freedreno/a3xx: fix gl_PointSize

If vertex writes pointsize, there are a few extra bits we need to turn
on in the cmdstream here and there.

Signed-off-by: Rob Clark <[email protected]>

5 files changed, 20 insertions, 13 deletions

diff --git a/src/gallium/drivers/freedreno/a3xx/fd3_compiler.c b/src/gallium/drivers/freedreno/a3xx/fd3_compiler.c
index 905af54e48d..14d95baa1fe 100644
--- a/src/gallium/drivers/freedreno/a3xx/fd3_compiler.c
+++ b/src/gallium/drivers/freedreno/a3xx/fd3_compiler.c
@@ -1997,8 +1997,10 @@ decl_out(struct fd3_compile_context *ctx, struct tgsi_full_declaration *decl)
 		switch (name) {
 		case TGSI_SEMANTIC_POSITION:
 			so->writes_pos = true;
-			/* fallthrough */
+			break;
 		case TGSI_SEMANTIC_PSIZE:
+			so->writes_psize = true;
+			break;
 		case TGSI_SEMANTIC_COLOR:
 		case TGSI_SEMANTIC_GENERIC:
 		case TGSI_SEMANTIC_FOG:
@@ -2013,7 +2015,7 @@ decl_out(struct fd3_compile_context *ctx, struct tgsi_full_declaration *decl)
 		case TGSI_SEMANTIC_POSITION:
 			comp = 2;  /* tgsi will write to .z component */
 			so->writes_pos = true;
-			/* fallthrough */
+			break;
 		case TGSI_SEMANTIC_COLOR:
 			break;
 		default:
diff --git a/src/gallium/drivers/freedreno/a3xx/fd3_compiler_old.c b/src/gallium/drivers/freedreno/a3xx/fd3_compiler_old.c
index 998b8e9cf7c..d40ee5c5310 100644
--- a/src/gallium/drivers/freedreno/a3xx/fd3_compiler_old.c
+++ b/src/gallium/drivers/freedreno/a3xx/fd3_compiler_old.c
@@ -1377,8 +1377,10 @@ decl_out(struct fd3_compile_context *ctx, struct tgsi_full_declaration *decl)
 		switch (name) {
 		case TGSI_SEMANTIC_POSITION:
 			so->writes_pos = true;
-			/* fallthrough */
+			break;
 		case TGSI_SEMANTIC_PSIZE:
+			so->writes_psize = true;
+			break;
 		case TGSI_SEMANTIC_COLOR:
 		case TGSI_SEMANTIC_GENERIC:
 		case TGSI_SEMANTIC_FOG:
@@ -1393,7 +1395,7 @@ decl_out(struct fd3_compile_context *ctx, struct tgsi_full_declaration *decl)
 		case TGSI_SEMANTIC_POSITION:
 			comp = 2;  /* tgsi will write to .z component */
 			so->writes_pos = true;
-			/* fallthrough */
+			break;
 		case TGSI_SEMANTIC_COLOR:
 			break;
 		default:
diff --git a/src/gallium/drivers/freedreno/a3xx/fd3_emit.c b/src/gallium/drivers/freedreno/a3xx/fd3_emit.c
index 50271fa137c..619ac1ee353 100644
--- a/src/gallium/drivers/freedreno/a3xx/fd3_emit.c
+++ b/src/gallium/drivers/freedreno/a3xx/fd3_emit.c
@@ -448,19 +448,20 @@ fd3_emit_state(struct fd_context *ctx, struct fd_ringbuffer *ring,
 	}
 
 	if (dirty & (FD_DIRTY_RASTERIZER | FD_DIRTY_PROG)) {
-		struct fd3_rasterizer_stateobj *rasterizer =
-				fd3_rasterizer_stateobj(ctx->rasterizer);
-		uint32_t stride_in_vpc = 0;
+		uint32_t val = fd3_rasterizer_stateobj(ctx->rasterizer)
+				->pc_prim_vtx_cntl;
 
 		if (!key.binning_pass) {
-			stride_in_vpc = align(fp->total_in, 4) / 4;
+			uint32_t stride_in_vpc = align(fp->total_in, 4) / 4;
 			if (stride_in_vpc > 0)
 				stride_in_vpc = MAX2(stride_in_vpc, 2);
+			val |= A3XX_PC_PRIM_VTX_CNTL_STRIDE_IN_VPC(stride_in_vpc);
 		}
 
+		val |= COND(vp->writes_psize, A3XX_PC_PRIM_VTX_CNTL_PSIZE);
+
 		OUT_PKT0(ring, REG_A3XX_PC_PRIM_VTX_CNTL, 1);
-		OUT_RING(ring, rasterizer->pc_prim_vtx_cntl |
-				A3XX_PC_PRIM_VTX_CNTL_STRIDE_IN_VPC(stride_in_vpc));
+		OUT_RING(ring, val);
 	}
 
 	if (dirty & FD_DIRTY_SCISSOR) {
diff --git a/src/gallium/drivers/freedreno/a3xx/fd3_program.c b/src/gallium/drivers/freedreno/a3xx/fd3_program.c
index 6fc39a96380..4cdd9387f9d 100644
--- a/src/gallium/drivers/freedreno/a3xx/fd3_program.c
+++ b/src/gallium/drivers/freedreno/a3xx/fd3_program.c
@@ -494,13 +494,15 @@ fd3_program_emit(struct fd_ringbuffer *ring,
 	if (key.binning_pass) {
 		OUT_PKT0(ring, REG_A3XX_VPC_ATTR, 2);
 		OUT_RING(ring, A3XX_VPC_ATTR_THRDASSIGN(1) |
-				A3XX_VPC_ATTR_LMSIZE(1));
+				A3XX_VPC_ATTR_LMSIZE(1) |
+				COND(vp->writes_psize, A3XX_VPC_ATTR_PSIZE));
 		OUT_RING(ring, 0x00000000);
 	} else {
 		OUT_PKT0(ring, REG_A3XX_VPC_ATTR, 2);
 		OUT_RING(ring, A3XX_VPC_ATTR_TOTALATTR(fp->total_in) |
 				A3XX_VPC_ATTR_THRDASSIGN(1) |
-				A3XX_VPC_ATTR_LMSIZE(1));
+				A3XX_VPC_ATTR_LMSIZE(1) |
+				COND(vp->writes_psize, A3XX_VPC_ATTR_PSIZE));
 		OUT_RING(ring, A3XX_VPC_PACK_NUMFPNONPOSVAR(fp->total_in) |
 				A3XX_VPC_PACK_NUMNONPOSVSVAR(fp->total_in));
 
diff --git a/src/gallium/drivers/freedreno/a3xx/fd3_program.h b/src/gallium/drivers/freedreno/a3xx/fd3_program.h
index 8f491b05e4d..8d4fd57ae75 100644
--- a/src/gallium/drivers/freedreno/a3xx/fd3_program.h
+++ b/src/gallium/drivers/freedreno/a3xx/fd3_program.h
@@ -92,7 +92,7 @@ struct fd3_shader_variant {
 		fd3_semantic semantic;
 		uint8_t regid;
 	} outputs[16];
-	bool writes_pos;
+	bool writes_pos, writes_psize;
 
 	/* vertices/inputs: */
 	unsigned inputs_count;