Commits

Branch selector

 main

User selector

ngimel

Datepicker

All time

Commit History

Commits on Sep 16, 2022

• Stride sort (pytorch#1249)

  ngimel

  committedSep 16, 2022
  e6490aa

Commits on Sep 14, 2022

• Fix caching of simplify_loops results (pytorch#1226)

  ngimel

  committedSep 14, 2022
  dedca39

Commits on Sep 13, 2022

• Reland pytorch#1195 heuristics PR (pytorch#1204)

  ngimel

  committedSep 13, 2022
  e109c10

Commits on Sep 12, 2022

• Autotune (pytorch#1195)

  ngimel

  committedSep 12, 2022
  395bbc8

Commits on Sep 8, 2022

• Use only reduction deps to determine if reduction is inner (pytorch#1140)

  ngimel

  committedSep 8, 2022
  e1ff6b6

Commits on Aug 27, 2022

• Make test_symbolic run on cuda (pytorch#1067)

  ngimel

  committedAug 27, 2022
  c7b34d7

• Don't decompose addmm, take 2 (pytorch#1060)

  ngimel

  committedAug 27, 2022
  dcd6eb2

Commits on Aug 26, 2022

• preserve strides and alignment when creating static inputs for cudagr… (pytorch#1031)

  ngimel

  committedAug 26, 2022
  da8bb9a

Commits on Aug 25, 2022

• don't turn AliasedLayout into fixed for extern kernels (pytorch#1013)

  ngimel

  committedAug 25, 2022
  0b8e55e

Commits on Aug 17, 2022

• Fix type promotion for elementwise lowerings (pytorch#865)

  ngimel

  committedAug 17, 2022
  444e380

• Index put (pytorch#855)

  ngimel

  committedAug 17, 2022
  cda0450

Commits on Aug 16, 2022

• Promote integral inputs to int64 by default for sum (pytorch#848)

  ngimel

  committedAug 16, 2022
  76ef057

• fix arange default dtype (pytorch#833)

  ngimel

  committedAug 16, 2022
  b003624

• Index broadcast (pytorch#831)

  ngimel

  committedAug 16, 2022
  283acd4

Commits on Aug 13, 2022

• don't tile broadcasted tensors, those tiles are not needed (pytorch#807)

  ngimel

  committedAug 13, 2022
  65d8715

Commits on Aug 12, 2022

• improve splits heuristics to handle outer reductions (pytorch#786)

  ngimel

  committedAug 12, 2022
  7b596b6

Commits on Aug 9, 2022

• make precision args float16/float32/amp mutually exclusive (pytorch#747)

  ngimel

  committedAug 9, 2022
  7181a8d

Commits on Jul 31, 2022

• don't recompile if optimize_ctx is same as accuracy_ctx; option to dump kineto trace (pytorch#686)

  ngimel

  committedJul 31, 2022
  5040d49

Commits on Jul 28, 2022

• Additional indexing simplifications (pytorch#675)

  ngimel

  committedJul 28, 2022
  a3dab20

• Zero drop (pytorch#670)

  ngimel

  committedJul 28, 2022
  a43631c

Commits on Jul 26, 2022

• simplify ModularIndexing even if base is an expr (pytorch#659)

  ngimel

  committedJul 26, 2022
  b56a00f

Commits on Jul 15, 2022

• enable specifying batch size in the command line (pytorch#581)

  ngimel

  committedJul 15, 2022
  e82fce0

Commits on Jul 14, 2022

• make triton generate kernels with the same src code each time (pytorch#575)

  ngimel

  committedJul 14, 2022
  31528f9

Commits on Jul 12, 2022

• cache triton kernels (pytorch#558)

  ngimel

  committedJul 12, 2022
  62e7e1d

Commits on Jul 10, 2022

• fix bmm -> mm optimization error (pytorch#548)

  ngimel

  committedJul 10, 2022
  de2072e

Commits on Jul 5, 2022

• Reduction deps (pytorch#502)

  ngimel

  committedJul 5, 2022
  77a7808

Commits on Jun 30, 2022

• Reduction cache (pytorch#487)

  ngimel

  committedJun 30, 2022
  b22b6a5

Commits on Jun 13, 2022

• Fix for any decomposition (pytorch#384)

  ngimel

  committedJun 13, 2022
  e3fc57a

Commits on May 31, 2022

• Low precision support (pytorch#304)

  ngimel

  committedMay 31, 2022
  b78166b