radeonsi: precompute some fields for PA_CL_VS_OUT_CNTL in si_shader_selector

Reviewed-by: Samuel Pitoiset <[email protected]>
Reviewed-by: Nicolai Hähnle <[email protected]>

4 files changed, 25 insertions, 16 deletions

diff --git a/src/gallium/drivers/radeonsi/si_pipe.h b/src/gallium/drivers/radeonsi/si_pipe.h
index 791145e9c50..108929c10c6 100644
--- a/src/gallium/drivers/radeonsi/si_pipe.h
+++ b/src/gallium/drivers/radeonsi/si_pipe.h
@@ -67,6 +67,7 @@
 #define SI_CONTEXT_VGT_STREAMOUT_SYNC	(R600_CONTEXT_PRIVATE_FLAG << 13)
 
 #define SI_MAX_BORDER_COLORS	4096
+#define SIX_BITS		0x3F
 
 struct si_compute;
 struct hash_table;
diff --git a/src/gallium/drivers/radeonsi/si_shader.h b/src/gallium/drivers/radeonsi/si_shader.h
index f739769af3f..7c04b7e2535 100644
--- a/src/gallium/drivers/radeonsi/si_shader.h
+++ b/src/gallium/drivers/radeonsi/si_shader.h
@@ -326,6 +326,9 @@ struct si_shader_selector {
 	/* PIPE_SHADER_[VERTEX|FRAGMENT|...] */
 	unsigned	type;
 	bool		vs_needs_prolog;
+	unsigned	pa_cl_vs_out_cntl;
+	ubyte		clipdist_mask;
+	ubyte		culldist_mask;
 
 	/* GS parameters. */
 	unsigned	esgs_itemsize;
diff --git a/src/gallium/drivers/radeonsi/si_state.c b/src/gallium/drivers/radeonsi/si_state.c
index 458bc109eeb..3f471a3672e 100644
--- a/src/gallium/drivers/radeonsi/si_state.c
+++ b/src/gallium/drivers/radeonsi/si_state.c
@@ -663,22 +663,19 @@ static void si_emit_clip_state(struct si_context *sctx, struct r600_atom *atom)
 	radeon_emit_array(cs, (uint32_t*)sctx->clip_state.state.ucp, 6*4);
 }
 
-#define SIX_BITS 0x3F
-
 static void si_emit_clip_regs(struct si_context *sctx, struct r600_atom *atom)
 {
 	struct radeon_winsys_cs *cs = sctx->b.gfx.cs;
 	struct si_shader *vs = si_get_vs_state(sctx);
-	struct tgsi_shader_info *info = si_get_vs_info(sctx);
+	struct si_shader_selector *vs_sel = vs->selector;
+	struct tgsi_shader_info *info = &vs_sel->info;
 	struct si_state_rasterizer *rs = sctx->queued.named.rasterizer;
 	unsigned window_space =
 	   info->properties[TGSI_PROPERTY_VS_WINDOW_SPACE_POSITION];
-	unsigned clipdist_mask =
-		info->writes_clipvertex ? SIX_BITS : info->clipdist_writemask;
+	unsigned clipdist_mask = vs_sel->clipdist_mask;
 	unsigned ucp_mask = clipdist_mask ? 0 : rs->clip_plane_enable & SIX_BITS;
-	unsigned culldist_mask = info->culldist_writemask << info->num_written_clipdistance;
+	unsigned culldist_mask = vs_sel->culldist_mask;
 	unsigned total_mask;
-	bool misc_vec_ena;
 
 	if (vs->key.opt.hw_vs.clip_disable) {
 		assert(!info->culldist_writemask);
@@ -696,18 +693,10 @@ static void si_emit_clip_regs(struct si_context *sctx, struct r600_atom *atom)
 	clipdist_mask &= rs->clip_plane_enable;
 	culldist_mask |= clipdist_mask;
 
-	misc_vec_ena = info->writes_psize || info->writes_edgeflag ||
-		       info->writes_layer || info->writes_viewport_index;
-
 	radeon_set_context_reg(cs, R_02881C_PA_CL_VS_OUT_CNTL,
-		S_02881C_USE_VTX_POINT_SIZE(info->writes_psize) |
-		S_02881C_USE_VTX_EDGE_FLAG(info->writes_edgeflag) |
-		S_02881C_USE_VTX_RENDER_TARGET_INDX(info->writes_layer) |
-	        S_02881C_USE_VTX_VIEWPORT_INDX(info->writes_viewport_index) |
+		vs_sel->pa_cl_vs_out_cntl |
 		S_02881C_VS_OUT_CCDIST0_VEC_ENA((total_mask & 0x0F) != 0) |
 		S_02881C_VS_OUT_CCDIST1_VEC_ENA((total_mask & 0xF0) != 0) |
-		S_02881C_VS_OUT_MISC_VEC_ENA(misc_vec_ena) |
-		S_02881C_VS_OUT_MISC_SIDE_BUS_ENA(misc_vec_ena) |
 		clipdist_mask | (culldist_mask << 8));
 	radeon_set_context_reg(cs, R_028810_PA_CL_CLIP_CNTL,
 		rs->pa_cl_clip_cntl |
diff --git a/src/gallium/drivers/radeonsi/si_state_shaders.c b/src/gallium/drivers/radeonsi/si_state_shaders.c
index 078b5e6250d..2e33138695e 100644
--- a/src/gallium/drivers/radeonsi/si_state_shaders.c
+++ b/src/gallium/drivers/radeonsi/si_state_shaders.c
@@ -2087,6 +2087,22 @@ static void *si_create_shader_selector(struct pipe_context *ctx,
 		break;
 	}
 
+	/* PA_CL_VS_OUT_CNTL */
+	bool misc_vec_ena =
+		sel->info.writes_psize || sel->info.writes_edgeflag ||
+		sel->info.writes_layer || sel->info.writes_viewport_index;
+	sel->pa_cl_vs_out_cntl =
+		S_02881C_USE_VTX_POINT_SIZE(sel->info.writes_psize) |
+		S_02881C_USE_VTX_EDGE_FLAG(sel->info.writes_edgeflag) |
+		S_02881C_USE_VTX_RENDER_TARGET_INDX(sel->info.writes_layer) |
+		S_02881C_USE_VTX_VIEWPORT_INDX(sel->info.writes_viewport_index) |
+		S_02881C_VS_OUT_MISC_VEC_ENA(misc_vec_ena) |
+		S_02881C_VS_OUT_MISC_SIDE_BUS_ENA(misc_vec_ena);
+	sel->clipdist_mask = sel->info.writes_clipvertex ?
+				     SIX_BITS : sel->info.clipdist_writemask;
+	sel->culldist_mask = sel->info.culldist_writemask <<
+			     sel->info.num_written_clipdistance;
+
 	/* DB_SHADER_CONTROL */
 	sel->db_shader_control =
 		S_02880C_Z_EXPORT_ENABLE(sel->info.writes_z) |