Code

Use mask during load for Softmax#645

Code

Open

rahulbatra85 wants to merge 24 commits intomain_perfROCm/triton:main_perffrom softmax_use_maskROCm/triton:softmax_use_maskCopy head branch name to clipboard

+9,364

Commits

Commits on Jul 16, 2024

• Add Perf Kernels

  Show description for 2d2dbe1
  micmelesse

  committed
  2d2dbe1

  Copy full SHA for 2d2dbe1

• skip backward (#586)

  micmelesse

  committed
  17575ea

  Copy full SHA for 17575ea

• Change all block pointers to tensor pointers (#585)

  Show description for a3d784a
  

  vgokhale

  authored and

  micmelesse

  committed
  a3d784a

  Copy full SHA for a3d784a

• Add support for bshd layout (#587)

  Show description for aa6685a
  

  vgokhale

  authored and

  micmelesse

  committed
  aa6685a

  Copy full SHA for aa6685a

• Post-Merge CI (#612)

  Show description for dbe1173
  micmelesse

  committed
  dbe1173

  Copy full SHA for dbe1173

Commits on Jul 18, 2024

• Increase CI timeout (#615)

  Show description for 23ba546
  vgokhale

  authored
  23ba546

  Copy full SHA for 23ba546

Commits on Jul 19, 2024

• Couple of FA optimizations (#608)

  Show description for df4c4d3
  

  vgokhale

  and

  micmelesse

  authored
  df4c4d3

  Copy full SHA for df4c4d3

Commits on Jul 31, 2024

• streamk v0.1 (#619)

  Show description for 52a908f
  xiaohuguo2023

  authored
  52a908f

  Copy full SHA for 52a908f

Commits on Aug 6, 2024

• Add explicit multiply-reduce GEMM kernel (#621)

  Show description for 1d2e066
  brunomazzottiamd

  authored
  1d2e066

  Copy full SHA for 1d2e066

Commits on Aug 13, 2024

• Copy *tune_gemm* from triton-mlir branch to main_perf branch (#614)

  Show description for 11e4447
  brunomazzottiamd

  authored
  11e4447

  Copy full SHA for 11e4447

Commits on Aug 16, 2024

• Clean up *tune_gemm* script from main_perf branch (#629)

  Show description for 624335f
  brunomazzottiamd

  authored
  624335f

  Copy full SHA for 624335f

Commits on Aug 19, 2024

• [tune gemm v3.4] Add xcd-based pid remapping and change back to rocprofv1 (#630)

  Show description for 15cb3a8
  zhanglx13

  authored
  15cb3a8

  Copy full SHA for 15cb3a8

• add barrier to fix racing for spinning locks (#632)

  xiaohuguo2023

  authored
  177d0bd

  Copy full SHA for 177d0bd

Commits on Sep 6, 2024

• Softmax kernel

  

  Rahul Batra

  committed
  e42690d

  Copy full SHA for e42690d

• Merge pull request #634 from ROCm/main_perf-softmax

  Show description for 6d283a2
  rahulbatra85

  authored
  6d283a2

  Copy full SHA for 6d283a2

• Move utility tools from triton-mlir to main_perf branch (#635)

  Show description for 3704738
  zhanglx13

  authored
  3704738

  Copy full SHA for 3704738

• Add rmsnorm kernel

  

  Rahul Batra

  committed
  f80aed7

  Copy full SHA for f80aed7

• Merge branch 'main_perf' into main_perf-rmsnorm

  rahulbatra85

  authored
  9da4278

  Copy full SHA for 9da4278

Commits on Sep 7, 2024

• Merge pull request #633 from ROCm/main_perf-rmsnorm

  Show description for c4bd738
  rahulbatra85

  authored
  c4bd738

  Copy full SHA for c4bd738

Commits on Sep 13, 2024

• Online softmax implementation

  

  Rahul Batra

  committed
  a782caf

  Copy full SHA for a782caf

Commits on Sep 16, 2024

• Merge pull request #639 from ROCm/softmax_updates

  Show description for 96b3d37
  rahulbatra85

  authored
  96b3d37

  Copy full SHA for 96b3d37

Commits on Sep 24, 2024

• Use mask during load for Softmax

  

  Rahul Batra

  committed
  8939325

  Copy full SHA for 8939325

Commits on Sep 25, 2024

• Use tl.exp2

  

  Rahul Batra

  committed
  ba9a70c

  Copy full SHA for ba9a70c

• Use SW pipelining with loops

  

  Rahul Batra

  committed
  a542825

  Copy full SHA for a542825