fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-01-07 04:20:18 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions 2

radv: always write the sample positions when a new descriptor BO is created

Browse source 
This was completely broken, for example in the following scenario:
- submits something which needs sample positions (this creates a new
descriptor BO with sample positions)
- submits something which needs the tess rings (this creates a new
descriptor BO with tess rings but without sample positions, ie.
add_sample_positions would be FALSE)
- submits something which needs sample positions again (this won't
create a new descriptor BO because it incorrectly remembered that
sample positions were set)

Fix this by always writing the sample positions.

This should fix the following flakes:
- dEQP-VK.fragment_shading_barycentric.*.weights.pipeline_topology_dynamic.msaa_interpolate_at_sample.*
- dEQP-VK.pipeline.fast_linked_library.multisample_interpolation.sample_interpolate_at_distinct_values.*
- dEQP-VK.pipeline.fast_linked_library.multisample_interpolation.sample_interpolation_consistency.*
- dEQP-VK.draw.renderpass.linear_interpolation.*
- dEQP-VK.draw.dynamic_rendering.primary_cmd_buff.linear_interpolation.*

These flakes were extremely hard to reproduce!

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25529>
This commit is contained in:
Samuel Pitoiset 2023-10-05 12:21:49 +02:00 • committed by Marge Bot
parent 596b438936
commit 4192e01dcc
12 changed files with 12 additions and 72 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

3
src/amd/ci/radv-navi10-aco-flakes.txt
View file

@ -1,6 +1,3 @@
dEQP-VK.draw.renderpass.linear_interpolation.offset_max_2_samples
dEQP-VK.draw.dynamic_rendering.primary_cmd_buff.linear_interpolation.no_offset_2_samples
dEQP-VK.draw.dynamic_rendering.primary_cmd_buff.linear_interpolation.offset_min_2_samples
dEQP-VK.dynamic_rendering.basic.*
dEQP-VK.dynamic_rendering.primary_cmd_buff.*resuming.*
dEQP-VK.dynamic_rendering.primary_cmd_buff.basic.contents_secondary_cmdbuffers

2
src/amd/ci/radv-navi21-aco-fails.txt
View file

@ -1,4 +1,2 @@
# New CTS failures in 1.3.6.3
dEQP-VK.fragment_shading_barycentric.weights.pipeline_topology_dynamic.msaa_interpolate_at_sample.triangle_fan.noperspective,Fail
dEQP-VK.fragment_shading_barycentric.weights.pipeline_topology_dynamic.msaa_interpolate_at_sample.triangle_fan.perspective,Fail
dEQP-VK.fragment_shading_rate.renderpass2.monolithic.attachment_rate.misc.ro_ds_stencil_read_only_optimal,Crash

11
src/amd/ci/radv-navi21-aco-flakes.txt
View file

@ -1,5 +1,3 @@
dEQP-VK.draw.dynamic_rendering.complete_secondary_cmd_buff.linear_interpolation.offset_max_2_samples
dEQP-VK.draw.renderpass.linear_interpolation.no_offset_\d_samples
dEQP-VK.dynamic_rendering.basic.*
dEQP-VK.dynamic_rendering.primary_cmd_buff.basic.*
dEQP-VK.pipeline.monolithic.image.suballocation.sampling_type.separate.view_type.2d.format.a2b10g10r10_uint_pack32.count_4.size.13x23
@ -8,17 +6,8 @@ dEQP-VK.pipeline.monolithic.image.suballocation.sampling_type.combined.view_type
# Came with the split of the test lists in CI
dEQP-VK.rasterization.primitives_multisample_2_bit.static_stipple.bresenham_line_strip_wide
# This fail is 100% reproducible when running vkcts in parallel mode,
# but isn't reproducible when running with the normal full test list.
# Given that we want both CI and developers to be able to get clean
# runs, let's mark this test as a flake until it gets fixed.
#
# See: https://gitlab.freedesktop.org/mesa/mesa/-/issues/7220
dEQP-VK.draw.renderpass.linear_interpolation.no_offset_\d_samples
dEQP-VK.draw.renderpass.multi_draw.mosaic.indexed_mixed.max_draws.stride_extra_12.2_instances_base_3.with_tess.multiview.offset_6
dEQP-VK.pipeline.fast_linked_library.multisample_interpolation.sample_interpolate_at_ignores_centroid.128_128_1.samples_8
dEQP-VK.mesh_shader.ext.query.all_stats_query.triangles.host_reset.copy.wait.indirect_draw.32bit.with_availability.single_block.task_mesh.inside_rp.single_view.only_primary
# Too many of these tests to follow, so let's cast a broad net

6
src/amd/ci/radv-navi21-llvm-fails.txt
View file

@ -5,12 +5,6 @@ dEQP-VK.spirv_assembly.instruction.graphics.float_controls.fp32.input_args.tanh_
dEQP-VK.draw.renderpass.shader_invocation.helper_invocation,Fail
dEQP-VK.fragment_shading_barycentric.weights.pipeline_topology_dynamic.msaa_interpolate_at_offset.triangle_fan.perspective,Fail
dEQP-VK.fragment_shading_barycentric.weights.pipeline_topology_dynamic.msaa_interpolate_at_offset.triangle_list.perspective,Fail
dEQP-VK.fragment_shading_barycentric.weights.pipeline_topology_dynamic.msaa_interpolate_at_offset.triangle_list_with_adjacency.perspective,Fail
dEQP-VK.fragment_shading_barycentric.weights.pipeline_topology_dynamic.msaa_interpolate_at_offset.triangle_strip.perspective,Fail
dEQP-VK.fragment_shading_barycentric.weights.pipeline_topology_dynamic.msaa_interpolate_at_offset.triangle_strip_with_adjacency.perspective,Fail
dEQP-VK.pipeline.monolithic.multisample.storage_image.64x64_1.r32g32b32a32_sfloat.samples_8,Fail
dEQP-VK.pipeline.monolithic.multisample.storage_image.64x64_1.r8g8b8a8_unorm.samples_8,Fail
dEQP-VK.pipeline.monolithic.multisample.storage_image.64x64_4.r32g32b32a32_sfloat.samples_8,Fail

3
src/amd/ci/radv-polaris10-aco-fails.txt
View file

@ -17,6 +17,3 @@ dEQP-VK.texture.mipmap.3d.image_view_min_lod.base_level.nearest_nearest_integer_
dEQP-VK.texture.mipmap.cubemap.image_view_min_lod.base_level.linear_nearest,Fail
dEQP-VK.texture.mipmap.cubemap.image_view_min_lod.base_level.nearest_linear,Fail
dEQP-VK.texture.mipmap.cubemap.image_view_min_lod.base_level.nearest_nearest,Fail
dEQP-VK.draw.dynamic_rendering.primary_cmd_buff.linear_interpolation.offset_min_2_samples,Fail
dEQP-VK.draw.renderpass.linear_interpolation.offset_max_2_samples,Fail

2
src/amd/ci/radv-polaris10-aco-flakes.txt
View file

@ -1,3 +1 @@
dEQP-VK.dynamic_rendering.basic.*
dEQP-VK.draw.dynamic_rendering.primary_cmd_buff.linear_interpolation.offset_min_2_samples
dEQP-VK.draw.renderpass.linear_interpolation.offset_max_2_samples

4
src/amd/ci/radv-raven-flakes.txt
View file

@ -1,6 +1,2 @@
dEQP-VK.draw.dynamic_rendering.primary_cmd_buff.linear_interpolation.offset_max_2_samples
dEQP-VK.draw.dynamic_rendering.primary_cmd_buff.linear_interpolation.offset_max_8_samples
dEQP-VK.draw.dynamic_rendering.complete_secondary_cmd_buff.linear_interpolation.no_offset_4_samples
dEQP-VK.draw.dynamic_rendering.complete_secondary_cmd_buff.linear_interpolation.offset_max_4_samples
# since Debian 12 CI uprev
dEQP-VK.pipeline.pipeline_library.framebuffer_attachment.diff_attachments_2d_32x32_39x41_ms

2
src/amd/ci/radv-renoir-aco-flakes.txt
View file

@ -1,10 +1,8 @@
dEQP-VK.dynamic_rendering.basic.*
dEQP-VK.reconvergence.workgroup_uniform_control_flow_elect.compute.nesting4.2.35
dEQP-VK.draw.renderpass.linear_interpolation.offset_max_2_samples
dEQP-VK.dynamic_rendering.primary_cmd_buff.basic.2_cmdbuffers_resuming
dEQP-VK.dynamic_rendering.primary_cmd_buff.basic.2_secondary_cmdbuffers_resuming
dEQP-VK.dynamic_rendering.primary_cmd_buff.basic.2_secondary_2_primary_cmdbuffers_resuming
dEQP-VK.draw.dynamic_rendering.primary_cmd_buff.linear_interpolation.offset_min_2_samples
# https://gitlab.freedesktop.org/mesa/mesa/-/issues/8817
dEQP-VK.*framebuffer_attachment.diff_attachments.*

2
src/amd/ci/radv-vangogh-aco-fails.txt
View file

@ -1,5 +1,3 @@
dEQP-VK.draw.dynamic_rendering.primary_cmd_buff.linear_interpolation.no_offset_2_samples,Fail
# New CTS failures in 1.3.6.3
dEQP-VK.api.copy_and_blit.copy_commands2.image_to_buffer.2d_images.mip_copies_bc1_rgb_unorm_block_64x192_5_layers_compute,Crash
dEQP-VK.api.copy_and_blit.copy_commands2.image_to_buffer.2d_images.mip_copies_bc1_rgb_unorm_block_64x192_5_layers_universal,Crash

10
src/amd/ci/radv-vangogh-aco-flakes.txt
View file

@ -4,14 +4,4 @@ dEQP-VK.dynamic_rendering.basic.*
dEQP-VK.dynamic_rendering.primary_cmd_buff.basic.*
dEQP-VK.draw.dynamic_rendering.*
# This fail is 100% reproducible when running vkcts in parallel mode,
# but isn't reproducible when running with the normal full test list.
# Given that we want both CI and developers to be able to get clean
# runs, let's mark this test as a flake until it gets fixed.
#
# See: https://gitlab.freedesktop.org/mesa/mesa/-/issues/7220
dEQP-VK.draw.renderpass.linear_interpolation.no_offset_2_samples
dEQP-VK.draw.renderpass.linear_interpolation.no_offset_4_samples
dEQP-VK.draw.renderpass.linear_interpolation.no_offset_8_samples
dEQP-VK.pipeline.monolithic.image.suballocation.sampling_type.separate.view_type.2d_array.format.r4g4b4a4_unorm_pack16.count_4.size.32x16_array_of_6

2
src/amd/ci/radv-vega10-aco-flakes.txt
View file

@ -1,8 +1,6 @@
dEQP-VK.dynamic_rendering.basic.*
dEQP-VK.dynamic_rendering.primary.*
dEQP-VK.draw.renderpass.linear_interpolation.offset_max_2_samples
dEQP-VK.draw.dynamic_rendering.primary_cmd_buff.linear_interpolation.offset_min_2_samples
dEQP-VK.multiview.renderpass2.multisample.8_1_1_8
dEQP-VK.multiview.renderpass2.multisample.8
dEQP-VK.multiview.renderpass2.multisample.15_15_15_15

37
src/amd/vulkan/radv_queue.c
View file

@ -270,7 +270,7 @@ radv_queue_submit_empty(struct radv_queue *queue, struct vk_queue_submit *submis
}
static void
radv_fill_shader_rings(struct radv_device *device, uint32_t *map, bool add_sample_positions, uint32_t esgs_ring_size,
radv_fill_shader_rings(struct radv_device *device, uint32_t *map, uint32_t esgs_ring_size,
struct radeon_winsys_bo *esgs_ring_bo, uint32_t gsvs_ring_size,
struct radeon_winsys_bo *gsvs_ring_bo, struct radeon_winsys_bo *tess_rings_bo,
struct radeon_winsys_bo *task_rings_bo, struct radeon_winsys_bo *mesh_scratch_ring_bo,
@ -495,16 +495,14 @@ radv_fill_shader_rings(struct radv_device *device, uint32_t *map, bool add_sampl
desc += 4;
if (add_sample_positions) {
/* add sample positions after all rings */
memcpy(desc, device->sample_locations_1x, 8);
desc += 2;
memcpy(desc, device->sample_locations_2x, 16);
desc += 4;
memcpy(desc, device->sample_locations_4x, 32);
desc += 8;
memcpy(desc, device->sample_locations_8x, 64);
}
/* add sample positions after all rings */
memcpy(desc, device->sample_locations_1x, 8);
desc += 2;
memcpy(desc, device->sample_locations_2x, 16);
desc += 4;
memcpy(desc, device->sample_locations_4x, 32);
desc += 8;
memcpy(desc, device->sample_locations_8x, 64);
}
static void
@ -965,15 +963,7 @@ radv_update_preamble_cs(struct radv_queue_state *queue, struct radv_device *devi
tess_rings_bo != queue->tess_rings_bo || task_rings_bo != queue->task_rings_bo ||
mesh_scratch_ring_bo != queue->mesh_scratch_ring_bo || attr_ring_bo != queue->attr_ring_bo ||
add_sample_positions) {
uint32_t size = 0;
if (gsvs_ring_bo || esgs_ring_bo || tess_rings_bo || task_rings_bo || mesh_scratch_ring_bo || attr_ring_bo ||
add_sample_positions) {
size = 176; /* 2 dword + 2 padding + 4 dword * 10 */
if (add_sample_positions)
size += 128; /* 64+32+16+8 = 120 bytes */
} else if (scratch_bo) {
size = 8; /* 2 dword */
}
const uint32_t size = 304;
result = ws->buffer_create(ws, size, 4096, RADEON_DOMAIN_VRAM,
RADEON_FLAG_CPU_ACCESS | RADEON_FLAG_NO_INTERPROCESS_SHARING | RADEON_FLAG_READ_ONLY,
@ -1000,11 +990,8 @@ radv_update_preamble_cs(struct radv_queue_state *queue, struct radv_device *devi
map[1] = rsrc1;
}
if (esgs_ring_bo || gsvs_ring_bo || tess_rings_bo || task_rings_bo || mesh_scratch_ring_bo || attr_ring_bo ||
add_sample_positions)
radv_fill_shader_rings(device, map, add_sample_positions, needs->esgs_ring_size, esgs_ring_bo,
needs->gsvs_ring_size, gsvs_ring_bo, tess_rings_bo, task_rings_bo, mesh_scratch_ring_bo,
needs->attr_ring_size, attr_ring_bo);
radv_fill_shader_rings(device, map, needs->esgs_ring_size, esgs_ring_bo, needs->gsvs_ring_size, gsvs_ring_bo,
tess_rings_bo, task_rings_bo, mesh_scratch_ring_bo, needs->attr_ring_size, attr_ring_bo);
ws->buffer_unmap(descriptor_bo);
}