Improving Vladimirs alpha test fix a bit as it turns out r300Enable didnt correctly disable alpha test. Vertex programs with high vertex count now run a lot faster as software fallback is no longer on. Also disabling 'temp to result'-instructions as they probably violate the spec and dont seem to be something thats needed.

5 files changed, 15 insertions, 19 deletions

diff --git a/src/mesa/drivers/dri/r300/r300_cmdbuf.c b/src/mesa/drivers/dri/r300/r300_cmdbuf.c
index c9cd137f1c..08551b0f5f 100644
--- a/src/mesa/drivers/dri/r300/r300_cmdbuf.c
+++ b/src/mesa/drivers/dri/r300/r300_cmdbuf.c
@@ -634,7 +634,7 @@ void r300EmitWait(r300ContextPtr rmesa, GLuint flags)
 								1 * sizeof(int),
 								__FUNCTION__);
 		cmd[0].i = 0;
-		cmd[0].wait.cmd_type = RADEON_CMD_WAIT;
+		cmd[0].wait.cmd_type = R300_CMD_WAIT;
 		cmd[0].wait.flags = flags;
 	}
 }
diff --git a/src/mesa/drivers/dri/r300/r300_context.c b/src/mesa/drivers/dri/r300/r300_context.c
index b37e98440f..cd7b536592 100644
--- a/src/mesa/drivers/dri/r300/r300_context.c
+++ b/src/mesa/drivers/dri/r300/r300_context.c
@@ -117,7 +117,7 @@ static const struct tnl_pipeline_stage *r300_pipeline[] = {
 	&_tnl_fog_coordinate_stage,
 	&_tnl_texgen_stage,
 	&_tnl_texture_transform_stage,
-	&_tnl_vertex_program_stage,
+	//&_tnl_vertex_program_stage,
 
 	/* Try again to go to tcl?
 	 *     - no good for asymmetric-twoside (do with multipass)
diff --git a/src/mesa/drivers/dri/r300/r300_render.c b/src/mesa/drivers/dri/r300/r300_render.c
index 9b3d5c4e2b..19377c2d34 100644
--- a/src/mesa/drivers/dri/r300/r300_render.c
+++ b/src/mesa/drivers/dri/r300/r300_render.c
@@ -513,6 +513,7 @@ static void r300_render_vb_primitive(r300ContextPtr rmesa,
 	}
 	
 	if(num_verts > 65535){ /* not implemented yet */
+		WARN_ONCE("Too many elts\n");
 		return;
 	}
 	r300EmitElts(ctx, rmesa->state.Elts+start, num_verts);
diff --git a/src/mesa/drivers/dri/r300/r300_state.c b/src/mesa/drivers/dri/r300/r300_state.c
index 60ab460996..ad2cdf585a 100644
--- a/src/mesa/drivers/dri/r300/r300_state.c
+++ b/src/mesa/drivers/dri/r300/r300_state.c
@@ -459,12 +459,12 @@ static void r300Enable(GLcontext* ctx, GLenum cap, GLboolean state)
 		break;
 
 	case GL_ALPHA_TEST:
-		R200_STATECHANGE(r300, at);
+		R300_STATECHANGE(r300, at);
 		if (state) {
 			r300->hw.at.cmd[R300_AT_ALPHA_TEST] |=
 			    R300_ALPHA_TEST_ENABLE;
 		} else {
-			r300->hw.at.cmd[R300_AT_ALPHA_TEST] |=
+			r300->hw.at.cmd[R300_AT_ALPHA_TEST] &=
 			    ~R300_ALPHA_TEST_ENABLE;
 		}
 		break;
@@ -1353,7 +1353,8 @@ void r300_setup_textures(GLcontext *ctx)
 	}
 	
 	for(i=0; i < mtu; i++) {
-		
+		/*if(ctx->Texture.Unit[i].Enabled == 0)
+			continue;*/
 		if( ((r300->state.render_inputs & (_TNL_BIT_TEX0<<i))!=0) != ((ctx->Texture.Unit[i].Enabled)!=0) ) {
 			WARN_ONCE("Mismatch between render_inputs and ctx->Texture.Unit[i].Enabled value.\n");
 		}
@@ -1998,10 +1999,8 @@ void r300ResetHwState(r300ContextPtr r300)
 
 	r300_set_blend_state(ctx);
 
-	r300->hw.at.cmd[R300_AT_ALPHA_TEST] = 0;
 	r300AlphaFunc(ctx, ctx->Color.AlphaFunc, ctx->Color.AlphaRef);
-	if(ctx->Color.AlphaEnabled)
-		r300->hw.at.cmd[R300_AT_ALPHA_TEST] |= R300_ALPHA_TEST_ENABLE;
+	r300Enable(ctx, GL_ALPHA_TEST, ctx->Color.AlphaEnabled);
 		
 		/* Initialize magic registers
 		 TODO : learn what they really do, or get rid of
@@ -2171,10 +2170,6 @@ void r300ResetHwState(r300ContextPtr r300)
 	r300->hw.unk4BC8.cmd[2] = 0;
 	r300->hw.unk4BC8.cmd[3] = 0;
 
-	//r300AlphaFunc(ctx, ctx->Color.AlphaFunc, ctx->Color.AlphaRef);
-#if 0
-	r300->hw.at.cmd[R300_AT_ALPHA_TEST] = 0;
-#endif
 
 	r300->hw.at.cmd[R300_AT_UNKNOWN] = 0;
 	r300->hw.unk4BD8.cmd[1] = 0;
diff --git a/src/mesa/drivers/dri/r300/r300_vertexprog.c b/src/mesa/drivers/dri/r300/r300_vertexprog.c
index 54aea4917b..02e0b5b739 100644
--- a/src/mesa/drivers/dri/r300/r300_vertexprog.c
+++ b/src/mesa/drivers/dri/r300/r300_vertexprog.c
@@ -492,7 +492,7 @@ static void translate_program(struct r300_vertex_program *vp)
 		
 		for(i=0; i < operands; i++)
 			src[i]=vpi->SrcReg[i];
-		
+#if 1
 		if(operands == 3){ /* TODO: scalars */
 			if( CMP_SRCS(src[1], src[2]) || CMP_SRCS(src[0], src[2]) ){
 				o_inst->op=MAKE_VSF_OP(R300_VPI_OUT_OP_ADD, u_temp_i,
@@ -538,7 +538,7 @@ static void translate_program(struct r300_vertex_program *vp)
 				u_temp_i--;
 			}
 		}
-		
+#endif		
 		/* these ops need special handling.
 		   Ops that need temp vars should probably be given reg indexes starting at the end of tmp area. */
 		switch(vpi->Opcode){
@@ -759,8 +759,8 @@ static void translate_program(struct r300_vertex_program *vp)
 				exit(-1);
 			break;
 		}
-		next:
-				
+		next: ;
+#if 0				
 		/* If instruction writes to result and one of the inputs is tmp, we move it at the end of program */
 		if(vpi->DstReg.File == PROGRAM_OUTPUT){
 			for(operand_index=0; operand_index < operands; operand_index++)
@@ -770,13 +770,13 @@ static void translate_program(struct r300_vertex_program *vp)
 					break;
 				}
 		}
-		
+#endif		
 	}
-	
+#if 0	
 	/* Put "tmp to result" instructions in */
 	for(i=0; i < vp->t2rs; i++, o_inst++)
 		*o_inst=t2rs[i];
-		
+#endif		
 	vp->program.length=(o_inst - vp->program.body.i) * 4;
 	
 	if(u_temp_i < vp->num_temporaries)