[Mesa-dev] [PATCH 1/2] radeonsi: fix Polaris MSAA regression

[Nicolai Hähnle]

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

The regression was introduced by commit d938b8c. The problem here is that in
order to use the small primitive filter, we need to explicitly set the sample
locations to 0. But the DB doesn't properly process the change of sample
locations without a flush, and so we can end up with incorrect Z values.

Instead of doing a flush, just disable the small primitive filter when MSAA
is force-disabled.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=96908
Cc: 12.0 <mesa-stable at lists.freedesktop.org>
---
 src/gallium/drivers/radeonsi/si_state.c | 34 ++++++++++++++++++---------------
 src/gallium/drivers/radeonsi/sid.h      |  1 +
 2 files changed, 20 insertions(+), 15 deletions(-)

diff --git a/src/gallium/drivers/radeonsi/si_state.c b/src/gallium/drivers/radeonsi/si_state.c
index f801ca5..c5b61e9 100644
--- a/src/gallium/drivers/radeonsi/si_state.c
+++ b/src/gallium/drivers/radeonsi/si_state.c
@@ -2569,22 +2569,31 @@ static void si_emit_msaa_sample_locs(struct si_context *sctx,
 	if (nr_samples <= 1 && sctx->smoothing_enabled)
 		nr_samples = SI_NUM_SMOOTH_AA_SAMPLES;
 
-	/* The small primitive filter on Polaris requires explicitly setting
-	 * sample locations to 0 when MSAA is disabled.
+	/* On Polaris, the small primitive filter uses the sample locations
+	 * even when MSAA is off, so we need to make sure they're set to 0.
 	 */
-	if (sctx->b.family >= CHIP_POLARIS10) {
-		struct si_state_rasterizer *rs = sctx->queued.named.rasterizer;
-
-		if (!sctx->smoothing_enabled &&
-		    rs && !rs->multisample_enable)
-			nr_samples = 1;
-	}
-
 	if ((nr_samples > 1 || sctx->b.family >= CHIP_POLARIS10) &&
 	    (nr_samples != sctx->msaa_sample_locs.nr_samples)) {
 		sctx->msaa_sample_locs.nr_samples = nr_samples;
 		cayman_emit_msaa_sample_locs(cs, nr_samples);
 	}
+
+	if (sctx->b.family >= CHIP_POLARIS10) {
+		struct si_state_rasterizer *rs = sctx->queued.named.rasterizer;
+		unsigned small_prim_filter_cntl =
+			S_028830_SMALL_PRIM_FILTER_ENABLE(1) |
+			S_028830_LINE_FILTER_DISABLE(1); /* line bug */
+
+		/* The alternative of setting sample locations to 0 would
+		 * require a DB flush to avoid Z errors, see
+		 * https://bugs.freedesktop.org/show_bug.cgi?id=96908
+		 */
+		if (sctx->framebuffer.nr_samples > 1 && rs && !rs->multisample_enable)
+			small_prim_filter_cntl &= C_028830_SMALL_PRIM_FILTER_ENABLE;
+
+		radeon_set_context_reg(cs, R_028830_PA_SU_SMALL_PRIM_FILTER_CNTL,
+				       small_prim_filter_cntl);
+	}
 }
 
 static void si_emit_msaa_config(struct si_context *sctx, struct r600_atom *atom)
@@ -3957,11 +3966,6 @@ static void si_init_config(struct si_context *sctx)
 	if (sctx->b.family == CHIP_STONEY)
 		si_pm4_set_reg(pm4, R_028C40_PA_SC_SHADER_CONTROL, 0);
 
-	if (sctx->b.family >= CHIP_POLARIS10)
-		si_pm4_set_reg(pm4, R_028830_PA_SU_SMALL_PRIM_FILTER_CNTL,
-			       S_028830_SMALL_PRIM_FILTER_ENABLE(1) |
-			       S_028830_LINE_FILTER_DISABLE(1)); /* line bug */
-
 	si_pm4_set_reg(pm4, R_028080_TA_BC_BASE_ADDR, border_color_va >> 8);
 	if (sctx->b.chip_class >= CIK)
 		si_pm4_set_reg(pm4, R_028084_TA_BC_BASE_ADDR_HI, border_color_va >> 40);
diff --git a/src/gallium/drivers/radeonsi/sid.h b/src/gallium/drivers/radeonsi/sid.h
index f6a05c5..0d2ac37 100644
--- a/src/gallium/drivers/radeonsi/sid.h
+++ b/src/gallium/drivers/radeonsi/sid.h
@@ -7216,6 +7216,7 @@
 /*     */
 #define R_028830_PA_SU_SMALL_PRIM_FILTER_CNTL                           0x028830 /* Polaris */
 #define   S_028830_SMALL_PRIM_FILTER_ENABLE(x)                        (((x) & 0x1) << 0)
+#define   C_028830_SMALL_PRIM_FILTER_ENABLE                           0xFFFFFFFE
 #define   S_028830_TRIANGLE_FILTER_DISABLE(x)                         (((x) & 0x1) << 1)
 #define   S_028830_LINE_FILTER_DISABLE(x)                             (((x) & 0x1) << 2)
 #define   S_028830_POINT_FILTER_DISABLE(x)                            (((x) & 0x1) << 3)
-- 
2.7.4