23
I Use This!
Very High Activity

Commits : Listings

Analyzed about 21 hours ago. based on code collected 2 days ago.
Mar 26, 2023 — Mar 26, 2024
Commit Message Contributor Files Modified Lines Added Lines Removed Code Location Date
Integrate LLVM at llvm/llvm-project@3cf169ca160e More... 2 days ago
Migrate Protobuf DebugString calls More... 2 days ago
[XLA:GPU] Add GpuHloCostAnalysis option to estimate Triton Softmax fusions. More... 2 days ago
[XLA:GPU][NFC] Move tile_analysis* to symbolic_tile*. More... 2 days ago
[XLA:GPU] Implement EmitDot in the new MLIR emitters More... 2 days ago
[XLA:GPU] Add threadId mappings to the legacy emitters and add tests to the new ones. More... 2 days ago
Integrate LLVM at llvm/llvm-project@fa3d789df15b More... 2 days ago
[XLA:GPU] Remove FusionMergerTriton pass. More... 2 days ago
Do not produce negative strides when building symbolic maps More... 2 days ago
[XLA:GPU] Implement new Triton Softmax codegen based on symbolic tiles and indexing maps. More... 2 days ago
Automated Code Change More... 2 days ago
[XLA:GPU] Fix layout for kSend/kRecv on GPU to always be default More... 2 days ago
[XLA:GPU] Expand test coverage for SymbolicTile derivation. More... 2 days ago
Automated Code Change More... 2 days ago
compat: Update forward compatibility horizon to 2024-03-26 More... 2 days ago
Update GraphDef version to 1813. More... 2 days ago
[XLA:GPU] Add option to return FDO profile as textproto. More... 2 days ago
No public description More... 2 days ago
Add deferring transposes of the input of `stablehlo.maximum` to output. More... 3 days ago
PR #10892: Fix some clang-tidy reported issues in xla/service More... 3 days ago
Remove CHECKs from op_profile_builder.cc More... 3 days ago
Add two arguments in OpExpanderPass: `preserve_sharding` and `relay_control_dependency`. More... 3 days ago
Use the information in allow_spmd_sharding_propagation_to_output and allow_spmd_sharding_propagation_to_parameters to determine what input and output tuple elements we are allowed to modfy the shardings of. More... 3 days ago
Change tf.IfrtLoadVariableOp to receive tensor future from IfrtRestoreTensorRegistry instead of from resource manager. Correspondingly, IfrtLoadVariableRegistry is changed to store a future of array. This enables tfrt+ifrt to restore tensor asynchronously More... 3 days ago
Create a util function `BuildGetTupleElementsForTupleResults` with proper processing on the shardings. More... 3 days ago
PR #10835: [GPU] Add support of select operation in cuDNN fusions. More... 3 days ago
Ensure that the module we consume has no unused computations. This can causes issues as we clone modules to support try_multiple_mesh_shapes, and cloning an HLO module removes dead computations leading to mismatches. More... 3 days ago
PR #10925: Include FP8 GEMMs in IsCublasGemm More... 3 days ago
[xla][gpu] Change related to the handling of pipelined Send/Recv. More... 3 days ago
Log checkpoint hashing error. More... 3 days ago