r200: emit elts into a separate ELT bo

6 files changed, 89 insertions, 408 deletions

diff --git a/src/mesa/drivers/dri/r200/r200_cmdbuf.c b/src/mesa/drivers/dri/r200/r200_cmdbuf.c
index 6d421f5d5ec..5768f39e5c1 100644
--- a/src/mesa/drivers/dri/r200/r200_cmdbuf.c
+++ b/src/mesa/drivers/dri/r200/r200_cmdbuf.c
@@ -238,40 +238,61 @@ void r200EmitVbufPrim( r200ContextPtr rmesa,
    OUT_BATCH(primitive | R200_VF_PRIM_WALK_LIST | R200_VF_COLOR_ORDER_RGBA |
 	     (vertex_nr << R200_VF_VERTEX_NUMBER_SHIFT));
    END_BATCH();
-#if 0
-   cmd = (drm_radeon_cmd_header_t *)r200AllocCmdBuf( rmesa, VBUF_BUFSZ,
-						  __FUNCTION__ );
-   cmd[0].i = 0;
-   cmd[0].header.cmd_type = RADEON_CMD_PACKET3_CLIP;
-   cmd[1].i = R200_CP_CMD_3D_DRAW_VBUF_2;
-   cmd[2].i = (primitive | 
-	       R200_VF_PRIM_WALK_LIST |
-	       R200_VF_COLOR_ORDER_RGBA |
-	       (vertex_nr << R200_VF_VERTEX_NUMBER_SHIFT));
-#endif
 }
 
+static void r200FireEB(r200ContextPtr rmesa, int vertex_count, int type)
+{
+	BATCH_LOCALS(&rmesa->radeon);
+
+	if (vertex_count > 0) {
+		BEGIN_BATCH(8);
+		OUT_BATCH_PACKET3(R200_CP_CMD_3D_DRAW_INDX_2, 0);
+		OUT_BATCH(R200_VF_PRIM_WALK_IND |
+			  ((vertex_count + 0) << 16) |
+			  type);
+		
+		if (!rmesa->radeon.radeonScreen->kernel_mm) {
+			OUT_BATCH_PACKET3(R200_CP_CMD_INDX_BUFFER, 2);
+			OUT_BATCH((0x80 << 24) | (0 << 16) | 0x810);
+			OUT_BATCH_RELOC(rmesa->tcl.elt_dma_offset,
+					rmesa->tcl.elt_dma_bo,
+					rmesa->tcl.elt_dma_offset,
+					RADEON_GEM_DOMAIN_GTT, 0, 0);
+			OUT_BATCH(vertex_count/2);
+		} else {
+			OUT_BATCH_PACKET3(R200_CP_CMD_INDX_BUFFER, 2);
+			OUT_BATCH((0x80 << 24) | (0 << 16) | 0x810);
+			OUT_BATCH(rmesa->tcl.elt_dma_offset);
+			OUT_BATCH(vertex_count/2);
+			radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs,
+					      rmesa->tcl.elt_dma_bo,
+					      RADEON_GEM_DOMAIN_GTT, 0, 0);
+		}
+		END_BATCH();
+	}
+}
 
-void r200FlushElts( GLcontext *ctx )
+void r200FlushElts( r200ContextPtr rmesa )
 {
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   int *cmd = (int *)(rmesa->store.cmd_buf + rmesa->store.elts_start);
    int dwords;
-   int nr = (rmesa->store.cmd_used - (rmesa->store.elts_start + 12)) / 2;
+   int nr, elt_used = rmesa->tcl.elt_used;
 
    if (R200_DEBUG & (DEBUG_IOCTL|DEBUG_PRIMS))
-      fprintf(stderr, "%s\n", __FUNCTION__);
+     fprintf(stderr, "%s %x %d\n", __FUNCTION__, rmesa->tcl.hw_primitive, elt_used);
 
-   assert( rmesa->dma.flush == r200FlushElts );
-   rmesa->dma.flush = NULL;
+   assert( rmesa->tcl.flush == r200FlushElts );
+   rmesa->tcl.flush = NULL;
 
-   /* Cope with odd number of elts:
-    */
-   rmesa->store.cmd_used = (rmesa->store.cmd_used + 2) & ~2;
-   dwords = (rmesa->store.cmd_used - rmesa->store.elts_start) / 4;
+   elt_used = (elt_used + 2) & ~2;
 
-   cmd[1] |= (dwords - 3) << 16;
-   cmd[2] |= nr << R200_VF_VERTEX_NUMBER_SHIFT;
+   nr = elt_used / 2;
+
+   radeon_bo_unmap(rmesa->tcl.elt_dma_bo);
+
+   r200FireEB(rmesa, nr, rmesa->tcl.hw_primitive);
+
+   radeon_bo_unref(rmesa->tcl.elt_dma_bo);
+   rmesa->tcl.elt_dma_bo = NULL;
 
    if (R200_DEBUG & DEBUG_SYNC) {
       fprintf(stderr, "%s: Syncing\n", __FUNCTION__);
@@ -284,7 +305,6 @@ GLushort *r200AllocEltsOpenEnded( r200ContextPtr rmesa,
 				    GLuint primitive,
 				    GLuint min_nr )
 {
-   drm_radeon_cmd_header_t *cmd;
    GLushort *retval;
 
    if (R200_DEBUG & DEBUG_IOCTL)
@@ -293,29 +313,24 @@ GLushort *r200AllocEltsOpenEnded( r200ContextPtr rmesa,
    assert((primitive & R200_VF_PRIM_WALK_IND));
    
    r200EmitState( rmesa );
-   
-   //   cmd = (drm_radeon_cmd_header_t *)r200AllocCmdBuf( rmesa, ELTS_BUFSZ(min_nr),
-   //						__FUNCTION__ );
-   cmd[0].i = 0;
-   cmd[0].header.cmd_type = RADEON_CMD_PACKET3_CLIP;
-   cmd[1].i = R200_CP_CMD_3D_DRAW_INDX_2;
-   cmd[2].i = (primitive | 
-	       R200_VF_PRIM_WALK_IND |
-	       R200_VF_COLOR_ORDER_RGBA);
 
+   rmesa->tcl.elt_dma_bo = radeon_bo_open(rmesa->radeon.radeonScreen->bom,
+					  0, R200_ELT_BUF_SZ, 4,
+					  RADEON_GEM_DOMAIN_GTT, 0);
+   rmesa->tcl.elt_dma_offset = 0;
+   rmesa->tcl.elt_used = min_nr * 2;
+
+   radeon_bo_map(rmesa->tcl.elt_dma_bo, 1);
+   retval = rmesa->tcl.elt_dma_bo->ptr + rmesa->tcl.elt_dma_offset;
    
-   retval = (GLushort *)(cmd+3);
 
    if (R200_DEBUG & DEBUG_PRIMS)
-      fprintf(stderr, "%s: header 0x%x prim %x \n",
-	      __FUNCTION__,
-	      cmd[1].i, primitive);
+      fprintf(stderr, "%s: header prim %x \n",
+	      __FUNCTION__, primitive);
 
-   assert(!rmesa->dma.flush);
+   assert(!rmesa->tcl.flush);
    rmesa->radeon.glCtx->Driver.NeedFlush |= FLUSH_STORED_VERTICES;
-   rmesa->dma.flush = r200FlushElts;
-
-   rmesa->store.elts_start = ((char *)cmd) - rmesa->store.cmd_buf;
+   rmesa->tcl.flush = r200FlushElts;
 
    return retval;
 }
@@ -439,112 +454,14 @@ void r200EmitAOS(r200ContextPtr rmesa, GLuint nr, GLuint offset)
    }
    END_BATCH();
 }
-		       
-#if 0
-void r200EmitAOS( r200ContextPtr rmesa,
-		    struct radeon_dma_region **component,
-		    GLuint nr,
-		    GLuint offset )
-{
-   drm_radeon_cmd_header_t *cmd;
-   int sz = AOS_BUFSZ(nr);
-   int i;
-   int *tmp;
 
-   if (R200_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s nr arrays: %d\n", __FUNCTION__, nr);
-
-   cmd = (drm_radeon_cmd_header_t *)r200AllocCmdBuf( rmesa, sz, __FUNCTION__ );
-   cmd[0].i = 0;
-   cmd[0].header.cmd_type = RADEON_CMD_PACKET3;
-   cmd[1].i = R200_CP_CMD_3D_LOAD_VBPNTR | (((sz / sizeof(int)) - 3) << 16);
-   cmd[2].i = nr;
-   tmp = &cmd[0].i;
-   cmd += 3;
-
-   for (i = 0 ; i < nr ; i++) {
-      if (i & 1) {
-	 cmd[0].i |= ((component[i]->aos_stride << 24) | 
-		      (component[i]->aos_size << 16));
-	 cmd[2].i = (component[i]->aos_start + 
-		     offset * component[i]->aos_stride * 4);
-	 cmd += 3;
-      }
-      else {
-	 cmd[0].i = ((component[i]->aos_stride << 8) | 
-		     (component[i]->aos_size << 0));
-	 cmd[1].i = (component[i]->aos_start + 
-		     offset * component[i]->aos_stride * 4);
-      }
-   }
-
-   if (R200_DEBUG & DEBUG_VERTS) {
-      fprintf(stderr, "%s:\n", __FUNCTION__);
-      for (i = 0 ; i < sz ; i++)
-	 fprintf(stderr, "   %d: %x\n", i, tmp[i]);
-   }
-}
-#endif
-
-void r200EmitBlit( r200ContextPtr rmesa,
-		   GLuint color_fmt,
-		   GLuint src_pitch,
-		   GLuint src_offset,
-		   GLuint dst_pitch,
-		   GLuint dst_offset,
-		   GLint srcx, GLint srcy,
-		   GLint dstx, GLint dsty,
-		   GLuint w, GLuint h )
+void r200FireAOS(r200ContextPtr rmesa, int vertex_count, int type)
 {
-   drm_radeon_cmd_header_t *cmd;
+	BATCH_LOCALS(&rmesa->radeon);
 
-   if (R200_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s src %x/%x %d,%d dst: %x/%x %d,%d sz: %dx%d\n",
-	      __FUNCTION__, 
-	      src_pitch, src_offset, srcx, srcy,
-	      dst_pitch, dst_offset, dstx, dsty,
-	      w, h);
-
-   assert( (src_pitch & 63) == 0 );
-   assert( (dst_pitch & 63) == 0 );
-   assert( (src_offset & 1023) == 0 );
-   assert( (dst_offset & 1023) == 0 );
-   assert( w < (1<<16) );
-   assert( h < (1<<16) );
-
-   //   cmd = (drm_radeon_cmd_header_t *)r200AllocCmdBuf( rmesa, 8 * sizeof(int),
-   //						  __FUNCTION__ );
-
-
-   cmd[0].header.cmd_type = RADEON_CMD_PACKET3;
-   cmd[1].i = R200_CP_CMD_BITBLT_MULTI | (5 << 16);
-   cmd[2].i = (RADEON_GMC_SRC_PITCH_OFFSET_CNTL |
-	       RADEON_GMC_DST_PITCH_OFFSET_CNTL |
-	       RADEON_GMC_BRUSH_NONE |
-	       (color_fmt << 8) |
-	       RADEON_GMC_SRC_DATATYPE_COLOR |
-	       RADEON_ROP3_S |
-	       RADEON_DP_SRC_SOURCE_MEMORY |
-	       RADEON_GMC_CLR_CMP_CNTL_DIS |
-	       RADEON_GMC_WR_MSK_DIS );
-
-   cmd[3].i = ((src_pitch/64)<<22) | (src_offset >> 10);
-   cmd[4].i = ((dst_pitch/64)<<22) | (dst_offset >> 10);
-   cmd[5].i = (srcx << 16) | srcy;
-   cmd[6].i = (dstx << 16) | dsty; /* dst */
-   cmd[7].i = (w << 16) | h;
+	BEGIN_BATCH(3);
+	OUT_BATCH_PACKET3(R200_CP_CMD_3D_DRAW_VBUF_2, 0);
+	OUT_BATCH(R200_VF_PRIM_WALK_LIST | (vertex_count << 16) | type);
+	END_BATCH();
 }
 
-
-void r200EmitWait( r200ContextPtr rmesa, GLuint flags )
-{
-   drm_radeon_cmd_header_t *cmd;
-
-   assert( !(flags & ~(RADEON_WAIT_2D|RADEON_WAIT_3D)) );
-
-   //   cmd = (drm_radeon_cmd_header_t *)r200AllocCmdBuf( rmesa, 1 * sizeof(int),
-   //					   __FUNCTION__ );
-   cmd[0].i = 0;
-   cmd[0].wait.cmd_type = RADEON_CMD_WAIT;
-   cmd[0].wait.flags = flags;
-}
diff --git a/src/mesa/drivers/dri/r200/r200_context.h b/src/mesa/drivers/dri/r200/r200_context.h
index 2725dc495f6..607997fafc9 100644
--- a/src/mesa/drivers/dri/r200/r200_context.h
+++ b/src/mesa/drivers/dri/r200/r200_context.h
@@ -531,7 +531,7 @@ struct r200_state {
 
 #define R200_CMD_BUF_SZ  (16*1024) 
 
-
+#define R200_ELT_BUF_SZ  (16*1024) 
 /* r200_tcl.c
  */
 struct r200_tcl_info {
@@ -544,6 +544,11 @@ struct r200_tcl_info {
 
    GLuint *Elts;
 
+   struct radeon_bo *elt_dma_bo;
+   int elt_dma_offset; /** Offset into this buffer object, in bytes */
+   int elt_used;
+
+   void (*flush) (r200ContextPtr);
    struct radeon_dma_region vertex_data[15];
 };
 
diff --git a/src/mesa/drivers/dri/r200/r200_ioctl.c b/src/mesa/drivers/dri/r200/r200_ioctl.c
index 1fbbbfbb569..65786be96cb 100644
--- a/src/mesa/drivers/dri/r200/r200_ioctl.c
+++ b/src/mesa/drivers/dri/r200/r200_ioctl.c
@@ -85,252 +85,6 @@ static void r200BackUpAndEmitLostStateLocked( r200ContextPtr rmesa )
    rmesa->store = saved_store;
 }
 
-#if 0
-int r200FlushCmdBufLocked( r200ContextPtr rmesa, const char * caller )
-{
-   int ret, i;
-   drm_radeon_cmd_buffer_t cmd;
-
-   if (rmesa->radeon.lost_context)
-      r200BackUpAndEmitLostStateLocked( rmesa );
-
-   if (R200_DEBUG & DEBUG_IOCTL) {
-      fprintf(stderr, "%s from %s\n", __FUNCTION__, caller); 
-
-      if (0 & R200_DEBUG & DEBUG_VERBOSE) 
-	 for (i = 0 ; i < rmesa->store.cmd_used ; i += 4 )
-	    fprintf(stderr, "%d: %x\n", i/4, 
-		    *(int *)(&rmesa->store.cmd_buf[i]));
-   }
-
-   if (R200_DEBUG & DEBUG_DMA)
-      fprintf(stderr, "%s: Releasing %d buffers\n", __FUNCTION__,
-	      rmesa->dma.nr_released_bufs);
-
-
-   if (R200_DEBUG & DEBUG_SANITY) {
-      if (rmesa->radeon.state.scissor.enabled) 
-	 ret = r200SanityCmdBuffer( rmesa, 
-				    rmesa->radeon.state.scissor.numClipRects,
-				    rmesa->radeon.state.scissor.pClipRects);
-      else
-	 ret = r200SanityCmdBuffer( rmesa, 
-				    rmesa->radeon.numClipRects,
-				    rmesa->radeon.pClipRects);
-      if (ret) {
-	 fprintf(stderr, "drmSanityCommandWrite: %d\n", ret);	 
-	 goto out;
-      }
-   }
-
-
-   if (R200_DEBUG & DEBUG_MEMORY) {
-      if (! driValidateTextureHeaps( rmesa->radeon.texture_heaps, rmesa->radeon.nr_heaps,
-				     & rmesa->radeon.swapped ) ) {
-	 fprintf( stderr, "%s: texture memory is inconsistent - expect "
-		  "mangled textures\n", __FUNCTION__ );
-      }
-   }
-
-
-   cmd.bufsz = rmesa->store.cmd_used;
-   cmd.buf = rmesa->store.cmd_buf;
-
-   if (rmesa->radeon.state.scissor.enabled) {
-      cmd.nbox = rmesa->radeon.state.scissor.numClipRects;
-      cmd.boxes = (drm_clip_rect_t *)rmesa->radeon.state.scissor.pClipRects;
-   } else {
-      cmd.nbox = rmesa->radeon.numClipRects;
-      cmd.boxes = (drm_clip_rect_t *)rmesa->radeon.pClipRects;
-   }
-
-   ret = drmCommandWrite( rmesa->radeon.dri.fd,
-			  DRM_RADEON_CMDBUF,
-			  &cmd, sizeof(cmd) );
-
-   if (ret)
-      fprintf(stderr, "drmCommandWrite: %d\n", ret);
-
-   if (R200_DEBUG & DEBUG_SYNC) {
-      fprintf(stderr, "\nSyncing in %s\n\n", __FUNCTION__);
-      radeonWaitForIdleLocked( &rmesa->radeon );
-   }
-
-
- out:
-   rmesa->store.primnr = 0;
-   rmesa->store.statenr = 0;
-   rmesa->store.cmd_used = 0;
-   rmesa->dma.nr_released_bufs = 0;
-   rmesa->save_on_next_emit = 1;
-
-   return ret;
-}
-
-/* Note: does not emit any commands to avoid recursion on
- * r200AllocCmdBuf.
- */
-void r200FlushCmdBuf( r200ContextPtr rmesa, const char *caller )
-{
-   int ret;
-
-   LOCK_HARDWARE( &rmesa->radeon );
-
-   ret = rcommonFlushCmdBufLocked( rmesa, caller );
-
-   UNLOCK_HARDWARE( &rmesa->radeon );
-
-   if (ret) {
-      fprintf(stderr, "drmRadeonCmdBuffer: %d (exiting)\n", ret);
-      exit(ret);
-   }
-}
-#endif
-
-
-/* =============================================================
- * Hardware vertex buffer handling
- */
-
-#if 0
-void r200RefillCurrentDmaRegion( r200ContextPtr rmesa )
-{
-   struct radeon_dma_buffer *dmabuf;
-   int fd = rmesa->radeon.dri.fd;
-   int index = 0;
-   int size = 0;
-   drmDMAReq dma;
-   int ret;
-
-   if (R200_DEBUG & (DEBUG_IOCTL|DEBUG_DMA))
-      fprintf(stderr, "%s\n", __FUNCTION__);  
-
-   if (rmesa->dma.flush) {
-      rmesa->dma.flush( rmesa->radeon.glCtx );
-   }
-
-   if (rmesa->dma.current.buf)
-      r200ReleaseDmaRegion( rmesa, &rmesa->dma.current, __FUNCTION__ );
-
-   if (rmesa->dma.nr_released_bufs > 4)
-      rcommonFlushCmdBuf( &rmesa->radeon, __FUNCTION__ );
-
-   dma.context = rmesa->radeon.dri.hwContext;
-   dma.send_count = 0;
-   dma.send_list = NULL;
-   dma.send_sizes = NULL;
-   dma.flags = 0;
-   dma.request_count = 1;
-   dma.request_size = RADEON_BUFFER_SIZE;
-   dma.request_list = &index;
-   dma.request_sizes = &size;
-   dma.granted_count = 0;
-
-   LOCK_HARDWARE(&rmesa->radeon);	/* no need to validate */
-
-   while (1) {
-      ret = drmDMA( fd, &dma );
-      if (ret == 0)
-	 break;
-   
-      if (rmesa->dma.nr_released_bufs) {
-	 rcommonFlushCmdBuf( &rmesa->radeon, __FUNCTION__ );
-      }
-
-      if (rmesa->radeon.do_usleeps) {
-	 UNLOCK_HARDWARE( &rmesa->radeon );
-	 DO_USLEEP( 1 );
-	 LOCK_HARDWARE( &rmesa->radeon );
-      }
-   }
-
-   UNLOCK_HARDWARE(&rmesa->radeon);
-
-   if (R200_DEBUG & DEBUG_DMA)
-      fprintf(stderr, "Allocated buffer %d\n", index);
-
-   dmabuf = CALLOC_STRUCT( radeon_dma_buffer );
-   dmabuf->buf = &rmesa->radeon.radeonScreen->buffers->list[index];
-   dmabuf->refcount = 1;
-
-   rmesa->dma.current.buf = dmabuf;
-   rmesa->dma.current.address = dmabuf->buf->address;
-   rmesa->dma.current.end = dmabuf->buf->total;
-   rmesa->dma.current.start = 0;
-   rmesa->dma.current.ptr = 0;
-}
-
-void r200ReleaseDmaRegion( r200ContextPtr rmesa,
-			     struct radeon_dma_region *region,
-			     const char *caller )
-{
-   if (R200_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s from %s\n", __FUNCTION__, caller); 
-   
-   if (!region->buf)
-      return;
-
-   if (rmesa->dma.flush)
-      rmesa->dma.flush( rmesa->radeon.glCtx );
-
-   if (--region->buf->refcount == 0) {
-      drm_radeon_cmd_header_t *cmd;
-
-      if (R200_DEBUG & (DEBUG_IOCTL|DEBUG_DMA))
-	 fprintf(stderr, "%s -- DISCARD BUF %d\n", __FUNCTION__,
-		 region->buf->buf->idx);  
-      
-      cmd = (drm_radeon_cmd_header_t *)r200AllocCmdBuf( rmesa, sizeof(*cmd), 
-						     __FUNCTION__ );
-      cmd->dma.cmd_type = RADEON_CMD_DMA_DISCARD;
-      cmd->dma.buf_idx = region->buf->buf->idx;
-      FREE(region->buf);
-      rmesa->dma.nr_released_bufs++;
-   }
-
-   region->buf = NULL;
-   region->start = 0;
-}
-
-/* Allocates a region from rmesa->dma.current.  If there isn't enough
- * space in current, grab a new buffer (and discard what was left of current)
- */
-void r200AllocDmaRegion( r200ContextPtr rmesa, 
-			   struct radeon_dma_region *region,
-			   int bytes,
-			   int alignment )
-{
-   if (R200_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s %d\n", __FUNCTION__, bytes);
-
-   if (rmesa->dma.flush)
-      rmesa->dma.flush( rmesa->radeon.glCtx );
-
-   if (region->buf)
-      r200ReleaseDmaRegion( rmesa, region, __FUNCTION__ );
-
-   alignment--;
-   rmesa->dma.current.start = rmesa->dma.current.ptr = 
-      (rmesa->dma.current.ptr + alignment) & ~alignment;
-
-   if ( rmesa->dma.current.ptr + bytes > rmesa->dma.current.end ) 
-      r200RefillCurrentDmaRegion( rmesa );
-
-   region->start = rmesa->dma.current.start;
-   region->ptr = rmesa->dma.current.start;
-   region->end = rmesa->dma.current.start + bytes;
-   region->address = rmesa->dma.current.address;
-   region->buf = rmesa->dma.current.buf;
-   region->buf->refcount++;
-
-   rmesa->dma.current.ptr += bytes; /* bug - if alignment > 7 */
-   rmesa->dma.current.start = 
-      rmesa->dma.current.ptr = (rmesa->dma.current.ptr + 0x7) & ~0x7;  
-
-   assert( rmesa->dma.current.ptr <= rmesa->dma.current.end );
-}
-#endif
-
 /* ================================================================
  * Buffer clear
  */
@@ -526,6 +280,9 @@ void r200Flush( GLcontext *ctx )
    if (rmesa->dma.flush)
       rmesa->dma.flush( ctx );
 
+   if (rmesa->tcl.flush)
+      rmesa->tcl.flush( rmesa );
+
    r200EmitState( rmesa );
 
    if (rmesa->radeon.cmdbuf.cs->cdw)
diff --git a/src/mesa/drivers/dri/r200/r200_ioctl.h b/src/mesa/drivers/dri/r200/r200_ioctl.h
index 1ae0dd114b5..e3f0b66514d 100644
--- a/src/mesa/drivers/dri/r200/r200_ioctl.h
+++ b/src/mesa/drivers/dri/r200/r200_ioctl.h
@@ -53,7 +53,8 @@ extern void r200EmitVbufPrim( r200ContextPtr rmesa,
 				GLuint primitive,
 				GLuint vertex_nr );
 
-extern void r200FlushElts(GLcontext *ctx);
+extern void r200FlushElts(r200ContextPtr rmesa);
+
 
 extern GLushort *r200AllocEltsOpenEnded( r200ContextPtr rmesa,
 					   GLuint primitive,
diff --git a/src/mesa/drivers/dri/r200/r200_pixel.c b/src/mesa/drivers/dri/r200/r200_pixel.c
index 930d797f45c..eefa2a1e11a 100644
--- a/src/mesa/drivers/dri/r200/r200_pixel.c
+++ b/src/mesa/drivers/dri/r200/r200_pixel.c
@@ -156,6 +156,8 @@ r200TryReadPixels( GLcontext *ctx,
    GLuint cpp = rmesa->radeon.radeonScreen->cpp;
    GLint size = width * height * cpp;
 
+   return GL_FALSE;
+#if 0
    if (R200_DEBUG & DEBUG_PIXEL)
       fprintf(stderr, "%s\n", __FUNCTION__);
 
@@ -165,7 +167,6 @@ r200TryReadPixels( GLcontext *ctx,
 			 pitch * height * rmesa->radeon.radeonScreen->cpp ) ) {
       if (R200_DEBUG & DEBUG_PIXEL)
 	 fprintf(stderr, "%s: dest not GART\n", __FUNCTION__);
-      return GL_FALSE;
    }
 
    /* Need GL_PACK_INVERT_MESA to cope with upsidedown results from
@@ -262,7 +263,7 @@ r200TryReadPixels( GLcontext *ctx,
    UNLOCK_HARDWARE( &rmesa->radeon );
 
    r200Finish( ctx ); /* required by GL */
-
+#endif
    return GL_TRUE;
 }
 
@@ -305,7 +306,7 @@ static void do_draw_pix( GLcontext *ctx,
 
    if (R200_DEBUG & DEBUG_PIXEL)
       fprintf(stderr, "%s\n", __FUNCTION__);
-
+#if 0
    switch ( rmesa->radeon.radeonScreen->cpp ) {
    case 2:
       blit_format = R200_CP_COLOR_FORMAT_RGB565;
@@ -366,6 +367,7 @@ static void do_draw_pix( GLcontext *ctx,
    rcommonFlushCmdBufLocked( &rmesa->radeon, __FUNCTION__ );
    radeonWaitForIdleLocked( &rmesa->radeon ); /* required by GL */
    UNLOCK_HARDWARE( &rmesa->radeon );
+#endif
 }
 
 
@@ -431,7 +433,7 @@ r200TryDrawPixels( GLcontext *ctx,
       return GL_FALSE;
    }
 
-   if ( r200IsGartMemory(rmesa, pixels, size) )
+   if (0)// r200IsGartMemory(rmesa, pixels, size) )
    {
       do_draw_pix( ctx, x, y, width, height, pitch, pixels, planemask );
       return GL_TRUE;
diff --git a/src/mesa/drivers/dri/r200/r200_tcl.c b/src/mesa/drivers/dri/r200/r200_tcl.c
index 0f15fd457e5..891a942e387 100644
--- a/src/mesa/drivers/dri/r200/r200_tcl.c
+++ b/src/mesa/drivers/dri/r200/r200_tcl.c
@@ -142,23 +142,22 @@ static GLboolean discrete_prim[0x10] = {
 
 static GLushort *r200AllocElts( r200ContextPtr rmesa, GLuint nr ) 
 {
-   fprintf(stderr,"alloc elts\n");
-   if (rmesa->dma.flush == r200FlushElts &&
-       rmesa->store.cmd_used + nr*2 < R200_CMD_BUF_SZ) {
+  fprintf(stderr,"alloc elts %d %d\n", nr, rmesa->tcl.elt_used);
+   if (rmesa->tcl.flush == r200FlushElts &&
+       rmesa->tcl.elt_used + nr*2 < R200_ELT_BUF_SZ) {
 
-      GLushort *dest = (GLushort *)(rmesa->store.cmd_buf +
-				    rmesa->store.cmd_used);
+      GLushort *dest = (GLushort *)(rmesa->tcl.elt_dma_bo->ptr +
+				    rmesa->tcl.elt_used);
 
-      rmesa->store.cmd_used += nr*2;
+      rmesa->tcl.elt_used += nr*2;
 
       return dest;
    }
    else {
-      if (rmesa->dma.flush)
-	 rmesa->dma.flush( rmesa->radeon.glCtx );
+      if (rmesa->tcl.flush)
+	 rmesa->tcl.flush( rmesa );
 
-      rcommonEnsureCmdBufSpace(rmesa, AOS_BUFSZ(rmesa->tcl.nr_aos_components) +
-			     rmesa->hw.max_state_size + ELTS_BUFSZ(nr) );
+      rcommonEnsureCmdBufSpace(rmesa, AOS_BUFSZ(rmesa->tcl.nr_aos_components));
 
       r200EmitAOS( rmesa,
 		   rmesa->tcl.nr_aos_components, 0 );