Pull Requests openxla/xla

[xla:gpu] Add device comms requirements to clique requests

#36417 opened 2026-01-14 19:44 by ezhulenev

Enhance Symbolic Expression Parser

#36416 opened 2026-01-14 19:29 by copybara-service[bot]

Add pjrt extension for accelerator metadata

#36415 opened 2026-01-14 18:35 by copybara-service[bot]

Check if caller's struct has space to hold the new field.

#36414 opened 2026-01-14 18:12 by copybara-service[bot]

[ROCm][XLA:GPU] Add ROCm-specific tuning for transpose emitter

#36413 opened 2026-01-14 17:55 by nurmukhametov

[Autotuner] Keep scaled dot configs in sync with gemm_fusion_autotuner.cc.

#36412 opened 2026-01-14 17:10 by copybara-service[bot]

Implement `NamedSharding::ToTileAssignment` converting `NamedSharding` to `TileAssignment` representation, required for implementing methods involving tile based indexing `TileIndexForDevice`, `TileOffsetForDevice` etc, e.g: cl/855697055

#36410 opened 2026-01-14 15:19 by copybara-service[bot]

Implement validation checks for `named_sharding.cc`.

#36409 opened 2026-01-14 14:55 by copybara-service[bot]

[XLA:GPU] Migrate runtime_intrinsics_test to HloPjRtTestBase.

#36407 opened 2026-01-14 14:51 by copybara-service[bot]

PR #36401: [XLA:GPU] Add TMA config to autotuning when possible

#36406 opened 2026-01-14 14:26 by copybara-service[bot]

Add `ScanExpander` to rewrite HLO `scan` into a `while` loop.

#36404 opened 2026-01-14 13:43 by copybara-service[bot]

Move thunk proto definitions into separate files

#36402 opened 2026-01-14 10:53 by copybara-service[bot]

[XLA:GPU] Add TMA config to autotuning when possible

#36401 opened 2026-01-14 10:38 by and-ivanov

Migrate gpu_compiler_test to HloPjrtTestBase.

#36399 opened 2026-01-14 10:14 by copybara-service[bot]

PR #36046: [ROCm] Fix failing unit tests on ROCm platform

#36397 opened 2026-01-14 08:15 by copybara-service[bot]

Refactor RunAsync to take std::vector<PjRtStreamExecutorExecutionInput>

#36394 opened 2026-01-14 05:08 by copybara-service[bot]

Integrate LLVM at llvm/llvm-project@95da1354d03b

#36393 opened 2026-01-14 04:54 by copybara-service[bot]

[READABILITY] Use idiomatic boolean conversion for std::optional checks

#36392 opened 2026-01-14 03:59 by copybara-service[bot]

Move collective_ops_e2e_test_base to xla/service/gpu/tests/

#36389 opened 2026-01-14 01:07 by copybara-service[bot]

[xla:gpu] Remove persistent collective cliques

#36388 opened 2026-01-14 00:55 by ezhulenev

[XLA:GPU] Migrate nccl_group_execution_test to always use PjRt for its test backend.

#36386 opened 2026-01-13 23:47 by copybara-service[bot]

Try to detect if on CUDA at runtime

#36385 opened 2026-01-13 23:06 by copybara-service[bot]

[ReplicaGroupV3][Refactor][10/n] Use CollectiveDeviceList accessor methods for HLOInstruction.

#36382 opened 2026-01-13 22:14 by copybara-service[bot]

PR #34853: [xla:emitters][oneAPI] Enable math.log1p in MathToLLVMConversionPatterns

#36380 opened 2026-01-13 21:58 by copybara-service[bot]

[Scythe] delete unused `GetParticipatingFlattenedIdGroups` function.

#36378 opened 2026-01-13 21:28 by copybara-service[bot]

More testing to understand kokoro-only flake.

#36377 opened 2026-01-13 21:12 by copybara-service[bot]

PR #32217: [Refactor] Completely Remove AsyncStreamKind

#36376 opened 2026-01-13 21:11 by copybara-service[bot]

PR #35456: Manage events for intra-process data transfers after enqueue of dependencies instead of completion

#36375 opened 2026-01-13 21:06 by copybara-service[bot]

Move ShapedSlice from gpu specific directory

#36374 opened 2026-01-13 20:44 by copybara-service[bot]

[PJRT] Propagate size to unmap_fn for HostMemoryAllocator.

#36372 opened 2026-01-13 19:51 by copybara-service[bot]