Skip to content

Pull requests: openxla/xla

New pull request New
668 Open 44,433 Closed
668 Open 44,433 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
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

Clear-up mlir context pool in the autotuner tool.
#45843 opened Jul 17, 2026 by copybara-service Bot Loading…
@shyamliagrawal
Automated Code Change
#45842 opened Jul 17, 2026 by copybara-service Bot Loading…
Lower approx_top_k to stable_sort+slice on GPU
#45841 opened Jul 17, 2026 by copybara-service Bot Loading…
@apivovarov
[XLA:GPU] Update AddMandatoryConstraints to support async computations with tuple shapes.
#45840 opened Jul 17, 2026 by copybara-service Bot Loading…
@olegshyshkov
Fix spelling mistake
#45838 opened Jul 16, 2026 by copybara-service Bot Loading…
@jakeharmon8
[xla:gpu] Fine grained control of collectives pipelining
#45836 opened Jul 16, 2026 by ezhulenev Member Loading…
2
Bump rules_ml_toolchain to version with C++ modules support
#45835 opened Jul 16, 2026 by copybara-service Bot Loading…
1
Refactor: Extract HLO preparation steps in LayoutAssignment.
#45834 opened Jul 16, 2026 by copybara-service Bot Loading…
[XLA] Add flag to track modifying passes in HLO pass pipeline
#45833 opened Jul 16, 2026 by copybara-service Bot Loading…
Update convolution_type_canonicalizer to handle newly added dot variants
#45832 opened Jul 16, 2026 by copybara-service Bot Loading…
[XLA:GPU] Verify DGRAD feature matching in ConvKindAssignment
#45831 opened Jul 16, 2026 by copybara-service Bot Loading…
@derdrdirk
[XLA:GPU] Skip operand swapping for non-FPROP convolutions in ConvOperandSwapper
#45830 opened Jul 16, 2026 by copybara-service Bot Loading…
@derdrdirk
[XLA:GPU] Support convolution_kind in MakeConvolveHlo helper.
#45829 opened Jul 16, 2026 by copybara-service Bot Loading…
@derdrdirk
[XLA:GPU] Updates HloVerifier to pass the instruction's convolution_kind into ShapeInference::InferConvolveShape.
#45828 opened Jul 16, 2026 by copybara-service Bot Loading…
@derdrdirk
Disable xla_gpu_experimental_ragged_all_to_all_use_barrier_with_nccl by default in the ragged_all_to_all_unit_tests. This requires symmetric memory which is not available on all platforms.
#45826 opened Jul 16, 2026 by copybara-service Bot Loading…
Add shape inference for backward convolutions (DGRAD and WGRAD).
#45825 opened Jul 16, 2026 by copybara-service Bot Loading…
@derdrdirk
[XLA:GPU] Add ConvolutionKind support to ShapeInference::InferConvolveShape. Currently HloVerifier only verifies fordward convolutions (not DGRAD or WGRAD backwards convolutions).
#45824 opened Jul 16, 2026 by copybara-service Bot Loading…
@derdrdirk
Integrate LLVM at llvm/llvm-project@9ec797443297
#45823 opened Jul 16, 2026 by copybara-service Bot Loading…
[XLA:GPU] Add RHS transpose fold test for GEMM rewriter
#45822 opened Jul 16, 2026 by Vaisman Contributor Loading…
[HLO] Validate the dimensions attribute in the HLO parser to avoid crashes
#45821 opened Jul 16, 2026 by EylonKrause Contributor Loading…
Internal code change.
#45819 opened Jul 16, 2026 by copybara-service Bot Loading…
Reverts 05a551d3073814f0284164fde944fadbd84c2abf
#45818 opened Jul 16, 2026 by copybara-service Bot Loading…
@dextero
Improve compilation performance by skipping checks when logging is disabled.
#45817 opened Jul 16, 2026 by copybara-service Bot Loading…
[XLA:GPU] Don't let rank-0 dimension-size scalars inherit op sharding
#45816 opened Jul 16, 2026 by nurmukhametov Contributor Loading…
[ROCm] Remove native suppoprt for exp gfx1250 because of accuracy issue
#45815 opened Jul 16, 2026 by zoranjovanovic-ns Contributor Loading…
ProTip! Type g i on any issue or pull request to go back to the issue listing page.