Mesa (main): radeonsi: wait for PS idle in si_set_framebuffer_state

[GitLab Mirror]

Module: Mesa
Branch: main
Commit: cf9ee6d43203803dfad3780582da0872ac534e27
URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=cf9ee6d43203803dfad3780582da0872ac534e27

Author: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer at amd.com>
Date:   Mon May 16 21:06:36 2022 +0200

radeonsi: wait for PS idle in si_set_framebuffer_state

This is needed to avoid write-after-read hazards in
texture -> render transitions.

This fixes fbo-depth tests that were flaky on GPUs (at
least sienna_cichlid and vega20).

Reviewed-by: Mihai Preda <mhpreda at gmail.com>
Reviewed-by: Marek Olšák <marek.olsak at amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16561>

---

 src/gallium/drivers/radeonsi/ci/gfx9-vega20-flakes.csv | 12 ------------
 src/gallium/drivers/radeonsi/si_state.c                |  5 ++++-
 2 files changed, 4 insertions(+), 13 deletions(-)

diff --git a/src/gallium/drivers/radeonsi/ci/gfx9-vega20-flakes.csv b/src/gallium/drivers/radeonsi/ci/gfx9-vega20-flakes.csv
index bddcf453ee6..ce8429c79ad 100644
--- a/src/gallium/drivers/radeonsi/ci/gfx9-vega20-flakes.csv
+++ b/src/gallium/drivers/radeonsi/ci/gfx9-vega20-flakes.csv
@@ -6,27 +6,15 @@ spec@!opengl 1.1 at depthstencil-default_fb-blit
 spec@!opengl 1.1 at depthstencil-default_fb-copypixels samples=2
 spec@!opengl 1.1 at depthstencil-default_fb-copypixels samples=4
 spec@!opengl 1.1 at depthstencil-default_fb-copypixels
-spec at arb_depth_buffer_float@fbo-depth-gl_depth32f_stencil8-blit
-spec at arb_depth_buffer_float@fbo-depth-gl_depth32f_stencil8-copypixels
-spec at arb_depth_buffer_float@fbo-depth-gl_depth_component32f-blit
-spec at arb_depth_buffer_float@fbo-depth-gl_depth_component32f-copypixels
 spec at arb_depth_buffer_float@fbo-depthstencil-gl_depth32f_stencil8-blit
 spec at arb_depth_buffer_float@fbo-depthstencil-gl_depth32f_stencil8-copypixels
 spec at arb_depth_buffer_float@fbo-stencil-gl_depth32f_stencil8-blit
-spec at arb_depth_texture@fbo-depth-gl_depth_component16-blit
-spec at arb_depth_texture@fbo-depth-gl_depth_component16-copypixels
-spec at arb_depth_texture@fbo-depth-gl_depth_component24-blit
-spec at arb_depth_texture@fbo-depth-gl_depth_component24-copypixels
-spec at arb_depth_texture@fbo-depth-gl_depth_component32-blit
-spec at arb_depth_texture@fbo-depth-gl_depth_component32-copypixels
 spec at arb_texture_stencil8@fbo-stencil-blit
 spec at ext_framebuffer_object@fbo-stencil-gl_stencil_index1-blit
 spec at ext_framebuffer_object@fbo-stencil-gl_stencil_index16-copypixels
 spec at ext_framebuffer_object@fbo-stencil-gl_stencil_index4-blit
 spec at ext_framebuffer_object@fbo-stencil-gl_stencil_index4-copypixels
 spec at ext_framebuffer_object@fbo-stencil-gl_stencil_index8-blit
-spec at ext_packed_depth_stencil@fbo-depth-gl_depth24_stencil8-blit
-spec at ext_packed_depth_stencil@fbo-depth-gl_depth24_stencil8-copypixels
 spec at ext_packed_depth_stencil@fbo-depthstencil-gl_depth24_stencil8-blit
 spec at ext_packed_depth_stencil@fbo-depthstencil-gl_depth24_stencil8-copypixels
 spec at ext_packed_depth_stencil@fbo-stencil-gl_depth24_stencil8-copypixels
diff --git a/src/gallium/drivers/radeonsi/si_state.c b/src/gallium/drivers/radeonsi/si_state.c
index a91145a836c..3cf7fc2f214 100644
--- a/src/gallium/drivers/radeonsi/si_state.c
+++ b/src/gallium/drivers/radeonsi/si_state.c
@@ -2971,6 +2971,9 @@ static void si_set_framebuffer_state(struct pipe_context *ctx,
     * - FB write -> shader read
     * - shader write -> FB read
     *
+    * Wait for draws because of possible transitions:
+    * - texture -> render (eg: glBlitFramebuffer(with src=dst) then glDraw*)
+    *
     * DB caches are flushed on demand (using si_decompress_textures).
     *
     * When MSAA is enabled, CB and TC caches are flushed on demand
@@ -2986,7 +2989,7 @@ static void si_set_framebuffer_state(struct pipe_context *ctx,
                                  sctx->framebuffer.all_DCC_pipe_aligned);
    }
 
-   sctx->flags |= SI_CONTEXT_CS_PARTIAL_FLUSH;
+   sctx->flags |= SI_CONTEXT_CS_PARTIAL_FLUSH | SI_CONTEXT_PS_PARTIAL_FLUSH;
 
    /* u_blitter doesn't invoke depth decompression when it does multiple
     * blits in a row, but the only case when it matters for DB is when