radeonsi: add drirc option "radeonsi_assume_no_z_fights"

This option enables a performance optimization where typical non-blending
draws with depth buffer may be rasterized out-of-order (on VI+, multi-SE
chips).

This optimization can lead to incorrect results when an applications
renders multiple objects with the same Z value at the same pixel, so we
will never enable it by default. But there may be applications that could
benefit from white-listing.

Reviewed-by: Marek Olšák <[email protected]>
Tested-by: Dieter Nützel <[email protected]>

4 files changed, 8 insertions, 4 deletions

diff --git a/src/gallium/drivers/radeonsi/driinfo_radeonsi.h b/src/gallium/drivers/radeonsi/driinfo_radeonsi.h
index af6284a7787..8be85289a0c 100644
--- a/src/gallium/drivers/radeonsi/driinfo_radeonsi.h
+++ b/src/gallium/drivers/radeonsi/driinfo_radeonsi.h
@@ -1,4 +1,5 @@
 // DriConf options specific to radeonsi
 DRI_CONF_SECTION_PERFORMANCE
     DRI_CONF_RADEONSI_ENABLE_SISCHED("false")
+    DRI_CONF_RADEONSI_ASSUME_NO_Z_FIGHTS("false")
 DRI_CONF_SECTION_END
diff --git a/src/gallium/drivers/radeonsi/si_pipe.c b/src/gallium/drivers/radeonsi/si_pipe.c
index 68d63692e4f..d6de1525717 100644
--- a/src/gallium/drivers/radeonsi/si_pipe.c
+++ b/src/gallium/drivers/radeonsi/si_pipe.c
@@ -1048,6 +1048,8 @@ struct pipe_screen *radeonsi_screen_create(struct radeon_winsys *ws,
 	sscreen->has_out_of_order_rast = sscreen->b.chip_class >= VI &&
 					 sscreen->b.info.max_se >= 2 &&
 					 !(sscreen->b.debug_flags & DBG_NO_OUT_OF_ORDER);
+	sscreen->assume_no_z_fights =
+		driQueryOptionb(config->options, "radeonsi_assume_no_z_fights");
 	sscreen->has_msaa_sample_loc_bug = (sscreen->b.family >= CHIP_POLARIS10 &&
 					    sscreen->b.family <= CHIP_POLARIS12) ||
 					   sscreen->b.family == CHIP_VEGA10 ||
diff --git a/src/gallium/drivers/radeonsi/si_pipe.h b/src/gallium/drivers/radeonsi/si_pipe.h
index 6d9d3def7b5..3d33e4f0ffa 100644
--- a/src/gallium/drivers/radeonsi/si_pipe.h
+++ b/src/gallium/drivers/radeonsi/si_pipe.h
@@ -95,6 +95,7 @@ struct si_screen {
 	bool				has_distributed_tess;
 	bool				has_draw_indirect_multi;
 	bool				has_out_of_order_rast;
+	bool				assume_no_z_fights;
 	bool				has_msaa_sample_loc_bug;
 	bool				dpbb_allowed;
 	bool				dfsm_allowed;
diff --git a/src/gallium/drivers/radeonsi/si_state.c b/src/gallium/drivers/radeonsi/si_state.c
index 9287086038d..66228af1d23 100644
--- a/src/gallium/drivers/radeonsi/si_state.c
+++ b/src/gallium/drivers/radeonsi/si_state.c
@@ -1094,6 +1094,7 @@ static bool si_order_invariant_stencil_state(const struct pipe_stencil_state *st
 static void *si_create_dsa_state(struct pipe_context *ctx,
 				 const struct pipe_depth_stencil_alpha_state *state)
 {
+	struct si_context *sctx = (struct si_context *)ctx;
 	struct si_state_dsa *dsa = CALLOC_STRUCT(si_state_dsa);
 	struct si_pm4_state *pm4 = &dsa->pm4;
 	unsigned db_depth_control;
@@ -1186,13 +1187,12 @@ static void *si_create_dsa_state(struct pipe_context *ctx,
 		(state->depth.func == PIPE_FUNC_ALWAYS ||
 		 state->depth.func == PIPE_FUNC_NEVER);
 
-	const bool assume_no_z_fights = false;
-
 	dsa->order_invariance[1].pass_last =
-		assume_no_z_fights && !dsa->stencil_write_enabled &&
+		sctx->screen->assume_no_z_fights &&
+		!dsa->stencil_write_enabled &&
 		dsa->depth_write_enabled && zfunc_is_ordered;
 	dsa->order_invariance[0].pass_last =
-		assume_no_z_fights &&
+		sctx->screen->assume_no_z_fights &&
 		dsa->depth_write_enabled && zfunc_is_ordered;
 
 	return dsa;