Modify to make perfetto and AGI run ok #71
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
offset field in kgsl_command_object is NOT used by KGSL, so we should offset directly to iova. Fixes weird hangs on KGSL. E.g. fixes the hang in: dEQP-VK.memory.pipeline_barrier.transfer_dst_storage_texel_buffer.1024 cc: mesa-stable Signed-off-by: Danylo Piliaiev <[email protected]> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20795> (cherry picked from commit 926f626)
Field must be disabled if any render targets have integer format, additionally for Gfx12+ field must be disabled when num multisamples > 1 or forced multisample count > 1. Cc: mesa-stable Signed-off-by: Tapani Pälli <[email protected]> Reviewed-by: Lionel Landwerlin <[email protected]> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20671> (cherry picked from commit 9b37ef4)
Field must be disabled if any render targets have integer format. Cc: mesa-stable Signed-off-by: Tapani Pälli <[email protected]> Reviewed-by: Lionel Landwerlin <[email protected]> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20671> (cherry picked from commit 58dd9d5)
Field must be disabled if any render targets have integer format, additionally for Gfx12+ field must be disabled when num multisamples > 1 or forced multisample count > 1. Cc: mesa-stable Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/7892 Signed-off-by: Tapani Pälli <[email protected]> Reviewed-by: Lionel Landwerlin <[email protected]> Reviewed-by: Kenneth Graunke <[email protected]> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20671> (cherry picked from commit 247c06d)
makes encoded videos resemble the input again :) Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/7992 Fixes: c4482a3 ("radeonsi/vcn: enable multi-slice encoding") Reviewed-by: Ruijing Dong <[email protected]> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20714> (cherry picked from commit e8a6063)
ensure that these have a lifetime great enough to be presented fixes #7781 cc: mesa-stable Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20793> (cherry picked from commit 020db79)
this is an artifact of very old code before the dynamic state was set for all graphics pipelines now the checks only cause blend constants to not be updated, which triggers bugs and validation failures cc: mesa-stable Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20799> (cherry picked from commit b4d18f2)
this is illegal and causes validation errors cc: mesa-stable Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20799> (cherry picked from commit 5d44318)
Fixes: befbd54 - glthread: don't use atomics for refcounting to decrease overhead on AMD Zen Reviewed-by: Pierre-Eric Pelloux-Prayer <[email protected]> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20804> (cherry picked from commit 4d4995b)
It was wrong. For example, if the draw vertex count was 10 and the upload vertex count was 150, u_vbuf wouldn't unroll the draw and would instead memcpy 150 vertices. This fixes that case. Fixes: 068a3bf - util: move and adjust the vertex upload heuristic equation from u_vbuf Reviewed-by: Pierre-Eric Pelloux-Prayer <[email protected]> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20824> (cherry picked from commit 4f6e785)
…exture This wasn't implemented, but the spec requires it. Fixes: 1fe7b1f - mesa: implement ARB_bindless_texture Reviewed-by: Pierre-Eric Pelloux-Prayer <[email protected]> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20824> (cherry picked from commit 7215262)
Surprisingly, the GL compatibility profile allows these in both glEnableClientState and glEnable. Fixes: 0b1dd18 - glthread: track which vertex array attribs are enabled Acked-by: Pierre-Eric Pelloux-Prayer <[email protected]> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20824> (cherry picked from commit 777166c)
unlocks were incorrectly added to paths using dri2_egl_display() as well as those using dri2_egl_display_lock() pthread_mutex_unlock() when unlocked is documented by posix as being undefined behaviour. On OpenBSD pthread_mutex_unlock() will call abort(3) if this happens. Fixes: f1efe03 ("egl/dri2: Add display lock") Reviewed-by: Rob Clark <[email protected]> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20712> (cherry picked from commit 0594b3c)
the outer swapchain object is persistent, which means checking it will never yield an update after the first check fixes #8122 Fixes: b2739c9 ("zink: set surface->dt when updating swapchain" Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20814> (cherry picked from commit 474ed4b)
This reverts commit 6d02054. v_cvt_pk_u8_f32 returns 0xff instead of v_cvt_u32_f32 & 0xff if the input is larger than 255. Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/8128 Cc: mesa-stable Reviewed-by: Timur Kristóf <[email protected]> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20829> (cherry picked from commit e527f68)
When this was last changed, the smoothing_enabled flag seems to have been forgotten about, breaking line-smoothing (and probably also polygon smoothing). Fixes: 4147add ("radeonsi: update db_eqaa even if msaa is disabled") Reviewed-by: Marek Olšák <[email protected]> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20810> (cherry picked from commit 9f4f131) Conflicts: src/amd/ci/radeonsi-raven-fails.txt src/amd/ci/radeonsi-stoney-fails.txt
This has to be before the MI_BATCH_BUFFER_END otherwise it has no effect. This also was messing around with you batch length alignment. Signed-off-by: Lionel Landwerlin <[email protected]> Fixes: b9aa66d ("anv: disable preemption for 3DPRIMITIVE during streamout") Reviewed-by: Tapani Pälli <[email protected]> Reviewed-by: Paulo Zanoni <[email protected]> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20802> (cherry picked from commit 6f02f9d) Conflicts: src/intel/vulkan/genX_gpu_memcpy.c
Clarify the relationship between zink heaps and vulkan memory type indices, and resolve the issues from mixing the two up. Closes: #7588, #7813 Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20264> (cherry picked from commit f6d3a57)
Trival renames to correctly identify vulkan memory type indices aren't the same as zink heaps Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20264> (cherry picked from commit e20e8f2)
VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT = 0x01 has been incidently the correct memory type index, but isn't guaranteed to be, which is why it hasn't caused issues yet Fixes: f9515d9 ("zink: allocate/place memory using memoryTypeIndex directly") Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20264> (cherry picked from commit c71287e)
…_mask I've been running into failures with tests like : dEQP-VK.robustness.robustness2.bind.notemplate.rgba32i.unroll.nonvolatile.uniform_buffer_dynamic.no_fmt_qual.len_4.samples_1.1d.frag With the load_global_const_block_intel NIR intrinsic, you can load a vec8/vec16 with a predicate. The predicate is correctly uniformized to feed into the SEND instruction's flag register. The problem is that a series of optimization first remove the find_live_channel and then changes the broadcast into a simple MOV instruction, on the assumption that the first channel is always active if there is not control flow. This is correct. But after that the cmod optimzation will remove this instruction : mov.nz.f0.0(16) null:D, vgrf16+0.0<0>:D NoMask because it seems to be equivalent to : cmp.g.f0.0(16) vgrf16:D, vgrf12:D, 63d In this case vgrf16 is the predicate to the load block SEND instruction. Since the execution mask is different between both, some of the channels of the SEND instruction end up not being loaded or loaded with the wrong predication and we end up with incorrect UBO data. Signed-off-by: Lionel Landwerlin <[email protected]> Cc: mesa-stable Reviewed-by: Marcin Ślusarz <[email protected]> Reviewed-by: Ian Romanick <[email protected]> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20852> (cherry picked from commit a50d2fd)
Fixes: bbad550 ("radv/winsys: fill real info for CHIP_GFX1100") Signed-off-by: Samuel Pitoiset <[email protected]> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20850> (cherry picked from commit bf3c14b)
Started showing up when nir_opt_large_constants call was moved in 88756ce. Fixes dEQP-VK.mesh_shader.ext.smoke.monolithic.fullscreen_gradient* Suggested-by: Kenneth Graunke <[email protected]> Fixes: 88756ce ("intel/compiler: Run nir_opt_large_constants before scalarizing consts") Reviewed-by: Lionel Landwerlin <[email protected]> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20876> (cherry picked from commit 536a2ac)
Fixes validation failures: Test case 'dEQP-GLES31.functional.android_extension_pack.shaders.es32.extension_directive.oes_sample_variables'.. MESA: error: Validation Error: [ UNASSIGNED-CoreValidation-Shader-InconsistentSpirv ] Object 0: handle = 0x563a1838b790, type = VK_OBJECT_TYPE_DEVICE; | MessageID = 0x6bbb14 | SPIR-V module not valid: [VUID-StandaloneSpirv-Flat-04744] Fragment OpEntryPoint operand 31 with Input interfaces with integer or float type must have a Flat decoration for Entry Point id 4. %gl_SampleId = OpVariable %_ptr_Input_uint Input Test case 'KHR-GL46.shader_ballot_tests.ShaderBallotAvailability'.. MESA: error: Validation Error: [ UNASSIGNED-CoreValidation-Shader-InconsistentSpirv ] Object 0: handle = 0x5558e12f17e0, type = VK_OBJECT_TYPE_DEVICE; | MessageID = 0x6bbb14 | SPIR-V module not valid: [VUID-StandaloneSpirv-Flat-04744] Fragment OpEntryPoint operand 28 with Input interfaces with integer or float type must have a Flat decoration for Entry Point id 4. %gl_SubgroupLocalInvocationId = OpVariable %_ptr_Input_uint Input Cc: mesa-stable Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20756> (cherry picked from commit 2a33d50)
GL lets you set a silly state, so do something plausible instead of undefined. Cc: mesa-stable Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20756> (cherry picked from commit adf8104)
MemoryModelVulkan was left out for CSes using it. Cc: mesa-stable Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20756> (cherry picked from commit 1e4deb3)
Issues including: 1. static declaration of **** follows non-static declaration for function timespec_get 2. ldd error: undefined symbol for functions major and minor 3. ffs and ffsll redefinition issue 4. ldd error: undefined symbol issue for functions build_id and futex 5. TLS symbol error: cannot load any more object with static TLS 6. Intel "Anvil" Vulkan driver requires the dl_iterate_phdr function 7. called object type 'uint32_t' is not a function or function pointer
Android clang add ANDROID note section, which doesn't satify 20 size. This is a workaround for the issue.
Video playback will not work without this patch. INTERNAL: Add importBuffer & release support After we rebased our allocator & mapper from 2.0 to 4.0, we use allocator 4.0 to allocate buffer, which can't be accessed by mapper 2.0 directly. To use the buffer allocated by allocator 4.0, we need importBuffer before lock, and release after unlock. Tracked-On: OAM-95916 Signed-off-by: Ren Chenglei <[email protected]>
This change is to help add NV12, NV12_Y_TILED_INTEL and P010 formats for Android. TESTS: The following CtsViewTestCases cases could pass android.view.cts.PixelCopyTest#testVideoProducer android.view.cts.SurfaceViewSyncTest#testVideoSurfaceViewCornerCoverage android.view.cts.SurfaceViewSyncTest#testVideoSurfaceViewTranslate android.view.cts.SurfaceViewSyncTest#testVideoSurfaceViewRotated android.view.cts.SurfaceViewSyncTest#testVideoSurfaceViewEdgeCoverage Signed-off-by: Lin Johnson <[email protected]> Signed-off-by: Kevin Strasser <[email protected]> Signed-off-by: Chenglei Ren <[email protected]>
This is to help revert part of 57d416d, which causes CTS dEQP regression. Test: dEQP-VK.memory.external_memory_host.* Signed-off-by: Ren Chenglei <[email protected]>
We should confirm height & width valid before call create_resource, otherwise virglrenderer driver will run into error. Tracked-On: OAM-97927 Signed-off-by: Ren Chenglei <[email protected]>
On Android, there will be garbage when screen rotation during video playback. This is to revert d245d7b Signed-off-by: Chenglei Ren <[email protected]> Tracked-On: OAM-99738
This patch try to fix a deadlock issue when runing android.media.cts.HeifWriterTest#testInputSurface_Grid_Handler In this case, there are two deadlock threads involved, the HeifWriter thread and a drawframe thread. The drawframe thread will try to draw a frame and send it to HeifWriter. By calling st_api_make_currentst() it will pending in surface dequeubuffer() with hold of EGL lock, and will block HeifWriter thread executing, the buffer will be returned after HeifWriter handled, In this case HeifWriter waiting for the EGL lock, and drawframe thread wait for buffer with EGL lock hold, so deadlock produced. In case this is not the first time run of make_current(), allocate buffer can be postponed, and allocate buffer in a context without hold of EGL lock. Tracked-On: OAM-96997 Signed-off-by: Yang, Dong <[email protected]> INTERNEL: allocate buffer if there isn't _ColorReadBuffer allocate buffer if there isn't _ColorReadBuffer Tracked-On: OAM-99287 Signed-off-by: Yang, Dong <[email protected]>
Currently the iris driver is adding the buffer objects to zombie list without checking if it is busy or not. It checks for it after 1 second which adds delay to buffer release. This fix checks if the bo is busy or not before adding it to zombie list. Without this fix, the applications expecting immediate buffer release would fail. The fix is identified while debugging below cts tests: android.graphics.cts.BitmapTest#testDrawingHardwareBitmapNotLeaking android.graphics.cts.BitmapTest#testHardwareBitmapNotLeaking Tracked-On: OAM-105862 Signed-off-by: Sai Teja Pottumuttu <[email protected]>
Previously buffer info getting for YUV format by gralloc4 is not complete for egl image creation. Need calculate extra chroma information from layout info. Tracked-On: OAM-106557 Signed-off-by: chenyanxzhu <[email protected]>
This is to help fix Coverity issue in mesa project. Tracked-On: OAM-105550 Signed-off-by: Ren Chenglei <[email protected]>
Tracked-On: OAM-106859 Signed-off-by: Kalenahally Madegowda, NandishaX <[email protected]> Signed-off-by: Kothapeta, BikshapathiX <[email protected]>
Tracked-On: OAM-106720 Signed-off-by: HeYue <[email protected]>
This reverts commit 554b19b. Vulkan failed to run with device enumerate failure on mesa23 due to this patch. Tracked-On: OAM-106720 Signed-off-by: Fei Jiang [email protected]
issue:projectceladon#63 case: testEncode[8(c2.intel.avc.encoder_video/avc_720x480_true_TEST_MODE_DETAILED)](android.media.encoder.cts.VideoEncoderTest) This reverts commit 3d6c7fb. Tracked-On: OAM-108751 Signed-off-by: zhangyichix <[email protected]>
This reverts commit 3d6c7fb. Tracked-On: OAM-108904 Signed-off-by: Jia, Lin A <[email protected]>
It is observed that in display resolutions where width is not equal to stride, vulkan rendering is being distorted. This is happening due to a stride calculation mismatch between minigbm and mesa. This fix makes sure that the stride calculated in minigbm is passed to anv and isl. The fix resolves the following two android CTS issues android.graphics.cts.VulkanPreTransformTest#testVulkanPreTransformNotSetToMatchCurrentTransform android.graphics.cts.VulkanPreTransformTest#testVulkanPreTransformSetToMatchCurrentTransform Tracked-On: OAM-106986 Signed-off-by: Sai Teja Pottumuttu <[email protected]>
chenyanxzhu
force-pushed
the
zhucx_master_agi
branch
from
b6dc368
to
92bc462
Compare
1, Build android-runnable pps-producer. Make pps_producer binary build by android makefile other than meson.build. Otherwise it will fail to run inside android. 2, Fix AGI fail issue: Add GpuCounterDescriptor in data source register stage. 3, Comment util_perfetto_init in eglGetDisplay, which will create a new thread. In the case of system start, zygote will stuck when there's more than 1 thread in waitUntilAllThreadsStopped. Instead, move the init op into util_perfetto_trace_begin. Tracked-On: OAM-110834 Signed-off-by: chenyanxzhu <[email protected]>
chenyanxzhu
force-pushed
the
zhucx_master_agi
branch
from
92bc462
to
0489015
Compare
sysopenci
force-pushed
the
celadon/s/mr0/master
branch
2 times, most recently
from
43d488a
to
ccc6905
Compare
sysopenci
force-pushed
the
celadon/s/mr0/master
branch
from
ccc6905
to
17a6b4f
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
1, Build android-runnable pps-producer. Make pps_producer binary
build by android makefile other than meson.build. Otherwise it
will fail to run inside android.
2, Fix AGI fail issue: Add GpuCounterDescriptor in data source
register stage.
3, Downgrade perfetto api version inside mesa from 29 to 15.
Currently the perfetto version used by perfetto/traced is 15.
4, Comment util_perfetto_init in eglGetDisplay, which will create
a new thread. In the case of system start, zygote will stuck when
there's more than 1 thread in waitUntilAllThreadsStopped. Instead,
move the init op into util_perfetto_trace_begin.
Tracked-On: