Skip to content

Pull requests: openxla/xla

New pull request New
1,623 Open 14,955 Closed
1,623 Open 14,955 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

Reverts 3e3efb1ba1dd6c19e218320e2c0248f112586e56
#16964 opened Sep 9, 2024 by copybara-service bot Loading…
@klucke
Use ScopedActivateContext in GpuCudaMallocAsyncAllocator instead of the ScopedActivateExecutorContext wrapper class.
#16963 opened Sep 9, 2024 by copybara-service bot Loading…
@klucke
Use ScopedActivateContext instead of wrapper ScopedActivateExecutorContext class in rocm classes.
#16962 opened Sep 9, 2024 by copybara-service bot Loading…
@klucke
Add more corner cases in tests for Div() and tighten the error tolerances for AbsComplex on TPU.
#16961 opened Sep 9, 2024 by copybara-service bot Loading…
Use ScopedActivateContext instead of the ScopedActivateExecutorContext wrapper class.
#16960 opened Sep 9, 2024 by copybara-service bot Loading…
@klucke
[cuda] Cache the results of slow cuda_asm_compiler functions
#16959 opened Sep 9, 2024 by sergey-kozub Loading…
@sergey-kozub
1
Allow fusion to accept more operands than parameters. This relaxation is mainly for window prefetch. With this change, now, the MSA can just append operands, without inserting a custom op inside the fusion to consume the appended operands.
#16958 opened Sep 9, 2024 by copybara-service bot Loading…
[IFRT] Verify that devices is not nullptr when used in Sharding
#16957 opened Sep 9, 2024 by copybara-service bot Loading…
Make it possible to avoid logging of the errors if needed
#16956 opened Sep 9, 2024 by copybara-service bot Loading…
@denisvnukov
Reduces logging verbosity when skipping setting shardings.
#16955 opened Sep 9, 2024 by copybara-service bot Loading…
Avoid dynamic-slice-fusion when reduce-scatter has tuple output
#16954 opened Sep 9, 2024 by shraiysh Loading…
De-duplicate code to read a NUMA node into a common library to be used by both ROCm and CUDA executors.
#16952 opened Sep 9, 2024 by copybara-service bot Loading…
@klucke
Automated rollback of 38 IncludeCleaner findings
#16951 opened Sep 9, 2024 by copybara-service bot Loading…
Clarify that sm_90a should only be used on Hopper
#16950 opened Sep 9, 2024 by dimvar Loading…
Revert header cleanup
#16949 opened Sep 9, 2024 by copybara-service bot Loading…
@akuegel
Remove unused code in driver & executor files.
#16948 opened Sep 9, 2024 by copybara-service bot Loading…
@klucke
Integrate LLVM at llvm/llvm-project@3940a1ba1454
#16946 opened Sep 9, 2024 by copybara-service bot Loading…
Automated Code Change
#16945 opened Sep 9, 2024 by copybara-service bot Loading…
Handle when JAX saves the manual axes on the CallOp of the shmap_body.
#16944 opened Sep 9, 2024 by copybara-service bot Loading…
@bartchr808
#sdy Handle redundant shard maps during import.
#16943 opened Sep 9, 2024 by copybara-service bot Loading…
@bartchr808
Rename GemmFusion to TritonFusionRewriter.
#16939 opened Sep 9, 2024 by copybara-service bot Loading…
@derdrdirk
Add NANOO FP8 support for collaborative communication unit tests
#16938 opened Sep 9, 2024 by ScXfjiang Loading…
[XLA:GPU][IndexAnalysis] Unify printing of indexing map class and attribute.
#16937 opened Sep 9, 2024 by copybara-service bot Loading…
@pifon2a
[XLA:GPU] Fix power of 2 padding for tiled transpose.
#16933 opened Sep 9, 2024 by copybara-service bot Loading…
@olegshyshkov
Automated Code Change
#16922 opened Sep 9, 2024 by copybara-service bot Loading…
ProTip! Add no:assignee to see everything that’s not assigned.