[DO NOT MERGE] Try for OAM-80992 random building issue #117
Conversation
Fixes: cb2322c Signed-off-by: Jonathan Marek <[email protected]> (cherry picked from commit 8eca6df)
In freedreno_gmem.c, gmem_align of 0x8000 is used. Alignment used here should be the same. Fixes: 912a9c8 Signed-off-by: Jonathan Marek <[email protected]> (cherry picked from commit 4f23767)
Fixes: 3a273a4 Signed-off-by: Jonathan Marek <[email protected]> (cherry picked from commit 6c0fefb)
Now that freedreno has create_with_modifiers(), this "hack" is needed to make some cases work. Copied from vc4. Fixes: 41ddf1d Signed-off-by: Jonathan Marek <[email protected]> (cherry picked from commit e3591b0)
The optimization in 4cd1a0b introduced a replacement of : cmp(8).z.f0.0 vgrf11.y:D, vgrf10.xxxx:D, vgrf2.xyyy:D ... cmp(8).nz.f0.0 null.x:D, vgrf11.yyyy:D, 0D By : cmp(8).z.f0.0 vgrf15.x:D, vgrf10.xxxx:D, vgrf2.yyyy:D ... mov(8) vgrf11.y:D, vgrf15.yyyy:D The first cmp instruction is storing in x while the second mov is sourcing from y. We need to take into account where the replacement on the scan_inst destination is going to store thing so that the replacement mov can source things from the correct location. Signed-off-by: Lionel Landwerlin <[email protected]> Fixes: 4cd1a0b ("i965/vec4: Propagate conditional modifiers from more compares to other compares") Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=109759 Reviewed-by: Ian Romanick <[email protected]> (cherry picked from commit 6e18414)
Added check for higher compat profile being allowed before assigning certain extensions. Fixes: 272fe94 (mesa: enable ARB_texture_buffer_* extensions in the Compatibility profile) Signed-off-by: Danylo Piliaiev <[email protected]> Signed-off-by: Yevhenii Kolesnikov <[email protected]> Reviewed-by: Timothy Arceri <[email protected]> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=107052 (cherry picked from commit 07f4b4e)
Some types of params such as some builtins are always padded. We need to keep track of this so we can restore the list correctly. Here we also remove a couple of cache entries that are not actually required as they get rebuilt by the _mesa_add_parameter() calls. This patch fixes a bunch of arb_texture_multisample and arb_sample_shading piglit tests for the radeonsi NIR backend. Fixes: edded12 ("mesa: rework ParameterList to allow packing") Reviewed-by: Marek Olšák <[email protected]> (cherry picked from commit 7536af6)
call XShmDetach to allow X server to free shared memory Fixes: bcd80be "drisw/glx: use XShm if possible" Signed-off-by: Ray Zhang <[email protected]> Reviewed-by: Dave Airlie <[email protected]> (cherry picked from commit b344e32)
Avoids regression on:
KHR-GLES*.core.tessellation_shader.single.xfb_captures_data_from_correct_stage
that is uncovered by the following patch.
"glsl: fix recording of variables for XFB in TCS shaders"
v2: Rebased over glsl: fix recording of variables for XFB in TCS shaders
v3: Move this patch before "glsl: fix recording of variables for XFB in TCS
shaders" to avoid temporal regressions. (Illia Mirkin)
Cc: 19.0 <[email protected]>
Reviewed-by: Timothy Arceri <[email protected]>
(cherry picked from commit bf1f494)
This is purely for conformance, since it's not actually possible to do
XFB on TCS output varyings. However we do have to make sure we record
the names correctly, and this removes an extra level of array-ness from
the names in question.
Fixes KHR-GL45.tessellation_shader.single.xfb_captures_data_from_correct_stage
v2: Add comment to the new program_resource_visitor::process function.
(Ilia Mirkin)
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=108457
Signed-off-by: Ilia Mirkin <[email protected]>
Cc: 19.0 <[email protected]>
Reviewed-by: Timothy Arceri <[email protected]>
(cherry picked from commit 4eec3a2)
Fix anv_extrypoints.{c,h} and anv_extensions.{c,h} missing dependencies
Rename the variable labels according to targets and python scripts
Align the building rules as per Automake for simplification
Fixes building errors during rebuils due to missing dependencies
(v2) Fixed a missing $(VULKAN_API_XML) reference
Fixes: 9a508b7 ("android: anv/extensions: fix generated sources build")
Fixes: dd088d4 ("anv/extensions: Generate a header file with extension tables")
Signed-off-by: Mauro Rossi <[email protected]>
Reviewed-by: Tapani Pälli <[email protected]>
Reviewed-by: Eric Engestrom <[email protected]>
Cc: "19.0" <[email protected]>
(cherry picked from commit 14e7e26)
Fixes undefined reference building errors for XML_* functions Signed-off-by: Mauro Rossi <[email protected]> Reviewed-by: Tapani Pälli <[email protected]> Cc: "19.0" <[email protected]> (cherry picked from commit ec0f465)
Check GetWindowInfo and ignore the computed sizes if there is an error. Fixes a regression caused by earlier commit when using old wine gallium nine patches. Should also address a crash at window destruction. Related issues: iXit/Mesa-3D#331 iXit/Mesa-3D#332 Cc: [email protected] Fixes: 2318ca6 ("st/nine: Handle window resize when a presentation buffer is used") Signed-off-by: Axel Davy <[email protected]> (cherry picked from commit 86666f0)
Apparently instead of returning error when passing a quality level different than 0 for D3DMULTISAMPLE_NONE, we should pass. Fixes: iXit/Mesa-3D#340 Cc: [email protected] Signed-off-by: Axel Davy <[email protected]> (cherry picked from commit 1d363d4)
…port We were accidentally not counting those surfaces Fixes: ddc4069 "anv: Implement VK_KHR_maintenance3" Reviewed-by: Caio Marcelo de Oliveira Filho <[email protected]> Reviewed-by: Lionel Landwerlin <[email protected]> (cherry picked from commit 5049fbd)
No idea how this fell through the cracks besides the fact that the sampler bound at 0 almost always works and the CTS isn't amazing. In any case, this appears to have been broken for almost forever. Reviewed-by: Samuel Iglesias Gonsálvez <[email protected]> Reviewed-by: Lionel Landwerlin <[email protected]> Cc: [email protected] (cherry picked from commit ca295dd)
When we have a larger sampler index, we get into the "high sampler" scenario and need an instruction header. Even in SIMD8, this pushes the instruction over the sampler message size maximum of 11 registers. Instead, we have to lower TXD to TXL. Fixes: cb98e07 "intel/fs: Support min_lod parameters on texture..." Reviewed-by: Lionel Landwerlin <[email protected]> Reviewed-by: Ian Romanick <[email protected]> (cherry picked from commit 5c96120)
This does not seem to fix anything ATM but is the right thing todo. Signed-off-by: Tapani Pälli <[email protected]> Fixes: f3e91e7 ("anv: add nir lowering pass for ycbcr textures") Reviewed-by: Lionel Landwerlin <[email protected]> (cherry picked from commit 33bf3d5)
This function was never used, and isn't properly guarded by HAVE_LIBDRM, breaking the build on systems that don't have libdrm. Let's just remove it. Fixes: 7552fcb "egl: add base EGL_EXT_device_base implementation" Reported-by: Timo Aaltonen <[email protected]> Signed-off-by: Eric Engestrom <[email protected]> Acked-by: Emil Velikov <[email protected]> (cherry picked from commit bcc4bfc)
If alignement is 0, offets returned by radv_cmd_buffer_upload_alloc() are always 0. These two virtual addresses were pointing at the same location. Cc: 18.3 19.0 <[email protected]> Signed-off-by: Samuel Pitoiset <[email protected]> Reviewed-by: Bas Nieuwenhuizen <[email protected]> (cherry picked from commit c2a1486)
We can't pull it from the variable type because it might be an array of blocks and not just the one block. While we're here, throw in some error checking. Reviewed-by: Lionel Landwerlin <[email protected]> Cc: [email protected] (cherry picked from commit f1dbc7e)
Fixes: 6ac2d16 ("i965/fs: Fix extract_i8/u8 to a 64-bit destination") Reviewed-by: Jason Ekstrand <[email protected]> (cherry picked from commit 4aaf139)
In the old code, we would generate the exact same instruction for extract_u8(some_u64, 0) and extract_u8(some_u64, 1). The mask-a-word trick only works for even numbered bytes. This fixes the (new) piglit test tests/spec/arb_gpu_shader_int64/execution/fs-ushr-and-mask.shader_test. v2: Use a SHR instead of an AND. This saves an instruction compared to using two moves. Suggested by Jason. Fixes: 6ac2d16 ("i965/fs: Fix extract_i8/u8 to a 64-bit destination") Reviewed-by: Jason Ekstrand <[email protected]> (cherry picked from commit 55e6454)
This variable is now unused, so let's remove it. Fixes: db77573 (virgl: modify how we handle GL_MAP_FLUSH_EXPLICIT_BIT) Reviewed-by: Gurchetan Singh <[email protected]> (cherry picked from commit 44620d4)
Pull the common code out of the two entrypoints into the helper which fetches the push descriptor set for us. Now that it does more than just get a thing, call it anv_cmd_buffer_push_descriptor_set. Cc: "19.0" <[email protected]> Reviewed-by: Lionel Landwerlin <[email protected]>
Technically, descriptor set layouts aren't required to survive past the function they're passed into so we need to reference them. Cc: "19.0" <[email protected]> Reviewed-by: Lionel Landwerlin <[email protected]>
This fixes a rendering issue where UBO updates aren't always picked up by drawing calls. This issue effected the Webots robotics simulator. VMware bug 2175527. Testing Done: Webots replay, piglit, misc Linux games Reviewed-by: Thomas Hellstrom <[email protected]> (cherry picked from commit d4381cf)
With createImage(), the caller was expected to set a SHARED flag if they needed the ability to get a GEM handle. DRI3, wayland, and gbm all set it, EGL_MESA_drm_image passes it through, and surfaceless doesn't need it because there's no way to request a handle. With the new createImageWithModifiers() DRI method to replace it, the expectation is that you'll always be able to share the buffer, so the flag is unnecessary in its arguments. However, we do need to tell gallium about this expectation. Without this, kmscube's modifiers path using gbm_bo_create_with_modifiers(&modifier, 1) instead of gbm_bo_create(SCANOUT | SHARED) will call the driver's resource_create() function wtih PIPE_BIND_SHARED unset, so the driver (particularly renderonly drivers) may allocate in such a way that it can't return an answer from gbm_bo_get_handle(). I used to have a hack in v3d using count==1 && modifier==LINEAR to indicate that you wanted SHARED anyway, but that was dropped recently. Fixes: 59527a3 ("v3d: Restructure RO allocations using resource_from_handle.") Reviewed-by: Kristian H. Kristensen <[email protected]> (cherry picked from commit fafead7)
This change makes following test pass: dEQP-VK.api.info.device.extensions Originally-from: Tapani Pälli <[email protected]> Test: [CTS 9.0_r8] dEQP-VK.api.info.device.extensions Signed-off-by: Kevin Strasser <[email protected]>
(cover letter https://patchwork.freedesktop.org/series/51006/) FROMLIST: i965: SIMD32 heuristics debug flag Added a new DEBUG_HEUR32 flag to INTEL_DEBUG flags for enabling SIMD32 selection heuristics. (am from https://patchwork.freedesktop.org/patch/256764/) FROMLIST: i965: SIMD32 heuristics control data Added a new structure for holding SIMD32 heuristics control data. The control data itself will be fetched from drirc. (am from https://patchwork.freedesktop.org/patch/256806/) FROMLIST: i965: SIMD32 heuristics control data from drirc To be able to test the heuristics with different parameters, they can be controlled via environment variables through drirc. (am from https://patchwork.freedesktop.org/patch/256788/) FROMLIST: mesa: Helper functions for counting set bits in a mask (am from https://patchwork.freedesktop.org/patch/256765/) FROMLIST: i965/fs: Save the instruction count of each dispatch width The SIMD32 selection heuristics will use this information for deciding whether SIMD32 shaders should be used. (am from https://patchwork.freedesktop.org/patch/256793/) FROMLIST: i965/fs: SIMD32 selection heuristic based on grouped texture fetches The function goes through the compiled shader and checks how many grouped texture fetches there are. This is a simple heuristic which gets rid of most of the regressions when enabling SIMD32 shaders but still retains some of the benefits. (am from https://patchwork.freedesktop.org/patch/256798/) FROMLIST: i965/fs: Enable all SIMD32 heuristics There are three simple heuristics for SIMD32 shader enabling: - How many MRTs does the shader write into? - How many grouped texture fetches does the shader have? - How many instructions does the SIMD32 shader have compared to the SIMD16 shader? For testing purposes, the heuristics can be controlled via these environment variables: simd32_heuristic_mrt_check - Enables MRT write check - Default: true simd32_heuristic_max_mrts - How many MRT writes the heuristic allows - Default: 1 simd32_heuristic_grouped_check - Enables grouped texture fetch check - Default: true simd32_heuristic_grouped_sends - How many grouped texture fetches the heuristic allows - Default: 6 simd32_heuristic_inst_check - Enables SIMD32 vs. SIMD16 instruction count check - Default: true simd32_heuristic_inst_ratio - SIMD32 vs. SIMD16 instruction count ratio the heuristic allows - Default: 2.3 SIMD32 shaders will not be compiled also when SIMD16 compilation fails or spills. (am from https://patchwork.freedesktop.org/patch/256766/)
This is needed to be in agreement with spec requirements: KhronosGroup/OpenGL-API#46 Piers Daniell: "We discussed this in the OpenGL/ES working group meeting and agreed that eliminating unused elements from the interface block array is not desirable. There is no statement in the spec that this takes place and it would be highly implementation dependent if it happens. If the application has an "interface" in the shader they need to match up with the API it would be quite confusing to have the binding point get compacted. So the answer is no, the binding points aren't affected by unused elements in the interface block array." Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=109532 Reported-By: Ilia Mirkin <[email protected]> Signed-off-by: Andrii Simiklit <[email protected]> TEST=[CTS 9.0r6} dEQP-GLES31.functional.ssbo.layout.random.all_per_block_buffers#18 (am from https://gitlab.freedesktop.org/mesa/mesa/merge_requests/332) Signed-off-by: Kevin Strasser <[email protected]>
…of ssbo/ubo This is needed to fix these tests: piglit.spec.arb_shader_storage_buffer_object.compiler.unused-array-element_frag piglit.spec.arb_shader_storage_buffer_object.compiler.unused-array-element_comp Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=109532 Reported-By: Ilia Mirkin <[email protected]> Signed-off-by: Andrii Simiklit <[email protected]> TEST=[CTS 9.0r6} dEQP-GLES31.functional.ssbo.layout.random.all_per_block_buffers#18 (am from https://gitlab.freedesktop.org/mesa/mesa/merge_requests/332) Signed-off-by: Kevin Strasser <[email protected]>
|
Please don't merge this PR, I put here for building test. We have one random vk building issue, which is only reproduced with buildbot. I have to create PR and build with buildbot. |
|
It looks strange, libmesa_vulkan_util should come within libmesa_vulkan_common |
|
@tpalli, yes, the reproduced rate is very low. Not sure, if it is caused by building rule changes on Q. Maybe some sub-module is built parallel. |
|
@tpalli, I have updated the patch. With the new changes, we have tried 15 buildings. And issue seems gone. Do we still need those two flags on Android? |
Yes we do need these flags, they are also used with i965 driver. Problem is that otherwise some other hash algorithm (default from toolchain) might get chosen and we explicitly want 'sha1'. I don't understand how having these flags could cause the linking issue. Maybe you were just lucky in this case? |
|
@tpalli, I tried to reproduce this issue on buildbot without any changes, and the reproduced rate is about 40%. Now with this changes, after 15 building, and no issue now. |
-Wl is the prefix used to give linker parameters when linker is run indirectly (via compiler), so it's not a parameter itself, just the indication that we want to give parameter '--build-id'
I do find it very strange if the linker parameter affects this. You could try if Vulkan apps work without these flags. When enabling Vulkan I had to put this flag, otherwise the build-id check would fail (see anv_physical_device_init_uuids in anv_device.c). |
|
@tpalli, that make sense! If possible, let's keep the origin patch. I updated the patch, I added "LOCAL_LDFLAGS += -Wl,--no-threads", which means we don't use multi-threads to do link. I triggered about 20 buildings, and not reproduced the random failure issue. Please take a review the latest patch. Thanks a lot! |
|
All right, adding '--no-threads' is fine for me, it will mean slower linking but that will affect only the vulkan driver. It could possibly indicate a bug in Q toolchain as this has been working well so far. |
|
Thanks @tpalli. Will we keep this as a internal changes, if yes, I will update the PR commit info, and get it merged asap. As the random failure break Cactus team's auto building. |
|
@tpalli & @YuanjunHuang FYI. So I reorg the generation order, we hope it has been generated before we call it. I will trigger many many buildings with build bot to check if issue is still here. |
We encounter vk random building issue on Android Q, please refer to following error log. https://buildbot.sh.intel.com/abspb/builders/build-q/builds/13888/steps/build_all/logs/stdio
|
I tried to move the generated file code to libmesa_vulkan_common, but issue is still here, :( |
|
@tpalli & @YuanjunHuang, what do you think of this changes, could we use this as a hot fix for cactus build? |
|
Are you able to access the build machine? It would be interesting to see what symbols 'libmesa_vulkan_util.a' and 'libmesa_vulkan_common.a' have in them, you can list the symbols as example like: "readelf -s --wide libmesa_vulkan_common.a'" |
|
Overall it looks to me that this is likely a bug in Android Q toolchain as the dependencies are set correctly in mk files. Using a workaround is ok for me but it might become painful to update that whenever there are changes, such as new error messages from new VK extensions. |
|
I have synced with Cactus team, when the build finished, the whole out folder will be deleted. Currently, we can't get 'libmesa_vulkan_util.a' and 'libmesa_vulkan_common.a'. |
strassek
force-pushed
the
master
branch
3 times, most recently
from
a69fa2f
to
6959b05
Compare
We encounter vk random building issue on Android Q, please refer
to following error log.
https://buildbot.sh.intel.com/abspb/builders/build-q/builds/13888/steps/build_all/logs/stdio