so, screen->eng3d, NV40TCL_DMA_COLOR2, 2);
+	so_data  (so, screen->base.channel->vram->handle);
+	so_data  (so, screen->base.channel->vram->handle);
+
+	so_method(so, screen->eng3d, 0x1ea4, 3);
+	so_data  (so, 0x00000010);
+	so_data  (so, 0x01000100);
+	so_data  (so, 0xff800006);
+
+	/* vtxprog output routing */
+	so_method(so, screen->eng3d, 0x1fc4, 1);
+	so_data  (so, 0x06144321);
+	so_method(so, screen->eng3d, 0x1fc8, 2);
+	so_data  (so, 0xedcba987);
+	so_data  (so, 0x00000021);
+	so_method(so, screen->eng3d, 0x1fd0, 1);
+	so_data  (so, 0x00171615);
+	so_method(so, screen->eng3d, 0x1fd4, 1);
+	so_data  (so, 0x001b1a19);
+
+	so_method(so, screen->eng3d, 0x1ef8, 1);
+	so_data  (so, 0x0020ffff);
+	so_method(so, screen->eng3d, 0x1d64, 1);
+	so_data  (so, 0x00d30000);
+	so_method(so, screen->eng3d, 0x1e94, 1);
+	so_data  (so, 0x00000001);
+}
+
 struct pipe_screen *
-nv30_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev)
+nvfx_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev)
 {
-	struct nv30_screen *screen = CALLOC_STRUCT(nv30_screen);
+	struct nvfx_screen *screen = CALLOC_STRUCT(nvfx_screen);
 	struct nouveau_channel *chan;
 	struct pipe_screen *pscreen;
 	struct nouveau_stateobj *so;
-	unsigned rankine_class = 0;
-	int ret, i;
+	unsigned eng3d_class = 0;
+	int ret;
 
 	if (!screen)
 		return NULL;
+
 	pscreen = &screen->base.base;
 
 	ret = nouveau_screen_init(&screen->base, dev);
 	if (ret) {
-		nv30_screen_destroy(pscreen);
+		nvfx_screen_destroy(pscreen);
 		return NULL;
 	}
 	chan = screen->base.channel;
 
 	pscreen->winsys = ws;
-	pscreen->destroy = nv30_screen_destroy;
-	pscreen->get_param = nv30_screen_get_param;
-	pscreen->get_paramf = nv30_screen_get_paramf;
-	pscreen->is_format_supported = nv30_screen_surface_format_supported;
-	pscreen->context_create = nv30_create;
-
-	nv30_screen_init_miptree_functions(pscreen);
+	pscreen->destroy = nvfx_screen_destroy;
+	pscreen->get_param = nvfx_screen_get_param;
+	pscreen->get_paramf = nvfx_screen_get_paramf;
+	pscreen->is_format_supported = nvfx_screen_surface_format_supported;
+	pscreen->context_create = nvfx_create;
 
-	/* 3D object */
 	switch (dev->chipset & 0xf0) {
 	case 0x30:
 		if (NV30TCL_CHIPSET_3X_MASK & (1 << (dev->chipset & 0x0f)))
-			rankine_class = 0x0397;
-		else
-		if (NV34TCL_CHIPSET_3X_MASK & (1 << (dev->chipset & 0x0f)))
-			rankine_class = 0x0697;
-		else
-		if (NV35TCL_CHIPSET_3X_MASK & (1 << (dev->chipset & 0x0f)))
-			rankine_class = 0x0497;
+			eng3d_class = 0x0397;
+		else if (NV34TCL_CHIPSET_3X_MASK & (1 << (dev->chipset & 0x0f)))
+			eng3d_class = 0x0697;
+		else if (NV35TCL_CHIPSET_3X_MASK & (1 << (dev->chipset & 0x0f)))
+			eng3d_class = 0x0497;
 		break;
-	default:
+	case 0x40:
+		if (NV4X_GRCLASS4097_CHIPSETS & (1 << (dev->chipset & 0x0f)))
+			eng3d_class = NV40TCL;
+		else if (NV4X_GRCLASS4497_CHIPSETS & (1 << (dev->chipset & 0x0f)))
+			eng3d_class = NV44TCL;
+		screen->is_nv4x = ~0;
+		break;
+	case 0x60:
+		if (NV6X_GRCLASS4497_CHIPSETS & (1 << (dev->chipset & 0x0f)))
+			eng3d_class = NV44TCL;
+		screen->is_nv4x = ~0;
 		break;
 	}
 
-	if (!rankine_class) {
-		NOUVEAU_ERR("Unknown nv3x chipset: nv%02x\n", dev->chipset);
+	if (!eng3d_class) {
+		NOUVEAU_ERR("Unknown nv3x/nv4x chipset: nv%02x\n", dev->chipset);
 		return NULL;
 	}
 
-	ret = nouveau_grobj_alloc(chan, 0xbeef3097, rankine_class,
-				  &screen->rankine);
+	nvfx_screen_init_miptree_functions(pscreen);
+
+	ret = nouveau_grobj_alloc(chan, 0xbeef3097, eng3d_class, &screen->eng3d);
 	if (ret) {
 		NOUVEAU_ERR("Error creating 3D object: %d\n", ret);
 		return FALSE;
@@ -245,13 +362,13 @@ nv30_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev)
 
 	/* 2D engine setup */
 	screen->eng2d = nv04_surface_2d_init(&screen->base);
-	screen->eng2d->buf = nv30_surface_buffer;
+	screen->eng2d->buf = nvfx_surface_buffer;
 
 	/* Notifier for sync purposes */
 	ret = nouveau_notifier_alloc(chan, 0xbeef0301, 1, &screen->sync);
 	if (ret) {
 		NOUVEAU_ERR("Error creating notifier object: %d\n", ret);
-		nv30_screen_destroy(pscreen);
+		nvfx_screen_destroy(pscreen);
 		return NULL;
 	}
 
@@ -259,99 +376,54 @@ nv30_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev)
 	ret = nouveau_notifier_alloc(chan, 0xbeef0302, 32, &screen->query);
 	if (ret) {
 		NOUVEAU_ERR("Error initialising query objects: %d\n", ret);
-		nv30_screen_destroy(pscreen);
+		nvfx_screen_destroy(pscreen);
 		return NULL;
 	}
 
 	ret = nouveau_resource_init(&screen->query_heap, 0, 32);
 	if (ret) {
 		NOUVEAU_ERR("Error initialising query object heap: %d\n", ret);
-		nv30_screen_destroy(pscreen);
+		nvfx_screen_destroy(pscreen);
 		return NULL;
 	}
 
 	/* Vtxprog resources */
-	if (nouveau_resource_init(&screen->vp_exec_heap, 0, 256) ||
+	if (nouveau_resource_init(&screen->vp_exec_heap, 0, screen->is_nv4x ? 512 : 256) ||
 	    nouveau_resource_init(&screen->vp_data_heap, 0, 256)) {
-		nv30_screen_destroy(pscreen);
+		nvfx_screen_destroy(pscreen);
 		return NULL;
 	}
 
-	/* Static rankine initialisation */
-	so = so_new(36, 60, 0);
-	so_method(so, screen->rankine, NV34TCL_DMA_NOTIFY, 1);
+	/* Static eng3d initialisation */
+	/* make the so big and don't worry about exact values
+	   since we it will be thrown away immediately after use */
+	so = so_new(256, 256, 0);
+	so_method(so, screen->eng3d, NV34TCL_DMA_NOTIFY, 1);
 	so_data  (so, screen->sync->handle);
-	so_method(so, screen->rankine, NV34TCL_DMA_TEXTURE0, 2);
+	so_method(so, screen->eng3d, NV34TCL_DMA_TEXTURE0, 2);
 	so_data  (so, chan->vram->handle);
 	so_data  (so, chan->gart->handle);
-	so_method(so, screen->rankine, NV34TCL_DMA_COLOR1, 1);
+	so_method(so, screen->eng3d, NV34TCL_DMA_COLOR1, 1);
 	so_data  (so, chan->vram->handle);
-	so_method(so, screen->rankine, NV34TCL_DMA_COLOR0, 2);
+	so_method(so, screen->eng3d, NV34TCL_DMA_COLOR0, 2);
 	so_data  (so, chan->vram->handle);
 	so_data  (so, chan->vram->handle);
-	so_method(so, screen->rankine, NV34TCL_DMA_VTXBUF0, 2);
+	so_method(so, screen->eng3d, NV34TCL_DMA_VTXBUF0, 2);
 	so_data  (so, chan->vram->handle);
 	so_data  (so, chan->gart->handle);
-/*	so_method(so, screen->rankine, NV34TCL_DMA_FENCE, 2);
-	so_data  (so, 0);
-	so_data  (so, screen->query->handle);*/
-	so_method(so, screen->rankine, NV34TCL_DMA_IN_MEMORY7, 1);
-	so_data  (so, chan->vram->handle);
-	so_method(so, screen->rankine, NV34TCL_DMA_IN_MEMORY8, 1);
-	so_data  (so, chan->vram->handle);
-
-	for (i=1; i<8; i++) {
-		so_method(so, screen->rankine, NV34TCL_VIEWPORT_CLIP_HORIZ(i), 1);
-		so_data  (so, 0);
-		so_method(so, screen->rankine, NV34TCL_VIEWPORT_CLIP_VERT(i), 1);
-		so_data  (so, 0);
-	}
-
-	so_method(so, screen->rankine, 0x220, 1);
-	so_data  (so, 1);
-
-	so_method(so, screen->rankine, 0x03b0, 1);
-	so_data  (so, 0x00100000);
-	so_method(so, screen->rankine, 0x1454, 1);
-	so_data  (so, 0);
-	so_method(so, screen->rankine, 0x1d80, 1);
-	so_data  (so, 3);
-	so_method(so, screen->rankine, 0x1450, 1);
-	so_data  (so, 0x00030004);
 
-	/* NEW */
-	so_method(so, screen->rankine, 0x1e98, 1);
-	so_data  (so, 0);
-	so_method(so, screen->rankine, 0x17e0, 3);
-	so_data  (so, fui(0.0));
-	so_data  (so, fui(0.0));
-	so_data  (so, fui(1.0));
-	so_method(so, screen->rankine, 0x1f80, 16);
-	for (i=0; i<16; i++) {
-		so_data  (so, (i==8) ? 0x0000ffff : 0);
-	}
-
-	so_method(so, screen->rankine, 0x120, 3);
+	so_method(so, screen->eng3d, NV34TCL_DMA_FENCE, 2);
 	so_data  (so, 0);
-	so_data  (so, 1);
-	so_data  (so, 2);
+	so_data  (so, screen->query->handle);
 
-	so_method(so, screen->rankine, 0x1d88, 1);
-	so_data  (so, 0x00001200);
-
-	so_method(so, screen->rankine, NV34TCL_RC_ENABLE, 1);
-	so_data  (so, 0);
-
-	so_method(so, screen->rankine, NV34TCL_DEPTH_RANGE_NEAR, 2);
-	so_data  (so, fui(0.0));
-	so_data  (so, fui(1.0));
-
-	so_method(so, screen->rankine, NV34TCL_MULTISAMPLE_CONTROL, 1);
-	so_data  (so, 0xffff0000);
+	so_method(so, screen->eng3d, NV34TCL_DMA_IN_MEMORY7, 2);
+	so_data  (so, chan->vram->handle);
+	so_data  (so, chan->vram->handle);
 
-	/* enables use of vp rather than fixed-function somehow */
-	so_method(so, screen->rankine, 0x1e94, 1);
-	so_data  (so, 0x13);
+	if(!screen->is_nv4x)
+		nv30_screen_init(screen, so);
+	else
+		nv40_screen_init(screen, so);
 
 	so_emit(chan, so);
 	so_ref(NULL, &so);
diff --git a/src/gallium/drivers/nv30/nv30_screen.h b/src/gallium/drivers/nvfx/nvfx_screen.h
index b7856cdf00..c0b4b9899d 100644
--- a/src/gallium/drivers/nv30/nv30_screen.h
+++ b/src/gallium/drivers/nvfx/nvfx_screen.h
@@ -1,20 +1,21 @@
-#ifndef __NV30_SCREEN_H__
-#define __NV30_SCREEN_H__
+#ifndef __NVFX_SCREEN_H__
+#define __NVFX_SCREEN_H__
 
 #include "nouveau/nouveau_screen.h"
+#include "nv04_surface_2d.h"
 
-#include "nouveau/nv04_surface_2d.h"
-
-struct nv30_screen {
+struct nvfx_screen {
 	struct nouveau_screen base;
 
 	struct nouveau_winsys *nvws;
 
-	struct nv30_context *cur_ctx;
+	struct nvfx_context *cur_ctx;
+
+	unsigned is_nv4x; /* either 0 or ~0 */
 
 	/* HW graphics objects */
 	struct nv04_surface_2d *eng2d;
-	struct nouveau_grobj *rankine;
+	struct nouveau_grobj *eng3d;
 	struct nouveau_notifier *sync;
 
 	/* Query object resources */
@@ -26,13 +27,13 @@ struct nv30_screen {
 	struct nouveau_resource *vp_data_heap;
 
 	/* Current 3D state of channel */
-	struct nouveau_stateobj *state[NV30_STATE_MAX];
+	struct nouveau_stateobj *state[NVFX_STATE_MAX];
 };
 
-static INLINE struct nv30_screen *
-nv30_screen(struct pipe_screen *screen)
+static INLINE struct nvfx_screen *
+nvfx_screen(struct pipe_screen *screen)
 {
-	return (struct nv30_screen *)screen;
+	return (struct nvfx_screen *)screen;
 }
 
 #endif
diff --git a/src/gallium/drivers/nvfx/nvfx_shader.h b/src/gallium/drivers/nvfx/nvfx_shader.h
new file mode 100644
index 0000000000..0b2f044f7f
--- /dev/null
+++ b/src/gallium/drivers/nvfx/nvfx_shader.h
@@ -0,0 +1,429 @@
+#ifndef __NVFX_SHADER_H__
+#define __NVFX_SHADER_H__
+
+/* this will resolve to either the NV30 or the NV40 version
+ * depending on the current hardware */
+/* unusual, but very fast and compact method */
+#define NVFX_VP(c) ((NV30_VP_##c) + (nvfx->is_nv4x & ((NV40_VP_##c) - (NV30_VP_##c))))
+
+#define NVFX_VP_INST_SLOT_VEC 0
+#define NVFX_VP_INST_SLOT_SCA 1
+
+#define NVFX_VP_INST_IN_POS  0    /* These seem to match the bindings specified in */
+#define NVFX_VP_INST_IN_WEIGHT  1    /* the ARB_v_p spec (2.14.3.1) */
+#define NVFX_VP_INST_IN_NORMAL  2
+#define NVFX_VP_INST_IN_COL0  3    /* Should probably confirm them all though */
+#define NVFX_VP_INST_IN_COL1  4
+#define NVFX_VP_INST_IN_FOGC  5
+#define NVFX_VP_INST_IN_TC0  8
+#define NVFX_VP_INST_IN_TC(n)  (8+n)
+
+#define NVFX_VP_INST_SCA_OP_NOP 0x00
+#define NVFX_VP_INST_SCA_OP_MOV 0x01
+#define NVFX_VP_INST_SCA_OP_RCP 0x02
+#define NVFX_VP_INST_SCA_OP_RCC 0x03
+#define NVFX_VP_INST_SCA_OP_RSQ 0x04
+#define NVFX_VP_INST_SCA_OP_EXP 0x05
+#define NVFX_VP_INST_SCA_OP_LOG 0x06
+#define NVFX_VP_INST_SCA_OP_LIT 0x07
+#define NVFX_VP_INST_SCA_OP_BRA 0x09
+#define NVFX_VP_INST_SCA_OP_CAL 0x0B
+#define NVFX_VP_INST_SCA_OP_RET 0x0C
+#define NVFX_VP_INST_SCA_OP_LG2 0x0D
+#define NVFX_VP_INST_SCA_OP_EX2 0x0E
+#define NVFX_VP_INST_SCA_OP_SIN 0x0F
+#define NVFX_VP_INST_SCA_OP_COS 0x10
+
+#define NV40_VP_INST_SCA_OP_PUSHA 0x13
+#define NV40_VP_INST_SCA_OP_POPA 0x14
+
+#define NVFX_VP_INST_VEC_OP_NOP 0x00
+#define NVFX_VP_INST_VEC_OP_MOV 0x01
+#define NVFX_VP_INST_VEC_OP_MUL 0x02
+#define NVFX_VP_INST_VEC_OP_ADD 0x03
+#define NVFX_VP_INST_VEC_OP_MAD 0x04
+#define NVFX_VP_INST_VEC_OP_DP3 0x05
+#define NVFX_VP_INST_VEC_OP_DPH 0x06
+#define NVFX_VP_INST_VEC_OP_DP4 0x07
+#define NVFX_VP_INST_VEC_OP_DST 0x08
+#define NVFX_VP_INST_VEC_OP_MIN 0x09
+#define NVFX_VP_INST_VEC_OP_MAX 0x0A
+#define NVFX_VP_INST_VEC_OP_SLT 0x0B
+#define NVFX_VP_INST_VEC_OP_SGE 0x0C
+#define NVFX_VP_INST_VEC_OP_ARL 0x0D
+#define NVFX_VP_INST_VEC_OP_FRC 0x0E
+#define NVFX_VP_INST_VEC_OP_FLR 0x0F
+#define NVFX_VP_INST_VEC_OP_SEQ 0x10
+#define NVFX_VP_INST_VEC_OP_SFL 0x11
+#define NVFX_VP_INST_VEC_OP_SGT 0x12
+#define NVFX_VP_INST_VEC_OP_SLE 0x13
+#define NVFX_VP_INST_VEC_OP_SNE 0x14
+#define NVFX_VP_INST_VEC_OP_STR 0x15
+#define NVFX_VP_INST_VEC_OP_SSG 0x16
+#define NVFX_VP_INST_VEC_OP_ARR 0x17
+#define NVFX_VP_INST_VEC_OP_ARA 0x18
+
+#define NV40_VP_INST_VEC_OP_TXL 0x19
+
+/* DWORD 3 */
+#define NVFX_VP_INST_LAST                           (1 << 0)
+
+/*
+ * Each fragment program opcode appears to be comprised of 4 32-bit values.
+ *
+ *   0 - Opcode, output reg/mask, ATTRIB source
+ *   1 - Source 0
+ *   2 - Source 1
+ *   3 - Source 2
+ *
+ * There appears to be no special difference between result regs and temp regs.
+ *     result.color == R0.xyzw
+ *     result.depth == R1.z
+ * When the fragprog contains instructions to write depth, NV30_TCL_PRIMITIVE_3D_UNK1D78=0
+ * otherwise it is set to 1.
+ *
+ * Constants are inserted directly after the instruction that uses them.
+ *
+ * It appears that it's not possible to use two input registers in one
+ * instruction as the input sourcing is done in the instruction dword
+ * and not the source selection dwords.  As such instructions such as:
+ *
+ *     ADD result.color, fragment.color, fragment.texcoord[0];
+ *
+ * must be split into two MOV's and then an ADD (nvidia does this) but
+ * I'm not sure why it's not just one MOV and then source the second input
+ * in the ADD instruction..
+ *
+ * Negation of the full source is done with NV30_FP_REG_NEGATE, arbitrary
+ * negation requires multiplication with a const.
+ *
+ * Arbitrary swizzling is supported with the exception of SWIZZLE_ZERO/SWIZZLE_ONE
+ * The temp/result regs appear to be initialised to (0.0, 0.0, 0.0, 0.0) as SWIZZLE_ZERO
+ * is implemented simply by not writing to the relevant components of the destination.
+ *
+ * Conditional execution
+ *   TODO
+ *
+ * Non-native instructions:
+ *   LIT
+ *   LRP - MAD+MAD
+ *   SUB - ADD, negate second source
+ *   RSQ - LG2 + EX2
+ *   POW - LG2 + MUL + EX2
+ *   SCS - COS + SIN
+ *   XPD
+ *
+ * NV40 Looping
+ *   Loops appear to be fairly expensive on NV40 at least, the proprietary
+ *   driver goes to a lot of effort to avoid using the native looping
+ *   instructions.  If the total number of *executed* instructions between
+ *   REP/ENDREP or LOOP/ENDLOOP is <=500, the driver will unroll the loop.
+ *   The maximum loop count is 255.
+ *
+ */
+
+//== Opcode / Destination selection ==
+#define NVFX_FP_OP_PROGRAM_END          (1 << 0)
+#define NVFX_FP_OP_OUT_REG_SHIFT        1
+#define NV30_FP_OP_OUT_REG_MASK          (31 << 1)  /* uncertain */
+#define NV40_FP_OP_OUT_REG_MASK          (63 << 1)
+/* Needs to be set when writing outputs to get expected result.. */
+#define NVFX_FP_OP_OUT_REG_HALF          (1 << 7)
+#define NVFX_FP_OP_COND_WRITE_ENABLE        (1 << 8)
+#define NVFX_FP_OP_OUTMASK_SHIFT        9
+#define NVFX_FP_OP_OUTMASK_MASK          (0xF << 9)
+#  define NVFX_FP_OP_OUT_X  (1<<9)
+#  define NVFX_FP_OP_OUT_Y  (1<<10)
+#  define NVFX_FP_OP_OUT_Z  (1<<11)
+#  define NVFX_FP_OP_OUT_W  (1<<12)
+/* Uncertain about these, especially the input_src values.. it's possible that
+ * they can be dynamically changed.
+ */
+#define NVFX_FP_OP_INPUT_SRC_SHIFT        13
+#define NVFX_FP_OP_INPUT_SRC_MASK        (15 << 13)
+#  define NVFX_FP_OP_INPUT_SRC_POSITION  0x0
+#  define NVFX_FP_OP_INPUT_SRC_COL0  0x1
+#  define NVFX_FP_OP_INPUT_SRC_COL1  0x2
+#  define NVFX_FP_OP_INPUT_SRC_FOGC  0x3
+#  define NVFX_FP_OP_INPUT_SRC_TC0    0x4
+#  define NVFX_FP_OP_INPUT_SRC_TC(n)  (0x4 + n)
+#  define NV40_FP_OP_INPUT_SRC_FACING  0xE
+#define NVFX_FP_OP_TEX_UNIT_SHIFT        17
+#define NVFX_FP_OP_TEX_UNIT_MASK        (0xF << 17) /* guess */
+#define NVFX_FP_OP_PRECISION_SHIFT        22
+#define NVFX_FP_OP_PRECISION_MASK        (3 << 22)
+#   define NVFX_FP_PRECISION_FP32  0
+#   define NVFX_FP_PRECISION_FP16  1
+#   define NVFX_FP_PRECISION_FX12  2
+#define NVFX_FP_OP_OPCODE_SHIFT          24
+#define NVFX_FP_OP_OPCODE_MASK          (0x3F << 24)
+/* NV30/NV40 fragment program opcodes */
+#define NVFX_FP_OP_OPCODE_NOP 0x00
+#define NVFX_FP_OP_OPCODE_MOV 0x01
+#define NVFX_FP_OP_OPCODE_MUL 0x02
+#define NVFX_FP_OP_OPCODE_ADD 0x03
+#define NVFX_FP_OP_OPCODE_MAD 0x04
+#define NVFX_FP_OP_OPCODE_DP3 0x05
+#define NVFX_FP_OP_OPCODE_DP4 0x06
+#define NVFX_FP_OP_OPCODE_DST 0x07
+#define NVFX_FP_OP_OPCODE_MIN 0x08
+#define NVFX_FP_OP_OPCODE_MAX 0x09
+#define NVFX_FP_OP_OPCODE_SLT 0x0A
+#define NVFX_FP_OP_OPCODE_SGE 0x0B
+#define NVFX_FP_OP_OPCODE_SLE 0x0C
+#define NVFX_FP_OP_OPCODE_SGT 0x0D
+#define NVFX_FP_OP_OPCODE_SNE 0x0E
+#define NVFX_FP_OP_OPCODE_SEQ 0x0F
+#define NVFX_FP_OP_OPCODE_FRC 0x10
+#define NVFX_FP_OP_OPCODE_FLR 0x11
+#define NVFX_FP_OP_OPCODE_KIL 0x12
+#define NVFX_FP_OP_OPCODE_PK4B 0x13
+#define NVFX_FP_OP_OPCODE_UP4B 0x14
+#define NVFX_FP_OP_OPCODE_DDX 0x15 /* can only write XY */
+#define NVFX_FP_OP_OPCODE_DDY 0x16 /* can only write XY */
+#define NVFX_FP_OP_OPCODE_TEX 0x17
+#define NVFX_FP_OP_OPCODE_TXP 0x18
+#define NVFX_FP_OP_OPCODE_TXD 0x19
+#define NVFX_FP_OP_OPCODE_RCP 0x1A
+#define NVFX_FP_OP_OPCODE_EX2 0x1C
+#define NVFX_FP_OP_OPCODE_LG2 0x1D
+#define NVFX_FP_OP_OPCODE_STR 0x20
+#define NVFX_FP_OP_OPCODE_SFL 0x21
+#define NVFX_FP_OP_OPCODE_COS 0x22
+#define NVFX_FP_OP_OPCODE_SIN 0x23
+#define NVFX_FP_OP_OPCODE_PK2H 0x24
+#define NVFX_FP_OP_OPCODE_UP2H 0x25
+#define NVFX_FP_OP_OPCODE_PK4UB 0x27
+#define NVFX_FP_OP_OPCODE_UP4UB 0x28
+#define NVFX_FP_OP_OPCODE_PK2US 0x29
+#define NVFX_FP_OP_OPCODE_UP2US 0x2A
+#define NVFX_FP_OP_OPCODE_DP2A 0x2E
+#define NVFX_FP_OP_OPCODE_TXB 0x31
+#define NVFX_FP_OP_OPCODE_DIV 0x3A
+
+/* NV30 only fragment program opcodes */
+#define NVFX_FP_OP_OPCODE_RSQ_NV30 0x1B
+#define NVFX_FP_OP_OPCODE_LIT_NV30 0x1E
+#define NVFX_FP_OP_OPCODE_LRP_NV30 0x1F
+#define NVFX_FP_OP_OPCODE_POW_NV30 0x26
+#define NVFX_FP_OP_OPCODE_RFL_NV30 0x36
+
+/* NV40 only fragment program opcodes */
+#define NVFX_FP_OP_OPCODE_TXL_NV40 0x31
+/* The use of these instructions appears to be indicated by bit 31 of DWORD 2.*/
+#define NV40_FP_OP_BRA_OPCODE_BRK                                    0x0
+#define NV40_FP_OP_BRA_OPCODE_CAL                                    0x1
+#define NV40_FP_OP_BRA_OPCODE_IF                                     0x2
+#define NV40_FP_OP_BRA_OPCODE_LOOP                                   0x3
+#define NV40_FP_OP_BRA_OPCODE_REP                                    0x4
+#define NV40_FP_OP_BRA_OPCODE_RET                                    0x5
+
+#define NVFX_FP_OP_OUT_SAT          (1 << 31)
+
+/* high order bits of SRC0 */
+#define NVFX_FP_OP_OUT_ABS          (1 << 29)
+#define NVFX_FP_OP_COND_SWZ_W_SHIFT        27
+#define NVFX_FP_OP_COND_SWZ_W_MASK        (3 << 27)
+#define NVFX_FP_OP_COND_SWZ_Z_SHIFT        25
+#define NVFX_FP_OP_COND_SWZ_Z_MASK        (3 << 25)
+#define NVFX_FP_OP_COND_SWZ_Y_SHIFT        23
+#define NVFX_FP_OP_COND_SWZ_Y_MASK        (3 << 23)
+#define NVFX_FP_OP_COND_SWZ_X_SHIFT        21
+#define NVFX_FP_OP_COND_SWZ_X_MASK        (3 << 21)
+#define NVFX_FP_OP_COND_SWZ_ALL_SHIFT        21
+#define NVFX_FP_OP_COND_SWZ_ALL_MASK        (0xFF << 21)
+#define NVFX_FP_OP_COND_SHIFT          18
+#define NVFX_FP_OP_COND_MASK          (0x07 << 18)
+#  define NVFX_FP_OP_COND_FL  0
+#  define NVFX_FP_OP_COND_LT  1
+#  define NVFX_FP_OP_COND_EQ  2
+#  define NVFX_FP_OP_COND_LE  3
+#  define NVFX_FP_OP_COND_GT  4
+#  define NVFX_FP_OP_COND_NE  5
+#  define NVFX_FP_OP_COND_GE  6
+#  define NVFX_FP_OP_COND_TR  7
+
+/* high order bits of SRC1 */
+#define NV40_FP_OP_OPCODE_IS_BRANCH                                      (1<<31)
+#define NVFX_FP_OP_DST_SCALE_SHIFT        28
+#define NVFX_FP_OP_DST_SCALE_MASK        (3 << 28)
+#define NVFX_FP_OP_DST_SCALE_1X                                                0
+#define NVFX_FP_OP_DST_SCALE_2X                                                1
+#define NVFX_FP_OP_DST_SCALE_4X                                                2
+#define NVFX_FP_OP_DST_SCALE_8X                                                3
+#define NVFX_FP_OP_DST_SCALE_INV_2X                                            5
+#define NVFX_FP_OP_DST_SCALE_INV_4X                                            6
+#define NVFX_FP_OP_DST_SCALE_INV_8X                                            7
+
+/* SRC1 LOOP */
+#define NV40_FP_OP_LOOP_INCR_SHIFT                                            19
+#define NV40_FP_OP_LOOP_INCR_MASK                                   (0xFF << 19)
+#define NV40_FP_OP_LOOP_INDEX_SHIFT                                           10
+#define NV40_FP_OP_LOOP_INDEX_MASK                                  (0xFF << 10)
+#define NV40_FP_OP_LOOP_COUNT_SHIFT                                            2
+#define NV40_FP_OP_LOOP_COUNT_MASK                                   (0xFF << 2)
+
+/* SRC1 IF */
+#define NV40_FP_OP_ELSE_ID_SHIFT                                               2
+#define NV40_FP_OP_ELSE_ID_MASK                                      (0xFF << 2)
+
+/* SRC1 CAL */
+#define NV40_FP_OP_IADDR_SHIFT                                                 2
+#define NV40_FP_OP_IADDR_MASK                                        (0xFF << 2)
+
+/* SRC1 REP
+ *   I have no idea why there are 3 count values here..  but they
+ *   have always been filled with the same value in my tests so
+ *   far..
+ */
+#define NV40_FP_OP_REP_COUNT1_SHIFT                                            2
+#define NV40_FP_OP_REP_COUNT1_MASK                                   (0xFF << 2)
+#define NV40_FP_OP_REP_COUNT2_SHIFT                                           10
+#define NV40_FP_OP_REP_COUNT2_MASK                                  (0xFF << 10)
+#define NV40_FP_OP_REP_COUNT3_SHIFT                                           19
+#define NV40_FP_OP_REP_COUNT3_MASK                                  (0xFF << 19)
+
+/* SRC2 REP/IF */
+#define NV40_FP_OP_END_ID_SHIFT                                                2
+#define NV40_FP_OP_END_ID_MASK                                       (0xFF << 2)
+
+/* high order bits of SRC2 */
+#define NVFX_FP_OP_INDEX_INPUT          (1 << 30)
+#define NV40_FP_OP_ADDR_INDEX_SHIFT        19
+#define NV40_FP_OP_ADDR_INDEX_MASK        (0xF << 19)
+
+//== Register selection ==
+#define NVFX_FP_REG_TYPE_SHIFT           0
+#define NVFX_FP_REG_TYPE_MASK           (3 << 0)
+#  define NVFX_FP_REG_TYPE_TEMP   0
+#  define NVFX_FP_REG_TYPE_INPUT  1
+#  define NVFX_FP_REG_TYPE_CONST  2
+#define NVFX_FP_REG_SRC_SHIFT            2
+#define NV30_FP_REG_SRC_MASK              (31 << 2)
+#define NV40_FP_REG_SRC_MASK              (63 << 2)
+#define NVFX_FP_REG_SRC_HALF            (1 << 8)
+#define NVFX_FP_REG_SWZ_ALL_SHIFT        9
+#define NVFX_FP_REG_SWZ_ALL_MASK        (255 << 9)
+#define NVFX_FP_REG_SWZ_X_SHIFT          9
+#define NVFX_FP_REG_SWZ_X_MASK          (3 << 9)
+#define NVFX_FP_REG_SWZ_Y_SHIFT          11
+#define NVFX_FP_REG_SWZ_Y_MASK          (3 << 11)
+#define NVFX_FP_REG_SWZ_Z_SHIFT          13
+#define NVFX_FP_REG_SWZ_Z_MASK          (3 << 13)
+#define NVFX_FP_REG_SWZ_W_SHIFT          15
+#define NVFX_FP_REG_SWZ_W_MASK          (3 << 15)
+#  define NVFX_FP_SWIZZLE_X  0
+#  define NVFX_FP_SWIZZLE_Y  1
+#  define NVFX_FP_SWIZZLE_Z  2
+#  define NVFX_FP_SWIZZLE_W  3
+#define NVFX_FP_REG_NEGATE          (1 << 17)
+
+#define NVFXSR_NONE	0
+#define NVFXSR_OUTPUT	1
+#define NVFXSR_INPUT	2
+#define NVFXSR_TEMP	3
+#define NVFXSR_CONST	4
+
+#define NVFX_COND_FL  0
+#define NVFX_COND_LT  1
+#define NVFX_COND_EQ  2
+#define NVFX_COND_LE  3
+#define NVFX_COND_GT  4
+#define NVFX_COND_NE  5
+#define NVFX_COND_GE  6
+#define NVFX_COND_TR  7
+
+/* Yes, this are ordered differently... */
+
+#define NVFX_VP_MASK_X 8
+#define NVFX_VP_MASK_Y 4
+#define NVFX_VP_MASK_Z 2
+#define NVFX_VP_MASK_W 1
+#define NVFX_VP_MASK_ALL 0xf
+
+#define NVFX_FP_MASK_X 1
+#define NVFX_FP_MASK_Y 2
+#define NVFX_FP_MASK_Z 4
+#define NVFX_FP_MASK_W 8
+#define NVFX_FP_MASK_ALL 0xf
+
+#define NVFX_SWZ_X 0
+#define NVFX_SWZ_Y 1
+#define NVFX_SWZ_Z 2
+#define NVFX_SWZ_W 3
+
+#define swz(s,x,y,z,w) nvfx_sr_swz((s), NVFX_SWZ_##x, NVFX_SWZ_##y, NVFX_SWZ_##z, NVFX_SWZ_##w)
+#define neg(s) nvfx_sr_neg((s))
+#define abs(s) nvfx_sr_abs((s))
+#define scale(s,v) nvfx_sr_scale((s), NVFX_FP_OP_DST_SCALE_##v)
+
+struct nvfx_sreg {
+	int type;
+	int index;
+
+	int dst_scale;
+
+	int negate;
+	int abs;
+	int swz[4];
+
+	int cc_update;
+	int cc_update_reg;
+	int cc_test;
+	int cc_test_reg;
+	int cc_swz[4];
+};
+
+static INLINE struct nvfx_sreg
+nvfx_sr(int type, int index)
+{
+	struct nvfx_sreg temp = {
+		.type = type,
+		.index = index,
+		.dst_scale = 0,
+		.abs = 0,
+		.negate = 0,
+		.swz = { 0, 1, 2, 3 },
+		.cc_update = 0,
+		.cc_update_reg = 0,
+		.cc_test = NVFX_COND_TR,
+		.cc_test_reg = 0,
+		.cc_swz = { 0, 1, 2, 3 },
+	};
+	return temp;
+}
+
+static INLINE struct nvfx_sreg
+nvfx_sr_swz(struct nvfx_sreg src, int x, int y, int z, int w)
+{
+	struct nvfx_sreg dst = src;
+
+	dst.swz[NVFX_SWZ_X] = src.swz[x];
+	dst.swz[NVFX_SWZ_Y] = src.swz[y];
+	dst.swz[NVFX_SWZ_Z] = src.swz[z];
+	dst.swz[NVFX_SWZ_W] = src.swz[w];
+	return dst;
+}
+
+static INLINE struct nvfx_sreg
+nvfx_sr_neg(struct nvfx_sreg src)
+{
+	src.negate = !src.negate;
+	return src;
+}
+
+static INLINE struct nvfx_sreg
+nvfx_sr_abs(struct nvfx_sreg src)
+{
+	src.abs = 1;
+	return src;
+}
+
+static INLINE struct nvfx_sreg
+nvfx_sr_scale(struct nvfx_sreg src, int scale)
+{
+	src.dst_scale = scale;
+	return src;
+}
+
+#endif
diff --git a/src/gallium/drivers/nvfx/nvfx_state.c b/src/gallium/drivers/nvfx/nvfx_state.c
new file mode 100644
index 0000000000..32a8199752
--- /dev/null
+++ b/src/gallium/drivers/nvfx/nvfx_state.c
@@ -0,0 +1,652 @@
+#include "pipe/p_state.h"
+#include "pipe/p_defines.h"
+#include "util/u_inlines.h"
+
+#include "draw/draw_context.h"
+
+#include "tgsi/tgsi_parse.h"
+
+#include "nvfx_context.h"
+#include "nvfx_state.h"
+#include "nvfx_tex.h"
+
+static void *
+nvfx_blend_state_create(struct pipe_context *pipe,
+			const struct pipe_blend_state *cso)
+{
+	struct nvfx_context *nvfx = nvfx_context(pipe);
+	struct nouveau_grobj *eng3d = nvfx->screen->eng3d;
+	struct nvfx_blend_state *bso = CALLOC(1, sizeof(*bso));
+	struct nouveau_stateobj *so = so_new(5, 8, 0);
+
+	if (cso->rt[0].blend_enable) {
+		so_method(so, eng3d, NV34TCL_BLEND_FUNC_ENABLE, 3);
+		so_data  (so, 1);
+		so_data  (so, (nvgl_blend_func(cso->rt[0].alpha_src_factor) << 16) |
+			       nvgl_blend_func(cso->rt[0].rgb_src_factor));
+		so_data  (so, nvgl_blend_func(cso->rt[0].alpha_dst_factor) << 16 |
+			      nvgl_blend_func(cso->rt[0].rgb_dst_factor));
+		if(nvfx->screen->base.device->chipset < 0x40) {
+			so_method(so, eng3d, NV34TCL_BLEND_EQUATION, 1);
+			so_data  (so, nvgl_blend_eqn(cso->rt[0].rgb_func));
+		} else {
+			so_method(so, eng3d, NV40TCL_BLEND_EQUATION, 1);
+			so_data  (so, nvgl_blend_eqn(cso->rt[0].alpha_func) << 16 |
+			      nvgl_blend_eqn(cso->rt[0].rgb_func));
+		}
+	} else {
+		so_method(so, eng3d, NV34TCL_BLEND_FUNC_ENABLE, 1);
+		so_data  (so, 0);
+	}
+
+	so_method(so, eng3d, NV34TCL_COLOR_MASK, 1);
+	so_data  (so, (((cso->rt[0].colormask & PIPE_MASK_A) ? (0x01 << 24) : 0) |
+	       ((cso->rt[0].colormask & PIPE_MASK_R) ? (0x01 << 16) : 0) |
+	       ((cso->rt[0].colormask & PIPE_MASK_G) ? (0x01 <<  8) : 0) |
+	       ((cso->rt[0].colormask & PIPE_MASK_B) ? (0x01 <<  0) : 0)));
+
+	/* TODO: add NV40 MRT color mask */
+
+	if (cso->logicop_enable) {
+		so_method(so, eng3d, NV34TCL_COLOR_LOGIC_OP_ENABLE, 2);
+		so_data  (so, 1);
+		so_data  (so, nvgl_logicop_func(cso->logicop_func));
+	} else {
+		so_method(so, eng3d, NV34TCL_COLOR_LOGIC_OP_ENABLE, 1);
+		so_data  (so, 0);
+	}
+
+	so_method(so, eng3d, NV34TCL_DITHER_ENABLE, 1);
+	so_data  (so, cso->dither ? 1 : 0);
+
+	so_ref(so, &bso->so);
+	so_ref(NULL, &so);
+	bso->pipe = *cso;
+	return (void *)bso;
+}
+
+static void
+nvfx_blend_state_bind(struct pipe_context *pipe, void *hwcso)
+{
+	struct nvfx_context *nvfx = nvfx_context(pipe);
+
+	nvfx->blend = hwcso;
+	nvfx->dirty |= NVFX_NEW_BLEND;
+}
+
+static void
+nvfx_blend_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	struct nvfx_blend_state *bso = hwcso;
+
+	so_ref(NULL, &bso->so);
+	FREE(bso);
+}
+
+static void *
+nvfx_sampler_state_create(struct pipe_context *pipe,
+			  const struct pipe_sampler_state *cso)
+{
+	struct nvfx_context *nvfx = nvfx_context(pipe);
+	struct nvfx_sampler_state *ps;
+
+	ps = MALLOC(sizeof(struct nvfx_sampler_state));
+
+	/* on nv30, we use this as an internal flag */
+	ps->fmt = cso->normalized_coords ? 0 : NV40TCL_TEX_FORMAT_RECT;
+	ps->en = 0;
+	ps->filt = nvfx_tex_filter(cso);
+	ps->wrap = (nvfx_tex_wrap_mode(cso->wrap_s) << NV34TCL_TX_WRAP_S_SHIFT) |
+		    (nvfx_tex_wrap_mode(cso->wrap_t) << NV34TCL_TX_WRAP_T_SHIFT) |
+		    (nvfx_tex_wrap_mode(cso->wrap_r) << NV34TCL_TX_WRAP_R_SHIFT) |
+		    nvfx_tex_wrap_compare_mode(cso);
+	ps->bcol = nvfx_tex_border_color(cso->border_color);
+
+	if(nvfx->is_nv4x)
+		nv40_sampler_state_init(pipe, ps, cso);
+	else
+		nv30_sampler_state_init(pipe, ps, cso);
+
+	return (void *)ps;
+}
+
+static void
+nvfx_sampler_state_bind(struct pipe_context *pipe, unsigned nr, void **sampler)
+{
+	struct nvfx_context *nvfx = nvfx_context(pipe);
+	unsigned unit;
+
+	for (unit = 0; unit < nr; unit++) {
+		nvfx->tex_sampler[unit] = sampler[unit];
+		nvfx->dirty_samplers |= (1 << unit);
+	}
+
+	for (unit = nr; unit < nvfx->nr_samplers; unit++) {
+		nvfx->tex_sampler[unit] = NULL;
+		nvfx->dirty_samplers |= (1 << unit);
+	}
+
+	nvfx->nr_samplers = nr;
+	nvfx->dirty |= NVFX_NEW_SAMPLER;
+}
+
+static void
+nvfx_sampler_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	FREE(hwcso);
+}
+
+static void
+nvfx_set_fragment_sampler_views(struct pipe_context *pipe,
+				unsigned nr,
+				struct pipe_sampler_view **views)
+{
+	struct nvfx_context *nvfx = nvfx_context(pipe);
+	unsigned unit;
+
+	for (unit = 0; unit < nr; unit++) {
+		pipe_sampler_view_reference(&nv30->fragment_sampler_views[unit],
+                                            views[unit]);
+		pipe_texture_reference((struct pipe_texture **)
+				       &nvfx->tex_miptree[unit], miptree[unit]);
+		nvfx->dirty_samplers |= (1 << unit);
+	}
+
+	for (unit = nr; unit < nvfx->nr_textures; unit++) {
+		pipe_sampler_view_reference(&nv30->fragment_sampler_views[unit],
+                                            NULL);
+		pipe_texture_reference((struct pipe_texture **)
+				       &nvfx->tex_miptree[unit], NULL);
+		nvfx->dirty_samplers |= (1 << unit);
+	}
+
+	nvfx->nr_textures = nr;
+	nvfx->dirty |= NVFX_NEW_SAMPLER;
+}
+
+
+static struct pipe_sampler_view *
+nv30_create_sampler_view(struct pipe_context *pipe,
+			 struct pipe_texture *texture,
+			 const struct pipe_sampler_view *templ)
+{
+	struct pipe_sampler_view *view = CALLOC_STRUCT(pipe_sampler_view);
+
+	if (view) {
+		*view = *templ;
+		view->reference.count = 1;
+		view->texture = NULL;
+		pipe_texture_reference(&view->texture, texture);
+		view->context = pipe;
+	}
+
+	return view;
+}
+
+
+static void
+nv30_sampler_view_destroy(struct pipe_context *pipe,
+			  struct pipe_sampler_view *view)
+{
+	pipe_texture_reference(&view->texture, NULL);
+	FREE(view);
+}
+
+static void *
+nvfx_rasterizer_state_create(struct pipe_context *pipe,
+			     const struct pipe_rasterizer_state *cso)
+{
+	struct nvfx_context *nvfx = nvfx_context(pipe);
+	struct nvfx_rasterizer_state *rsso = CALLOC(1, sizeof(*rsso));
+	struct nouveau_stateobj *so = so_new(9, 19, 0);
+	struct nouveau_grobj *eng3d = nvfx->screen->eng3d;
+
+	/*XXX: ignored:
+	 * 	light_twoside
+	 * 	point_smooth -nohw
+	 * 	multisample
+	 */
+
+	so_method(so, eng3d, NV34TCL_SHADE_MODEL, 1);
+	so_data  (so, cso->flatshade ? NV34TCL_SHADE_MODEL_FLAT :
+				       NV34TCL_SHADE_MODEL_SMOOTH);
+
+	so_method(so, eng3d, NV34TCL_LINE_WIDTH, 2);
+	so_data  (so, (unsigned char)(cso->line_width * 8.0) & 0xff);
+	so_data  (so, cso->line_smooth ? 1 : 0);
+	so_method(so, eng3d, NV34TCL_LINE_STIPPLE_ENABLE, 2);
+	so_data  (so, cso->line_stipple_enable ? 1 : 0);
+	so_data  (so, (cso->line_stipple_pattern << 16) |
+		       cso->line_stipple_factor);
+
+	so_method(so, eng3d, NV34TCL_POINT_SIZE, 1);
+	so_data  (so, fui(cso->point_size));
+
+	so_method(so, eng3d, NV34TCL_POLYGON_MODE_FRONT, 6);
+	if (cso->front_winding == PIPE_WINDING_CCW) {
+		so_data(so, nvgl_polygon_mode(cso->fill_ccw));
+		so_data(so, nvgl_polygon_mode(cso->fill_cw));
+		switch (cso->cull_mode) {
+		case PIPE_WINDING_CCW:
+			so_data(so, NV34TCL_CULL_FACE_FRONT);
+			break;
+		case PIPE_WINDING_CW:
+			so_data(so, NV34TCL_CULL_FACE_BACK);
+			break;
+		case PIPE_WINDING_BOTH:
+			so_data(so, NV34TCL_CULL_FACE_FRONT_AND_BACK);
+			break;
+		default:
+			so_data(so, NV34TCL_CULL_FACE_BACK);
+			break;
+		}
+		so_data(so, NV34TCL_FRONT_FACE_CCW);
+	} else {
+		so_data(so, nvgl_polygon_mode(cso->fill_cw));
+		so_data(so, nvgl_polygon_mode(cso->fill_ccw));
+		switch (cso->cull_mode) {
+		case PIPE_WINDING_CCW:
+			so_data(so, NV34TCL_CULL_FACE_BACK);
+			break;
+		case PIPE_WINDING_CW:
+			so_data(so, NV34TCL_CULL_FACE_FRONT);
+			break;
+		case PIPE_WINDING_BOTH:
+			so_data(so, NV34TCL_CULL_FACE_FRONT_AND_BACK);
+			break;
+		default:
+			so_data(so, NV34TCL_CULL_FACE_BACK);
+			break;
+		}
+		so_data(so, NV34TCL_FRONT_FACE_CW);
+	}
+	so_data(so, cso->poly_smooth ? 1 : 0);
+	so_data(so, (cso->cull_mode != PIPE_WINDING_NONE) ? 1 : 0);
+
+	so_method(so, eng3d, NV34TCL_POLYGON_STIPPLE_ENABLE, 1);
+	so_data  (so, cso->poly_stipple_enable ? 1 : 0);
+
+	so_method(so, eng3d, NV34TCL_POLYGON_OFFSET_POINT_ENABLE, 3);
+	if ((cso->offset_cw && cso->fill_cw == PIPE_POLYGON_MODE_POINT) ||
+	    (cso->offset_ccw && cso->fill_ccw == PIPE_POLYGON_MODE_POINT))
+		so_data(so, 1);
+	else
+		so_data(so, 0);
+	if ((cso->offset_cw && cso->fill_cw == PIPE_POLYGON_MODE_LINE) ||
+	    (cso->offset_ccw && cso->fill_ccw == PIPE_POLYGON_MODE_LINE))
+		so_data(so, 1);
+	else
+		so_data(so, 0);
+	if ((cso->offset_cw && cso->fill_cw == PIPE_POLYGON_MODE_FILL) ||
+	    (cso->offset_ccw && cso->fill_ccw == PIPE_POLYGON_MODE_FILL))
+		so_data(so, 1);
+	else
+		so_data(so, 0);
+	if (cso->offset_cw || cso->offset_ccw) {
+		so_method(so, eng3d, NV34TCL_POLYGON_OFFSET_FACTOR, 2);
+		so_data  (so, fui(cso->offset_scale));
+		so_data  (so, fui(cso->offset_units * 2));
+	}
+
+	so_method(so, eng3d, NV34TCL_POINT_SPRITE, 1);
+	if (cso->point_quad_rasterization) {
+		unsigned psctl = (1 << 0), i;
+
+		for (i = 0; i < 8; i++) {
+			if ((cso->sprite_coord_enable >> i) & 1)
+				psctl |= (1 << (8 + i));
+		}
+
+		so_data(so, psctl);
+	} else {
+		so_data(so, 0);
+	}
+
+	so_ref(so, &rsso->so);
+	so_ref(NULL, &so);
+	rsso->pipe = *cso;
+	return (void *)rsso;
+}
+
+static void
+nvfx_rasterizer_state_bind(struct pipe_context *pipe, void *hwcso)
+{
+	struct nvfx_context *nvfx = nvfx_context(pipe);
+
+	nvfx->rasterizer = hwcso;
+	nvfx->dirty |= NVFX_NEW_RAST;
+	nvfx->draw_dirty |= NVFX_NEW_RAST;
+}
+
+static void
+nvfx_rasterizer_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	struct nvfx_rasterizer_state *rsso = hwcso;
+
+	so_ref(NULL, &rsso->so);
+	FREE(rsso);
+}
+
+static void *
+nvfx_depth_stencil_alpha_state_create(struct pipe_context *pipe,
+			const struct pipe_depth_stencil_alpha_state *cso)
+{
+	struct nvfx_context *nvfx = nvfx_context(pipe);
+	struct nvfx_zsa_state *zsaso = CALLOC(1, sizeof(*zsaso));
+	struct nouveau_stateobj *so = so_new(6, 20, 0);
+	struct nouveau_grobj *eng3d = nvfx->screen->eng3d;
+
+	so_method(so, eng3d, NV34TCL_DEPTH_FUNC, 3);
+	so_data  (so, nvgl_comparison_op(cso->depth.func));
+	so_data  (so, cso->depth.writemask ? 1 : 0);
+	so_data  (so, cso->depth.enabled ? 1 : 0);
+
+	so_method(so, eng3d, NV34TCL_ALPHA_FUNC_ENABLE, 3);
+	so_data  (so, cso->alpha.enabled ? 1 : 0);
+	so_data  (so, nvgl_comparison_op(cso->alpha.func));
+	so_data  (so, float_to_ubyte(cso->alpha.ref_value));
+
+	if (cso->stencil[0].enabled) {
+		so_method(so, eng3d, NV34TCL_STENCIL_FRONT_ENABLE, 3);
+		so_data  (so, cso->stencil[0].enabled ? 1 : 0);
+		so_data  (so, cso->stencil[0].writemask);
+		so_data  (so, nvgl_comparison_op(cso->stencil[0].func));
+		so_method(so, eng3d, NV34TCL_STENCIL_FRONT_FUNC_MASK, 4);
+		so_data  (so, cso->stencil[0].valuemask);
+		so_data  (so, nvgl_stencil_op(cso->stencil[0].fail_op));
+		so_data  (so, nvgl_stencil_op(cso->stencil[0].zfail_op));
+		so_data  (so, nvgl_stencil_op(cso->stencil[0].zpass_op));
+	} else {
+		so_method(so, eng3d, NV34TCL_STENCIL_FRONT_ENABLE, 1);
+		so_data  (so, 0);
+	}
+
+	if (cso->stencil[1].enabled) {
+		so_method(so, eng3d, NV34TCL_STENCIL_BACK_ENABLE, 3);
+		so_data  (so, cso->stencil[1].enabled ? 1 : 0);
+		so_data  (so, cso->stencil[1].writemask);
+		so_data  (so, nvgl_comparison_op(cso->stencil[1].func));
+		so_method(so, eng3d, NV34TCL_STENCIL_BACK_FUNC_MASK, 4);
+		so_data  (so, cso->stencil[1].valuemask);
+		so_data  (so, nvgl_stencil_op(cso->stencil[1].fail_op));
+		so_data  (so, nvgl_stencil_op(cso->stencil[1].zfail_op));
+		so_data  (so, nvgl_stencil_op(cso->stencil[1].zpass_op));
+	} else {
+		so_method(so, eng3d, NV34TCL_STENCIL_BACK_ENABLE, 1);
+		so_data  (so, 0);
+	}
+
+	so_ref(so, &zsaso->so);
+	so_ref(NULL, &so);
+	zsaso->pipe = *cso;
+	return (void *)zsaso;
+}
+
+static void
+nvfx_depth_stencil_alpha_state_bind(struct pipe_context *pipe, void *hwcso)
+{
+	struct nvfx_context *nvfx = nvfx_context(pipe);
+
+	nvfx->zsa = hwcso;
+	nvfx->dirty |= NVFX_NEW_ZSA;
+}
+
+static void
+nvfx_depth_stencil_alpha_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	struct nvfx_zsa_state *zsaso = hwcso;
+
+	so_ref(NULL, &zsaso->so);
+	FREE(zsaso);
+}
+
+static void *
+nvfx_vp_state_create(struct pipe_context *pipe,
+		     const struct pipe_shader_state *cso)
+{
+	struct nvfx_context *nvfx = nvfx_context(pipe);
+	struct nvfx_vertex_program *vp;
+
+	vp = CALLOC(1, sizeof(struct nvfx_vertex_program));
+	vp->pipe.tokens = tgsi_dup_tokens(cso->tokens);
+	vp->draw = draw_create_vertex_shader(nvfx->draw, &vp->pipe);
+
+	return (void *)vp;
+}
+
+static void
+nvfx_vp_state_bind(struct pipe_context *pipe, void *hwcso)
+{
+	struct nvfx_context *nvfx = nvfx_context(pipe);
+
+	nvfx->vertprog = hwcso;
+	nvfx->dirty |= NVFX_NEW_VERTPROG;
+	nvfx->draw_dirty |= NVFX_NEW_VERTPROG;
+}
+
+static void
+nvfx_vp_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	struct nvfx_context *nvfx = nvfx_context(pipe);
+	struct nvfx_vertex_program *vp = hwcso;
+
+	draw_delete_vertex_shader(nvfx->draw, vp->draw);
+	nvfx_vertprog_destroy(nvfx, vp);
+	FREE((void*)vp->pipe.tokens);
+	FREE(vp);
+}
+
+static void *
+nvfx_fp_state_create(struct pipe_context *pipe,
+		     const struct pipe_shader_state *cso)
+{
+	struct nvfx_fragment_program *fp;
+
+	fp = CALLOC(1, sizeof(struct nvfx_fragment_program));
+	fp->pipe.tokens = tgsi_dup_tokens(cso->tokens);
+
+	tgsi_scan_shader(fp->pipe.tokens, &fp->info);
+
+	return (void *)fp;
+}
+
+static void
+nvfx_fp_state_bind(struct pipe_context *pipe, void *hwcso)
+{
+	struct nvfx_context *nvfx = nvfx_context(pipe);
+
+	nvfx->fragprog = hwcso;
+	nvfx->dirty |= NVFX_NEW_FRAGPROG;
+}
+
+static void
+nvfx_fp_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	struct nvfx_context *nvfx = nvfx_context(pipe);
+	struct nvfx_fragment_program *fp = hwcso;
+
+	nvfx_fragprog_destroy(nvfx, fp);
+	FREE((void*)fp->pipe.tokens);
+	FREE(fp);
+}
+
+static void
+nvfx_set_blend_color(struct pipe_context *pipe,
+		     const struct pipe_blend_color *bcol)
+{
+	struct nvfx_context *nvfx = nvfx_context(pipe);
+
+	nvfx->blend_colour = *bcol;
+	nvfx->dirty |= NVFX_NEW_BCOL;
+}
+
+static void
+nvfx_set_stencil_ref(struct pipe_context *pipe,
+		     const struct pipe_stencil_ref *sr)
+{
+	struct nvfx_context *nvfx = nvfx_context(pipe);
+
+	nvfx->stencil_ref = *sr;
+	nvfx->dirty |= NVFX_NEW_SR;
+}
+
+static void
+nvfx_set_clip_state(struct pipe_context *pipe,
+		    const struct pipe_clip_state *clip)
+{
+	struct nvfx_context *nvfx = nvfx_context(pipe);
+
+	nvfx->clip = *clip;
+	nvfx->dirty |= NVFX_NEW_UCP;
+	nvfx->draw_dirty |= NVFX_NEW_UCP;
+}
+
+static void
+nvfx_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index,
+			 struct pipe_buffer *buf )
+{
+	struct nvfx_context *nvfx = nvfx_context(pipe);
+
+	nvfx->constbuf[shader] = buf;
+	nvfx->constbuf_nr[shader] = buf->size / (4 * sizeof(float));
+
+	if (shader == PIPE_SHADER_VERTEX) {
+		nvfx->dirty |= NVFX_NEW_VERTPROG;
+	} else
+	if (shader == PIPE_SHADER_FRAGMENT) {
+		nvfx->dirty |= NVFX_NEW_FRAGPROG;
+	}
+}
+
+static void
+nvfx_set_framebuffer_state(struct pipe_context *pipe,
+			   const struct pipe_framebuffer_state *fb)
+{
+	struct nvfx_context *nvfx = nvfx_context(pipe);
+
+	nvfx->framebuffer = *fb;
+	nvfx->dirty |= NVFX_NEW_FB;
+}
+
+static void
+nvfx_set_polygon_stipple(struct pipe_context *pipe,
+			 const struct pipe_poly_stipple *stipple)
+{
+	struct nvfx_context *nvfx = nvfx_context(pipe);
+
+	memcpy(nvfx->stipple, stipple->stipple, 4 * 32);
+	nvfx->dirty |= NVFX_NEW_STIPPLE;
+}
+
+static void
+nvfx_set_scissor_state(struct pipe_context *pipe,
+		       const struct pipe_scissor_state *s)
+{
+	struct nvfx_context *nvfx = nvfx_context(pipe);
+
+	nvfx->scissor = *s;
+	nvfx->dirty |= NVFX_NEW_SCISSOR;
+}
+
+static void
+nvfx_set_viewport_state(struct pipe_context *pipe,
+			const struct pipe_viewport_state *vpt)
+{
+	struct nvfx_context *nvfx = nvfx_context(pipe);
+
+	nvfx->viewport = *vpt;
+	nvfx->dirty |= NVFX_NEW_VIEWPORT;
+	nvfx->draw_dirty |= NVFX_NEW_VIEWPORT;
+}
+
+static void
+nvfx_set_vertex_buffers(struct pipe_context *pipe, unsigned count,
+			const struct pipe_vertex_buffer *vb)
+{
+	struct nvfx_context *nvfx = nvfx_context(pipe);
+
+	memcpy(nvfx->vtxbuf, vb, sizeof(*vb) * count);
+	nvfx->vtxbuf_nr = count;
+
+	nvfx->dirty |= NVFX_NEW_ARRAYS;
+	nvfx->draw_dirty |= NVFX_NEW_ARRAYS;
+}
+
+static void *
+nvfx_vtxelts_state_create(struct pipe_context *pipe,
+			  unsigned num_elements,
+			  const struct pipe_vertex_element *elements)
+{
+	struct nvfx_vtxelt_state *cso = CALLOC_STRUCT(nvfx_vtxelt_state);
+
+	assert(num_elements < 16); /* not doing fallbacks yet */
+	cso->num_elements = num_elements;
+	memcpy(cso->pipe, elements, num_elements * sizeof(*elements));
+
+/*	nvfx_vtxelt_construct(cso);*/
+
+	return (void *)cso;
+}
+
+static void
+nvfx_vtxelts_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	FREE(hwcso);
+}
+
+static void
+nvfx_vtxelts_state_bind(struct pipe_context *pipe, void *hwcso)
+{
+	struct nvfx_context *nvfx = nvfx_context(pipe);
+
+	nvfx->vtxelt = hwcso;
+	nvfx->dirty |= NVFX_NEW_ARRAYS;
+	/*nvfx->draw_dirty |= NVFX_NEW_ARRAYS;*/
+}
+
+void
+nvfx_init_state_functions(struct nvfx_context *nvfx)
+{
+	nvfx->pipe.create_blend_state = nvfx_blend_state_create;
+	nvfx->pipe.bind_blend_state = nvfx_blend_state_bind;
+	nvfx->pipe.delete_blend_state = nvfx_blend_state_delete;
+
+	nvfx->pipe.create_sampler_state = nvfx_sampler_state_create;
+	nvfx->pipe.bind_fragment_sampler_states = nvfx_sampler_state_bind;
+	nvfx->pipe.delete_sampler_state = nvfx_sampler_state_delete;
+	nvfx->pipe.set_fragment_sampler_textures = nvfx_set_sampler_texture;
+
+	nvfx->pipe.create_rasterizer_state = nvfx_rasterizer_state_create;
+	nvfx->pipe.bind_rasterizer_state = nvfx_rasterizer_state_bind;
+	nvfx->pipe.delete_rasterizer_state = nvfx_rasterizer_state_delete;
+
+	nvfx->pipe.create_depth_stencil_alpha_state =
+		nvfx_depth_stencil_alpha_state_create;
+	nvfx->pipe.bind_depth_stencil_alpha_state =
+		nvfx_depth_stencil_alpha_state_bind;
+	nvfx->pipe.delete_depth_stencil_alpha_state =
+		nvfx_depth_stencil_alpha_state_delete;
+
+	nvfx->pipe.create_vs_state = nvfx_vp_state_create;
+	nvfx->pipe.bind_vs_state = nvfx_vp_state_bind;
+	nvfx->pipe.delete_vs_state = nvfx_vp_state_delete;
+
+	nvfx->pipe.create_fs_state = nvfx_fp_state_create;
+	nvfx->pipe.bind_fs_state = nvfx_fp_state_bind;
+	nvfx->pipe.delete_fs_state = nvfx_fp_state_delete;
+
+	nvfx->pipe.set_blend_color = nvfx_set_blend_color;
+        nvfx->pipe.set_stencil_ref = nvfx_set_stencil_ref;
+	nvfx->pipe.set_clip_state = nvfx_set_clip_state;
+	nvfx->pipe.set_constant_buffer = nvfx_set_constant_buffer;
+	nvfx->pipe.set_framebuffer_state = nvfx_set_framebuffer_state;
+	nvfx->pipe.set_polygon_stipple = nvfx_set_polygon_stipple;
+	nvfx->pipe.set_scissor_state = nvfx_set_scissor_state;
+	nvfx->pipe.set_viewport_state = nvfx_set_viewport_state;
+
+	nvfx->pipe.create_vertex_elements_state = nvfx_vtxelts_state_create;
+	nvfx->pipe.delete_vertex_elements_state = nvfx_vtxelts_state_delete;
+	nvfx->pipe.bind_vertex_elements_state = nvfx_vtxelts_state_bind;
+
+	nvfx->pipe.set_vertex_buffers = nvfx_set_vertex_buffers;
+}
diff --git a/src/gallium/drivers/nv40/nv40_state.h b/src/gallium/drivers/nvfx/nvfx_state.h
index e2e69420ea..e585246879 100644
--- a/src/gallium/drivers/nv40/nv40_state.h
+++ b/src/gallium/drivers/nvfx/nvfx_state.h
@@ -1,29 +1,21 @@
-#ifndef __NV40_STATE_H__
-#define __NV40_STATE_H__
+#ifndef __NVFX_STATE_H__
+#define __NVFX_STATE_H__
 
 #include "pipe/p_state.h"
 #include "tgsi/tgsi_scan.h"
 
-struct nv40_sampler_state {
-	uint32_t fmt;
-	uint32_t wrap;
-	uint32_t en;
-	uint32_t filt;
-	uint32_t bcol;
-};
-
-struct nv40_vertex_program_exec {
+struct nvfx_vertex_program_exec {
 	uint32_t data[4];
 	boolean has_branch_offset;
 	int const_index;
 };
 
-struct nv40_vertex_program_data {
+struct nvfx_vertex_program_data {
 	int index; /* immediates == -1 */
 	float value[4];
 };
 
-struct nv40_vertex_program {
+struct nvfx_vertex_program {
 	struct pipe_shader_state pipe;
 
 	struct draw_vertex_shader *draw;
@@ -32,9 +24,9 @@ struct nv40_vertex_program {
 
 	struct pipe_clip_state ucp;
 
-	struct nv40_vertex_program_exec *insns;
+	struct nvfx_vertex_program_exec *insns;
 	unsigned nr_insns;
-	struct nv40_vertex_program_data *consts;
+	struct nvfx_vertex_program_data *consts;
 	unsigned nr_consts;
 
 	struct nouveau_resource *exec;
@@ -49,12 +41,12 @@ struct nv40_vertex_program {
 	struct nouveau_stateobj *so;
 };
 
-struct nv40_fragment_program_data {
+struct nvfx_fragment_program_data {
 	unsigned offset;
 	unsigned index;
 };
 
-struct nv40_fragment_program {
+struct nvfx_fragment_program {
 	struct pipe_shader_state pipe;
 	struct tgsi_shader_info info;
 
@@ -64,7 +56,7 @@ struct nv40_fragment_program {
 	uint32_t *insn;
 	int       insn_len;
 
-	struct nv40_fragment_program_data *consts;
+	struct nvfx_fragment_program_data *consts;
 	unsigned nr_consts;
 
 	struct pipe_buffer *buffer;
@@ -73,9 +65,9 @@ struct nv40_fragment_program {
 	struct nouveau_stateobj *so;
 };
 
-#define NV40_MAX_TEXTURE_LEVELS  16
+#define NVFX_MAX_TEXTURE_LEVELS  16
 
-struct nv40_miptree {
+struct nvfx_miptree {
 	struct pipe_texture base;
 	struct nouveau_bo *bo;
 
@@ -85,7 +77,7 @@ struct nv40_miptree {
 	struct {
 		uint pitch;
 		uint *image_offset;
-	} level[NV40_MAX_TEXTURE_LEVELS];
+	} level[NVFX_MAX_TEXTURE_LEVELS];
 };
 
 #endif
diff --git a/src/gallium/drivers/nvfx/nvfx_state_blend.c b/src/gallium/drivers/nvfx/nvfx_state_blend.c
new file mode 100644
index 0000000000..03b6ef8117
--- /dev/null
+++ b/src/gallium/drivers/nvfx/nvfx_state_blend.c
@@ -0,0 +1,41 @@
+#include "nvfx_context.h"
+
+static boolean
+nvfx_state_blend_validate(struct nvfx_context *nvfx)
+{
+	so_ref(nvfx->blend->so, &nvfx->state.hw[NVFX_STATE_BLEND]);
+	return TRUE;
+}
+
+struct nvfx_state_entry nvfx_state_blend = {
+	.validate = nvfx_state_blend_validate,
+	.dirty = {
+		.pipe = NVFX_NEW_BLEND,
+		.hw = NVFX_STATE_BLEND
+	}
+};
+
+static boolean
+nvfx_state_blend_colour_validate(struct nvfx_context *nvfx)
+{
+	struct nouveau_stateobj *so = so_new(1, 1, 0);
+	struct pipe_blend_color *bcol = &nvfx->blend_colour;
+
+	so_method(so, nvfx->screen->eng3d, NV34TCL_BLEND_COLOR, 1);
+	so_data  (so, ((float_to_ubyte(bcol->color[3]) << 24) |
+		       (float_to_ubyte(bcol->color[0]) << 16) |
+		       (float_to_ubyte(bcol->color[1]) <<  8) |
+		       (float_to_ubyte(bcol->color[2]) <<  0)));
+
+	so_ref(so, &nvfx->state.hw[NVFX_STATE_BCOL]);
+	so_ref(NULL, &so);
+	return TRUE;
+}
+
+struct nvfx_state_entry nvfx_state_blend_colour = {
+	.validate = nvfx_state_blend_colour_validate,
+	.dirty = {
+		.pipe = NVFX_NEW_BCOL,
+		.hw = NVFX_STATE_BCOL
+	}
+};
diff --git a/src/gallium/drivers/nvfx/nvfx_state_emit.c b/src/gallium/drivers/nvfx/nvfx_state_emit.c
new file mode 100644
index 0000000000..72537388ea
--- /dev/null
+++ b/src/gallium/drivers/nvfx/nvfx_state_emit.c
@@ -0,0 +1,179 @@
+#include "nvfx_context.h"
+#include "nvfx_state.h"
+#include "draw/draw_context.h"
+
+#define RENDER_STATES(name, vbo) \
+static struct nvfx_state_entry *name##render_states[] = { \
+	&nvfx_state_framebuffer, \
+	&nvfx_state_rasterizer, \
+	&nvfx_state_scissor, \
+	&nvfx_state_stipple, \
+	&nvfx_state_fragprog, \
+	&nvfx_state_fragtex, \
+	&nvfx_state_vertprog, \
+	&nvfx_state_blend, \
+	&nvfx_state_blend_colour, \
+	&nvfx_state_zsa, \
+	&nvfx_state_sr, \
+	&nvfx_state_viewport, \
+	&nvfx_state_##vbo, \
+	NULL \
+}
+
+RENDER_STATES(, vbo);
+RENDER_STATES(swtnl_, vtxfmt);
+
+static void
+nvfx_state_do_validate(struct nvfx_context *nvfx,
+		       struct nvfx_state_entry **states)
+{
+	while (*states) {
+		struct nvfx_state_entry *e = *states;
+
+		if (nvfx->dirty & e->dirty.pipe) {
+			if (e->validate(nvfx))
+				nvfx->state.dirty |= (1ULL << e->dirty.hw);
+		}
+
+		states++;
+	}
+	nvfx->dirty = 0;
+}
+
+void
+nvfx_state_emit(struct nvfx_context *nvfx)
+{
+	struct nvfx_state *state = &nvfx->state;
+	struct nvfx_screen *screen = nvfx->screen;
+	struct nouveau_channel *chan = screen->base.channel;
+	struct nouveau_grobj *eng3d = screen->eng3d;
+	unsigned i;
+	uint64_t states;
+
+	/* XXX: race conditions
+	 */
+	if (nvfx != screen->cur_ctx) {
+		for (i = 0; i < NVFX_STATE_MAX; i++) {
+			if (state->hw[i] && screen->state[i] != state->hw[i])
+				state->dirty |= (1ULL << i);
+		}
+
+		screen->cur_ctx = nvfx;
+	}
+
+	for (i = 0, states = state->dirty; states; i++) {
+		if (!(states & (1ULL << i)))
+			continue;
+		so_ref (state->hw[i], &nvfx->screen->state[i]);
+		if (state->hw[i])
+			so_emit(chan, nvfx->screen->state[i]);
+		states &= ~(1ULL << i);
+	}
+
+	/* TODO: could nv30 need this or something similar too? */
+	if(nvfx->is_nv4x) {
+		if (state->dirty & ((1ULL << NVFX_STATE_FRAGPROG) |
+				    (1ULL << NVFX_STATE_FRAGTEX0))) {
+			BEGIN_RING(chan, eng3d, NV40TCL_TEX_CACHE_CTL, 1);
+			OUT_RING  (chan, 2);
+			BEGIN_RING(chan, eng3d, NV40TCL_TEX_CACHE_CTL, 1);
+			OUT_RING  (chan, 1);
+		}
+	}
+	state->dirty = 0;
+}
+
+void
+nvfx_state_flush_notify(struct nouveau_channel *chan)
+{
+	struct nvfx_context *nvfx = chan->user_private;
+	struct nvfx_state *state = &nvfx->state;
+	unsigned i, samplers;
+
+	so_emit_reloc_markers(chan, state->hw[NVFX_STATE_FB]);
+	for (i = 0, samplers = state->fp_samplers; i < 16 && samplers; i++) {
+		if (!(samplers & (1 << i)))
+			continue;
+		so_emit_reloc_markers(chan,
+				      state->hw[NVFX_STATE_FRAGTEX0+i]);
+		samplers &= ~(1ULL << i);
+	}
+	so_emit_reloc_markers(chan, state->hw[NVFX_STATE_FRAGPROG]);
+	if (state->hw[NVFX_STATE_VTXBUF] && nvfx->render_mode == HW)
+		so_emit_reloc_markers(chan, state->hw[NVFX_STATE_VTXBUF]);
+}
+
+boolean
+nvfx_state_validate(struct nvfx_context *nvfx)
+{
+	boolean was_sw = nvfx->fallback_swtnl ? TRUE : FALSE;
+
+	if (nvfx->render_mode != HW) {
+		/* Don't even bother trying to go back to hw if none
+		 * of the states that caused swtnl previously have changed.
+		 */
+		if ((nvfx->fallback_swtnl & nvfx->dirty)
+				!= nvfx->fallback_swtnl)
+			return FALSE;
+
+		/* Attempt to go to hwtnl again */
+		nvfx->pipe.flush(&nvfx->pipe, 0, NULL);
+		nvfx->dirty |= (NVFX_NEW_VIEWPORT |
+				NVFX_NEW_VERTPROG |
+				NVFX_NEW_ARRAYS);
+		nvfx->render_mode = HW;
+	}
+
+	nvfx_state_do_validate(nvfx, render_states);
+
+	if (nvfx->fallback_swtnl || nvfx->fallback_swrast)
+		return FALSE;
+
+	if (was_sw)
+		NOUVEAU_ERR("swtnl->hw\n");
+
+	return TRUE;
+}
+
+boolean
+nvfx_state_validate_swtnl(struct nvfx_context *nvfx)
+{
+	struct draw_context *draw = nvfx->draw;
+
+	/* Setup for swtnl */
+	if (nvfx->render_mode == HW) {
+		NOUVEAU_ERR("hw->swtnl 0x%08x\n", nvfx->fallback_swtnl);
+		nvfx->pipe.flush(&nvfx->pipe, 0, NULL);
+		nvfx->dirty |= (NVFX_NEW_VIEWPORT |
+				NVFX_NEW_VERTPROG |
+				NVFX_NEW_ARRAYS);
+		nvfx->render_mode = SWTNL;
+	}
+
+	if (nvfx->draw_dirty & NVFX_NEW_VERTPROG)
+		draw_bind_vertex_shader(draw, nvfx->vertprog->draw);
+
+	if (nvfx->draw_dirty & NVFX_NEW_RAST)
+		draw_set_rasterizer_state(draw, &nvfx->rasterizer->pipe);
+
+	if (nvfx->draw_dirty & NVFX_NEW_UCP)
+		draw_set_clip_state(draw, &nvfx->clip);
+
+	if (nvfx->draw_dirty & NVFX_NEW_VIEWPORT)
+		draw_set_viewport_state(draw, &nvfx->viewport);
+
+	if (nvfx->draw_dirty & NVFX_NEW_ARRAYS) {
+		draw_set_vertex_buffers(draw, nvfx->vtxbuf_nr, nvfx->vtxbuf);
+		draw_set_vertex_elements(draw, nvfx->vtxelt->num_elements, nvfx->vtxelt->pipe);
+	}
+
+	nvfx_state_do_validate(nvfx, swtnl_render_states);
+
+	if (nvfx->fallback_swrast) {
+		NOUVEAU_ERR("swtnl->swrast 0x%08x\n", nvfx->fallback_swrast);
+		return FALSE;
+	}
+
+	nvfx->draw_dirty = 0;
+	return TRUE;
+}
diff --git a/src/gallium/drivers/nvfx/nvfx_state_fb.c b/src/gallium/drivers/nvfx/nvfx_state_fb.c
new file mode 100644
index 0000000000..dd64ba4193
--- /dev/null
+++ b/src/gallium/drivers/nvfx/nvfx_state_fb.c
@@ -0,0 +1,234 @@
+#include "nvfx_context.h"
+#include "nouveau/nouveau_util.h"
+
+static struct pipe_buffer *
+nvfx_do_surface_buffer(struct pipe_surface *surface)
+{
+	struct nvfx_miptree *mt = (struct nvfx_miptree *)surface->texture;
+	return mt->buffer;
+}
+
+#define nvfx_surface_buffer(ps) nouveau_bo(nvfx_do_surface_buffer(ps))
+
+static boolean
+nvfx_state_framebuffer_validate(struct nvfx_context *nvfx)
+{
+	struct pipe_framebuffer_state *fb = &nvfx->framebuffer;
+	struct nouveau_channel *chan = nvfx->screen->base.channel;
+	struct nouveau_grobj *eng3d = nvfx->screen->eng3d;
+	struct nv04_surface *rt[4], *zeta = NULL;
+	uint32_t rt_enable = 0, rt_format = 0;
+	int i, colour_format = 0, zeta_format = 0;
+	int depth_only = 0;
+	struct nouveau_stateobj *so = so_new(18, 24, 10);
+	unsigned rt_flags = NOUVEAU_BO_RDWR | NOUVEAU_BO_VRAM;
+	unsigned w = fb->width;
+	unsigned h = fb->height;
+	int colour_bits = 32, zeta_bits = 32;
+
+	if(!nvfx->is_nv4x)
+		assert(fb->nr_cbufs <= 2);
+	else
+		assert(fb->nr_cbufs <= 4);
+
+	for (i = 0; i < fb->nr_cbufs; i++) {
+		if (colour_format) {
+			assert(colour_format == fb->cbufs[i]->format);
+		} else {
+			colour_format = fb->cbufs[i]->format;
+			rt_enable |= (NV34TCL_RT_ENABLE_COLOR0 << i);
+			rt[i] = (struct nv04_surface *)fb->cbufs[i];
+		}
+	}
+
+	if (rt_enable & (NV34TCL_RT_ENABLE_COLOR1 |
+			 NV40TCL_RT_ENABLE_COLOR2 | NV40TCL_RT_ENABLE_COLOR3))
+		rt_enable |= NV34TCL_RT_ENABLE_MRT;
+
+	if (fb->zsbuf) {
+		zeta_format = fb->zsbuf->format;
+		zeta = (struct nv04_surface *)fb->zsbuf;
+	}
+
+	if (rt_enable & (NV34TCL_RT_ENABLE_COLOR0 | NV34TCL_RT_ENABLE_COLOR1 |
+		NV40TCL_RT_ENABLE_COLOR2 | NV40TCL_RT_ENABLE_COLOR3)) {
+		/* Render to at least a colour buffer */
+		if (!(rt[0]->base.texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) {
+			assert(!(fb->width & (fb->width - 1)) && !(fb->height & (fb->height - 1)));
+			for (i = 1; i < fb->nr_cbufs; i++)
+				assert(!(rt[i]->base.texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR));
+
+			rt_format = NV34TCL_RT_FORMAT_TYPE_SWIZZLED |
+				(log2i(rt[0]->base.width) << NV34TCL_RT_FORMAT_LOG2_WIDTH_SHIFT) |
+				(log2i(rt[0]->base.height) << NV34TCL_RT_FORMAT_LOG2_HEIGHT_SHIFT);
+		}
+		else
+			rt_format = NV34TCL_RT_FORMAT_TYPE_LINEAR;
+	} else if (fb->zsbuf) {
+		depth_only = 1;
+
+		/* Render to depth buffer only */
+		if (!(zeta->base.texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) {
+			assert(!(fb->width & (fb->width - 1)) && !(fb->height & (fb->height - 1)));
+
+			rt_format = NV34TCL_RT_FORMAT_TYPE_SWIZZLED |
+				(log2i(zeta->base.width) << NV34TCL_RT_FORMAT_LOG2_WIDTH_SHIFT) |
+				(log2i(zeta->base.height) << NV34TCL_RT_FORMAT_LOG2_HEIGHT_SHIFT);
+		}
+		else
+			rt_format = NV34TCL_RT_FORMAT_TYPE_LINEAR;
+	} else {
+		return FALSE;
+	}
+
+	switch (colour_format) {
+	case PIPE_FORMAT_B8G8R8X8_UNORM:
+		rt_format |= NV34TCL_RT_FORMAT_COLOR_X8R8G8B8;
+		break;
+	case PIPE_FORMAT_B8G8R8A8_UNORM:
+	case 0:
+		rt_format |= NV34TCL_RT_FORMAT_COLOR_A8R8G8B8;
+		break;
+	case PIPE_FORMAT_B5G6R5_UNORM:
+		rt_format |= NV34TCL_RT_FORMAT_COLOR_R5G6B5;
+		colour_bits = 16;
+		break;
+	default:
+		assert(0);
+	}
+
+	switch (zeta_format) {
+	case PIPE_FORMAT_Z16_UNORM:
+		rt_format |= NV34TCL_RT_FORMAT_ZETA_Z16;
+		zeta_bits = 16;
+		break;
+	case PIPE_FORMAT_S8Z24_UNORM:
+	case PIPE_FORMAT_X8Z24_UNORM:
+	case 0:
+		rt_format |= NV34TCL_RT_FORMAT_ZETA_Z24S8;
+		break;
+	default:
+		assert(0);
+	}
+
+	if ((!nvfx->is_nv4x) && colour_bits > zeta_bits) {
+		/* TODO: does this limitation really exist?
+		   TODO: can it be worked around somehow? */
+		return FALSE;
+	}
+
+	if ((rt_enable & NV34TCL_RT_ENABLE_COLOR0)
+		|| ((!nvfx->is_nv4x) && depth_only)) {
+		struct nv04_surface *rt0 = (depth_only ? zeta : rt[0]);
+		uint32_t pitch = rt0->pitch;
+
+		if(!nvfx->is_nv4x)
+		{
+			if (zeta) {
+				pitch |= (zeta->pitch << 16);
+			} else {
+				pitch |= (pitch << 16);
+			}
+		}
+
+		so_method(so, eng3d, NV34TCL_DMA_COLOR0, 1);
+		so_reloc (so, nvfx_surface_buffer(&rt0->base), 0,
+			      rt_flags | NOUVEAU_BO_OR,
+			      chan->vram->handle, chan->gart->handle);
+		so_method(so, eng3d, NV34TCL_COLOR0_PITCH, 2);
+		so_data  (so, pitch);
+		so_reloc (so, nvfx_surface_buffer(&rt[0]->base),
+			      rt0->base.offset, rt_flags | NOUVEAU_BO_LOW,
+			      0, 0);
+	}
+
+	if (rt_enable & NV34TCL_RT_ENABLE_COLOR1) {
+		so_method(so, eng3d, NV34TCL_DMA_COLOR1, 1);
+		so_reloc (so, nvfx_surface_buffer(&rt[1]->base), 0,
+			      rt_flags | NOUVEAU_BO_OR,
+			      chan->vram->handle, chan->gart->handle);
+		so_method(so, eng3d, NV34TCL_COLOR1_OFFSET, 2);
+		so_reloc (so, nvfx_surface_buffer(&rt[1]->base),
+			      rt[1]->base.offset, rt_flags | NOUVEAU_BO_LOW,
+			      0, 0);
+		so_data  (so, rt[1]->pitch);
+	}
+
+	if(nvfx->is_nv4x)
+	{
+		if (rt_enable & NV40TCL_RT_ENABLE_COLOR2) {
+			so_method(so, eng3d, NV40TCL_DMA_COLOR2, 1);
+			so_reloc (so, nvfx_surface_buffer(&rt[2]->base), 0,
+				      rt_flags | NOUVEAU_BO_OR,
+				      chan->vram->handle, chan->gart->handle);
+			so_method(so, eng3d, NV40TCL_COLOR2_OFFSET, 1);
+			so_reloc (so, nvfx_surface_buffer(&rt[2]->base),
+				      rt[2]->base.offset, rt_flags | NOUVEAU_BO_LOW,
+				      0, 0);
+			so_method(so, eng3d, NV40TCL_COLOR2_PITCH, 1);
+			so_data  (so, rt[2]->pitch);
+		}
+
+		if (rt_enable & NV40TCL_RT_ENABLE_COLOR3) {
+			so_method(so, eng3d, NV40TCL_DMA_COLOR3, 1);
+			so_reloc (so, nvfx_surface_buffer(&rt[3]->base), 0,
+				      rt_flags | NOUVEAU_BO_OR,
+				      chan->vram->handle, chan->gart->handle);
+			so_method(so, eng3d, NV40TCL_COLOR3_OFFSET, 1);
+			so_reloc (so, nvfx_surface_buffer(&rt[3]->base),
+				      rt[3]->base.offset, rt_flags | NOUVEAU_BO_LOW,
+				      0, 0);
+			so_method(so, eng3d, NV40TCL_COLOR3_PITCH, 1);
+			so_data  (so, rt[3]->pitch);
+		}
+	}
+
+	if (zeta_format) {
+		so_method(so, eng3d, NV34TCL_DMA_ZETA, 1);
+		so_reloc (so, nvfx_surface_buffer(&zeta->base), 0,
+			      rt_flags | NOUVEAU_BO_OR,
+			      chan->vram->handle, chan->gart->handle);
+		so_method(so, eng3d, NV34TCL_ZETA_OFFSET, 1);
+		/* TODO: reverse engineer LMA */
+		so_reloc (so, nvfx_surface_buffer(&zeta->base),
+			      zeta->base.offset, rt_flags | NOUVEAU_BO_LOW, 0, 0);
+	        if(nvfx->is_nv4x) {
+			so_method(so, eng3d, NV40TCL_ZETA_PITCH, 1);
+			so_data  (so, zeta->pitch);
+		}
+	}
+
+	so_method(so, eng3d, NV34TCL_RT_ENABLE, 1);
+	so_data  (so, rt_enable);
+	so_method(so, eng3d, NV34TCL_RT_HORIZ, 3);
+	so_data  (so, (w << 16) | 0);
+	so_data  (so, (h << 16) | 0);
+	so_data  (so, rt_format);
+	so_method(so, eng3d, NV34TCL_VIEWPORT_HORIZ, 2);
+	so_data  (so, (w << 16) | 0);
+	so_data  (so, (h << 16) | 0);
+	so_method(so, eng3d, NV34TCL_VIEWPORT_CLIP_HORIZ(0), 2);
+	so_data  (so, ((w - 1) << 16) | 0);
+	so_data  (so, ((h - 1) << 16) | 0);
+	so_method(so, eng3d, 0x1d88, 1);
+	so_data  (so, (1 << 12) | h);
+
+	if(!nvfx->is_nv4x) {
+		/* Wonder why this is needed, context should all be set to zero on init */
+		/* TODO: we can most likely remove this, after putting it in context init */
+		so_method(so, eng3d, NV34TCL_VIEWPORT_TX_ORIGIN, 1);
+		so_data  (so, 0);
+	}
+
+	so_ref(so, &nvfx->state.hw[NVFX_STATE_FB]);
+	so_ref(NULL, &so);
+	return TRUE;
+}
+
+struct nvfx_state_entry nvfx_state_framebuffer = {
+	.validate = nvfx_state_framebuffer_validate,
+	.dirty = {
+		.pipe = NVFX_NEW_FB,
+		.hw = NVFX_STATE_FB
+	}
+};
diff --git a/src/gallium/drivers/nvfx/nvfx_state_rasterizer.c b/src/gallium/drivers/nvfx/nvfx_state_rasterizer.c
new file mode 100644
index 0000000000..0d35ecbf20
--- /dev/null
+++ b/src/gallium/drivers/nvfx/nvfx_state_rasterizer.c
@@ -0,0 +1,17 @@
+#include "nvfx_context.h"
+
+static boolean
+nvfx_state_rasterizer_validate(struct nvfx_context *nvfx)
+{
+	so_ref(nvfx->rasterizer->so,
+	       &nvfx->state.hw[NVFX_STATE_RAST]);
+	return TRUE;
+}
+
+struct nvfx_state_entry nvfx_state_rasterizer = {
+	.validate = nvfx_state_rasterizer_validate,
+	.dirty = {
+		.pipe = NVFX_NEW_RAST,
+		.hw = NVFX_STATE_RAST
+	}
+};
diff --git a/src/gallium/drivers/nvfx/nvfx_state_scissor.c b/src/gallium/drivers/nvfx/nvfx_state_scissor.c
new file mode 100644
index 0000000000..940d8cb5c0
--- /dev/null
+++ b/src/gallium/drivers/nvfx/nvfx_state_scissor.c
@@ -0,0 +1,36 @@
+#include "nvfx_context.h"
+
+static boolean
+nvfx_state_scissor_validate(struct nvfx_context *nvfx)
+{
+	struct pipe_rasterizer_state *rast = &nvfx->rasterizer->pipe;
+	struct pipe_scissor_state *s = &nvfx->scissor;
+	struct nouveau_stateobj *so;
+
+	if (nvfx->state.hw[NVFX_STATE_SCISSOR] &&
+	    (rast->scissor == 0 && nvfx->state.scissor_enabled == 0))
+		return FALSE;
+	nvfx->state.scissor_enabled = rast->scissor;
+
+	so = so_new(1, 2, 0);
+	so_method(so, nvfx->screen->eng3d, NV34TCL_SCISSOR_HORIZ, 2);
+	if (nvfx->state.scissor_enabled) {
+		so_data  (so, ((s->maxx - s->minx) << 16) | s->minx);
+		so_data  (so, ((s->maxy - s->miny) << 16) | s->miny);
+	} else {
+		so_data  (so, 4096 << 16);
+		so_data  (so, 4096 << 16);
+	}
+
+	so_ref(so, &nvfx->state.hw[NVFX_STATE_SCISSOR]);
+	so_ref(NULL, &so);
+	return TRUE;
+}
+
+struct nvfx_state_entry nvfx_state_scissor = {
+	.validate = nvfx_state_scissor_validate,
+	.dirty = {
+		.pipe = NVFX_NEW_SCISSOR | NVFX_NEW_RAST,
+		.hw = NVFX_STATE_SCISSOR
+	}
+};
diff --git a/src/gallium/drivers/nvfx/nvfx_state_stipple.c b/src/gallium/drivers/nvfx/nvfx_state_stipple.c
new file mode 100644
index 0000000000..57cd3c936a
--- /dev/null
+++ b/src/gallium/drivers/nvfx/nvfx_state_stipple.c
@@ -0,0 +1,40 @@
+#include "nvfx_context.h"
+
+static boolean
+nvfx_state_stipple_validate(struct nvfx_context *nvfx)
+{
+	struct pipe_rasterizer_state *rast = &nvfx->rasterizer->pipe;
+	struct nouveau_grobj *eng3d = nvfx->screen->eng3d;
+	struct nouveau_stateobj *so;
+
+	if (nvfx->state.hw[NVFX_STATE_STIPPLE] &&
+	   (rast->poly_stipple_enable == 0 && nvfx->state.stipple_enabled == 0))
+		return FALSE;
+
+	if (rast->poly_stipple_enable) {
+		unsigned i;
+
+		so = so_new(2, 33, 0);
+		so_method(so, eng3d, NV34TCL_POLYGON_STIPPLE_ENABLE, 1);
+		so_data  (so, 1);
+		so_method(so, eng3d, NV34TCL_POLYGON_STIPPLE_PATTERN(0), 32);
+		for (i = 0; i < 32; i++)
+			so_data(so, nvfx->stipple[i]);
+	} else {
+		so = so_new(1, 1, 0);
+		so_method(so, eng3d, NV34TCL_POLYGON_STIPPLE_ENABLE, 1);
+		so_data  (so, 0);
+	}
+
+	so_ref(so, &nvfx->state.hw[NVFX_STATE_STIPPLE]);
+	so_ref(NULL, &so);
+	return TRUE;
+}
+
+struct nvfx_state_entry nvfx_state_stipple = {
+	.validate = nvfx_state_stipple_validate,
+	.dirty = {
+		.pipe = NVFX_NEW_STIPPLE | NVFX_NEW_RAST,
+		.hw = NVFX_STATE_STIPPLE,
+	}
+};
diff --git a/src/gallium/drivers/nvfx/nvfx_state_viewport.c b/src/gallium/drivers/nvfx/nvfx_state_viewport.c
new file mode 100644
index 0000000000..82e0e9220b
--- /dev/null
+++ b/src/gallium/drivers/nvfx/nvfx_state_viewport.c
@@ -0,0 +1,51 @@
+#include "nvfx_context.h"
+
+static boolean
+nvfx_state_viewport_validate(struct nvfx_context *nvfx)
+{
+	struct pipe_viewport_state *vpt = &nvfx->viewport;
+	struct nouveau_stateobj *so;
+
+	if (nvfx->state.hw[NVFX_STATE_VIEWPORT] &&
+	    !(nvfx->dirty & NVFX_NEW_VIEWPORT))
+		return FALSE;
+
+	so = so_new(2, 9, 0);
+	so_method(so, nvfx->screen->eng3d,
+		  NV34TCL_VIEWPORT_TRANSLATE_X, 8);
+	if(nvfx->render_mode == HW) {
+		so_data  (so, fui(vpt->translate[0]));
+		so_data  (so, fui(vpt->translate[1]));
+		so_data  (so, fui(vpt->translate[2]));
+		so_data  (so, fui(vpt->translate[3]));
+		so_data  (so, fui(vpt->scale[0]));
+		so_data  (so, fui(vpt->scale[1]));
+		so_data  (so, fui(vpt->scale[2]));
+		so_data  (so, fui(vpt->scale[3]));
+		so_method(so, nvfx->screen->eng3d, 0x1d78, 1);
+		so_data  (so, 1);
+	} else {
+		so_data  (so, fui(0.0f));
+		so_data  (so, fui(0.0f));
+		so_data  (so, fui(0.0f));
+		so_data  (so, fui(0.0f));
+		so_data  (so, fui(1.0f));
+		so_data  (so, fui(1.0f));
+		so_data  (so, fui(1.0f));
+		so_data  (so, fui(1.0f));
+		so_method(so, nvfx->screen->eng3d, 0x1d78, 1);
+		so_data  (so, nvfx->is_nv4x ? 0x110 : 1);
+	}
+
+	so_ref(so, &nvfx->state.hw[NVFX_STATE_VIEWPORT]);
+	so_ref(NULL, &so);
+	return TRUE;
+}
+
+struct nvfx_state_entry nvfx_state_viewport = {
+	.validate = nvfx_state_viewport_validate,
+	.dirty = {
+		.pipe = NVFX_NEW_VIEWPORT,
+		.hw = NVFX_STATE_VIEWPORT
+	}
+};
diff --git a/src/gallium/drivers/nvfx/nvfx_state_zsa.c b/src/gallium/drivers/nvfx/nvfx_state_zsa.c
new file mode 100644
index 0000000000..c84fd041c1
--- /dev/null
+++ b/src/gallium/drivers/nvfx/nvfx_state_zsa.c
@@ -0,0 +1,41 @@
+#include "nvfx_context.h"
+
+static boolean
+nvfx_state_zsa_validate(struct nvfx_context *nvfx)
+{
+	so_ref(nvfx->zsa->so,
+	       &nvfx->state.hw[NVFX_STATE_ZSA]);
+	return TRUE;
+}
+
+struct nvfx_state_entry nvfx_state_zsa = {
+	.validate = nvfx_state_zsa_validate,
+	.dirty = {
+		.pipe = NVFX_NEW_ZSA,
+		.hw = NVFX_STATE_ZSA
+	}
+};
+
+static boolean
+nvfx_state_sr_validate(struct nvfx_context *nvfx)
+{
+	struct nouveau_stateobj *so = so_new(2, 2, 0);
+	struct pipe_stencil_ref *sr = &nvfx->stencil_ref;
+
+	so_method(so, nvfx->screen->eng3d, NV34TCL_STENCIL_FRONT_FUNC_REF, 1);
+	so_data  (so, sr->ref_value[0]);
+	so_method(so, nvfx->screen->eng3d, NV34TCL_STENCIL_BACK_FUNC_REF, 1);
+	so_data  (so, sr->ref_value[1]);
+
+	so_ref(so, &nvfx->state.hw[NVFX_STATE_SR]);
+	so_ref(NULL, &so);
+	return TRUE;
+}
+
+struct nvfx_state_entry nvfx_state_sr = {
+	.validate = nvfx_state_sr_validate,
+	.dirty = {
+		.pipe = NVFX_NEW_SR,
+		.hw = NVFX_STATE_SR
+	}
+};
diff --git a/src/gallium/drivers/nv30/nv30_surface.c b/src/gallium/drivers/nvfx/nvfx_surface.c
index bc18e577ee..8a05ad0a57 100644
--- a/src/gallium/drivers/nv30/nv30_surface.c
+++ b/src/gallium/drivers/nvfx/nvfx_surface.c
@@ -1,9 +1,9 @@
 
 /**************************************************************************
- * 
+ *
  * Copyright 2003 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
@@ -11,11 +11,11 @@
  * 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.
@@ -23,40 +23,40 @@
  * 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 "nv30_context.h"
+#include "nvfx_context.h"
 #include "pipe/p_defines.h"
 #include "util/u_inlines.h"
 #include "util/u_tile.h"
 
 static void
-nv30_surface_copy(struct pipe_context *pipe,
+nvfx_surface_copy(struct pipe_context *pipe,
 		  struct pipe_surface *dest, unsigned destx, unsigned desty,
 		  struct pipe_surface *src, unsigned srcx, unsigned srcy,
 		  unsigned width, unsigned height)
 {
-	struct nv30_context *nv30 = nv30_context(pipe);
-	struct nv04_surface_2d *eng2d = nv30->screen->eng2d;
+	struct nvfx_context *nvfx = nvfx_context(pipe);
+	struct nv04_surface_2d *eng2d = nvfx->screen->eng2d;
 
 	eng2d->copy(eng2d, dest, destx, desty, src, srcx, srcy, width, height);
 }
 
 static void
-nv30_surface_fill(struct pipe_context *pipe, struct pipe_surface *dest,
+nvfx_surface_fill(struct pipe_context *pipe, struct pipe_surface *dest,
 		  unsigned destx, unsigned desty, unsigned width,
 		  unsigned height, unsigned value)
 {
-	struct nv30_context *nv30 = nv30_context(pipe);
-	struct nv04_surface_2d *eng2d = nv30->screen->eng2d;
+	struct nvfx_context *nvfx = nvfx_context(pipe);
+	struct nv04_surface_2d *eng2d = nvfx->screen->eng2d;
 
 	eng2d->fill(eng2d, dest, destx, desty, width, height, value);
 }
 
 void
-nv30_init_surface_functions(struct nv30_context *nv30)
+nvfx_init_surface_functions(struct nvfx_context *nvfx)
 {
-	nv30->pipe.surface_copy = nv30_surface_copy;
-	nv30->pipe.surface_fill = nv30_surface_fill;
+	nvfx->pipe.surface_copy = nvfx_surface_copy;
+	nvfx->pipe.surface_fill = nvfx_surface_fill;
 }
diff --git a/src/gallium/drivers/nvfx/nvfx_tex.h b/src/gallium/drivers/nvfx/nvfx_tex.h
new file mode 100644
index 0000000000..69187a79e7
--- /dev/null
+++ b/src/gallium/drivers/nvfx/nvfx_tex.h
@@ -0,0 +1,133 @@
+#ifndef NVFX_TEX_H_
+#define NVFX_TEX_H_
+
+static inline unsigned
+nvfx_tex_wrap_mode(unsigned wrap) {
+	unsigned ret;
+
+	switch (wrap) {
+	case PIPE_TEX_WRAP_REPEAT:
+		ret = NV34TCL_TX_WRAP_S_REPEAT;
+		break;
+	case PIPE_TEX_WRAP_MIRROR_REPEAT:
+		ret = NV34TCL_TX_WRAP_S_MIRRORED_REPEAT;
+		break;
+	case PIPE_TEX_WRAP_CLAMP_TO_EDGE:
+		ret = NV34TCL_TX_WRAP_S_CLAMP_TO_EDGE;
+		break;
+	case PIPE_TEX_WRAP_CLAMP_TO_BORDER:
+		ret = NV34TCL_TX_WRAP_S_CLAMP_TO_BORDER;
+		break;
+	case PIPE_TEX_WRAP_CLAMP:
+		ret = NV34TCL_TX_WRAP_S_CLAMP;
+		break;
+	case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE:
+		ret = NV40TCL_TEX_WRAP_S_MIRROR_CLAMP_TO_EDGE;
+		break;
+	case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER:
+		ret = NV40TCL_TEX_WRAP_S_MIRROR_CLAMP_TO_BORDER;
+		break;
+	case PIPE_TEX_WRAP_MIRROR_CLAMP:
+		ret = NV40TCL_TEX_WRAP_S_MIRROR_CLAMP;
+		break;
+	default:
+		NOUVEAU_ERR("unknown wrap mode: %d\n", wrap);
+		ret = NV34TCL_TX_WRAP_S_REPEAT;
+		break;
+	}
+
+	return ret >> NV34TCL_TX_WRAP_S_SHIFT;
+}
+
+static inline unsigned
+nvfx_tex_wrap_compare_mode(const struct pipe_sampler_state* cso)
+{
+	if (cso->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) {
+		switch (cso->compare_func) {
+		case PIPE_FUNC_NEVER:
+			return NV34TCL_TX_WRAP_RCOMP_NEVER;
+		case PIPE_FUNC_GREATER:
+			return NV34TCL_TX_WRAP_RCOMP_GREATER;
+		case PIPE_FUNC_EQUAL:
+			return NV34TCL_TX_WRAP_RCOMP_EQUAL;
+		case PIPE_FUNC_GEQUAL:
+			return NV34TCL_TX_WRAP_RCOMP_GEQUAL;
+		case PIPE_FUNC_LESS:
+			return NV34TCL_TX_WRAP_RCOMP_LESS;
+		case PIPE_FUNC_NOTEQUAL:
+			return NV34TCL_TX_WRAP_RCOMP_NOTEQUAL;
+		case PIPE_FUNC_LEQUAL:
+			return NV34TCL_TX_WRAP_RCOMP_LEQUAL;
+		case PIPE_FUNC_ALWAYS:
+			return NV34TCL_TX_WRAP_RCOMP_ALWAYS;
+		default:
+			break;
+		}
+	}
+	return 0;
+}
+
+static inline unsigned nvfx_tex_filter(const struct pipe_sampler_state* cso)
+{
+	unsigned filter = 0;
+	switch (cso->mag_img_filter) {
+	case PIPE_TEX_FILTER_LINEAR:
+		filter |= NV34TCL_TX_FILTER_MAGNIFY_LINEAR;
+		break;
+	case PIPE_TEX_FILTER_NEAREST:
+	default:
+		filter |= NV34TCL_TX_FILTER_MAGNIFY_NEAREST;
+		break;
+	}
+
+	switch (cso->min_img_filter) {
+	case PIPE_TEX_FILTER_LINEAR:
+		switch (cso->min_mip_filter) {
+		case PIPE_TEX_MIPFILTER_NEAREST:
+			filter |= NV34TCL_TX_FILTER_MINIFY_LINEAR_MIPMAP_NEAREST;
+			break;
+		case PIPE_TEX_MIPFILTER_LINEAR:
+			filter |= NV34TCL_TX_FILTER_MINIFY_LINEAR_MIPMAP_LINEAR;
+			break;
+		case PIPE_TEX_MIPFILTER_NONE:
+		default:
+			filter |= NV34TCL_TX_FILTER_MINIFY_LINEAR;
+			break;
+		}
+		break;
+	case PIPE_TEX_FILTER_NEAREST:
+	default:
+		switch (cso->min_mip_filter) {
+		case PIPE_TEX_MIPFILTER_NEAREST:
+			filter |= NV34TCL_TX_FILTER_MINIFY_NEAREST_MIPMAP_NEAREST;
+		break;
+		case PIPE_TEX_MIPFILTER_LINEAR:
+			filter |= NV34TCL_TX_FILTER_MINIFY_NEAREST_MIPMAP_LINEAR;
+			break;
+		case PIPE_TEX_MIPFILTER_NONE:
+		default:
+			filter |= NV34TCL_TX_FILTER_MINIFY_NEAREST;
+			break;
+		}
+		break;
+	}
+	return filter;
+}
+
+static inline unsigned nvfx_tex_border_color(const float* border_color)
+{
+	return ((float_to_ubyte(border_color[3]) << 24) |
+		    (float_to_ubyte(border_color[0]) << 16) |
+		    (float_to_ubyte(border_color[1]) <<  8) |
+		    (float_to_ubyte(border_color[2]) <<  0));
+}
+
+struct nvfx_sampler_state {
+	uint32_t fmt;
+	uint32_t wrap;
+	uint32_t en;
+	uint32_t filt;
+	uint32_t bcol;
+};
+
+#endif /* NVFX_TEX_H_ */
diff --git a/src/gallium/drivers/nv30/nv30_transfer.c b/src/gallium/drivers/nvfx/nvfx_transfer.c
index cfc109bb74..409b354d58 100644
--- a/src/gallium/drivers/nv30/nv30_transfer.c
+++ b/src/gallium/drivers/nvfx/nvfx_transfer.c
@@ -5,18 +5,18 @@
 #include "util/u_memory.h"
 #include "util/u_math.h"
 #include "nouveau/nouveau_winsys.h"
-#include "nv30_context.h"
-#include "nv30_screen.h"
-#include "nv30_state.h"
+#include "nvfx_context.h"
+#include "nvfx_screen.h"
+#include "nvfx_state.h"
 
-struct nv30_transfer {
+struct nvfx_transfer {
 	struct pipe_transfer base;
 	struct pipe_surface *surface;
 	boolean direct;
 };
 
 static void
-nv30_compatible_transfer_tex(struct pipe_texture *pt, unsigned width, unsigned height,
+nvfx_compatible_transfer_tex(struct pipe_texture *pt, unsigned width, unsigned height,
                              struct pipe_texture *template)
 {
 	memset(template, 0, sizeof(struct pipe_texture));
@@ -33,17 +33,17 @@ nv30_compatible_transfer_tex(struct pipe_texture *pt, unsigned width, unsigned h
 }
 
 static struct pipe_transfer *
-nv30_transfer_new(struct pipe_context *pcontext, struct pipe_texture *pt,
+nvfx_transfer_new(struct pipe_context *pcontext, struct pipe_texture *pt,
 		  unsigned face, unsigned level, unsigned zslice,
 		  enum pipe_transfer_usage usage,
 		  unsigned x, unsigned y, unsigned w, unsigned h)
 {
         struct pipe_screen *pscreen = pcontext->screen;
-	struct nv30_miptree *mt = (struct nv30_miptree *)pt;
-	struct nv30_transfer *tx;
+	struct nvfx_miptree *mt = (struct nvfx_miptree *)pt;
+	struct nvfx_transfer *tx;
 	struct pipe_texture tx_tex_template, *tx_tex;
 
-	tx = CALLOC_STRUCT(nv30_transfer);
+	tx = CALLOC_STRUCT(nvfx_transfer);
 	if (!tx)
 		return NULL;
 
@@ -72,7 +72,7 @@ nv30_transfer_new(struct pipe_context *pcontext, struct pipe_texture *pt,
 
 	tx->direct = false;
 
-	nv30_compatible_transfer_tex(pt, w, h, &tx_tex_template);
+	nvfx_compatible_transfer_tex(pt, w, h, &tx_tex_template);
 
 	tx_tex = pscreen->texture_create(pscreen, &tx_tex_template);
 	if (!tx_tex)
@@ -81,7 +81,7 @@ nv30_transfer_new(struct pipe_context *pcontext, struct pipe_texture *pt,
 		return NULL;
 	}
 
-	tx->base.stride = ((struct nv30_miptree*)tx_tex)->level[0].pitch;
+	tx->base.stride = ((struct nvfx_miptree*)tx_tex)->level[0].pitch;
 
 	tx->surface = pscreen->get_tex_surface(pscreen, tx_tex,
 	                                       0, 0, 0,
@@ -97,7 +97,7 @@ nv30_transfer_new(struct pipe_context *pcontext, struct pipe_texture *pt,
 	}
 
 	if (usage & PIPE_TRANSFER_READ) {
-		struct nv30_screen *nvscreen = nv30_screen(pscreen);
+		struct nvfx_screen *nvscreen = nvfx_screen(pscreen);
 		struct pipe_surface *src;
 
 		src = pscreen->get_tex_surface(pscreen, pt,
@@ -118,14 +118,14 @@ nv30_transfer_new(struct pipe_context *pcontext, struct pipe_texture *pt,
 }
 
 static void
-nv30_transfer_del(struct pipe_context *pcontext,
+nvfx_transfer_del(struct pipe_context *pcontext,
                   struct pipe_transfer *ptx)
 {
-	struct nv30_transfer *tx = (struct nv30_transfer *)ptx;
+	struct nvfx_transfer *tx = (struct nvfx_transfer *)ptx;
 
 	if (!tx->direct && (ptx->usage & PIPE_TRANSFER_WRITE)) {
 		struct pipe_screen *pscreen = pcontext->screen;
-		struct nv30_screen *nvscreen = nv30_screen(pscreen);
+		struct nvfx_screen *nvscreen = nvfx_screen(pscreen);
 		struct pipe_surface *dst;
 
 		dst = pscreen->get_tex_surface(pscreen, ptx->texture,
@@ -147,12 +147,12 @@ nv30_transfer_del(struct pipe_context *pcontext,
 }
 
 static void *
-nv30_transfer_map(struct pipe_context *pcontext, struct pipe_transfer *ptx)
+nvfx_transfer_map(struct pipe_context *pcontext, struct pipe_transfer *ptx)
 {
         struct pipe_screen *pscreen = pcontext->screen;
-	struct nv30_transfer *tx = (struct nv30_transfer *)ptx;
+	struct nvfx_transfer *tx = (struct nvfx_transfer *)ptx;
 	struct nv04_surface *ns = (struct nv04_surface *)tx->surface;
-	struct nv30_miptree *mt = (struct nv30_miptree *)tx->surface->texture;
+	struct nvfx_miptree *mt = (struct nvfx_miptree *)tx->surface->texture;
 	void *map = pipe_buffer_map(pscreen, mt->buffer,
 	                            pipe_transfer_buffer_flags(ptx));
 
@@ -163,20 +163,20 @@ nv30_transfer_map(struct pipe_context *pcontext, struct pipe_transfer *ptx)
 }
 
 static void
-nv30_transfer_unmap(struct pipe_context *pcontext, struct pipe_transfer *ptx)
+nvfx_transfer_unmap(struct pipe_context *pcontext, struct pipe_transfer *ptx)
 {
-        struct pipe_screen *pscreen = pcontext->screen;
-	struct nv30_transfer *tx = (struct nv30_transfer *)ptx;
-	struct nv30_miptree *mt = (struct nv30_miptree *)tx->surface->texture;
+	struct pipe_screen *pscreen = pcontext->screen;
+	struct nvfx_transfer *tx = (struct nvfx_transfer *)ptx;
+	struct nvfx_miptree *mt = (struct nvfx_miptree *)tx->surface->texture;
 
 	pipe_buffer_unmap(pscreen, mt->buffer);
 }
 
 void
-nv30_init_transfer_functions(struct nv30_context *nv30)
+nvfx_init_transfer_functions(struct nvfx_context *nvfx)
 {
-	nv30->pipe.get_tex_transfer = nv30_transfer_new;
-	nv30->pipe.tex_transfer_destroy = nv30_transfer_del;
-	nv30->pipe.transfer_map = nv30_transfer_map;
-	nv30->pipe.transfer_unmap = nv30_transfer_unmap;
+	nvfx->pipe.get_tex_transfer = nvfx_transfer_new;
+	nvfx->pipe.tex_transfer_destroy = nvfx_transfer_del;
+	nvfx->pipe.transfer_map = nvfx_transfer_map;
+	nvfx->pipe.transfer_unmap = nvfx_transfer_unmap;
 }
diff --git a/src/gallium/drivers/nv30/nv30_vbo.c b/src/gallium/drivers/nvfx/nvfx_vbo.c
index f3856bb5a5..257087f8f6 100644
--- a/src/gallium/drivers/nv30/nv30_vbo.c
+++ b/src/gallium/drivers/nvfx/nvfx_vbo.c
@@ -3,17 +3,24 @@
 #include "util/u_inlines.h"
 #include "util/u_format.h"
 
-#include "nv30_context.h"
-#include "nv30_state.h"
+#include "nvfx_context.h"
+#include "nvfx_state.h"
 
 #include "nouveau/nouveau_channel.h"
 #include "nouveau/nouveau_pushbuf.h"
 #include "nouveau/nouveau_util.h"
 
-#define FORCE_SWTNL 0
+static boolean
+nvfx_force_swtnl(struct nvfx_context *nvfx)
+{
+	static int force_swtnl = -1;
+	if(force_swtnl < 0)
+		force_swtnl = debug_get_bool_option("NOUVEAU_SWTNL", 0);
+	return force_swtnl;
+}
 
 static INLINE int
-nv30_vbo_format_to_hw(enum pipe_format pipe, unsigned *fmt, unsigned *ncomp)
+nvfx_vbo_format_to_hw(enum pipe_format pipe, unsigned *fmt, unsigned *ncomp)
 {
 	switch (pipe) {
 	case PIPE_FORMAT_R32_FLOAT:
@@ -69,15 +76,15 @@ nv30_vbo_format_to_hw(enum pipe_format pipe, unsigned *fmt, unsigned *ncomp)
 }
 
 static boolean
-nv30_vbo_set_idxbuf(struct nv30_context *nv30, struct pipe_buffer *ib,
+nvfx_vbo_set_idxbuf(struct nvfx_context *nvfx, struct pipe_buffer *ib,
 		    unsigned ib_size)
 {
-	struct pipe_screen *pscreen = &nv30->screen->base.base;
+	struct pipe_screen *pscreen = &nvfx->screen->base.base;
 	unsigned type;
 
 	if (!ib) {
-		nv30->idxbuf = NULL;
-		nv30->idxbuf_format = 0xdeadbeef;
+		nvfx->idxbuf = NULL;
+		nvfx->idxbuf_format = 0xdeadbeef;
 		return FALSE;
 	}
 
@@ -95,27 +102,27 @@ nv30_vbo_set_idxbuf(struct nv30_context *nv30, struct pipe_buffer *ib,
 		return FALSE;
 	}
 
-	if (ib != nv30->idxbuf ||
-	    type != nv30->idxbuf_format) {
-		nv30->dirty |= NV30_NEW_ARRAYS;
-		nv30->idxbuf = ib;
-		nv30->idxbuf_format = type;
+	if (ib != nvfx->idxbuf ||
+	    type != nvfx->idxbuf_format) {
+		nvfx->dirty |= NVFX_NEW_ARRAYS;
+		nvfx->idxbuf = ib;
+		nvfx->idxbuf_format = type;
 	}
 
 	return TRUE;
 }
 
 static boolean
-nv30_vbo_static_attrib(struct nv30_context *nv30, struct nouveau_stateobj *so,
+nvfx_vbo_static_attrib(struct nvfx_context *nvfx, struct nouveau_stateobj *so,
 		       int attrib, struct pipe_vertex_element *ve,
 		       struct pipe_vertex_buffer *vb)
 {
-	struct pipe_screen *pscreen = nv30->pipe.screen;
-	struct nouveau_grobj *rankine = nv30->screen->rankine;
+	struct pipe_screen *pscreen = nvfx->pipe.screen;
+	struct nouveau_grobj *eng3d = nvfx->screen->eng3d;
 	unsigned type, ncomp;
 	void *map;
 
-	if (nv30_vbo_format_to_hw(ve->src_format, &type, &ncomp))
+	if (nvfx_vbo_format_to_hw(ve->src_format, &type, &ncomp))
 		return FALSE;
 
 	map  = pipe_buffer_map(pscreen, vb->buffer, PIPE_BUFFER_USAGE_CPU_READ);
@@ -128,25 +135,25 @@ nv30_vbo_static_attrib(struct nv30_context *nv30, struct nouveau_stateobj *so,
 
 		switch (ncomp) {
 		case 4:
-			so_method(so, rankine, NV34TCL_VTX_ATTR_4F_X(attrib), 4);
+			so_method(so, eng3d, NV34TCL_VTX_ATTR_4F_X(attrib), 4);
 			so_data  (so, fui(v[0]));
 			so_data  (so, fui(v[1]));
 			so_data  (so, fui(v[2]));
 			so_data  (so, fui(v[3]));
 			break;
 		case 3:
-			so_method(so, rankine, NV34TCL_VTX_ATTR_3F_X(attrib), 3);
+			so_method(so, eng3d, NV34TCL_VTX_ATTR_3F_X(attrib), 3);
 			so_data  (so, fui(v[0]));
 			so_data  (so, fui(v[1]));
 			so_data  (so, fui(v[2]));
 			break;
 		case 2:
-			so_method(so, rankine, NV34TCL_VTX_ATTR_2F_X(attrib), 2);
+			so_method(so, eng3d, NV34TCL_VTX_ATTR_2F_X(attrib), 2);
 			so_data  (so, fui(v[0]));
 			so_data  (so, fui(v[1]));
 			break;
 		case 1:
-			so_method(so, rankine, NV34TCL_VTX_ATTR_1F(attrib), 1);
+			so_method(so, eng3d, NV34TCL_VTX_ATTR_1F(attrib), 1);
 			so_data  (so, fui(v[0]));
 			break;
 		default:
@@ -165,26 +172,26 @@ nv30_vbo_static_attrib(struct nv30_context *nv30, struct nouveau_stateobj *so,
 }
 
 void
-nv30_draw_arrays(struct pipe_context *pipe,
+nvfx_draw_arrays(struct pipe_context *pipe,
 		 unsigned mode, unsigned start, unsigned count)
 {
-	struct nv30_context *nv30 = nv30_context(pipe);
-	struct nv30_screen *screen = nv30->screen;
+	struct nvfx_context *nvfx = nvfx_context(pipe);
+	struct nvfx_screen *screen = nvfx->screen;
 	struct nouveau_channel *chan = screen->base.channel;
-	struct nouveau_grobj *rankine = screen->rankine;
+	struct nouveau_grobj *eng3d = screen->eng3d;
 	unsigned restart = 0;
 
-	nv30_vbo_set_idxbuf(nv30, NULL, 0);
-	if (FORCE_SWTNL || !nv30_state_validate(nv30)) {
-		/*return nv30_draw_elements_swtnl(pipe, NULL, 0,
-						mode, start, count);*/
-		return;
+	nvfx_vbo_set_idxbuf(nvfx, NULL, 0);
+	if (nvfx_force_swtnl(nvfx) || !nvfx_state_validate(nvfx)) {
+		nvfx_draw_elements_swtnl(pipe, NULL, 0,
+                                           mode, start, count);
+                return;
 	}
 
 	while (count) {
 		unsigned vc, nr;
 
-		nv30_state_emit(nv30);
+		nvfx_state_emit(nvfx);
 
 		vc = nouveau_vbuf_split(AVAIL_RING(chan), 6, 256,
 					mode, start, count, &restart);
@@ -193,12 +200,12 @@ nv30_draw_arrays(struct pipe_context *pipe,
 			continue;
 		}
 
-		BEGIN_RING(chan, rankine, NV34TCL_VERTEX_BEGIN_END, 1);
+		BEGIN_RING(chan, eng3d, NV34TCL_VERTEX_BEGIN_END, 1);
 		OUT_RING  (chan, nvgl_primitive(mode));
 
 		nr = (vc & 0xff);
 		if (nr) {
-			BEGIN_RING(chan, rankine, NV34TCL_VB_VERTEX_BATCH, 1);
+			BEGIN_RING(chan, eng3d, NV34TCL_VB_VERTEX_BATCH, 1);
 			OUT_RING  (chan, ((nr - 1) << 24) | start);
 			start += nr;
 		}
@@ -209,14 +216,14 @@ nv30_draw_arrays(struct pipe_context *pipe,
 
 			nr -= push;
 
-			BEGIN_RING_NI(chan, rankine, NV34TCL_VB_VERTEX_BATCH, push);
+			BEGIN_RING_NI(chan, eng3d, NV34TCL_VB_VERTEX_BATCH, push);
 			while (push--) {
 				OUT_RING(chan, ((0x100 - 1) << 24) | start);
 				start += 0x100;
 			}
 		}
 
-		BEGIN_RING(chan, rankine, NV34TCL_VERTEX_BEGIN_END, 1);
+		BEGIN_RING(chan, eng3d, NV34TCL_VERTEX_BEGIN_END, 1);
 		OUT_RING  (chan, 0);
 
 		count -= vc;
@@ -227,18 +234,18 @@ nv30_draw_arrays(struct pipe_context *pipe,
 }
 
 static INLINE void
-nv30_draw_elements_u08(struct nv30_context *nv30, void *ib,
+nvfx_draw_elements_u08(struct nvfx_context *nvfx, void *ib,
 		       unsigned mode, unsigned start, unsigned count)
 {
-	struct nv30_screen *screen = nv30->screen;
+	struct nvfx_screen *screen = nvfx->screen;
 	struct nouveau_channel *chan = screen->base.channel;
-	struct nouveau_grobj *rankine = screen->rankine;
+	struct nouveau_grobj *eng3d = screen->eng3d;
 
 	while (count) {
 		uint8_t *elts = (uint8_t *)ib + start;
 		unsigned vc, push, restart = 0;
 
-		nv30_state_emit(nv30);
+		nvfx_state_emit(nvfx);
 
 		vc = nouveau_vbuf_split(AVAIL_RING(chan), 6, 2,
 					mode, start, count, &restart);
@@ -248,11 +255,11 @@ nv30_draw_elements_u08(struct nv30_context *nv30, void *ib,
 		}
 		count -= vc;
 
-		BEGIN_RING(chan, rankine, NV34TCL_VERTEX_BEGIN_END, 1);
+		BEGIN_RING(chan, eng3d, NV34TCL_VERTEX_BEGIN_END, 1);
 		OUT_RING  (chan, nvgl_primitive(mode));
 
 		if (vc & 1) {
-			BEGIN_RING(chan, rankine, NV34TCL_VB_ELEMENT_U32, 1);
+			BEGIN_RING(chan, eng3d, NV34TCL_VB_ELEMENT_U32, 1);
 			OUT_RING  (chan, elts[0]);
 			elts++; vc--;
 		}
@@ -262,7 +269,7 @@ nv30_draw_elements_u08(struct nv30_context *nv30, void *ib,
 
 			push = MIN2(vc, 2047 * 2);
 
-			BEGIN_RING_NI(chan, rankine, NV34TCL_VB_ELEMENT_U16, push >> 1);
+			BEGIN_RING_NI(chan, eng3d, NV34TCL_VB_ELEMENT_U16, push >> 1);
 			for (i = 0; i < push; i+=2)
 				OUT_RING(chan, (elts[i+1] << 16) | elts[i]);
 
@@ -270,7 +277,7 @@ nv30_draw_elements_u08(struct nv30_context *nv30, void *ib,
 			elts += push;
 		}
 
-		BEGIN_RING(chan, rankine, NV34TCL_VERTEX_BEGIN_END, 1);
+		BEGIN_RING(chan, eng3d, NV34TCL_VERTEX_BEGIN_END, 1);
 		OUT_RING  (chan, 0);
 
 		start = restart;
@@ -278,18 +285,18 @@ nv30_draw_elements_u08(struct nv30_context *nv30, void *ib,
 }
 
 static INLINE void
-nv30_draw_elements_u16(struct nv30_context *nv30, void *ib,
+nvfx_draw_elements_u16(struct nvfx_context *nvfx, void *ib,
 		       unsigned mode, unsigned start, unsigned count)
 {
-	struct nv30_screen *screen = nv30->screen;
+	struct nvfx_screen *screen = nvfx->screen;
 	struct nouveau_channel *chan = screen->base.channel;
-	struct nouveau_grobj *rankine = screen->rankine;
+	struct nouveau_grobj *eng3d = screen->eng3d;
 
 	while (count) {
 		uint16_t *elts = (uint16_t *)ib + start;
 		unsigned vc, push, restart = 0;
 
-		nv30_state_emit(nv30);
+		nvfx_state_emit(nvfx);
 
 		vc = nouveau_vbuf_split(AVAIL_RING(chan), 6, 2,
 					mode, start, count, &restart);
@@ -299,11 +306,11 @@ nv30_draw_elements_u16(struct nv30_context *nv30, void *ib,
 		}
 		count -= vc;
 
-		BEGIN_RING(chan, rankine, NV34TCL_VERTEX_BEGIN_END, 1);
+		BEGIN_RING(chan, eng3d, NV34TCL_VERTEX_BEGIN_END, 1);
 		OUT_RING  (chan, nvgl_primitive(mode));
 
 		if (vc & 1) {
-			BEGIN_RING(chan, rankine, NV34TCL_VB_ELEMENT_U32, 1);
+			BEGIN_RING(chan, eng3d, NV34TCL_VB_ELEMENT_U32, 1);
 			OUT_RING  (chan, elts[0]);
 			elts++; vc--;
 		}
@@ -313,7 +320,7 @@ nv30_draw_elements_u16(struct nv30_context *nv30, void *ib,
 
 			push = MIN2(vc, 2047 * 2);
 
-			BEGIN_RING_NI(chan, rankine, NV34TCL_VB_ELEMENT_U16, push >> 1);
+			BEGIN_RING_NI(chan, eng3d, NV34TCL_VB_ELEMENT_U16, push >> 1);
 			for (i = 0; i < push; i+=2)
 				OUT_RING(chan, (elts[i+1] << 16) | elts[i]);
 
@@ -321,7 +328,7 @@ nv30_draw_elements_u16(struct nv30_context *nv30, void *ib,
 			elts += push;
 		}
 
-		BEGIN_RING(chan, rankine, NV34TCL_VERTEX_BEGIN_END, 1);
+		BEGIN_RING(chan, eng3d, NV34TCL_VERTEX_BEGIN_END, 1);
 		OUT_RING  (chan, 0);
 
 		start = restart;
@@ -329,18 +336,18 @@ nv30_draw_elements_u16(struct nv30_context *nv30, void *ib,
 }
 
 static INLINE void
-nv30_draw_elements_u32(struct nv30_context *nv30, void *ib,
+nvfx_draw_elements_u32(struct nvfx_context *nvfx, void *ib,
 		       unsigned mode, unsigned start, unsigned count)
 {
-	struct nv30_screen *screen = nv30->screen;
+	struct nvfx_screen *screen = nvfx->screen;
 	struct nouveau_channel *chan = screen->base.channel;
-	struct nouveau_grobj *rankine = screen->rankine;
+	struct nouveau_grobj *eng3d = screen->eng3d;
 
 	while (count) {
 		uint32_t *elts = (uint32_t *)ib + start;
 		unsigned vc, push, restart = 0;
 
-		nv30_state_emit(nv30);
+		nvfx_state_emit(nvfx);
 
 		vc = nouveau_vbuf_split(AVAIL_RING(chan), 5, 1,
 					mode, start, count, &restart);
@@ -350,20 +357,20 @@ nv30_draw_elements_u32(struct nv30_context *nv30, void *ib,
 		}
 		count -= vc;
 
-		BEGIN_RING(chan, rankine, NV34TCL_VERTEX_BEGIN_END, 1);
+		BEGIN_RING(chan, eng3d, NV34TCL_VERTEX_BEGIN_END, 1);
 		OUT_RING  (chan, nvgl_primitive(mode));
 
 		while (vc) {
 			push = MIN2(vc, 2047);
 
-			BEGIN_RING_NI(chan, rankine, NV34TCL_VB_ELEMENT_U32, push);
+			BEGIN_RING_NI(chan, eng3d, NV34TCL_VB_ELEMENT_U32, push);
 			OUT_RINGp    (chan, elts, push);
 
 			vc -= push;
 			elts += push;
 		}
 
-		BEGIN_RING(chan, rankine, NV34TCL_VERTEX_BEGIN_END, 1);
+		BEGIN_RING(chan, eng3d, NV34TCL_VERTEX_BEGIN_END, 1);
 		OUT_RING  (chan, 0);
 
 		start = restart;
@@ -371,11 +378,11 @@ nv30_draw_elements_u32(struct nv30_context *nv30, void *ib,
 }
 
 static void
-nv30_draw_elements_inline(struct pipe_context *pipe,
+nvfx_draw_elements_inline(struct pipe_context *pipe,
 			  struct pipe_buffer *ib, unsigned ib_size,
 			  unsigned mode, unsigned start, unsigned count)
 {
-	struct nv30_context *nv30 = nv30_context(pipe);
+	struct nvfx_context *nvfx = nvfx_context(pipe);
 	struct pipe_screen *pscreen = pipe->screen;
 	void *map;
 
@@ -387,13 +394,13 @@ nv30_draw_elements_inline(struct pipe_context *pipe,
 
 	switch (ib_size) {
 	case 1:
-		nv30_draw_elements_u08(nv30, map, mode, start, count);
+		nvfx_draw_elements_u08(nvfx, map, mode, start, count);
 		break;
 	case 2:
-		nv30_draw_elements_u16(nv30, map, mode, start, count);
+		nvfx_draw_elements_u16(nvfx, map, mode, start, count);
 		break;
 	case 4:
-		nv30_draw_elements_u32(nv30, map, mode, start, count);
+		nvfx_draw_elements_u32(nvfx, map, mode, start, count);
 		break;
 	default:
 		NOUVEAU_ERR("invalid idxbuf fmt %d\n", ib_size);
@@ -404,19 +411,19 @@ nv30_draw_elements_inline(struct pipe_context *pipe,
 }
 
 static void
-nv30_draw_elements_vbo(struct pipe_context *pipe,
+nvfx_draw_elements_vbo(struct pipe_context *pipe,
 		       unsigned mode, unsigned start, unsigned count)
 {
-	struct nv30_context *nv30 = nv30_context(pipe);
-	struct nv30_screen *screen = nv30->screen;
+	struct nvfx_context *nvfx = nvfx_context(pipe);
+	struct nvfx_screen *screen = nvfx->screen;
 	struct nouveau_channel *chan = screen->base.channel;
-	struct nouveau_grobj *rankine = screen->rankine;
+	struct nouveau_grobj *eng3d = screen->eng3d;
 	unsigned restart = 0;
 
 	while (count) {
 		unsigned nr, vc;
 
-		nv30_state_emit(nv30);
+		nvfx_state_emit(nvfx);
 
 		vc = nouveau_vbuf_split(AVAIL_RING(chan), 6, 256,
 					mode, start, count, &restart);
@@ -425,12 +432,12 @@ nv30_draw_elements_vbo(struct pipe_context *pipe,
 			continue;
 		}
 
-		BEGIN_RING(chan, rankine, NV34TCL_VERTEX_BEGIN_END, 1);
+		BEGIN_RING(chan, eng3d, NV34TCL_VERTEX_BEGIN_END, 1);
 		OUT_RING  (chan, nvgl_primitive(mode));
 
 		nr = (vc & 0xff);
 		if (nr) {
-			BEGIN_RING(chan, rankine, NV34TCL_VB_INDEX_BATCH, 1);
+			BEGIN_RING(chan, eng3d, NV34TCL_VB_INDEX_BATCH, 1);
 			OUT_RING  (chan, ((nr - 1) << 24) | start);
 			start += nr;
 		}
@@ -441,14 +448,14 @@ nv30_draw_elements_vbo(struct pipe_context *pipe,
 
 			nr -= push;
 
-			BEGIN_RING_NI(chan, rankine, NV34TCL_VB_INDEX_BATCH, push);
+			BEGIN_RING_NI(chan, eng3d, NV34TCL_VB_INDEX_BATCH, push);
 			while (push--) {
 				OUT_RING(chan, ((0x100 - 1) << 24) | start);
 				start += 0x100;
 			}
 		}
 
-		BEGIN_RING(chan, rankine, NV34TCL_VERTEX_BEGIN_END, 1);
+		BEGIN_RING(chan, eng3d, NV34TCL_VERTEX_BEGIN_END, 1);
 		OUT_RING  (chan, 0);
 
 		count -= vc;
@@ -457,24 +464,24 @@ nv30_draw_elements_vbo(struct pipe_context *pipe,
 }
 
 void
-nv30_draw_elements(struct pipe_context *pipe,
+nvfx_draw_elements(struct pipe_context *pipe,
 		   struct pipe_buffer *indexBuffer, unsigned indexSize,
 		   unsigned mode, unsigned start, unsigned count)
 {
-	struct nv30_context *nv30 = nv30_context(pipe);
+	struct nvfx_context *nvfx = nvfx_context(pipe);
 	boolean idxbuf;
 
-	idxbuf = nv30_vbo_set_idxbuf(nv30, indexBuffer, indexSize);
-	if (FORCE_SWTNL || !nv30_state_validate(nv30)) {
-		/*return nv30_draw_elements_swtnl(pipe, NULL, 0,
-						mode, start, count);*/
+	idxbuf = nvfx_vbo_set_idxbuf(nvfx, indexBuffer, indexSize);
+	if (nvfx_force_swtnl(nvfx) || !nvfx_state_validate(nvfx)) {
+		nvfx_draw_elements_swtnl(pipe, indexBuffer, indexSize,
+                                           mode, start, count);
 		return;
 	}
 
 	if (idxbuf) {
-		nv30_draw_elements_vbo(pipe, mode, start, count);
+		nvfx_draw_elements_vbo(pipe, mode, start, count);
 	} else {
-		nv30_draw_elements_inline(pipe, indexBuffer, indexSize,
+		nvfx_draw_elements_inline(pipe, indexBuffer, indexSize,
 					  mode, start, count);
 	}
 
@@ -482,49 +489,50 @@ nv30_draw_elements(struct pipe_context *pipe,
 }
 
 static boolean
-nv30_vbo_validate(struct nv30_context *nv30)
+nvfx_vbo_validate(struct nvfx_context *nvfx)
 {
 	struct nouveau_stateobj *vtxbuf, *vtxfmt, *sattr = NULL;
-	struct nouveau_grobj *rankine = nv30->screen->rankine;
-	struct pipe_buffer *ib = nv30->idxbuf;
-	unsigned ib_format = nv30->idxbuf_format;
+	struct nouveau_grobj *eng3d = nvfx->screen->eng3d;
+	struct pipe_buffer *ib = nvfx->idxbuf;
+	unsigned ib_format = nvfx->idxbuf_format;
 	unsigned vb_flags = NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD;
 	int hw;
 
 	vtxbuf = so_new(3, 17, 18);
-	so_method(vtxbuf, rankine, NV34TCL_VTXBUF_ADDRESS(0), nv30->vtxelt->num_elements);
+	so_method(vtxbuf, eng3d, NV34TCL_VTXBUF_ADDRESS(0), nvfx->vtxelt->num_elements);
 	vtxfmt = so_new(1, 16, 0);
-	so_method(vtxfmt, rankine, NV34TCL_VTXFMT(0), nv30->vtxelt->num_elements);
+	so_method(vtxfmt, eng3d, NV34TCL_VTXFMT(0), nvfx->vtxelt->num_elements);
 
-	for (hw = 0; hw < nv30->vtxelt->num_elements; hw++) {
+	for (hw = 0; hw < nvfx->vtxelt->num_elements; hw++) {
 		struct pipe_vertex_element *ve;
 		struct pipe_vertex_buffer *vb;
 		unsigned type, ncomp;
 
-		ve = &nv30->vtxelt->pipe[hw];
-		vb = &nv30->vtxbuf[ve->vertex_buffer_index];
+		ve = &nvfx->vtxelt->pipe[hw];
+		vb = &nvfx->vtxbuf[ve->vertex_buffer_index];
 
 		if (!vb->stride) {
 			if (!sattr)
 				sattr = so_new(16, 16 * 4, 0);
 
-			if (nv30_vbo_static_attrib(nv30, sattr, hw, ve, vb)) {
+			if (nvfx_vbo_static_attrib(nvfx, sattr, hw, ve, vb)) {
 				so_data(vtxbuf, 0);
 				so_data(vtxfmt, NV34TCL_VTXFMT_TYPE_FLOAT);
 				continue;
 			}
 		}
 
-		if (nv30_vbo_format_to_hw(ve->src_format, &type, &ncomp)) {
-			/*nv30->fallback_swtnl |= NV30_NEW_ARRAYS;*/
+		if (nvfx_vbo_format_to_hw(ve->src_format, &type, &ncomp)) {
+			nvfx->fallback_swtnl |= NVFX_NEW_ARRAYS;
 			so_ref(NULL, &vtxbuf);
 			so_ref(NULL, &vtxfmt);
 			return FALSE;
 		}
 
-		so_reloc(vtxbuf, nouveau_bo(vb->buffer), vb->buffer_offset +
-				 ve->src_offset, vb_flags | NOUVEAU_BO_LOW |
-				 NOUVEAU_BO_OR, 0, NV34TCL_VTXBUF_ADDRESS_DMA1);
+		so_reloc(vtxbuf, nouveau_bo(vb->buffer),
+				 vb->buffer_offset + ve->src_offset,
+				 vb_flags | NOUVEAU_BO_LOW | NOUVEAU_BO_OR,
+				 0, NV34TCL_VTXBUF_ADDRESS_DMA1);
 		so_data (vtxfmt, ((vb->stride << NV34TCL_VTXFMT_STRIDE_SHIFT) |
 				  (ncomp << NV34TCL_VTXFMT_SIZE_SHIFT) | type));
 	}
@@ -532,31 +540,31 @@ nv30_vbo_validate(struct nv30_context *nv30)
 	if (ib) {
 		struct nouveau_bo *bo = nouveau_bo(ib);
 
-		so_method(vtxbuf, rankine, NV34TCL_IDXBUF_ADDRESS, 2);
+		so_method(vtxbuf, eng3d, NV34TCL_IDXBUF_ADDRESS, 2);
 		so_reloc (vtxbuf, bo, 0, vb_flags | NOUVEAU_BO_LOW, 0, 0);
 		so_reloc (vtxbuf, bo, ib_format, vb_flags | NOUVEAU_BO_OR,
 				  0, NV34TCL_IDXBUF_FORMAT_DMA1);
 	}
 
-	so_method(vtxbuf, rankine, 0x1710, 1);
+	so_method(vtxbuf, eng3d, 0x1710, 1);
 	so_data  (vtxbuf, 0);
 
-	so_ref(vtxbuf, &nv30->state.hw[NV30_STATE_VTXBUF]);
+	so_ref(vtxbuf, &nvfx->state.hw[NVFX_STATE_VTXBUF]);
 	so_ref(NULL, &vtxbuf);
-	nv30->state.dirty |= (1ULL << NV30_STATE_VTXBUF);
-	so_ref(vtxfmt, &nv30->state.hw[NV30_STATE_VTXFMT]);
+	nvfx->state.dirty |= (1ULL << NVFX_STATE_VTXBUF);
+	so_ref(vtxfmt, &nvfx->state.hw[NVFX_STATE_VTXFMT]);
 	so_ref(NULL, &vtxfmt);
-	nv30->state.dirty |= (1ULL << NV30_STATE_VTXFMT);
-	so_ref(sattr, &nv30->state.hw[NV30_STATE_VTXATTR]);
+	nvfx->state.dirty |= (1ULL << NVFX_STATE_VTXFMT);
+	so_ref(sattr, &nvfx->state.hw[NVFX_STATE_VTXATTR]);
 	so_ref(NULL, &sattr);
-	nv30->state.dirty |= (1ULL << NV30_STATE_VTXATTR);
+	nvfx->state.dirty |= (1ULL << NVFX_STATE_VTXATTR);
 	return FALSE;
 }
 
-struct nv30_state_entry nv30_state_vbo = {
-	.validate = nv30_vbo_validate,
+struct nvfx_state_entry nvfx_state_vbo = {
+	.validate = nvfx_vbo_validate,
 	.dirty = {
-		.pipe = NV30_NEW_ARRAYS,
+		.pipe = NVFX_NEW_ARRAYS,
 		.hw = 0,
 	}
 };
diff --git a/src/gallium/drivers/nv40/nv40_vertprog.c b/src/gallium/drivers/nvfx/nvfx_vertprog.c
index c93c5d127c..2d243be16a 100644
--- a/src/gallium/drivers/nv40/nv40_vertprog.c
+++ b/src/gallium/drivers/nvfx/nvfx_vertprog.c
@@ -5,10 +5,11 @@
 
 #include "pipe/p_shader_tokens.h"
 #include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_dump.h"
 #include "tgsi/tgsi_util.h"
 
-#include "nv40_context.h"
-#include "nv40_state.h"
+#include "nvfx_context.h"
+#include "nvfx_state.h"
 
 /* TODO (at least...):
  *  1. Indexed consts  + ARL
@@ -21,76 +22,62 @@
  *  4. bugs
  */
 
-#define SWZ_X 0
-#define SWZ_Y 1
-#define SWZ_Z 2
-#define SWZ_W 3
-#define MASK_X 8
-#define MASK_Y 4
-#define MASK_Z 2
-#define MASK_W 1
-#define MASK_ALL (MASK_X|MASK_Y|MASK_Z|MASK_W)
-#define DEF_SCALE 0
-#define DEF_CTEST 0
-#include "nv40_shader.h"
+#include "nv30_vertprog.h"
+#include "nv40_vertprog.h"
 
-#define swz(s,x,y,z,w) nv40_sr_swz((s), SWZ_##x, SWZ_##y, SWZ_##z, SWZ_##w)
-#define neg(s) nv40_sr_neg((s))
-#define abs(s) nv40_sr_abs((s))
+#define NVFX_VP_INST_DEST_CLIP(n) ((~0 - 6) + (n))
 
-#define NV40_VP_INST_DEST_CLIP(n) ((~0 - 6) + (n))
+struct nvfx_vpc {
+	struct nvfx_vertex_program *vp;
 
-struct nv40_vpc {
-	struct nv40_vertex_program *vp;
-
-	struct nv40_vertex_program_exec *vpi;
+	struct nvfx_vertex_program_exec *vpi;
 
 	unsigned r_temps;
 	unsigned r_temps_discard;
-	struct nv40_sreg r_result[PIPE_MAX_SHADER_OUTPUTS];
-	struct nv40_sreg *r_address;
-	struct nv40_sreg *r_temp;
+	struct nvfx_sreg r_result[PIPE_MAX_SHADER_OUTPUTS];
+	struct nvfx_sreg *r_address;
+	struct nvfx_sreg *r_temp;
 
-	struct nv40_sreg *imm;
+	struct nvfx_sreg *imm;
 	unsigned nr_imm;
 
 	unsigned hpos_idx;
 };
 
-static struct nv40_sreg
-temp(struct nv40_vpc *vpc)
+static struct nvfx_sreg
+temp(struct nvfx_vpc *vpc)
 {
 	int idx = ffs(~vpc->r_temps) - 1;
 
 	if (idx < 0) {
 		NOUVEAU_ERR("out of temps!!\n");
 		assert(0);
-		return nv40_sr(NV40SR_TEMP, 0);
+		return nvfx_sr(NVFXSR_TEMP, 0);
 	}
 
 	vpc->r_temps |= (1 << idx);
 	vpc->r_temps_discard |= (1 << idx);
-	return nv40_sr(NV40SR_TEMP, idx);
+	return nvfx_sr(NVFXSR_TEMP, idx);
 }
 
 static INLINE void
-release_temps(struct nv40_vpc *vpc)
+release_temps(struct nvfx_vpc *vpc)
 {
 	vpc->r_temps &= ~vpc->r_temps_discard;
 	vpc->r_temps_discard = 0;
 }
 
-static struct nv40_sreg
-constant(struct nv40_vpc *vpc, int pipe, float x, float y, float z, float w)
+static struct nvfx_sreg
+constant(struct nvfx_vpc *vpc, int pipe, float x, float y, float z, float w)
 {
-	struct nv40_vertex_program *vp = vpc->vp;
-	struct nv40_vertex_program_data *vpd;
+	struct nvfx_vertex_program *vp = vpc->vp;
+	struct nvfx_vertex_program_data *vpd;
 	int idx;
 
 	if (pipe >= 0) {
 		for (idx = 0; idx < vp->nr_consts; idx++) {
 			if (vp->consts[idx].index == pipe)
-				return nv40_sr(NV40SR_CONST, idx);
+				return nvfx_sr(NVFXSR_CONST, idx);
 		}
 	}
 
@@ -103,70 +90,70 @@ constant(struct nv40_vpc *vpc, int pipe, float x, float y, float z, float w)
 	vpd->value[1] = y;
 	vpd->value[2] = z;
 	vpd->value[3] = w;
-	return nv40_sr(NV40SR_CONST, idx);
+	return nvfx_sr(NVFXSR_CONST, idx);
 }
 
 #define arith(cc,s,o,d,m,s0,s1,s2) \
-	nv40_vp_arith((cc), (s), NV40_VP_INST_##o, (d), (m), (s0), (s1), (s2))
+	nvfx_vp_arith(nvfx, (cc), NVFX_VP_INST_SLOT_##s, NVFX_VP_INST_##s##_OP_##o, (d), (m), (s0), (s1), (s2))
 
 static void
-emit_src(struct nv40_vpc *vpc, uint32_t *hw, int pos, struct nv40_sreg src)
+emit_src(struct nvfx_context* nvfx, struct nvfx_vpc *vpc, uint32_t *hw, int pos, struct nvfx_sreg src)
 {
-	struct nv40_vertex_program *vp = vpc->vp;
+	struct nvfx_vertex_program *vp = vpc->vp;
 	uint32_t sr = 0;
 
 	switch (src.type) {
-	case NV40SR_TEMP:
-		sr |= (NV40_VP_SRC_REG_TYPE_TEMP << NV40_VP_SRC_REG_TYPE_SHIFT);
-		sr |= (src.index << NV40_VP_SRC_TEMP_SRC_SHIFT);
+	case NVFXSR_TEMP:
+		sr |= (NVFX_VP(SRC_REG_TYPE_TEMP) << NVFX_VP(SRC_REG_TYPE_SHIFT));
+		sr |= (src.index << NVFX_VP(SRC_TEMP_SRC_SHIFT));
 		break;
-	case NV40SR_INPUT:
-		sr |= (NV40_VP_SRC_REG_TYPE_INPUT <<
-		       NV40_VP_SRC_REG_TYPE_SHIFT);
+	case NVFXSR_INPUT:
+		sr |= (NVFX_VP(SRC_REG_TYPE_INPUT) <<
+		       NVFX_VP(SRC_REG_TYPE_SHIFT));
 		vp->ir |= (1 << src.index);
-		hw[1] |= (src.index << NV40_VP_INST_INPUT_SRC_SHIFT);
+		hw[1] |= (src.index << NVFX_VP(INST_INPUT_SRC_SHIFT));
 		break;
-	case NV40SR_CONST:
-		sr |= (NV40_VP_SRC_REG_TYPE_CONST <<
-		       NV40_VP_SRC_REG_TYPE_SHIFT);
+	case NVFXSR_CONST:
+		sr |= (NVFX_VP(SRC_REG_TYPE_CONST) <<
+		       NVFX_VP(SRC_REG_TYPE_SHIFT));
 		assert(vpc->vpi->const_index == -1 ||
 		       vpc->vpi->const_index == src.index);
 		vpc->vpi->const_index = src.index;
 		break;
-	case NV40SR_NONE:
-		sr |= (NV40_VP_SRC_REG_TYPE_INPUT <<
-		       NV40_VP_SRC_REG_TYPE_SHIFT);
+	case NVFXSR_NONE:
+		sr |= (NVFX_VP(SRC_REG_TYPE_INPUT) <<
+		       NVFX_VP(SRC_REG_TYPE_SHIFT));
 		break;
 	default:
 		assert(0);
 	}
 
 	if (src.negate)
-		sr |= NV40_VP_SRC_NEGATE;
+		sr |= NVFX_VP(SRC_NEGATE);
 
 	if (src.abs)
 		hw[0] |= (1 << (21 + pos));
 
-	sr |= ((src.swz[0] << NV40_VP_SRC_SWZ_X_SHIFT) |
-	       (src.swz[1] << NV40_VP_SRC_SWZ_Y_SHIFT) |
-	       (src.swz[2] << NV40_VP_SRC_SWZ_Z_SHIFT) |
-	       (src.swz[3] << NV40_VP_SRC_SWZ_W_SHIFT));
+	sr |= ((src.swz[0] << NVFX_VP(SRC_SWZ_X_SHIFT)) |
+	       (src.swz[1] << NVFX_VP(SRC_SWZ_Y_SHIFT)) |
+	       (src.swz[2] << NVFX_VP(SRC_SWZ_Z_SHIFT)) |
+	       (src.swz[3] << NVFX_VP(SRC_SWZ_W_SHIFT)));
 
 	switch (pos) {
 	case 0:
-		hw[1] |= ((sr & NV40_VP_SRC0_HIGH_MASK) >>
-			  NV40_VP_SRC0_HIGH_SHIFT) << NV40_VP_INST_SRC0H_SHIFT;
-		hw[2] |= (sr & NV40_VP_SRC0_LOW_MASK) <<
-			  NV40_VP_INST_SRC0L_SHIFT;
+		hw[1] |= ((sr & NVFX_VP(SRC0_HIGH_MASK)) >>
+			  NVFX_VP(SRC0_HIGH_SHIFT)) << NVFX_VP(INST_SRC0H_SHIFT);
+		hw[2] |= (sr & NVFX_VP(SRC0_LOW_MASK)) <<
+			  NVFX_VP(INST_SRC0L_SHIFT);
 		break;
 	case 1:
-		hw[2] |= sr << NV40_VP_INST_SRC1_SHIFT;
+		hw[2] |= sr << NVFX_VP(INST_SRC1_SHIFT);
 		break;
 	case 2:
-		hw[2] |= ((sr & NV40_VP_SRC2_HIGH_MASK) >>
-			  NV40_VP_SRC2_HIGH_SHIFT) << NV40_VP_INST_SRC2H_SHIFT;
-		hw[3] |= (sr & NV40_VP_SRC2_LOW_MASK) <<
-			  NV40_VP_INST_SRC2L_SHIFT;
+		hw[2] |= ((sr & NVFX_VP(SRC2_HIGH_MASK)) >>
+			  NVFX_VP(SRC2_HIGH_SHIFT)) << NVFX_VP(INST_SRC2H_SHIFT);
+		hw[3] |= (sr & NVFX_VP(SRC2_LOW_MASK)) <<
+			  NVFX_VP(INST_SRC2L_SHIFT);
 		break;
 	default:
 		assert(0);
@@ -174,78 +161,114 @@ emit_src(struct nv40_vpc *vpc, uint32_t *hw, int pos, struct nv40_sreg src)
 }
 
 static void
-emit_dst(struct nv40_vpc *vpc, uint32_t *hw, int slot, struct nv40_sreg dst)
+emit_dst(struct nvfx_context* nvfx, struct nvfx_vpc *vpc, uint32_t *hw, int slot, struct nvfx_sreg dst)
 {
-	struct nv40_vertex_program *vp = vpc->vp;
+	struct nvfx_vertex_program *vp = vpc->vp;
 
 	switch (dst.type) {
-	case NV40SR_TEMP:
-		hw[3] |= NV40_VP_INST_DEST_MASK;
-		if (slot == 0) {
-			hw[0] |= (dst.index <<
-				  NV40_VP_INST_VEC_DEST_TEMP_SHIFT);
-		} else {
-			hw[3] |= (dst.index << 
-				  NV40_VP_INST_SCA_DEST_TEMP_SHIFT);
+	case NVFXSR_TEMP:
+		if(!nvfx->is_nv4x)
+			hw[0] |= (dst.index << NV30_VP_INST_DEST_TEMP_ID_SHIFT);
+		else {
+			hw[3] |= NV40_VP_INST_DEST_MASK;
+			if (slot == 0) {
+				hw[0] |= (dst.index <<
+					  NV40_VP_INST_VEC_DEST_TEMP_SHIFT);
+			} else {
+				hw[3] |= (dst.index <<
+					  NV40_VP_INST_SCA_DEST_TEMP_SHIFT);
+			}
 		}
 		break;
-	case NV40SR_OUTPUT:
+	case NVFXSR_OUTPUT:
+		/* TODO: this may be wrong because on nv30 COL0 and BFC0 are swapped */
 		switch (dst.index) {
-		case NV40_VP_INST_DEST_COL0 : vp->or |= (1 << 0); break;
-		case NV40_VP_INST_DEST_COL1 : vp->or |= (1 << 1); break;
-		case NV40_VP_INST_DEST_BFC0 : vp->or |= (1 << 2); break;
-		case NV40_VP_INST_DEST_BFC1 : vp->or |= (1 << 3); break;
-		case NV40_VP_INST_DEST_FOGC : vp->or |= (1 << 4); break;
-		case NV40_VP_INST_DEST_PSZ  : vp->or |= (1 << 5); break;
-		case NV40_VP_INST_DEST_TC(0): vp->or |= (1 << 14); break;
-		case NV40_VP_INST_DEST_TC(1): vp->or |= (1 << 15); break;
-		case NV40_VP_INST_DEST_TC(2): vp->or |= (1 << 16); break;
-		case NV40_VP_INST_DEST_TC(3): vp->or |= (1 << 17); break;
-		case NV40_VP_INST_DEST_TC(4): vp->or |= (1 << 18); break;
-		case NV40_VP_INST_DEST_TC(5): vp->or |= (1 << 19); break;
-		case NV40_VP_INST_DEST_TC(6): vp->or |= (1 << 20); break;
-		case NV40_VP_INST_DEST_TC(7): vp->or |= (1 << 21); break;
-		case NV40_VP_INST_DEST_CLIP(0):
+		case NVFX_VP_INST_DEST_CLIP(0):
 			vp->or |= (1 << 6);
-			vp->clip_ctrl |= NV40TCL_CLIP_PLANE_ENABLE_PLANE0;
-			dst.index = NV40_VP_INST_DEST_FOGC;
+			vp->clip_ctrl |= NV34TCL_VP_CLIP_PLANES_ENABLE_PLANE0;
+			dst.index = NVFX_VP(INST_DEST_FOGC);
 			break;
-		case NV40_VP_INST_DEST_CLIP(1):
+		case NVFX_VP_INST_DEST_CLIP(1):
 			vp->or |= (1 << 7);
-			vp->clip_ctrl |= NV40TCL_CLIP_PLANE_ENABLE_PLANE1;
-			dst.index = NV40_VP_INST_DEST_FOGC;
+			vp->clip_ctrl |= NV34TCL_VP_CLIP_PLANES_ENABLE_PLANE1;
+			dst.index = NVFX_VP(INST_DEST_FOGC);
 			break;
-		case NV40_VP_INST_DEST_CLIP(2):
+		case NVFX_VP_INST_DEST_CLIP(2):
 			vp->or |= (1 << 8);
-			vp->clip_ctrl |= NV40TCL_CLIP_PLANE_ENABLE_PLANE2;
-			dst.index = NV40_VP_INST_DEST_FOGC;
+			vp->clip_ctrl |= NV34TCL_VP_CLIP_PLANES_ENABLE_PLANE2;
+			dst.index = NVFX_VP(INST_DEST_FOGC);
 			break;
-		case NV40_VP_INST_DEST_CLIP(3):
+		case NVFX_VP_INST_DEST_CLIP(3):
 			vp->or |= (1 << 9);
-			vp->clip_ctrl |= NV40TCL_CLIP_PLANE_ENABLE_PLANE3;
-			dst.index = NV40_VP_INST_DEST_PSZ;
+			vp->clip_ctrl |= NV34TCL_VP_CLIP_PLANES_ENABLE_PLANE3;
+			dst.index = NVFX_VP(INST_DEST_PSZ);
 			break;
-		case NV40_VP_INST_DEST_CLIP(4):
+		case NVFX_VP_INST_DEST_CLIP(4):
 			vp->or |= (1 << 10);
-			vp->clip_ctrl |= NV40TCL_CLIP_PLANE_ENABLE_PLANE4;
-			dst.index = NV40_VP_INST_DEST_PSZ;
+			vp->clip_ctrl |= NV34TCL_VP_CLIP_PLANES_ENABLE_PLANE4;
+			dst.index = NVFX_VP(INST_DEST_PSZ);
 			break;
-		case NV40_VP_INST_DEST_CLIP(5):
+		case NVFX_VP_INST_DEST_CLIP(5):
 			vp->or |= (1 << 11);
-			vp->clip_ctrl |= NV40TCL_CLIP_PLANE_ENABLE_PLANE5;
-			dst.index = NV40_VP_INST_DEST_PSZ;
+			vp->clip_ctrl |= NV34TCL_VP_CLIP_PLANES_ENABLE_PLANE5;
+			dst.index = NVFX_VP(INST_DEST_PSZ);
 			break;
 		default:
+			if(!nvfx->is_nv4x) {
+				switch (dst.index) {
+				case NV30_VP_INST_DEST_COL0 : vp->or |= (1 << 0); break;
+				case NV30_VP_INST_DEST_COL1 : vp->or |= (1 << 1); break;
+				case NV30_VP_INST_DEST_BFC0 : vp->or |= (1 << 2); break;
+				case NV30_VP_INST_DEST_BFC1 : vp->or |= (1 << 3); break;
+				case NV30_VP_INST_DEST_FOGC: vp->or |= (1 << 4); break;
+				case NV30_VP_INST_DEST_PSZ  : vp->or |= (1 << 5); break;
+				case NV30_VP_INST_DEST_TC(0): vp->or |= (1 << 14); break;
+				case NV30_VP_INST_DEST_TC(1): vp->or |= (1 << 15); break;
+				case NV30_VP_INST_DEST_TC(2): vp->or |= (1 << 16); break;
+				case NV30_VP_INST_DEST_TC(3): vp->or |= (1 << 17); break;
+				case NV30_VP_INST_DEST_TC(4): vp->or |= (1 << 18); break;
+				case NV30_VP_INST_DEST_TC(5): vp->or |= (1 << 19); break;
+				case NV30_VP_INST_DEST_TC(6): vp->or |= (1 << 20); break;
+				case NV30_VP_INST_DEST_TC(7): vp->or |= (1 << 21); break;
+				}
+			} else {
+				switch (dst.index) {
+				case NV40_VP_INST_DEST_COL0 : vp->or |= (1 << 0); break;
+				case NV40_VP_INST_DEST_COL1 : vp->or |= (1 << 1); break;
+				case NV40_VP_INST_DEST_BFC0 : vp->or |= (1 << 2); break;
+				case NV40_VP_INST_DEST_BFC1 : vp->or |= (1 << 3); break;
+				case NV40_VP_INST_DEST_FOGC: vp->or |= (1 << 4); break;
+				case NV40_VP_INST_DEST_PSZ  : vp->or |= (1 << 5); break;
+				case NV40_VP_INST_DEST_TC(0): vp->or |= (1 << 14); break;
+				case NV40_VP_INST_DEST_TC(1): vp->or |= (1 << 15); break;
+				case NV40_VP_INST_DEST_TC(2): vp->or |= (1 << 16); break;
+				case NV40_VP_INST_DEST_TC(3): vp->or |= (1 << 17); break;
+				case NV40_VP_INST_DEST_TC(4): vp->or |= (1 << 18); break;
+				case NV40_VP_INST_DEST_TC(5): vp->or |= (1 << 19); break;
+				case NV40_VP_INST_DEST_TC(6): vp->or |= (1 << 20); break;
+				case NV40_VP_INST_DEST_TC(7): vp->or |= (1 << 21); break;
+				}
+			}
 			break;
 		}
 
-		hw[3] |= (dst.index << NV40_VP_INST_DEST_SHIFT);
-		if (slot == 0) {
-			hw[0] |= NV40_VP_INST_VEC_RESULT;
-			hw[0] |= NV40_VP_INST_VEC_DEST_TEMP_MASK | (1<<20);
+		if(!nvfx->is_nv4x) {
+			hw[3] |= (dst.index << NV30_VP_INST_DEST_SHIFT);
+			hw[0] |= NV30_VP_INST_VEC_DEST_TEMP_MASK | (1<<20);
+
+			/*XXX: no way this is entirely correct, someone needs to
+			 *     figure out what exactly it is.
+			 */
+			hw[3] |= 0x800;
 		} else {
-			hw[3] |= NV40_VP_INST_SCA_RESULT;
-			hw[3] |= NV40_VP_INST_SCA_DEST_TEMP_MASK;
+			hw[3] |= (dst.index << NV40_VP_INST_DEST_SHIFT);
+			if (slot == 0) {
+				hw[0] |= NV40_VP_INST_VEC_RESULT;
+				hw[0] |= NV40_VP_INST_VEC_DEST_TEMP_MASK | (1<<20);
+			} else {
+				hw[3] |= NV40_VP_INST_SCA_RESULT;
+				hw[3] |= NV40_VP_INST_SCA_DEST_TEMP_MASK;
+			}
 		}
 		break;
 	default:
@@ -254,12 +277,12 @@ emit_dst(struct nv40_vpc *vpc, uint32_t *hw, int slot, struct nv40_sreg dst)
 }
 
 static void
-nv40_vp_arith(struct nv40_vpc *vpc, int slot, int op,
-	      struct nv40_sreg dst, int mask,
-	      struct nv40_sreg s0, struct nv40_sreg s1,
-	      struct nv40_sreg s2)
+nvfx_vp_arith(struct nvfx_context* nvfx, struct nvfx_vpc *vpc, int slot, int op,
+	      struct nvfx_sreg dst, int mask,
+	      struct nvfx_sreg s0, struct nvfx_sreg s1,
+	      struct nvfx_sreg s2)
 {
-	struct nv40_vertex_program *vp = vpc->vp;
+	struct nvfx_vertex_program *vp = vpc->vp;
 	uint32_t *hw;
 
 	vp->insns = realloc(vp->insns, ++vp->nr_insns * sizeof(*vpc->vpi));
@@ -269,35 +292,53 @@ nv40_vp_arith(struct nv40_vpc *vpc, int slot, int op,
 
 	hw = vpc->vpi->data;
 
-	hw[0] |= (NV40_VP_INST_COND_TR << NV40_VP_INST_COND_SHIFT);
-	hw[0] |= ((0 << NV40_VP_INST_COND_SWZ_X_SHIFT) |
-		  (1 << NV40_VP_INST_COND_SWZ_Y_SHIFT) |
-		  (2 << NV40_VP_INST_COND_SWZ_Z_SHIFT) |
-		  (3 << NV40_VP_INST_COND_SWZ_W_SHIFT));
-
-	if (slot == 0) {
-		hw[1] |= (op << NV40_VP_INST_VEC_OPCODE_SHIFT);
-		hw[3] |= NV40_VP_INST_SCA_DEST_TEMP_MASK;
-		hw[3] |= (mask << NV40_VP_INST_VEC_WRITEMASK_SHIFT);
-	} else {
-		hw[1] |= (op << NV40_VP_INST_SCA_OPCODE_SHIFT);
-		hw[0] |= (NV40_VP_INST_VEC_DEST_TEMP_MASK | (1 << 20));
-		hw[3] |= (mask << NV40_VP_INST_SCA_WRITEMASK_SHIFT);
+	hw[0] |= (NVFX_COND_TR << NVFX_VP(INST_COND_SHIFT));
+	hw[0] |= ((0 << NVFX_VP(INST_COND_SWZ_X_SHIFT)) |
+		  (1 << NVFX_VP(INST_COND_SWZ_Y_SHIFT)) |
+		  (2 << NVFX_VP(INST_COND_SWZ_Z_SHIFT)) |
+		  (3 << NVFX_VP(INST_COND_SWZ_W_SHIFT)));
+
+	if(!nvfx->is_nv4x) {
+		hw[1] |= (op << NV30_VP_INST_VEC_OPCODE_SHIFT);
+//		hw[3] |= NVFX_VP(INST_SCA_DEST_TEMP_MASK);
+//		hw[3] |= (mask << NVFX_VP(INST_VEC_WRITEMASK_SHIFT));
+
+		if (dst.type == NVFXSR_OUTPUT) {
+			if (slot)
+				hw[3] |= (mask << NV30_VP_INST_SDEST_WRITEMASK_SHIFT);
+			else
+				hw[3] |= (mask << NV30_VP_INST_VDEST_WRITEMASK_SHIFT);
+		} else {
+			if (slot)
+				hw[3] |= (mask << NV30_VP_INST_STEMP_WRITEMASK_SHIFT);
+			else
+				hw[3] |= (mask << NV30_VP_INST_VTEMP_WRITEMASK_SHIFT);
+		}
+	 } else {
+		if (slot == 0) {
+			hw[1] |= (op << NV40_VP_INST_VEC_OPCODE_SHIFT);
+			hw[3] |= NV40_VP_INST_SCA_DEST_TEMP_MASK;
+			hw[3] |= (mask << NV40_VP_INST_VEC_WRITEMASK_SHIFT);
+	    } else {
+			hw[1] |= (op << NV40_VP_INST_SCA_OPCODE_SHIFT);
+			hw[0] |= (NV40_VP_INST_VEC_DEST_TEMP_MASK | (1 << 20));
+			hw[3] |= (mask << NV40_VP_INST_SCA_WRITEMASK_SHIFT);
+		}
 	}
 
-	emit_dst(vpc, hw, slot, dst);
-	emit_src(vpc, hw, 0, s0);
-	emit_src(vpc, hw, 1, s1);
-	emit_src(vpc, hw, 2, s2);
+	emit_dst(nvfx, vpc, hw, slot, dst);
+	emit_src(nvfx, vpc, hw, 0, s0);
+	emit_src(nvfx, vpc, hw, 1, s1);
+	emit_src(nvfx, vpc, hw, 2, s2);
 }
 
-static INLINE struct nv40_sreg
-tgsi_src(struct nv40_vpc *vpc, const struct tgsi_full_src_register *fsrc) {
-	struct nv40_sreg src;
+static INLINE struct nvfx_sreg
+tgsi_src(struct nvfx_vpc *vpc, const struct tgsi_full_src_register *fsrc) {
+	struct nvfx_sreg src;
 
 	switch (fsrc->Register.File) {
 	case TGSI_FILE_INPUT:
-		src = nv40_sr(NV40SR_INPUT, fsrc->Register.Index);
+		src = nvfx_sr(NVFXSR_INPUT, fsrc->Register.Index);
 		break;
 	case TGSI_FILE_CONSTANT:
 		src = constant(vpc, fsrc->Register.Index, 0, 0, 0, 0);
@@ -322,9 +363,9 @@ tgsi_src(struct nv40_vpc *vpc, const struct tgsi_full_src_register *fsrc) {
 	return src;
 }
 
-static INLINE struct nv40_sreg
-tgsi_dst(struct nv40_vpc *vpc, const struct tgsi_full_dst_register *fdst) {
-	struct nv40_sreg dst;
+static INLINE struct nvfx_sreg
+tgsi_dst(struct nvfx_vpc *vpc, const struct tgsi_full_dst_register *fdst) {
+	struct nvfx_sreg dst;
 
 	switch (fdst->Register.File) {
 	case TGSI_FILE_OUTPUT:
@@ -349,52 +390,19 @@ tgsi_mask(uint tgsi)
 {
 	int mask = 0;
 
-	if (tgsi & TGSI_WRITEMASK_X) mask |= MASK_X;
-	if (tgsi & TGSI_WRITEMASK_Y) mask |= MASK_Y;
-	if (tgsi & TGSI_WRITEMASK_Z) mask |= MASK_Z;
-	if (tgsi & TGSI_WRITEMASK_W) mask |= MASK_W;
+	if (tgsi & TGSI_WRITEMASK_X) mask |= NVFX_VP_MASK_X;
+	if (tgsi & TGSI_WRITEMASK_Y) mask |= NVFX_VP_MASK_Y;
+	if (tgsi & TGSI_WRITEMASK_Z) mask |= NVFX_VP_MASK_Z;
+	if (tgsi & TGSI_WRITEMASK_W) mask |= NVFX_VP_MASK_W;
 	return mask;
 }
 
 static boolean
-src_native_swz(struct nv40_vpc *vpc, const struct tgsi_full_src_register *fsrc,
-	       struct nv40_sreg *src)
-{
-	const struct nv40_sreg none = nv40_sr(NV40SR_NONE, 0);
-	struct nv40_sreg tgsi = tgsi_src(vpc, fsrc);
-	uint mask = 0;
-	uint c;
-
-	for (c = 0; c < 4; c++) {
-		switch (tgsi_util_get_full_src_register_swizzle(fsrc, c)) {
-		case TGSI_SWIZZLE_X:
-		case TGSI_SWIZZLE_Y:
-		case TGSI_SWIZZLE_Z:
-		case TGSI_SWIZZLE_W:
-			mask |= tgsi_mask(1 << c);
-			break;
-		default:
-			assert(0);
-		}
-	}
-
-	if (mask == MASK_ALL)
-		return TRUE;
-
-	*src = temp(vpc);
-
-	if (mask)
-		arith(vpc, 0, OP_MOV, *src, mask, tgsi, none, none);
-
-	return FALSE;
-}
-
-static boolean
-nv40_vertprog_parse_instruction(struct nv40_vpc *vpc,
+nvfx_vertprog_parse_instruction(struct nvfx_context* nvfx, struct nvfx_vpc *vpc,
 				const struct tgsi_full_instruction *finst)
 {
-	struct nv40_sreg src[3], dst, tmp;
-	struct nv40_sreg none = nv40_sr(NV40SR_NONE, 0);
+	struct nvfx_sreg src[3], dst, tmp;
+	struct nvfx_sreg none = nvfx_sr(NVFXSR_NONE, 0);
 	int mask;
 	int ai = -1, ci = -1, ii = -1;
 	int i;
@@ -418,23 +426,12 @@ nv40_vertprog_parse_instruction(struct nv40_vpc *vpc,
 
 		switch (fsrc->Register.File) {
 		case TGSI_FILE_INPUT:
-		case TGSI_FILE_CONSTANT:
-		case TGSI_FILE_TEMPORARY:
-			if (!src_native_swz(vpc, fsrc, &src[i]))
-				continue;
-			break;
-		default:
-			break;
-		}
-
-		switch (fsrc->Register.File) {
-		case TGSI_FILE_INPUT:
 			if (ai == -1 || ai == fsrc->Register.Index) {
 				ai = fsrc->Register.Index;
 				src[i] = tgsi_src(vpc, fsrc);
 			} else {
 				src[i] = temp(vpc);
-				arith(vpc, 0, OP_MOV, src[i], MASK_ALL,
+				arith(vpc, VEC, MOV, src[i], NVFX_VP_MASK_ALL,
 				      tgsi_src(vpc, fsrc), none, none);
 			}
 			break;
@@ -445,7 +442,7 @@ nv40_vertprog_parse_instruction(struct nv40_vpc *vpc,
 				src[i] = tgsi_src(vpc, fsrc);
 			} else {
 				src[i] = temp(vpc);
-				arith(vpc, 0, OP_MOV, src[i], MASK_ALL,
+				arith(vpc, VEC, MOV, src[i], NVFX_VP_MASK_ALL,
 				      tgsi_src(vpc, fsrc), none, none);
 			}
 			break;
@@ -456,7 +453,7 @@ nv40_vertprog_parse_instruction(struct nv40_vpc *vpc,
 				src[i] = tgsi_src(vpc, fsrc);
 			} else {
 				src[i] = temp(vpc);
-				arith(vpc, 0, OP_MOV, src[i], MASK_ALL,
+				arith(vpc, VEC, MOV, src[i], NVFX_VP_MASK_ALL,
 				      tgsi_src(vpc, fsrc), none, none);
 			}
 			break;
@@ -474,93 +471,96 @@ nv40_vertprog_parse_instruction(struct nv40_vpc *vpc,
 
 	switch (finst->Instruction.Opcode) {
 	case TGSI_OPCODE_ABS:
-		arith(vpc, 0, OP_MOV, dst, mask, abs(src[0]), none, none);
+		arith(vpc, VEC, MOV, dst, mask, abs(src[0]), none, none);
 		break;
 	case TGSI_OPCODE_ADD:
-		arith(vpc, 0, OP_ADD, dst, mask, src[0], none, src[1]);
+		arith(vpc, VEC, ADD, dst, mask, src[0], none, src[1]);
 		break;
 	case TGSI_OPCODE_ARL:
-		arith(vpc, 0, OP_ARL, dst, mask, src[0], none, none);
+		arith(vpc, VEC, ARL, dst, mask, src[0], none, none);
 		break;
 	case TGSI_OPCODE_DP3:
-		arith(vpc, 0, OP_DP3, dst, mask, src[0], src[1], none);
+		arith(vpc, VEC, DP3, dst, mask, src[0], src[1], none);
 		break;
 	case TGSI_OPCODE_DP4:
-		arith(vpc, 0, OP_DP4, dst, mask, src[0], src[1], none);
+		arith(vpc, VEC, DP4, dst, mask, src[0], src[1], none);
 		break;
 	case TGSI_OPCODE_DPH:
-		arith(vpc, 0, OP_DPH, dst, mask, src[0], src[1], none);
+		arith(vpc, VEC, DPH, dst, mask, src[0], src[1], none);
 		break;
 	case TGSI_OPCODE_DST:
-		arith(vpc, 0, OP_DST, dst, mask, src[0], src[1], none);
+		arith(vpc, VEC, DST, dst, mask, src[0], src[1], none);
 		break;
 	case TGSI_OPCODE_EX2:
-		arith(vpc, 1, OP_EX2, dst, mask, none, none, src[0]);
+		arith(vpc, SCA, EX2, dst, mask, none, none, src[0]);
 		break;
 	case TGSI_OPCODE_EXP:
-		arith(vpc, 1, OP_EXP, dst, mask, none, none, src[0]);
+		arith(vpc, SCA, EXP, dst, mask, none, none, src[0]);
 		break;
 	case TGSI_OPCODE_FLR:
-		arith(vpc, 0, OP_FLR, dst, mask, src[0], none, none);
+		arith(vpc, VEC, FLR, dst, mask, src[0], none, none);
 		break;
 	case TGSI_OPCODE_FRC:
-		arith(vpc, 0, OP_FRC, dst, mask, src[0], none, none);
+		arith(vpc, VEC, FRC, dst, mask, src[0], none, none);
 		break;
 	case TGSI_OPCODE_LG2:
-		arith(vpc, 1, OP_LG2, dst, mask, none, none, src[0]);
+		arith(vpc, SCA, LG2, dst, mask, none, none, src[0]);
 		break;
 	case TGSI_OPCODE_LIT:
-		arith(vpc, 1, OP_LIT, dst, mask, none, none, src[0]);
+		arith(vpc, SCA, LIT, dst, mask, none, none, src[0]);
 		break;
 	case TGSI_OPCODE_LOG:
-		arith(vpc, 1, OP_LOG, dst, mask, none, none, src[0]);
+		arith(vpc, SCA, LOG, dst, mask, none, none, src[0]);
 		break;
 	case TGSI_OPCODE_MAD:
-		arith(vpc, 0, OP_MAD, dst, mask, src[0], src[1], src[2]);
+		arith(vpc, VEC, MAD, dst, mask, src[0], src[1], src[2]);
 		break;
 	case TGSI_OPCODE_MAX:
-		arith(vpc, 0, OP_MAX, dst, mask, src[0], src[1], none);
+		arith(vpc, VEC, MAX, dst, mask, src[0], src[1], none);
 		break;
 	case TGSI_OPCODE_MIN:
-		arith(vpc, 0, OP_MIN, dst, mask, src[0], src[1], none);
+		arith(vpc, VEC, MIN, dst, mask, src[0], src[1], none);
 		break;
 	case TGSI_OPCODE_MOV:
-		arith(vpc, 0, OP_MOV, dst, mask, src[0], none, none);
+		arith(vpc, VEC, MOV, dst, mask, src[0], none, none);
 		break;
 	case TGSI_OPCODE_MUL:
-		arith(vpc, 0, OP_MUL, dst, mask, src[0], src[1], none);
+		arith(vpc, VEC, MUL, dst, mask, src[0], src[1], none);
 		break;
 	case TGSI_OPCODE_POW:
 		tmp = temp(vpc);
-		arith(vpc, 1, OP_LG2, tmp, MASK_X, none, none,
+		arith(vpc, SCA, LG2, tmp, NVFX_VP_MASK_X, none, none,
 		      swz(src[0], X, X, X, X));
-		arith(vpc, 0, OP_MUL, tmp, MASK_X, swz(tmp, X, X, X, X),
+		arith(vpc, VEC, MUL, tmp, NVFX_VP_MASK_X, swz(tmp, X, X, X, X),
 		      swz(src[1], X, X, X, X), none);
-		arith(vpc, 1, OP_EX2, dst, mask, none, none,
+		arith(vpc, SCA, EX2, dst, mask, none, none,
 		      swz(tmp, X, X, X, X));
 		break;
 	case TGSI_OPCODE_RCP:
-		arith(vpc, 1, OP_RCP, dst, mask, none, none, src[0]);
+		arith(vpc, SCA, RCP, dst, mask, none, none, src[0]);
 		break;
 	case TGSI_OPCODE_RET:
 		break;
 	case TGSI_OPCODE_RSQ:
-		arith(vpc, 1, OP_RSQ, dst, mask, none, none, abs(src[0]));
+		arith(vpc, SCA, RSQ, dst, mask, none, none, abs(src[0]));
 		break;
 	case TGSI_OPCODE_SGE:
-		arith(vpc, 0, OP_SGE, dst, mask, src[0], src[1], none);
+		arith(vpc, VEC, SGE, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_SGT:
+		arith(vpc, VEC, SGT, dst, mask, src[0], src[1], none);
 		break;
 	case TGSI_OPCODE_SLT:
-		arith(vpc, 0, OP_SLT, dst, mask, src[0], src[1], none);
+		arith(vpc, VEC, SLT, dst, mask, src[0], src[1], none);
 		break;
 	case TGSI_OPCODE_SUB:
-		arith(vpc, 0, OP_ADD, dst, mask, src[0], none, neg(src[1]));
+		arith(vpc, VEC, ADD, dst, mask, src[0], none, neg(src[1]));
 		break;
 	case TGSI_OPCODE_XPD:
 		tmp = temp(vpc);
-		arith(vpc, 0, OP_MUL, tmp, mask,
+		arith(vpc, VEC, MUL, tmp, mask,
 		      swz(src[0], Z, X, Y, Y), swz(src[1], Y, Z, X, X), none);
-		arith(vpc, 0, OP_MAD, dst, (mask & ~MASK_W),
+		arith(vpc, VEC, MAD, dst, (mask & ~NVFX_VP_MASK_W),
 		      swz(src[0], Y, Z, X, X), swz(src[1], Z, X, Y, Y),
 		      neg(tmp));
 		break;
@@ -574,7 +574,7 @@ nv40_vertprog_parse_instruction(struct nv40_vpc *vpc,
 }
 
 static boolean
-nv40_vertprog_parse_decl_output(struct nv40_vpc *vpc,
+nvfx_vertprog_parse_decl_output(struct nvfx_context* nvfx, struct nvfx_vpc *vpc,
 				const struct tgsi_full_declaration *fdec)
 {
 	unsigned idx = fdec->Range.First;
@@ -582,15 +582,15 @@ nv40_vertprog_parse_decl_output(struct nv40_vpc *vpc,
 
 	switch (fdec->Semantic.Name) {
 	case TGSI_SEMANTIC_POSITION:
-		hw = NV40_VP_INST_DEST_POS;
+		hw = NVFX_VP(INST_DEST_POS);
 		vpc->hpos_idx = idx;
 		break;
 	case TGSI_SEMANTIC_COLOR:
 		if (fdec->Semantic.Index == 0) {
-			hw = NV40_VP_INST_DEST_COL0;
+			hw = NVFX_VP(INST_DEST_COL0);
 		} else
 		if (fdec->Semantic.Index == 1) {
-			hw = NV40_VP_INST_DEST_COL1;
+			hw = NVFX_VP(INST_DEST_COL1);
 		} else {
 			NOUVEAU_ERR("bad colour semantic index\n");
 			return FALSE;
@@ -598,24 +598,24 @@ nv40_vertprog_parse_decl_output(struct nv40_vpc *vpc,
 		break;
 	case TGSI_SEMANTIC_BCOLOR:
 		if (fdec->Semantic.Index == 0) {
-			hw = NV40_VP_INST_DEST_BFC0;
+			hw = NVFX_VP(INST_DEST_BFC0);
 		} else
 		if (fdec->Semantic.Index == 1) {
-			hw = NV40_VP_INST_DEST_BFC1;
+			hw = NVFX_VP(INST_DEST_BFC1);
 		} else {
 			NOUVEAU_ERR("bad bcolour semantic index\n");
 			return FALSE;
 		}
 		break;
 	case TGSI_SEMANTIC_FOG:
-		hw = NV40_VP_INST_DEST_FOGC;
+		hw = NVFX_VP(INST_DEST_FOGC);
 		break;
 	case TGSI_SEMANTIC_PSIZE:
-		hw = NV40_VP_INST_DEST_PSZ;
+		hw = NVFX_VP(INST_DEST_PSZ);
 		break;
 	case TGSI_SEMANTIC_GENERIC:
 		if (fdec->Semantic.Index <= 7) {
-			hw = NV40_VP_INST_DEST_TC(fdec->Semantic.Index);
+			hw = NVFX_VP(INST_DEST_TC(fdec->Semantic.Index));
 		} else {
 			NOUVEAU_ERR("bad generic semantic index\n");
 			return FALSE;
@@ -630,12 +630,12 @@ nv40_vertprog_parse_decl_output(struct nv40_vpc *vpc,
 		return FALSE;
 	}
 
-	vpc->r_result[idx] = nv40_sr(NV40SR_OUTPUT, hw);
+	vpc->r_result[idx] = nvfx_sr(NVFXSR_OUTPUT, hw);
 	return TRUE;
 }
 
 static boolean
-nv40_vertprog_prepare(struct nv40_vpc *vpc)
+nvfx_vertprog_prepare(struct nvfx_context* nvfx, struct nvfx_vpc *vpc)
 {
 	struct tgsi_parse_context p;
 	int high_temp = -1, high_addr = -1, nr_imm = 0, i;
@@ -670,7 +670,7 @@ nv40_vertprog_prepare(struct nv40_vpc *vpc)
 				break;
 #endif
 			case TGSI_FILE_OUTPUT:
-				if (!nv40_vertprog_parse_decl_output(vpc, fdec))
+				if (!nvfx_vertprog_parse_decl_output(nvfx, vpc, fdec))
 					return FALSE;
 				break;
 			default:
@@ -691,7 +691,7 @@ nv40_vertprog_prepare(struct nv40_vpc *vpc)
 				if (fdst->Register.Index > high_addr)
 					high_addr = fdst->Register.Index;
 			}
-		
+
 		}
 			break;
 #endif
@@ -702,18 +702,18 @@ nv40_vertprog_prepare(struct nv40_vpc *vpc)
 	tgsi_parse_free(&p);
 
 	if (nr_imm) {
-		vpc->imm = CALLOC(nr_imm, sizeof(struct nv40_sreg));
+		vpc->imm = CALLOC(nr_imm, sizeof(struct nvfx_sreg));
 		assert(vpc->imm);
 	}
 
 	if (++high_temp) {
-		vpc->r_temp = CALLOC(high_temp, sizeof(struct nv40_sreg));
+		vpc->r_temp = CALLOC(high_temp, sizeof(struct nvfx_sreg));
 		for (i = 0; i < high_temp; i++)
 			vpc->r_temp[i] = temp(vpc);
 	}
 
 	if (++high_addr) {
-		vpc->r_address = CALLOC(high_addr, sizeof(struct nv40_sreg));
+		vpc->r_address = CALLOC(high_addr, sizeof(struct nvfx_sreg));
 		for (i = 0; i < high_addr; i++)
 			vpc->r_address[i] = temp(vpc);
 	}
@@ -723,20 +723,20 @@ nv40_vertprog_prepare(struct nv40_vpc *vpc)
 }
 
 static void
-nv40_vertprog_translate(struct nv40_context *nv40,
-			struct nv40_vertex_program *vp)
+nvfx_vertprog_translate(struct nvfx_context *nvfx,
+			struct nvfx_vertex_program *vp)
 {
 	struct tgsi_parse_context parse;
-	struct nv40_vpc *vpc = NULL;
-	struct nv40_sreg none = nv40_sr(NV40SR_NONE, 0);
+	struct nvfx_vpc *vpc = NULL;
+	struct nvfx_sreg none = nvfx_sr(NVFXSR_NONE, 0);
 	int i;
 
-	vpc = CALLOC(1, sizeof(struct nv40_vpc));
+	vpc = CALLOC(1, sizeof(struct nvfx_vpc));
 	if (!vpc)
 		return;
 	vpc->vp = vp;
 
-	if (!nv40_vertprog_prepare(vpc)) {
+	if (!nvfx_vertprog_prepare(nvfx, vpc)) {
 		FREE(vpc);
 		return;
 	}
@@ -775,7 +775,7 @@ nv40_vertprog_translate(struct nv40_context *nv40,
 		{
 			const struct tgsi_full_instruction *finst;
 			finst = &parse.FullToken.FullInstruction;
-			if (!nv40_vertprog_parse_instruction(vpc, finst))
+			if (!nvfx_vertprog_parse_instruction(nvfx, vpc, finst))
 				goto out_err;
 		}
 			break;
@@ -785,74 +785,74 @@ nv40_vertprog_translate(struct nv40_context *nv40,
 	}
 
 	/* Write out HPOS if it was redirected to a temp earlier */
-	if (vpc->r_result[vpc->hpos_idx].type != NV40SR_OUTPUT) {
-		struct nv40_sreg hpos = nv40_sr(NV40SR_OUTPUT,
-						NV40_VP_INST_DEST_POS);
-		struct nv40_sreg htmp = vpc->r_result[vpc->hpos_idx];
+	if (vpc->r_result[vpc->hpos_idx].type != NVFXSR_OUTPUT) {
+		struct nvfx_sreg hpos = nvfx_sr(NVFXSR_OUTPUT,
+						NVFX_VP(INST_DEST_POS));
+		struct nvfx_sreg htmp = vpc->r_result[vpc->hpos_idx];
 
-		arith(vpc, 0, OP_MOV, hpos, MASK_ALL, htmp, none, none);
+		arith(vpc, VEC, MOV, hpos, NVFX_VP_MASK_ALL, htmp, none, none);
 	}
 
 	/* Insert code to handle user clip planes */
 	for (i = 0; i < vp->ucp.nr; i++) {
-		struct nv40_sreg cdst = nv40_sr(NV40SR_OUTPUT,
-						NV40_VP_INST_DEST_CLIP(i));
-		struct nv40_sreg ceqn = constant(vpc, -1,
-						 nv40->clip.ucp[i][0],
-						 nv40->clip.ucp[i][1],
-						 nv40->clip.ucp[i][2],
-						 nv40->clip.ucp[i][3]);
-		struct nv40_sreg htmp = vpc->r_result[vpc->hpos_idx];
+		struct nvfx_sreg cdst = nvfx_sr(NVFXSR_OUTPUT,
+						NVFX_VP_INST_DEST_CLIP(i));
+		struct nvfx_sreg ceqn = constant(vpc, -1,
+						 nvfx->clip.ucp[i][0],
+						 nvfx->clip.ucp[i][1],
+						 nvfx->clip.ucp[i][2],
+						 nvfx->clip.ucp[i][3]);
+		struct nvfx_sreg htmp = vpc->r_result[vpc->hpos_idx];
 		unsigned mask;
 
 		switch (i) {
-		case 0: case 3: mask = MASK_Y; break;
-		case 1: case 4: mask = MASK_Z; break;
-		case 2: case 5: mask = MASK_W; break;
+		case 0: case 3: mask = NVFX_VP_MASK_Y; break;
+		case 1: case 4: mask = NVFX_VP_MASK_Z; break;
+		case 2: case 5: mask = NVFX_VP_MASK_W; break;
 		default:
 			NOUVEAU_ERR("invalid clip dist #%d\n", i);
 			goto out_err;
 		}
 
-		arith(vpc, 0, OP_DP4, cdst, mask, htmp, ceqn, none);
+		arith(vpc, VEC, DP4, cdst, mask, htmp, ceqn, none);
 	}
 
-	vp->insns[vp->nr_insns - 1].data[3] |= NV40_VP_INST_LAST;
+	vp->insns[vp->nr_insns - 1].data[3] |= NVFX_VP_INST_LAST;
 	vp->translated = TRUE;
 out_err:
 	tgsi_parse_free(&parse);
 	if (vpc->r_temp)
-		FREE(vpc->r_temp); 
+		FREE(vpc->r_temp);
 	if (vpc->r_address)
-		FREE(vpc->r_address); 
-	if (vpc->imm)	
-		FREE(vpc->imm); 
+		FREE(vpc->r_address);
+	if (vpc->imm)
+		FREE(vpc->imm);
 	FREE(vpc);
 }
 
 static boolean
-nv40_vertprog_validate(struct nv40_context *nv40)
-{ 
-	struct pipe_screen *pscreen = nv40->pipe.screen;
-	struct nv40_screen *screen = nv40->screen;
+nvfx_vertprog_validate(struct nvfx_context *nvfx)
+{
+	struct pipe_screen *pscreen = nvfx->pipe.screen;
+	struct nvfx_screen *screen = nvfx->screen;
 	struct nouveau_channel *chan = screen->base.channel;
-	struct nouveau_grobj *curie = screen->curie;
-	struct nv40_vertex_program *vp;
+	struct nouveau_grobj *eng3d = screen->eng3d;
+	struct nvfx_vertex_program *vp;
 	struct pipe_buffer *constbuf;
 	boolean upload_code = FALSE, upload_data = FALSE;
 	int i;
 
-	if (nv40->render_mode == HW) {
-		vp = nv40->vertprog;
-		constbuf = nv40->constbuf[PIPE_SHADER_VERTEX];
+	if (nvfx->render_mode == HW) {
+		vp = nvfx->vertprog;
+		constbuf = nvfx->constbuf[PIPE_SHADER_VERTEX];
 
-		if ((nv40->dirty & NV40_NEW_UCP) ||
-		    memcmp(&nv40->clip, &vp->ucp, sizeof(vp->ucp))) {
-			nv40_vertprog_destroy(nv40, vp);
-			memcpy(&vp->ucp, &nv40->clip, sizeof(vp->ucp));
+		if ((nvfx->dirty & NVFX_NEW_UCP) ||
+		    memcmp(&nvfx->clip, &vp->ucp, sizeof(vp->ucp))) {
+			nvfx_vertprog_destroy(nvfx, vp);
+			memcpy(&vp->ucp, &nvfx->clip, sizeof(vp->ucp));
 		}
 	} else {
-		vp = nv40->swtnl.vertprog;
+		vp = nvfx->swtnl.vertprog;
 		constbuf = NULL;
 	}
 
@@ -860,24 +860,24 @@ nv40_vertprog_validate(struct nv40_context *nv40)
 	if (vp->translated)
 		goto check_gpu_resources;
 
-	nv40->fallback_swtnl &= ~NV40_NEW_VERTPROG;
-	nv40_vertprog_translate(nv40, vp);
+	nvfx->fallback_swtnl &= ~NVFX_NEW_VERTPROG;
+		nvfx_vertprog_translate(nvfx, vp);
 	if (!vp->translated) {
-		nv40->fallback_swtnl |= NV40_NEW_VERTPROG;
-		return FALSE;
+		nvfx->fallback_swtnl |= NVFX_NEW_VERTPROG;
+			return FALSE;
 	}
 
 check_gpu_resources:
 	/* Allocate hw vtxprog exec slots */
 	if (!vp->exec) {
-		struct nouveau_resource *heap = nv40->screen->vp_exec_heap;
+		struct nouveau_resource *heap = nvfx->screen->vp_exec_heap;
 		struct nouveau_stateobj *so;
 		uint vplen = vp->nr_insns;
 
 		if (nouveau_resource_alloc(heap, vplen, vp, &vp->exec)) {
 			while (heap->next && heap->size < vplen) {
-				struct nv40_vertex_program *evict;
-				
+				struct nvfx_vertex_program *evict;
+
 				evict = heap->next->priv;
 				nouveau_resource_free(&evict->exec);
 			}
@@ -887,12 +887,14 @@ check_gpu_resources:
 		}
 
 		so = so_new(3, 4, 0);
-		so_method(so, curie, NV40TCL_VP_START_FROM_ID, 1);
+		so_method(so, eng3d, NV34TCL_VP_START_FROM_ID, 1);
 		so_data  (so, vp->exec->start);
-		so_method(so, curie, NV40TCL_VP_ATTRIB_EN, 2);
-		so_data  (so, vp->ir);
-		so_data  (so, vp->or);
-		so_method(so, curie,  NV40TCL_CLIP_PLANE_ENABLE, 1);
+		if(nvfx->is_nv4x) {
+			so_method(so, eng3d, NV40TCL_VP_ATTRIB_EN, 2);
+			so_data  (so, vp->ir);
+			so_data  (so, vp->or);
+		}
+		so_method(so, eng3d,  NV34TCL_VP_CLIP_PLANES_ENABLE, 1);
 		so_data  (so, vp->clip_ctrl);
 		so_ref(so, &vp->so);
 		so_ref(NULL, &so);
@@ -902,12 +904,12 @@ check_gpu_resources:
 
 	/* Allocate hw vtxprog const slots */
 	if (vp->nr_consts && !vp->data) {
-		struct nouveau_resource *heap = nv40->screen->vp_data_heap;
+		struct nouveau_resource *heap = nvfx->screen->vp_data_heap;
 
 		if (nouveau_resource_alloc(heap, vp->nr_consts, vp, &vp->data)) {
 			while (heap->next && heap->size < vp->nr_consts) {
-				struct nv40_vertex_program *evict;
-				
+				struct nvfx_vertex_program *evict;
+
 				evict = heap->next->priv;
 				nouveau_resource_free(&evict->data);
 			}
@@ -929,7 +931,7 @@ check_gpu_resources:
 	 */
 	if (vp->exec_start != vp->exec->start) {
 		for (i = 0; i < vp->nr_insns; i++) {
-			struct nv40_vertex_program_exec *vpi = &vp->insns[i];
+			struct nvfx_vertex_program_exec *vpi = &vp->insns[i];
 
 			if (vpi->has_branch_offset) {
 				assert(0);
@@ -941,13 +943,13 @@ check_gpu_resources:
 
 	if (vp->nr_consts && vp->data_start != vp->data->start) {
 		for (i = 0; i < vp->nr_insns; i++) {
-			struct nv40_vertex_program_exec *vpi = &vp->insns[i];
+			struct nvfx_vertex_program_exec *vpi = &vp->insns[i];
 
 			if (vpi->const_index >= 0) {
-				vpi->data[1] &= ~NV40_VP_INST_CONST_SRC_MASK;
+				vpi->data[1] &= ~NVFX_VP(INST_CONST_SRC_MASK);
 				vpi->data[1] |=
 					(vpi->const_index + vp->data->start) <<
-					NV40_VP_INST_CONST_SRC_SHIFT;
+					NVFX_VP(INST_CONST_SRC_SHIFT);
 
 			}
 		}
@@ -965,7 +967,7 @@ check_gpu_resources:
 		}
 
 		for (i = 0; i < vp->nr_consts; i++) {
-			struct nv40_vertex_program_data *vpd = &vp->consts[i];
+			struct nvfx_vertex_program_data *vpd = &vp->consts[i];
 
 			if (vpd->index >= 0) {
 				if (!upload_data &&
@@ -976,13 +978,13 @@ check_gpu_resources:
 				       4 * sizeof(float));
 			}
 
-			BEGIN_RING(chan, curie, NV40TCL_VP_UPLOAD_CONST_ID, 5);
+			BEGIN_RING(chan, eng3d, NV34TCL_VP_UPLOAD_CONST_ID, 5);
 			OUT_RING  (chan, i + vp->data->start);
 			OUT_RINGp (chan, (uint32_t *)vpd->value, 4);
 		}
 
 		if (constbuf)
-			pscreen->buffer_unmap(pscreen, constbuf);
+			pipe_buffer_unmap(pscreen, constbuf);
 	}
 
 	/* Upload vtxprog */
@@ -995,16 +997,16 @@ check_gpu_resources:
 			NOUVEAU_MSG("VP %d: 0x%08x\n", i, vp->insns[i].data[3]);
 		}
 #endif
-		BEGIN_RING(chan, curie, NV40TCL_VP_UPLOAD_FROM_ID, 1);
+		BEGIN_RING(chan, eng3d, NV34TCL_VP_UPLOAD_FROM_ID, 1);
 		OUT_RING  (chan, vp->exec->start);
 		for (i = 0; i < vp->nr_insns; i++) {
-			BEGIN_RING(chan, curie, NV40TCL_VP_UPLOAD_INST(0), 4);
+			BEGIN_RING(chan, eng3d, NV34TCL_VP_UPLOAD_INST(0), 4);
 			OUT_RINGp (chan, vp->insns[i].data, 4);
 		}
 	}
 
-	if (vp->so != nv40->state.hw[NV40_STATE_VERTPROG]) {
-		so_ref(vp->so, &nv40->state.hw[NV40_STATE_VERTPROG]);
+	if (vp->so != nvfx->state.hw[NVFX_STATE_VERTPROG]) {
+		so_ref(vp->so, &nvfx->state.hw[NVFX_STATE_VERTPROG]);
 		return TRUE;
 	}
 
@@ -1012,7 +1014,7 @@ check_gpu_resources:
 }
 
 void
-nv40_vertprog_destroy(struct nv40_context *nv40, struct nv40_vertex_program *vp)
+nvfx_vertprog_destroy(struct nvfx_context *nvfx, struct nvfx_vertex_program *vp)
 {
 	vp->translated = FALSE;
 
@@ -1038,11 +1040,10 @@ nv40_vertprog_destroy(struct nv40_context *nv40, struct nv40_vertex_program *vp)
 	so_ref(NULL, &vp->so);
 }
 
-struct nv40_state_entry nv40_state_vertprog = {
-	.validate = nv40_vertprog_validate,
+struct nvfx_state_entry nvfx_state_vertprog = {
+	.validate = nvfx_vertprog_validate,
 	.dirty = {
-		.pipe = NV40_NEW_VERTPROG | NV40_NEW_UCP,
-		.hw = NV40_STATE_VERTPROG,
+		.pipe = NVFX_NEW_VERTPROG | NVFX_NEW_UCP,
+		.hw = NVFX_STATE_VERTPROG,
 	}
 };
-
diff --git a/src/gallium/drivers/r300/Makefile b/src/gallium/drivers/r300/Makefile
index 61b54af4dd..a6529b2060 100644
--- a/src/gallium/drivers/r300/Makefile
+++ b/src/gallium/drivers/r300/Makefile
@@ -14,6 +14,7 @@ C_SOURCES = \
 	r300_query.c \
 	r300_render.c \
 	r300_screen.c \
+	r300_screen_buffer.c \
 	r300_state.c \
 	r300_state_derived.c \
 	r300_state_invariant.c \
diff --git a/src/gallium/drivers/r300/r300_context.c b/src/gallium/drivers/r300/r300_context.c
index 8606c0004e..d994a46ccf 100644
--- a/src/gallium/drivers/r300/r300_context.c
+++ b/src/gallium/drivers/r300/r300_context.c
@@ -24,6 +24,7 @@
 
 #include "util/u_memory.h"
 #include "util/u_simple_list.h"
+#include "util/u_upload_mgr.h"
 
 #include "r300_blit.h"
 #include "r300_context.h"
@@ -55,6 +56,9 @@ static void r300_destroy_context(struct pipe_context* context)
         FREE(query);
     }
 
+    u_upload_destroy(r300->upload_vb);
+    u_upload_destroy(r300->upload_ib);
+
     FREE(r300->blend_color_state.state);
     FREE(r300->clip_state.state);
     FREE(r300->fb_state.state);
@@ -72,8 +76,7 @@ r300_is_texture_referenced(struct pipe_context *pipe,
                            struct pipe_texture *texture,
                            unsigned face, unsigned level)
 {
-    return pipe->is_buffer_referenced(pipe,
-                                      ((struct r300_texture *)texture)->buffer);
+    return 0;
 }
 
 static unsigned int
@@ -157,16 +160,14 @@ struct pipe_context* r300_create_context(struct pipe_screen* screen,
 {
     struct r300_context* r300 = CALLOC_STRUCT(r300_context);
     struct r300_screen* r300screen = r300_screen(screen);
-    struct radeon_winsys* radeon_winsys = r300screen->radeon_winsys;
+    struct r300_winsys_screen *rws = r300screen->rws;
 
     if (!r300)
         return NULL;
 
-    r300screen->ctx = (struct pipe_context*)r300;
-
-    r300->winsys = radeon_winsys;
+    r300->rws = rws;
 
-    r300->context.winsys = (struct pipe_winsys*)radeon_winsys;
+    r300->context.winsys = (struct pipe_winsys*)rws;
     r300->context.screen = screen;
     r300->context.priv = priv;
 
@@ -212,16 +213,33 @@ struct pipe_context* r300_create_context(struct pipe_screen* screen,
 
     r300_init_transfer_functions(r300);
 
-    /* r300_init_surface_functions(r300); */
-
     r300_init_state_functions(r300);
 
     r300->invariant_state.dirty = TRUE;
 
-    r300->winsys->set_flush_cb(r300->winsys, r300_flush_cb, r300);
+    rws->set_flush_cb(r300->rws, r300_flush_cb, r300);
     r300->dirty_hw++;
 
     r300->blitter = util_blitter_create(&r300->context);
 
+    r300->upload_ib = u_upload_create(screen,
+				      32 * 1024, 16,
+				      PIPE_BUFFER_USAGE_INDEX);
+
+    if (r300->upload_ib == NULL)
+        goto no_upload_ib;
+
+    r300->upload_vb = u_upload_create(screen,
+				      128 * 1024, 16,
+				      PIPE_BUFFER_USAGE_VERTEX);
+    if (r300->upload_vb == NULL)
+        goto no_upload_vb;
+
     return &r300->context;
+
+ no_upload_ib:
+    u_upload_destroy(r300->upload_ib);
+ no_upload_vb:
+    FREE(r300);
+    return NULL;
 }
diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h
index 7b86669888..2763103fda 100644
--- a/src/gallium/drivers/r300/r300_context.h
+++ b/src/gallium/drivers/r300/r300_context.h
@@ -32,6 +32,7 @@
 
 #include "r300_screen.h"
 
+struct u_upload_mgr;
 struct r300_context;
 
 struct r300_fragment_shader;
@@ -268,7 +269,7 @@ struct r300_texture {
     boolean is_npot;
 
     /* Pipe buffer backing this texture. */
-    struct pipe_buffer* buffer;
+    struct r300_winsys_buffer *buffer;
 
     /* Registers carrying texture format data. */
     struct r300_texture_format_state state;
@@ -302,7 +303,7 @@ struct r300_context {
     struct pipe_context context;
 
     /* The interface to the windowing system, etc. */
-    struct radeon_winsys* winsys;
+    struct r300_winsys_screen *rws;
     /* Draw module. Used mostly for SW TCL. */
     struct draw_context* draw;
     /* Accelerated blit support. */
@@ -369,6 +370,7 @@ struct r300_context {
     int vertex_buffer_max_index;
     /* Vertex elements for Gallium. */
     struct r300_vertex_element_state *velems;
+    bool any_user_vbs;
 
     /* Vertex info for Draw. */
     struct vertex_info vertex_info;
@@ -389,6 +391,9 @@ struct r300_context {
     uint32_t zbuffer_bpp;
     /* Whether scissor is enabled. */
     boolean scissor_enabled;
+    /* upload managers */
+    struct u_upload_mgr *upload_vb;
+    struct u_upload_mgr *upload_ib;
 };
 
 /* Convenience cast wrapper. */
diff --git a/src/gallium/drivers/r300/r300_cs.h b/src/gallium/drivers/r300/r300_cs.h
index 151f72b0fe..ad07efbffd 100644
--- a/src/gallium/drivers/r300/r300_cs.h
+++ b/src/gallium/drivers/r300/r300_cs.h
@@ -51,7 +51,7 @@
 
 #define CS_LOCALS(context) \
     struct r300_context* const cs_context_copy = (context); \
-    struct radeon_winsys* cs_winsys = cs_context_copy->winsys; \
+    struct r300_winsys_screen *cs_winsys = cs_context_copy->rws; \
     int cs_count = 0; (void) cs_count;
 
 #define CHECK_CS(size) \
@@ -105,22 +105,34 @@
     cs_count--; \
 } while (0)
 
-#define OUT_CS_RELOC(bo, offset, rd, wd, flags) do { \
+#define OUT_CS_BUF_RELOC(bo, offset, rd, wd, flags) do { \
     DBG(cs_context_copy, DBG_CS, "r300: writing relocation for buffer %p, offset %d, " \
             "domains (%d, %d, %d)\n", \
         bo, offset, rd, wd, flags); \
     assert(bo); \
     cs_winsys->write_cs_dword(cs_winsys, offset); \
-    cs_winsys->write_cs_reloc(cs_winsys, bo, rd, wd, flags); \
+    r300_buffer_write_reloc(cs_winsys, r300_buffer(bo), rd, wd, flags);	\
     cs_count -= 3; \
 } while (0)
 
-#define OUT_CS_RELOC_NO_OFFSET(bo, rd, wd, flags) do { \
+
+#define OUT_CS_TEX_RELOC(tex, offset, rd, wd, flags) do { \
+    DBG(cs_context_copy, DBG_CS, "r300: writing relocation for texture %p, offset %d, " \
+            "domains (%d, %d, %d)\n", \
+        tex, offset, rd, wd, flags); \
+    assert(tex); \
+    cs_winsys->write_cs_dword(cs_winsys, offset); \
+    r300_texture_write_reloc(cs_winsys, tex, rd, wd, flags);	\
+    cs_count -= 3; \
+} while (0)
+
+
+#define OUT_CS_BUF_RELOC_NO_OFFSET(bo, rd, wd, flags) do { \
     DBG(cs_context_copy, DBG_CS, "r300: writing relocation for buffer %p, " \
             "domains (%d, %d, %d)\n", \
         bo, rd, wd, flags); \
     assert(bo); \
-    cs_winsys->write_cs_reloc(cs_winsys, bo, rd, wd, flags); \
+    r300_buffer_write_reloc(cs_winsys, r300_buffer(bo), rd, wd, flags);	\
     cs_count -= 2; \
 } while (0)
 
diff --git a/src/gallium/drivers/r300/r300_debug.c b/src/gallium/drivers/r300/r300_debug.c
index b881730848..d6177577c8 100644
--- a/src/gallium/drivers/r300/r300_debug.c
+++ b/src/gallium/drivers/r300/r300_debug.c
@@ -37,6 +37,7 @@ static struct debug_option debug_options[] = {
     { "draw", DBG_DRAW, "Draw and emit" },
     { "tex", DBG_TEX, "Textures" },
     { "fall", DBG_FALL, "Fallbacks" },
+    { "anisohq", DBG_ANISOHQ, "High quality anisotropic filtering (for benchmarking purposes only!)" },
 
     { "all", ~0, "Convenience option that enables all debug flags" },
 
diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index fc8a8a2773..0a3f0f45c0 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -32,6 +32,8 @@
 #include "r300_emit.h"
 #include "r300_fs.h"
 #include "r300_screen.h"
+#include "r300_screen_buffer.h"
+#include "r300_state_inlines.h"
 #include "r300_vs.h"
 
 void r300_emit_blend_state(struct r300_context* r300,
@@ -415,10 +417,10 @@ void r300_emit_fb_state(struct r300_context* r300, unsigned size, void* state)
         assert(tex && tex->buffer && "cbuf is marked, but NULL!");
 
         OUT_CS_REG_SEQ(R300_RB3D_COLOROFFSET0 + (4 * i), 1);
-        OUT_CS_RELOC(tex->buffer, surf->offset, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+        OUT_CS_TEX_RELOC(tex, surf->offset, 0, RADEON_GEM_DOMAIN_VRAM, 0);
 
         OUT_CS_REG_SEQ(R300_RB3D_COLORPITCH0 + (4 * i), 1);
-        OUT_CS_RELOC(tex->buffer, tex->fb_state.colorpitch[surf->level],
+        OUT_CS_TEX_RELOC(tex, tex->fb_state.colorpitch[surf->level],
                      0, RADEON_GEM_DOMAIN_VRAM, 0);
 
         OUT_CS_REG(R300_US_OUT_FMT_0 + (4 * i), tex->fb_state.us_out_fmt);
@@ -434,12 +436,12 @@ void r300_emit_fb_state(struct r300_context* r300, unsigned size, void* state)
         assert(tex && tex->buffer && "zsbuf is marked, but NULL!");
 
         OUT_CS_REG_SEQ(R300_ZB_DEPTHOFFSET, 1);
-        OUT_CS_RELOC(tex->buffer, surf->offset, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+        OUT_CS_TEX_RELOC(tex, surf->offset, 0, RADEON_GEM_DOMAIN_VRAM, 0);
 
         OUT_CS_REG(R300_ZB_FORMAT, tex->fb_state.zb_format);
 
         OUT_CS_REG_SEQ(R300_ZB_DEPTHPITCH, 1);
-        OUT_CS_RELOC(tex->buffer, tex->fb_state.depthpitch[surf->level],
+        OUT_CS_TEX_RELOC(tex, tex->fb_state.depthpitch[surf->level],
                      0, RADEON_GEM_DOMAIN_VRAM, 0);
     }
 
@@ -448,7 +450,7 @@ void r300_emit_fb_state(struct r300_context* r300, unsigned size, void* state)
     END_CS;
 }
 
-static void r300_emit_query_start(struct r300_context *r300)
+void r300_emit_query_start(struct r300_context *r300)
 {
     struct r300_capabilities *caps = r300_screen(r300->context.screen)->caps;
     struct r300_query *query = r300->query_current;
@@ -491,13 +493,13 @@ static void r300_emit_query_finish(struct r300_context *r300,
             /* pipe 3 only */
             OUT_CS_REG(R300_SU_REG_DEST, 1 << 3);
             OUT_CS_REG_SEQ(R300_ZB_ZPASS_ADDR, 1);
-            OUT_CS_RELOC(r300->oqbo, query->offset + (sizeof(uint32_t) * 3),
+            OUT_CS_BUF_RELOC(r300->oqbo, query->offset + (sizeof(uint32_t) * 3),
                     0, RADEON_GEM_DOMAIN_GTT, 0);
         case 3:
             /* pipe 2 only */
             OUT_CS_REG(R300_SU_REG_DEST, 1 << 2);
             OUT_CS_REG_SEQ(R300_ZB_ZPASS_ADDR, 1);
-            OUT_CS_RELOC(r300->oqbo, query->offset + (sizeof(uint32_t) * 2),
+            OUT_CS_BUF_RELOC(r300->oqbo, query->offset + (sizeof(uint32_t) * 2),
                     0, RADEON_GEM_DOMAIN_GTT, 0);
         case 2:
             /* pipe 1 only */
@@ -505,13 +507,13 @@ static void r300_emit_query_finish(struct r300_context *r300,
             OUT_CS_REG(R300_SU_REG_DEST,
                     1 << (caps->high_second_pipe ? 3 : 1));
             OUT_CS_REG_SEQ(R300_ZB_ZPASS_ADDR, 1);
-            OUT_CS_RELOC(r300->oqbo, query->offset + (sizeof(uint32_t) * 1),
+            OUT_CS_BUF_RELOC(r300->oqbo, query->offset + (sizeof(uint32_t) * 1),
                     0, RADEON_GEM_DOMAIN_GTT, 0);
         case 1:
             /* pipe 0 only */
             OUT_CS_REG(R300_SU_REG_DEST, 1 << 0);
             OUT_CS_REG_SEQ(R300_ZB_ZPASS_ADDR, 1);
-            OUT_CS_RELOC(r300->oqbo, query->offset + (sizeof(uint32_t) * 0),
+            OUT_CS_BUF_RELOC(r300->oqbo, query->offset + (sizeof(uint32_t) * 0),
                     0, RADEON_GEM_DOMAIN_GTT, 0);
             break;
         default:
@@ -533,7 +535,7 @@ static void rv530_emit_query_single(struct r300_context *r300,
     BEGIN_CS(8);
     OUT_CS_REG(RV530_FG_ZBREG_DEST, RV530_FG_ZBREG_DEST_PIPE_SELECT_0);
     OUT_CS_REG_SEQ(R300_ZB_ZPASS_ADDR, 1);
-    OUT_CS_RELOC(r300->oqbo, query->offset, 0, RADEON_GEM_DOMAIN_GTT, 0);
+    OUT_CS_BUF_RELOC(r300->oqbo, query->offset, 0, RADEON_GEM_DOMAIN_GTT, 0);
     OUT_CS_REG(RV530_FG_ZBREG_DEST, RV530_FG_ZBREG_DEST_PIPE_SELECT_ALL);
     END_CS;
 }
@@ -546,10 +548,10 @@ static void rv530_emit_query_double(struct r300_context *r300,
     BEGIN_CS(14);
     OUT_CS_REG(RV530_FG_ZBREG_DEST, RV530_FG_ZBREG_DEST_PIPE_SELECT_0);
     OUT_CS_REG_SEQ(R300_ZB_ZPASS_ADDR, 1);
-    OUT_CS_RELOC(r300->oqbo, query->offset, 0, RADEON_GEM_DOMAIN_GTT, 0);
+    OUT_CS_BUF_RELOC(r300->oqbo, query->offset, 0, RADEON_GEM_DOMAIN_GTT, 0);
     OUT_CS_REG(RV530_FG_ZBREG_DEST, RV530_FG_ZBREG_DEST_PIPE_SELECT_1);
     OUT_CS_REG_SEQ(R300_ZB_ZPASS_ADDR, 1);
-    OUT_CS_RELOC(r300->oqbo, query->offset + sizeof(uint32_t), 0, RADEON_GEM_DOMAIN_GTT, 0);
+    OUT_CS_BUF_RELOC(r300->oqbo, query->offset + sizeof(uint32_t), 0, RADEON_GEM_DOMAIN_GTT, 0);
     OUT_CS_REG(RV530_FG_ZBREG_DEST, RV530_FG_ZBREG_DEST_PIPE_SELECT_ALL);
     END_CS;
 }
@@ -747,7 +749,7 @@ void r300_emit_textures_state(struct r300_context *r300,
             OUT_CS_REG(R300_TX_FORMAT2_0 + (i * 4), texstate->format[2]);
 
             OUT_CS_REG_SEQ(R300_TX_OFFSET_0 + (i * 4), 1);
-            OUT_CS_RELOC(allstate->textures[i]->buffer, texstate->tile_config,
+            OUT_CS_TEX_RELOC(allstate->textures[i], texstate->tile_config,
                          RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0, 0);
         }
     }
@@ -788,8 +790,8 @@ void r300_emit_aos(struct r300_context* r300, unsigned offset)
     }
 
     for (i = 0; i < aos_count; i++) {
-        OUT_CS_RELOC_NO_OFFSET(vbuf[velem[i].vertex_buffer_index].buffer,
-                               RADEON_GEM_DOMAIN_GTT, 0, 0);
+        OUT_CS_BUF_RELOC_NO_OFFSET(vbuf[velem[i].vertex_buffer_index].buffer,
+				   RADEON_GEM_DOMAIN_GTT, 0, 0);
     }
     END_CS;
 }
@@ -814,7 +816,7 @@ void r300_emit_vertex_buffer(struct r300_context* r300)
     OUT_CS(r300->vertex_info.size |
             (r300->vertex_info.size << 8));
     OUT_CS(r300->vbo_offset);
-    OUT_CS_RELOC(r300->vbo, 0, RADEON_GEM_DOMAIN_GTT, 0, 0);
+    OUT_CS_BUF_RELOC(r300->vbo, 0, RADEON_GEM_DOMAIN_GTT, 0, 0);
     END_CS;
 }
 
@@ -1009,16 +1011,22 @@ void r300_emit_buffer_validate(struct r300_context *r300,
     unsigned i;
     boolean invalid = FALSE;
 
+    /* upload buffers first */
+    if (r300->any_user_vbs) {
+        r300_upload_user_buffers(r300);
+        r300->any_user_vbs = false;
+    }
+
     /* Clean out BOs. */
-    r300->winsys->reset_bos(r300->winsys);
+    r300->rws->reset_bos(r300->rws);
 
 validate:
     /* Color buffers... */
     for (i = 0; i < fb->nr_cbufs; i++) {
         tex = (struct r300_texture*)fb->cbufs[i]->texture;
         assert(tex && tex->buffer && "cbuf is marked, but NULL!");
-        if (!r300->winsys->add_buffer(r300->winsys, tex->buffer,
-                    0, RADEON_GEM_DOMAIN_VRAM)) {
+        if (!r300_add_texture(r300->rws, tex,
+			      0, RADEON_GEM_DOMAIN_VRAM)) {
             r300->context.flush(&r300->context, 0, NULL);
             goto validate;
         }
@@ -1027,8 +1035,8 @@ validate:
     if (fb->zsbuf) {
         tex = (struct r300_texture*)fb->zsbuf->texture;
         assert(tex && tex->buffer && "zsbuf is marked, but NULL!");
-        if (!r300->winsys->add_buffer(r300->winsys, tex->buffer,
-                    0, RADEON_GEM_DOMAIN_VRAM)) {
+        if (!r300_add_texture(r300->rws, tex,
+			      0, RADEON_GEM_DOMAIN_VRAM)) {
             r300->context.flush(&r300->context, 0, NULL);
             goto validate;
         }
@@ -1038,24 +1046,24 @@ validate:
         if (!r300->fragment_sampler_views[i])
             continue;
         tex = (struct r300_texture *)r300->fragment_sampler_views[i]->texture;
-        if (!r300->winsys->add_buffer(r300->winsys, tex->buffer,
-                    RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0)) {
+        if (!r300_add_texture(r300->rws, tex,
+			      RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0)) {
             r300->context.flush(&r300->context, 0, NULL);
             goto validate;
         }
     }
     /* ...occlusion query buffer... */
     if (r300->dirty_state & R300_NEW_QUERY) {
-        if (!r300->winsys->add_buffer(r300->winsys, r300->oqbo,
-                    0, RADEON_GEM_DOMAIN_GTT)) {
+        if (!r300_add_buffer(r300->rws, r300->oqbo,
+			     0, RADEON_GEM_DOMAIN_GTT)) {
             r300->context.flush(&r300->context, 0, NULL);
             goto validate;
         }
     }
     /* ...vertex buffer for SWTCL path... */
     if (r300->vbo) {
-        if (!r300->winsys->add_buffer(r300->winsys, r300->vbo,
-                    RADEON_GEM_DOMAIN_GTT, 0)) {
+        if (!r300_add_buffer(r300->rws, r300->vbo,
+			     RADEON_GEM_DOMAIN_GTT, 0)) {
             r300->context.flush(&r300->context, 0, NULL);
             goto validate;
         }
@@ -1065,23 +1073,22 @@ validate:
         for (i = 0; i < r300->velems->count; i++) {
             pbuf = vbuf[velem[i].vertex_buffer_index].buffer;
 
-            if (!r300->winsys->add_buffer(r300->winsys, pbuf,
-                                          RADEON_GEM_DOMAIN_GTT, 0)) {
-                r300->context.flush(&r300->context, 0, NULL);
+            if (!r300_add_buffer(r300->rws, pbuf,
+				 RADEON_GEM_DOMAIN_GTT, 0)) {
+		r300->context.flush(&r300->context, 0, NULL);
                 goto validate;
             }
         }
     }
     /* ...and index buffer for HWTCL path. */
     if (index_buffer) {
-        if (!r300->winsys->add_buffer(r300->winsys, index_buffer,
-                                      RADEON_GEM_DOMAIN_GTT, 0)) {
+        if (!r300_add_buffer(r300->rws, index_buffer,
+			     RADEON_GEM_DOMAIN_GTT, 0)) {
             r300->context.flush(&r300->context, 0, NULL);
             goto validate;
         }
     }
-
-    if (!r300->winsys->validate(r300->winsys)) {
+    if (!r300->rws->validate(r300->rws)) {
         r300->context.flush(&r300->context, 0, NULL);
         if (invalid) {
             /* Well, hell. */
diff --git a/src/gallium/drivers/r300/r300_emit.h b/src/gallium/drivers/r300/r300_emit.h
index 449e640a88..7db2fc6a1a 100644
--- a/src/gallium/drivers/r300/r300_emit.h
+++ b/src/gallium/drivers/r300/r300_emit.h
@@ -57,8 +57,7 @@ void r500_emit_fs_constant_buffer(struct r300_context* r300,
 
 void r300_emit_fb_state(struct r300_context* r300, unsigned size, void* state);
 
-void r300_emit_query_begin(struct r300_context* r300,
-                           struct r300_query* query);
+void r300_emit_query_start(struct r300_context* r300);
 
 void r300_emit_query_end(struct r300_context* r300);
 
diff --git a/src/gallium/drivers/r300/r300_reg.h b/src/gallium/drivers/r300/r300_reg.h
index c67cc86871..1c2b252887 100644
--- a/src/gallium/drivers/r300/r300_reg.h
+++ b/src/gallium/drivers/r300/r300_reg.h
@@ -1500,6 +1500,10 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #	define R300_ANISO_THRESHOLD_MASK       (7<<17)
 
 #	define R500_MACRO_SWITCH               (1<<22)
+#       define R500_TX_MAX_ANISO(x)            ((x) << 23)
+#       define R500_TX_MAX_ANISO_MASK          (63 << 23)
+#       define R500_TX_ANISO_HIGH_QUALITY      (1 << 30)
+
 #	define R500_BORDER_FIX                 (1<<31)
 
 #define R300_TX_FORMAT0_0                   0x4480
diff --git a/src/gallium/drivers/r300/r300_render.c b/src/gallium/drivers/r300/r300_render.c
index ccf39876a5..47100c83b0 100644
--- a/src/gallium/drivers/r300/r300_render.c
+++ b/src/gallium/drivers/r300/r300_render.c
@@ -30,10 +30,12 @@
 
 #include "util/u_format.h"
 #include "util/u_memory.h"
+#include "util/u_upload_mgr.h"
 #include "util/u_prim.h"
 
 #include "r300_cs.h"
 #include "r300_context.h"
+#include "r300_screen_buffer.h"
 #include "r300_emit.h"
 #include "r300_reg.h"
 #include "r300_render.h"
@@ -123,7 +125,7 @@ static uint32_t r300_provoking_vertex_fixes(struct r300_context *r300,
 static boolean r300_reserve_cs_space(struct r300_context *r300,
                                      unsigned dwords)
 {
-    if (!r300->winsys->check_cs(r300->winsys, dwords)) {
+    if (!r300->rws->check_cs(r300->rws, dwords)) {
         r300->context.flush(&r300->context, 0, NULL);
         return TRUE;
     }
@@ -131,9 +133,37 @@ static boolean r300_reserve_cs_space(struct r300_context *r300,
 }
 
 static boolean immd_is_good_idea(struct r300_context *r300,
-                                      unsigned count)
+                                 unsigned count)
 {
-    return count <= 4;
+    struct pipe_vertex_element* velem;
+    struct pipe_vertex_buffer* vbuf;
+    boolean checked[PIPE_MAX_ATTRIBS] = {0};
+    unsigned vertex_element_count = r300->velems->count;
+    unsigned i, vbi;
+
+    if (count > 4) {
+        return FALSE;
+    }
+
+    /* We shouldn't map buffers referenced by CS, busy buffers,
+     * and ones placed in VRAM. */
+    /* XXX Check for VRAM buffers. */
+    for (i = 0; i < vertex_element_count; i++) {
+        velem = &r300->velems->velem[i];
+        vbi = velem->vertex_buffer_index;
+
+        if (!checked[vbi]) {
+            vbuf = &r300->vertex_buffer[vbi];
+
+            if (r300_buffer_is_referenced(r300,
+					  vbuf->buffer)) {
+                /* It's a very bad idea to map it... */
+                return FALSE;
+            }
+            checked[vbi] = TRUE;
+        }
+    }
+    return TRUE;
 }
 
 static void r300_emit_draw_arrays_immediate(struct r300_context *r300,
@@ -316,8 +346,8 @@ static void r300_emit_draw_elements(struct r300_context *r300,
     OUT_CS(R300_INDX_BUFFER_ONE_REG_WR | (R300_VAP_PORT_IDX0 >> 2) |
            (0 << R300_INDX_BUFFER_SKIP_SHIFT));
     OUT_CS(offset_dwords << 2);
-    OUT_CS_RELOC(indexBuffer, count_dwords,
-        RADEON_GEM_DOMAIN_GTT, 0, 0);
+    OUT_CS_BUF_RELOC(indexBuffer, count_dwords,
+		     RADEON_GEM_DOMAIN_GTT, 0, 0);
 
     END_CS;
 }
@@ -384,12 +414,16 @@ void r300_draw_range_elements(struct pipe_context* pipe,
 
     r300_update_derived_state(r300);
 
+    r300_upload_index_buffer(r300, &indexBuffer, indexSize, start, count);
+
     /* 128 dwords for emit_aos and emit_draw_elements */
     r300_reserve_cs_space(r300, r300_get_num_dirty_dwords(r300) + 128);
     r300_emit_buffer_validate(r300, TRUE, indexBuffer);
     r300_emit_dirty_state(r300);
     r300_emit_aos(r300, 0);
 
+    u_upload_flush(r300->upload_vb);
+    u_upload_flush(r300->upload_ib);
     if (alt_num_verts || count <= 65535) {
         r300_emit_draw_elements(r300, indexBuffer, indexSize, minIndex,
                                 maxIndex, mode, start, count);
@@ -412,7 +446,7 @@ void r300_draw_range_elements(struct pipe_context* pipe,
     }
 
     if (indexBuffer != orgIndexBuffer) {
-        pipe->screen->buffer_destroy(indexBuffer);
+        pipe_buffer_reference( &indexBuffer, NULL );
     }
 }
 
@@ -476,6 +510,7 @@ void r300_draw_arrays(struct pipe_context* pipe, unsigned mode,
                 }
             } while (count);
         }
+	u_upload_flush(r300->upload_vb);
     }
 }
 
diff --git a/src/gallium/drivers/r300/r300_screen.c b/src/gallium/drivers/r300/r300_screen.c
index 64d1909a38..3e31688f8e 100644
--- a/src/gallium/drivers/r300/r300_screen.c
+++ b/src/gallium/drivers/r300/r300_screen.c
@@ -23,13 +23,13 @@
 
 #include "util/u_format.h"
 #include "util/u_memory.h"
-#include "util/u_simple_screen.h"
 
 #include "r300_context.h"
 #include "r300_texture.h"
 
 #include "radeon_winsys.h"
-#include "r300_winsys.h"
+
+#include "r300_screen_buffer.h"
 
 /* Return the identifier behind whom the brave coders responsible for this
  * amalgamation of code, sweat, and duct tape, routinely obscure their names.
@@ -254,15 +254,19 @@ static boolean r300_is_format_supported(struct pipe_screen* screen,
 static void r300_destroy_screen(struct pipe_screen* pscreen)
 {
     struct r300_screen* r300screen = r300_screen(pscreen);
+    struct r300_winsys_screen *rws = r300_winsys_screen(pscreen);
+
+    if (rws)
+      rws->destroy(rws);
 
     FREE(r300screen->caps);
     FREE(r300screen);
 }
 
-struct pipe_screen* r300_create_screen(struct radeon_winsys* radeon_winsys)
+struct pipe_screen* r300_create_screen(struct r300_winsys_screen *rws)
 {
-    struct r300_screen* r300screen = CALLOC_STRUCT(r300_screen);
-    struct r300_capabilities* caps = CALLOC_STRUCT(r300_capabilities);
+    struct r300_screen *r300screen = CALLOC_STRUCT(r300_screen);
+    struct r300_capabilities *caps = CALLOC_STRUCT(r300_capabilities);
 
     if (!r300screen || !caps) {
         FREE(r300screen);
@@ -270,16 +274,16 @@ struct pipe_screen* r300_create_screen(struct radeon_winsys* radeon_winsys)
         return NULL;
     }
 
-    caps->pci_id = radeon_winsys->pci_id;
-    caps->num_frag_pipes = radeon_winsys->gb_pipes;
-    caps->num_z_pipes = radeon_winsys->z_pipes;
+    caps->pci_id = rws->get_value(rws, R300_VID_PCI_ID);
+    caps->num_frag_pipes = rws->get_value(rws, R300_VID_GB_PIPES);
+    caps->num_z_pipes = rws->get_value(rws, R300_VID_Z_PIPES);
 
     r300_init_debug(r300screen);
     r300_parse_chipset(caps);
 
     r300screen->caps = caps;
-    r300screen->radeon_winsys = radeon_winsys;
-    r300screen->screen.winsys = (struct pipe_winsys*)radeon_winsys;
+    r300screen->rws = rws;
+    r300screen->screen.winsys = (struct pipe_winsys*)rws;
     r300screen->screen.destroy = r300_destroy_screen;
     r300screen->screen.get_name = r300_get_name;
     r300screen->screen.get_vendor = r300_get_vendor;
@@ -289,8 +293,13 @@ struct pipe_screen* r300_create_screen(struct radeon_winsys* radeon_winsys)
     r300screen->screen.context_create = r300_create_context;
 
     r300_init_screen_texture_functions(&r300screen->screen);
-    u_simple_screen_init(&r300screen->screen);
 
+    r300_screen_init_buffer_functions(r300screen);
     return &r300screen->screen;
 }
 
+struct r300_winsys_screen *
+r300_winsys_screen(struct pipe_screen *screen)
+{
+    return r300_screen(screen)->rws;
+}
diff --git a/src/gallium/drivers/r300/r300_screen.h b/src/gallium/drivers/r300/r300_screen.h
index 484bde6a6b..1ccc0bfb7a 100644
--- a/src/gallium/drivers/r300/r300_screen.h
+++ b/src/gallium/drivers/r300/r300_screen.h
@@ -36,11 +36,7 @@ struct r300_screen {
     /* Parent class */
     struct pipe_screen screen;
 
-    struct radeon_winsys* radeon_winsys;
-
-    /* XXX This hack will be removed once texture transfers become part of
-     * pipe_context. */
-    struct pipe_context* ctx;
+    struct r300_winsys_screen *rws;
 
     /* Chipset capabilities */
     struct r300_capabilities* caps;
@@ -55,9 +51,6 @@ static INLINE struct r300_screen* r300_screen(struct pipe_screen* screen) {
     return (struct r300_screen*)screen;
 }
 
-/* Creates a new r300 screen. */
-struct pipe_screen* r300_create_screen(struct radeon_winsys* radeon_winsys);
-
 /* Debug functionality. */
 
 /**
@@ -77,6 +70,7 @@ struct pipe_screen* r300_create_screen(struct radeon_winsys* radeon_winsys);
 #define DBG_DRAW    0x0000010
 #define DBG_TEX     0x0000020
 #define DBG_FALL    0x0000040
+#define DBG_ANISOHQ 0x0000080
 /*@}*/
 
 static INLINE boolean SCREEN_DBG_ON(struct r300_screen * screen, unsigned flags)
diff --git a/src/gallium/drivers/r300/r300_screen_buffer.c b/src/gallium/drivers/r300/r300_screen_buffer.c
new file mode 100644
index 0000000000..b97d0d76a4
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_screen_buffer.c
@@ -0,0 +1,314 @@
+/*
+ * Copyright 2010 Red Hat 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
+ * on 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
+ * THE AUTHOR(S) AND/OR THEIR 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.
+ *
+ * Authors: Dave Airlie
+ */
+#include <stdio.h>
+
+#include "util/u_inlines.h"
+#include "util/u_format.h"
+#include "util/u_memory.h"
+#include "util/u_upload_mgr.h"
+#include "util/u_math.h"
+
+#include "r300_screen_buffer.h"
+
+#include "r300_winsys.h"
+
+boolean r300_buffer_is_referenced(struct r300_context *r300,
+				  struct pipe_buffer *buf)
+{
+    struct r300_buffer *rbuf = r300_buffer(buf);
+    if (r300_buffer_is_user_buffer(buf))
+	return FALSE;
+
+    return r300->rws->is_buffer_referenced(r300->rws, rbuf->buf);
+    
+}
+int r300_upload_index_buffer(struct r300_context *r300,
+			     struct pipe_buffer **index_buffer,
+			     unsigned index_size,
+			     unsigned start,
+			     unsigned count)
+{
+   struct pipe_buffer *upload_buffer = NULL;
+   unsigned index_offset = start * index_size;
+   int ret = 0;
+
+    if (r300_buffer_is_user_buffer(*index_buffer)) {
+	ret = u_upload_buffer(r300->upload_ib,
+			      index_offset,
+			      count * index_size,
+			      *index_buffer,
+			      &index_offset,
+			      &upload_buffer);
+	if (ret) {
+	    goto done;
+	}
+	*index_buffer = upload_buffer;
+    }
+ done:
+    //    if (upload_buffer)
+    //	pipe_buffer_reference(&upload_buffer, NULL);
+    return ret;
+}
+
+int r300_upload_user_buffers(struct r300_context *r300)
+{
+    enum pipe_error ret = PIPE_OK;
+    int i, nr;
+
+    nr = r300->vertex_buffer_count;
+
+    for (i = 0; i < nr; i++) {
+
+	if (r300_buffer_is_user_buffer(r300->vertex_buffer[i].buffer)) {
+	    struct pipe_buffer *upload_buffer = NULL;
+	    unsigned offset = 0; /*r300->vertex_buffer[i].buffer_offset * 4;*/
+	    unsigned size = r300->vertex_buffer[i].buffer->size;
+	    unsigned upload_offset;
+	    ret = u_upload_buffer(r300->upload_vb,
+				  offset, size,
+				  r300->vertex_buffer[i].buffer,
+				  &upload_offset, &upload_buffer);
+	    if (ret)
+		return ret;
+
+	    pipe_buffer_reference(&r300->vertex_buffer[i].buffer, NULL);
+	    r300->vertex_buffer[i].buffer = upload_buffer;
+	    r300->vertex_buffer[i].buffer_offset = upload_offset;
+	}
+    }
+    return ret;
+}
+
+static struct r300_winsys_buffer *
+r300_winsys_buffer_create(struct r300_screen *r300screen,
+			  unsigned alignment,
+			  unsigned usage,
+			  unsigned size)
+{
+    struct r300_winsys_screen *rws = r300screen->rws;
+    struct r300_winsys_buffer *buf;
+
+    buf = rws->buffer_create(rws, alignment, usage, size);
+    return buf;
+}
+
+static void r300_winsys_buffer_destroy(struct r300_screen *r300screen,
+				       struct r300_buffer *rbuf)
+{
+    struct r300_winsys_screen *rws = r300screen->rws;
+
+    if (rbuf->buf) {
+	rws->buffer_reference(rws, &rbuf->buf, NULL);
+	rbuf->buf = NULL;
+    }
+}
+
+static struct pipe_buffer *r300_buffer_create(struct pipe_screen *screen,
+					      unsigned alignment,
+					      unsigned usage,
+					      unsigned size)
+{
+    struct r300_screen *r300screen = r300_screen(screen);
+    struct r300_buffer *rbuf;
+
+    rbuf = CALLOC_STRUCT(r300_buffer);
+    if (!rbuf)
+	goto error1;
+
+    rbuf->magic = R300_BUFFER_MAGIC;
+
+    pipe_reference_init(&rbuf->base.reference, 1);
+    rbuf->base.screen = screen;
+    rbuf->base.alignment = alignment;
+    rbuf->base.usage = usage;
+    rbuf->base.size = size;
+
+    rbuf->buf = r300_winsys_buffer_create(r300screen,
+					  alignment,
+					  usage,
+					  size);
+
+    if (!rbuf->buf)
+	goto error2;
+
+    return &rbuf->base;
+error2:
+    FREE(rbuf);
+error1:
+    return NULL;
+}
+
+
+static struct pipe_buffer *r300_user_buffer_create(struct pipe_screen *screen,
+						   void *ptr,
+						   unsigned bytes)
+{
+    struct r300_buffer *rbuf;
+
+    rbuf = CALLOC_STRUCT(r300_buffer);
+    if (!rbuf)
+	goto no_rbuf;
+
+    rbuf->magic = R300_BUFFER_MAGIC;
+
+    pipe_reference_init(&rbuf->base.reference, 1);
+    rbuf->base.screen = screen;
+    rbuf->base.alignment = 1;
+    rbuf->base.usage = 0;
+    rbuf->base.size = bytes;
+
+    rbuf->user_buffer = ptr;
+    return &rbuf->base;
+
+no_rbuf:
+    return NULL;
+}
+
+static void r300_buffer_destroy(struct pipe_buffer *buf)
+{
+    struct r300_screen *r300screen = r300_screen(buf->screen);
+    struct r300_buffer *rbuf = r300_buffer(buf);
+
+    r300_winsys_buffer_destroy(r300screen, rbuf);
+    FREE(rbuf);
+}
+
+static void *
+r300_buffer_map_range(struct pipe_screen *screen,
+		      struct pipe_buffer *buf,
+		      unsigned offset, unsigned length,
+		      unsigned usage )
+{
+    struct r300_screen *r300screen = r300_screen(screen);
+    struct r300_winsys_screen *rws = r300screen->rws;
+    struct r300_buffer *rbuf = r300_buffer(buf);
+    void *map;
+    int flush = 0;
+    int i;
+
+    if (rbuf->user_buffer)
+	return rbuf->user_buffer;
+
+    if (rbuf->base.usage & PIPE_BUFFER_USAGE_CONSTANT)
+	goto just_map;
+
+    /* check if the mapping is to a range we already flushed */
+    if (usage & PIPE_BUFFER_USAGE_DISCARD) {
+	for (i = 0; i < rbuf->num_ranges; i++) {
+
+	    if ((offset >= rbuf->ranges[i].start) &&
+		(offset < rbuf->ranges[i].end))
+		flush = 1;
+	    
+	    if (flush) {
+		/* unreference this hw buffer and allocate a new one */
+		rws->buffer_reference(rws, &rbuf->buf, NULL);
+
+		rbuf->num_ranges = 0;
+		rbuf->map = NULL;
+		rbuf->buf = r300_winsys_buffer_create(r300screen,
+						      rbuf->base.alignment,
+						      rbuf->base.usage,
+						      rbuf->base.size);
+		break;
+	    }
+	}
+    }
+just_map:
+    map = rws->buffer_map(rws, rbuf->buf, usage | R300_USAGE_FLAG_DONT_SYNC);
+   
+    return map;
+}
+
+static void 
+r300_buffer_flush_mapped_range( struct pipe_screen *screen,
+				struct pipe_buffer *buf,
+				unsigned offset,
+				unsigned length )
+{
+    struct r300_buffer *rbuf = r300_buffer(buf);
+    int i;
+
+    if (rbuf->user_buffer)
+	return;
+
+    if (rbuf->base.usage & PIPE_BUFFER_USAGE_CONSTANT)
+	return;
+
+    /* mark the range as used */
+    for(i = 0; i < rbuf->num_ranges; ++i) {
+	if(offset <= rbuf->ranges[i].end && rbuf->ranges[i].start <= (offset+length)) {
+	    rbuf->ranges[i].start = MIN2(rbuf->ranges[i].start, offset);
+	    rbuf->ranges[i].end   = MAX2(rbuf->ranges[i].end, (offset+length));
+	    return;
+	}
+    }
+
+    rbuf->ranges[rbuf->num_ranges].start = offset;
+    rbuf->ranges[rbuf->num_ranges].end = offset+length;
+    rbuf->num_ranges++;
+}
+
+static void *
+r300_buffer_map(struct pipe_screen *screen,
+		struct pipe_buffer *buf,
+		unsigned usage)
+{
+    struct r300_screen *r300screen = r300_screen(screen);
+    struct r300_winsys_screen *rws = r300screen->rws;
+    struct r300_buffer *rbuf = r300_buffer(buf);
+    void *map;
+
+   if (rbuf->user_buffer)
+      return rbuf->user_buffer;
+
+    map = rws->buffer_map(rws, rbuf->buf, usage);
+
+    return map;
+}
+
+static void
+r300_buffer_unmap(struct pipe_screen *screen,
+		  struct pipe_buffer *buf)
+{
+    struct r300_screen *r300screen = r300_screen(screen);
+    struct r300_winsys_screen *rws = r300screen->rws;
+    struct r300_buffer *rbuf = r300_buffer(buf);
+
+    if (rbuf->buf) {
+        rws->buffer_unmap(rws, rbuf->buf);
+    }
+}
+
+void r300_screen_init_buffer_functions(struct r300_screen *r300screen)
+{
+    r300screen->screen.buffer_create = r300_buffer_create;
+    r300screen->screen.user_buffer_create = r300_user_buffer_create;
+    r300screen->screen.buffer_map = r300_buffer_map;
+    r300screen->screen.buffer_map_range = r300_buffer_map_range;
+    r300screen->screen.buffer_flush_mapped_range = r300_buffer_flush_mapped_range;
+    r300screen->screen.buffer_unmap = r300_buffer_unmap;
+    r300screen->screen.buffer_destroy = r300_buffer_destroy;
+}
diff --git a/src/gallium/drivers/r300/r300_screen_buffer.h b/src/gallium/drivers/r300/r300_screen_buffer.h
new file mode 100644
index 0000000000..0cf349c25c
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_screen_buffer.h
@@ -0,0 +1,99 @@
+#ifndef R300_SCREEN_BUFFER_H
+#define R300_SCREEN_BUFFER_H
+#include <stdio.h>
+#include "pipe/p_compiler.h"
+#include "pipe/p_state.h"
+#include "r300_screen.h"
+
+#include "r300_winsys.h"
+#include "r300_context.h"
+
+#define R300_BUFFER_MAGIC 0xabcd1234
+
+struct r300_buffer_range {
+    uint32_t start;
+    uint32_t end;
+};
+#define R300_BUFFER_MAX_RANGES 32
+
+struct r300_buffer
+{
+    struct pipe_buffer base;
+
+    uint32_t magic;
+
+    struct r300_winsys_buffer *buf;
+
+    void *user_buffer;
+    struct r300_buffer_range ranges[R300_BUFFER_MAX_RANGES];
+    unsigned num_ranges;
+
+    void *map;
+};
+
+static INLINE struct r300_buffer *
+r300_buffer(struct pipe_buffer *buffer)
+{
+    if (buffer) {
+	assert(((struct r300_buffer *)buffer)->magic == R300_BUFFER_MAGIC);
+	return (struct r300_buffer *)buffer;
+    }
+    return NULL;
+}
+
+static INLINE boolean 
+r300_buffer_is_user_buffer(struct pipe_buffer *buffer)
+{
+    return r300_buffer(buffer)->user_buffer ? true : false;
+}
+
+static INLINE boolean r300_add_buffer(struct r300_winsys_screen *rws,
+				      struct pipe_buffer *buffer,
+				      int rd, int wr)
+{
+    struct r300_buffer *buf = r300_buffer(buffer);
+
+    if (!buf->buf)
+	return true;
+
+    return rws->add_buffer(rws, buf->buf, rd, wr);
+}
+
+
+static INLINE boolean r300_add_texture(struct r300_winsys_screen *rws,
+				       struct r300_texture *tex,
+				       int rd, int wr)
+{
+    return rws->add_buffer(rws, tex->buffer, rd, wr);
+}
+
+void r300_screen_init_buffer_functions(struct r300_screen *r300screen);
+
+static INLINE void r300_buffer_write_reloc(struct r300_winsys_screen *rws,
+				      struct r300_buffer *buf,
+				      uint32_t rd, uint32_t wd, uint32_t flags)
+{
+    if (!buf->buf)
+	return;
+
+    rws->write_cs_reloc(rws, buf->buf, rd, wd, flags);
+}
+
+static INLINE void r300_texture_write_reloc(struct r300_winsys_screen *rws,
+					    struct r300_texture *texture,
+					    uint32_t rd, uint32_t wd, uint32_t flags)
+{
+    rws->write_cs_reloc(rws, texture->buffer, rd, wd, flags);
+}
+
+int r300_upload_user_buffers(struct r300_context *r300);
+
+int r300_upload_index_buffer(struct r300_context *r300,
+			     struct pipe_buffer **index_buffer,
+			     unsigned index_size,
+			     unsigned start,
+			     unsigned count);
+
+boolean r300_buffer_is_referenced(struct r300_context *r300,
+				  struct pipe_buffer *buf);
+#endif
diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c
index d4b7376f12..71ac89715c 100644
--- a/src/gallium/drivers/r300/r300_state.c
+++ b/src/gallium/drivers/r300/r300_state.c
@@ -34,6 +34,7 @@
 #include "r300_context.h"
 #include "r300_reg.h"
 #include "r300_screen.h"
+#include "r300_screen_buffer.h"
 #include "r300_state_inlines.h"
 #include "r300_fs.h"
 #include "r300_vs.h"
@@ -525,7 +526,7 @@ static void r300_fb_update_tiling_flags(struct r300_context *r300,
         tex = (struct r300_texture*)old_state->cbufs[i]->texture;
 
         if (tex) {
-            r300->winsys->buffer_set_tiling(r300->winsys, tex->buffer,
+            r300->rws->buffer_set_tiling(r300->rws, tex->buffer,
                                             tex->pitch[0],
                                             tex->microtile != 0,
                                             tex->macrotile != 0);
@@ -537,7 +538,7 @@ static void r300_fb_update_tiling_flags(struct r300_context *r300,
         tex = (struct r300_texture*)old_state->zsbuf->texture;
 
         if (tex) {
-            r300->winsys->buffer_set_tiling(r300->winsys, tex->buffer,
+            r300->rws->buffer_set_tiling(r300->rws, tex->buffer,
                                             tex->pitch[0],
                                             tex->microtile != 0,
                                             tex->macrotile != 0);
@@ -549,7 +550,7 @@ static void r300_fb_update_tiling_flags(struct r300_context *r300,
         tex = (struct r300_texture*)new_state->cbufs[i]->texture;
         level = new_state->cbufs[i]->level;
 
-        r300->winsys->buffer_set_tiling(r300->winsys, tex->buffer,
+        r300->rws->buffer_set_tiling(r300->rws, tex->buffer,
                                         tex->pitch[level],
                                         tex->microtile != 0,
                                         tex->mip_macrotile[level] != 0);
@@ -558,7 +559,7 @@ static void r300_fb_update_tiling_flags(struct r300_context *r300,
         tex = (struct r300_texture*)new_state->zsbuf->texture;
         level = new_state->zsbuf->level;
 
-        r300->winsys->buffer_set_tiling(r300->winsys, tex->buffer,
+        r300->rws->buffer_set_tiling(r300->rws, tex->buffer,
                                         tex->pitch[level],
                                         tex->microtile != 0,
                                         tex->mip_macrotile[level] != 0);
@@ -849,6 +850,7 @@ static void*
 {
     struct r300_context* r300 = r300_context(pipe);
     struct r300_sampler_state* sampler = CALLOC_STRUCT(r300_sampler_state);
+    boolean is_r500 = r300_screen(pipe->screen)->caps->is_r500;
     int lod_bias;
     union util_color uc;
 
@@ -864,6 +866,8 @@ static void*
                                                    state->min_mip_filter,
                                                    state->max_anisotropy > 0);
 
+    sampler->filter0 |= r300_anisotropy(state->max_anisotropy);
+
     /* Unfortunately, r300-r500 don't support floating-point mipmap lods. */
     /* We must pass these to the merge function to clamp them properly. */
     sampler->min_lod = MAX2((unsigned)state->min_lod, 0);
@@ -873,7 +877,13 @@ static void*
 
     sampler->filter1 |= lod_bias << R300_LOD_BIAS_SHIFT;
 
-    sampler->filter1 |= r300_anisotropy(state->max_anisotropy);
+    /* This is very high quality anisotropic filtering for R5xx.
+     * It's good for benchmarking the performance of texturing but
+     * in practice we don't want to slow down the driver because it's
+     * a pretty good performance killer. Feel free to play with it. */
+    if (DBG_ON(r300, DBG_ANISOHQ) && is_r500) {
+        sampler->filter1 |= r500_anisotropy(state->max_anisotropy);
+    }
 
     util_pack_color(state->border_color, PIPE_FORMAT_B8G8R8A8_UNORM, &uc);
     sampler->border_color = uc.ui;
@@ -1061,17 +1071,30 @@ static void r300_set_vertex_buffers(struct pipe_context* pipe,
                                     const struct pipe_vertex_buffer* buffers)
 {
     struct r300_context* r300 = r300_context(pipe);
-    unsigned i, max_index = (1 << 24) - 1;
+    int i;
+    unsigned max_index = (1 << 24) - 1;
+    boolean any_user_buffer = false;
 
-    memcpy(r300->vertex_buffer, buffers,
-        sizeof(struct pipe_vertex_buffer) * count);
+    if (count == r300->vertex_buffer_count &&
+	memcmp(r300->vertex_buffer, buffers, count * sizeof(buffers[0])) == 0)
+        return;
 
     for (i = 0; i < count; i++) {
+	pipe_buffer_reference(&r300->vertex_buffer[i].buffer, buffers[i].buffer);
+	if (r300_buffer_is_user_buffer(buffers[i].buffer))
+	    any_user_buffer = true;
         max_index = MIN2(buffers[i].max_index, max_index);
     }
 
+    for ( ; i < r300->vertex_buffer_count; i++)
+	pipe_buffer_reference(&r300->vertex_buffer[i].buffer, NULL);
+
+    memcpy(r300->vertex_buffer, buffers,
+	   sizeof(struct pipe_vertex_buffer) * count);
+
     r300->vertex_buffer_count = count;
     r300->vertex_buffer_max_index = max_index;
+    r300->any_user_vbs = any_user_buffer;
 
     if (r300->draw) {
         draw_flush(r300->draw);
diff --git a/src/gallium/drivers/r300/r300_state_inlines.h b/src/gallium/drivers/r300/r300_state_inlines.h
index a32924ed0a..8485d4f8f9 100644
--- a/src/gallium/drivers/r300/r300_state_inlines.h
+++ b/src/gallium/drivers/r300/r300_state_inlines.h
@@ -327,6 +327,18 @@ static INLINE uint32_t r300_anisotropy(unsigned max_aniso)
     }
 }
 
+static INLINE uint32_t r500_anisotropy(unsigned max_aniso)
+{
+    if (!max_aniso) {
+        return 0;
+    }
+    max_aniso -= 1;
+
+    // Map the range [0, 15] to [0, 63].
+    return R500_TX_MAX_ANISO(MIN2((unsigned)(max_aniso*4.2001), 63)) |
+           R500_TX_ANISO_HIGH_QUALITY;;
+}
+
 /* Non-CSO state. (For now.) */
 
 static INLINE uint32_t r300_translate_gb_pipes(int pipe_count)
diff --git a/src/gallium/drivers/r300/r300_texture.c b/src/gallium/drivers/r300/r300_texture.c
index 04124afd4d..7c7656068b 100644
--- a/src/gallium/drivers/r300/r300_texture.c
+++ b/src/gallium/drivers/r300/r300_texture.c
@@ -773,7 +773,7 @@ static struct pipe_texture* r300_texture_create(struct pipe_screen* screen,
 {
     struct r300_texture* tex = CALLOC_STRUCT(r300_texture);
     struct r300_screen* rscreen = r300_screen(screen);
-    struct radeon_winsys* winsys = (struct radeon_winsys*)screen->winsys;
+    struct r300_winsys_screen *rws = (struct r300_winsys_screen *)screen->winsys;
 
     if (!tex) {
         return NULL;
@@ -790,13 +790,13 @@ static struct pipe_texture* r300_texture_create(struct pipe_screen* screen,
     r300_setup_miptree(rscreen, tex);
     r300_setup_texture_state(rscreen, tex);
 
-    tex->buffer = screen->buffer_create(screen, 2048,
-                                        PIPE_BUFFER_USAGE_PIXEL,
-                                        tex->size);
-    winsys->buffer_set_tiling(winsys, tex->buffer,
-                              tex->pitch[0],
-                              tex->microtile != R300_BUFFER_LINEAR,
-                              tex->macrotile != R300_BUFFER_LINEAR);
+    tex->buffer = rws->buffer_create(rws, 2048,
+				     PIPE_BUFFER_USAGE_PIXEL,
+				     tex->size);
+    rws->buffer_set_tiling(rws, tex->buffer,
+			   tex->pitch[0],
+			   tex->microtile != R300_BUFFER_LINEAR,
+			   tex->macrotile != R300_BUFFER_LINEAR);
 
     if (!tex->buffer) {
         FREE(tex);
@@ -809,9 +809,9 @@ static struct pipe_texture* r300_texture_create(struct pipe_screen* screen,
 static void r300_texture_destroy(struct pipe_texture* texture)
 {
     struct r300_texture* tex = (struct r300_texture*)texture;
+    struct r300_winsys_screen *rws = (struct r300_winsys_screen *)texture->screen->winsys;
 
-    pipe_buffer_reference(&tex->buffer, NULL);
-
+    rws->buffer_reference(rws, &tex->buffer, NULL);
     FREE(tex);
 }
 
@@ -857,9 +857,9 @@ static struct pipe_texture*
                              const struct pipe_texture* base,
                              struct winsys_handle *whandle)
 {
-    struct radeon_winsys* winsys = (struct radeon_winsys*)screen->winsys;
+    struct r300_winsys_screen *rws = (struct r300_winsys_screen*)screen->winsys;
     struct r300_screen* rscreen = r300_screen(screen);
-    struct pipe_buffer *buffer;
+    struct r300_winsys_buffer *buffer;
     struct r300_texture* tex;
     unsigned stride;
 
@@ -870,7 +870,7 @@ static struct pipe_texture*
         return NULL;
     }
 
-    buffer = winsys->buffer_from_handle(winsys, screen, whandle, &stride);
+    buffer = rws->buffer_from_handle(rws, screen, whandle, &stride);
     if (!buffer) {
         return NULL;
     }
@@ -901,7 +901,7 @@ static boolean
                             struct pipe_texture *texture,
                             struct winsys_handle *whandle)
 {
-    struct radeon_winsys* winsys = (struct radeon_winsys*)screen->winsys;
+    struct r300_winsys_screen *rws = (struct r300_winsys_screen *)screen->winsys;
     struct r300_texture* tex = (struct r300_texture*)texture;
     unsigned stride;
 
@@ -911,7 +911,7 @@ static boolean
 
     stride = r300_texture_get_stride(r300_screen(screen), tex, 0);
 
-    winsys->buffer_get_handle(winsys, tex->buffer, stride, whandle);
+    rws->buffer_get_handle(rws, tex->buffer, stride, whandle);
 
     return TRUE;
 }
@@ -977,3 +977,25 @@ void r300_init_screen_texture_functions(struct pipe_screen* screen)
     screen->video_surface_destroy= r300_video_surface_destroy;
 }
 
+boolean r300_get_texture_buffer(struct pipe_screen* screen,
+                                struct pipe_texture* texture,
+                                struct r300_winsys_buffer** buffer,
+                                unsigned* stride)
+{
+    struct r300_texture* tex = (struct r300_texture*)texture;
+    struct r300_winsys_screen *rws = (struct r300_winsys_screen *)screen->winsys;
+    struct r300_winsys_buffer *buf;
+
+    if (!tex) {
+        return FALSE;
+    }
+
+    rws->buffer_reference(rws, &buf, tex->buffer);
+
+    if (stride) {
+        *stride = r300_texture_get_stride(r300_screen(screen), tex, 0);
+    }
+
+    *buffer = buf;
+    return TRUE;
+}
diff --git a/src/gallium/drivers/r300/r300_texture.h b/src/gallium/drivers/r300/r300_texture.h
index 138b62784e..60c7fa8342 100644
--- a/src/gallium/drivers/r300/r300_texture.h
+++ b/src/gallium/drivers/r300/r300_texture.h
@@ -63,8 +63,8 @@ r300_video_surface(struct pipe_video_surface *pvs)
 /* Used internally for texture_is_referenced()
  */
 boolean r300_get_texture_buffer(struct pipe_screen* screen,
-                                struct pipe_texture* texture,
-                                struct pipe_buffer** buffer,
+                                struct pipe_texture *texture,
+                                struct r300_winsys_buffer** buffer,
                                 unsigned* stride);
 
 #endif /* R300_TEXTURE_H */
diff --git a/src/gallium/drivers/r300/r300_transfer.c b/src/gallium/drivers/r300/r300_transfer.c
index 495e3dee76..987a040698 100644
--- a/src/gallium/drivers/r300/r300_transfer.c
+++ b/src/gallium/drivers/r300/r300_transfer.c
@@ -26,6 +26,8 @@
 #include "r300_texture.h"
 #include "r300_screen.h"
 
+#include "r300_winsys.h"
+
 #include "util/u_memory.h"
 #include "util/u_format.h"
 
@@ -225,6 +227,7 @@ static void r300_tex_transfer_destroy(struct pipe_context *ctx,
 static void* r300_transfer_map(struct pipe_context *ctx,
                                struct pipe_transfer *transfer)
 {
+    struct r300_winsys_screen *rws = (struct r300_winsys_screen *)ctx->winsys;
     struct r300_transfer *r300transfer = r300_transfer(transfer);
     struct r300_texture *tex = (struct r300_texture*)transfer->texture;
     char *map;
@@ -233,12 +236,12 @@ static void* r300_transfer_map(struct pipe_context *ctx,
     if (r300transfer->detiled_texture) {
         /* The detiled texture is of the same size as the region being mapped
          * (no offset needed). */
-        return pipe_buffer_map(ctx->screen,
+        return rws->buffer_map(rws,
                                r300transfer->detiled_texture->buffer,
                                pipe_transfer_buffer_flags(transfer));
     } else {
         /* Tiling is disabled. */
-        map = pipe_buffer_map(ctx->screen, tex->buffer,
+        map = rws->buffer_map(rws, tex->buffer,
                               pipe_transfer_buffer_flags(transfer));
 
         if (!map) {
@@ -254,13 +257,14 @@ static void* r300_transfer_map(struct pipe_context *ctx,
 static void r300_transfer_unmap(struct pipe_context *ctx,
                                 struct pipe_transfer *transfer)
 {
+    struct r300_winsys_screen *rws = (struct r300_winsys_screen *)ctx->winsys;
     struct r300_transfer *r300transfer = r300_transfer(transfer);
     struct r300_texture *tex = (struct r300_texture*)transfer->texture;
 
     if (r300transfer->detiled_texture) {
-        pipe_buffer_unmap(ctx->screen, r300transfer->detiled_texture->buffer);
+	rws->buffer_unmap(rws, r300transfer->detiled_texture->buffer);
     } else {
-        pipe_buffer_unmap(ctx->screen, tex->buffer);
+        rws->buffer_unmap(rws, tex->buffer);
     }
 }
 
diff --git a/src/gallium/drivers/r300/r300_vs.c b/src/gallium/drivers/r300/r300_vs.c
index 379939ac75..bd6b95dccb 100644
--- a/src/gallium/drivers/r300/r300_vs.c
+++ b/src/gallium/drivers/r300/r300_vs.c
@@ -34,8 +34,6 @@
 
 #include "radeon_compiler.h"
 
-#include "util/u_math.h"
-
 /* Convert info about VS output semantics into r300_shader_semantics. */
 static void r300_shader_read_vs_outputs(
     struct tgsi_shader_info* info,
diff --git a/src/gallium/drivers/r300/r300_winsys.h b/src/gallium/drivers/r300/r300_winsys.h
index ddf2b79003..e5183a8239 100644
--- a/src/gallium/drivers/r300/r300_winsys.h
+++ b/src/gallium/drivers/r300/r300_winsys.h
@@ -23,10 +23,6 @@
 #ifndef R300_WINSYS_H
 #define R300_WINSYS_H
 
-#ifdef __cplusplus
-extern "C" {
-#endif
-
 /* The public interface header for the r300 pipe driver.
  * Any winsys hosting this pipe needs to implement r300_winsys and then
  * call r300_create_screen to start things. */
@@ -34,14 +30,146 @@ extern "C" {
 #include "pipe/p_defines.h"
 #include "pipe/p_state.h"
 
-struct radeon_winsys;
+struct r300_winsys_screen;
 
 /* Creates a new r300 screen. */
-struct pipe_screen* r300_create_screen(struct radeon_winsys* radeon_winsys);
+struct pipe_screen* r300_create_screen(struct r300_winsys_screen *rws);
+
+struct r300_winsys_buffer;
+
+
+boolean r300_get_texture_buffer(struct pipe_screen* screen,
+                                struct pipe_texture* texture,
+                                struct r300_winsys_buffer** buffer,
+                                unsigned *stride);
+
+enum r300_value_id {
+    R300_VID_PCI_ID,
+    R300_VID_GB_PIPES,
+    R300_VID_Z_PIPES,
+};
+
+#define R300_USAGE_FLAG_DONT_SYNC (1 << 17)
+
+struct r300_winsys_screen {
+    void (*destroy)(struct r300_winsys_screen *ws);
+    
+    /**
+     * Buffer management. Buffer attributes are mostly fixed over its lifetime.
+     *
+     * Remember that gallium gets to choose the interface it needs, and the
+     * window systems must then implement that interface (rather than the
+     * other way around...).
+     *
+     * usage is a bitmask of R300_WINSYS_BUFFER_USAGE_PIXEL/VERTEX/INDEX/CONSTANT. This
+     * usage argument is only an optimization hint, not a guarantee, therefore
+     * proper behavior must be observed in all circumstances.
+     *
+     * alignment indicates the client's alignment requirements, eg for
+     * SSE instructions.
+     */
+    struct r300_winsys_buffer *(*buffer_create)(struct r300_winsys_screen *ws,
+						unsigned alignment,
+						unsigned usage,
+						unsigned size);
+    
+    /**
+     * Map the entire data store of a buffer object into the client's address.
+     * flags is bitmask of R300_WINSYS_BUFFER_USAGE_CPU_READ/WRITE flags.
+     */
+    void *(*buffer_map)( struct r300_winsys_screen *ws,
+			 struct r300_winsys_buffer *buf,
+			 unsigned usage);
+
+    void (*buffer_unmap)( struct r300_winsys_screen *ws,
+			  struct r300_winsys_buffer *buf );
+
+    void (*buffer_destroy)( struct r300_winsys_buffer *buf );
+
+
+    void (*buffer_reference)(struct r300_winsys_screen *rws,
+			     struct r300_winsys_buffer **pdst,
+			     struct r300_winsys_buffer *src);
+
+    boolean (*buffer_references)(struct r300_winsys_buffer *a,
+				 struct r300_winsys_buffer *b);
+
+    void (*buffer_flush_range)(struct r300_winsys_screen *rws,
+			       struct r300_winsys_buffer *buf,
+			       unsigned offset,
+			       unsigned length);
+
+    /* Add a pipe_buffer to the list of buffer objects to validate. */
+    boolean (*add_buffer)(struct r300_winsys_screen *winsys,
+                          struct r300_winsys_buffer *buf,
+                          uint32_t rd,
+                          uint32_t wd);
+
+
+    /* Revalidate all currently setup pipe_buffers.
+     * Returns TRUE if a flush is required. */
+    boolean (*validate)(struct r300_winsys_screen* winsys);
+
+    /* Check to see if there's room for commands. */
+    boolean (*check_cs)(struct r300_winsys_screen* winsys, int size);
+
+    /* Start a command emit. */
+    void (*begin_cs)(struct r300_winsys_screen* winsys,
+                     int size,
+                     const char* file,
+                     const char* function,
+                     int line);
+
+    /* Write a dword to the command buffer. */
+    void (*write_cs_dword)(struct r300_winsys_screen* winsys, uint32_t dword);
+
+    /* Write a relocated dword to the command buffer. */
+    void (*write_cs_reloc)(struct r300_winsys_screen *winsys,
+                           struct r300_winsys_buffer *buf,
+                           uint32_t rd,
+                           uint32_t wd,
+                           uint32_t flags);
+
+    /* Finish a command emit. */
+    void (*end_cs)(struct r300_winsys_screen* winsys,
+                   const char* file,
+                   const char* function,
+                   int line);
+
+    /* Flush the CS. */
+    void (*flush_cs)(struct r300_winsys_screen* winsys);
+
+    /* winsys flush - callback from winsys when flush required */
+    void (*set_flush_cb)(struct r300_winsys_screen *winsys,
+			 void (*flush_cb)(void *), void *data);
+
+    void (*reset_bos)(struct r300_winsys_screen *winsys);
+
+    void (*buffer_set_tiling)(struct r300_winsys_screen *winsys,
+                              struct r300_winsys_buffer *buffer,
+                              uint32_t pitch,
+                              boolean microtiled,
+                              boolean macrotiled);
+
+    uint32_t (*get_value)(struct r300_winsys_screen *winsys,
+			  enum r300_value_id vid);
+
+    struct r300_winsys_buffer *(*buffer_from_handle)(struct r300_winsys_screen *winsys,
+						     struct pipe_screen *screen,
+						     struct winsys_handle *whandle,
+						     unsigned *stride);
+    boolean (*buffer_get_handle)(struct r300_winsys_screen *winsys,
+				 struct r300_winsys_buffer *buffer,
+				 unsigned stride,
+				 struct winsys_handle *whandle);
+
+    boolean (*is_buffer_referenced)(struct r300_winsys_screen *winsys,
+                                    struct r300_winsys_buffer *buffer);
 
+  
+};
 
-#ifdef __cplusplus
-}
-#endif
+struct r300_winsys_screen *
+r300_winsys_screen(struct pipe_screen *screen);
 
 #endif /* R300_WINSYS_H */
diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c
index da8529c154..2aff6118f4 100644
--- a/src/gallium/drivers/softpipe/sp_texture.c
+++ b/src/gallium/drivers/softpipe/sp_texture.c
@@ -36,7 +36,6 @@
 #include "util/u_format.h"
 #include "util/u_math.h"
 #include "util/u_memory.h"
-#include "util/u_simple_screen.h"
 
 #include "sp_context.h"
 #include "sp_texture.h"
diff --git a/src/gallium/drivers/svga/svga_screen_texture.c b/src/gallium/drivers/svga/svga_screen_texture.c
index 107e4a3962..4a058eda88 100644
--- a/src/gallium/drivers/svga/svga_screen_texture.c
+++ b/src/gallium/drivers/svga/svga_screen_texture.c
@@ -41,8 +41,6 @@
 #include "svga_debug.h"
 #include "svga_screen_buffer.h"
 
-#include <util/u_string.h>
-
 
 /* XXX: This isn't a real hardware flag, but just a hack for kernel to
  * know about primary surfaces. Find a better way to accomplish this.
diff --git a/src/gallium/drivers/trace/tr_drm.c b/src/gallium/drivers/trace/tr_drm.c
index 906b3262e4..eaa47df406 100644
--- a/src/gallium/drivers/trace/tr_drm.c
+++ b/src/gallium/drivers/trace/tr_drm.c
@@ -30,9 +30,6 @@
 #include "util/u_memory.h"
 #include "tr_drm.h"
 #include "tr_screen.h"
-#include "tr_context.h"
-#include "tr_buffer.h"
-#include "tr_texture.h"
 #include "tr_public.h"
 
 struct trace_drm_api
diff --git a/src/gallium/state_trackers/glx/xlib/Makefile b/src/gallium/state_trackers/glx/xlib/Makefile
index 7b2adc62c3..8c7cc524df 100644
--- a/src/gallium/state_trackers/glx/xlib/Makefile
+++ b/src/gallium/state_trackers/glx/xlib/Makefile
@@ -5,7 +5,8 @@ LIBNAME = xlib
 
 LIBRARY_INCLUDES = \
 	-I$(TOP)/include \
-	-I$(TOP)/src/mesa
+	-I$(TOP)/src/mesa \
+	$(X_CFLAGS)
 
 C_SOURCES = \
 	glx_api.c \
diff --git a/src/gallium/winsys/drm/nouveau/dri/Makefile b/src/gallium/winsys/drm/nouveau/dri/Makefile
index 7e95f79d03..50ac3f203e 100644
--- a/src/gallium/winsys/drm/nouveau/dri/Makefile
+++ b/src/gallium/winsys/drm/nouveau/dri/Makefile
@@ -6,8 +6,7 @@ LIBNAME = nouveau_dri.so
 PIPE_DRIVERS = \
 	$(TOP)/src/gallium/state_trackers/dri/libdridrm.a \
 	$(TOP)/src/gallium/winsys/drm/nouveau/drm/libnouveaudrm.a \
-	$(TOP)/src/gallium/drivers/nv30/libnv30.a \
-	$(TOP)/src/gallium/drivers/nv40/libnv40.a \
+	$(TOP)/src/gallium/drivers/nvfx/libnvfx.a \
 	$(TOP)/src/gallium/drivers/nv50/libnv50.a \
 	$(TOP)/src/gallium/drivers/nouveau/libnouveau.a
 
diff --git a/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c
index 21517b4bb5..716d4bacd3 100644
--- a/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c
+++ b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c
@@ -86,11 +86,9 @@ nouveau_drm_create_screen(struct drm_api *api, int fd,
 
 	switch (dev->chipset & 0xf0) {
 	case 0x30:
-		init = nv30_screen_create;
-		break;
 	case 0x40:
 	case 0x60:
-		init = nv40_screen_create;
+		init = nvfx_screen_create;
 		break;
 	case 0x50:
 	case 0x80:
diff --git a/src/gallium/winsys/drm/nouveau/egl/Makefile b/src/gallium/winsys/drm/nouveau/egl/Makefile
index 2c35260332..47d1127615 100644
--- a/src/gallium/winsys/drm/nouveau/egl/Makefile
+++ b/src/gallium/winsys/drm/nouveau/egl/Makefile
@@ -7,8 +7,7 @@ EGL_DRIVER_LIBS = -ldrm_nouveau
 
 EGL_DRIVER_PIPES = \
 	$(TOP)/src/gallium/winsys/drm/nouveau/drm/libnouveaudrm.a \
-	$(TOP)/src/gallium/drivers/nv30/libnv30.a \
-	$(TOP)/src/gallium/drivers/nv40/libnv40.a \
+	$(TOP)/src/gallium/drivers/nvfx/libnvfx.a \
 	$(TOP)/src/gallium/drivers/nv50/libnv50.a \
 	$(TOP)/src/gallium/drivers/nouveau/libnouveau.a \
 	$(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a
diff --git a/src/gallium/winsys/drm/nouveau/xorg/Makefile b/src/gallium/winsys/drm/nouveau/xorg/Makefile
index 179b50230b..f7f6fe17dd 100644
--- a/src/gallium/winsys/drm/nouveau/xorg/Makefile
+++ b/src/gallium/winsys/drm/nouveau/xorg/Makefile
@@ -18,8 +18,7 @@ INCLUDES = \
 LIBS = \
 	$(TOP)/src/gallium/state_trackers/xorg/libxorgtracker.a \
 	$(TOP)/src/gallium/winsys/drm/nouveau/drm/libnouveaudrm.a \
-	$(TOP)/src/gallium/drivers/nv30/libnv30.a \
-	$(TOP)/src/gallium/drivers/nv40/libnv40.a \
+	$(TOP)/src/gallium/drivers/nvfx/libnvfx.a \
 	$(TOP)/src/gallium/drivers/nv50/libnv50.a \
 	$(TOP)/src/gallium/drivers/nouveau/libnouveau.a \
 	$(GALLIUM_AUXILIARIES)
diff --git a/src/gallium/winsys/drm/radeon/core/Makefile b/src/gallium/winsys/drm/radeon/core/Makefile
index 860cbb6dbf..13bbbf730d 100644
--- a/src/gallium/winsys/drm/radeon/core/Makefile
+++ b/src/gallium/winsys/drm/radeon/core/Makefile
@@ -5,7 +5,7 @@ include $(TOP)/configs/current
 LIBNAME = radeonwinsys
 
 C_SOURCES = \
-	radeon_buffer.c \
+	radeon_drm_buffer.c \
 	radeon_drm.c \
 	radeon_r300.c
 
diff --git a/src/gallium/winsys/drm/radeon/core/radeon_buffer.c b/src/gallium/winsys/drm/radeon/core/radeon_buffer.c
index daa032af6f..25b58b2926 100644
--- a/src/gallium/winsys/drm/radeon/core/radeon_buffer.c
+++ b/src/gallium/winsys/drm/radeon/core/radeon_buffer.c
@@ -213,6 +213,18 @@ static void radeon_buffer_unmap(struct pipe_winsys *ws,
     }
 }
 
+static boolean radeon_is_buffer_referenced(struct radeon_winsys *ws,
+                                           struct pipe_buffer *buffer)
+{
+    struct radeon_pipe_buffer *radeon_buffer =
+        (struct radeon_pipe_buffer*)buffer;
+    uint32_t domain;
+
+    /* Referenced by CS or HW. */
+    return radeon_bo_is_referenced_by_cs(radeon_buffer->bo, ws->priv->cs) ||
+           radeon_bo_is_busy(radeon_buffer->bo, &domain);
+}
+
 static void radeon_buffer_set_tiling(struct radeon_winsys *ws,
                                      struct pipe_buffer *buffer,
                                      uint32_t pitch,
@@ -370,5 +382,7 @@ struct radeon_winsys* radeon_pipe_winsys(int fd)
     radeon_ws->buffer_from_handle = radeon_buffer_from_handle;
     radeon_ws->buffer_get_handle = radeon_buffer_get_handle;
 
+    radeon_ws->is_buffer_referenced = radeon_is_buffer_referenced;
+
     return radeon_ws;
 }
diff --git a/src/gallium/winsys/drm/radeon/core/radeon_buffer.h b/src/gallium/winsys/drm/radeon/core/radeon_buffer.h
index f776e2d900..e1fcfcfcca 100644
--- a/src/gallium/winsys/drm/radeon/core/radeon_buffer.h
+++ b/src/gallium/winsys/drm/radeon/core/radeon_buffer.h
@@ -30,48 +30,56 @@
 #ifndef RADEON_BUFFER_H
 #define RADEON_BUFFER_H
 
+#include <stdio.h>
+
+#include "pipe/p_defines.h"
+#include "util/u_inlines.h"
+
 #include "pipebuffer/pb_buffer.h"
+#include "pipebuffer/pb_bufmgr.h"
 
 #include "radeon_bo.h"
 #include "radeon_cs.h"
 
 #include "radeon_winsys.h"
 
-struct radeon_pipe_buffer {
-    struct pipe_buffer  base;
-    /* Pointer to GPU-backed BO. */
-    struct radeon_bo    *bo;
-    /* Pointer to fallback PB buffer. */
-    struct pb_buffer    *pb;
-    boolean flinked;
-    uint32_t flink;
-};
 
 #define RADEON_MAX_BOS 24
 
-struct radeon_winsys_priv {
-    /* DRM FD */
-    int fd;
+static INLINE struct pb_buffer *
+radeon_pb_buffer(struct r300_winsys_buffer *buffer)
+{
+    return (struct pb_buffer *)buffer;
+}
 
-    /* Radeon BO manager. */
-    struct radeon_bo_manager* bom;
+static INLINE struct r300_winsys_buffer *
+radeon_libdrm_winsys_buffer(struct pb_buffer *buffer)
+{
+    return (struct r300_winsys_buffer *)buffer;
+}
 
-    /* Radeon CS manager. */
-    struct radeon_cs_manager* csm;
+struct pb_manager *
+radeon_drm_bufmgr_create(struct radeon_libdrm_winsys *rws);
 
-    /* Current CS. */
-    struct radeon_cs* cs;
+boolean radeon_drm_bufmgr_add_buffer(struct pb_buffer *_buf,
+				     uint32_t rd, uint32_t wd);
 
-    /* Flush CB */
-    void (*flush_cb)(void *);
-    void *flush_data;
-};
 
-struct radeon_winsys* radeon_pipe_winsys(int fb);
-#if 0
-struct pipe_surface *radeon_surface_from_handle(struct radeon_context *radeon_context,
-                                             uint32_t handle,
-                                             enum pipe_format format,
-                                             int w, int h, int pitch);
-#endif
+void radeon_drm_bufmgr_write_reloc(struct pb_buffer *_buf,
+				   uint32_t rd, uint32_t wd,
+				   uint32_t flags);
+
+struct radeon_libdrm_winsys* radeon_pipe_winsys(int fd);
+
+struct pb_buffer *radeon_drm_bufmgr_create_buffer_from_handle(struct pb_manager *_mgr,
+							      uint32_t handle);
+
+void radeon_drm_bufmgr_set_tiling(struct pb_buffer *_buf, boolean microtiled, boolean macrotiled, uint32_t pitch);
+
+void radeon_drm_bufmgr_flush_maps(struct pb_manager *_mgr);
+
+boolean radeon_drm_bufmgr_get_handle(struct pb_buffer *_buf,
+				     struct winsys_handle *whandle);
+
+boolean radeon_drm_bufmgr_is_buffer_referenced(struct pb_buffer *_buf);
 #endif
diff --git a/src/gallium/winsys/drm/radeon/core/radeon_drm.c b/src/gallium/winsys/drm/radeon/core/radeon_drm.c
index 97edb6a47e..d70173e805 100644
--- a/src/gallium/winsys/drm/radeon/core/radeon_drm.c
+++ b/src/gallium/winsys/drm/radeon/core/radeon_drm.c
@@ -41,8 +41,22 @@
 #include "xf86drm.h"
 #include <sys/ioctl.h>
 
+static struct radeon_libdrm_winsys *
+radeon_winsys_create(int fd)
+{
+    struct radeon_libdrm_winsys *rws;
+
+    rws = CALLOC_STRUCT(radeon_libdrm_winsys);
+    if (rws == NULL) {
+        return NULL;
+    }
+
+    rws->fd = fd;
+    return rws;
+}
+
 /* Helper function to do the ioctls needed for setup and init. */
-static void do_ioctls(int fd, struct radeon_winsys* winsys)
+static void do_ioctls(int fd, struct radeon_libdrm_winsys* winsys)
 {
     struct drm_radeon_gem_info gem_info = {0};
     struct drm_radeon_info info = {0};
@@ -133,19 +147,28 @@ struct pipe_screen* radeon_create_screen(struct drm_api* api,
                                          int drmFB,
                                          struct drm_create_screen_arg *arg)
 {
-    struct radeon_winsys* rwinsys = radeon_pipe_winsys(drmFB);
-    do_ioctls(drmFB, rwinsys);
+    struct radeon_libdrm_winsys* rws; 
+    boolean ret;
+
+    rws = radeon_winsys_create(drmFB);
+    if (!rws)
+	return NULL;
+
+    do_ioctls(drmFB, rws);
 
     /* The state tracker can organize a softpipe fallback if no hw
      * driver is found.
      */
-    if (is_r3xx(rwinsys->pci_id)) {
-        radeon_setup_winsys(drmFB, rwinsys);
-        return r300_create_screen(rwinsys);
-    } else {
-        FREE(rwinsys);
-        return NULL;
+    if (is_r3xx(rws->pci_id)) {
+        ret = radeon_setup_winsys(drmFB, rws);
+	if (ret == FALSE)
+	    goto fail;
+        return r300_create_screen(&rws->base);
     }
+
+fail:
+    FREE(rws);
+    return NULL;
 }
 
 static void radeon_drm_api_destroy(struct drm_api *api)
diff --git a/src/gallium/winsys/drm/radeon/core/radeon_drm.h b/src/gallium/winsys/drm/radeon/core/radeon_drm.h
index 78451b6f01..2dc077c052 100644
--- a/src/gallium/winsys/drm/radeon/core/radeon_drm.h
+++ b/src/gallium/winsys/drm/radeon/core/radeon_drm.h
@@ -37,6 +37,22 @@ struct pipe_screen* radeon_create_screen(struct drm_api* api,
                                          int drmFB,
 					 struct drm_create_screen_arg *arg);
 
+boolean radeon_buffer_from_texture(struct drm_api* api,
+                                   struct pipe_screen* screen,
+                                   struct pipe_texture* texture,
+                                   struct pipe_buffer** buffer,
+                                   unsigned* stride);
+
+boolean radeon_handle_from_buffer(struct drm_api* api,
+                                  struct pipe_screen* screen,
+                                  struct pipe_buffer* buffer,
+                                  unsigned* handle);
+
+boolean radeon_global_handle_from_buffer(struct drm_api* api,
+                                         struct pipe_screen* screen,
+                                         struct pipe_buffer* buffer,
+                                         unsigned* handle);
+
 void radeon_destroy_drm_api(struct drm_api* api);
 
 /* Guess at whether this chipset should use r300g.
diff --git a/src/gallium/winsys/drm/radeon/core/radeon_drm_buffer.c b/src/gallium/winsys/drm/radeon/core/radeon_drm_buffer.c
new file mode 100644
index 0000000000..cc56a2bb8f
--- /dev/null
+++ b/src/gallium/winsys/drm/radeon/core/radeon_drm_buffer.c
@@ -0,0 +1,368 @@
+
+#include <sys/ioctl.h>
+#include "radeon_drm.h"
+#include "radeon_bo_gem.h"
+#include "radeon_cs_gem.h"
+#include "radeon_buffer.h"
+
+#include "util/u_inlines.h"
+#include "util/u_memory.h"
+#include "util/u_simple_list.h"
+#include "pipebuffer/pb_buffer.h"
+#include "pipebuffer/pb_bufmgr.h"
+
+#include "radeon_winsys.h"
+struct radeon_drm_bufmgr;
+
+struct radeon_drm_buffer {
+    struct pb_buffer base;
+    struct radeon_drm_bufmgr *mgr;
+
+    struct radeon_bo *bo;
+
+    boolean flinked;
+    uint32_t flink;
+
+    boolean mapped;
+    struct radeon_drm_buffer *next, *prev;
+};
+
+extern const struct pb_vtbl radeon_drm_buffer_vtbl;
+
+
+static INLINE struct radeon_drm_buffer *
+radeon_drm_buffer(struct pb_buffer *buf)
+{
+    assert(buf);
+    assert(buf->vtbl == &radeon_drm_buffer_vtbl);
+    return (struct radeon_drm_buffer *)buf;
+}
+
+struct radeon_drm_bufmgr {
+    struct pb_manager base;
+    struct radeon_libdrm_winsys *rws;
+    struct radeon_drm_buffer buffer_map_list;
+};
+
+static INLINE struct radeon_drm_bufmgr *
+radeon_drm_bufmgr(struct pb_manager *mgr)
+{
+    assert(mgr);
+    return (struct radeon_drm_bufmgr *)mgr;
+}
+
+static void
+radeon_drm_buffer_destroy(struct pb_buffer *_buf)
+{
+    struct radeon_drm_buffer *buf = radeon_drm_buffer(_buf);
+
+    if (buf->mapped) {
+	remove_from_list(buf);
+	radeon_bo_unmap(buf->bo);
+	buf->mapped = false;
+    }
+    radeon_bo_unref(buf->bo);
+
+    FREE(buf);
+}
+
+static void *
+radeon_drm_buffer_map(struct pb_buffer *_buf,
+		      unsigned flags)
+{
+    struct radeon_drm_buffer *buf = radeon_drm_buffer(_buf);
+    int write;
+
+    if (buf->mapped)
+	return buf->bo->ptr;
+    
+    if (flags & PIPE_BUFFER_USAGE_DONTBLOCK) {
+        uint32_t domain;
+
+        if (radeon_bo_is_busy(buf->bo, &domain))
+            return NULL;
+    }
+
+
+    if (radeon_bo_is_referenced_by_cs(buf->bo, buf->mgr->rws->cs)) {
+        buf->mgr->rws->flush_cb(buf->mgr->rws->flush_data);
+    }
+
+    if (flags & PIPE_BUFFER_USAGE_CPU_WRITE) {
+        write = 1;
+    }
+
+    if (radeon_bo_map(buf->bo, write)) {
+        return NULL;
+    }
+    buf->mapped = true;
+    insert_at_tail(&buf->mgr->buffer_map_list, buf);
+    return buf->bo->ptr;
+}
+
+static void
+radeon_drm_buffer_unmap(struct pb_buffer *_buf)
+{
+    (void)_buf;
+}
+
+static void
+radeon_drm_buffer_get_base_buffer(struct pb_buffer *buf,
+				  struct pb_buffer **base_buf,
+				  unsigned *offset)
+{
+    *base_buf = buf;
+    *offset = 0;
+}
+
+
+static enum pipe_error
+radeon_drm_buffer_validate(struct pb_buffer *_buf, 
+			   struct pb_validate *vl,
+			   unsigned flags)
+{
+   /* Always pinned */
+   return PIPE_OK;
+}
+
+static void
+radeon_drm_buffer_fence(struct pb_buffer *buf,
+			struct pipe_fence_handle *fence)
+{
+}
+
+const struct pb_vtbl radeon_drm_buffer_vtbl = {
+    radeon_drm_buffer_destroy,
+    radeon_drm_buffer_map,
+    radeon_drm_buffer_unmap,
+    radeon_drm_buffer_validate,
+    radeon_drm_buffer_fence,
+    radeon_drm_buffer_get_base_buffer,
+};
+
+
+static uint32_t radeon_domain_from_usage(unsigned usage)
+{
+    uint32_t domain = 0;
+
+    if (usage & PIPE_BUFFER_USAGE_GPU_WRITE) {
+        domain |= RADEON_GEM_DOMAIN_VRAM;
+    }
+    if (usage & PIPE_BUFFER_USAGE_PIXEL) {
+        domain |= RADEON_GEM_DOMAIN_VRAM;
+    }
+    if (usage & PIPE_BUFFER_USAGE_VERTEX) {
+        domain |= RADEON_GEM_DOMAIN_GTT;
+    }
+    if (usage & PIPE_BUFFER_USAGE_INDEX) {
+        domain |= RADEON_GEM_DOMAIN_GTT;
+    }
+
+    return domain;
+}
+
+struct pb_buffer *radeon_drm_bufmgr_create_buffer_from_handle(struct pb_manager *_mgr,
+							      uint32_t handle)
+{
+    struct radeon_drm_bufmgr *mgr = radeon_drm_bufmgr(_mgr);
+    struct radeon_libdrm_winsys *rws = mgr->rws;
+    struct radeon_drm_buffer *buf;
+    struct radeon_bo *bo;
+
+    bo = radeon_bo_open(rws->bom, handle, 0,
+			0, 0, 0);
+    if (bo == NULL)
+	return NULL;
+
+    buf = CALLOC_STRUCT(radeon_drm_buffer);
+    if (!buf) {
+	radeon_bo_unref(bo);
+	return NULL;
+    }
+
+    make_empty_list(buf);
+
+    pipe_reference_init(&buf->base.base.reference, 1);
+    buf->base.base.alignment = 0;
+    buf->base.base.usage = PIPE_BUFFER_USAGE_PIXEL;
+    buf->base.base.size = 0;
+    buf->base.vtbl = &radeon_drm_buffer_vtbl;
+    buf->mgr = mgr;
+
+    buf->bo = bo;
+
+    return &buf->base;
+}
+
+static struct pb_buffer *
+radeon_drm_bufmgr_create_buffer(struct pb_manager *_mgr,
+				pb_size size,
+				const struct pb_desc *desc)
+{
+    struct radeon_drm_bufmgr *mgr = radeon_drm_bufmgr(_mgr);
+    struct radeon_libdrm_winsys *rws = mgr->rws;
+    struct radeon_drm_buffer *buf;
+    uint32_t domain;
+
+    buf = CALLOC_STRUCT(radeon_drm_buffer);
+    if (!buf)
+	goto error1;
+
+    pipe_reference_init(&buf->base.base.reference, 1);
+    buf->base.base.alignment = desc->alignment;
+    buf->base.base.usage = desc->usage;
+    buf->base.base.size = size;
+    buf->base.vtbl = &radeon_drm_buffer_vtbl;
+    buf->mgr = mgr;
+
+    make_empty_list(buf);
+    domain = radeon_domain_from_usage(desc->usage);
+    buf->bo = radeon_bo_open(rws->bom, 0, size,
+			     desc->alignment, domain, 0);
+    if (buf->bo == NULL)
+	goto error2;
+
+    return &buf->base;
+
+ error2:
+    FREE(buf);
+ error1:
+    return NULL; 
+}
+
+static void
+radeon_drm_bufmgr_flush(struct pb_manager *mgr)
+{
+    /* NOP */
+}
+
+static void
+radeon_drm_bufmgr_destroy(struct pb_manager *_mgr)
+{
+    struct radeon_drm_bufmgr *mgr = radeon_drm_bufmgr(_mgr);
+    FREE(mgr);
+}
+
+struct pb_manager *
+radeon_drm_bufmgr_create(struct radeon_libdrm_winsys *rws)
+{
+    struct radeon_drm_bufmgr *mgr;
+
+    mgr = CALLOC_STRUCT(radeon_drm_bufmgr);
+    if (!mgr)
+	return NULL;
+
+    mgr->base.destroy = radeon_drm_bufmgr_destroy;
+    mgr->base.create_buffer = radeon_drm_bufmgr_create_buffer;
+    mgr->base.flush = radeon_drm_bufmgr_flush;
+
+    mgr->rws = rws;
+    make_empty_list(&mgr->buffer_map_list);
+    return &mgr->base;
+}
+
+static struct radeon_drm_buffer *get_drm_buffer(struct pb_buffer *_buf)
+{
+    struct radeon_drm_buffer *buf;
+    if (_buf->vtbl == &radeon_drm_buffer_vtbl) {
+        buf = radeon_drm_buffer(_buf);
+    } else {
+	struct pb_buffer *base_buf;
+	pb_size offset;
+	pb_get_base_buffer(_buf, &base_buf, &offset);
+
+	buf = radeon_drm_buffer(base_buf);
+    }
+    return buf;
+}
+
+boolean radeon_drm_bufmgr_get_handle(struct pb_buffer *_buf,
+				     struct winsys_handle *whandle)
+{
+    int retval, fd;
+    struct drm_gem_flink flink;
+    struct radeon_drm_buffer *buf = get_drm_buffer(_buf);
+    if (whandle->type == DRM_API_HANDLE_TYPE_SHARED) {
+	if (!buf->flinked) {
+	    fd = buf->mgr->rws->fd;
+	    flink.handle = buf->bo->handle;
+
+	    retval = ioctl(fd, DRM_IOCTL_GEM_FLINK, &flink);
+	    if (retval) {
+		return false;
+	    }
+
+	    buf->flinked = TRUE;
+	    buf->flink = flink.name;
+	}
+	whandle->handle = buf->flink;
+    } else if (whandle->type == DRM_API_HANDLE_TYPE_KMS) {
+	whandle->handle = buf->bo->handle;
+    }
+    return TRUE;
+}
+					   
+
+void radeon_drm_bufmgr_set_tiling(struct pb_buffer *_buf, boolean microtiled, boolean macrotiled, uint32_t pitch)
+{
+    struct radeon_drm_buffer *buf = get_drm_buffer(_buf);
+    uint32_t flags = 0;
+
+    if (microtiled)
+	flags |= RADEON_BO_FLAGS_MICRO_TILE;
+    if (macrotiled)
+	flags |= RADEON_BO_FLAGS_MACRO_TILE;
+
+    radeon_bo_set_tiling(buf->bo, flags, pitch);
+
+}
+
+boolean radeon_drm_bufmgr_add_buffer(struct pb_buffer *_buf,
+				     uint32_t rd, uint32_t wd)
+{
+    struct radeon_drm_buffer *buf = get_drm_buffer(_buf);
+    radeon_cs_space_add_persistent_bo(buf->mgr->rws->cs, buf->bo,
+					  rd, wd);
+    return true;
+}
+
+void radeon_drm_bufmgr_write_reloc(struct pb_buffer *_buf,
+				   uint32_t rd, uint32_t wd,
+				   uint32_t flags)
+{
+    struct radeon_drm_buffer *buf = get_drm_buffer(_buf);
+    int retval;
+
+    retval = radeon_cs_write_reloc(buf->mgr->rws->cs,
+				   buf->bo, rd, wd, flags);
+    if (retval) {
+        debug_printf("radeon: Relocation of %p (%d, %d, %d) failed!\n",
+		     buf, rd, wd, flags);
+    }
+}
+
+boolean radeon_drm_bufmgr_is_buffer_referenced(struct pb_buffer *_buf)
+{
+    struct radeon_drm_buffer *buf = get_drm_buffer(_buf);
+    uint32_t domain;
+
+    return (radeon_bo_is_referenced_by_cs(buf->bo, buf->mgr->rws->cs) ||
+	    radeon_bo_is_busy(buf->bo, &domain));
+}
+					    
+
+void radeon_drm_bufmgr_flush_maps(struct pb_manager *_mgr)
+{
+    struct radeon_drm_bufmgr *mgr = radeon_drm_bufmgr(_mgr);
+    struct radeon_drm_buffer *rpb, *t_rpb;
+
+    foreach_s(rpb, t_rpb, &mgr->buffer_map_list) {
+	rpb->mapped = 0;
+	radeon_bo_unmap(rpb->bo);
+	remove_from_list(rpb);
+    }
+
+    make_empty_list(&mgr->buffer_map_list);
+
+    
+}
diff --git a/src/gallium/winsys/drm/radeon/core/radeon_r300.c b/src/gallium/winsys/drm/radeon/core/radeon_r300.c
index 122bd21354..5b82a776a8 100644
--- a/src/gallium/winsys/drm/radeon/core/radeon_r300.c
+++ b/src/gallium/winsys/drm/radeon/core/radeon_r300.c
@@ -23,31 +23,139 @@
 #include "radeon_r300.h"
 #include "radeon_buffer.h"
 
+#include "radeon_bo_gem.h"
 #include "radeon_cs_gem.h"
+#include "state_tracker/drm_api.h"
 
-static void radeon_set_flush_cb(struct radeon_winsys *winsys,
+static struct r300_winsys_buffer *
+radeon_r300_winsys_buffer_create(struct r300_winsys_screen *rws,
+				 unsigned alignment,
+				 unsigned usage,
+				 unsigned size)
+{
+    struct radeon_libdrm_winsys *ws = radeon_winsys_screen(rws);
+    struct pb_desc desc;
+    struct pb_manager *provider;
+    struct pb_buffer *buffer;
+
+    memset(&desc, 0, sizeof(desc));
+    desc.alignment = alignment;
+    desc.usage = usage;
+
+    if (usage & PIPE_BUFFER_USAGE_CONSTANT)
+        provider = ws->mman;
+    else
+        provider = ws->kman;
+    buffer = provider->create_buffer(provider, size, &desc);
+    if (!buffer)
+	return NULL;
+
+    return radeon_libdrm_winsys_buffer(buffer);
+}
+
+static void radeon_r300_winsys_buffer_destroy(struct r300_winsys_buffer *buf)
+{
+    struct pb_buffer *_buf = radeon_pb_buffer(buf);
+
+    pb_destroy(_buf);
+}
+static void radeon_r300_winsys_buffer_set_tiling(struct r300_winsys_screen *rws,
+						  struct r300_winsys_buffer *buf,
+						  uint32_t pitch,
+						  boolean microtiled,
+						  boolean macrotiled)
+{
+    struct pb_buffer *_buf = radeon_pb_buffer(buf);
+    radeon_drm_bufmgr_set_tiling(_buf, microtiled, macrotiled, pitch);
+}
+
+static void *radeon_r300_winsys_buffer_map(struct r300_winsys_screen *ws,
+					   struct r300_winsys_buffer *buf,
+					   unsigned usage)
+{
+    struct pb_buffer *_buf = radeon_pb_buffer(buf);
+    
+    return pb_map(_buf, usage);
+}
+
+static void radeon_r300_winsys_buffer_unmap(struct r300_winsys_screen *ws,
+					    struct r300_winsys_buffer *buf)
+{
+    struct pb_buffer *_buf = radeon_pb_buffer(buf);
+
+    pb_unmap(_buf);
+}
+
+static void radeon_r300_winsys_buffer_reference(struct r300_winsys_screen *rws,
+						struct r300_winsys_buffer **pdst,
+						struct r300_winsys_buffer *src)
+{
+    struct pb_buffer *_src = radeon_pb_buffer(src);
+    struct pb_buffer *_dst = radeon_pb_buffer(*pdst);
+
+    pb_reference(&_dst, _src);
+
+    *pdst = radeon_libdrm_winsys_buffer(_dst);
+}
+
+static boolean radeon_r300_winsys_is_buffer_referenced(struct r300_winsys_screen *rws,
+						       struct r300_winsys_buffer *buf)
+{
+    struct pb_buffer *_buf = radeon_pb_buffer(buf);
+
+    return radeon_drm_bufmgr_is_buffer_referenced(_buf);
+}
+
+static struct r300_winsys_buffer *radeon_r300_winsys_buffer_from_handle(struct r300_winsys_screen *rws,
+									struct pipe_screen *screen,
+									struct winsys_handle *whandle,
+									unsigned *stride)
+{
+    struct radeon_libdrm_winsys *ws = radeon_winsys_screen(rws);
+    struct pb_buffer *_buf;
+
+    _buf = radeon_drm_bufmgr_create_buffer_from_handle(ws->kman, whandle->handle);
+    *stride = whandle->stride;
+    return radeon_libdrm_winsys_buffer(_buf);
+}
+
+static boolean radeon_r300_winsys_buffer_get_handle(struct r300_winsys_screen *rws,
+						    struct r300_winsys_buffer *buffer,
+						    unsigned stride,
+						    struct winsys_handle *whandle)
+{
+    struct pb_buffer *_buf = radeon_pb_buffer(buffer);
+    boolean ret;
+    ret = radeon_drm_bufmgr_get_handle(_buf, whandle);
+    if (ret)
+	whandle->stride = stride;
+    return ret;
+}
+
+static void radeon_set_flush_cb(struct r300_winsys_screen *rws,
                                 void (*flush_cb)(void *),
                                 void *data)
 {
-    winsys->priv->flush_cb = flush_cb;
-    winsys->priv->flush_data = data;
-    radeon_cs_space_set_flush(winsys->priv->cs, flush_cb, data);
+    struct radeon_libdrm_winsys *ws = radeon_winsys_screen(rws);
+    ws->flush_cb = flush_cb;
+    ws->flush_data = data;
+    radeon_cs_space_set_flush(ws->cs, flush_cb, data);
 }
 
-static boolean radeon_add_buffer(struct radeon_winsys* winsys,
-                                 struct pipe_buffer* pbuffer,
+static boolean radeon_add_buffer(struct r300_winsys_screen *rws,
+                                 struct r300_winsys_buffer *buf,
                                  uint32_t rd,
                                  uint32_t wd)
 {
-    struct radeon_bo* bo = ((struct radeon_pipe_buffer*)pbuffer)->bo;
+    struct pb_buffer *_buf = radeon_pb_buffer(buf);
 
-    radeon_cs_space_add_persistent_bo(winsys->priv->cs, bo, rd, wd);
-    return TRUE;
+    return radeon_drm_bufmgr_add_buffer(_buf, rd, wd);
 }
 
-static boolean radeon_validate(struct radeon_winsys* winsys)
+static boolean radeon_validate(struct r300_winsys_screen *rws)
 {
-    if (radeon_cs_space_check(winsys->priv->cs) < 0) {
+    struct radeon_libdrm_winsys *ws = radeon_winsys_screen(rws);
+    if (radeon_cs_space_check(ws->cs) < 0) {
         return FALSE;
     }
 
@@ -55,108 +163,175 @@ static boolean radeon_validate(struct radeon_winsys* winsys)
     return TRUE;
 }
 
-static boolean radeon_check_cs(struct radeon_winsys* winsys, int size)
+static boolean radeon_check_cs(struct r300_winsys_screen *rws, int size)
 {
-    struct radeon_cs* cs = winsys->priv->cs;
+    struct radeon_libdrm_winsys *ws = radeon_winsys_screen(rws);
+    struct radeon_cs *cs = ws->cs;
 
-    return radeon_validate(winsys) && cs->cdw + size <= cs->ndw;
+    return radeon_validate(rws) && cs->cdw + size <= cs->ndw;
 }
 
-static void radeon_begin_cs(struct radeon_winsys* winsys,
+static void radeon_begin_cs(struct r300_winsys_screen *rws,
                             int size,
                             const char* file,
                             const char* function,
                             int line)
 {
-    radeon_cs_begin(winsys->priv->cs, size, file, function, line);
+    struct radeon_libdrm_winsys *ws = radeon_winsys_screen(rws);
+    radeon_cs_begin(ws->cs, size, file, function, line);
 }
 
-static void radeon_write_cs_dword(struct radeon_winsys* winsys,
+static void radeon_write_cs_dword(struct r300_winsys_screen *rws,
                                   uint32_t dword)
 {
-    radeon_cs_write_dword(winsys->priv->cs, dword);
+    struct radeon_libdrm_winsys *ws = radeon_winsys_screen(rws);
+    radeon_cs_write_dword(ws->cs, dword);
 }
 
-static void radeon_write_cs_reloc(struct radeon_winsys* winsys,
-                                  struct pipe_buffer* pbuffer,
+static void radeon_write_cs_reloc(struct r300_winsys_screen *rws,
+                                  struct r300_winsys_buffer *buf,
                                   uint32_t rd,
                                   uint32_t wd,
                                   uint32_t flags)
 {
-    int retval = 0;
-    struct radeon_pipe_buffer* radeon_buffer =
-        (struct radeon_pipe_buffer*)pbuffer;
-
-    assert(!radeon_buffer->pb);
-
-    retval = radeon_cs_write_reloc(winsys->priv->cs, radeon_buffer->bo,
-                                   rd, wd, flags);
-
-    if (retval) {
-        debug_printf("radeon: Relocation of %p (%d, %d, %d) failed!\n",
-                pbuffer, rd, wd, flags);
-    }
+    struct pb_buffer *_buf = radeon_pb_buffer(buf);
+    radeon_drm_bufmgr_write_reloc(_buf, rd, wd, flags);
 }
 
-static void radeon_reset_bos(struct radeon_winsys *winsys)
+static void radeon_reset_bos(struct r300_winsys_screen *rws)
 {
-    radeon_cs_space_reset_bos(winsys->priv->cs);
+    struct radeon_libdrm_winsys *ws = radeon_winsys_screen(rws);
+    radeon_cs_space_reset_bos(ws->cs);
 }
 
-static void radeon_end_cs(struct radeon_winsys* winsys,
+static void radeon_end_cs(struct r300_winsys_screen *rws,
                           const char* file,
                           const char* function,
                           int line)
 {
-    radeon_cs_end(winsys->priv->cs, file, function, line);
+    struct radeon_libdrm_winsys *ws = radeon_winsys_screen(rws);
+    radeon_cs_end(ws->cs, file, function, line);
 }
 
-static void radeon_flush_cs(struct radeon_winsys* winsys)
+static void radeon_flush_cs(struct r300_winsys_screen *rws)
 {
+    struct radeon_libdrm_winsys *ws = radeon_winsys_screen(rws);
     int retval;
 
     /* Don't flush a zero-sized CS. */
-    if (!winsys->priv->cs->cdw) {
+    if (!ws->cs->cdw) {
         return;
     }
 
+    radeon_drm_bufmgr_flush_maps(ws->kman);
     /* Emit the CS. */
-    retval = radeon_cs_emit(winsys->priv->cs);
+    retval = radeon_cs_emit(ws->cs);
     if (retval) {
         debug_printf("radeon: Bad CS, dumping...\n");
-        radeon_cs_print(winsys->priv->cs, stderr);
+        radeon_cs_print(ws->cs, stderr);
     }
 
     /* Reset CS.
      * Someday, when we care about performance, we should really find a way
      * to rotate between two or three CS objects so that the GPU can be
      * spinning through one CS while another one is being filled. */
-    radeon_cs_erase(winsys->priv->cs);
+    radeon_cs_erase(ws->cs);
 }
 
-void
-radeon_setup_winsys(int fd, struct radeon_winsys* winsys)
+static uint32_t radeon_get_value(struct r300_winsys_screen *rws,
+			     enum r300_value_id id)
 {
-    struct radeon_winsys_priv* priv = winsys->priv;
+    struct radeon_libdrm_winsys *ws = (struct radeon_libdrm_winsys *)rws;
 
-    priv->csm = radeon_cs_manager_gem_ctor(fd);
+    switch(id) {
+    case R300_VID_PCI_ID:
+	return ws->pci_id;
+    case R300_VID_GB_PIPES:
+	return ws->gb_pipes;
+    case R300_VID_Z_PIPES:
+	return ws->z_pipes;
+    }
+    return 0;
+}
+
+static void
+radeon_winsys_destroy(struct r300_winsys_screen *rws)
+{
+    struct radeon_libdrm_winsys *ws = (struct radeon_libdrm_winsys *)rws;
+    radeon_cs_destroy(ws->cs);
+
+    ws->kman->destroy(ws->kman);
+    ws->mman->destroy(ws->mman);
+
+    radeon_bo_manager_gem_dtor(ws->bom);
+    radeon_cs_manager_gem_dtor(ws->csm);
+}
+
+boolean
+radeon_setup_winsys(int fd, struct radeon_libdrm_winsys* ws)
+{
+    
+    ws->csm = radeon_cs_manager_gem_ctor(fd);
+    if (!ws->csm)
+	goto fail;
+    ws->bom = radeon_bo_manager_gem_ctor(fd);
+    if (!ws->bom)
+	goto fail;
+    ws->kman = radeon_drm_bufmgr_create(ws);
+    if (!ws->kman)
+	goto fail;
+
+    ws->mman = pb_malloc_bufmgr_create();
+    if (!ws->mman)
+	goto fail;
 
     /* Size limit on IBs is 64 kibibytes. */
-    priv->cs = radeon_cs_create(priv->csm, 1024 * 64 / 4);
-    radeon_cs_set_limit(priv->cs,
-            RADEON_GEM_DOMAIN_GTT, winsys->gart_size);
-    radeon_cs_set_limit(priv->cs,
-            RADEON_GEM_DOMAIN_VRAM, winsys->vram_size);
-
-    winsys->add_buffer = radeon_add_buffer;
-    winsys->validate = radeon_validate;
-
-    winsys->check_cs = radeon_check_cs;
-    winsys->begin_cs = radeon_begin_cs;
-    winsys->write_cs_dword = radeon_write_cs_dword;
-    winsys->write_cs_reloc = radeon_write_cs_reloc;
-    winsys->end_cs = radeon_end_cs;
-    winsys->flush_cs = radeon_flush_cs;
-    winsys->reset_bos = radeon_reset_bos;
-    winsys->set_flush_cb = radeon_set_flush_cb;
+    ws->cs = radeon_cs_create(ws->csm, 1024 * 64 / 4);
+    if (!ws->cs)
+	goto fail;
+    radeon_cs_set_limit(ws->cs,
+            RADEON_GEM_DOMAIN_GTT, ws->gart_size);
+    radeon_cs_set_limit(ws->cs,
+            RADEON_GEM_DOMAIN_VRAM, ws->vram_size);
+
+    ws->base.add_buffer = radeon_add_buffer;
+    ws->base.validate = radeon_validate;
+    ws->base.destroy = radeon_winsys_destroy;
+    ws->base.check_cs = radeon_check_cs;
+    ws->base.begin_cs = radeon_begin_cs;
+    ws->base.write_cs_dword = radeon_write_cs_dword;
+    ws->base.write_cs_reloc = radeon_write_cs_reloc;
+    ws->base.end_cs = radeon_end_cs;
+    ws->base.flush_cs = radeon_flush_cs;
+    ws->base.reset_bos = radeon_reset_bos;
+    ws->base.set_flush_cb = radeon_set_flush_cb;
+    ws->base.get_value = radeon_get_value;
+
+    ws->base.buffer_create = radeon_r300_winsys_buffer_create;
+    ws->base.buffer_destroy = radeon_r300_winsys_buffer_destroy;
+    ws->base.buffer_set_tiling = radeon_r300_winsys_buffer_set_tiling;
+    ws->base.buffer_map = radeon_r300_winsys_buffer_map;
+    ws->base.buffer_unmap = radeon_r300_winsys_buffer_unmap;
+    ws->base.buffer_reference = radeon_r300_winsys_buffer_reference;
+    ws->base.buffer_from_handle = radeon_r300_winsys_buffer_from_handle;
+    ws->base.buffer_get_handle = radeon_r300_winsys_buffer_get_handle;
+    ws->base.is_buffer_referenced = radeon_r300_winsys_is_buffer_referenced;
+    return TRUE;
+
+fail:
+    if (ws->csm)
+	radeon_cs_manager_gem_dtor(ws->csm);
+
+    if (ws->bom)
+	radeon_bo_manager_gem_dtor(ws->bom);
+
+
+    if (ws->kman)
+	ws->kman->destroy(ws->kman);
+    if (ws->mman)
+	ws->mman->destroy(ws->mman);
+
+    if (ws->cs)
+	radeon_cs_destroy(ws->cs);
+    return FALSE;
 }
diff --git a/src/gallium/winsys/drm/radeon/core/radeon_r300.h b/src/gallium/winsys/drm/radeon/core/radeon_r300.h
index e655dc32c8..2703464ad8 100644
--- a/src/gallium/winsys/drm/radeon/core/radeon_r300.h
+++ b/src/gallium/winsys/drm/radeon/core/radeon_r300.h
@@ -25,6 +25,6 @@
 
 #include "radeon_winsys.h"
 
-void radeon_setup_winsys(int fd, struct radeon_winsys* winsys);
+boolean radeon_setup_winsys(int fd, struct radeon_libdrm_winsys* winsys);
 
 #endif /* RADEON_R300_H */
diff --git a/src/gallium/winsys/drm/radeon/core/radeon_winsys.h b/src/gallium/winsys/drm/radeon/core/radeon_winsys.h
index 37eeb45979..16cc701ad6 100644
--- a/src/gallium/winsys/drm/radeon/core/radeon_winsys.h
+++ b/src/gallium/winsys/drm/radeon/core/radeon_winsys.h
@@ -30,16 +30,15 @@
 #ifndef RADEON_WINSYS_H
 #define RADEON_WINSYS_H
 
-#include "util/u_simple_screen.h"
+#include "r300_winsys.h"
 
-struct radeon_winsys_priv;
-
-struct radeon_winsys {
+struct radeon_libdrm_winsys {
     /* Parent class. */
-    struct pipe_winsys base;
+    struct r300_winsys_screen base;
+
+    struct pb_manager *kman;
 
-    /* Winsys private */
-    struct radeon_winsys_priv* priv;
+    struct pb_manager *mman;
 
     /* PCI ID */
     uint32_t pci_id;
@@ -56,68 +55,27 @@ struct radeon_winsys {
     /* VRAM size. */
     uint32_t vram_size;
 
-    /* Create a buffer from a winsys handle. */
-    struct pipe_buffer *(*buffer_from_handle)(struct radeon_winsys *winsys,
-                                              struct pipe_screen *screen,
-                                              struct winsys_handle *whandle,
-                                              unsigned *stride);
-
-    /* Get the handle from a buffer. */
-    boolean (*buffer_get_handle)(struct radeon_winsys *winsys,
-                                 struct pipe_buffer *buffer,
-                                 unsigned stride,
-                                 struct winsys_handle *whandle);
-
-    /* Add a pipe_buffer to the list of buffer objects to validate. */
-    boolean (*add_buffer)(struct radeon_winsys* winsys,
-                          struct pipe_buffer* pbuffer,
-                          uint32_t rd,
-                          uint32_t wd);
-
-    /* Revalidate all currently setup pipe_buffers.
-     * Returns TRUE if a flush is required. */
-    boolean (*validate)(struct radeon_winsys* winsys);
-
-    /* Check to see if there's room for commands. */
-    boolean (*check_cs)(struct radeon_winsys* winsys, int size);
-
-    /* Start a command emit. */
-    void (*begin_cs)(struct radeon_winsys* winsys,
-                     int size,
-                     const char* file,
-                     const char* function,
-                     int line);
-
-    /* Write a dword to the command buffer. */
-    void (*write_cs_dword)(struct radeon_winsys* winsys, uint32_t dword);
-
-    /* Write a relocated dword to the command buffer. */
-    void (*write_cs_reloc)(struct radeon_winsys* winsys,
-                           struct pipe_buffer* bo,
-                           uint32_t rd,
-                           uint32_t wd,
-                           uint32_t flags);
-
-    /* Finish a command emit. */
-    void (*end_cs)(struct radeon_winsys* winsys,
-                   const char* file,
-                   const char* function,
-                   int line);
-
-    /* Flush the CS. */
-    void (*flush_cs)(struct radeon_winsys* winsys);
-
-    /* winsys flush - callback from winsys when flush required */
-    void (*set_flush_cb)(struct radeon_winsys *winsys,
-			 void (*flush_cb)(void *), void *data);
-
-    void (*reset_bos)(struct radeon_winsys *winsys);
-
-    void (*buffer_set_tiling)(struct radeon_winsys* winsys,
-                              struct pipe_buffer* buffer,
-                              uint32_t pitch,
-                              boolean microtiled,
-                              boolean macrotiled);
+    /* DRM FD */
+    int fd;
+
+    /* Radeon BO manager. */
+    struct radeon_bo_manager *bom;
+
+    /* Radeon CS manager. */
+    struct radeon_cs_manager *csm;
+
+    /* Current CS. */
+    struct radeon_cs *cs;
+
+    /* Flush CB */
+    void (*flush_cb)(void *);
+    void *flush_data;
 };
 
+static INLINE struct radeon_libdrm_winsys *
+radeon_winsys_screen(struct r300_winsys_screen *base)
+{
+  return (struct radeon_libdrm_winsys *)base;
+}
+
 #endif
diff --git a/src/gallium/winsys/xlib/Makefile b/src/gallium/winsys/xlib/Makefile
index 68542b488d..18357a4112 100644
--- a/src/gallium/winsys/xlib/Makefile
+++ b/src/gallium/winsys/xlib/Makefile
@@ -6,7 +6,8 @@ LIBNAME = ws_xlib
 LIBRARY_INCLUDES = \
 	-I$(TOP)/src/gallium/include \
 	-I$(TOP)/src/gallium/drivers \
-	-I$(TOP)/src/gallium/auxiliary
+	-I$(TOP)/src/gallium/auxiliary \
+	$(X_CFLAGS)
 
 C_SOURCES = \
 	xlib_sw_winsys.c 
diff --git a/src/mesa/SConscript b/src/mesa/SConscript
index e80ec5ee88..86da6f58bd 100644
--- a/src/mesa/SConscript
+++ b/src/mesa/SConscript
@@ -250,6 +250,7 @@ if env['platform'] != 'winddk':
 		'glapi/glapi.c',
 		'glapi/glapi_dispatch.c',
 		'glapi/glapi_entrypoint.c',
+		'glapi/glapi_execmem.c',
 		'glapi/glapi_getproc.c',
 		'glapi/glapi_nop.c',
 		'glapi/glthread.c',
diff --git a/src/mesa/drivers/dri/common/dri_sw.c b/src/mesa/drivers/dri/common/dri_sw.c
new file mode 100644
index 0000000000..b7f9036f47
--- /dev/null
+++ b/src/mesa/drivers/dri/common/dri_sw.c
@@ -0,0 +1,267 @@
+/*
+ * Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * Copyright 2010 George Sapountzis <gsapountzis@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * \file dri_sw.c
+ *
+ * DRISW utility functions, i.e. dri_util.c stripped from drm-specific bits.
+ */
+
+#include "dri_sw.h"
+#include "utils.h"
+
+
+/**
+ * Screen functions
+ */
+
+static void
+setupLoaderExtensions(__DRIscreen *psp,
+		      const __DRIextension **extensions)
+{
+    int i;
+
+    for (i = 0; extensions[i]; i++) {
+	if (strcmp(extensions[i]->name, __DRI_SWRAST_LOADER) == 0)
+	    psp->swrast_loader = (__DRIswrastLoaderExtension *) extensions[i];
+    }
+}
+
+static __DRIscreen *
+driCreateNewScreen(int scrn, const __DRIextension **extensions,
+		   const __DRIconfig ***driver_configs, void *data)
+{
+    static const __DRIextension *emptyExtensionList[] = { NULL };
+    __DRIscreen *psp;
+
+    (void) data;
+
+    psp = CALLOC_STRUCT(__DRIscreenRec);
+    if (!psp)
+	return NULL;
+
+    setupLoaderExtensions(psp, extensions);
+
+    psp->extensions = emptyExtensionList;
+    psp->myNum = scrn;
+
+    *driver_configs = driDriverAPI.InitScreen(psp);
+
+    if (*driver_configs == NULL) {
+	FREE(psp);
+	return NULL;
+    }
+
+    return psp;
+}
+
+static void driDestroyScreen(__DRIscreen *psp)
+{
+    if (psp) {
+	driDriverAPI.DestroyScreen(psp);
+
+	FREE(psp);
+    }
+}
+
+static const __DRIextension **driGetExtensions(__DRIscreen *psp)
+{
+    return psp->extensions;
+}
+
+
+/**
+ * Context functions
+ */
+
+static __DRIcontext *
+driCreateNewContext(__DRIscreen *psp, const __DRIconfig *config,
+		    __DRIcontext *shared, void *data)
+{
+    __DRIcontext *pcp;
+    void * const shareCtx = (shared != NULL) ? shared->driverPrivate : NULL;
+
+    pcp = CALLOC_STRUCT(__DRIcontextRec);
+    if (!pcp)
+	return NULL;
+
+    pcp->loaderPrivate = data;
+
+    pcp->driScreenPriv = psp;
+    pcp->driDrawablePriv = NULL;
+    pcp->driReadablePriv = NULL;
+
+    if (!driDriverAPI.CreateContext(&config->modes, pcp, shareCtx)) {
+	FREE(pcp);
+	return NULL;
+    }
+
+    return pcp;
+}
+
+static void
+driDestroyContext(__DRIcontext *pcp)
+{
+    if (pcp) {
+	driDriverAPI.DestroyContext(pcp);
+	FREE(pcp);
+    }
+}
+
+static int
+driCopyContext(__DRIcontext *dst, __DRIcontext *src, unsigned long mask)
+{
+    return GL_FALSE;
+}
+
+static void dri_get_drawable(__DRIdrawable *pdp);
+static void dri_put_drawable(__DRIdrawable *pdp);
+
+static int driBindContext(__DRIcontext *pcp,
+			  __DRIdrawable *pdp,
+			  __DRIdrawable *prp)
+{
+    /* Bind the drawable to the context */
+    if (pcp) {
+	pcp->driDrawablePriv = pdp;
+	pcp->driReadablePriv = prp;
+	if (pdp) {
+	    dri_get_drawable(pdp);
+	}
+	if ( prp && pdp != prp ) {
+	    dri_get_drawable(prp);
+	}
+    }
+
+    return driDriverAPI.MakeCurrent(pcp, pdp, prp);
+}
+
+static int driUnbindContext(__DRIcontext *pcp)
+{
+    __DRIdrawable *pdp;
+    __DRIdrawable *prp;
+
+    if (pcp == NULL)
+	return GL_FALSE;
+
+    pdp = pcp->driDrawablePriv;
+    prp = pcp->driReadablePriv;
+
+    /* already unbound */
+    if (!pdp && !prp)
+	return GL_TRUE;
+
+    driDriverAPI.UnbindContext(pcp);
+
+    dri_put_drawable(pdp);
+
+    if (prp != pdp) {
+	dri_put_drawable(prp);
+    }
+
+    pcp->driDrawablePriv = NULL;
+    pcp->driReadablePriv = NULL;
+
+    return GL_TRUE;
+}
+
+
+/**
+ * Drawable functions
+ */
+
+static void dri_get_drawable(__DRIdrawable *pdp)
+{
+    pdp->refcount++;
+}
+
+static void dri_put_drawable(__DRIdrawable *pdp)
+{
+    if (pdp) {
+	pdp->refcount--;
+	if (pdp->refcount)
+	    return;
+
+	driDriverAPI.DestroyBuffer(pdp);
+
+	FREE(pdp);
+    }
+}
+
+static __DRIdrawable *
+driCreateNewDrawable(__DRIscreen *psp,
+		     const __DRIconfig *config, void *data)
+{
+    __DRIdrawable *pdp;
+
+    pdp = CALLOC_STRUCT(__DRIdrawableRec);
+    if (!pdp)
+	return NULL;
+
+    pdp->loaderPrivate = data;
+
+    pdp->driScreenPriv = psp;
+
+    dri_get_drawable(pdp);
+
+    if (!driDriverAPI.CreateBuffer(psp, pdp, &config->modes, GL_FALSE)) {
+	FREE(pdp);
+	return NULL;
+    }
+
+    return pdp;
+}
+
+static void
+driDestroyDrawable(__DRIdrawable *pdp)
+{
+    dri_put_drawable(pdp);
+}
+
+static void driSwapBuffers(__DRIdrawable *pdp)
+{
+    driDriverAPI.SwapBuffers(pdp);
+}
+
+const __DRIcoreExtension driCoreExtension = {
+    { __DRI_CORE, __DRI_CORE_VERSION },
+    NULL, /* driCreateNewScreen */
+    driDestroyScreen,
+    driGetExtensions,
+    driGetConfigAttrib,
+    driIndexConfigAttrib,
+    NULL, /* driCreateNewDrawable */
+    driDestroyDrawable,
+    driSwapBuffers,
+    driCreateNewContext,
+    driCopyContext,
+    driDestroyContext,
+    driBindContext,
+    driUnbindContext
+};
+
+const __DRIswrastExtension driSWRastExtension = {
+    { __DRI_SWRAST, __DRI_SWRAST_VERSION },
+    driCreateNewScreen,
+    driCreateNewDrawable
+};
diff --git a/src/mesa/drivers/dri/common/dri_sw.h b/src/mesa/drivers/dri/common/dri_sw.h
new file mode 100644
index 0000000000..e353e26b34
--- /dev/null
+++ b/src/mesa/drivers/dri/common/dri_sw.h
@@ -0,0 +1,112 @@
+/*
+ * Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * Copyright 2010 George Sapountzis <gsapountzis@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+
+#ifndef _DRI_SW_H
+#define _DRI_SW_H
+
+#include <GL/gl.h>
+#include <GL/internal/glcore.h>
+#include <GL/internal/dri_interface.h>
+
+
+/**
+ * Extensions
+ */
+extern const __DRIcoreExtension driCoreExtension;
+extern const __DRIswrastExtension driSWRastExtension;
+
+
+/**
+ * Data types
+ */
+struct __DRIscreenRec {
+    int myNum;
+
+    int fd;
+
+    void *private;
+
+    const __DRIextension **extensions;
+
+    const __DRIswrastLoaderExtension *swrast_loader;
+};
+
+struct __DRIcontextRec {
+
+    void *driverPrivate;
+
+    void *loaderPrivate;
+
+    __DRIdrawable *driDrawablePriv;
+
+    __DRIdrawable *driReadablePriv;
+
+    __DRIscreen *driScreenPriv;
+};
+
+struct __DRIdrawableRec {
+
+    void *driverPrivate;
+
+    void *loaderPrivate;
+
+    __DRIscreen *driScreenPriv;
+
+    int refcount;
+};
+
+
+/**
+ * Driver callback functions
+ */
+struct __DriverAPIRec {
+    const __DRIconfig **(*InitScreen) (__DRIscreen * priv);
+
+    void (*DestroyScreen)(__DRIscreen *driScrnPriv);
+
+    GLboolean (*CreateContext)(const __GLcontextModes *glVis,
+                               __DRIcontext *driContextPriv,
+                               void *sharedContextPrivate);
+
+    void (*DestroyContext)(__DRIcontext *driContextPriv);
+
+    GLboolean (*CreateBuffer)(__DRIscreen *driScrnPriv,
+                              __DRIdrawable *driDrawPriv,
+                              const __GLcontextModes *glVis,
+                              GLboolean pixmapBuffer);
+
+    void (*DestroyBuffer)(__DRIdrawable *driDrawPriv);
+
+    void (*SwapBuffers)(__DRIdrawable *driDrawPriv);
+
+    GLboolean (*MakeCurrent)(__DRIcontext *driContextPriv,
+                             __DRIdrawable *driDrawPriv,
+                             __DRIdrawable *driReadPriv);
+
+    GLboolean (*UnbindContext)(__DRIcontext *driContextPriv);
+};
+
+extern const struct __DriverAPIRec driDriverAPI;
+
+#endif /* _DRI_SW_H */
diff --git a/src/mesa/drivers/dri/common/dri_util.c b/src/mesa/drivers/dri/common/dri_util.c
index 75c98825b7..badbb5ff82 100644
--- a/src/mesa/drivers/dri/common/dri_util.c
+++ b/src/mesa/drivers/dri/common/dri_util.c
@@ -47,28 +47,6 @@ const __DRIextension driReadDrawableExtension = {
     __DRI_READ_DRAWABLE, __DRI_READ_DRAWABLE_VERSION
 };
 
-/**
- * Print message to \c stderr if the \c LIBGL_DEBUG environment variable
- * is set. 
- * 
- * Is called from the drivers.
- * 
- * \param f \c printf like format string.
- */
-void
-__driUtilMessage(const char *f, ...)
-{
-    va_list args;
-
-    if (getenv("LIBGL_DEBUG")) {
-        fprintf(stderr, "libGL: ");
-        va_start(args, f);
-        vfprintf(stderr, f, args);
-        va_end(args);
-        fprintf(stderr, "\n");
-    }
-}
-
 GLint
 driIntersectArea( drm_clip_rect_t rect1, drm_clip_rect_t rect2 )
 {
diff --git a/src/mesa/drivers/dri/common/dri_util.h b/src/mesa/drivers/dri/common/dri_util.h
index 99c0f1e442..f63583cebc 100644
--- a/src/mesa/drivers/dri/common/dri_util.h
+++ b/src/mesa/drivers/dri/common/dri_util.h
@@ -552,10 +552,6 @@ struct __DRIscreenRec {
 };
 
 extern void
-__driUtilMessage(const char *f, ...);
-
-
-extern void
 __driUtilUpdateDrawableInfo(__DRIdrawable *pdp);
 
 extern float
diff --git a/src/mesa/drivers/dri/common/utils.c b/src/mesa/drivers/dri/common/utils.c
index b85b364c57..0dd879abc9 100644
--- a/src/mesa/drivers/dri/common/utils.c
+++ b/src/mesa/drivers/dri/common/utils.c
@@ -37,6 +37,29 @@
 #include "utils.h"
 
 
+/**
+ * Print message to \c stderr if the \c LIBGL_DEBUG environment variable
+ * is set. 
+ * 
+ * Is called from the drivers.
+ * 
+ * \param f \c printf like format string.
+ */
+void
+__driUtilMessage(const char *f, ...)
+{
+    va_list args;
+
+    if (getenv("LIBGL_DEBUG")) {
+        fprintf(stderr, "libGL: ");
+        va_start(args, f);
+        vfprintf(stderr, f, args);
+        va_end(args);
+        fprintf(stderr, "\n");
+    }
+}
+
+
 unsigned
 driParseDebugString( const char * debug, 
 		     const struct dri_debug_control * control  )
@@ -230,9 +253,6 @@ void driInitSingleExtension( GLcontext * ctx,
 /**
  * Utility function used by drivers to test the verions of other components.
  *
- * If one of the version requirements is not met, a message is logged using
- * \c __driUtilMessage.
- *
  * \param driver_name  Name of the driver.  Used in error messages.
  * \param driActual    Actual DRI version supplied __driCreateNewScreen.
  * \param driExpected  Minimum DRI version required by the driver.
@@ -244,7 +264,7 @@ void driInitSingleExtension( GLcontext * ctx,
  * \returns \c GL_TRUE if all version requirements are met.  Otherwise,
  *          \c GL_FALSE is returned.
  * 
- * \sa __driCreateNewScreen, driCheckDriDdxDrmVersions2, __driUtilMessage
+ * \sa __driCreateNewScreen, driCheckDriDdxDrmVersions2
  *
  * \todo
  * Now that the old \c driCheckDriDdxDrmVersions function is gone, this
diff --git a/src/mesa/drivers/dri/common/utils.h b/src/mesa/drivers/dri/common/utils.h
index 02ca3feb73..de6070c398 100644
--- a/src/mesa/drivers/dri/common/utils.h
+++ b/src/mesa/drivers/dri/common/utils.h
@@ -69,6 +69,9 @@ struct __DRIutilversionRec2 {
     int    patch;        /**< Patch-level. */
 };
 
+extern void
+__driUtilMessage(const char *f, ...);
+
 extern unsigned driParseDebugString( const char * debug,
     const struct dri_debug_control * control );
 
diff --git a/src/mesa/drivers/dri/common/xmlconfig.c b/src/mesa/drivers/dri/common/xmlconfig.c
index 477259ea7e..de4500a39b 100644
--- a/src/mesa/drivers/dri/common/xmlconfig.c
+++ b/src/mesa/drivers/dri/common/xmlconfig.c
@@ -36,7 +36,7 @@
 #include <unistd.h>
 #include <errno.h>
 #include "main/imports.h"
-#include "dri_util.h"
+#include "utils.h"
 #include "xmlconfig.h"
 
 #undef GET_PROGRAM_NAME
diff --git a/src/mesa/drivers/dri/mach64/mach64_context.c b/src/mesa/drivers/dri/mach64/mach64_context.c
index 77e7e53ce0..73b1e08d4b 100644
--- a/src/mesa/drivers/dri/mach64/mach64_context.c
+++ b/src/mesa/drivers/dri/mach64/mach64_context.c
@@ -31,6 +31,7 @@
 
 #include "main/glheader.h"
 #include "main/context.h"
+#include "main/extensions.h"
 #include "main/simple_list.h"
 #include "main/imports.h"
 
diff --git a/src/mesa/drivers/dri/r300/r300_blit.c b/src/mesa/drivers/dri/r300/r300_blit.c
index d870c7f852..fa60628a5e 100644
--- a/src/mesa/drivers/dri/r300/r300_blit.c
+++ b/src/mesa/drivers/dri/r300/r300_blit.c
@@ -582,12 +582,6 @@ unsigned r300_blit(GLcontext *ctx,
     if (dst_pitch % 2 > 0)
         ++dst_pitch;
 
-    /* Rendering to small buffer doesn't work.
-     * Looks like a hw limitation.
-     */
-    if (dst_pitch < 32)
-        return 0;
-
     /* Need to clamp the region size to make sure
      * we don't read outside of the source buffer
      * or write outside of the destination buffer.
diff --git a/src/mesa/drivers/dri/radeon/radeon_pixel_read.c b/src/mesa/drivers/dri/radeon/radeon_pixel_read.c
index b180c1d9a5..dadb8002c7 100644
--- a/src/mesa/drivers/dri/radeon/radeon_pixel_read.c
+++ b/src/mesa/drivers/dri/radeon/radeon_pixel_read.c
@@ -50,22 +50,33 @@ static gl_format gl_format_and_type_to_mesa_format(GLenum format, GLenum type)
             break;
         case GL_RGBA:
             switch (type) {
-                case GL_UNSIGNED_BYTE:
-                    return MESA_FORMAT_RGBA8888_REV;
                 case GL_FLOAT:
                     return MESA_FORMAT_RGBA_FLOAT32;
+                case GL_UNSIGNED_SHORT_5_5_5_1:
+                    return MESA_FORMAT_RGBA5551;
+                case GL_UNSIGNED_INT_8_8_8_8:
+                    return MESA_FORMAT_RGBA8888;
+                case GL_UNSIGNED_BYTE:
+                case GL_UNSIGNED_INT_8_8_8_8_REV:
+                    return MESA_FORMAT_RGBA8888_REV;
+            }
+            break;
+        case GL_BGRA:
+            switch (type) {
                 case GL_UNSIGNED_SHORT_4_4_4_4:
-                    return MESA_FORMAT_ARGB4444;
+                    return MESA_FORMAT_ARGB4444_REV;
                 case GL_UNSIGNED_SHORT_4_4_4_4_REV:
                     return MESA_FORMAT_ARGB4444;
                 case GL_UNSIGNED_SHORT_5_5_5_1:
-                    return MESA_FORMAT_RGBA5551;
-                case GL_UNSIGNED_SHORT_1_5_5_5_REV:
                     return MESA_FORMAT_ARGB1555_REV;
+                case GL_UNSIGNED_SHORT_1_5_5_5_REV:
+                    return MESA_FORMAT_ARGB1555;
                 case GL_UNSIGNED_INT_8_8_8_8:
-                    return MESA_FORMAT_ARGB8888;
-                case GL_UNSIGNED_INT_8_8_8_8_REV:
                     return MESA_FORMAT_ARGB8888_REV;
+                case GL_UNSIGNED_BYTE:
+                case GL_UNSIGNED_INT_8_8_8_8_REV:
+                    return MESA_FORMAT_ARGB8888;
+
             }
             break;
     }
diff --git a/src/mesa/drivers/dri/radeon/radeon_tex_copy.c b/src/mesa/drivers/dri/radeon/radeon_tex_copy.c
index e57d77e7ef..29fd31ac23 100644
--- a/src/mesa/drivers/dri/radeon/radeon_tex_copy.c
+++ b/src/mesa/drivers/dri/radeon/radeon_tex_copy.c
@@ -28,6 +28,7 @@
 #include "radeon_common.h"
 #include "radeon_texture.h"
 
+#include "main/enums.h"
 #include "main/image.h"
 #include "main/teximage.h"
 #include "main/texstate.h"
@@ -59,18 +60,27 @@ do_copy_texsubimage(GLcontext *ctx,
     }
 
     if (_mesa_get_format_bits(timg->base.TexFormat, GL_DEPTH_BITS) > 0) {
-        rrb = radeon_renderbuffer(ctx->ReadBuffer->_DepthBuffer);
+        if (ctx->ReadBuffer->_DepthBuffer && ctx->ReadBuffer->_DepthBuffer->Wrapped) {
+            rrb = radeon_renderbuffer(ctx->ReadBuffer->_DepthBuffer->Wrapped);
+        } else {
+            rrb = radeon_renderbuffer(ctx->ReadBuffer->_DepthBuffer);
+        }
         flip_y = ctx->ReadBuffer->Attachment[BUFFER_DEPTH].Type == GL_NONE;
     } else {
         rrb = radeon_renderbuffer(ctx->ReadBuffer->_ColorReadBuffer);
         flip_y = ctx->ReadBuffer->Attachment[BUFFER_COLOR0].Type == GL_NONE;
     }
 
+    // This is software renderbuffer, fallback to swrast
+    if (!rrb) {
+        return GL_FALSE;
+    }
+
     if (!timg->mt) {
         radeon_validate_texture_miptree(ctx, &tobj->base);
     }
 
-    assert(rrb && rrb->bo);
+    assert(rrb->bo);
     assert(timg->mt);
     assert(timg->mt->bo);
     assert(timg->base.Width >= dstx + width);
@@ -174,6 +184,10 @@ radeonCopyTexImage2D(GLcontext *ctx, GLenum target, GLint level,
     return;
 
 fail:
+    radeon_print(RADEON_FALLBACKS, RADEON_NORMAL,
+                 "Falling back to sw for glCopyTexImage2D (internalFormat %s, border %d)\n",
+                 _mesa_lookup_enum_by_nr(internalFormat), border);
+
     _mesa_meta_CopyTexImage2D(ctx, target, level, internalFormat, x, y,
                               width, height, border);
 }
@@ -192,7 +206,8 @@ radeonCopyTexSubImage2D(GLcontext *ctx, GLenum target, GLint level,
                              radeon_tex_obj(texObj), (radeon_texture_image *)texImage,
                              xoffset, yoffset, x, y, width, height)) {
 
-       //DEBUG_FALLBACKS
+        radeon_print(RADEON_FALLBACKS, RADEON_NORMAL,
+                     "Falling back to sw for glCopyTexSubImage2D\n");
 
         _mesa_meta_CopyTexSubImage2D(ctx, target, level,
                                      xoffset, yoffset, x, y, width, height);
diff --git a/src/mesa/drivers/dri/swrast/Makefile b/src/mesa/drivers/dri/swrast/Makefile
index 771169c1ff..cc59eefdb2 100644
--- a/src/mesa/drivers/dri/swrast/Makefile
+++ b/src/mesa/drivers/dri/swrast/Makefile
@@ -17,7 +17,8 @@ ASM_SOURCES =
 
 SWRAST_COMMON_SOURCES = \
 	../../common/driverfuncs.c \
-	../common/utils.c
+	../common/utils.c \
+	../common/dri_sw.c
 
 include ../Makefile.template
 
diff --git a/src/mesa/drivers/dri/swrast/swrast.c b/src/mesa/drivers/dri/swrast/swrast.c
index 03c672ecf1..e9ca99a86f 100644
--- a/src/mesa/drivers/dri/swrast/swrast.c
+++ b/src/mesa/drivers/dri/swrast/swrast.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (C) 2008 George Sapountzis <gsap7@yahoo.gr>
+ * Copyright 2008, 2010 George Sapountzis <gsapountzis@gmail.com>
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -54,17 +54,9 @@
  * Screen and config-related functions
  */
 
-static void
-setupLoaderExtensions(__DRIscreen *psp,
-		      const __DRIextension **extensions)
-{
-    int i;
-
-    for (i = 0; extensions[i]; i++) {
-	if (strcmp(extensions[i]->name, __DRI_SWRAST_LOADER) == 0)
-	    psp->swrast_loader = (__DRIswrastLoaderExtension *) extensions[i];
-    }
-}
+static const __DRIextension *dri_screen_extensions[] = {
+    NULL
+};
 
 static __DRIconfig **
 swrastFillInModes(__DRIscreen *psp,
@@ -143,26 +135,14 @@ swrastFillInModes(__DRIscreen *psp,
     return configs;
 }
 
-static __DRIscreen *
-driCreateNewScreen(int scrn, const __DRIextension **extensions,
-		   const __DRIconfig ***driver_configs, void *data)
+static const __DRIconfig **
+dri_init_screen(__DRIscreen * psp)
 {
-    static const __DRIextension *emptyExtensionList[] = { NULL };
-    __DRIscreen *psp;
     __DRIconfig **configs8, **configs16, **configs24, **configs32;
 
-    (void) data;
-
     TRACE;
 
-    psp = calloc(1, sizeof(*psp));
-    if (!psp)
-	return NULL;
-
-    setupLoaderExtensions(psp, extensions);
-
-    psp->num = scrn;
-    psp->extensions = emptyExtensionList;
+    psp->extensions = dri_screen_extensions;
 
     configs8  = swrastFillInModes(psp,  8,  8, 0, 1);
     configs16 = swrastFillInModes(psp, 16, 16, 0, 1);
@@ -171,28 +151,15 @@ driCreateNewScreen(int scrn, const __DRIextension **extensions,
 
     configs16 = driConcatConfigs(configs8, configs16);
     configs24 = driConcatConfigs(configs16, configs24);
-    *driver_configs = (const __DRIconfig **)
-       driConcatConfigs(configs24, configs32);
-
-    driInitExtensions( NULL, NULL, GL_FALSE );
-
-    return psp;
-}
-
-static void driDestroyScreen(__DRIscreen *psp)
-{
-    TRACE;
+    configs32 = driConcatConfigs(configs24, configs32);
 
-    if (psp) {
-	free(psp);
-    }
+    return (const __DRIconfig **)configs32;
 }
 
-static const __DRIextension **driGetExtensions(__DRIscreen *psp)
+static void
+dri_destroy_screen(__DRIscreen * sPriv)
 {
     TRACE;
-
-    return psp->extensions;
 }
 
 
@@ -336,94 +303,116 @@ swrast_new_renderbuffer(const GLvisual *visual, GLboolean front)
     return xrb;
 }
 
-static __DRIdrawable *
-driCreateNewDrawable(__DRIscreen *screen,
-		     const __DRIconfig *config, void *data)
+static GLboolean
+dri_create_buffer(__DRIscreen * sPriv,
+		  __DRIdrawable * dPriv,
+		  const __GLcontextModes * visual, GLboolean isPixmap)
 {
-    __DRIdrawable *buf;
+    struct dri_drawable *drawable = NULL;
+    GLframebuffer *fb;
     struct swrast_renderbuffer *frontrb, *backrb;
 
     TRACE;
 
-    buf = calloc(1, sizeof *buf);
-    if (!buf)
-	return NULL;
+    drawable = CALLOC_STRUCT(dri_drawable);
+    if (drawable == NULL)
+	goto drawable_fail;
 
-    buf->loaderPrivate = data;
+    dPriv->driverPrivate = drawable;
+    drawable->dPriv = dPriv;
 
-    buf->driScreenPriv = screen;
+    drawable->row = malloc(MAX_WIDTH * 4);
+    if (drawable->row == NULL)
+	goto drawable_fail;
 
-    buf->row = malloc(MAX_WIDTH * 4);
+    fb = &drawable->Base;
 
     /* basic framebuffer setup */
-    _mesa_initialize_window_framebuffer(&buf->Base, &config->modes);
+    _mesa_initialize_window_framebuffer(fb, visual);
 
     /* add front renderbuffer */
-    frontrb = swrast_new_renderbuffer(&config->modes, GL_TRUE);
-    _mesa_add_renderbuffer(&buf->Base, BUFFER_FRONT_LEFT, &frontrb->Base);
+    frontrb = swrast_new_renderbuffer(visual, GL_TRUE);
+    _mesa_add_renderbuffer(fb, BUFFER_FRONT_LEFT, &frontrb->Base);
 
     /* add back renderbuffer */
-    if (config->modes.doubleBufferMode) {
-	backrb = swrast_new_renderbuffer(&config->modes, GL_FALSE);
-	_mesa_add_renderbuffer(&buf->Base, BUFFER_BACK_LEFT, &backrb->Base);
+    if (visual->doubleBufferMode) {
+	backrb = swrast_new_renderbuffer(visual, GL_FALSE);
+	_mesa_add_renderbuffer(fb, BUFFER_BACK_LEFT, &backrb->Base);
     }
 
     /* add software renderbuffers */
-    _mesa_add_soft_renderbuffers(&buf->Base,
+    _mesa_add_soft_renderbuffers(fb,
 				 GL_FALSE, /* color */
-				 config->modes.haveDepthBuffer,
-				 config->modes.haveStencilBuffer,
-				 config->modes.haveAccumBuffer,
+				 visual->haveDepthBuffer,
+				 visual->haveStencilBuffer,
+				 visual->haveAccumBuffer,
 				 GL_FALSE, /* alpha */
 				 GL_FALSE /* aux bufs */);
 
-    return buf;
+    return GL_TRUE;
+
+drawable_fail:
+
+    if (drawable)
+	free(drawable->row);
+
+    FREE(drawable);
+
+    return GL_FALSE;
 }
 
 static void
-driDestroyDrawable(__DRIdrawable *buf)
+dri_destroy_buffer(__DRIdrawable * dPriv)
 {
     TRACE;
 
-    if (buf) {
-	struct gl_framebuffer *fb = &buf->Base;
+    if (dPriv) {
+	struct dri_drawable *drawable = dri_drawable(dPriv);
+	GLframebuffer *fb;
 
-	free(buf->row);
+	free(drawable->row);
+
+	fb = &drawable->Base;
 
 	fb->DeletePending = GL_TRUE;
 	_mesa_reference_framebuffer(&fb, NULL);
     }
 }
 
-static void driSwapBuffers(__DRIdrawable *buf)
+static void
+dri_swap_buffers(__DRIdrawable * dPriv)
 {
-    GET_CURRENT_CONTEXT(ctx);
+    __DRIscreen *sPriv = dPriv->driScreenPriv;
 
-    struct swrast_renderbuffer *frontrb =
-	swrast_renderbuffer(buf->Base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer);
-    struct swrast_renderbuffer *backrb =
-	swrast_renderbuffer(buf->Base.Attachment[BUFFER_BACK_LEFT].Renderbuffer);
+    GET_CURRENT_CONTEXT(ctx);
 
-    __DRIscreen *screen = buf->driScreenPriv;
+    struct dri_drawable *drawable = dri_drawable(dPriv);
+    GLframebuffer *fb;
+    struct swrast_renderbuffer *frontrb, *backrb;
 
     TRACE;
 
+    fb = &drawable->Base;
+
+    frontrb = swrast_renderbuffer(fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer);
+    backrb = swrast_renderbuffer(fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer);
+
     /* check for signle-buffered */
     if (backrb == NULL)
 	return;
 
     /* check if swapping currently bound buffer */
-    if (ctx && ctx->DrawBuffer == &(buf->Base)) {
+    if (ctx && ctx->DrawBuffer == fb) {
 	/* flush pending rendering */
 	_mesa_notifySwapBuffers(ctx);
     }
 
-    screen->swrast_loader->putImage(buf, __DRI_SWRAST_IMAGE_OP_SWAP,
-				    0, 0,
-				    frontrb->Base.Width,
-				    frontrb->Base.Height,
-				    backrb->Base.Data,
-				    buf->loaderPrivate);
+    sPriv->swrast_loader->putImage(dPriv, __DRI_SWRAST_IMAGE_OP_SWAP,
+				   0, 0,
+				   frontrb->Base.Width,
+				   frontrb->Base.Height,
+				   backrb->Base.Data,
+				   dPriv->loaderPrivate);
 }
 
 
@@ -434,13 +423,13 @@ static void driSwapBuffers(__DRIdrawable *buf)
 static void
 get_window_size( GLframebuffer *fb, GLsizei *w, GLsizei *h )
 {
-    __DRIdrawable *buf = swrast_drawable(fb);
-    __DRIscreen *screen = buf->driScreenPriv;
+    __DRIdrawable *dPriv = swrast_drawable(fb)->dPriv;
+    __DRIscreen *sPriv = dPriv->driScreenPriv;
     int x, y;
 
-    screen->swrast_loader->getDrawableInfo(buf,
-					   &x, &y, w, h,
-					   buf->loaderPrivate);
+    sPriv->swrast_loader->getDrawableInfo(dPriv,
+					  &x, &y, w, h,
+					  dPriv->loaderPrivate);
 }
 
 static void
@@ -502,37 +491,40 @@ swrast_init_driver_functions(struct dd_function_table *driver)
  * Context-related functions.
  */
 
-static __DRIcontext *
-driCreateNewContext(__DRIscreen *screen, const __DRIconfig *config,
-		    __DRIcontext *shared, void *data)
+static GLboolean
+dri_create_context(const __GLcontextModes * visual,
+		   __DRIcontext * cPriv, void *sharedContextPrivate)
 {
-    __DRIcontext *ctx;
-    GLcontext *mesaCtx;
+    struct dri_context *ctx = NULL;
+    struct dri_context *share = (struct dri_context *)sharedContextPrivate;
+    GLcontext *mesaCtx = NULL;
+    GLcontext *sharedCtx = NULL;
     struct dd_function_table functions;
 
     TRACE;
 
-    ctx = calloc(1, sizeof *ctx);
-    if (!ctx)
-	return NULL;
-
-    ctx->loaderPrivate = data;
+    ctx = CALLOC_STRUCT(dri_context);
+    if (ctx == NULL)
+	goto context_fail;
 
-    ctx->driScreenPriv = screen;
+    cPriv->driverPrivate = ctx;
+    ctx->cPriv = cPriv;
 
     /* build table of device driver functions */
     _mesa_init_driver_functions(&functions);
     swrast_init_driver_functions(&functions);
 
-    if (!_mesa_initialize_context(&ctx->Base, &config->modes,
-				  shared ? &shared->Base : NULL,
-				  &functions, (void *) ctx)) {
-      free(ctx);
-      return NULL;
+    if (share) {
+	sharedCtx = &share->Base;
     }
 
     mesaCtx = &ctx->Base;
 
+    /* basic context setup */
+    if (!_mesa_initialize_context(mesaCtx, visual, sharedCtx, &functions, (void *) cPriv)) {
+	goto context_fail;
+    }
+
     /* do bounds checking to prevent segfaults and server crashes! */
     mesaCtx->Const.CheckArrayBounds = GL_TRUE;
 
@@ -558,17 +550,28 @@ driCreateNewContext(__DRIscreen *screen, const __DRIconfig *config,
 
     _mesa_meta_init(mesaCtx);
 
-    return ctx;
+    driInitExtensions( mesaCtx, NULL, GL_FALSE );
+
+    return GL_TRUE;
+
+context_fail:
+
+    FREE(ctx);
+
+    return GL_FALSE;
 }
 
 static void
-driDestroyContext(__DRIcontext *ctx)
+dri_destroy_context(__DRIcontext * cPriv)
 {
-    GLcontext *mesaCtx;
     TRACE;
 
-    if (ctx) {
+    if (cPriv) {
+	struct dri_context *ctx = dri_context(cPriv);
+	GLcontext *mesaCtx;
+
 	mesaCtx = &ctx->Base;
+
         _mesa_meta_free(mesaCtx);
 	_swsetup_DestroyContext( mesaCtx );
 	_swrast_DestroyContext( mesaCtx );
@@ -578,26 +581,22 @@ driDestroyContext(__DRIcontext *ctx)
     }
 }
 
-static int
-driCopyContext(__DRIcontext *dst, __DRIcontext *src, unsigned long mask)
-{
-    TRACE;
-
-    _mesa_copy_context(&src->Base, &dst->Base, mask);
-    return GL_TRUE;
-}
-
-static int driBindContext(__DRIcontext *ctx,
-			  __DRIdrawable *draw,
-			  __DRIdrawable *read)
+static GLboolean
+dri_make_current(__DRIcontext * cPriv,
+		 __DRIdrawable * driDrawPriv,
+		 __DRIdrawable * driReadPriv)
 {
     GLcontext *mesaCtx;
     GLframebuffer *mesaDraw;
     GLframebuffer *mesaRead;
     TRACE;
 
-    if (ctx) {
-	if (!draw || !read)
+    if (cPriv) {
+	struct dri_context *ctx = dri_context(cPriv);
+	struct dri_drawable *draw = dri_drawable(driDrawPriv);
+	struct dri_drawable *read = dri_drawable(driReadPriv);
+
+	if (!driDrawPriv || !driReadPriv)
 	    return GL_FALSE;
 
 	mesaCtx = &ctx->Base;
@@ -614,7 +613,7 @@ static int driBindContext(__DRIcontext *ctx,
 	_glapi_check_multithread();
 
 	swrast_check_and_update_window_size(mesaCtx, mesaDraw);
-	if (read != draw)
+	if (mesaRead != mesaDraw)
 	    swrast_check_and_update_window_size(mesaCtx, mesaRead);
 
 	_mesa_make_current( mesaCtx,
@@ -629,35 +628,25 @@ static int driBindContext(__DRIcontext *ctx,
     return GL_TRUE;
 }
 
-static int driUnbindContext(__DRIcontext *ctx)
+static GLboolean
+dri_unbind_context(__DRIcontext * cPriv)
 {
     TRACE;
-    (void) ctx;
+    (void) cPriv;
     return GL_TRUE;
 }
 
 
-static const __DRIcoreExtension driCoreExtension = {
-    { __DRI_CORE, __DRI_CORE_VERSION },
-    NULL, /* driCreateNewScreen */
-    driDestroyScreen,
-    driGetExtensions,
-    driGetConfigAttrib,
-    driIndexConfigAttrib,
-    NULL, /* driCreateNewDrawable */
-    driDestroyDrawable,
-    driSwapBuffers,
-    driCreateNewContext,
-    driCopyContext,
-    driDestroyContext,
-    driBindContext,
-    driUnbindContext
-};
-
-static const __DRIswrastExtension driSWRastExtension = {
-    { __DRI_SWRAST, __DRI_SWRAST_VERSION },
-    driCreateNewScreen,
-    driCreateNewDrawable
+const struct __DriverAPIRec driDriverAPI = {
+    .InitScreen = dri_init_screen,
+    .DestroyScreen = dri_destroy_screen,
+    .CreateContext = dri_create_context,
+    .DestroyContext = dri_destroy_context,
+    .CreateBuffer = dri_create_buffer,
+    .DestroyBuffer = dri_destroy_buffer,
+    .SwapBuffers = dri_swap_buffers,
+    .MakeCurrent = dri_make_current,
+    .UnbindContext = dri_unbind_context,
 };
 
 /* This is the table of extensions that the loader will dlsym() for. */
diff --git a/src/mesa/drivers/dri/swrast/swrast_priv.h b/src/mesa/drivers/dri/swrast/swrast_priv.h
index 4722007f95..77670d89a5 100644
--- a/src/mesa/drivers/dri/swrast/swrast_priv.h
+++ b/src/mesa/drivers/dri/swrast/swrast_priv.h
@@ -3,6 +3,7 @@
  * Version:  7.1
  *
  * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright 2008, 2010 George Sapountzis <gsapountzis@gmail.com>
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -22,11 +23,6 @@
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  */
 
-/*
- * Authors:
- *    George Sapountzis <gsap7@yahoo.gr>
- */
-
 
 #ifndef _SWRAST_PRIV_H
 #define _SWRAST_PRIV_H
@@ -34,6 +30,7 @@
 #include <GL/gl.h>
 #include <GL/internal/dri_interface.h>
 #include "main/mtypes.h"
+#include "dri_sw.h"
 
 
 /**
@@ -58,33 +55,51 @@
 /**
  * Data types
  */
-struct __DRIscreenRec {
-    int num;
-
-    const __DRIextension **extensions;
+struct dri_context
+{
+    /* mesa, base class, must be first */
+    GLcontext Base;
 
-    const __DRIswrastLoaderExtension *swrast_loader;
+    /* dri */
+    __DRIcontext *cPriv;
 };
 
-struct __DRIcontextRec {
-    GLcontext Base;
-
-    void *loaderPrivate;
+static INLINE struct dri_context *
+dri_context(__DRIcontext * driContextPriv)
+{
+    return (struct dri_context *)driContextPriv->driverPrivate;
+}
 
-    __DRIscreen *driScreenPriv;
-};
+static INLINE struct dri_context *
+swrast_context(GLcontext *ctx)
+{
+    return (struct dri_context *) ctx;
+}
 
-struct __DRIdrawableRec {
+struct dri_drawable
+{
+    /* mesa, base class, must be first */
     GLframebuffer Base;
 
-    void *loaderPrivate;
-
-    __DRIscreen *driScreenPriv;
+    /* dri */
+    __DRIdrawable *dPriv;
 
     /* scratch row for optimized front-buffer rendering */
     char *row;
 };
 
+static INLINE struct dri_drawable *
+dri_drawable(__DRIdrawable * driDrawPriv)
+{
+    return (struct dri_drawable *)driDrawPriv->driverPrivate;
+}
+
+static INLINE struct dri_drawable *
+swrast_drawable(GLframebuffer *fb)
+{
+    return (struct dri_drawable *) fb;
+}
+
 struct swrast_renderbuffer {
     struct gl_renderbuffer Base;
 
@@ -94,18 +109,6 @@ struct swrast_renderbuffer {
     GLuint bpp;
 };
 
-static INLINE __DRIcontext *
-swrast_context(GLcontext *ctx)
-{
-    return (__DRIcontext *) ctx;
-}
-
-static INLINE __DRIdrawable *
-swrast_drawable(GLframebuffer *fb)
-{
-    return (__DRIdrawable *) fb;
-}
-
 static INLINE struct swrast_renderbuffer *
 swrast_renderbuffer(struct gl_renderbuffer *rb)
 {
diff --git a/src/mesa/drivers/dri/swrast/swrast_span.c b/src/mesa/drivers/dri/swrast/swrast_span.c
index 5290dc82b9..c5681e34a9 100644
--- a/src/mesa/drivers/dri/swrast/swrast_span.c
+++ b/src/mesa/drivers/dri/swrast/swrast_span.c
@@ -3,6 +3,7 @@
  * Version:  7.1
  *
  * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright 2008, 2010 George Sapountzis <gsapountzis@gmail.com>
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -22,11 +23,6 @@
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  */
 
-/*
- * Authors:
- *    George Sapountzis <gsap7@yahoo.gr>
- */
-
 #include "swrast_priv.h"
 
 #define YFLIP(_xrb, Y) ((_xrb)->Base.Height - (Y) - 1)
diff --git a/src/mesa/drivers/dri/swrast/swrast_spantemp.h b/src/mesa/drivers/dri/swrast/swrast_spantemp.h
index 879a0c12e7..079726ae4a 100644
--- a/src/mesa/drivers/dri/swrast/swrast_spantemp.h
+++ b/src/mesa/drivers/dri/swrast/swrast_spantemp.h
@@ -39,8 +39,8 @@
 static INLINE void
 PUT_PIXEL( GLcontext *glCtx, GLint x, GLint y, GLubyte *p )
 {
-    __DRIcontext *ctx = swrast_context(glCtx);
-    __DRIdrawable *draw = swrast_drawable(glCtx->DrawBuffer);
+    __DRIcontext *ctx = swrast_context(glCtx)->cPriv;
+    __DRIdrawable *draw = swrast_drawable(glCtx->DrawBuffer)->dPriv;
 
     __DRIscreen *screen = ctx->driScreenPriv;
 
@@ -53,8 +53,8 @@ PUT_PIXEL( GLcontext *glCtx, GLint x, GLint y, GLubyte *p )
 static INLINE void
 GET_PIXEL( GLcontext *glCtx, GLint x, GLint y, GLubyte *p )
 {
-    __DRIcontext *ctx = swrast_context(glCtx);
-    __DRIdrawable *read = swrast_drawable(glCtx->ReadBuffer);
+    __DRIcontext *ctx = swrast_context(glCtx)->cPriv;
+    __DRIdrawable *read = swrast_drawable(glCtx->ReadBuffer)->dPriv;
 
     __DRIscreen *screen = ctx->driScreenPriv;
 
@@ -65,8 +65,8 @@ GET_PIXEL( GLcontext *glCtx, GLint x, GLint y, GLubyte *p )
 static INLINE void
 PUT_ROW( GLcontext *glCtx, GLint x, GLint y, GLuint n, char *row )
 {
-    __DRIcontext *ctx = swrast_context(glCtx);
-    __DRIdrawable *draw = swrast_drawable(glCtx->DrawBuffer);
+    __DRIcontext *ctx = swrast_context(glCtx)->cPriv;
+    __DRIdrawable *draw = swrast_drawable(glCtx->DrawBuffer)->dPriv;
 
     __DRIscreen *screen = ctx->driScreenPriv;
 
@@ -78,8 +78,8 @@ PUT_ROW( GLcontext *glCtx, GLint x, GLint y, GLuint n, char *row )
 static INLINE void
 GET_ROW( GLcontext *glCtx, GLint x, GLint y, GLuint n, char *row )
 {
-    __DRIcontext *ctx = swrast_context(glCtx);
-    __DRIdrawable *read = swrast_drawable(glCtx->ReadBuffer);
+    __DRIcontext *ctx = swrast_context(glCtx)->cPriv;
+    __DRIdrawable *read = swrast_drawable(glCtx->ReadBuffer)->dPriv;
 
     __DRIscreen *screen = ctx->driScreenPriv;
 
diff --git a/src/mesa/glapi/glapi_entrypoint.c b/src/mesa/glapi/glapi_entrypoint.c
index 5e6e5995f2..c4f43f66a1 100644
--- a/src/mesa/glapi/glapi_entrypoint.c
+++ b/src/mesa/glapi/glapi_entrypoint.c
@@ -65,7 +65,7 @@ get_entrypoint_address(GLuint functionOffset)
 #endif
 
 
-#if defined(PTHREADS) || defined(GLX_USE_TLS)
+#if defined(USE_X86_ASM)
 
 /**
  * Perform platform-specific GL API entry-point fixups.
@@ -73,7 +73,7 @@ get_entrypoint_address(GLuint functionOffset)
 static void
 init_glapi_relocs( void )
 {
-#if defined(USE_X86_ASM) && defined(GLX_USE_TLS) && !defined(GLX_X86_READONLY_TEXT)
+#if defined(GLX_USE_TLS) && !defined(GLX_X86_READONLY_TEXT)
     extern unsigned long _x86_get_dispatch(void);
     char run_time_patch[] = {
        0x65, 0xa1, 0, 0, 0, 0 /* movl %gs:0,%eax */
@@ -88,8 +88,63 @@ init_glapi_relocs( void )
 	curr_func += DISPATCH_FUNCTION_SIZE;
     }
 #endif
-#ifdef USE_SPARC_ASM
-    extern void __glapi_sparc_icache_flush(unsigned int *);
+}
+
+
+/**
+ * Generate a dispatch function (entrypoint) which jumps through
+ * the given slot number (offset) in the current dispatch table.
+ * We need assembly language in order to accomplish this.
+ */
+_glapi_proc
+generate_entrypoint(GLuint functionOffset)
+{
+   /* 32 is chosen as something of a magic offset.  For x86, the dispatch
+    * at offset 32 is the first one where the offset in the
+    * "jmp OFFSET*4(%eax)" can't be encoded in a single byte.
+    */
+   const GLubyte * const template_func = gl_dispatch_functions_start 
+     + (DISPATCH_FUNCTION_SIZE * 32);
+   GLubyte * const code = (GLubyte *) _glapi_exec_malloc(DISPATCH_FUNCTION_SIZE);
+
+
+   if ( code != NULL ) {
+      (void) memcpy(code, template_func, DISPATCH_FUNCTION_SIZE);
+      fill_in_entrypoint_offset( (_glapi_proc) code, functionOffset );
+   }
+
+   return (_glapi_proc) code;
+}
+
+
+/**
+ * This function inserts a new dispatch offset into the assembly language
+ * stub that was generated with the preceeding function.
+ */
+void
+fill_in_entrypoint_offset(_glapi_proc entrypoint, GLuint offset)
+{
+   GLubyte * const code = (GLubyte *) entrypoint;
+
+#if defined(GLX_USE_TLS)
+   *((unsigned int *)(code +  8)) = 4 * offset;
+#elif defined(THREADS)
+   *((unsigned int *)(code + 11)) = 4 * offset;
+   *((unsigned int *)(code + 22)) = 4 * offset;
+#else
+   *((unsigned int *)(code +  7)) = 4 * offset;
+#endif
+}
+
+
+#elif defined(USE_SPARC_ASM)
+
+extern void __glapi_sparc_icache_flush(unsigned int *);
+
+static void
+init_glapi_relocs( void )
+{
+#if defined(PTHREADS) || defined(GLX_USE_TLS)
     static const unsigned int template[] = {
 #ifdef GLX_USE_TLS
 	0x05000000, /* sethi %hi(_glapi_tls_Dispatch), %g2 */
@@ -155,7 +210,7 @@ init_glapi_relocs( void )
     int idx;
 #endif
 
-#if defined(GLX_USE_TLS)
+#ifdef GLX_USE_TLS
     code[0] = template[0] | (dispatch >> 10);
     code[1] = template[1];
     __glapi_sparc_icache_flush(&code[0]);
@@ -215,51 +270,10 @@ init_glapi_relocs( void )
 #endif
 }
 
-void
-init_glapi_relocs_once( void )
-{
-   static pthread_once_t once_control = PTHREAD_ONCE_INIT;
-   pthread_once( & once_control, init_glapi_relocs );
-}
-
-#else
-
-void
-init_glapi_relocs_once( void ) { }
-
-#endif /* defined(PTHREADS) || defined(GLX_USE_TLS) */
-
-
-#ifdef USE_SPARC_ASM
-extern void __glapi_sparc_icache_flush(unsigned int *);
-#endif
 
-/**
- * Generate a dispatch function (entrypoint) which jumps through
- * the given slot number (offset) in the current dispatch table.
- * We need assembly language in order to accomplish this.
- */
 _glapi_proc
 generate_entrypoint(GLuint functionOffset)
 {
-#if defined(USE_X86_ASM)
-   /* 32 is chosen as something of a magic offset.  For x86, the dispatch
-    * at offset 32 is the first one where the offset in the
-    * "jmp OFFSET*4(%eax)" can't be encoded in a single byte.
-    */
-   const GLubyte * const template_func = gl_dispatch_functions_start 
-     + (DISPATCH_FUNCTION_SIZE * 32);
-   GLubyte * const code = (GLubyte *) malloc(DISPATCH_FUNCTION_SIZE);
-
-
-   if ( code != NULL ) {
-      (void) memcpy(code, template_func, DISPATCH_FUNCTION_SIZE);
-      fill_in_entrypoint_offset( (_glapi_proc) code, functionOffset );
-   }
-
-   return (_glapi_proc) code;
-#elif defined(USE_SPARC_ASM)
-
 #if defined(PTHREADS) || defined(GLX_USE_TLS)
    static const unsigned int template[] = {
       0x07000000, /* sethi %hi(0), %g3 */
@@ -274,7 +288,7 @@ generate_entrypoint(GLuint functionOffset)
    extern unsigned int __glapi_sparc_pthread_stub;
    unsigned long call_dest = (unsigned long ) &__glapi_sparc_pthread_stub;
 #endif
-   unsigned int *code = (unsigned int *) malloc(sizeof(template));
+   unsigned int *code = (unsigned int *) _glapi_exec_malloc(sizeof(template));
    if (code) {
       code[0] = template[0] | (functionOffset & 0x3fffff);
       code[1] = template[1];
@@ -287,45 +301,52 @@ generate_entrypoint(GLuint functionOffset)
    }
    return (_glapi_proc) code;
 #endif
-
-#else
-   (void) functionOffset;
-   return NULL;
-#endif /* USE_*_ASM */
 }
 
 
-/**
- * This function inserts a new dispatch offset into the assembly language
- * stub that was generated with the preceeding function.
- */
 void
 fill_in_entrypoint_offset(_glapi_proc entrypoint, GLuint offset)
 {
-#if defined(USE_X86_ASM)
-   GLubyte * const code = (GLubyte *) entrypoint;
-
-#if DISPATCH_FUNCTION_SIZE == 32
-   *((unsigned int *)(code + 11)) = 4 * offset;
-   *((unsigned int *)(code + 22)) = 4 * offset;
-#elif DISPATCH_FUNCTION_SIZE == 16 && defined( GLX_USE_TLS )
-   *((unsigned int *)(code +  8)) = 4 * offset;
-#elif DISPATCH_FUNCTION_SIZE == 16
-   *((unsigned int *)(code +  7)) = 4 * offset;
-#else
-# error Invalid DISPATCH_FUNCTION_SIZE!
-#endif
-
-#elif defined(USE_SPARC_ASM)
    unsigned int *code = (unsigned int *) entrypoint;
+
    code[0] &= ~0x3fffff;
    code[0] |= (offset * sizeof(void *)) & 0x3fffff;
    __glapi_sparc_icache_flush(&code[0]);
-#else
+}
+
+
+#else /* USE_*_ASM */
 
+static void
+init_glapi_relocs( void )
+{
+}
+
+
+_glapi_proc
+generate_entrypoint(GLuint functionOffset)
+{
+   (void) functionOffset;
+   return NULL;
+}
+
+
+void
+fill_in_entrypoint_offset(_glapi_proc entrypoint, GLuint offset)
+{
    /* an unimplemented architecture */
    (void) entrypoint;
    (void) offset;
+}
 
 #endif /* USE_*_ASM */
+
+
+void
+init_glapi_relocs_once( void )
+{
+#if defined(PTHREADS) || defined(GLX_USE_TLS)
+   static pthread_once_t once_control = PTHREAD_ONCE_INIT;
+   pthread_once( & once_control, init_glapi_relocs );
+#endif
 }
diff --git a/src/mesa/glapi/glapi_execmem.c b/src/mesa/glapi/glapi_execmem.c
new file mode 100644
index 0000000000..6a1fac597f
--- /dev/null
+++ b/src/mesa/glapi/glapi_execmem.c
@@ -0,0 +1,127 @@
+/*
+ * Mesa 3-D graphics library
+ * Version:  6.5
+ *
+ * Copyright (C) 1999-2005  Brian Paul   All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+
+/**
+ * \file glapi_execmem.c
+ *
+ * Function for allocating executable memory for dispatch stubs.
+ *
+ * Copied from main/execmem.c and simplified for dispatch stubs.
+ */
+
+
+#ifdef HAVE_DIX_CONFIG_H
+#include <dix-config.h>
+#include "glapi/mesa.h"
+#else
+#include "main/compiler.h"
+#endif
+
+#include "glapi/glthread.h"
+
+
+#if defined(__linux__) || defined(__OpenBSD__) || defined(_NetBSD__) || defined(__sun)
+
+#include <unistd.h>
+#include <sys/mman.h>
+
+#ifdef MESA_SELINUX
+#include <selinux/selinux.h>
+#endif
+
+
+#ifndef MAP_ANONYMOUS
+#define MAP_ANONYMOUS MAP_ANON
+#endif
+
+
+#define EXEC_MAP_SIZE (4*1024)
+
+_glthread_DECLARE_STATIC_MUTEX(exec_mutex);
+
+static unsigned int head = 0;
+
+static unsigned char *exec_mem = NULL;
+
+
+/*
+ * Dispatch stubs are of fixed size and never freed. Thus, we do not need to
+ * overlay a heap, we just mmap a page and manage through an index.
+ */
+
+static int
+init_map(void)
+{
+#ifdef MESA_SELINUX
+   if (is_selinux_enabled()) {
+      if (!security_get_boolean_active("allow_execmem") ||
+	  !security_get_boolean_pending("allow_execmem"))
+         return 0;
+   }
+#endif
+
+   if (!exec_mem)
+      exec_mem = mmap(NULL, EXEC_MAP_SIZE, PROT_EXEC | PROT_READ | PROT_WRITE,
+		      MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
+
+   return (exec_mem != MAP_FAILED);
+}
+
+
+void *
+_glapi_exec_malloc(unsigned int size)
+{
+   void *addr = NULL;
+
+   _glthread_LOCK_MUTEX(exec_mutex);
+
+   if (!init_map())
+      goto bail;
+
+   /* free space check, assumes no integer overflow */
+   if (head + size > EXEC_MAP_SIZE)
+      goto bail;
+
+   /* allocation, assumes proper addr and size alignement */
+   addr = exec_mem + head;
+   head += size;
+
+bail:
+   _glthread_UNLOCK_MUTEX(exec_mutex);
+
+   return addr;
+}
+
+
+#else
+
+void *
+_glapi_exec_malloc(unsigned int size)
+{
+   return malloc(size);
+}
+
+
+#endif
diff --git a/src/mesa/glapi/glapi_getproc.c b/src/mesa/glapi/glapi_getproc.c
index 295657875d..c73e8dd3b0 100644
--- a/src/mesa/glapi/glapi_getproc.c
+++ b/src/mesa/glapi/glapi_getproc.c
@@ -200,12 +200,6 @@ struct _glapi_function {
 };
 
 
-/*
- * Number of extension functions which we can dynamically add at runtime.
- */
-#define MAX_EXTENSION_FUNCS 300
-
-
 static struct _glapi_function ExtEntryTable[MAX_EXTENSION_FUNCS];
 static GLuint NumExtEntryPoints = 0;
 
diff --git a/src/mesa/glapi/glapi_priv.h b/src/mesa/glapi/glapi_priv.h
index 7cd81ee8dc..0e2de460f2 100644
--- a/src/mesa/glapi/glapi_priv.h
+++ b/src/mesa/glapi/glapi_priv.h
@@ -28,6 +28,9 @@
 
 #include "glthread.h"
 
+
+/* getproc */
+
 extern void
 _glapi_check_table_not_null(const struct _glapi_table *table);
 
@@ -36,6 +39,14 @@ extern void
 _glapi_check_table(const struct _glapi_table *table);
 
 
+/* execmem */
+
+extern void *
+_glapi_exec_malloc(GLuint size);
+
+
+/* entrypoint */
+
 extern void
 init_glapi_relocs_once(void);
 
@@ -52,15 +63,35 @@ extern _glapi_proc
 get_entrypoint_address(GLuint functionOffset);
 
 
-#if defined(USE_X64_64_ASM) && defined(GLX_USE_TLS)
-# define DISPATCH_FUNCTION_SIZE  16
-#elif defined(USE_X86_ASM)
-# if defined(THREADS) && !defined(GLX_USE_TLS)
+/**
+ * Size (in bytes) of dispatch function (entrypoint).
+ */
+#if defined(USE_X86_ASM)
+# if defined(GLX_USE_TLS)
+#  define DISPATCH_FUNCTION_SIZE  16
+# elif defined(THREADS)
 #  define DISPATCH_FUNCTION_SIZE  32
 # else
 #  define DISPATCH_FUNCTION_SIZE  16
 # endif
 #endif
 
+#if defined(USE_X64_64_ASM)
+# if defined(GLX_USE_TLS)
+#  define DISPATCH_FUNCTION_SIZE  16
+# endif
+#endif
+
+
+/**
+ * Number of extension functions which we can dynamically add at runtime.
+ *
+ * Number of extension functions is also subject to the size of backing exec
+ * mem we allocate. For the common case of dispatch stubs with size 16 bytes,
+ * the two limits will be hit simultaneously. For larger dispatch function
+ * sizes, MAX_EXTENSION_FUNCS is effectively reduced.
+ */
+#define MAX_EXTENSION_FUNCS 256
+
 
 #endif
diff --git a/src/mesa/sources.mak b/src/mesa/sources.mak
index 74885548e5..d59e24de1f 100644
--- a/src/mesa/sources.mak
+++ b/src/mesa/sources.mak
@@ -89,6 +89,7 @@ GLAPI_SOURCES = \
 	glapi/glapi.c \
 	glapi/glapi_dispatch.c \
 	glapi/glapi_entrypoint.c \
+	glapi/glapi_execmem.c \
 	glapi/glapi_getproc.c \
 	glapi/glapi_nop.c \
 	glapi/glthread.c