issues
search
openxla
/
xla
A machine learning compiler for GPUs, CPUs, and ML accelerators
Apache License 2.0
2.64k
stars
418
forks
source link
issues
Newest
Newest
Most commented
Recently updated
Oldest
Least commented
Least recently updated
Introduce a more fine-grained target for stream_executor/platform/initialize.h.
#17904
copybara-service[bot]
closed
1 week ago
0
Split GpuCommandBuffer into a CUDA and a ROCm specific version
#17903
copybara-service[bot]
opened
1 week ago
0
Add an algebraic simplification pattern for multiply(add(conv(input, filter), bias), broadcast(constant)) -> add(conv(input, multiply(filter, broadcast(constant))), multiply(bias, broadcast(constant)))
#17902
copybara-service[bot]
closed
1 week ago
0
If the auto-sharding pass cannot find a solution for any of the mesh shapes tried, return an error message listing the errors encountered for each of the mesh shapes. This is intended to make debugging easier.
#17901
copybara-service[bot]
closed
1 week ago
0
[ROCm] Fix build break in executor and kernel test introduced in f896afd
#17900
hsharsha
closed
1 week ago
0
[XLA:GPU] Fix `GpuFloatSupport` for reductions.
#17899
copybara-service[bot]
closed
1 week ago
0
Host Offloading: Process "MoveToHost" instructions in the order they are executed.
#17898
copybara-service[bot]
closed
1 week ago
0
Abandons Auto Sharding for any partial mesh shape that results in a suboptimal solution.
#17897
copybara-service[bot]
closed
1 week ago
0
[TEST] Debug linking of mlir_fusion_opt
#17896
copybara-service[bot]
opened
1 week ago
0
Reverts 693ee2e13225331bebc946442af7e2d59355adea
#17895
copybara-service[bot]
opened
1 week ago
0
Reverts da687fc523c261a6d51f28cc35ef7d188a2a6391
#17894
copybara-service[bot]
closed
1 week ago
0
[XLA:GPU] Move (gated) call to `FusionBlockLevelRewriter` after all possible
#17893
copybara-service[bot]
closed
1 week ago
0
PR #17819: Added use_enabled_free_threading flag to build nanobind with NB_FREE_THREADED=1
#17892
copybara-service[bot]
closed
1 week ago
0
#sdy add JAX Shardy support for memories.
#17891
copybara-service[bot]
closed
3 days ago
0
[ROCM] Add nanoo fp8 support in type traits
#17890
ScXfjiang
closed
3 days ago
3
[XLA:GPU][Cleanup] Remove pre-Ampere paths in GEMM fusion autotuner.
#17889
copybara-service[bot]
closed
1 week ago
0
Integrate LLVM at llvm/llvm-project@00128a20eec2
#17888
copybara-service[bot]
closed
1 week ago
0
[XLA:GPU] Regression in FP8 matmul scaling fusion
#17887
balancap
opened
1 week ago
0
[NVIDIA] Optimize deterministic scalar scatter
#17886
serach24
opened
1 week ago
2
Fix nextafter for FP8 FNUZ types.
#17885
copybara-service[bot]
closed
1 week ago
0
[HLO Componentization] Create hlo/parser sub-component (Phase II).
#17884
copybara-service[bot]
closed
1 week ago
0
Avoid acquiring BufferSequencingEvent::mu_ twice in the case where the thread_pool executes the callbacks inline.
#17883
copybara-service[bot]
closed
1 week ago
0
Add Stat type for Source Stack to show in trace viewer
#17882
copybara-service[bot]
closed
1 week ago
0
[HLO Componentization] Create hlo/tools sub-component (Phase I).
#17881
copybara-service[bot]
closed
1 week ago
0
PR #17453: Reorder Collective Optimization Passes
#17880
copybara-service[bot]
closed
1 week ago
0
Remove unneeded dso_loader dependencies.
#17879
copybara-service[bot]
closed
1 week ago
0
[XLA:GPU] Fix comments in collective select folder
#17878
copybara-service[bot]
closed
1 week ago
0
Add missing cuda-only tags to stream_executor/cuda targets
#17877
copybara-service[bot]
closed
1 week ago
0
PR #17858: [ROCm] Fix build brake caused by missing dso_loader.h includes
#17876
copybara-service[bot]
opened
1 week ago
1
Integrate LLVM at llvm/llvm-project@485237413577
#17875
copybara-service[bot]
opened
1 week ago
0
Remove AutoShardingResult in favor of a boolean now that the value kModuleUnchangedNoShardingPerformed of the enum is unused, effectively making it a boolean. Also simplified away some dead code.
#17874
copybara-service[bot]
closed
1 week ago
0
[XLA:GPU] Remove the now obsolete `--xla_gpu_enable_triton_hopper` flag.
#17873
copybara-service[bot]
closed
1 week ago
0
Internal change: fix non-copyable object from TF_ASSIGN_OR_RETURN
#17872
copybara-service[bot]
opened
1 week ago
0
Clear caches on jax exit.
#17871
copybara-service[bot]
closed
1 week ago
0
[XLA] Fix typos in comments and clean up includes in HloRematerialization.
#17870
copybara-service[bot]
closed
1 week ago
0
Make rocm files use tsl DsoLoader functions instead of the stream_executor wrappers.
#17869
copybara-service[bot]
closed
1 week ago
0
Always use provided tmpdir if specified by user
#17868
copybara-service[bot]
opened
1 week ago
0
Simplify error handling in auto-sharding.
#17867
copybara-service[bot]
closed
1 week ago
0
[XLA:Python] Fix more bugs in the weakref_lru_cache implementation.
#17866
copybara-service[bot]
closed
1 week ago
0
[ROCm] Fixed linker issues with rocblas_get_version_string_size and r…
#17865
zoranjovanovic-ns
closed
1 week ago
0
Associates names with individual input sharding combinations (rather than strategies).
#17864
copybara-service[bot]
closed
1 week ago
0
[XLA:CPU] Return error when trying to create a view of an unaligned buffer.
#17863
copybara-service[bot]
closed
1 week ago
1
Integrate LLVM at llvm/llvm-project@6292f117c39b
#17862
copybara-service[bot]
closed
1 week ago
0
Integrate LLVM at llvm/llvm-project@6292f117c39b
#17861
copybara-service[bot]
opened
1 week ago
0
Delete xla_client.execute_with_python_values.
#17860
copybara-service[bot]
closed
1 week ago
0
Revert "Remove unneeded dso_loader.h include from ROCM files."
#17859
hsharsha
closed
1 week ago
2
[ROCm] Fix build brake caused by missing dso_loader.h includes
#17858
mmakevic-amd
closed
1 week ago
1
PR #17776: Relax the error tolerance of UnaryElementwiseTest.ElementwiseFusionExecutesCorrectly
#17857
copybara-service[bot]
closed
1 week ago
0
[XLA:GPU] Unify DimVar, RangeVar and RTVar.
#17856
copybara-service[bot]
closed
1 week ago
0
[XLA:CPU][oneDNN] Add post-ops for oneDNN Convolutions
#17855
akhilgoe
opened
1 week ago
0
Previous
Next