Commits

Branch selector

 main

User selector

All users

Datepicker

All time

Commit History

Commits on Oct 23, 2025

• [AMD] Fix branch condition in BufferLoadToLocalOpConversion (#8501)

  Show description for bad2576
  kelesvol

  authored
  bad2576

  Copy full SHA for bad2576

Commits on Oct 22, 2025

• [GLUON] add device-side TMA (#8505)

  Show description for 56c6468
  hgl71964

  authored
  56c6468

  Copy full SHA for 56c6468

• [AxisInfo] Make unrealized_conversion_cast handling more robust (#8507)

  Show description for fe57b25
  matthias-springer

  authored
  fe57b25

  Copy full SHA for fe57b25

Commits on Oct 21, 2025

• [BACKEND] Generalise maybeDeduplicate to all layouts (#8492)

  Show description for c172d53
  lezcano

  authored
  c172d53

  Copy full SHA for c172d53

Commits on Oct 20, 2025

• [AMD] Expose buffer ops small tensor analysis to Python (#8471)

  Show description for 8f5aa60
  njriasan

  authored
  8f5aa60

  Copy full SHA for 8f5aa60

• [NVIDIA] Add native FP4 scaled_dot for SM120 (#8494)

  Show description for ff63ee2
  ita9naiwa

  authored
  ff63ee2

  Copy full SHA for ff63ee2

• [GLUON] Improve layout check message (#8456)

  Show description for b3e233e
  hgl71964

  authored
  b3e233e

  Copy full SHA for b3e233e

• [NVIDIA] Rewrite getSM120DotScaledScaleLayout and Refactor MMAv2 (#8482)

  Show description for 814b862
  ita9naiwa

  authored
  814b862

  Copy full SHA for 814b862

Commits on Oct 19, 2025

• [BACKEND] Implement BF16x3 trick (#7592)

  Show description for 33e7dc2
  plotfi

  authored
  33e7dc2

  Copy full SHA for 33e7dc2

• [AMD] Support Skinny Blocks for TDM on gfx1250 (#8479)

  Show description for 3251bb8
  knwng

  authored
  3251bb8

  Copy full SHA for 3251bb8

Commits on Oct 18, 2025

• [AMD] Explicitly set ConvertToBufferOps small tensor option (#8470)

  Show description for 90666a8
  yangshuxin

  authored
  90666a8

  Copy full SHA for 90666a8

• [mxfp] fix bf16 x mxfp4 bug with SUBTILE_FACTOR > 1 (#8478)

  Show description for 2172bb4
  jongsoo-openai

  authored
  2172bb4

  Copy full SHA for 2172bb4

• [BACKEND] Prevent combinatory explosion when checking tmem_load uses (#8477)

  ThomasRaoux

  authored
  a4ab31d

  Copy full SHA for a4ab31d

• [PTXAS] Upgrade ptxas to 12.9.86 for blackwell (#8476)

  Show description for 3b3c852
  ThomasRaoux

  authored
  3b3c852

  Copy full SHA for 3b3c852

Commits on Oct 17, 2025

• [KERNELS] Support `y_indx` and uniform distribution (#8472)

  Jokeren

  authored
  273649e

  Copy full SHA for 273649e

• [TRITON_KERNELS] Apply MXFP4 Hopper layout on A100 (#8474)

  Show description for 28533b1
  hyouklee0

  authored
  28533b1

  Copy full SHA for 28533b1

• [WS] Recognize warp-specialized nested loops in `AssignLatencies` (#8451)

  Show description for 3200f5a
  masahi

  authored
  3200f5a

  Copy full SHA for 3200f5a

• Revert "[LAYOUTS] Generate distributed layouts for `tcgen05.ld/st` generically (#8421)" (#8469)

  Show description for f5d04d5
  lezcano

  authored
  f5d04d5

  Copy full SHA for f5d04d5

• [AMD] Add gfx1250 support for ds_read_tr (#8461)

  Show description for 34676a2
  nzaghen

  authored
  34676a2

  Copy full SHA for 34676a2

• [gluon] Basic multi-cta support (#8468)

  Show description for 3c8c0cf
  peterbell10

  authored
  3c8c0cf

  Copy full SHA for 3c8c0cf

• [AMD] fix ptr-canonicalization pass segmentation fault (#8465)

  Show description for 24bd281
  yangshuxin

  authored
  24bd281

  Copy full SHA for 24bd281

Commits on Oct 16, 2025

• [FRONTEND] Improve errors for TMA desc failure (#8462)

  Show description for e5e0081
  ThomasRaoux

  authored
  e5e0081

  Copy full SHA for e5e0081

• Do not use C++20 designed initializers in `Transforms/ConcurrencySanitizer.cpp` (#8455)

  Show description for 5745035
  anmyachev

  authored
  5745035

  Copy full SHA for 5745035

• [TRITON_KERNELS] basic expert parallelism implementation (#8448)

  Show description for 35bb255
  

  ptillet

  and

  Jokeren

  authored
  35bb255

  Copy full SHA for 35bb255

• [AMD] Lift unneeded ds_read_tr lowering restrictions (#8442)

  Show description for 2875f40
  nzaghen

  authored
  2875f40

  Copy full SHA for 2875f40

Commits on Oct 15, 2025

• [Triton Plugin] Reorder the configuration sequence of Triton, third-party libraries, and Triton plugins (#8397)

  Show description for e9f8ec9
  

  lechenyu

  and

  Jokeren

  authored
  e9f8ec9

  Copy full SHA for e9f8ec9

• [WS] Swap ordering of `OptimizePartitionWarps` and SWP (#8415)

  Show description for d13acf9
  masahi

  authored
  d13acf9

  Copy full SHA for d13acf9

• [NFC] Kill unused functions (#8446)

  lezcano

  authored
  f634d49

  Copy full SHA for f634d49

• [AMD] Use v_dot for bf16 multiplication on gfx11/gfx12 (#8444)

  Show description for 2e20593
  

  ptrojahn

  and

  paultrojahnamd

  authored
  2e20593

  Copy full SHA for 2e20593

• [LAYOUTS] Generate distributed layouts for `tcgen05.ld/st` generically (#8421)

  Show description for ea01a7e
  lezcano

  authored
  ea01a7e

  Copy full SHA for ea01a7e

• [TESTING] Add tb=short to CI (#8440)

  Show description for b9ffc97
  lezcano

  authored
  b9ffc97

  Copy full SHA for b9ffc97

Commits on Oct 14, 2025

• [AMD] Add f16 gemm gluon example for gfx1250 (#8443)

  Show description for 6fce184
  

  PMylon

  and

  borontion

  authored
  6fce184

  Copy full SHA for 6fce184

• [DOC] Clarify lhs_scale and rhs_scale requirements in dot_scaled (#8433)

  Show description for e45315f
  ita9naiwa

  authored
  e45315f

  Copy full SHA for e45315f

• Do not use C++20 designed initializers in `TritonNVIDIAGPUToLLVM/DotOpToLLVM/MMAHelpers.h` (#8441)

  Show description for 626ff18
  anmyachev

  authored
  626ff18

  Copy full SHA for 626ff18

Commits on Oct 13, 2025

• [LAYOUTS] Fix backwards propagation for Fp4ToFp (#8438)

  Show description for 3516eac
  lezcano

  authored
  3516eac

  Copy full SHA for 3516eac

Pagination

Previous

Next