Commits

Branch selector

 master

User selector

All users

All time

Commit History

Commits on Apr 20, 2026

• [XLA:GPU] Emit bitcast with the same logic as tiling, i.e., transpose, reshape, trt.

  Show description for 7fd8f68
  tensorflower-gardener

  committed
  7fd8f68

  Copy full SHA for 7fd8f68

• [XLA] Treat computations with a different number of parameters as different.

  Show description for 37072e6
  

  dimitar-asenov

  authored and

  tensorflower-gardener

  committed
  37072e6

  Copy full SHA for 37072e6

• [XLA:CPU] Add FAST_COMPILE tests to CI using test_cpu_fast_compile argument.

  Show description for b383e6b
  

  seantalts

  authored and

  tensorflower-gardener

  committed
  b383e6b

  Copy full SHA for b383e6b

• Remove AffineMap serialization/printing from indexing_map_serialization

  Show description for 95a1412
  tensorflower-gardener

  committed
  95a1412

  Copy full SHA for 95a1412

• Move constants to the right when canonicalize a chain of products

  Show description for 43eaa0c
  tensorflower-gardener

  committed
  43eaa0c

  Copy full SHA for 43eaa0c

• Add missing minus parsing to SymbolicExpr serializer

  Show description for 7e67e9c
  tensorflower-gardener

  committed
  7e67e9c

  Copy full SHA for 7e67e9c

• Delete a comment that was mis-introduced from a previous change.

  Show description for ce992a6
  

  h-joo

  authored and

  tensorflower-gardener

  committed
  ce992a6

  Copy full SHA for ce992a6

• PR #40702: [ROCm] Label FP8 fast accumulation as supported on ROCm 7

  Show description for c0b10ef
  

  amd-jianli12

  authored and

  tensorflower-gardener

  committed
  c0b10ef

  Copy full SHA for c0b10ef

• [xla_compile] Various fixes to xla_compile:

  Show description for b824d9e
  

  khasanovaa

  authored and

  tensorflower-gardener

  committed
  b824d9e

  Copy full SHA for b824d9e

• PR #38777: [ROCm] Add error checking and status propagation to ROCm profiler

  Show description for 1760ed1
  

  magaonka-amd

  authored and

  tensorflower-gardener

  committed
  1760ed1

  Copy full SHA for 1760ed1

• Remove unused AffineMap/AffineExpr includes and update comments.

  Show description for 81f4d3a
  tensorflower-gardener

  committed
  81f4d3a

  Copy full SHA for 81f4d3a

• Remove all remaining SymbolicExpr/AffineExp conversions

  Show description for ad71023
  tensorflower-gardener

  committed
  ad71023

  Copy full SHA for ad71023

• [XLA:GPU]: Add error context also on allocation errors.

  Show description for 679ada2
  

  sohaibiftikhar

  authored and

  tensorflower-gardener

  committed
  679ada2

  Copy full SHA for 679ada2

• Update autotune cache key for triton integration 1.25.

  Show description for acd3506
  

  vwbaker

  authored and

  tensorflower-gardener

  committed
  acd3506

  Copy full SHA for acd3506

• PR #41079: [XLA:GPU] Migrate MemzeroCmd into MemzeroThunk

  Show description for 2d8736e
  

  shawnwang18

  authored and

  tensorflower-gardener

  committed
  2d8736e

  Copy full SHA for 2d8736e

• PR #40852: [XLA:GPU] Fix missing P2P peer access in CudaMemoryReservation::SetAccess

  Show description for 06178bb
  

  shawnwang18

  authored and

  tensorflower-gardener

  committed
  06178bb

  Copy full SHA for 06178bb

• Remove TfrtGpuClient support from se_gpu_pjrt_compiler.

  Show description for 36d1ba3
  

  beckerhe

  authored and

  tensorflower-gardener

  committed
  36d1ba3

  Copy full SHA for 36d1ba3

• PR #39201: [xla:gpu] Split legacy custom calls into LegacyCustomCallThunk

  Show description for 3ab77b9
  

  ezhulenev

  authored and

  tensorflower-gardener

  committed
  3ab77b9

  Copy full SHA for 3ab77b9

• PR #41134: [tsl] Add Untie() to tsl::TiedRef

  Show description for 2f504cc
  

  ezhulenev

  authored and

  tensorflower-gardener

  committed
  2f504cc

  Copy full SHA for 2f504cc

• Automated Code Change

  Show description for f44becd
  tensorflower-gardener

  committed
  f44becd

  Copy full SHA for f44becd

• Reverts 46b5a5ca079c80cb23577ea8c3831062b08ab962

  Show description for e0d6a6c
  

  beckerhe

  authored and

  tensorflower-gardener

  committed
  e0d6a6c

  Copy full SHA for e0d6a6c

• Automated Code Change

  Show description for c56b864
  tensorflower-gardener

  committed
  c56b864

  Copy full SHA for c56b864

• Reverts 6f3131d781264c1c7aafddeedd98a7f49e81d085

  Show description for 665458a
  tensorflower-gardener

  committed
  665458a

  Copy full SHA for 665458a

Commits on Apr 19, 2026

• Remove a typo

  Show description for 0cee052
  

  amitsabne1

  authored and

  tensorflower-gardener

  committed
  0cee052

  Copy full SHA for 0cee052

Commits on Apr 18, 2026

• Refactor `HloModule::GetFingerprint128` to avoid large string allocations.

  Show description for 2cd48d2
  

  allanrenucci

  authored and

  tensorflower-gardener

  committed
  2cd48d2

  Copy full SHA for 2cd48d2

• Reverts c926ad1c96d55c6d2b967db6df64893729232f87

  Show description for 6ba21a5
  tensorflower-gardener

  committed
  6ba21a5

  Copy full SHA for 6ba21a5

• Don't delegate `qint8` weights with non-zero `zero_point`.

  Show description for 5a6a435
  

  gonnet

  authored and

  tensorflower-gardener

  committed
  5a6a435

  Copy full SHA for 5a6a435

• [XLA:CPU] Cap simplification pipeline fixed-point iterations for FAST_COMPILE.

  Show description for 4235de8
  

  seantalts

  authored and

  tensorflower-gardener

  committed
  4235de8

  Copy full SHA for 4235de8

• Automated Code Change

  Show description for 04dff64
  tensorflower-gardener

  committed
  04dff64

  Copy full SHA for 04dff64

• Automated Code Change

  Show description for 57f6700
  tensorflower-gardener

  committed
  57f6700

  Copy full SHA for 57f6700

• Automated Code Change

  Show description for 1255efa
  tensorflower-gardener

  committed
  1255efa

  Copy full SHA for 1255efa

• Automated Code Change

  Show description for edbe95d
  tensorflower-gardener

  committed
  edbe95d

  Copy full SHA for edbe95d

• [XLA] Stop moving operations into conditionals with empty tuple operands.

  Show description for c926ad1
  

  blakehechtman

  authored and

  tensorflower-gardener

  committed
  c926ad1

  Copy full SHA for c926ad1

Commits on Apr 17, 2026

• PR #41127: [xla:gpu] Extract nccl_types library

  Show description for e9d1938
  

  ezhulenev

  authored and

  tensorflower-gardener

  committed
  e9d1938

  Copy full SHA for e9d1938

• [IFRT IR] Add a method that the IFRT IR compiler that only creates a compiled program, but not an executable.

  Show description for c809154
  

  ICGog

  authored and

  tensorflower-gardener

  committed
  c809154

  Copy full SHA for c809154