| COMMIT |
0.70 |
Refactor Subgraph construction to use std::make_unique |
|
Explicitly states 'This change is AI gen |
2026-04-13 |
| PR |
0.15 |
Upgrade curl to 8.18.0 and update libcurl integration |
|
Free-text uses terse, domain-specific ch |
2026-04-13 |
| COMMIT |
0.10 |
PR #40594: [XLA:GPU] Implement PutSignal, Signal, and WaitSi |
|
Slightly more formal, but domain-specifi |
2026-04-10 |
| PR |
0.10 |
[Fix compilation error using MSVC] Add Tensor and Helper for |
|
Direct, technical and minimal phrasing; |
2026-04-12 |
| PR |
0.05 |
Fix AArch64 CPUIDInfo init |
|
Brief, informal explanation with typos; |
2025-10-14 |
| COMMIT |
0.00 |
PR #40751: [xla:gpu] Delete special case for calls to comman |
|
Casual tone, domain-specific context, no |
2026-04-13 |
| COMMIT |
0.00 |
Roll-forward with fix. Allow async slice conversion candidat |
|
Standard brief technical commit, no AI s |
2026-04-13 |
| COMMIT |
0.00 |
Move rules_ml_toolchain loading from WORKSPACE and workspace |
|
Somewhat formal but uses project-specifi |
2026-04-13 |
| COMMIT |
0.00 |
Remove unused SynchronousMemZero from PluggableDeviceStreamE |
|
Terse, technical, typical of human commi |
2026-04-13 |
| COMMIT |
0.00 |
[XLA:GPU] Use maximum clique to allocate scratch buffer. |
|
Short, technical commit, no signs of AI |
2026-04-13 |
| COMMIT |
0.00 |
Check if hlo_live_range_ in buffer assignment is valid befor |
|
Direct and minimal technical update, no |
2026-04-13 |
| COMMIT |
0.00 |
PR #39926: Improvements to the HBM OOM Error page (Error E10 |
|
Uses template, but free-text is concise |
2026-04-13 |
| COMMIT |
0.00 |
Handle unregistered dialects as part of unstable dialects |
|
Brief, technical, human-like summary. |
2026-04-13 |
| COMMIT |
0.00 |
[XLA:MSA] Minor test updates, make operand op code optional |
|
Standard terse commit, human style, no A |
2026-04-13 |
| COMMIT |
0.00 |
Fix unchecked return values of MarkSubgraphAsDelegationSkipp |
|
Terse, technical commit message typical |
2026-04-13 |
| COMMIT |
0.00 |
Replace `MultiplyAndCheckOverflow` with `CheckedInt` in full |
|
Concise technical phrasing, no AI signal |
2026-04-13 |
| COMMIT |
0.00 |
Pjrt: Pipe `total_allocation_bytes`, `indefinite_allocations |
|
Jargon-rich, specific, human-written sty |
2026-04-13 |
| COMMIT |
0.00 |
Handle `HloShardingV3` for `Manual`/`Unreduced` Subgroups |
|
Brief, domain-specific, lacks AI hallmar |
2026-04-13 |
| COMMIT |
0.00 |
Add more performance microbenchmarks for SlinkyThreadPool |
|
Short, informal, fits human commit norms |
2026-04-13 |
| COMMIT |
0.00 |
Make GetCompiledMemoryStats export additional fields. |
|
Template changelog, concise edit summary |
2026-04-13 |
| COMMIT |
0.00 |
PR #40676: [xla:gpu] Add connected-components permute under |
|
PR includes template, human-authored det |
2026-04-13 |
| COMMIT |
0.00 |
pjrt_compiler: Provide a mechanism for fallible casts to PjR |
|
Technical explanation, natural human wri |
2026-04-13 |
| COMMIT |
0.00 |
Handle size 1 axis in `IsManual`/`IsUnreduced` checks |
|
Minimalist, domain-specific, clearly hum |
2026-04-13 |
| COMMIT |
0.00 |
[stablehlo] Only call isLegalLocation in debug mode. |
|
Informal, justified change, typical huma |
2026-04-13 |
| COMMIT |
0.00 |
Handle axis of size 1 in `V3ToV2Sharding` |
|
Technical jargon and typo indicate human |
2026-04-13 |
| COMMIT |
0.00 |
PR #33240: Adding a delayMoveToHost heuristic to LHS and rel |
|
Structured with template sections, some |
2026-04-13 |
| COMMIT |
0.00 |
Only implicitly convert from type that are convertible to th |
|
Direct, human phrasing and clear enginee |
2026-04-13 |
| COMMIT |
0.00 |
Migrate indexing analysis utils to use SymbolicExpr/Symbolic |
|
Mostly technical details, domain-specifi |
2026-04-13 |
| COMMIT |
0.00 |
PR #40679: [xla:gpu] Remove wait_on_operation_queues field |
|
Technical with issue references and conc |
2026-04-13 |
| COMMIT |
0.00 |
[XLA:GPU] Use scratch buffer in one-shot RA2A. |
|
Short, human technical commit with no AI |
2026-04-13 |
| COMMIT |
0.00 |
Early sharding conversion in `HandleAllReduce` which was any |
|
Informal tone with technical context, li |
2026-04-13 |
| COMMIT |
0.00 |
Remove deprecated mlir::AffineExpr overloads from RangeEvalu |
|
Technical and concise, no AI hallmarks. |
2026-04-13 |
| COMMIT |
0.00 |
[XLA:GPU] Add scratch memory management for GPU collectives |
|
Direct and technical with no AI indicato |
2026-04-13 |
| COMMIT |
0.00 |
[NFC] Delete platform specific TSL casts libraries (part 1). |
|
Concise, domain-specific commit, no sign |
2026-04-13 |
| COMMIT |
0.00 |
Automated Code Change |
|
Generic automated change message; no AI |
2026-04-13 |
| COMMIT |
0.00 |
Automated Code Change |
|
Generic automated change message; no AI |
2026-04-13 |
| COMMIT |
0.00 |
[XLA] Provide a default unimplemented implementation for `Sy |
|
Direct, terse, domain-specific phrasing. |
2026-04-13 |
| COMMIT |
0.00 |
[XLA:GPU] Integrate TritonXLAConvertUnsupportedTypesPass int |
|
Straightforward technical changelog with |
2026-04-13 |
| COMMIT |
0.00 |
PR #40502: [ROCm] Fix Group-gemm e2e tests to meet gfx950 hi |
|
Includes domain detail, typos, and infor |
2026-04-13 |
| COMMIT |
0.00 |
[XLA:GPU] Add nozapfhahn tags to tests that time out in cove |
|
Direct, domain-specific description; no |
2026-04-13 |
| COMMIT |
0.00 |
[NFC] Delete platform specific TSL casts libraries (part 1). |
|
Terse, human-written, with technical abb |
2026-04-13 |
| COMMIT |
0.00 |
[mpmd] Remove deprecated attributes. |
|
Very brief, direct, and domain-specific. |
2026-04-13 |
| COMMIT |
0.00 |
Automated Code Change |
|
Generic automated change message; no AI |
2026-04-13 |
| COMMIT |
0.00 |
Automated Code Change |
|
Generic automated change message; no AI |
2026-04-13 |
| COMMIT |
0.00 |
Automated Code Change |
|
Minimal, formulaic commit message; lacks |
2026-04-13 |
| COMMIT |
0.00 |
Automated Code Change |
|
Minimal, formulaic commit message; lacks |
2026-04-13 |
| COMMIT |
0.00 |
Automated Code Change |
|
Minimal, formulaic commit message; lacks |
2026-04-13 |
| COMMIT |
0.00 |
Migrate SimplifyAffine to use SymbolicMap |
|
Terse, specific, domain-relevant with mi |
2026-04-13 |
| COMMIT |
0.00 |
Remove deprecated tsl::errors::Is* functions. |
|
Concise, technical, domain-specific, no |
2026-04-13 |
| COMMIT |
0.00 |
PR #40747: [XLA:GPU] Fix GetAlgorithms for cuBLASLt fp8 gemm |
|
Domain-jargon, informal, no excessive po |
2026-04-13 |
| COMMIT |
0.00 |
PR #40664: [XLA:GPU][oneAPI] Fix autotuner failures in unit |
|
Direct, technical explanation; no AI-sty |
2026-04-13 |
| COMMIT |
0.00 |
PR #40701: RedzoneAllocatorKernel: update name to avoid conf |
|
Casual, domain-specific justification; l |
2026-04-13 |
| COMMIT |
0.00 |
PR #40519: Preserve the MoveToDevice barrier when rewriting |
|
Technical, domain-typical, shows human c |
2026-04-13 |
| COMMIT |
0.00 |
Restrict target platform for rocm builds. |
|
Brief, issue-focused, domain-specific; l |
2026-04-13 |
| COMMIT |
0.00 |
PR #40749: Add missing NCCL one-sided comm symbols to nccl.s |
|
Concise, technical, domain-specific; tem |
2026-04-13 |
| COMMIT |
0.00 |
PR #40722: [XLA:GPU] Migrate Memset32Cmd into Memset32BitVal |
|
Technical, terse with clear domain jargo |
2026-04-13 |
| COMMIT |
0.00 |
PR #40515: [XLA:GPU][codegen] Reconcile types for thread / b |
|
Direct, technical, references specific t |
2026-04-13 |
| COMMIT |
0.00 |
Automated Code Change |
|
Very terse, automated marker; no AI sign |
2026-04-13 |
| COMMIT |
0.00 |
Automated Code Change |
|
Very terse, automated marker; no AI sign |
2026-04-13 |
| COMMIT |
0.00 |
Make select emitters async |
|
Short, technical commit convention; huma |
2026-04-13 |
| COMMIT |
0.00 |
Automated Code Change |
|
Very terse, automated marker; no AI sign |
2026-04-13 |
| COMMIT |
0.00 |
Automated Code Change |
|
Very terse, automated marker; no AI sign |
2026-04-13 |
| COMMIT |
0.00 |
Automated Code Change |
|
Very terse, automated marker; no AI sign |
2026-04-13 |
| COMMIT |
0.00 |
[XLA:MemorySpaceAssignment] Re-reserve allocations for color |
|
Bulleted technical explanation; brief/in |
2026-04-12 |
| COMMIT |
0.00 |
[XLA:GPU] Use a map from the parallel dim id to the Symbolic |
|
Technical terminology and concise phrasi |
2026-04-12 |
| COMMIT |
0.00 |
Reverts 7baf72cd5425d5f24c737f4379957eb0deb71288 |
|
Standard revert commit, brief and clear. |
2026-04-11 |
| COMMIT |
0.00 |
Automated Code Change |
|
'Automated Code Change' hints automation |
2026-04-11 |
| COMMIT |
0.00 |
Automated Code Change |
|
'Automated Code Change' hints automation |
2026-04-11 |
| COMMIT |
0.00 |
Automated Code Change |
|
'Automated Code Change' hints automation |
2026-04-11 |
| COMMIT |
0.00 |
Automated Code Change |
|
'Automated Code Change' hints automation |
2026-04-11 |
| COMMIT |
0.00 |
Add integer overflow checks and improve error handling in Si |
|
Technical change with domain-specific te |
2026-04-11 |
| COMMIT |
0.00 |
Automated Code Change |
|
'Automated Code Change' hints automation |
2026-04-11 |
| COMMIT |
0.00 |
Plumb the `num_warmup_batch_threads` BatchFunction op attrib |
|
Specific technical reference and informa |
2026-04-11 |
| COMMIT |
0.00 |
Swap the order of ABSL_DEPRECATED and ABSL_REFACTOR_INLINE i |
|
Concise technical language and specific |
2026-04-10 |
| COMMIT |
0.00 |
Add `tsl::io::EnsureTrailingSlash` function to `tsl/platform |
|
Brief, technical commit message with dom |
2026-04-10 |
| COMMIT |
0.00 |
Fix invalid-null-argument crash in GraphConstructor and tens |
|
Technical changelog with domain detail; |
2026-04-10 |
| COMMIT |
0.00 |
[XLA Test Fixture Migration] Migrate collective ops test to |
|
Template-style message with migration de |
2026-04-10 |
| COMMIT |
0.00 |
Add a container for CUDA 13.2 and CUDNN 9.15 |
|
Simple addition note; domain abbreviatio |
2026-04-10 |
| COMMIT |
0.00 |
Replace `MultiplyAndCheckOverflow` with `CheckedInt` in Embe |
|
Concise technical language; domain abbre |
2026-04-10 |
| COMMIT |
0.00 |
Fix host tracing routine for benchmarks. |
|
Terse, technical fix with jargon; typica |
2026-04-10 |
| COMMIT |
0.00 |
Integrate LLVM at llvm/llvm-project@815edc3ff646 |
|
Structured technical integration; domain |
2026-04-10 |
| COMMIT |
0.00 |
Add num_warmup_batch_threads as a batching op attribute. |
|
Brief feature addition, template-driven |
2026-04-10 |
| COMMIT |
0.00 |
introduce KernelCompiler interface |
|
Informal tone, domain references, incomp |
2026-04-10 |
| COMMIT |
0.00 |
Integrate StableHLO at openxla/stablehlo@3a8886de |
|
Structured technical integration, domain |
2026-04-10 |
| COMMIT |
0.00 |
[IFRT] Format SerDes test names using test parameters |
|
Concise technical summary, human domain |
2026-04-10 |
| COMMIT |
0.00 |
Remove race condition in memory_usage_monitor_test |
|
Casual technical explanation, contains d |
2026-04-10 |
| COMMIT |
0.00 |
Add `Literal::MakeUnique` as a syntactic sugar of `Literal:: |
|
Detailed, human-like explanation with do |
2026-04-10 |
| COMMIT |
0.00 |
[XLA:MSA] Skip uses that are not allowed in VMem when block |
|
Very terse commit, typical human enginee |
2026-04-10 |
| COMMIT |
0.00 |
[XLA:MSA] Skip use update if the producing instruction is al |
|
Short, technical message, no AI signals |
2026-04-10 |
| COMMIT |
0.00 |
Use `pkg @ url` for local wheel overrides again, instead of |
|
Informal phrasing, clear human decision |
2026-04-10 |
| COMMIT |
0.00 |
Add a test to verify that Literal::Make() returns an error w |
|
Brief technical summary, standard human |
2026-04-10 |
| COMMIT |
0.00 |
Add precision checking test for default dot algorithm. |
|
Contains technical details and variable |
2026-04-10 |
| COMMIT |
0.00 |
PR #40653: [ROCm] Add missing ROCm dependencies to collectiv |
|
Contains domain error logs and tool-spec |
2026-04-10 |
| COMMIT |
0.00 |
Reverts 6f0ba297cea5e0c6ed75eee398e56b2efae181d8 |
|
Standard revert message, human terse sty |
2026-04-10 |
| COMMIT |
0.00 |
[XLA:GPU] Add tile propagation for reshape (only cases for 1 |
|
Technical shorthand and terse, domain-sp |
2026-04-10 |
| COMMIT |
0.00 |
[XLA:GPU] Emit dot via the new tiling. |
|
Terse message with technical jargon; typ |
2026-04-10 |
| COMMIT |
0.00 |
update agents.md to stop insisting on TF_ prefixes |
|
Casual tone and abbreviations; clear sig |
2026-04-10 |
| COMMIT |
0.00 |
PR #40499: [ROCm] Add waves_per_eu support to Triton GEMM co |
|
Detailed, technical explanation with abb |
2026-04-10 |
| COMMIT |
0.00 |
Pass forward `GpuTopology` in `StreamExecutorGpuCompiler::Co |
|
Direct and technical; contains domain-sp |
2026-04-10 |
| COMMIT |
0.00 |
[XLA:GPU] Use LSA API to store the result in Ragged-all-to-a |
|
Terse, technical; typical engineering co |
2026-04-10 |
| COMMIT |
0.00 |
[XLA:GPU] set reduction tile sizes for tiling space |
|
Brief, technical context with a casual h |
2026-04-10 |
| COMMIT |
0.00 |
PR #36893: [ROCm] CI: Unify ROCm CI into a single workflow f |
|
Structured but stays technical; template |
2026-04-10 |
| COMMIT |
0.00 |
PR #33613: [NVIDIA GPU] Minimize collective resource for opt |
|
Somewhat formal, but technical and domai |
2026-04-10 |
| PR |
0.00 |
XLA:PJRT: Optimize 1-byte transposes with 16x16 SIMD kernels |
|
— |
2026-02-18 |
| PR |
0.00 |
Fix out-of-bounds access and stack overflow in XLA parsing. |
|
— |
2026-04-13 |
| PR |
0.00 |
Introducing option to control the number of window prefetche |
|
— |
2026-04-08 |
| PR |
0.00 |
PR #40667: [xla:gpu] Cleanup GPU thunk APIs |
|
— |
2026-04-13 |
| PR |
0.00 |
PR #40751: [xla:gpu] Delete special case for calls to comman |
|
— |
2026-04-13 |
| PR |
0.00 |
Prevent metadata id overlap between the cupti collector and |
|
— |
2026-04-13 |
| PR |
0.00 |
Roll-forward with fix. Allow async slice conversion candidat |
|
— |
2026-04-13 |
| PR |
0.00 |
This change introduces a GetFD method to the SubProcess clas |
|
— |
2026-04-13 |
| PR |
0.00 |
PR #40738: [xla:gpu] Optimize hot path for requesting GpuCli |
|
— |
2026-04-13 |
| PR |
0.00 |
Move rules_ml_toolchain loading from WORKSPACE and workspace |
|
— |
2026-04-10 |
| PR |
0.00 |
Remove unused SynchronousMemZero from PluggableDeviceStreamE |
|
— |
2026-04-13 |
| PR |
0.00 |
triton thunk emitter |
|
— |
2026-04-13 |
| PR |
0.00 |
Allow configuration of the `num_warmup_batch_threads` BatchF |
|
— |
2026-04-13 |
| PR |
0.00 |
[XLA:GPU] Use maximum clique to allocate scratch buffer. |
|
— |
2026-04-13 |
| PR |
0.00 |
make sort emitter async |
|
— |
2026-04-13 |
| PR |
0.00 |
[XLA] Make sure that executable serialization is determinist |
|
— |
2026-04-09 |
| PR |
0.00 |
Check if hlo_live_range_ in buffer assignment is valid befor |
|
— |
2026-04-10 |
| PR |
0.00 |
Disable colocate_predecessor_trees pass |
|
— |
2024-04-03 |
| PR |
0.00 |
PR #39926: Improvements to the HBM OOM Error page (Error E10 |
|
— |
2026-04-08 |
| PR |
0.00 |
Migrate to xla::HostMemoryAllocator |
|
— |
2026-04-08 |
| PR |
0.00 |
port EmitRngGetAndUpdateStateLLVMIR to produce KernelDefinit |
|
— |
2026-04-10 |
| PR |
0.00 |
[XLA:GPU] Express bitcast tiling via transpose and reshape t |
|
— |
2026-04-13 |
| PR |
0.00 |
Refactor Subgraph construction to use std::make_unique |
|
— |
2026-04-11 |
| PR |
0.00 |
Handle unregistered dialects as part of unstable dialects |
|
— |
2026-04-09 |
| PR |
0.00 |
[XLA:LAYOUT_ASSIGNMENT] Use alias analysis for layout assign |
|
— |
2026-03-21 |
| PR |
0.00 |
Use unique module IDs in PJRT dump paths. |
|
— |
2026-04-06 |
| PR |
0.00 |
[XLA:MSA] Minor test updates, make operand op code optional |
|
— |
2026-02-18 |
| PR |
0.00 |
Use constexpr `absl::string_view` instead of `char` as per g |
|
— |
2026-04-13 |
| PR |
0.00 |
Fix unchecked return values of MarkSubgraphAsDelegationSkipp |
|
— |
2026-04-11 |
| PR |
0.00 |
Refactor `SpmdPartitioner::ConvertUnreducedSubgroup` to use |
|
— |
2026-04-13 |
| PR |
0.00 |
Replace `MultiplyAndCheckOverflow` with `CheckedInt` in full |
|
— |
2026-04-08 |
| PR |
0.00 |
Wrap clang-cl.exe with a retry mechanism. |
|
— |
2026-04-11 |
| PR |
0.00 |
PR #40716: [XLA:GPU] Migrate MemcpyDeviceToDeviceCmd to Devi |
|
— |
2026-04-13 |
| PR |
0.00 |
Integrate LLVM at llvm/llvm-project@2cf353b5e856 |
|
— |
2026-04-13 |
| PR |
0.00 |
Pjrt: Pipe `total_allocation_bytes`, `indefinite_allocations |
|
— |
2026-04-08 |
| PR |
0.00 |
Handle `HloShardingV3` for `Manual`/`Unreduced` Subgroups |
|
— |
2026-04-13 |
| PR |
0.00 |
Move autotuner pass |
|
— |
2026-04-10 |
| PR |
0.00 |
Add more performance microbenchmarks for SlinkyThreadPool |
|
— |
2026-04-13 |
| PR |
0.00 |
Make GetCompiledMemoryStats export additional fields. |
|
— |
2026-04-08 |
| PR |
0.00 |
PR #40676: [xla:gpu] Add connected-components permute under |
|
— |
2026-04-13 |
| PR |
0.00 |
pjrt_compiler: Provide a mechanism for fallible casts to PjR |
|
— |
2026-04-10 |
| PR |
0.00 |
Handle size 1 axis in `IsManual`/`IsUnreduced` checks |
|
— |
2026-04-13 |
| PR |
0.00 |
Remove last uses of AddConstraints and FromTensorSize Affine |
|
— |
2026-04-09 |
| PR |
0.00 |
Migrate remaining codegen users to use IndexingMap::Symbolic |
|
— |
2026-04-09 |
| PR |
0.00 |
Migrate last constructors from indexing_map_serialization |
|
— |
2026-04-09 |
| PR |
0.00 |
Remove IndexingMap::GetAffineMap() and all cache-related log |
|
— |
2026-04-07 |
| PR |
0.00 |
[XLA:GPU] Emit bitcast with the new tiling. |
|
— |
2026-04-13 |
| PR |
0.00 |
[XLA:GPU] Updated reshape tile propagation to support `kColl |
|
— |
2026-04-10 |
| PR |
0.00 |
Integrate LLVM at llvm/llvm-project@1ed2769c681e |
|
— |
2026-04-13 |
| PR |
0.00 |
Add missing minus parsing to SymbolicExpr serializer |
|
— |
2026-04-13 |
| PR |
0.00 |
Integrate LLVM at llvm/llvm-project@1ed2769c681e |
|
— |
2026-04-13 |
| PR |
0.00 |
Remove deprecated AffineMap-based constructors and methods f |
|
— |
2026-04-09 |
| PR |
0.00 |
Remove AffineMap serialization/printing from indexing_map_se |
|
— |
2026-04-13 |
| PR |
0.00 |
[stablehlo] Only call isLegalLocation in debug mode. |
|
— |
2026-04-13 |
| PR |
0.00 |
[XLA:GPU][NFCI] cleanups after removing nested fusions |
|
— |
2026-04-10 |
| PR |
0.00 |
Handle axis of size 1 in `V3ToV2Sharding` |
|
— |
2026-04-13 |
| PR |
0.00 |
Convert unreduced and manual shardings in the format picker, |
|
— |
2026-04-09 |
| PR |
0.00 |
Migrate and remove IndexingMap::ConstraintsSatisfied(AffineE |
|
— |
2026-04-08 |
| PR |
0.00 |
Migrate and remove IndexingMap::GetConstraints |
|
— |
2026-04-08 |
| PR |
0.00 |
Migrate codegen/tiling to SymbolicMap |
|
— |
2026-03-26 |
| PR |
0.00 |
PR #33240: Adding a delayMoveToHost heuristic to LHS and rel |
|
— |
2026-04-13 |
| PR |
0.00 |
[XLA:GPU] Implement all-reduce via the new tiling. |
|
— |
2026-04-13 |
| PR |
0.00 |
Update Bazel version to 8.6.0. |
|
— |
2026-04-13 |
| PR |
0.00 |
Fix: Correctly handle replicated to partial replicated shard |
|
— |
2026-04-13 |
| PR |
0.00 |
[XLA:PJRT] Align collective memory with respect to library s |
|
— |
2026-04-13 |
| PR |
0.00 |
Remove unused dependencies on tsl/platform:types |
|
— |
2026-04-13 |
| PR |
0.00 |
Only implicitly convert from type that are convertible to th |
|
— |
2026-04-10 |
| PR |
0.00 |
Add `aggregate_working_set_profile` to ProfileOptions. |
|
— |
2026-04-13 |
| PR |
0.00 |
Migrate indexing analysis utils to use SymbolicExpr/Symbolic |
|
— |
2026-04-08 |
| PR |
0.00 |
PR #40679: [xla:gpu] Remove wait_on_operation_queues field |
|
— |
2026-04-13 |
| PR |
0.00 |
[XLA:GPU] do not create regions for reduce arguments |
|
— |
2026-04-13 |
| PR |
0.00 |
[XLA:GPU] Use scratch buffer in one-shot RA2A. |
|
— |
2026-04-12 |
| PR |
0.00 |
Add aggregation mode to cupti collector. |
|
— |
2026-03-18 |
| PR |
0.00 |
Early sharding conversion in `HandleAllReduce` which was any |
|
— |
2026-04-13 |
| PR |
0.00 |
Replicate the scalar scaling factor for 4-bit tensor-wise qu |
|
— |
2026-03-18 |
| PR |
0.00 |
Remove all remaining SymbolicExpr/AffineExp conversions |
|
— |
2026-04-13 |
| PR |
0.00 |
Use `HasPartialReplication` and Convert to v2 for `GroupShar |
|
— |
2026-04-13 |
| PR |
0.00 |
Remove deprecated mlir::AffineExpr overloads from RangeEvalu |
|
— |
2026-04-08 |
| PR |
0.00 |
Prevent metadata id overlap between the cupti collector and |
|
— |
2026-04-09 |
| PR |
0.00 |
[XLA:GPU] Add scratch memory management for GPU collectives |
|
— |
2026-04-10 |
| PR |
0.00 |
Standardize XProf hostname resolution to prioritize system h |
|
— |
2026-04-13 |
| PR |
0.00 |
[NFC] Delete platform specific TSL casts libraries (part 1). |
|
— |
2026-04-13 |
| PR |
0.00 |
Automated Code Change |
|
— |
2026-04-12 |
| PR |
0.00 |
Refactor: Use GetV2Sharding helper to avoid unnecessary HloS |
|
— |
2026-04-13 |
| PR |
0.00 |
[NFC] Delete platform specific TSL casts libraries (part 1). |
|
— |
2026-04-13 |
| PR |
0.00 |
Remove unused AffineMap/AffineExpr includes and update comme |
|
— |
2026-04-13 |
| PR |
0.00 |
Automated Code Change |
|
— |
2026-04-12 |
| PR |
0.00 |
[XLA] Provide a default unimplemented implementation for `Sy |
|
— |
2026-04-13 |
| PR |
0.00 |
[XLA:GPU] Integrate TritonXLAConvertUnsupportedTypesPass int |
|
— |
2026-04-13 |
| PR |
0.00 |
Move gemm fusion's IsSupportedTritonInstruction over to the |
|
— |
2026-03-12 |
| PR |
0.00 |
PR #40502: [ROCm] Fix Group-gemm e2e tests to meet gfx950 hi |
|
— |
2026-04-13 |
| PR |
0.00 |
[XLA:GPU] Add nozapfhahn tags to tests that time out in cove |
|
— |
2026-04-13 |
| PR |
0.00 |
[NFC] Delete platform specific TSL casts libraries (part 1). |
|
— |
2026-04-10 |
| PR |
0.00 |
[mpmd] Remove deprecated attributes. |
|
— |
2026-04-08 |
| PR |
0.00 |
Automated Code Change |
|
— |
2026-04-12 |
| PR |
0.00 |
Automated Code Change |
|
— |
2026-04-11 |
| PR |
0.00 |
Automated Code Change |
|
— |
2026-04-13 |
