-fix texrect fallback when using vtxfmt_a (glitz, xgl, etc.)

-struct vertex_buffer to struct radeon_vertex_buffer conversion
-vertex attrib arrays did not supersede conventinal arrays when enabled

5 files changed, 102 insertions, 480 deletions

diff --git a/src/mesa/drivers/dri/r300/r300_context.h b/src/mesa/drivers/dri/r300/r300_context.h
index f948752e750..4a7772aaa7a 100644
--- a/src/mesa/drivers/dri/r300/r300_context.h
+++ b/src/mesa/drivers/dri/r300/r300_context.h
@@ -714,7 +714,6 @@ struct r300_fragment_program {
 #define REG_COLOR0	1
 #define REG_TEX0	2
 
-#ifdef USER_BUFFERS
 struct dt {
 	GLint size;
 	GLenum type;
@@ -736,7 +735,6 @@ struct radeon_vertex_buffer {
 	GLsizei LockCount;
 	int lock_uptodate;
 };
-#endif
 
 struct r300_aos_rec {
 	GLuint offset;
@@ -759,9 +757,7 @@ struct r300_state {
 	struct r300_pfs_compile_state pfs_compile;
 	struct r300_dma_region aos[R300_MAX_AOS_ARRAYS];
 	int aos_count;
-#ifdef USER_BUFFERS
 	struct radeon_vertex_buffer VB;
-#endif
 
 	GLuint *Elts;
 	struct r300_dma_region elt_dma;
@@ -777,6 +773,9 @@ struct r300_state {
 	
 };
 
+#define R300_FALLBACK_NONE 0
+#define R300_FALLBACK_TCL 1
+#define R300_FALLBACK_RAST 2
 
 /**
  * R300 context structure.
@@ -808,9 +807,6 @@ struct r300_context {
 	GLuint prefer_gart_client_texturing;
 
 #ifdef USER_BUFFERS
-	key_t mm_ipc_key;
-	int mm_shm_id;
-	int mm_sem_id;
 	struct radeon_memory_manager *rmm;
 	GLvector4f dummy_attrib[_TNL_ATTRIB_MAX];
 	GLvector4f *temp_attrib[_TNL_ATTRIB_MAX];
@@ -847,11 +843,13 @@ extern GLboolean r300CreateContext(const __GLcontextModes * glVisual,
 void r300_translate_vertex_shader(struct r300_vertex_program *vp);
 extern void r300InitShaderFuncs(struct dd_function_table *functions);
 extern int r300VertexProgUpdateParams(GLcontext *ctx, struct r300_vertex_program *vp, float *dst);
-extern GLboolean r300Fallback(GLcontext *ctx);
+extern int r300Fallback(GLcontext *ctx);
+
+extern void radeon_vb_to_rvb(r300ContextPtr rmesa, struct radeon_vertex_buffer *rvb, struct vertex_buffer *vb);
+extern GLboolean r300_run_vb_render(GLcontext *ctx, struct tnl_pipeline_stage *stage);
 
 #ifdef RADEON_VTXFMT_A
 extern void radeon_init_vtxfmt_a(r300ContextPtr rmesa);
-extern GLboolean r300_run_vb_render_vtxfmt_a(GLcontext *ctx, struct tnl_pipeline_stage *stage);
 #endif
 
 #ifdef HW_VBOS
diff --git a/src/mesa/drivers/dri/r300/r300_maos.c b/src/mesa/drivers/dri/r300/r300_maos.c
index d9f8c031751..c2ff9c5d983 100644
--- a/src/mesa/drivers/dri/r300/r300_maos.c
+++ b/src/mesa/drivers/dri/r300/r300_maos.c
@@ -255,306 +255,8 @@ void r300EmitElts(GLcontext * ctx, void *elts, unsigned long n_elts, int elt_siz
 /* Emit vertex data to GART memory (unless immediate mode)
  * Route inputs to the vertex processor
  */
-void r300EmitArrays(GLcontext * ctx, GLboolean immd)
-{
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	r300ContextPtr r300 = rmesa;
-	struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
-	GLuint nr = 0;
-	GLuint count = VB->Count;
-	GLuint dw,mask;
-	GLuint vic_1 = 0;	/* R300_VAP_INPUT_CNTL_1 */
-	GLuint aa_vap_reg = 0; /* VAP register assignment */
-	GLuint i;
-	GLuint inputs = 0;
-	
-
-#define CONFIGURE_AOS(r, f, v, sz, cn) { \
-		if (RADEON_DEBUG & DEBUG_STATE) \
-			fprintf(stderr, "Enabling "#v "\n"); \
-		if (++nr >= R300_MAX_AOS_ARRAYS) { \
-			fprintf(stderr, "Aieee! AOS array count exceeded!\n"); \
-			exit(-1); \
-		} \
-		\
-		if (hw_tcl_on == GL_FALSE) \
-			rmesa->state.aos[nr-1].aos_reg = aa_vap_reg++; \
-		rmesa->state.aos[nr-1].aos_format = f; \
-		if (immd) { \
-			rmesa->state.aos[nr-1].aos_size = 4; \
-			rmesa->state.aos[nr-1].aos_stride = 4; \
-			rmesa->state.aos[nr-1].aos_offset = 0; \
-		} else { \
-			emit_vector(ctx, \
-						&rmesa->state.aos[nr-1], \
-						v->data, \
-						sz, \
-						v->stride, \
-						cn); \
-		rmesa->state.vap_reg.r=rmesa->state.aos[nr-1].aos_reg; \
-		} \
-}
-
-	if (hw_tcl_on) {
-		GLuint InputsRead = CURRENT_VERTEX_SHADER(ctx)->Base.InputsRead;
-		struct r300_vertex_program *prog=(struct r300_vertex_program *)CURRENT_VERTEX_SHADER(ctx);
-		if (InputsRead & (1<<VERT_ATTRIB_POS)) {
-			inputs |= _TNL_BIT_POS;
-			rmesa->state.aos[nr++].aos_reg = prog->inputs[VERT_ATTRIB_POS];
-		}
-		if (InputsRead & (1<<VERT_ATTRIB_NORMAL)) {
-			inputs |= _TNL_BIT_NORMAL;
-			rmesa->state.aos[nr++].aos_reg = prog->inputs[VERT_ATTRIB_NORMAL];
-		}
-		if (InputsRead & (1<<VERT_ATTRIB_COLOR0)) {
-			inputs |= _TNL_BIT_COLOR0;
-			rmesa->state.aos[nr++].aos_reg = prog->inputs[VERT_ATTRIB_COLOR0];
-		}
-		if (InputsRead & (1<<VERT_ATTRIB_COLOR1)) {
-			inputs |= _TNL_BIT_COLOR1;
-			rmesa->state.aos[nr++].aos_reg = prog->inputs[VERT_ATTRIB_COLOR1];
-		}
-		if (InputsRead & (1<<VERT_ATTRIB_FOG)) {
-			inputs |= _TNL_BIT_FOG;
-			rmesa->state.aos[nr++].aos_reg = prog->inputs[VERT_ATTRIB_FOG];
-		}
-		if(ctx->Const.MaxTextureUnits > 8) { /* Not sure if this can even happen... */
-			fprintf(stderr, "%s: Cant handle that many inputs\n", __FUNCTION__);
-			exit(-1);
-		}
-		for (i=0;i<ctx->Const.MaxTextureUnits;i++) {
-			if (InputsRead & (1<<(VERT_ATTRIB_TEX0+i))) {
-				inputs |= _TNL_BIT_TEX0<<i;
-				rmesa->state.aos[nr++].aos_reg = prog->inputs[VERT_ATTRIB_TEX0+i];
-			}
-		}
-		nr = 0;
-	} else {
-		inputs = TNL_CONTEXT(ctx)->render_inputs;
-	}
-	rmesa->state.render_inputs = inputs;
-
-	if (inputs & _TNL_BIT_POS) {
-		CONFIGURE_AOS(i_coords,	AOS_FORMAT_FLOAT,
-						VB->ObjPtr,
-						immd ? 4 : VB->ObjPtr->size,
-						count);
-
-		vic_1 |= R300_INPUT_CNTL_POS;
-	}
-
-	if (inputs & _TNL_BIT_NORMAL) {
-		CONFIGURE_AOS(i_normal,	AOS_FORMAT_FLOAT,
-						VB->NormalPtr,
-						immd ? 4 : VB->NormalPtr->size,
-						count);
-
-		vic_1 |= R300_INPUT_CNTL_NORMAL;
-	}
-
-	if (inputs & _TNL_BIT_COLOR0) {
-		int emitsize=4;
-
-		if (!immd) {
-			if (VB->ColorPtr[0]->size == 4 &&
-			    (VB->ColorPtr[0]->stride != 0 ||
-			     VB->ColorPtr[0]->data[0][3] != 1.0)) {
-				emitsize = 4;
-			} else {
-				emitsize = 3;
-			}
-		}
-
-		CONFIGURE_AOS(i_color[0], AOS_FORMAT_FLOAT_COLOR,
-						VB->ColorPtr[0],
-						immd ? 4 : emitsize,
-						count);
-
-		vic_1 |= R300_INPUT_CNTL_COLOR;
-	}
-
-	if (inputs & _TNL_BIT_COLOR1) {
-		CONFIGURE_AOS(i_color[1], AOS_FORMAT_FLOAT_COLOR,
-						VB->SecondaryColorPtr[0],
-						immd ? 4 : VB->SecondaryColorPtr[0]->size,
-						count);
-	}
-
-#if 0
-	if (inputs & _TNL_BIT_FOG) {
-		CONFIGURE_AOS(	AOS_FORMAT_FLOAT,
-						VB->FogCoordPtr,
-						immd ? 4 : VB->FogCoordPtr->size,
-						count);
-	}
-#endif
-
-	r300->state.texture.tc_count = 0;
-	for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
-		if (inputs & (_TNL_BIT_TEX0 << i)) {
-			CONFIGURE_AOS(i_tex[i], AOS_FORMAT_FLOAT,
-							VB->TexCoordPtr[i],
-							immd ? 4 : VB->TexCoordPtr[i]->size,
-							count);
-
-			vic_1 |= R300_INPUT_CNTL_TC0 << i;
-			r300->state.texture.tc_count++;
-		}
-	}
-	
-#define SHOW_INFO(n) do { \
-	if (RADEON_DEBUG & DEBUG_ALL) { \
-	fprintf(stderr, "RR[%d] - sz=%d, reg=%d, fmt=%d -- st=%d, of=0x%08x\n", \
-		n, \
-		r300->state.aos[n].aos_size, \
-		r300->state.aos[n].aos_reg, \
-		r300->state.aos[n].aos_format, \
-		r300->state.aos[n].aos_stride, \
-		r300->state.aos[n].aos_offset); \
-	} \
-} while(0);
 
-	/* setup INPUT_ROUTE */
-	R300_STATECHANGE(r300, vir[0]);
-	for(i=0;i+1<nr;i+=2){
-		SHOW_INFO(i)
-		SHOW_INFO(i+1)
-		dw=(r300->state.aos[i].aos_size-1)
-		| ((r300->state.aos[i].aos_reg)<<8)
-		| (r300->state.aos[i].aos_format<<14)
-		| (((r300->state.aos[i+1].aos_size-1)
-		| ((r300->state.aos[i+1].aos_reg)<<8)
-		| (r300->state.aos[i+1].aos_format<<14))<<16);
-
-		if(i+2==nr){
-			dw|=(1<<(13+16));
-			}
-		r300->hw.vir[0].cmd[R300_VIR_CNTL_0+(i>>1)]=dw;
-		}
-	if(nr & 1){
-		SHOW_INFO(nr-1)
-		dw=(r300->state.aos[nr-1].aos_size-1)
-		| (r300->state.aos[nr-1].aos_format<<14)
-		| ((r300->state.aos[nr-1].aos_reg)<<8)
-		| (1<<13);
-		r300->hw.vir[0].cmd[R300_VIR_CNTL_0+(nr>>1)]=dw;
-		//fprintf(stderr, "vir0 dw=%08x\n", dw);
-		}
-	/* Set the rest of INPUT_ROUTE_0 to 0 */
-	//for(i=((count+1)>>1); i<8; i++)r300->hw.vir[0].cmd[R300_VIR_CNTL_0+i]=(0x0);
-	((drm_r300_cmd_header_t*)r300->hw.vir[0].cmd)->packet0.count = (nr+1)>>1;
-
-
-	/* Mesa assumes that all missing components are from (0, 0, 0, 1) */
-#define ALL_COMPONENTS ((R300_INPUT_ROUTE_SELECT_X<<R300_INPUT_ROUTE_X_SHIFT) \
-		| (R300_INPUT_ROUTE_SELECT_Y<<R300_INPUT_ROUTE_Y_SHIFT) \
-		| (R300_INPUT_ROUTE_SELECT_Z<<R300_INPUT_ROUTE_Z_SHIFT) \
-		| (R300_INPUT_ROUTE_SELECT_W<<R300_INPUT_ROUTE_W_SHIFT))
-
-#define ALL_DEFAULT ((R300_INPUT_ROUTE_SELECT_ZERO<<R300_INPUT_ROUTE_X_SHIFT) \
-		| (R300_INPUT_ROUTE_SELECT_ZERO<<R300_INPUT_ROUTE_Y_SHIFT) \
-		| (R300_INPUT_ROUTE_SELECT_ZERO<<R300_INPUT_ROUTE_Z_SHIFT) \
-		| (R300_INPUT_ROUTE_SELECT_ONE<<R300_INPUT_ROUTE_W_SHIFT))
-
-	R300_STATECHANGE(r300, vir[1]);
-
-	for(i=0;i+1<nr;i+=2){
-		/* do i first.. */
-		mask=(1<<(r300->state.aos[i].aos_size*3))-1;
-		dw=(ALL_COMPONENTS & mask)
-		| (ALL_DEFAULT & ~mask)
-		| R300_INPUT_ROUTE_ENABLE;
-
-		/* i+1 */
-		mask=(1<<(r300->state.aos[i+1].aos_size*3))-1;
-		dw|=(
-		(ALL_COMPONENTS & mask)
-		| (ALL_DEFAULT & ~mask)
-		| R300_INPUT_ROUTE_ENABLE
-		)<<16;
-
-		r300->hw.vir[1].cmd[R300_VIR_CNTL_0+(i>>1)]=dw;
-		}
-	if(nr & 1){
-		mask=(1<<(r300->state.aos[nr-1].aos_size*3))-1;
-		dw=(ALL_COMPONENTS & mask)
-		| (ALL_DEFAULT & ~mask)
-		| R300_INPUT_ROUTE_ENABLE;
-		r300->hw.vir[1].cmd[R300_VIR_CNTL_0+(nr>>1)]=dw;
-		//fprintf(stderr, "vir1 dw=%08x\n", dw);
-		}
-	/* Set the rest of INPUT_ROUTE_1 to 0 */
-	//for(i=((count+1)>>1); i<8; i++)r300->hw.vir[1].cmd[R300_VIR_CNTL_0+i]=0x0;
-	((drm_r300_cmd_header_t*)r300->hw.vir[1].cmd)->packet0.count = (nr+1)>>1;
-
-	/* Set up input_cntl */
-	/* I don't think this is needed for vertex buffers, but it doesn't hurt anything */
-	R300_STATECHANGE(r300, vic);
-	r300->hw.vic.cmd[R300_VIC_CNTL_0]=0x5555;  /* Hard coded value, no idea what it means */
-	r300->hw.vic.cmd[R300_VIC_CNTL_1]=vic_1;
-
-#if 0
-	r300->hw.vic.cmd[R300_VIC_CNTL_1]=0;
-
-	if(r300->state.render_inputs & _TNL_BIT_POS)
-		r300->hw.vic.cmd[R300_VIC_CNTL_1]|=R300_INPUT_CNTL_POS;
-
-	if(r300->state.render_inputs & _TNL_BIT_NORMAL)
-		r300->hw.vic.cmd[R300_VIC_CNTL_1]|=R300_INPUT_CNTL_NORMAL;
-
-	if(r300->state.render_inputs & _TNL_BIT_COLOR0)
-		r300->hw.vic.cmd[R300_VIC_CNTL_1]|=R300_INPUT_CNTL_COLOR;
-
-	for(i=0;i < ctx->Const.MaxTextureUnits;i++)
-		if(r300->state.render_inputs & (_TNL_BIT_TEX0<<i))
-			r300->hw.vic.cmd[R300_VIC_CNTL_1]|=(R300_INPUT_CNTL_TC0<<i);
-#endif
-
-	/* Stage 3: VAP output */
-	
-	R300_STATECHANGE(r300, vof);
-	
-	r300->hw.vof.cmd[R300_VOF_CNTL_0]=0;
-	r300->hw.vof.cmd[R300_VOF_CNTL_1]=0;
-	if (hw_tcl_on){
-		GLuint OutputsWritten = CURRENT_VERTEX_SHADER(ctx)->Base.OutputsWritten;
-		
-		if(OutputsWritten & (1<<VERT_RESULT_HPOS))
-			r300->hw.vof.cmd[R300_VOF_CNTL_0] |= R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT;
-		if(OutputsWritten & (1<<VERT_RESULT_COL0))
-			r300->hw.vof.cmd[R300_VOF_CNTL_0] |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_PRESENT;
-		if(OutputsWritten & (1<<VERT_RESULT_COL1))
-			r300->hw.vof.cmd[R300_VOF_CNTL_0] |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_1_PRESENT;
-		/*if(OutputsWritten & (1<<VERT_RESULT_BFC0))
-			r300->hw.vof.cmd[R300_VOF_CNTL_0] |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_2_PRESENT;
-		if(OutputsWritten & (1<<VERT_RESULT_BFC1))
-			r300->hw.vof.cmd[R300_VOF_CNTL_0] |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_3_PRESENT;*/
-		//if(OutputsWritten & (1<<VERT_RESULT_FOGC))
-
-		if(OutputsWritten & (1<<VERT_RESULT_PSIZ))
-			r300->hw.vof.cmd[R300_VOF_CNTL_0] |= R300_VAP_OUTPUT_VTX_FMT_0__PT_SIZE_PRESENT;
-
-		for(i=0;i < ctx->Const.MaxTextureUnits;i++)
-			if(OutputsWritten & (1<<(VERT_RESULT_TEX0+i)))
-				r300->hw.vof.cmd[R300_VOF_CNTL_1] |= (4<<(3*i));
-	} else {
-		if(inputs & _TNL_BIT_POS)
-			r300->hw.vof.cmd[R300_VOF_CNTL_0] |= R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT;
-		if(inputs & _TNL_BIT_COLOR0)
-			r300->hw.vof.cmd[R300_VOF_CNTL_0] |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_PRESENT;
-		if(inputs & _TNL_BIT_COLOR1)
-			r300->hw.vof.cmd[R300_VOF_CNTL_0] |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_1_PRESENT;
-
-		for(i=0;i < ctx->Const.MaxTextureUnits;i++)
-			if(inputs & (_TNL_BIT_TEX0<<i))
-				r300->hw.vof.cmd[R300_VOF_CNTL_1]|=(4<<(3*i));
-	}
-
-	rmesa->state.aos_count = nr;
-}
-
-#ifdef RADEON_VTXFMT_A
-void r300EmitArraysVtx(GLcontext * ctx, GLboolean immd)
+void r300EmitArrays(GLcontext * ctx, GLboolean immd)
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 	r300ContextPtr r300 = rmesa;
@@ -568,7 +270,6 @@ void r300EmitArraysVtx(GLcontext * ctx, GLboolean immd)
 	GLuint i;
 	GLuint inputs = 0;
 	
-#undef CONFIGURE_AOS
 #define CONFIGURE_AOS(r, f, v, sz, cn) { \
 		if (RADEON_DEBUG & DEBUG_STATE) \
 			fprintf(stderr, "Enabling "#v "\n"); \
@@ -872,7 +573,6 @@ void r300EmitArraysVtx(GLcontext * ctx, GLboolean immd)
 
 	rmesa->state.aos_count = nr;
 }
-#endif
 
 #ifdef USER_BUFFERS
 void r300UseArrays(GLcontext * ctx)
diff --git a/src/mesa/drivers/dri/r300/r300_maos.h b/src/mesa/drivers/dri/r300/r300_maos.h
index c75589085bf..7235af6ef6b 100644
--- a/src/mesa/drivers/dri/r300/r300_maos.h
+++ b/src/mesa/drivers/dri/r300/r300_maos.h
@@ -43,10 +43,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 extern void r300EmitElts(GLcontext * ctx, void *elts, unsigned long n_elts, int elt_size);
 extern void r300EmitArrays(GLcontext * ctx, GLboolean immd);
 
-#ifdef RADEON_VTXFMT_A
-extern void r300EmitArraysVtx(GLcontext * ctx, GLboolean immd);
-#endif
-
 #ifdef USER_BUFFERS
 void r300UseArrays(GLcontext * ctx);
 #endif
diff --git a/src/mesa/drivers/dri/r300/r300_render.c b/src/mesa/drivers/dri/r300/r300_render.c
index ee24f14f615..acdecb54ec2 100644
--- a/src/mesa/drivers/dri/r300/r300_render.c
+++ b/src/mesa/drivers/dri/r300/r300_render.c
@@ -282,99 +282,6 @@ static void r300_render_vb_primitive(r300ContextPtr rmesa,
 	int prim)
 {
    int type, num_verts;
-   LOCAL_VARS
-
-   type=r300_get_primitive_type(rmesa, ctx, prim);
-   num_verts=r300_get_num_verts(rmesa, ctx, end-start, prim);
-
-   if(type<0 || num_verts <= 0)return;
-
-   if(rmesa->state.Elts){
-	r300EmitAOS(rmesa, rmesa->state.aos_count, 0);
-#if 0
-	int i;
-	start_index32_packet(num_verts, type);
-	for(i=0; i < num_verts; i++)
-		e32(rmesa->state.Elts[start+i]); /* start ? */
-#else
-	if(num_verts == 1){
-		start_index32_packet(num_verts, type);
-		e32(rmesa->state.Elts[start]);
-		return;
-	}
-	
-	if(num_verts > 65535){ /* not implemented yet */
-		WARN_ONCE("Too many elts\n");
-		return;
-	}
-	r300EmitElts(ctx, rmesa->state.Elts+start, num_verts, 4);
-	fire_EB(PASS_PREFIX GET_START(&(rmesa->state.elt_dma)), num_verts, type, 4);
-#endif
-   }else{
-	   r300EmitAOS(rmesa, rmesa->state.aos_count, start);
-	   fire_AOS(PASS_PREFIX num_verts, type);
-   }
-}
-
-static GLboolean r300_run_vb_render(GLcontext *ctx,
-				 struct tnl_pipeline_stage *stage)
-{
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	TNLcontext *tnl = TNL_CONTEXT(ctx);
-	struct vertex_buffer *VB = &tnl->vb;
-	int i;
-	LOCAL_VARS
-   
-	if (RADEON_DEBUG & DEBUG_PRIMS)
-		fprintf(stderr, "%s\n", __FUNCTION__);
-	
-	
-	//r300UpdateShaders(rmesa);
-	
-   	r300ReleaseArrays(ctx);
-	r300EmitArrays(ctx, GL_FALSE);
-	
-	r300UpdateShaderStates(rmesa);
-
-	reg_start(R300_RB3D_DSTCACHE_CTLSTAT,0);
-	e32(0x0000000a);
-
-	reg_start(0x4f18,0);
-	e32(0x00000003);
-	r300EmitState(rmesa);
-	
-	rmesa->state.Elts = VB->Elts;
-
-	for(i=0; i < VB->PrimitiveCount; i++){
-		GLuint prim = VB->Primitive[i].mode;
-		GLuint start = VB->Primitive[i].start;
-		GLuint length = VB->Primitive[i].count;
-		
-		r300_render_vb_primitive(rmesa, ctx, start, start + length, prim);
-	}
-	
-	reg_start(R300_RB3D_DSTCACHE_CTLSTAT,0);
-	e32(0x0000000a);
-
-	reg_start(0x4f18,0);
-	e32(0x00000003);
-		
-#ifdef USER_BUFFERS
-	r300UseArrays(ctx);
-#endif
-
-	return GL_FALSE;
-}
-
-#ifdef RADEON_VTXFMT_A
-
-static void r300_render_vb_primitive_vtxfmt_a(r300ContextPtr rmesa,
-	GLcontext *ctx,
-	int start,
-	int end,
-	int prim)
-{
-   int type, num_verts;
 
    type=r300_get_primitive_type(rmesa, ctx, prim);
    num_verts=r300_get_num_verts(rmesa, ctx, end-start, prim);
@@ -416,6 +323,7 @@ void dump_array(struct r300_dma_region *rvb, int count)
 	int *out = (int *)(rvb->address + rvb->start);
 	int i, ci;
 	
+	fprintf(stderr, "stride %d:", rvb->aos_stride);
 	for (i=0; i < count; i++) {
 		fprintf(stderr, "{");
 		if (rvb->aos_format == AOS_FORMAT_FLOAT)
@@ -437,7 +345,7 @@ void dump_dt(struct dt *dt, int count)
 	int *out = dt->data;
 	int i, ci;
 	
-	fprintf(stderr, "base at %p ", out);
+	fprintf(stderr, "stride %d", dt->stride);
 	
 	for (i=0; i < count; i++){
 		fprintf(stderr, "{");
@@ -456,23 +364,27 @@ void dump_dt(struct dt *dt, int count)
 }
 #endif
 
-/*static */GLboolean r300_run_vb_render_vtxfmt_a(GLcontext *ctx,
+GLboolean r300_run_vb_render(GLcontext *ctx,
 				 struct tnl_pipeline_stage *stage)
 {
-   r300ContextPtr rmesa = R300_CONTEXT(ctx);
-   //TNLcontext *tnl = TNL_CONTEXT(ctx);
-   struct radeon_vertex_buffer *VB = &rmesa->state.VB; //&tnl->vb;
-   int i;
-   LOCAL_VARS
+	r300ContextPtr rmesa = R300_CONTEXT(ctx);
+	struct radeon_vertex_buffer *VB = &rmesa->state.VB;
+	int i;
+	LOCAL_VARS
    
 	if (RADEON_DEBUG & DEBUG_PRIMS)
 		fprintf(stderr, "%s\n", __FUNCTION__);
+
+	if (stage) {
+ 		TNLcontext *tnl = TNL_CONTEXT(ctx);
+		radeon_vb_to_rvb(rmesa, VB, &tnl->vb);
+	}
 	
 	r300UpdateShaders(rmesa);
 	if (rmesa->state.VB.LockCount == 0 || 1) {
  	  	r300ReleaseArrays(ctx);
-		r300EmitArraysVtx(ctx, GL_FALSE);
-		
+		r300EmitArrays(ctx, GL_FALSE);
+
 		r300UpdateShaderStates(rmesa);
 	} else {
 		/* TODO: Figure out why do we need these. */
@@ -496,7 +408,7 @@ void dump_dt(struct dt *dt, int count)
 #endif
 #if 0
  	  	r300ReleaseArrays(ctx);
-		r300EmitArraysVtx(ctx, GL_FALSE);
+		r300EmitArrays(ctx, GL_FALSE);
 			
 		fprintf(stderr, "after:\n");
 		for(i=0; i < rmesa->state.aos_count; i++){
@@ -522,7 +434,7 @@ void dump_dt(struct dt *dt, int count)
 		GLuint start = VB->Primitive[i].start;
 		GLuint length = VB->Primitive[i].count;
 		
-		r300_render_vb_primitive_vtxfmt_a(rmesa, ctx, start, start + length, prim);
+		r300_render_vb_primitive(rmesa, ctx, start, start + length, prim);
 	}
 
 	reg_start(R300_RB3D_DSTCACHE_CTLSTAT,0);
@@ -536,19 +448,19 @@ void dump_dt(struct dt *dt, int count)
 #endif
 	return GL_FALSE;
 }
-#endif
 
 #define FALLBACK_IF(expr) \
 do {										\
 	if (expr) {								\
 		if (1 || RADEON_DEBUG & DEBUG_FALLBACKS)			\
 			WARN_ONCE("fallback:%s\n", #expr);			\
-		return GL_TRUE;							\
+		return R300_FALLBACK_RAST;							\
 	}									\
 } while(0)
 
-GLboolean r300Fallback(GLcontext *ctx)
+int r300Fallback(GLcontext *ctx)
 {
+	int i;
 
 	//FALLBACK_IF(ctx->RenderMode != GL_RENDER);  // We do not do SELECT or FEEDBACK (yet ?)
 	
@@ -576,8 +488,12 @@ GLboolean r300Fallback(GLcontext *ctx)
 	/* Rest could be done with vertex fragments */
 	if (ctx->Extensions.NV_point_sprite || ctx->Extensions.ARB_point_sprite)
 		FALLBACK_IF(ctx->Point.PointSprite); // GL_POINT_SPRITE_NV
-	
-	return GL_FALSE;
+
+	for (i = 0; i < ctx->Const.MaxTextureUnits; i++)
+		if (ctx->Texture.Unit[i]._ReallyEnabled & TEXTURE_RECT_BIT)
+			return R300_FALLBACK_TCL;
+		
+	return R300_FALLBACK_NONE;
 }
 
 /**
@@ -593,9 +509,9 @@ static GLboolean r300_run_render(GLcontext *ctx,
 	if (RADEON_DEBUG & DEBUG_PRIMS)
 		fprintf(stderr, "%s\n", __FUNCTION__);
 
-	if (r300Fallback(ctx))
+	if (r300Fallback(ctx) >= R300_FALLBACK_RAST)
 		return GL_TRUE;
-	
+
 	return r300_run_vb_render(ctx, stage);
 }
 
@@ -613,7 +529,6 @@ static GLboolean r300_run_tcl_render(GLcontext *ctx,
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 	struct r300_vertex_program *vp;
-	int i;
    
    	hw_tcl_on=future_hw_tcl_on;
    
@@ -622,17 +537,11 @@ static GLboolean r300_run_tcl_render(GLcontext *ctx,
 	if(hw_tcl_on == GL_FALSE)
 		return GL_TRUE;
 	
-	if (r300Fallback(ctx)) {
+	if (r300Fallback(ctx) >= R300_FALLBACK_TCL) {
 		hw_tcl_on = GL_FALSE;
 		return GL_TRUE;
 	}
 	
-	for (i = 0; i < ctx->Const.MaxTextureUnits; i++) /* XXX: Needs to be part of r300Fallback */
-		if (ctx->Texture.Unit[i]._ReallyEnabled & TEXTURE_RECT_BIT) {
-			hw_tcl_on = GL_FALSE;
-			return GL_TRUE;
-		}
-	
 	r300UpdateShaders(rmesa);
 
 	vp = (struct r300_vertex_program *)CURRENT_VERTEX_SHADER(ctx);
diff --git a/src/mesa/drivers/dri/r300/radeon_vtxfmt_a.c b/src/mesa/drivers/dri/r300/radeon_vtxfmt_a.c
index 053af1ac661..5ea9bb4eb45 100644
--- a/src/mesa/drivers/dri/r300/radeon_vtxfmt_a.c
+++ b/src/mesa/drivers/dri/r300/radeon_vtxfmt_a.c
@@ -45,15 +45,55 @@
 #include "api_validate.h"
 #include "state.h"
 
+#define CONV_VB(a, b) rvb->AttribPtr[(a)].size = vb->b->size, \
+			rvb->AttribPtr[(a)].type = GL_FLOAT, \
+			rvb->AttribPtr[(a)].stride = vb->b->stride, \
+			rvb->AttribPtr[(a)].data = vb->b->data
+
+void radeon_vb_to_rvb(r300ContextPtr rmesa, struct radeon_vertex_buffer *rvb, struct vertex_buffer *vb)
+{
+	int i;
+	GLcontext *ctx;
+	ctx = rmesa->radeon.glCtx;
+	
+	memset(rvb, 0, sizeof(*rvb));
+	
+	rvb->Elts = vb->Elts;
+	rvb->elt_size = 4;
+	rvb->elt_min = 0;
+	rvb->elt_max = vb->Count;
+	
+	rvb->Count = vb->Count;
+	
+	CONV_VB(VERT_ATTRIB_POS, ObjPtr);
+	CONV_VB(VERT_ATTRIB_NORMAL, NormalPtr);
+	CONV_VB(VERT_ATTRIB_COLOR0, ColorPtr[0]);
+	CONV_VB(VERT_ATTRIB_COLOR1, SecondaryColorPtr[0]);
+	CONV_VB(VERT_ATTRIB_FOG, FogCoordPtr);
+	
+	for (i=0; i < MAX_TEXTURE_COORD_UNITS; i++)
+		CONV_VB(VERT_ATTRIB_TEX0 + i, TexCoordPtr[i]);
+	
+	rvb->Primitive = vb->Primitive;
+	rvb->PrimitiveCount = vb->PrimitiveCount;
+	rvb->LockFirst = rvb->LockCount = 0;
+	rvb->lock_uptodate = GL_FALSE;
+}
+
 #ifdef RADEON_VTXFMT_A
 
 extern void _tnl_array_init( GLcontext *ctx );
 
-#define CONV(a, b) rmesa->state.VB.AttribPtr[(a)].size = ctx->Array.b.Size, \
-			rmesa->state.VB.AttribPtr[(a)].data = ctx->Array.b.BufferObj->Name ? \
-			(void *)ADD_POINTERS(ctx->Array.b.Ptr, ctx->Array.b.BufferObj->Data) : (void *)ctx->Array.b.Ptr, \
-			rmesa->state.VB.AttribPtr[(a)].stride = ctx->Array.b.StrideB, \
-			rmesa->state.VB.AttribPtr[(a)].type = ctx->Array.b.Type
+#define CONV(a, b) do { \
+			if (ctx->Array.b.Enabled) { \
+				rmesa->state.VB.AttribPtr[(a)].size = ctx->Array.b.Size; \
+				rmesa->state.VB.AttribPtr[(a)].data = ctx->Array.b.BufferObj->Name ? \
+				(void *)ADD_POINTERS(ctx->Array.b.Ptr, ctx->Array.b.BufferObj->Data) : (void *)ctx->Array.b.Ptr; \
+				rmesa->state.VB.AttribPtr[(a)].stride = ctx->Array.b.StrideB; \
+				rmesa->state.VB.AttribPtr[(a)].type = ctx->Array.b.Type; \
+				enabled |= 1 << (a); \
+			} \
+		   } while (0)
 
 static int setup_arrays(r300ContextPtr rmesa, GLint start)
 {
@@ -63,40 +103,25 @@ static int setup_arrays(r300ContextPtr rmesa, GLint start)
 	GLuint enabled = 0;
 	
 	ctx = rmesa->radeon.glCtx;
-	if (r300Fallback(ctx))
-		return -1;
+	i = r300Fallback(ctx);
+	if (i)
+		return i;
 
 	memset(rmesa->state.VB.AttribPtr, 0, VERT_ATTRIB_MAX*sizeof(struct dt));
 	
 	CONV(VERT_ATTRIB_POS, Vertex);
-	if (ctx->Array.Vertex.Enabled)
-		enabled |= 1 << VERT_ATTRIB_POS;
-	
 	CONV(VERT_ATTRIB_NORMAL, Normal);
-	if (ctx->Array.Normal.Enabled)
-		enabled |= 1 << VERT_ATTRIB_NORMAL;
-	
 	CONV(VERT_ATTRIB_COLOR0, Color);
-	if (ctx->Array.Color.Enabled)
-		enabled |= 1 << VERT_ATTRIB_COLOR0;
-	
 	CONV(VERT_ATTRIB_COLOR1, SecondaryColor);
-	if (ctx->Array.SecondaryColor.Enabled)
-		enabled |= 1 << VERT_ATTRIB_COLOR1;
-	
 	CONV(VERT_ATTRIB_FOG, FogCoord);
-	if (ctx->Array.FogCoord.Enabled)
-		enabled |= 1 << VERT_ATTRIB_FOG;
 	
-	for (i=0; i < MAX_TEXTURE_COORD_UNITS; i++) {
+	for (i=0; i < MAX_TEXTURE_COORD_UNITS; i++)
 		CONV(VERT_ATTRIB_TEX0 + i, TexCoord[i]);
-		
-		if(ctx->Array.TexCoord[i].Enabled) {
-			enabled |= 1 << (VERT_ATTRIB_TEX0+i);
-		}
-		
-	}
-
+	
+	if (ctx->VertexProgram._Enabled)
+		for (i=0; i < VERT_ATTRIB_MAX; i++)
+			CONV(i, VertexAttrib[i]);
+	
 	for (i=0; i < VERT_ATTRIB_MAX; i++) {
 		if (enabled & (1 << i)) {
 			rmesa->state.VB.AttribPtr[i].data += rmesa->state.VB.AttribPtr[i].stride * start;
@@ -113,12 +138,12 @@ static int setup_arrays(r300ContextPtr rmesa, GLint start)
 		if(rmesa->state.VB.AttribPtr[i].type != GL_UNSIGNED_BYTE &&
 			rmesa->state.VB.AttribPtr[i].type != GL_FLOAT){
 			WARN_ONCE("Unsupported format %d at index %d\n", rmesa->state.VB.AttribPtr[i].type, i);
-			return -1;
+			return R300_FALLBACK_TCL;
 		}
 		if(rmesa->state.VB.AttribPtr[i].type == GL_UNSIGNED_BYTE &&
 			rmesa->state.VB.AttribPtr[i].size != 4){
 			WARN_ONCE("Unsupported component count for ub colors\n");
-			return -1;
+			return R300_FALLBACK_TCL;
 		}
 		
 		/*fprintf(stderr, "%d: ", i);
@@ -143,7 +168,7 @@ static int setup_arrays(r300ContextPtr rmesa, GLint start)
 		fprintf(stderr, "Stride %d ", rmesa->state.VB.AttribPtr[i].stride);
 		fprintf(stderr, "\n");*/
 	}
-	return 0;
+	return R300_FALLBACK_NONE;
 }
 
 void radeon_init_vtxfmt_a(r300ContextPtr rmesa);
@@ -180,12 +205,6 @@ static void radeonDrawElements( GLenum mode, GLsizei count, GLenum type, const G
 		return;
 	
 	FLUSH_CURRENT( ctx, 0 );
-	/*
-		fprintf(stderr, "dt at %s:\n", __FUNCTION__);
-		for(i=0; i < VERT_ATTRIB_MAX; i++){
-			fprintf(stderr, "dt %d:", i);
-			dump_dt(&rmesa->state.VB.AttribPtr[i], rmesa->state.VB.Count);
-		}*/
 	r300ReleaseDmaRegion(rmesa, &rvb, __FUNCTION__);
 	
 	switch (type) {
@@ -276,7 +295,7 @@ static void radeonDrawElements( GLenum mode, GLsizei count, GLenum type, const G
 			rmesa->state.VB.Count = rmesa->state.VB.LockCount;
 			
 			r300ReleaseArrays(ctx);
-			r300EmitArraysVtx(ctx, GL_FALSE);
+			r300EmitArrays(ctx, GL_FALSE);
 			
 			rmesa->state.VB.lock_uptodate = GL_TRUE;
 		}
@@ -292,7 +311,7 @@ static void radeonDrawElements( GLenum mode, GLsizei count, GLenum type, const G
 			return;
 		}
 	} else {
-		if (setup_arrays(rmesa, min))
+		if (setup_arrays(rmesa, min) >= R300_FALLBACK_TCL)
 			goto fallback;
 		
 		rmesa->state.VB.Count = max - min + 1;
@@ -313,7 +332,7 @@ static void radeonDrawElements( GLenum mode, GLsizei count, GLenum type, const G
 	rmesa->state.VB.Elts = ptr;
 	rmesa->state.VB.elt_size = elt_size;
 	
-	r300_run_vb_render_vtxfmt_a(ctx, NULL);
+	r300_run_vb_render(ctx, NULL);
 	
 	if(rvb.buf)
 		radeon_mm_use(rmesa, rvb.buf->id);
@@ -438,7 +457,7 @@ static void radeonDrawRangeElements(GLenum mode, GLuint min, GLuint max, GLsizei
 			rmesa->state.VB.Count = rmesa->state.VB.LockCount;
 			
 			r300ReleaseArrays(ctx);
-			r300EmitArraysVtx(ctx, GL_FALSE);
+			r300EmitArrays(ctx, GL_FALSE);
 			
 			rmesa->state.VB.lock_uptodate = GL_TRUE;
 		}
@@ -454,7 +473,7 @@ static void radeonDrawRangeElements(GLenum mode, GLuint min, GLuint max, GLsizei
 			return;
 		}*/
 	} else {
-		if (setup_arrays(rmesa, min))
+		if (setup_arrays(rmesa, min) >= R300_FALLBACK_TCL)
 			goto fallback;
 		
 		rmesa->state.VB.Count = max - min + 1;
@@ -477,7 +496,7 @@ static void radeonDrawRangeElements(GLenum mode, GLuint min, GLuint max, GLsizei
 	rmesa->state.VB.elt_min = min;
 	rmesa->state.VB.elt_max = max;
 	
-	r300_run_vb_render_vtxfmt_a(ctx, NULL);
+	r300_run_vb_render(ctx, NULL);
 	
 	if(rvb.buf)
 		radeon_mm_use(rmesa, rvb.buf->id);
@@ -531,7 +550,7 @@ static void radeonDrawArrays( GLenum mode, GLint start, GLsizei count )
 			rmesa->state.VB.Count = rmesa->state.VB.LockCount;
 			
 			r300ReleaseArrays(ctx);
-			r300EmitArraysVtx(ctx, GL_FALSE);
+			r300EmitArrays(ctx, GL_FALSE);
 			
 			rmesa->state.VB.lock_uptodate = GL_TRUE;
 		}
@@ -547,7 +566,7 @@ static void radeonDrawArrays( GLenum mode, GLint start, GLsizei count )
 			goto fallback;
 		}
 	} else {
-		if (setup_arrays(rmesa, start))
+		if (setup_arrays(rmesa, start) >= R300_FALLBACK_TCL)
 			goto fallback;
 		
 		rmesa->state.VB.Count = count;
@@ -570,7 +589,7 @@ static void radeonDrawArrays( GLenum mode, GLint start, GLsizei count )
 	rmesa->state.VB.elt_min = 0;
 	rmesa->state.VB.elt_max = 0;
 	
-	r300_run_vb_render_vtxfmt_a(ctx, NULL);
+	r300_run_vb_render(ctx, NULL);
 
 	return ;