Fix build break in tfrt_gpu_buffer_test using absl_testing::StatusIs#534
Merged
hsharsha merged 1 commit intoJan 20, 2026
Merged
Conversation
Collaborator
Author
|
upstream PR openxla#36599 |
i-chaochen
approved these changes
Jan 20, 2026
Collaborator
i-chaochen
left a comment
i-chaochen
left a comment
There was a problem hiding this comment.
is this supposed to be failed on NV side as well?
Collaborator
Author
|
It should fail on NV side as well. This test is not picked up as it has no_oss tag |
Collaborator
hmmmm....maybe the real "fix" is to add |
hsharsha
force-pushed
the
jax_0.8.2_fix_build_break_absl_testing
branch
from
3bb81d5 to
f8f771b
Compare
hsharsha
force-pushed
the
jax_0.8.2_fix_build_break_absl_testing
branch
from
f8f771b to
2b383d5
Compare
nurmukhametov
pushed a commit
that referenced
this pull request
Jan 22, 2026
nurmukhametov
pushed a commit
that referenced
this pull request
Jan 22, 2026
nurmukhametov
pushed a commit
that referenced
this pull request
Jan 23, 2026
nurmukhametov
pushed a commit
that referenced
this pull request
Jan 23, 2026
nurmukhametov
pushed a commit
that referenced
this pull request
Jan 23, 2026
nurmukhametov
pushed a commit
that referenced
this pull request
Jan 23, 2026
nurmukhametov
pushed a commit
that referenced
this pull request
Jan 23, 2026
nurmukhametov
pushed a commit
that referenced
this pull request
Jan 23, 2026
nurmukhametov
pushed a commit
that referenced
this pull request
Jan 26, 2026
i-chaochen
added a commit
that referenced
this pull request
Feb 9, 2026
* [ROCm] Build infrastructure and CI scripts * Fix infinite recursion in HloInstruction::Accept/Visit const wrappers (#470) The const wrapper methods for Accept() and Visit() were calling themselves instead of the template versions, causing infinite recursion and stack overflow. * Mark nvshmem tests as cuda-only (#458) * Skipped CanNotEmitTritonCustomCallOnPreAmpereGpu test for ROCM. * Make device_count_ atomic (#343) * Make device_count_ atomic * Use relaxed memory order * Fix build error * [ROCm] Enable embeded bitcode libs and inprocess lld (#507) Added TF_ROCM_INPROCESS_LLD and TF_ROCM_EMBEDDED_DEVICE_LIB form 0.6.0 otherwise identical to openxla#32439. Env vars only needed for 0.8.0. * [ROCm] Pass warp size to Triton compilation pipeline * [ROCm] Add FNUZ FP8 type support in Triton * [ROCm] Temporary workaround for column reduction warp size * PR openxla#36046: [ROCm] Fix failing unit tests on ROCm platform Imported from GitHub PR openxla#36046 📝 Summary of Changes - layout_assignment tests are marked cuda-only. - sample_file_test needs higher autotuner level for MIOpen to return conv algorithm. Earlier this was coming from GetDebugOptionsForTest. - buffer_debug_log test is made gpu agnostic by using cannonical gpu name. - cublas_gemm_rewriter_test_amdgpu_any fix unit test to remove padding for ROCm as introduced in openxla#33854 - gpu_kernel_tiling_test_amdgpu_any is updated to respect higher launch dimensions now supported by hipruntime - Mark dynamic_shared_memory_test as cuda-only - Add arch specific checks for barriers to sorting.hlo 🎯 Justification Fixes failing unit tests on ROCm platform * Fix build break in tfrt_gpu_buffer_test using absl_testing::StatusIs (#534) * Port transpose changes from v0.8.0 to v0.8.2 (#526) It should be dropped after the rebase on top of 330a305 * [ROCm] Fix failing test TritonEmitterTest/RocmWarpSizeIsSetCorrectly (#545) * [ROCm] Fix failing test TritonEmitterTest/RocmWarpSizeIsSetCorrectly Define valid tile parameters and non-zero shared memory. * Update xla/backends/gpu/codegen/triton/fusion_emitter_device_test.cc Co-authored-by: Maxime France-Pillois <[email protected]> * Update xla/backends/gpu/codegen/triton/fusion_emitter_device_test.cc Co-authored-by: Maxime France-Pillois <[email protected]> --------- Co-authored-by: Maxime France-Pillois <[email protected]> * Fix MIOpen linking for RNN kernels Add explicit linkopts to miopen cc_library target to ensure libMIOpen.so is properly linked at runtime. This fixes AttributeError: module 'jaxlib.gpu_rnn' has no attribute 'compute_rnn_workspace_reserve_space_sizes' in experimental_rnn_test in JAX. Without this change, the _rnn.so shared library fails to load MIOpen symbols properly, causing RNN test failures. * Force rbe incompatible tests to be executed locally (#485) * [ROCm] Add missing cuda-only tag * enable mx datatype for rocm (#462) * enable mx datatype for rocm * add // TF_ROCM_VERSION >= 70000 * fix unit test build * Add rocprofiler-sdk (v3) integration with roctracer fallback Integrate rocprofiler-sdk for ROCm profiling with fallback to roctracer (v1) when rocprofiler-sdk is not available. * [ROCm] Always process convolutions through MIOpen backend for decomposition Override AddConvAndGemmAutotuningPass in AMDGPUCompiler to ensure convolutions are always sent to MIOpen for processing, regardless of xla_gpu_autotune_level. This is required because MIOpen handles decomposition of unsupported fused convolutions back to regular convs, which must happen even when autotuning is disabled. Fixes cudnn_fused_conv_rewriter_autotune_disabled_test failures on ROCm. * Changed error value for SplitK test in fusion_emitter_device_legacy_port_test.cc (#538) * [ROCm] Add PJRT_Triton_Extension support (#548) This change is PJRT_Triton_Extension support for ROCm as counterpart of that for CUDA. Pallas Triton calls are lowered to HSACO directly rather than PTX on ROCm platform. * Fix expected output in fusion_emitter_int4_device_test for ROCm. * skip conditional graph tests * Fixed missing rtne in Triton to pass support_test. * [ROCm] Add rocm-only tag to triton_rocm target Fix dependency validation by tagging triton_rocm as rocm-only since it depends on the rocm-only amdgpu_backend target. * Avoid upcast of lib func operands to F32 for F16 type. * Modify fusion_emitter_large_test to work on ROCm. (#568) * Modify fusion_emitter_large_test to work on ROCm. * Fix fall-through warning in support_legacy.cc * Fixed dot_algorithms_test. Updated support_legacy and test itself. * Modified triton_fusion_numerics_verifier_test to work on ROCm. * [ROCm] Use shared AsBlasLtEpilogue in GemmWorkspaceRewriter Replace the duplicate with the shared function to fix the issue and prevent future divergence. The duplicate AsBlasLtEpilogue in gemm_workspace_rewriter.cc was missing SILU epilogue support, breaking ROCm Swish fusion tests. This duplicate was introduced in PR openxla#35132. * Sync mgpu tests with xla_mgpu config * [ROCm] Fix RocmWarpSizeIsSetCorrectly test to use new dump file naming After commit 4ce9326, Triton pass dumps use the naming pattern {module}.{kernel}.{pass_manager_name}.txt instead of *.triton-passes.log. Update the test to match the new convention. * Enable hlo_runner_main_gpu for rocm * enable hipblaslt as a default choice and disable nccl comm split to avoid hanging * Add flag to control swish activation fusion. (#577) Add flag to control swish activation fusion. * Improve test strategy for swish fusion flag (#585) Move tests to a more suitable file. * Revert "Fix infinite recursion in HloInstruction::Accept/Visit const wrappers (#470)" This reverts commit 21a2d57. * Disable hipblaslt as default choice * Execute test directly if running on system without GPU (#608) * Execute test directly if running on system without GPU * Address review comments * Address review comments * Remove non-existent test targets from ROCm CI exclusion list The following targets no longer exist in their respective BUILD files and were causing Bazel target pattern parsing failures. * Bundle librocm_smi64.so for MI200 lit tests MI200 lit tests use hlo-opt which links against ROCm libraries. When running on remote workers without ROCm installed, hlo-opt fails with: "error while loading shared libraries: librocm_smi64.so.1" The _tools_on_path rule bundles libraries into lit_lib/ by extracting them from CcInfo.linking_context.linker_inputs[].dynamic_library. However, ROCm's cc_library targets with .so files in srcs don't populate dynamic_library (unlike CUDA which uses cc_import). Add a new rocm_smi_import target using cc_import, which properly exposes the shared library via CcInfo. Use this target in lit.bzl so librocm_smi64.so.1 gets bundled into lit_lib/ and is available at runtime via hlo-opt's rpath. --------- Co-authored-by: Pham Binh <[email protected]> Co-authored-by: Alex <[email protected]> Co-authored-by: Zoran Jovanovic <[email protected]> Co-authored-by: Dragan Mladjenovic <[email protected]> Co-authored-by: Harsha H S <[email protected]> Co-authored-by: Maxime France-Pillois <[email protected]> Co-authored-by: magaonka-amd <[email protected]> Co-authored-by: Xuefei Jiang <[email protected]> Co-authored-by: cj401-amd <[email protected]> Co-authored-by: zoranjovanovic-ns <[email protected]> Co-authored-by: Jian Li <[email protected]> Co-authored-by: Chao Chen <[email protected]> Co-authored-by: Alexandros Theodoridis <[email protected]> Co-authored-by: Milica Makevic <[email protected]>
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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.
📝 Summary of Changes
Use absl_testing::StatusIs instead of testing::status::StatusIs and extend the test to all supported gpu architectures.
🎯 Justification
Fixes build break
🚀 Kind of Contribution
Please remove what does not apply: 🐛 Bug Fix, 🧪 Tests
Submission Checklist