Commits

Branch selector

 main

User selector

All users

All time

Commit History

Commits on May 17, 2026

• Automated Code Change

  Show description for 1ac2f29
  Google-ML-Automation

  committed
  1ac2f29

  Copy full SHA for 1ac2f29

Commits on May 16, 2026

• Automated Code Change

  Show description for 3518079
  Google-ML-Automation

  committed
  3518079

  Copy full SHA for 3518079

• Integrate LLVM at llvm/llvm-project@88afb5af47a9

  Show description for 946508c
  Google-ML-Automation

  committed
  946508c

  Copy full SHA for 946508c

• Enable `xla_enable_rgv3_materialization` in the default HLO module config for SPMD tests and assert on `device_assignment` for `OriginalValueWithTrivialShardingAnnotation` and `OriginalValueWithTra…

  Show description for f2baa01
  

  Kingston Mandisodza

  authored and

  Google-ML-Automation

  committed
  f2baa01

  Copy full SHA for f2baa01

• [XLA:GPU] fusion_block_level_rewriter: do not force multi-output fusions

  Show description for ccaeb5f
  

  metaflow

  authored and

  Google-ML-Automation

  committed
  ccaeb5f

  Copy full SHA for ccaeb5f

• PR #41038: [ROCm] Use hermetic llvm to compile xla

  Show description for 4faa405
  

  alekstheod

  authored and

  Google-ML-Automation

  committed
  4faa405

  Copy full SHA for 4faa405

• Skip `NoReshardOnBroadcastDims` for V3 and write a V3 test for the same behaviour. This is because `NoReshardOnBroadcastDims` generates a collective op rather than a copy due to the inclusion of in…

  Show description for ecf7836
  

  Kingston Mandisodza

  authored and

  Google-ML-Automation

  committed
  ecf7836

  Copy full SHA for ecf7836

• Automated Code Change

  Show description for 6228a2a
  Google-ML-Automation

  committed
  6228a2a

  Copy full SHA for 6228a2a

• Add GetMemorySpaceKindIds() to PjRtTopologyDescription.

  Show description for 3340bb4
  

  pschuh

  authored and

  Google-ML-Automation

  committed
  3340bb4

  Copy full SHA for 3340bb4

• Optimize ArgMinMax (Reduce) when fed by scalar constant via broadcast, and simple ReduceWindow implementing cumsum when fed by scalar constant.

  Show description for 49f67c8
  

  jbspooner

  authored and

  Google-ML-Automation

  committed
  49f67c8

  Copy full SHA for 49f67c8

Commits on May 15, 2026

• PR #42464: [XLA:GPU][oneAPI] Part4 of oneAPI plugin: SE GPU PJRT CLIENT changes

  Show description for b3c28f5
  

  ashiqimranintel

  authored and

  Google-ML-Automation

  committed
  b3c28f5

  Copy full SHA for b3c28f5

• PR #34219: Document "--xla_syntax_sugar_async_ops" flag

  Show description for f4d25f7
  

  pavithraes

  authored and

  Google-ML-Automation

  committed
  f4d25f7

  Copy full SHA for f4d25f7

• [GPU] Migrate RequiresCollectiveSymmetricMemorySpace to unified collective buffer analysis

  Show description for 646df5f
  

  apivovarov

  authored and

  Google-ML-Automation

  committed
  646df5f

  Copy full SHA for 646df5f

• PR #40593: [XLA:GPU] Ensure that the host and target platforms match for AOT compilation

  Show description for e96b7ec
  

  akhilgoe

  authored and

  Google-ML-Automation

  committed
  e96b7ec

  Copy full SHA for e96b7ec

• Support `PartialReplication` in `HloSharding::NumTiles`

  Show description for beabb66
  

  KanishAnand

  authored and

  Google-ML-Automation

  committed
  beabb66

  Copy full SHA for beabb66

• Refactor `UniqueTiledDim` to re-use `HloSharding::TiledDataRank` function

  Show description for 4179405
  

  KanishAnand

  authored and

  Google-ML-Automation

  committed
  4179405

  Copy full SHA for 4179405

• Implement MakeCanonicalShapeForMemorySpace for GPU and c-api.

  Show description for 1c0eb97
  

  pschuh

  authored and

  Google-ML-Automation

  committed
  1c0eb97

  Copy full SHA for 1c0eb97

• make compilation of llvm ir to ptx async for Triton

  Show description for 31ea922
  

  ermilovmaxim

  authored and

  Google-ML-Automation

  committed
  31ea922

  Copy full SHA for 31ea922

• PR #37936: [NVIDIA GPU] Reduce rendezvous frequency in gpu executable

  Show description for b5cbfba
  

  Tixxx

  authored and

  Google-ML-Automation

  committed
  b5cbfba

  Copy full SHA for b5cbfba

• Integrate LLVM at llvm/llvm-project@4f52e220a69f

  Show description for 92c55b4
  Google-ML-Automation

  committed
  92c55b4

  Copy full SHA for 92c55b4

• The fusion_constant_sinking pass is updated to support sinking constants into custom fusions. Sinking is permitted only if both the custom fusion and the constant operand have a MUST_FUSE frontend …

  Show description for e42e4b5
  Google-ML-Automation

  committed
  e42e4b5

  Copy full SHA for e42e4b5

• CUDA 13 Release build changes

  Show description for a42b06b
  

  dmiltr3

  authored and

  Google-ML-Automation

  committed
  a42b06b

  Copy full SHA for a42b06b

• strip_memory_placement_annotations_test uses HloTestBase, but it's deprecated.

  Show description for 5f49eb7
  

  nvgrw

  authored and

  Google-ML-Automation

  committed
  5f49eb7

  Copy full SHA for 5f49eb7

• collect spilled register stats for CustomKernelThunks

  Show description for fc8e289
  

  ermilovmaxim

  authored and

  Google-ML-Automation

  committed
  fc8e289

  Copy full SHA for fc8e289

• Fix `AnalyzeTileAssignment` in `tile_assignment.cc` to construct the mesh based on sub dims rather than tile assignment.

  Show description for def8c4b
  

  Kingston Mandisodza

  authored and

  Google-ML-Automation

  committed
  def8c4b

  Copy full SHA for def8c4b

• [PART 2] Externalize HLO isolation test into a new open-source API.

  Show description for 2d5b2f3
  Google-ML-Automation

  committed
  2d5b2f3

  Copy full SHA for 2d5b2f3

• Make it possible for users of unsafe_client_pointer to fetch a PJRT_Api* from a PJRT_Client*.

  Show description for cdde79b
  

  pschuh

  authored and

  Google-ML-Automation

  committed
  cdde79b

  Copy full SHA for cdde79b

• Revert zstd.patch update due to source of changes being reverted in https://github.com/openxla/xla/pull/42661.

  Show description for 1abf5a2
  

  belitskiy

  authored and

  Google-ML-Automation

  committed
  1abf5a2

  Copy full SHA for 1abf5a2

• [XLA:GPU] Handle zero tile sizes in XTile emitters

  Show description for 0c35d85
  

  vickyliu-go4it

  authored and

  Google-ML-Automation

  committed
  0c35d85

  Copy full SHA for 0c35d85

• Integrate LLVM at llvm/llvm-project@adb5802b494c

  Show description for 90dd6d3
  Google-ML-Automation

  committed
  90dd6d3

  Copy full SHA for 90dd6d3

• Support `HloShardingV3` in `MakeTiledPartitionOrdinals` where specific removal of replicated dimensions is not required for v3 sharding as `HloShardingV3::dimensions()` only contains non-replicated…

  Show description for dfb2813
  

  KanishAnand

  authored and

  Google-ML-Automation

  committed
  dfb2813

  Copy full SHA for dfb2813

• [XLA:CPU] Implement StripMemoryPlacementAnnotations pass and make layout source of truth for output memory kinds

  Show description for c56c6be
  

  seantalts

  authored and

  Google-ML-Automation

  committed
  c56c6be

  Copy full SHA for c56c6be

• Re-enable CuDNN fusion tests and relax workspace size check.

  Show description for dba8ab6
  

  beckerhe

  authored and

  Google-ML-Automation

  committed
  dba8ab6

  Copy full SHA for dba8ab6

• Automated Code Change

  Show description for f8c9fc9
  Google-ML-Automation

  committed
  f8c9fc9

  Copy full SHA for f8c9fc9

• PR #41817: [XLA:GPU] Migrate RaggedAllToAll command buffer support to thunk

  Show description for 1d8807b
  

  shawnwang18

  authored and

  Google-ML-Automation

  committed
  1d8807b

  Copy full SHA for 1d8807b