Commits

Branch selector

 main

User selector

All users

Datepicker

All time

Commit History

Commits on Mar 16, 2025

• [AMD] Use -triton-licm to replace general -licm (#6212)

  Show description for 01bb1d7
  antiagainst

  authored
  01bb1d7

  Copy full SHA for 01bb1d7

Commits on Mar 15, 2025

• [Blackwell] Support mixed precision in mxfp tutorial (#6204)

  Show description for d392203
  masahi

  authored
  d392203

  Copy full SHA for d392203

• [AMD] Enable automatic loop fusion (#5876)

  Show description for 8f5984c
  antiagainst

  authored
  8f5984c

  Copy full SHA for 8f5984c

• [NFC] Move non-experimental tensor descriptor tests to their own file (#6195)

  peterbell10

  authored
  8f1c71d

  Copy full SHA for 8f1c71d

• [BC-break] Remove experimental prefix from tensor descriptor ops (#6194)

  Show description for 6d454dd
  peterbell10

  authored
  6d454dd

  Copy full SHA for 6d454dd

• [TEST][NFC] Remove redundant helper functions (#6208)

  Jokeren

  authored
  a95a013

  Copy full SHA for a95a013

• [AMD] Fix test_matmul.py num_stage with large block (#6210)

  Show description for 37c154d
  jungpark-mlir

  authored
  37c154d

  Copy full SHA for 37c154d

• [AMD] Extend RangeAnalysis with support for tt.SplitOp (#6133)

  Show description for fa151fb
  makslevental

  authored
  fa151fb

  Copy full SHA for fa151fb

Commits on Mar 14, 2025

• Revert "[LAYOUTS] Move order to LinearEncoding implementation (#6181)" (#6209)

  Show description for 54ee837
  lezcano

  authored
  54ee837

  Copy full SHA for 54ee837

• [Blackwell] Relax the condition to pipeline tc_gen5_mma_scaled (#6197)

  Show description for 3c58ce9
  ThomasRaoux

  authored
  3c58ce9

  Copy full SHA for 3c58ce9

• Propagate disallow_acc_multi_buffer when flattening loops (#6199)

  ThomasRaoux

  authored
  c3cd412

  Copy full SHA for c3cd412

• Support creating int1 tensors in interpreter (#6058)

  Show description for 1a91751
  saagarjha

  authored
  1a91751

  Copy full SHA for 1a91751

• [LAYOUTS] Move order to LinearEncoding implementation (#6181)

  Show description for 1217ffc
  lezcano

  authored
  1217ffc

  Copy full SHA for 1217ffc

• [Blackwell] Fix alignment of tmem allocations (#6205)

  Show description for 2dc516b
  ThomasRaoux

  authored
  2dc516b

  Copy full SHA for 2dc516b

Commits on Mar 13, 2025

• [AMD] Support [b]f16 <-> OCP [b]f8 conversions on gfx950 (#6185)

  Show description for 9794310
  yiqian1

  authored
  9794310

  Copy full SHA for 9794310

• [AMD] Update the GPU product names (#6193)

  Show description for ab91a66
  yiqian1

  authored
  ab91a66

  Copy full SHA for ab91a66

• Fixed `grid` Lambda Function in `LayerNorm` and type anotation of `broadcast_impl_shape` and `get_block_shapes` (#6189)

  Show description for 795e68e
  

  p81sunshine

  and

  Jokeren

  authored
  795e68e

  Copy full SHA for 795e68e

• [NVIDIA] Propagate TMA encoding from MMA operand (#6045)

  Show description for 7f906ad
  

  peterbell10

  and

  masahi

  authored
  7f906ad

  Copy full SHA for 7f906ad

• [AMD] Allow tl.assume in the main loop body when pipelining (#6162)

  Show description for 3e80da2
  

  njriasan

  and

  Nick Riasanovsky

  authored
  3e80da2

  Copy full SHA for 3e80da2

• Disable multithreaded MLIR compilation with MLIR_DISABLE_MULTITHREADING (#6184)

  Show description for 80cbb0c
  htyu

  authored
  80cbb0c

  Copy full SHA for 80cbb0c

• [AMD] Enable LDS transpose load for fp8 types on gfx950 (#6159)

  Show description for f1a60b9
  

  3 peopleauthored
  f1a60b9

  Copy full SHA for f1a60b9

Commits on Mar 12, 2025

• [BACKEND] Replacing IRRewriterBase with OpBuilder in couple of places (#6187)

  Show description for 6bf8e08
  pawelszczerbuk

  authored
  6bf8e08

  Copy full SHA for 6bf8e08

• Update ptxas to version 12.8.93 (#6149)

  Show description for 1c5ea22
  ThomasRaoux

  authored
  1c5ea22

  Copy full SHA for 1c5ea22

• [AMD] Fix one cluster pingpong with extra ops after local load (#6183)

  Show description for fe4fcdf
  jungpark-mlir

  authored
  fe4fcdf

  Copy full SHA for fe4fcdf

• [AMD] Disable loop pipelining when there is assert or print (#6180)

  Show description for 7380904
  ThomasRaoux

  authored
  7380904

  Copy full SHA for 7380904

Commits on Mar 11, 2025

• [TritonGPU] Add `ttng.arrive_barrier` (#6174)

  Show description for 361cfe4
  Mogball

  authored
  361cfe4

  Copy full SHA for 361cfe4

• [TritonGPU] Fix extra space in local_alloc asm format (NFC) (#6173)

  Show description for d9f10eb
  Mogball

  authored
  d9f10eb

  Copy full SHA for d9f10eb

• [BACKEND] Fix clang warnings (#6163)

  chsigg

  authored
  4e5e465

  Copy full SHA for 4e5e465

• [LAYOUTS] Kill getThreadsPerWarp & ...WithUniqueData (#6170)

  Show description for 18b8d1d
  lezcano

  authored
  18b8d1d

  Copy full SHA for 18b8d1d

• [LAYOUTS] Cache LinearEncoding creation (#6169)

  Show description for b28f9ad
  lezcano

  authored
  b28f9ad

  Copy full SHA for b28f9ad

• [AMD] Error out if exceeding max threads per block (#6125)

  Show description for 8c840b2
  schung-amd

  authored
  8c840b2

  Copy full SHA for 8c840b2

Commits on Mar 10, 2025

• [OPTIMIZER] Customize LICM to hoist loads from loops (#6051)

  Show description for 58f7ef6
  

  ardaunal

  and

  peterbell10

  authored
  58f7ef6

  Copy full SHA for 58f7ef6

• [BACKEND] Bump to llvm/llvm-project@2619c2ed58 (#6148)

  Show description for e350beb
  yiqian1

  authored
  e350beb

  Copy full SHA for e350beb

• [PIPELINE] AssignLatencies for mmav5 (#6077)

  Show description for 9ce44c8
  

  pawelszczerbuk

  and

  ThomasRaoux

  authored
  9ce44c8

  Copy full SHA for 9ce44c8

• Annotations for elementary dtypes and pointers (#6152)

  Show description for cb78503
  apgoucher

  authored
  cb78503

  Copy full SHA for cb78503

Pagination

Previous

Next