[Mesa-dev] [PATCH 2/2] radeonsi: use trapezoid distribution for tess on Fiji and Polaris

[Nicolai Hähnle]

From: Nicolai Hähnle <nicolai.haehnle at amd.com>

This yields a small performance improvement in Unigine Heaven.
---
 src/gallium/drivers/radeonsi/si_state.c         | 22 +++++++++++++++++-----
 src/gallium/drivers/radeonsi/si_state_shaders.c | 10 +++++++---
 2 files changed, 24 insertions(+), 8 deletions(-)

diff --git a/src/gallium/drivers/radeonsi/si_state.c b/src/gallium/drivers/radeonsi/si_state.c
index 6be2f4b..1cef1dc 100644
--- a/src/gallium/drivers/radeonsi/si_state.c
+++ b/src/gallium/drivers/radeonsi/si_state.c
@@ -3839,16 +3839,28 @@ static void si_init_config(struct si_context *sctx)
 	}
 
 	if (sctx->b.chip_class >= VI) {
+		unsigned vgt_tess_distribution;
+
 		si_pm4_set_reg(pm4, R_028424_CB_DCC_CONTROL,
 			       S_028424_OVERWRITE_COMBINER_MRT_SHARING_DISABLE(1) |
 			       S_028424_OVERWRITE_COMBINER_WATERMARK(4));
 		si_pm4_set_reg(pm4, R_028C58_VGT_VERTEX_REUSE_BLOCK_CNTL, 30);
 		si_pm4_set_reg(pm4, R_028C5C_VGT_OUT_DEALLOC_CNTL, 32);
-		si_pm4_set_reg(pm4, R_028B50_VGT_TESS_DISTRIBUTION,
-		               S_028B50_ACCUM_ISOLINE(32) |
-		               S_028B50_ACCUM_TRI(11) |
-		               S_028B50_ACCUM_QUAD(11) |
-		               S_028B50_DONUT_SPLIT(16));
+
+		vgt_tess_distribution =
+			S_028B50_ACCUM_ISOLINE(32) |
+			S_028B50_ACCUM_TRI(11) |
+			S_028B50_ACCUM_QUAD(11) |
+			S_028B50_DONUT_SPLIT(16);
+
+		/* Testing with Unigine Heaven extreme tesselation yielded best results
+		 * with TRAP_SPLIT = 3.
+		 */
+		if (sctx->b.family == CHIP_FIJI ||
+		    sctx->b.family >= CHIP_POLARIS10)
+			vgt_tess_distribution |= S_028B50_TRAP_SPLIT(3);
+
+		si_pm4_set_reg(pm4, R_028B50_VGT_TESS_DISTRIBUTION, vgt_tess_distribution);
 	}
 
 	if (sctx->b.family == CHIP_STONEY)
diff --git a/src/gallium/drivers/radeonsi/si_state_shaders.c b/src/gallium/drivers/radeonsi/si_state_shaders.c
index a7af76d..daef49d 100644
--- a/src/gallium/drivers/radeonsi/si_state_shaders.c
+++ b/src/gallium/drivers/radeonsi/si_state_shaders.c
@@ -300,9 +300,13 @@ static void si_set_tesseval_regs(struct si_screen *sscreen,
 	else
 		topology = V_028B6C_OUTPUT_TRIANGLE_CW;
 
-	if (sscreen->b.chip_class >= VI)
-		distribution_mode = V_028B6C_DISTRIBUTION_MODE_DONUTS;
-	else
+	if (sscreen->b.chip_class >= VI) {
+		if (sscreen->b.family == CHIP_FIJI ||
+		    sscreen->b.family >= CHIP_POLARIS10)
+			distribution_mode = V_028B6C_DISTRIBUTION_MODE_TRAPEZOIDS;
+		else
+			distribution_mode = V_028B6C_DISTRIBUTION_MODE_DONUTS;
+	} else
 		distribution_mode = V_028B6C_DISTRIBUTION_MODE_NO_DIST;
 
 	si_pm4_set_reg(pm4, R_028B6C_VGT_TF_PARAM,
-- 
2.7.4