per vp sw fallbacks

src/mesa/drivers/dri/r300/r300_context.h

@@ -583,6 +583,9 @@ extern int hw_tcl_on;
 
 #define CURRENT_VERTEX_SHADER(ctx) (ctx->VertexProgram._Current)
 
+/* Should but doesnt work */
+//#define CURRENT_VERTEX_SHADER(ctx) (R300_CONTEXT(ctx)->curr_vp)
+
 //#define TMU_ENABLED(ctx, unit) (hw_tcl_on ? ctx->Texture.Unit[unit]._ReallyEnabled && (OutputsWritten & (1<<(VERT_RESULT_TEX0+(unit)))) : 
 //	(r300->state.render_inputs & (_TNL_BIT_TEX0<<(unit))))
 //#define TMU_ENABLED(ctx, unit) (hw_tcl_on ? ctx->Texture.Unit[unit]._ReallyEnabled && OutputsWritten & (1<<(VERT_RESULT_TEX0+(unit))) : 
@@ -604,6 +607,9 @@ struct r300_vertex_program {
 	int num_temporaries; /* Number of temp vars used by program */
 	int inputs[VERT_ATTRIB_MAX];
 	int outputs[VERT_RESULT_MAX];
+	int native;
+	int ref_count;
+	int use_ref_count;
 };
 
 #if USE_ARB_F_P == 1
@@ -852,6 +858,7 @@ struct r300_context {
 	struct r300_hw_state hw;
 	struct r300_cmdbuf cmdbuf;
 	struct r300_state state;
+	struct vertex_program *curr_vp;
 
 	/* Vertex buffers
 	 */
@@ -917,4 +924,9 @@ extern void radeon_init_vtxfmt_a(r300ContextPtr rmesa);
 extern void r300_init_vbo_funcs(struct dd_function_table *functions);
 #endif
 
+#define RADEON_D_CAPTURE 0
+#define RADEON_D_PLAYBACK 1
+#define RADEON_D_PLAYBACK_RAW 2
+#define RADEON_D_T 3
+
 #endif				/* __R300_CONTEXT_H__ */

src/mesa/drivers/dri/r300/r300_reg.h

@@ -1275,6 +1275,8 @@ I am fairly certain that they are correct unless stated otherwise in comments.
 #define R300_RB3D_ZSTENCIL_FORMAT                   0x4F10
 #	define R300_DEPTH_FORMAT_16BIT_INT_Z     (0 << 0)
 #	define R300_DEPTH_FORMAT_24BIT_INT_Z     (2 << 0)
+ /* 16 bit format or some aditional bit ? */
+#	define R300_DEPTH_FORMAT_UNK32          (32 << 0)
 
 /* gap */
 #define R300_RB3D_DEPTHOFFSET               0x4F20

src/mesa/drivers/dri/r300/r300_render.c

@@ -543,7 +543,7 @@ static GLboolean r300_run_vb_render(GLcontext *ctx,
 		fprintf(stderr, "%s\n", __FUNCTION__);
 	
 	
-	r300UpdateShaders(rmesa);
+	//r300UpdateShaders(rmesa);
 	
    	r300ReleaseArrays(ctx);
 	r300EmitArrays(ctx, GL_FALSE);
@@ -567,13 +567,13 @@ static GLboolean r300_run_vb_render(GLcontext *ctx,
 		
 		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
@@ -690,6 +690,7 @@ void dump_dt(struct dt *dt, int count)
 	if (RADEON_DEBUG & DEBUG_PRIMS)
 		fprintf(stderr, "%s\n", __FUNCTION__);
 	
+	r300UpdateShaders(rmesa);
 	if (rmesa->state.VB.LockCount == 0) {
  	  	r300ReleaseArrays(ctx);
 		r300EmitArraysVtx(ctx, GL_FALSE);
@@ -845,6 +846,7 @@ static GLboolean r300_run_tcl_render(GLcontext *ctx,
 				 struct tnl_pipeline_stage *stage)
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
+	struct r300_vertex_program *vp;
    
    	hw_tcl_on=future_hw_tcl_on;
    
@@ -853,25 +855,23 @@ static GLboolean r300_run_tcl_render(GLcontext *ctx,
 	if(hw_tcl_on == GL_FALSE)
 		return GL_TRUE;
 	
-	//r300UpdateShaders(rmesa);
+	r300UpdateShaders(rmesa);
+
+	vp = (struct r300_vertex_program *)CURRENT_VERTEX_SHADER(ctx);
+#if 0 /* Draw every second request with software arb vp */
+	vp->native++;
+	vp->native &= 1;
+	//vp->native = GL_FALSE;
+#endif
+	if (vp->native == GL_FALSE) {
+		hw_tcl_on = GL_FALSE;
+		return GL_TRUE;
+	}
 	//r300UpdateShaderStates(rmesa);
 	
 	return r300_run_vb_render(ctx, stage);
 }
 
-static void r300_check_tcl_render(GLcontext *ctx, struct tnl_pipeline_stage *stage)
-{
-
-	if (RADEON_DEBUG & DEBUG_STATE)
-		fprintf(stderr, "%s\n", __FUNCTION__);
-
-	/* We only support rendering in hardware for now */
-	if (ctx->RenderMode != GL_RENDER) {
-		//stage->active = GL_FALSE;
-		return;
-	}
-}
-
 const struct tnl_pipeline_stage _r300_tcl_stage = {
 	"r300 tcl",
 	NULL,

src/mesa/drivers/dri/r300/r300_shader.c

@@ -11,13 +11,22 @@
 
 static void r300BindProgram(GLcontext *ctx, GLenum target, struct program *prog)
 {
-	/*
+	
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 	struct r300_vertex_program *vp=(void *)prog;
-	*/
+	
 	
 	switch(target){
 		case GL_VERTEX_PROGRAM_ARB:
+		rmesa->curr_vp = vp;
+		vp->ref_count++;
+#if 0
+		if((vp->ref_count % 1500) == 0) {
+			fprintf(stderr, "id %p, ref_count %d\n", vp, vp->ref_count);
+			_mesa_print_program(&vp->mesa_program.Base);
+		}
+#endif
+		
 #if USE_ARB_F_P == 1
 		case GL_FRAGMENT_PROGRAM_ARB:
 #endif
@@ -77,8 +86,11 @@ static struct program *r300NewProgram(GLcontext *ctx, GLenum target, GLuint id)
 
 static void r300DeleteProgram(GLcontext *ctx, struct program *prog)
 {
-	//r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	//struct r300_vertex_program *vp=(void *)prog;
+	r300ContextPtr rmesa = R300_CONTEXT(ctx);
+	struct r300_vertex_program *vp=(void *)prog;
+	
+	/*if(rmesa->curr_vp == vp)
+		rmesa->curr_vp = NULL;*/
 	
 	_mesa_delete_program(ctx, prog);
 }
@@ -86,16 +98,16 @@ static void r300DeleteProgram(GLcontext *ctx, struct program *prog)
 static void r300ProgramStringNotify(GLcontext *ctx, GLenum target, 
 				struct program *prog)
 {
-	/*struct r300_vertex_program *vp=(void *)prog;*/
+	struct r300_vertex_program *vp=(void *)prog;
 #if USE_ARB_F_P == 1
 	struct r300_fragment_program *fp = (struct r300_fragment_program *) prog;
 #endif
 	
 	switch(target) {
 	case GL_VERTEX_PROGRAM_ARB:
-		/*vp->translated=GL_FALSE;
-		translate_vertex_shader(vp);*/
-		//debug_vp(ctx, vp);
+		vp->translated = GL_FALSE;
+		memset(&vp->translated, 0, sizeof(struct r300_vertex_program) - sizeof(struct vertex_program));
+		/*translate_vertex_shader(vp);*/
 	break;
 	case GL_FRAGMENT_PROGRAM_ARB:
 #if USE_ARB_F_P == 1

src/mesa/drivers/dri/r300/r300_vertexprog.c

@@ -934,7 +934,10 @@ void translate_vertex_shader(struct r300_vertex_program *vp)
 	if(u_temp_i < vp->num_temporaries){
 		WARN_ONCE("Ran out of temps, num temps %d, us %d\n", vp->num_temporaries, u_temp_i);
 		vp->translated=GL_TRUE; //GL_FALSE; /* temps exhausted - program cannot be run */
-	}else
+		vp->native = GL_FALSE;
+	}else{
 		vp->translated=GL_TRUE;
+		vp->native = GL_TRUE;
+	}
 }