r300: add hw accelerated support for different vertex data formats

7 files changed, 485 insertions, 57 deletions

diff --git a/src/mesa/drivers/dri/r300/Makefile b/src/mesa/drivers/dri/r300/Makefile
index 62715e3b50d..bdb09624be4 100644
--- a/src/mesa/drivers/dri/r300/Makefile
+++ b/src/mesa/drivers/dri/r300/Makefile
@@ -37,6 +37,7 @@ RADEON_COMMON_SOURCES = \
 DRIVER_SOURCES = \
 		 radeon_screen.c \
 		 r300_context.c \
+		 r300_draw.c \
 		 r300_ioctl.c \
 		 r300_cmdbuf.c \
 		 r300_state.c \
diff --git a/src/mesa/drivers/dri/r300/r300_context.c b/src/mesa/drivers/dri/r300/r300_context.c
index dbd5ce589e3..b7911f23cce 100644
--- a/src/mesa/drivers/dri/r300/r300_context.c
+++ b/src/mesa/drivers/dri/r300/r300_context.c
@@ -153,10 +153,6 @@ const struct dri_extension gl_20_extension[] = {
 
 static const struct tnl_pipeline_stage *r300_pipeline[] = {
 
-	/* Try and go straight to t&l
-	 */
-	&_r300_tcl_stage,
-
 	/* Catch any t&l fallbacks
 	 */
 	&_tnl_vertex_transform_stage,
@@ -436,8 +432,11 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual,
 	_tnl_allow_pixel_fog(ctx, GL_FALSE);
 	_tnl_allow_vertex_fog(ctx, GL_TRUE);
 
-	if (!r300->options.hw_tcl_enabled)
+	if (r300->options.hw_tcl_enabled) {
+		r300InitDraw(ctx);
+	} else {
 		r300InitSwtcl(ctx);
+	}
 
 	radeon_fbo_init(&r300->radeon);
 	radeonInitSpanFuncs( ctx );
diff --git a/src/mesa/drivers/dri/r300/r300_context.h b/src/mesa/drivers/dri/r300/r300_context.h
index 7694fe4862f..11cfb55a9b5 100644
--- a/src/mesa/drivers/dri/r300/r300_context.h
+++ b/src/mesa/drivers/dri/r300/r300_context.h
@@ -617,6 +617,13 @@ struct r300_vertex_buffer {
 	GLubyte num_attribs;
 };
 
+struct r300_index_buffer {
+	GLvoid *ptr;
+	GLboolean is_32bit;
+	GLboolean free_needed;
+	GLuint count;
+};
+
 /**
  * \brief R300 context structure.
  */
@@ -644,6 +651,7 @@ struct r300_context {
 	
 	struct r300_swtcl_info swtcl;
 	struct r300_vertex_buffer vbuf;
+	struct r300_index_buffer ind_buf;
 	GLboolean vap_flush_needed;
 
 	uint32_t fallback;
@@ -666,6 +674,8 @@ extern int r300VertexProgUpdateParams(GLcontext * ctx,
 
 extern void r300InitShaderFunctions(r300ContextPtr r300);
 
+extern void r300InitDraw(GLcontext *ctx);
+
 #define r300PackFloat32 radeonPackFloat32
 #define r300PackFloat24 radeonPackFloat24
 
diff --git a/src/mesa/drivers/dri/r300/r300_draw.c b/src/mesa/drivers/dri/r300/r300_draw.c
new file mode 100644
index 00000000000..ba748787214
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/r300_draw.c
@@ -0,0 +1,435 @@
+/**************************************************************************
+ *
+ * Copyright 2009 Maciej Cencora
+ *
+ * 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 <stdlib.h>
+
+#include "main/glheader.h"
+#include "main/context.h"
+#include "main/state.h"
+#include "main/api_validate.h"
+#include "main/enums.h"
+
+#include "r300_reg.h"
+#include "r300_context.h"
+#include "r300_emit.h"
+#include "r300_render.h"
+#include "r300_state.h"
+#include "r300_tex.h"
+
+#include "tnl/tnl.h"
+#include "tnl/t_vp_build.h"
+#include "vbo/vbo_context.h"
+#include "swrast/swrast.h"
+#include "swrast_setup/swrast_setup.h"
+
+static void r300FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer *mesa_ind_buf, struct gl_buffer_object **bo, GLuint *nr_bo)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	struct r300_index_buffer *ind_buf = &r300->ind_buf;
+	GLvoid *src_ptr;
+
+	if (!mesa_ind_buf) {
+		ind_buf->ptr = NULL;
+		return;
+	}
+
+	ind_buf->count = mesa_ind_buf->count;
+	if (mesa_ind_buf->obj->Name && !mesa_ind_buf->obj->Pointer) {
+		bo[*nr_bo] = mesa_ind_buf->obj;
+		(*nr_bo)++;
+		ctx->Driver.MapBuffer(ctx, GL_ELEMENT_ARRAY_BUFFER, GL_READ_ONLY_ARB, mesa_ind_buf->obj);
+		assert(mesa_ind_buf->obj->Pointer != NULL);
+	}
+	src_ptr = ADD_POINTERS(mesa_ind_buf->obj->Pointer, mesa_ind_buf->ptr);
+
+	if (mesa_ind_buf->type == GL_UNSIGNED_BYTE) {
+		GLubyte *in = (GLubyte *)src_ptr;
+		GLuint *out = _mesa_malloc(sizeof(GLuint) * mesa_ind_buf->count);
+		int i;
+
+		for (i = 0; i < mesa_ind_buf->count; ++i) {
+			out[i] = (GLuint) in[i];
+		}
+
+		ind_buf->ptr = out;
+		ind_buf->free_needed = GL_TRUE;
+		ind_buf->is_32bit = GL_TRUE;
+	} else if (mesa_ind_buf->type == GL_UNSIGNED_SHORT) {
+		ind_buf->ptr = src_ptr;
+		ind_buf->free_needed = GL_FALSE;
+		ind_buf->is_32bit = GL_FALSE;
+	} else {
+		ind_buf->ptr = src_ptr;
+		ind_buf->free_needed = GL_FALSE;
+		ind_buf->is_32bit = GL_TRUE;
+	}
+}
+
+static int getTypeSize(GLenum type)
+{
+	switch (type) {
+		case GL_DOUBLE:
+			return sizeof(GLdouble);
+		case GL_FLOAT:
+			return sizeof(GLfloat);
+		case GL_INT:
+			return sizeof(GLint);
+		case GL_UNSIGNED_INT:
+			return sizeof(GLuint);
+		case GL_SHORT:
+			return sizeof(GLshort);
+		case GL_UNSIGNED_SHORT:
+			return sizeof(GLushort);
+		case GL_BYTE:
+			return sizeof(GLbyte);
+		case GL_UNSIGNED_BYTE:
+			return sizeof(GLubyte);
+		default:
+			assert(0);
+			return 0;
+	}
+}
+
+#define CONVERT( TYPE, MACRO ) do {		\
+	GLuint i, j, sz;				\
+	sz = input->Size;				\
+	if (input->Normalized) {			\
+		for (i = 0; i < count; i++) {		\
+			const TYPE *in = (TYPE *)src_ptr;		\
+			for (j = 0; j < sz; j++) {		\
+				*dst_ptr++ = MACRO(*in);		\
+				in++;				\
+			}					\
+			src_ptr += input->StrideB;			\
+		}						\
+	} else {					\
+		for (i = 0; i < count; i++) {		\
+			const TYPE *in = (TYPE *)src_ptr;		\
+			for (j = 0; j < sz; j++) {		\
+				*dst_ptr++ = (GLfloat)(*in);		\
+				in++;				\
+			}					\
+			src_ptr += input->StrideB;			\
+		}						\
+	}						\
+} while (0)
+
+static void r300TranslateAttrib(GLcontext *ctx, GLuint attr, int count, const struct gl_client_array *input, struct gl_buffer_object **bo, GLuint *nr_bo)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	struct r300_vertex_buffer *vbuf = &r300->vbuf;
+	struct vertex_attribute r300_attr;
+	const void *src_ptr;
+	GLenum type;
+
+	if (input->BufferObj->Name) {
+		if (!input->BufferObj->Pointer) {
+			bo[*nr_bo] = input->BufferObj;
+			(*nr_bo)++;
+			ctx->Driver.MapBuffer(ctx, GL_ARRAY_BUFFER, GL_READ_ONLY_ARB, input->BufferObj);
+			assert(input->BufferObj->Pointer != NULL);
+		}
+
+		src_ptr = ADD_POINTERS(input->BufferObj->Pointer, input->Ptr);
+	} else
+		src_ptr = input->Ptr;
+
+	if (input->Type == GL_DOUBLE || ((getTypeSize(input->Type) * input->Size) % 4 > 0)) {
+		if (RADEON_DEBUG & DEBUG_FALLBACKS) {
+			fprintf(stderr, "%s: Converting vertex attributes, attribute data format %x,", __FUNCTION__, input->Type);
+			fprintf(stderr, "stride %d, components %d\n", input->StrideB, input->Size);
+		}
+
+		GLfloat *dst_ptr, *tmp;
+		tmp = dst_ptr = _mesa_malloc(sizeof(GLfloat) * input->Size * count);
+
+		switch (input->Type) {
+			case GL_DOUBLE:
+				CONVERT(GLdouble, (GLfloat));
+				break;
+			case GL_UNSIGNED_INT:
+				CONVERT(GLuint, UINT_TO_FLOAT);
+				break;
+			case GL_INT:
+				CONVERT(GLint, INT_TO_FLOAT);
+				break;
+			case GL_UNSIGNED_SHORT:
+				CONVERT(GLushort, USHORT_TO_FLOAT);
+				break;
+			case GL_SHORT:
+				CONVERT(GLshort, SHORT_TO_FLOAT);
+				break;
+			case GL_UNSIGNED_BYTE:
+				assert(input->Format != GL_BGRA);
+				CONVERT(GLubyte, UBYTE_TO_FLOAT);
+				break;
+			case GL_BYTE:
+				CONVERT(GLbyte, BYTE_TO_FLOAT);
+				break;
+		}
+
+		type = GL_FLOAT;
+		r300_attr.free_needed = GL_TRUE;
+		r300_attr.data = tmp;
+		r300_attr.stride = sizeof(GLfloat) * input->Size;
+		r300_attr.dwords = input->Size;
+	} else {
+		type = input->Type;
+		r300_attr.free_needed = GL_FALSE;
+		r300_attr.data = (GLvoid *)src_ptr;
+		r300_attr.stride = input->StrideB;
+		r300_attr.dwords = getTypeSize(type) * input->Size / 4;
+	}
+
+	r300_attr.size = input->Size;
+	r300_attr.element = attr;
+	r300_attr.dst_loc = vbuf->num_attribs;
+
+	switch (type) {
+		case GL_FLOAT:
+			switch (input->Size) {
+				case 1: r300_attr.data_type = R300_DATA_TYPE_FLOAT_1; break;
+				case 2: r300_attr.data_type = R300_DATA_TYPE_FLOAT_2; break;
+				case 3: r300_attr.data_type = R300_DATA_TYPE_FLOAT_3; break;
+				case 4: r300_attr.data_type = R300_DATA_TYPE_FLOAT_4; break;
+			}
+			r300_attr._signed = 0;
+			r300_attr.normalize = 0;
+			break;
+		case GL_SHORT:
+			r300_attr._signed = 1;
+			r300_attr.normalize = input->Normalized;
+			if (input->Size == 2)
+				r300_attr.data_type = R300_DATA_TYPE_SHORT_2;
+			else if (input->Size == 4)
+				r300_attr.data_type = R300_DATA_TYPE_SHORT_4;
+			else
+				assert(0);
+			break;
+		case GL_BYTE:
+			assert(input->Size == 4);
+			r300_attr._signed = 1;
+			r300_attr.normalize = input->Normalized;
+			r300_attr.data_type = R300_DATA_TYPE_BYTE;
+			break;
+		case GL_UNSIGNED_SHORT:
+			r300_attr._signed = 0;
+			r300_attr.normalize = input->Normalized;
+			if (input->Size == 2)
+				r300_attr.data_type = R300_DATA_TYPE_SHORT_2;
+			else if (input->Size == 4)
+				r300_attr.data_type = R300_DATA_TYPE_SHORT_4;
+			else
+				assert(0);
+			break;
+		case GL_UNSIGNED_BYTE:
+			assert(input->Size == 4);
+			r300_attr._signed = 0;
+			r300_attr.normalize = input->Normalized;
+			if (input->Format == GL_BGRA)
+				r300_attr.data_type = R300_DATA_TYPE_D3DCOLOR;
+			else
+				r300_attr.data_type = R300_DATA_TYPE_BYTE;
+			break;
+
+		default:
+		case GL_DOUBLE:
+		case GL_INT:
+		case GL_UNSIGNED_INT:
+			assert(0);
+			break;
+	}
+
+	switch (input->Size) {
+		case 4:
+			r300_attr.swizzle = SWIZZLE_XYZW;
+			break;
+		case 3:
+			r300_attr.swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ONE);
+			break;
+		case 2:
+			r300_attr.swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_ZERO, SWIZZLE_ONE);
+			break;
+		case 1:
+			r300_attr.swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_ZERO, SWIZZLE_ZERO, SWIZZLE_ONE);
+			break;
+	}
+
+	r300_attr.write_mask = MASK_XYZW;
+
+	vbuf->attribs[vbuf->num_attribs] = r300_attr;
+	++vbuf->num_attribs;
+}
+
+static void r300SetVertexFormat(GLcontext *ctx, const struct gl_client_array *arrays[], int count, struct gl_buffer_object **bo, GLuint *nr_bo)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	struct r300_vertex_buffer *vbuf = &r300->vbuf;
+
+	{
+		int i, tmp;
+
+		tmp = r300->selected_vp->key.InputsRead;
+		i = 0;
+		vbuf->num_attribs = 0;
+		while (tmp) {
+			/* find first enabled bit */
+			while (!(tmp & 1)) {
+				tmp >>= 1;
+				++i;
+			}
+
+			r300TranslateAttrib(ctx, i, count, arrays[i], bo, nr_bo);
+
+			tmp >>= 1;
+			++i;
+		}
+	}
+
+	r300SwitchFallback(ctx, R300_FALLBACK_AOS_LIMIT, vbuf->num_attribs > R300_MAX_AOS_ARRAYS);
+	if (r300->fallback)
+		return;
+
+	{
+		int i;
+
+		for (i = 0; i < vbuf->num_attribs; i++) {
+			rcommon_emit_vector(ctx, &r300->radeon.tcl.aos[i],
+						vbuf->attribs[i].data, vbuf->attribs[i].dwords,
+						vbuf->attribs[i].stride, count);
+		}
+
+		r300->radeon.tcl.aos_count = vbuf->num_attribs;
+	}
+}
+
+static void r300FreeData(GLcontext *ctx, struct gl_buffer_object **bo, GLuint nr_bo)
+{
+	{
+		struct r300_vertex_buffer *vbuf = &R300_CONTEXT(ctx)->vbuf;
+		int i;
+
+		for (i = 0; i < vbuf->num_attribs; i++) {
+			if (vbuf->attribs[i].free_needed)
+				_mesa_free(vbuf->attribs[i].data);
+		}
+	}
+
+	{
+		struct r300_index_buffer *ind_buf = &R300_CONTEXT(ctx)->ind_buf;
+		if (ind_buf->free_needed)
+			_mesa_free(ind_buf->ptr);
+	}
+
+	{
+		int i;
+
+		for (i = 0; i < nr_bo; ++i) {
+			ctx->Driver.UnmapBuffer(ctx, 0, bo[i]);
+		}
+	}
+}
+
+static GLboolean r300TryDrawPrims(GLcontext *ctx,
+					 const struct gl_client_array *arrays[],
+					 const struct _mesa_prim *prim,
+					 GLuint nr_prims,
+					 const struct _mesa_index_buffer *ib,
+					 GLuint min_index,
+					 GLuint max_index )
+{
+	struct r300_context *r300 = R300_CONTEXT(ctx);
+	struct gl_buffer_object *bo[VERT_ATTRIB_MAX+1];
+	GLuint i, nr_bo = 0;
+
+	if (ctx->NewState)
+		_mesa_update_state( ctx );
+
+	if (r300->options.hw_tcl_enabled)
+		_tnl_UpdateFixedFunctionProgram(ctx);
+
+	r300UpdateShaders(r300);
+
+	r300SwitchFallback(ctx, R300_FALLBACK_INVALID_BUFFERS, !r300ValidateBuffers(ctx));
+
+	r300FixupIndexBuffer(ctx, ib, bo, &nr_bo);
+
+	r300SetVertexFormat(ctx, arrays, max_index + 1, bo, &nr_bo);
+
+	if (r300->fallback)
+		return GL_FALSE;
+
+	r300SetupVAP(ctx, r300->selected_vp->key.InputsRead, r300->selected_vp->key.OutputsWritten);
+
+	r300UpdateShaderStates(r300);
+
+	r300EmitCacheFlush(r300);
+	radeonEmitState(&r300->radeon);
+
+	for (i = 0; i < nr_prims; ++i) {
+		r300RunRenderPrimitive(ctx, prim[i].start, prim[i].start + prim[i].count, prim[i].mode);
+	}
+
+	r300EmitCacheFlush(r300);
+
+	radeonReleaseArrays(ctx, ~0);
+
+	r300FreeData(ctx, bo, nr_bo);
+
+	return GL_TRUE;
+}
+
+/* TODO: rebase if number of indices in any of primitives is > 8192 for 32bit indices or 16384 for 16bit indices */
+
+static void r300DrawPrims(GLcontext *ctx,
+			 const struct gl_client_array *arrays[],
+			 const struct _mesa_prim *prim,
+			 GLuint nr_prims,
+			 const struct _mesa_index_buffer *ib,
+			 GLuint min_index,
+			 GLuint max_index)
+{
+	GLboolean retval;
+
+	if (min_index) {
+		vbo_rebase_prims( ctx, arrays, prim, nr_prims, ib, min_index, max_index, r300DrawPrims );
+		return;
+	}
+
+	/* Make an attempt at drawing */
+	retval = r300TryDrawPrims(ctx, arrays, prim, nr_prims, ib, min_index, max_index);
+
+	/* If failed run tnl pipeline - it should take care of fallbacks */
+	if (!retval)
+		_tnl_draw_prims(ctx, arrays, prim, nr_prims, ib, min_index, max_index);
+}
+
+void r300InitDraw(GLcontext *ctx)
+{
+	struct vbo_context *vbo = vbo_context(ctx);
+
+	vbo->draw_prims = r300DrawPrims;
+}
diff --git a/src/mesa/drivers/dri/r300/r300_emit.c b/src/mesa/drivers/dri/r300/r300_emit.c
index 1e79a76b47f..d6c29ea3887 100644
--- a/src/mesa/drivers/dri/r300/r300_emit.c
+++ b/src/mesa/drivers/dri/r300/r300_emit.c
@@ -337,6 +337,11 @@ void r300EmitArrays(GLcontext * ctx)
 	    r300VAPOutputCntl1(ctx, OutputsWritten);
 
 	rmesa->radeon.tcl.aos_count = nr;
+
+	/* Fill index buffer info */
+	rmesa->ind_buf.ptr = vb->Elts;
+	rmesa->ind_buf.is_32bit = GL_TRUE;
+	rmesa->ind_buf.free_needed = GL_FALSE;
 }
 
 void r300EmitCacheFlush(r300ContextPtr rmesa)
diff --git a/src/mesa/drivers/dri/r300/r300_render.c b/src/mesa/drivers/dri/r300/r300_render.c
index 92310a02647..4982bc63b94 100644
--- a/src/mesa/drivers/dri/r300/r300_render.c
+++ b/src/mesa/drivers/dri/r300/r300_render.c
@@ -172,16 +172,19 @@ int r300NumVerts(r300ContextPtr rmesa, int num_verts, int prim)
 	return num_verts - verts_off;
 }
 
-static void r300EmitElts(GLcontext * ctx, void *elts, unsigned long n_elts)
+static void r300EmitElts(GLcontext * ctx, unsigned long n_elts)
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 	void *out;
+	GLbyte el_size;
+
+	el_size = rmesa->ind_buf.is_32bit ? 4 : 2;
 
 	radeonAllocDmaRegion(&rmesa->radeon, &rmesa->radeon.tcl.elt_dma_bo,
-			     &rmesa->radeon.tcl.elt_dma_offset, n_elts * 4, 4);
+			     &rmesa->radeon.tcl.elt_dma_offset, n_elts * el_size, 4);
 	radeon_bo_map(rmesa->radeon.tcl.elt_dma_bo, 1);
 	out = rmesa->radeon.tcl.elt_dma_bo->ptr + rmesa->radeon.tcl.elt_dma_offset;
-	memcpy(out, elts, n_elts * 4);
+	memcpy(out, rmesa->ind_buf.ptr, n_elts * el_size);
 	radeon_bo_unmap(rmesa->radeon.tcl.elt_dma_bo);
 }
 
@@ -191,12 +194,20 @@ static void r300FireEB(r300ContextPtr rmesa, int vertex_count, int type)
 
     r300_emit_scissor(rmesa->radeon.glCtx);
 	if (vertex_count > 0) {
+		int size;
+
 		BEGIN_BATCH(10);
 		OUT_BATCH_PACKET3(R300_PACKET3_3D_DRAW_INDX_2, 0);
-		OUT_BATCH(R300_VAP_VF_CNTL__PRIM_WALK_INDICES |
-			  ((vertex_count + 0) << 16) |
-			  type |
+		if (rmesa->ind_buf.is_32bit) {
+			size = vertex_count;
+			OUT_BATCH(R300_VAP_VF_CNTL__PRIM_WALK_INDICES |
+			  ((vertex_count + 0) << 16) | type |
 			  R300_VAP_VF_CNTL__INDEX_SIZE_32bit);
+		} else {
+			size = (vertex_count + 1) >> 1;
+			OUT_BATCH(R300_VAP_VF_CNTL__PRIM_WALK_INDICES |
+			   ((vertex_count + 0) << 16) | type);
+		}
 
 		if (!rmesa->radeon.radeonScreen->kernel_mm) {
 			OUT_BATCH_PACKET3(R300_PACKET3_INDX_BUFFER, 2);
@@ -206,13 +217,13 @@ static void r300FireEB(r300ContextPtr rmesa, int vertex_count, int type)
 					rmesa->radeon.tcl.elt_dma_bo,
 					rmesa->radeon.tcl.elt_dma_offset,
 					RADEON_GEM_DOMAIN_GTT, 0, 0);
-			OUT_BATCH(vertex_count);
+			OUT_BATCH(size);
 		} else {
 			OUT_BATCH_PACKET3(R300_PACKET3_INDX_BUFFER, 2);
 			OUT_BATCH(R300_INDX_BUFFER_ONE_REG_WR | (0 << R300_INDX_BUFFER_SKIP_SHIFT) |
 	    			 (R300_VAP_PORT_IDX0 >> 2));
 			OUT_BATCH(rmesa->radeon.tcl.elt_dma_offset);
-			OUT_BATCH(vertex_count);
+			OUT_BATCH(size);
 			radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs,
 					      rmesa->radeon.tcl.elt_dma_bo,
 					      RADEON_GEM_DOMAIN_GTT, 0, 0);
@@ -337,12 +348,10 @@ static void r300FireAOS(r300ContextPtr rmesa, int vertex_count, int type)
 	END_BATCH();
 }
 
-static void r300RunRenderPrimitive(r300ContextPtr rmesa, GLcontext * ctx,
-				   int start, int end, int prim)
+void r300RunRenderPrimitive(GLcontext * ctx, int start, int end, int prim)
 {
+	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 	int type, num_verts;
-	TNLcontext *tnl = TNL_CONTEXT(ctx);
-	struct vertex_buffer *vb = &tnl->vb;
 
 	type = r300PrimitiveType(rmesa, prim);
 	num_verts = r300NumVerts(rmesa, end - start, prim);
@@ -356,7 +365,7 @@ static void r300RunRenderPrimitive(r300ContextPtr rmesa, GLcontext * ctx,
 	 */
 	rcommonEnsureCmdBufSpace(&rmesa->radeon, 64, __FUNCTION__);
 
-	if (vb->Elts) {
+	if (rmesa->ind_buf.ptr) {
 		if (num_verts > 65535) {
 			/* not implemented yet */
 			WARN_ONCE("Too many elts\n");
@@ -373,7 +382,7 @@ static void r300RunRenderPrimitive(r300ContextPtr rmesa, GLcontext * ctx,
 		 * allocating the index array might actually evict the vertex
 		 * arrays. *sigh*
 		 */
-		r300EmitElts(ctx, vb->Elts, num_verts);
+		r300EmitElts(ctx, num_verts);
 		r300EmitAOS(rmesa, rmesa->radeon.tcl.aos_count, start);
 		r300FireEB(rmesa, num_verts, type);
 	} else {
@@ -383,8 +392,7 @@ static void r300RunRenderPrimitive(r300ContextPtr rmesa, GLcontext * ctx,
 	COMMIT_BATCH();
 }
 
-static void r300RunRender(GLcontext * ctx,
-			       struct tnl_pipeline_stage *stage)
+static void r300RunRender(GLcontext * ctx, struct tnl_pipeline_stage *stage)
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 	int i;
@@ -406,7 +414,7 @@ static void r300RunRender(GLcontext * ctx,
 		GLuint prim = _tnl_translate_prim(&vb->Primitive[i]);
 		GLuint start = vb->Primitive[i].start;
 		GLuint end = vb->Primitive[i].start + vb->Primitive[i].count;
-		r300RunRenderPrimitive(rmesa, ctx, start, end, prim);
+		r300RunRenderPrimitive(ctx, start, end, prim);
 	}
 
 	r300EmitCacheFlush(rmesa);
@@ -436,6 +444,8 @@ static const char *getFallbackString(uint32_t bit)
 			return "render mode != GL_RENDER";
 		case R300_FALLBACK_FRAGMENT_PROGRAM:
 			return "fragment program";
+		case R300_FALLBACK_AOS_LIMIT:
+			return "aos limit";
 		case R300_FALLBACK_INVALID_BUFFERS:
 			return "invalid buffers";
 		default:
@@ -504,30 +514,6 @@ static GLboolean r300RunNonTCLRender(GLcontext * ctx,
 	return GL_FALSE;
 }
 
-static GLboolean r300RunTCLRender(GLcontext * ctx,
-				  struct tnl_pipeline_stage *stage)
-{
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-
-	if (RADEON_DEBUG & DEBUG_PRIMS)
-		fprintf(stderr, "%s\n", __FUNCTION__);
-
-	if (rmesa->options.hw_tcl_enabled == GL_FALSE)
-		return GL_TRUE;
-
-	/* Call it here so we can fallback early */
-	r300UpdateShaders(rmesa);
-
-	r300SwitchFallback(ctx, R300_FALLBACK_INVALID_BUFFERS, !r300ValidateBuffers(ctx));
-
-	if (rmesa->fallback)
-		return GL_TRUE;
-
-	r300RunRender(ctx, stage);
-
-	return GL_FALSE;
-}
-
 const struct tnl_pipeline_stage _r300_render_stage = {
 	"r300 Hardware Rasterization",
 	NULL,
@@ -536,12 +522,3 @@ const struct tnl_pipeline_stage _r300_render_stage = {
 	NULL,
 	r300RunNonTCLRender
 };
-
-const struct tnl_pipeline_stage _r300_tcl_stage = {
-	"r300 Hardware Transform, Clipping and Lighting",
-	NULL,
-	NULL,
-	NULL,
-	NULL,
-	r300RunTCLRender
-};
diff --git a/src/mesa/drivers/dri/r300/r300_render.h b/src/mesa/drivers/dri/r300/r300_render.h
index 002cd613dc9..192d738f5ff 100644
--- a/src/mesa/drivers/dri/r300/r300_render.h
+++ b/src/mesa/drivers/dri/r300/r300_render.h
@@ -41,6 +41,7 @@
 #define R300_FALLBACK_STENCIL_TWOSIDE   (1 << 21)
 #define R300_FALLBACK_RENDER_MODE       (1 << 22)
 #define R300_FALLBACK_FRAGMENT_PROGRAM  (1 << 23)
+#define R300_FALLBACK_AOS_LIMIT         (1 << 30)
 #define R300_FALLBACK_INVALID_BUFFERS   (1 << 31)
 #define R300_RASTER_FALLBACK_MASK        0xffff0000
 
@@ -52,8 +53,8 @@
 
 extern const struct tnl_pipeline_stage _r300_render_stage;
 
-extern const struct tnl_pipeline_stage _r300_tcl_stage;
-
 extern void r300SwitchFallback(GLcontext *ctx, uint32_t bit, GLboolean mode);
 
+extern void r300RunRenderPrimitive(GLcontext * ctx, int start, int end, int prim);
+
 #endif