Pull Requests openxla/xla

Refine OpenXLA AI development documentation

#36104 opened 2026-01-08 07:23 by copybara-service[bot]

Use override_hostname from ProfileOptions

#36103 opened 2026-01-08 06:47 by copybara-service[bot]

[XLA:SchedulingGroups] Gaps should ignore false dependencies caused by opt-barriers and simple tuples. Gap search algorithm should skip them and go directly from the operand to the respective get-tuple-element, e.g.:

#36101 opened 2026-01-08 06:19 by copybara-service[bot]

Automated Code Change

#36098 opened 2026-01-08 04:16 by copybara-service[bot]

Automated Code Change

#36097 opened 2026-01-08 04:11 by copybara-service[bot]

PR #35750: [se] Add support for custom kernel argument packing via template specialization

#36095 opened 2026-01-08 04:07 by copybara-service[bot]

PR #36075: [xla:gpu] CommandBuffer: clean up child commands APIs

#36094 opened 2026-01-08 04:04 by copybara-service[bot]

Add support for Send and Recv Thunk in Command Buffer.

#36093 opened 2026-01-08 03:57 by copybara-service[bot]

Fixes memory space propagation for nested fusions and add cycle detection.

#36092 opened 2026-01-08 03:51 by copybara-service[bot]

Internal change only

#36085 opened 2026-01-08 02:27 by copybara-service[bot]

Expose total memory size in PjRt CompiledMemoryStats.

#36084 opened 2026-01-08 02:23 by copybara-service[bot]

switch from deprecated TF_CHECK_OK

#36083 opened 2026-01-08 00:51 by copybara-service[bot]

Add Execution cancellation to IFRT.

#36078 opened 2026-01-07 23:12 by copybara-service[bot]

[IFRT] Add an execution cancellation API

#36076 opened 2026-01-07 22:25 by copybara-service[bot]

[xla:gpu] CommandBuffer: clean up child commands APIs

#36075 opened 2026-01-07 22:07 by ezhulenev

[XLA] Add function to clone with provided computation.

#36074 opened 2026-01-07 21:53 by copybara-service[bot]

Integrate LLVM at llvm/llvm-project@53f84636eb86

#36073 opened 2026-01-07 21:47 by copybara-service[bot]

[XLA:HLO] Update async_chain_next_ if async-ops are being replaced.

#36072 opened 2026-01-07 21:44 by copybara-service[bot]

[XLA:GPU] Migrate gpu_cub_sort_test to HloPjRtTestBase.

#36071 opened 2026-01-07 20:32 by copybara-service[bot]

Add Shape to (Memset32|ReplicaId)Thunk buffer_uses

#36068 opened 2026-01-07 19:24 by copybara-service[bot]

Disable heuristic collective combining by default.

#36066 opened 2026-01-07 18:12 by copybara-service[bot]

[XLA:GPU][XTile] Move setting warp_specialization to lowering pass.

#36063 opened 2026-01-07 17:02 by copybara-service[bot]

[XLA:CPU] Don't create a elemental thunk for Transpose.

#36062 opened 2026-01-07 16:54 by copybara-service[bot]

Update hlo_sharding.cc to use num_dimension(s) rather than tile_assignment_.num_dimension(s)

#36061 opened 2026-01-07 16:38 by copybara-service[bot]

Integrate LLVM at llvm/llvm-project@c958a7c66ec4

#36060 opened 2026-01-07 16:20 by copybara-service[bot]

PR #35304: [XLA:GPU] [plugin] Adding sycl/oneapi plugin support for Intel GPU

#36059 opened 2026-01-07 16:02 by copybara-service[bot]

Change the `Platform::Id` definition to also define the platform name

#36058 opened 2026-01-07 15:59 by copybara-service[bot]

PR #35733: [XLA:GPU][oneAPI] SYCL executor and tests

#36057 opened 2026-01-07 15:57 by copybara-service[bot]

[XLA:GPU] Add Python stack traces to XLA NaN checker log messages

#36056 opened 2026-01-07 15:47 by copybara-service[bot]

Use `xla::llvm_ir::CreateMlirModuleOp()` to replace `builder.create<ModuleOp>()`

#36055 opened 2026-01-07 15:38 by copybara-service[bot]