r200/r300: swtcl fixups to use old dma buffers on top of BOs

14 files changed, 163 insertions, 311 deletions

diff --git a/src/mesa/drivers/dri/r200/r200_context.c b/src/mesa/drivers/dri/r200/r200_context.c
index c2b269880c..c2b5afc385 100644
--- a/src/mesa/drivers/dri/r200/r200_context.c
+++ b/src/mesa/drivers/dri/r200/r200_context.c
@@ -297,6 +297,7 @@ static void r200_init_vtbl(radeonContextPtr radeon)
    radeon->vtbl.update_draw_buffer = r200UpdateDrawBuffer;
    radeon->vtbl.emit_cs_header = r200_vtbl_emit_cs_header;
    radeon->vtbl.emit_state = r200_vtbl_emit_state;
+   radeon->vtbl.swtcl_flush = r200_swtcl_flush;
 }
 
 
@@ -389,7 +390,7 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual,
       rmesa->radeon.texture_depth = ( screen->cpp == 4 ) ?
 	 DRI_CONF_TEXTURE_DEPTH_32 : DRI_CONF_TEXTURE_DEPTH_16;
 
-   rmesa->swtcl.RenderIndex = ~0;
+   rmesa->radeon.swtcl.RenderIndex = ~0;
    rmesa->hw.all_dirty = 1;
 
    /* Set the maximum texture size small enough that we can guarentee that
diff --git a/src/mesa/drivers/dri/r200/r200_context.h b/src/mesa/drivers/dri/r200/r200_context.h
index 2df0ca1272..4b7ce5d1ef 100644
--- a/src/mesa/drivers/dri/r200/r200_context.h
+++ b/src/mesa/drivers/dri/r200/r200_context.h
@@ -551,40 +551,12 @@ struct r200_tcl_info {
 /* r200_swtcl.c
  */
 struct r200_swtcl_info {
-   GLuint RenderIndex;
-   
-   /**
-    * Size of a hardware vertex.  This is calculated when \c ::vertex_attrs is
-    * installed in the Mesa state vector.
-    */
-   GLuint vertex_size;
-
-   /**
-    * Attributes instructing the Mesa TCL pipeline where / how to put vertex
-    * data in the hardware buffer.
-    */
-   struct tnl_attr_map vertex_attrs[VERT_ATTRIB_MAX];
-
-   /**
-    * Number of elements of \c ::vertex_attrs that are actually used.
-    */
-   GLuint vertex_attr_count;
 
-   /**
-    * Cached pointer to the buffer where Mesa will store vertex data.
-    */
-   GLubyte *verts;
 
-   /* Fallback rasterization functions
-    */
    radeon_point_func draw_point;
    radeon_line_func draw_line;
    radeon_tri_func draw_tri;
 
-   GLuint hw_primitive;
-   GLenum render_primitive;
-   GLuint numverts;
-
    /**
     * Offset of the 4UB color data within a hardware (swtcl) vertex.
     */
@@ -599,8 +571,6 @@ struct r200_swtcl_info {
     * Should Mesa project vertex data or will the hardware do it?
     */
    GLboolean needproj;
-
-   struct radeon_bo *bo;
 };
 
 
diff --git a/src/mesa/drivers/dri/r200/r200_maos_arrays.c b/src/mesa/drivers/dri/r200/r200_maos_arrays.c
index 278e00442c..1bf7ef33f8 100644
--- a/src/mesa/drivers/dri/r200/r200_maos_arrays.c
+++ b/src/mesa/drivers/dri/r200/r200_maos_arrays.c
@@ -75,37 +75,29 @@ static void r200_emit_vecfog(GLcontext *ctx, struct radeon_aos *aos,
 {
 	radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
 	uint32_t *out;
-	uint32_t bo_size;
 	int i;
 	int size = 1;
 
-	memset(aos, 0, sizeof(struct radeon_aos));
 	if (stride == 0) {
-		bo_size = size * 4;
+		radeonAllocDmaRegion(rmesa, &aos->bo, &aos->offset, size * 4, 32);
 		count = 1;
 		aos->stride = 0;
 	} else {
-		bo_size = size * count * 4;
+		radeonAllocDmaRegion(rmesa, &aos->bo, &aos->offset, size * 4, 32);
 		aos->stride = size;
 	}
-	aos->bo = radeon_bo_open(rmesa->radeonScreen->bom,
-				 0, bo_size, 32, RADEON_GEM_DOMAIN_GTT, 0);
-	aos->offset = 0;
+
 	aos->components = size;
 	aos->count = count;
 
-	radeon_bo_map(aos->bo, 1);
 	out = (uint32_t*)((char*)aos->bo->ptr + aos->offset);
 	for (i = 0; i < count; i++) {
 	  out[0] = r200ComputeFogBlendFactor( ctx, *(GLfloat *)data );
 	  out++;
 	  data += stride;
 	}
-	radeon_bo_unmap(aos->bo);
 }
 
-
-
 /* Emit any changed arrays to new GART memory, re-emit a packet to
  * update the arrays.  
  */
@@ -230,6 +222,7 @@ after_emit:
       rmesa->hw.vtx.cmd[VTX_VTXFMT_1] = vfmt1;
    }
 
+   radeon_bo_unmap(rmesa->radeon.dma.current);
    rmesa->tcl.nr_aos_components = nr;
 }
 
@@ -240,7 +233,9 @@ void r200ReleaseArrays( GLcontext *ctx, GLuint newinputs )
    int i;
    for (i = 0; i < rmesa->tcl.nr_aos_components; i++) {
      if (rmesa->tcl.aos[i].bo) {
-       rmesa->tcl.aos[i].bo = radeon_bo_unref(rmesa->tcl.aos[i].bo);
+       radeon_bo_unref(rmesa->tcl.aos[i].bo);
+       rmesa->tcl.aos[i].bo = NULL;
      }
    }
+   radeonReleaseDmaRegion(&rmesa->radeon);
 }
diff --git a/src/mesa/drivers/dri/r200/r200_swtcl.c b/src/mesa/drivers/dri/r200/r200_swtcl.c
index ef725a0efb..a39072d1d2 100644
--- a/src/mesa/drivers/dri/r200/r200_swtcl.c
+++ b/src/mesa/drivers/dri/r200/r200_swtcl.c
@@ -56,27 +56,24 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "r200_tcl.h"
 
 
-static void flush_last_swtcl_prim( GLcontext *ctx  );
-
-
 /***********************************************************************
  *                         Initialization 
  ***********************************************************************/
 
 #define EMIT_ATTR( ATTR, STYLE, F0 )					\
 do {									\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].attrib = (ATTR);	\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].format = (STYLE);	\
-   rmesa->swtcl.vertex_attr_count++;					\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].attrib = (ATTR);	\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].format = (STYLE);	\
+   rmesa->radeon.swtcl.vertex_attr_count++;					\
    fmt_0 |= F0;								\
 } while (0)
 
 #define EMIT_PAD( N )							\
 do {									\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].attrib = 0;		\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].format = EMIT_PAD;	\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].offset = (N);		\
-   rmesa->swtcl.vertex_attr_count++;					\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].attrib = 0;		\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].format = EMIT_PAD;	\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].offset = (N);		\
+   rmesa->radeon.swtcl.vertex_attr_count++;					\
 } while (0)
 
 static void r200SetVertexFormat( GLcontext *ctx )
@@ -101,7 +98,7 @@ static void r200SetVertexFormat( GLcontext *ctx )
    }
 
    assert( VB->AttribPtr[VERT_ATTRIB_POS] != NULL );
-   rmesa->swtcl.vertex_attr_count = 0;
+   rmesa->radeon.swtcl.vertex_attr_count = 0;
 
    /* EMIT_ATTR's must be in order as they tell t_vertex.c how to
     * build up a hardware vertex.
@@ -194,12 +191,12 @@ static void r200SetVertexFormat( GLcontext *ctx )
       rmesa->hw.vtx.cmd[VTX_VTXFMT_0] = fmt_0;
       rmesa->hw.vtx.cmd[VTX_VTXFMT_1] = fmt_1;
 
-      rmesa->swtcl.vertex_size =
+      rmesa->radeon.swtcl.vertex_size =
 	  _tnl_install_attrs( ctx,
-			      rmesa->swtcl.vertex_attrs, 
-			      rmesa->swtcl.vertex_attr_count,
+			      rmesa->radeon.swtcl.vertex_attrs, 
+			      rmesa->radeon.swtcl.vertex_attr_count,
 			      NULL, 0 );
-      rmesa->swtcl.vertex_size /= 4;
+      rmesa->radeon.swtcl.vertex_size /= 4;
       RENDERINPUTS_COPY( rmesa->radeon.tnl_index_bitset, index_bitset );
    }
 }
@@ -270,63 +267,27 @@ void r200ChooseVertexState( GLcontext *ctx )
    }
 }
 
-
-/* Flush vertices in the current dma region.
- */
-static void flush_last_swtcl_prim(GLcontext *ctx)
+void r200_swtcl_flush(GLcontext *ctx, uint32_t current_offset)
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   if (R200_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   rmesa->radeon.dma.flush = NULL;
-
-   radeon_bo_unmap(rmesa->swtcl.bo);
-   rcommonEnsureCmdBufSpace(rmesa,
+   rcommonEnsureCmdBufSpace(&rmesa->radeon,
 			    rmesa->hw.max_state_size + (12*sizeof(int)),
 			    __FUNCTION__);
 
 
    r200EmitState(rmesa);
    r200EmitVertexAOS( rmesa,
-		      rmesa->swtcl.vertex_size,
-		      rmesa->swtcl.bo, 0);
+		      rmesa->radeon.swtcl.vertex_size,
+		      rmesa->radeon.dma.current,
+		      current_offset);
 
 		      
    r200EmitVbufPrim( rmesa,
-		     rmesa->swtcl.hw_primitive,
-		     rmesa->swtcl.numverts);
+		     rmesa->radeon.swtcl.hw_primitive,
+		     rmesa->radeon.swtcl.numverts);
 
-   radeon_bo_unref(rmesa->swtcl.bo);
-   //   COMMIT_BATCH();
-   rmesa->swtcl.numverts = 0;
 }
 
-
-/* Alloc space in the current dma region.
- */
-static INLINE void *
-r200AllocDmaLowVerts( r200ContextPtr rmesa, int nverts, int vsize )
-{
-   GLuint bytes = vsize * nverts;
-
-   if (rmesa->swtcl.flush)
-     rmesa->swtcl.flush(rmesa->radeon.glCtx);
-
-   rmesa->swtcl.bo = radeon_bo_open(rmesa->radeon.radeonScreen->bom,
-				    0, bytes, 4, RADEON_GEM_DOMAIN_GTT, 0);
-   radeon_bo_map(rmesa->swtcl.bo, 1);
-   if (rmesa->radeon.dma.flush == NULL) {
-     rmesa->radeon.glCtx->Driver.NeedFlush |= FLUSH_STORED_VERTICES;
-     rmesa->radeon.dma.flush = flush_last_swtcl_prim;
-   }
-   rmesa->swtcl.numverts = nverts;
-   return rmesa->swtcl.bo->ptr;
-
-   
-}
-
-
 /**************************************************************************/
 
 
@@ -374,11 +335,11 @@ static void r200ResetLineStipple( GLcontext *ctx );
 #undef LOCAL_VARS
 #undef ALLOC_VERTS
 #define CTX_ARG r200ContextPtr rmesa
-#define GET_VERTEX_DWORDS() rmesa->swtcl.vertex_size
-#define ALLOC_VERTS( n, size ) r200AllocDmaLowVerts( rmesa, n, size * 4 )
+#define GET_VERTEX_DWORDS() rmesa->radeon.swtcl.vertex_size
+#define ALLOC_VERTS( n, size ) rcommonAllocDmaLowVerts( &rmesa->radeon, n, size * 4 )
 #define LOCAL_VARS						\
    r200ContextPtr rmesa = R200_CONTEXT(ctx);		\
-   const char *r200verts = (char *)rmesa->swtcl.verts;
+   const char *r200verts = (char *)rmesa->radeon.swtcl.verts;
 #define VERT(x) (radeonVertex *)(r200verts + ((x) * vertsize * sizeof(int)))
 #define VERTEX radeonVertex 
 #define DO_DEBUG_VERTS (1 && (R200_DEBUG & DEBUG_VERTS))
@@ -438,7 +399,7 @@ static struct {
 #define VERT_Y(_v) _v->v.y
 #define VERT_Z(_v) _v->v.z
 #define AREA_IS_CCW( a ) (a < 0)
-#define GET_VERTEX(e) (rmesa->swtcl.verts + (e*rmesa->swtcl.vertex_size*sizeof(int)))
+#define GET_VERTEX(e) (rmesa->radeon.swtcl.verts + (e*rmesa->radeon.swtcl.vertex_size*sizeof(int)))
 
 #define VERT_SET_RGBA( v, c )  					\
 do {								\
@@ -495,7 +456,7 @@ do {							\
  ***********************************************************************/
 
 #define RASTERIZE(x) r200RasterPrimitive( ctx, reduced_hw_prim(ctx, x) )
-#define RENDER_PRIMITIVE rmesa->swtcl.render_primitive
+#define RENDER_PRIMITIVE rmesa->radeon.swtcl.render_primitive
 #undef TAG
 #define TAG(x) x
 #include "tnl_dd/t_dd_unfilled.h"
@@ -551,8 +512,8 @@ static void init_rast_tab( void )
 #undef LOCAL_VARS
 #define LOCAL_VARS						\
    r200ContextPtr rmesa = R200_CONTEXT(ctx);		\
-   const GLuint vertsize = rmesa->swtcl.vertex_size;		\
-   const char *r200verts = (char *)rmesa->swtcl.verts;		\
+   const GLuint vertsize = rmesa->radeon.swtcl.vertex_size;		\
+   const char *r200verts = (char *)rmesa->radeon.swtcl.verts;		\
    const GLuint * const elt = TNL_CONTEXT(ctx)->vb.Elts;	\
    const GLboolean stipple = ctx->Line.StippleFlag;		\
    (void) elt; (void) stipple;
@@ -587,7 +548,7 @@ void r200ChooseRenderState( GLcontext *ctx )
    if (flags & DD_TRI_LIGHT_TWOSIDE) index |= R200_TWOSIDE_BIT;
    if (flags & DD_TRI_UNFILLED)      index |= R200_UNFILLED_BIT;
 
-   if (index != rmesa->swtcl.RenderIndex) {
+   if (index != rmesa->radeon.swtcl.RenderIndex) {
       tnl->Driver.Render.Points = rast_tab[index].points;
       tnl->Driver.Render.Line = rast_tab[index].line;
       tnl->Driver.Render.ClippedLine = rast_tab[index].line;
@@ -604,7 +565,7 @@ void r200ChooseRenderState( GLcontext *ctx )
 	 tnl->Driver.Render.ClippedPolygon = _tnl_RenderClippedPolygon;
       }
 
-      rmesa->swtcl.RenderIndex = index;
+      rmesa->radeon.swtcl.RenderIndex = index;
    }
 }
 
@@ -618,7 +579,7 @@ static void r200RasterPrimitive( GLcontext *ctx, GLuint hwprim )
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
 
-   if (rmesa->swtcl.hw_primitive != hwprim) {
+   if (rmesa->radeon.swtcl.hw_primitive != hwprim) {
       /* need to disable perspective-correct texturing for point sprites */
       if ((hwprim & 0xf) == R200_VF_PRIM_POINT_SPRITES && ctx->Point.PointSprite) {
 	 if (rmesa->hw.set.cmd[SET_RE_CNTL] & R200_PERSPECTIVE_ENABLE) {
@@ -631,14 +592,14 @@ static void r200RasterPrimitive( GLcontext *ctx, GLuint hwprim )
 	 rmesa->hw.set.cmd[SET_RE_CNTL] |= R200_PERSPECTIVE_ENABLE;
       }
       R200_NEWPRIM( rmesa );
-      rmesa->swtcl.hw_primitive = hwprim;
+      rmesa->radeon.swtcl.hw_primitive = hwprim;
    }
 }
 
 static void r200RenderPrimitive( GLcontext *ctx, GLenum prim )
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   rmesa->swtcl.render_primitive = prim;
+   rmesa->radeon.swtcl.render_primitive = prim;
    if (prim < GL_TRIANGLES || !(ctx->_TriangleCaps & DD_TRI_UNFILLED)) 
       r200RasterPrimitive( ctx, reduced_hw_prim(ctx, prim) );
 }
@@ -691,7 +652,7 @@ void r200Fallback( GLcontext *ctx, GLuint bit, GLboolean mode )
 	 R200_FIREVERTICES( rmesa );
 	 TCL_FALLBACK( ctx, R200_TCL_FALLBACK_RASTER, GL_TRUE );
 	 _swsetup_Wakeup( ctx );
-	 rmesa->swtcl.RenderIndex = ~0;
+	 rmesa->radeon.swtcl.RenderIndex = ~0;
          if (R200_DEBUG & DEBUG_FALLBACKS) {
             fprintf(stderr, "R200 begin rasterization fallback: 0x%x %s\n",
                     bit, getFallbackString(bit));
@@ -776,7 +737,7 @@ r200PointsBitmap( GLcontext *ctx, GLint px, GLint py,
       vte |= R200_VTX_W0_FMT;
       vap &= ~R200_VAP_FORCE_W_TO_ONE;
 
-      rmesa->swtcl.vertex_size = 5;
+      rmesa->radeon.swtcl.vertex_size = 5;
 
       if ( (rmesa->hw.vtx.cmd[VTX_VTXFMT_0] != fmt_0)
 	   || (rmesa->hw.vtx.cmd[VTX_VTXFMT_1] != fmt_1) ) {
@@ -944,10 +905,10 @@ void r200InitSwtcl( GLcontext *ctx )
    _tnl_init_vertices( ctx, ctx->Const.MaxArrayLockSize + 12, 
 		       36 * sizeof(GLfloat) );
    
-   rmesa->swtcl.verts = (GLubyte *)tnl->clipspace.vertex_buf;
-   rmesa->swtcl.RenderIndex = ~0;
-   rmesa->swtcl.render_primitive = GL_TRIANGLES;
-   rmesa->swtcl.hw_primitive = 0;
+   rmesa->radeon.swtcl.verts = (GLubyte *)tnl->clipspace.vertex_buf;
+   rmesa->radeon.swtcl.RenderIndex = ~0;
+   rmesa->radeon.swtcl.render_primitive = GL_TRIANGLES;
+   rmesa->radeon.swtcl.hw_primitive = 0;
 }
 
 
diff --git a/src/mesa/drivers/dri/r200/r200_swtcl.h b/src/mesa/drivers/dri/r200/r200_swtcl.h
index 91c7c0ff8f..a4051a4b76 100644
--- a/src/mesa/drivers/dri/r200/r200_swtcl.h
+++ b/src/mesa/drivers/dri/r200/r200_swtcl.h
@@ -66,5 +66,5 @@ extern void r200PointsBitmap( GLcontext *ctx, GLint px, GLint py,
 			      const struct gl_pixelstore_attrib *unpack,
 			      const GLubyte *bitmap );
 
-
+void r200_swtcl_flush(GLcontext *ctx, uint32_t current_offset);
 #endif
diff --git a/src/mesa/drivers/dri/r200/r200_tcl.c b/src/mesa/drivers/dri/r200/r200_tcl.c
index 3c19e330f5..494a8f3d7b 100644
--- a/src/mesa/drivers/dri/r200/r200_tcl.c
+++ b/src/mesa/drivers/dri/r200/r200_tcl.c
@@ -157,7 +157,7 @@ static GLushort *r200AllocElts( r200ContextPtr rmesa, GLuint nr )
       if (rmesa->radeon.dma.flush)
 	 rmesa->radeon.dma.flush( rmesa->radeon.glCtx );
 
-      rcommonEnsureCmdBufSpace(rmesa, AOS_BUFSZ(rmesa->tcl.nr_aos_components));
+      rcommonEnsureCmdBufSpace(&rmesa->radeon, AOS_BUFSZ(rmesa->tcl.nr_aos_components), __FUNCTION__);
 
       r200EmitAOS( rmesa,
 		   rmesa->tcl.nr_aos_components, 0 );
@@ -190,7 +190,7 @@ static void r200EmitPrim( GLcontext *ctx,
    //   fprintf(stderr,"Emit prim %d\n", rmesa->tcl.nr_aos_components);
    rcommonEnsureCmdBufSpace( &rmesa->radeon,
 			     AOS_BUFSZ(rmesa->tcl.nr_aos_components) +
-			     rmesa->hw.max_state_size + VBUF_BUFSZ );
+			     rmesa->hw.max_state_size + VBUF_BUFSZ, __FUNCTION__ );
 
    r200EmitAOS( rmesa,
 		rmesa->tcl.nr_aos_components,
diff --git a/src/mesa/drivers/dri/r300/r300_context.c b/src/mesa/drivers/dri/r300/r300_context.c
index f57952d878..7e707a38b3 100644
--- a/src/mesa/drivers/dri/r300/r300_context.c
+++ b/src/mesa/drivers/dri/r300/r300_context.c
@@ -260,6 +260,7 @@ static void r300_init_vtbl(radeonContextPtr radeon)
    radeon->vtbl.emit_cs_header = r300_vtbl_emit_cs_header;
    radeon->vtbl.emit_state = r300_vtbl_emit_state;
    radeon->vtbl.flush_vertices = r300_vtbl_flush_vertices;
+   radeon->vtbl.swtcl_flush = r300_swtcl_flush;
 }
 
 
diff --git a/src/mesa/drivers/dri/r300/r300_context.h b/src/mesa/drivers/dri/r300/r300_context.h
index 155529a8a6..9e97a136e3 100644
--- a/src/mesa/drivers/dri/r300/r300_context.h
+++ b/src/mesa/drivers/dri/r300/r300_context.h
@@ -658,37 +658,7 @@ struct r300_state {
 /* r300_swtcl.c
  */
 struct r300_swtcl_info {
-   GLuint RenderIndex;
-
-   /**
-    * Size of a hardware vertex.  This is calculated when \c ::vertex_attrs is
-    * installed in the Mesa state vector.
-    */
-   GLuint vertex_size;
-
-   /**
-    * Attributes instructing the Mesa TCL pipeline where / how to put vertex
-    * data in the hardware buffer.
-    */
-   struct tnl_attr_map vertex_attrs[VERT_ATTRIB_MAX];
-
-   /**
-    * Number of elements of \c ::vertex_attrs that are actually used.
-    */
-   GLuint vertex_attr_count;
-
-   /**
-    * Cached pointer to the buffer where Mesa will store vertex data.
-    */
-   GLubyte *verts;
-
-   /* Fallback rasterization functions
-    */
-   GLuint hw_primitive;
-   GLenum render_primitive;
-   GLuint numverts;
-
-   /**
+  /*
     * Offset of the 4UB color data within a hardware (swtcl) vertex.
     */
    GLuint coloroffset;
@@ -697,8 +667,6 @@ struct r300_swtcl_info {
     * Offset of the 3UB specular color data within a hardware (swtcl) vertex.
     */
    GLuint specoffset;
-
-   struct radeon_bo *bo;
 };
 
 
diff --git a/src/mesa/drivers/dri/r300/r300_swtcl.c b/src/mesa/drivers/dri/r300/r300_swtcl.c
index ef65fbb127..73e3c51b9a 100644
--- a/src/mesa/drivers/dri/r300/r300_swtcl.c
+++ b/src/mesa/drivers/dri/r300/r300_swtcl.c
@@ -57,24 +57,21 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "r300_ioctl.h"
 #include "r300_emit.h"
 
-static void flush_last_swtcl_prim( GLcontext *ctx);
-
-
 void r300EmitVertexAOS(r300ContextPtr rmesa, GLuint vertex_size, struct radeon_bo *bo, GLuint offset);
 void r300EmitVbufPrim(r300ContextPtr rmesa, GLuint primitive, GLuint vertex_nr);
 #define EMIT_ATTR( ATTR, STYLE )					\
 do {									\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].attrib = (ATTR);	\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].format = (STYLE);	\
-   rmesa->swtcl.vertex_attr_count++;					\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].attrib = (ATTR);	\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].format = (STYLE);	\
+   rmesa->radeon.swtcl.vertex_attr_count++;					\
 } while (0)
 
 #define EMIT_PAD( N )							\
 do {									\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].attrib = 0;		\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].format = EMIT_PAD;	\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].offset = (N);		\
-   rmesa->swtcl.vertex_attr_count++;					\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].attrib = 0;		\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].format = EMIT_PAD;	\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].offset = (N);		\
+   rmesa->radeon.swtcl.vertex_attr_count++;					\
 } while (0)
 
 static void r300SetVertexFormat( GLcontext *ctx )
@@ -112,7 +109,7 @@ static void r300SetVertexFormat( GLcontext *ctx )
 	}
 
 	assert( VB->AttribPtr[VERT_ATTRIB_POS] != NULL );
-	rmesa->swtcl.vertex_attr_count = 0;
+	rmesa->radeon.swtcl.vertex_attr_count = 0;
 
 	/* EMIT_ATTR's must be in order as they tell t_vertex.c how to
 	 * build up a hardware vertex.
@@ -222,95 +219,20 @@ static void r300SetVertexFormat( GLcontext *ctx )
 	rmesa->hw.vof.cmd[R300_VOF_CNTL_0] = r300VAPOutputCntl0(ctx, OutputsWritten);
 	rmesa->hw.vof.cmd[R300_VOF_CNTL_1] = vap_fmt_1;
 
-	rmesa->swtcl.vertex_size =
+	rmesa->radeon.swtcl.vertex_size =
 		_tnl_install_attrs( ctx,
-				    rmesa->swtcl.vertex_attrs,
-				    rmesa->swtcl.vertex_attr_count,
+				    rmesa->radeon.swtcl.vertex_attrs,
+				    rmesa->radeon.swtcl.vertex_attr_count,
 				    NULL, 0 );
 
-	rmesa->swtcl.vertex_size /= 4;
+	rmesa->radeon.swtcl.vertex_size /= 4;
 
 	RENDERINPUTS_COPY( rmesa->tnl_index_bitset, index_bitset );
 
 
 	R300_STATECHANGE(rmesa, vte);
 	rmesa->hw.vte.cmd[1] = vte;
-	rmesa->hw.vte.cmd[2] = rmesa->swtcl.vertex_size;
-}
-
-
-/* Flush vertices in the current dma region.
- */
-static void flush_last_swtcl_prim( GLcontext *ctx  )
-{
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	struct radeon_dma *dma = &rmesa->radeon.dma;
-		
-
-	if (RADEON_DEBUG & DEBUG_IOCTL)
-		fprintf(stderr, "%s\n", __FUNCTION__);
-	dma->flush = NULL;
-
-	if (dma->current) {
-	    GLuint current_offset = dma->current_used;
-
-	    assert (dma->current_used +
-		    rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 ==
-		    dma->current_vertexptr);
-
-	    radeon_bo_unmap(dma->current);
-	    if (dma->current_used != dma->current_vertexptr) {
-		    dma->current_used = dma->current_vertexptr;
-
-		    rcommonEnsureCmdBufSpace(rmesa,
-					     rmesa->hw.max_state_size + (12*sizeof(int)),
-					     __FUNCTION__);
-		    r300EmitState(rmesa);
-		    r300EmitVertexAOS(rmesa,
-				      rmesa->swtcl.vertex_size,
-				      dma->current,
-				      current_offset);
-
-		    r300EmitVbufPrim(rmesa,
-				     rmesa->swtcl.hw_primitive,
-				     rmesa->swtcl.numverts);
-		    r300EmitCacheFlush(rmesa);
-		    COMMIT_BATCH();
-	    }
-	    radeonReleaseDmaRegion(&rmesa->radeon);
-	    rmesa->swtcl.numverts = 0;
-	}
-}
-
-/* Alloc space in the current dma region.
- */
-static void *
-r300AllocDmaLowVerts( r300ContextPtr rmesa, int nverts, int vsize )
-{
-	GLuint bytes = vsize * nverts;
-	void *head;
-
-	if (!rmesa->radeon.dma.current || rmesa->radeon.dma.current_vertexptr + bytes > rmesa->radeon.dma.current->size) {
-                radeonRefillCurrentDmaRegion( &rmesa->radeon, bytes);
-	}
-
-        if (!rmesa->radeon.dma.flush) {
-                rmesa->radeon.glCtx->Driver.NeedFlush |= FLUSH_STORED_VERTICES;
-                rmesa->radeon.dma.flush = flush_last_swtcl_prim;
-        }
-
-	ASSERT( vsize == rmesa->swtcl.vertex_size * 4 );
-        ASSERT( rmesa->radeon.dma.flush == flush_last_swtcl_prim );
-        ASSERT( rmesa->radeon.dma.current_used +
-                rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 ==
-                rmesa->radeon.dma.current_vertexptr );
-
-//	fprintf(stderr,"current %p %x\n", rmesa->radeon.dma.current->ptr,
-//		rmesa->radeon.dma.current_vertexptr);
-	head = (rmesa->radeon.dma.current->ptr + rmesa->radeon.dma.current_vertexptr);
-	rmesa->radeon.dma.current_vertexptr += bytes;
-	rmesa->swtcl.numverts += nverts;
-	return head;
+	rmesa->hw.vte.cmd[2] = rmesa->radeon.swtcl.vertex_size;
 }
 
 static GLuint reduced_prim[] = {
@@ -350,11 +272,11 @@ static void r300RenderPrimitive( GLcontext *ctx, GLenum prim );
 #undef LOCAL_VARS
 #undef ALLOC_VERTS
 #define CTX_ARG r300ContextPtr rmesa
-#define GET_VERTEX_DWORDS() rmesa->swtcl.vertex_size
-#define ALLOC_VERTS( n, size ) r300AllocDmaLowVerts( rmesa, n, size * 4 )
+#define GET_VERTEX_DWORDS() rmesa->radeon.swtcl.vertex_size
+#define ALLOC_VERTS( n, size ) rcommonAllocDmaLowVerts( &rmesa->radeon, n, size * 4 )
 #define LOCAL_VARS						\
    r300ContextPtr rmesa = R300_CONTEXT(ctx);		\
-   const char *r300verts = (char *)rmesa->swtcl.verts;
+   const char *r300verts = (char *)rmesa->radeon.swtcl.verts;
 #define VERT(x) (r300Vertex *)(r300verts + ((x) * vertsize * sizeof(int)))
 #define VERTEX r300Vertex
 #define DO_DEBUG_VERTS (1 && (RADEON_DEBUG & DEBUG_VERTS))
@@ -413,7 +335,7 @@ static struct {
 #define VERT_Y(_v) _v->v.y
 #define VERT_Z(_v) _v->v.z
 #define AREA_IS_CCW( a ) (a < 0)
-#define GET_VERTEX(e) (rmesa->swtcl.verts + (e*rmesa->swtcl.vertex_size*sizeof(int)))
+#define GET_VERTEX(e) (rmesa->radeon.swtcl.verts + (e*rmesa->radeon.swtcl.vertex_size*sizeof(int)))
 
 /* Only used to pull back colors into vertices (ie, we know color is
  * floating point).
@@ -459,7 +381,7 @@ do {							\
  ***********************************************************************/
 
 #define RASTERIZE(x) r300RasterPrimitive( ctx, reduced_prim[x] )
-#define RENDER_PRIMITIVE rmesa->swtcl.render_primitive
+#define RENDER_PRIMITIVE rmesa->radeon.swtcl.render_primitive
 #undef TAG
 #define TAG(x) x
 #include "tnl_dd/t_dd_unfilled.h"
@@ -516,8 +438,8 @@ static void init_rast_tab( void )
 #undef LOCAL_VARS
 #define LOCAL_VARS						\
    r300ContextPtr rmesa = R300_CONTEXT(ctx);		\
-   const GLuint vertsize = rmesa->swtcl.vertex_size;		\
-   const char *r300verts = (char *)rmesa->swtcl.verts;		\
+   const GLuint vertsize = rmesa->radeon.swtcl.vertex_size;		\
+   const char *r300verts = (char *)rmesa->radeon.swtcl.verts;		\
    const GLuint * const elt = TNL_CONTEXT(ctx)->vb.Elts;	\
    const GLboolean stipple = ctx->Line.StippleFlag;		\
    (void) elt; (void) stipple;
@@ -549,7 +471,7 @@ static void r300ChooseRenderState( GLcontext *ctx )
 	if (flags & DD_TRI_LIGHT_TWOSIDE) index |= R300_TWOSIDE_BIT;
 	if (flags & DD_TRI_UNFILLED)      index |= R300_UNFILLED_BIT;
 
-	if (index != rmesa->swtcl.RenderIndex) {
+	if (index != rmesa->radeon.swtcl.RenderIndex) {
 		tnl->Driver.Render.Points = rast_tab[index].points;
 		tnl->Driver.Render.Line = rast_tab[index].line;
 		tnl->Driver.Render.ClippedLine = rast_tab[index].line;
@@ -566,7 +488,7 @@ static void r300ChooseRenderState( GLcontext *ctx )
 			tnl->Driver.Render.ClippedPolygon = _tnl_RenderClippedPolygon;
 		}
 
-		rmesa->swtcl.RenderIndex = index;
+		rmesa->radeon.swtcl.RenderIndex = index;
 	}
 }
 
@@ -598,9 +520,9 @@ static void r300RasterPrimitive( GLcontext *ctx, GLuint hwprim )
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 
-	if (rmesa->swtcl.hw_primitive != hwprim) {
+	if (rmesa->radeon.swtcl.hw_primitive != hwprim) {
 	        R300_NEWPRIM( rmesa );
-		rmesa->swtcl.hw_primitive = hwprim;
+		rmesa->radeon.swtcl.hw_primitive = hwprim;
 	}
 }
 
@@ -608,7 +530,7 @@ static void r300RenderPrimitive(GLcontext *ctx, GLenum prim)
 {
 
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	rmesa->swtcl.render_primitive = prim;
+	rmesa->radeon.swtcl.render_primitive = prim;
 
 	if ((prim == GL_TRIANGLES) && (ctx->_TriangleCaps & DD_TRI_UNFILLED))
 	  return;
@@ -647,10 +569,10 @@ void r300InitSwtcl(GLcontext *ctx)
 	_tnl_init_vertices( ctx, ctx->Const.MaxArrayLockSize + 12,
 			    48 * sizeof(GLfloat) );
 
-	rmesa->swtcl.verts = (GLubyte *)tnl->clipspace.vertex_buf;
-	rmesa->swtcl.RenderIndex = ~0;
-	rmesa->swtcl.render_primitive = GL_TRIANGLES;
-	rmesa->swtcl.hw_primitive = 0;
+	rmesa->radeon.swtcl.verts = (GLubyte *)tnl->clipspace.vertex_buf;
+	rmesa->radeon.swtcl.RenderIndex = ~0;
+	rmesa->radeon.swtcl.render_primitive = GL_TRIANGLES;
+	rmesa->radeon.swtcl.hw_primitive = 0;
 
 	_tnl_invalidate_vertex_state( ctx, ~0 );
 	_tnl_invalidate_vertices( ctx, ~0 );
@@ -698,3 +620,24 @@ void r300EmitVbufPrim(r300ContextPtr rmesa, GLuint primitive, GLuint vertex_nr)
 	OUT_BATCH(R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_LIST | (num_verts << 16) | type);
 	END_BATCH();
 }
+
+void r300_swtcl_flush(GLcontext *ctx, uint32_t current_offset)
+{
+  r300ContextPtr rmesa = R300_CONTEXT(ctx);
+
+  rcommonEnsureCmdBufSpace(&rmesa->radeon,
+			   rmesa->hw.max_state_size + (12*sizeof(int)),
+			   __FUNCTION__);
+  r300EmitState(rmesa);
+  r300EmitVertexAOS(rmesa,
+		    rmesa->radeon.swtcl.vertex_size,
+		    rmesa->radeon.dma.current,
+		    current_offset);
+  
+  r300EmitVbufPrim(rmesa,
+		   rmesa->radeon.swtcl.hw_primitive,
+		   rmesa->radeon.swtcl.numverts);
+  r300EmitCacheFlush(rmesa);
+  COMMIT_BATCH();
+
+}
diff --git a/src/mesa/drivers/dri/r300/r300_swtcl.h b/src/mesa/drivers/dri/r300/r300_swtcl.h
index 55df53c1ad..23b4ce3469 100644
--- a/src/mesa/drivers/dri/r300/r300_swtcl.h
+++ b/src/mesa/drivers/dri/r300/r300_swtcl.h
@@ -42,4 +42,5 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 extern void r300InitSwtcl( GLcontext *ctx );
 extern void r300DestroySwtcl( GLcontext *ctx );
 
+extern void r300_swtcl_flush(GLcontext *ctx, uint32_t current_offset);
 #endif
diff --git a/src/mesa/drivers/dri/radeon/common_context.h b/src/mesa/drivers/dri/radeon/common_context.h
index 1ed33e2aec..90abca0bc1 100644
--- a/src/mesa/drivers/dri/radeon/common_context.h
+++ b/src/mesa/drivers/dri/radeon/common_context.h
@@ -11,6 +11,7 @@
 #include "radeon_screen.h"
 #include "radeon_drm.h"
 #include "dri_util.h"
+#include "tnl/t_vertex.h"
 
 /* This union is used to avoid warnings/miscompilation
    with float to uint32_t casts due to strict-aliasing */
@@ -250,13 +251,19 @@ struct radeon_dma {
 /* radeon_swtcl.c
  */
 struct radeon_swtcl_info {
-   struct radeon_bo *bo;
 
-   /* Fallback rasterization functions
-    */
-   GLuint hw_primitive;
-   GLenum render_primitive;
-   GLuint numverts;
+	GLuint RenderIndex;
+	GLuint vertex_size;
+	GLubyte *verts;
+
+	/* Fallback rasterization functions
+	 */
+	GLuint hw_primitive;
+	GLenum render_primitive;
+	GLuint numverts;
+
+	struct tnl_attr_map vertex_attrs[VERT_ATTRIB_MAX];
+	GLuint vertex_attr_count;
 
 };
 
@@ -416,7 +423,7 @@ struct radeon_context {
    /* Derived state - for r300 only */
    struct radeon_state state;
 
-   struct radeon_swtcl swtcl;
+   struct radeon_swtcl_info swtcl;
    /* Configuration cache
     */
    driOptionCache optionCache;
@@ -424,14 +431,15 @@ struct radeon_context {
    struct radeon_cmdbuf cmdbuf;
 
    struct {
-      void (*get_lock)(radeonContextPtr radeon);
-      void (*update_viewport_offset)(GLcontext *ctx);
-      void (*flush)(GLcontext *ctx);
-      void (*set_all_dirty)(GLcontext *ctx);
-      void (*update_draw_buffer)(GLcontext *ctx);
-      void (*emit_cs_header)(struct radeon_cs *cs, radeonContextPtr rmesa);
-      void (*emit_state)(radeonContextPtr rmesa);
-      void (*flush_vertices)(radeonContextPtr rmesa);
+	   void (*get_lock)(radeonContextPtr radeon);
+	   void (*update_viewport_offset)(GLcontext *ctx);
+	   void (*flush)(GLcontext *ctx);
+	   void (*set_all_dirty)(GLcontext *ctx);
+	   void (*update_draw_buffer)(GLcontext *ctx);
+	   void (*emit_cs_header)(struct radeon_cs *cs, radeonContextPtr rmesa);
+	   void (*emit_state)(radeonContextPtr rmesa);
+	   void (*flush_vertices)(radeonContextPtr rmesa);
+	   void (*swtcl_flush)(GLcontext *ctx, uint32_t offset);
    } vtbl;
 };
 
diff --git a/src/mesa/drivers/dri/radeon/common_misc.c b/src/mesa/drivers/dri/radeon/common_misc.c
index 1adcefb19a..9044967ac0 100644
--- a/src/mesa/drivers/dri/radeon/common_misc.c
+++ b/src/mesa/drivers/dri/radeon/common_misc.c
@@ -2420,46 +2420,46 @@ void radeonAllocDmaRegion(radeonContextPtr rmesa,
 
 void radeonReleaseDmaRegion(radeonContextPtr rmesa)
 {
-	rmesa->dma.nr_released_bufs++;
-	radeon_bo_unref(rmesa->dma.current);
+	if (rmesa->dma.current) {
+		rmesa->dma.nr_released_bufs++;
+		radeon_bo_unref(rmesa->dma.current);
+	}
 	rmesa->dma.current = NULL;
 }
 
-void rcommonEmitVertexAOS(radeonContextPtr rmesa, GLuint vertex_size, struct radeon_bo *bo, GLuint offset)
+
+/* Flush vertices in the current dma region.
+ */
+void rcommon_flush_last_swtcl_prim( GLcontext *ctx  )
 {
-	BATCH_LOCALS(rmesa);
+	radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+	struct radeon_dma *dma = &rmesa->dma;
+		
 
-	if (RADEON_DEBUG & DEBUG_VERTS)
-		fprintf(stderr, "%s:  vertex_size %d, offset 0x%x \n",
-			__FUNCTION__, vertex_size, offset);
+	if (RADEON_DEBUG & DEBUG_IOCTL)
+		fprintf(stderr, "%s\n", __FUNCTION__);
+	dma->flush = NULL;
 
-	BEGIN_BATCH(5);
-	OUT_BATCH_PACKET3(R300_PACKET3_3D_LOAD_VBPNTR, 2);
-	OUT_BATCH(1);
-	OUT_BATCH(vertex_size | (vertex_size << 8));
-	OUT_BATCH_RELOC(offset, bo, offset, RADEON_GEM_DOMAIN_GTT, 0, 0);
-	END_BATCH();
-}
+	if (dma->current) {
+	    GLuint current_offset = dma->current_used;
 
-void rcommonEmitVbufPrim(radeonContextPtr rmesa, GLuint primitive, GLuint vertex_nr)
-{
-	BATCH_LOCALS(rmesa);
-	int type, num_verts;
+	    assert (dma->current_used +
+		    rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 ==
+		    dma->current_vertexptr);
 
-	type = r300PrimitiveType(rmesa, primitive);
-	num_verts = r300NumVerts(rmesa, vertex_nr, primitive);
+	    radeon_bo_unmap(dma->current);
+	    if (dma->current_used != dma->current_vertexptr) {
+		    dma->current_used = dma->current_vertexptr;
 
-	BEGIN_BATCH(3);
-	OUT_BATCH_PACKET3(R300_PACKET3_3D_DRAW_VBUF_2, 0);
-	OUT_BATCH(R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_LIST | (num_verts << 16) | type);
-	END_BATCH();
+		    rmesa->vtbl.swtcl_flush(ctx, current_offset);
+	    }
+	    radeonReleaseDmaRegion(rmesa);
+	    rmesa->swtcl.numverts = 0;
+	}
 }
-			    
-
-
 /* Alloc space in the current dma region.
  */
-static void *
+void *
 rcommonAllocDmaLowVerts( radeonContextPtr rmesa, int nverts, int vsize )
 {
 	GLuint bytes = vsize * nverts;
@@ -2471,19 +2471,19 @@ rcommonAllocDmaLowVerts( radeonContextPtr rmesa, int nverts, int vsize )
 
         if (!rmesa->dma.flush) {
                 rmesa->glCtx->Driver.NeedFlush |= FLUSH_STORED_VERTICES;
-                rmesa->dma.flush = flush_last_swtcl_prim;
+                rmesa->dma.flush = rcommon_flush_last_swtcl_prim;
         }
 
 	ASSERT( vsize == rmesa->swtcl.vertex_size * 4 );
-        ASSERT( rmesa->radeon.dma.flush == flush_last_swtcl_prim );
-        ASSERT( rmesa->radeon.dma.current_used +
+        ASSERT( rmesa->dma.flush == rcommon_flush_last_swtcl_prim );
+        ASSERT( rmesa->dma.current_used +
                 rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 ==
-                rmesa->radeon.dma.current_vertexptr );
+                rmesa->dma.current_vertexptr );
 
 //	fprintf(stderr,"current %p %x\n", rmesa->radeon.dma.current->ptr,
 //		rmesa->radeon.dma.current_vertexptr);
-	head = (rmesa->radeon.dma.current->ptr + rmesa->radeon.dma.current_vertexptr);
-	rmesa->radeon.dma.current_vertexptr += bytes;
+	head = (rmesa->dma.current->ptr + rmesa->dma.current_vertexptr);
+	rmesa->dma.current_vertexptr += bytes;
 	rmesa->swtcl.numverts += nverts;
 	return head;
 }
diff --git a/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c b/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c
index 171818db37..5c6d93f573 100644
--- a/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c
+++ b/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c
@@ -44,6 +44,8 @@ struct cs_manager_legacy {
     /* hack for scratch stuff */
     uint32_t                    pending_age;
     uint32_t                    pending_count;
+
+
 };
 
 struct cs_reloc_legacy {
@@ -507,3 +509,4 @@ void radeon_cs_manager_legacy_dtor(struct radeon_cs_manager *csm)
 {
     free(csm);
 }
+
diff --git a/src/mesa/drivers/dri/radeon/radeon_screen.c b/src/mesa/drivers/dri/radeon/radeon_screen.c
index 320bf9244b..4ff4267348 100644
--- a/src/mesa/drivers/dri/radeon/radeon_screen.c
+++ b/src/mesa/drivers/dri/radeon/radeon_screen.c
@@ -1015,6 +1015,7 @@ radeonCreateScreen( __DRIscreenPrivate *sPriv )
      free(screen);
      return NULL;
    }
+
    return screen;
 }