Code

Merge Cutlass version 3.6#167

Code

Closed

aacostadiaz wants to merge 63 commits intocodeplaysoftware:sycl-developcodeplaysoftware/cutlass-fork:sycl-developfrom aacostadiaz:aacosta/3.6aacostadiaz/cutlass-fork:aacosta/3.6Copy head branch name to clipboard

+236,716-56,081

Commits

Commits on Aug 5, 2024

• Fix isnan namespace qualification in cutlass/functional.h (NVIDIA#1679)

  Show description for 19b4c5e
  mhoemmen

  authored
  19b4c5e

  Copy full SHA for 19b4c5e

Commits on Aug 6, 2024

• support data type w2 used in cutlass_library (NVIDIA#1517)

  gavinchen430

  authored
  e22ba59

  Copy full SHA for e22ba59

Commits on Aug 8, 2024

• 5476 cutlass 3x gemm kernels (NVIDIA#1695)

  Show description for 2049c6c
  depaulmillz

  authored
  2049c6c

  Copy full SHA for 2049c6c

• Add CLayout_64x208 (NVIDIA#1680)

  Show description for 7192f4a
  tridao

  authored
  7192f4a

  Copy full SHA for 7192f4a

Commits on Aug 12, 2024

• 3.5.1 plots and updated readme (NVIDIA#1708)

  Show description for 4e5a8f6
  depaulmillz

  authored
  4e5a8f6

  Copy full SHA for 4e5a8f6

Commits on Aug 14, 2024

• Update half.h (NVIDIA#1709)

  eqy

  authored
  fb17043

  Copy full SHA for fb17043

Commits on Aug 15, 2024

• update 3.5.1 readme/changelog

  hwu36

  committed
  8d8cfdf

  Copy full SHA for 8d8cfdf

• Merge pull request NVIDIA#1713 from NVIDIA/351_sparse_update

  Show description for 865be73
  d-k-b

  authored
  865be73

  Copy full SHA for 865be73

Commits on Aug 16, 2024

• fix uint128

  hwu36

  committed
  b0296bf

  Copy full SHA for b0296bf

• Add couple configs into generator.py for mixed input MM (NVIDIA#1350)

  Show description for 3f084f7
  

  alexsamardzic

  and

  hwu36

  authored
  3f084f7

  Copy full SHA for 3f084f7

• Merge pull request NVIDIA#1714 from NVIDIA/u128_div

  Show description for f93a691
  d-k-b

  authored
  f93a691

  Copy full SHA for f93a691

Commits on Aug 19, 2024

• Use CUDA runtime API to retrieve function pointer to driver API (NVIDIA#1700)

  Show description for 4dbf5db
  

  shunfan-shao

  and

  shunfans

  authored
  4dbf5db

  Copy full SHA for 4dbf5db

Commits on Aug 20, 2024

• minor fix for a double quote in CMakeLists.txt (NVIDIA#1727)

  Shreya-gaur

  authored
  f7b19de

  Copy full SHA for f7b19de

Commits on Aug 30, 2024

• Add support for mixed 4-bit/8-bit data types GEMM (NVIDIA#1413)

  Show description for e1976da
  

  3 peopleauthored
  e1976da

  Copy full SHA for e1976da

Commits on Sep 4, 2024

• Update barrier.h (NVIDIA#1782)

  Algy

  authored
  6c30441

  Copy full SHA for 6c30441

• Add Sm90LinCombPerColBias (NVIDIA#1774)

  Show description for 7369adc
  

  ucassjy

  and

  Jiayu Sun

  authored
  7369adc

  Copy full SHA for 7369adc

Commits on Sep 5, 2024

• Remove extraneous comma in declaration (NVIDIA#1776)

  saagarjha

  authored
  06e3377

  Copy full SHA for 06e3377

Commits on Sep 6, 2024

• set_slice3x3 -> set_slice_3x3 (NVIDIA#1784)

  lucifer1004

  authored
  82f5075

  Copy full SHA for 82f5075

• Support ComputeFn where output type differs from input type (NVIDIA#1771)

  Show description for 323c817
  tridao

  authored
  323c817

  Copy full SHA for 323c817

Commits on Sep 9, 2024

• fix assertion (NVIDIA#1790)

  seanxwzhang

  authored
  21d0534

  Copy full SHA for 21d0534

Commits on Sep 11, 2024

• Support for TMA Epilogue for Group Gemm and add pingpong ptr array & Group Gemm (NVIDIA#1795)

  Junkai-Wu

  authored
  dbdae51

  Copy full SHA for dbdae51

• Prefix a member template name with the template keyword. (NVIDIA#1796)

  Show description for 3a8c01a
  shumway

  authored
  3a8c01a

  Copy full SHA for 3a8c01a

Commits on Sep 16, 2024

• add publication: ‘EVT: Accelerating Deep Learning Training with Epilogue Visitor Tree’ (NVIDIA#1526)

  Show description for 9f68995
  

  reed-lau

  and

  hwu36

  authored
  9f68995

  Copy full SHA for 9f68995

• Fix MMA promotion interval assertions (NVIDIA#1641)

  LyricZhao

  authored
  1ebda1c

  Copy full SHA for 1ebda1c

Commits on Sep 18, 2024

• Add print_svg for mma (NVIDIA#1733)

  Show description for 2991ce1
  reed-lau

  authored
  2991ce1

  Copy full SHA for 2991ce1

Commits on Sep 19, 2024

• Adjust profiler space for SM89 (NVIDIA#1553)

  wenlei-bao

  authored
  44dae8b

  Copy full SHA for 44dae8b

Commits on Sep 25, 2024

• Add some can implement rules of hopper convolution. (NVIDIA#1835)

  Junkai-Wu

  authored
  e2b0789

  Copy full SHA for e2b0789

Commits on Oct 7, 2024

• Fix cute doc (NVIDIA#1529)

  jiweibo

  authored
  b27c49e

  Copy full SHA for b27c49e

• Fix typos in test/unit/conv/cache_testbed_output.h (NVIDIA#1652)

  Show description for 477a677
  

  alexander-zinoviev

  and

  Alexander Zinoviev

  authored
  477a677

  Copy full SHA for 477a677

• Fix typo in comment (NVIDIA#1787)

  sjfeng1999

  authored
  0837a2a

  Copy full SHA for 0837a2a

Commits on Oct 9, 2024

• CUTLASS 3.6.0 (NVIDIA#1850)

  Show description for cc3c29a
  

  3 peopleauthored
  cc3c29a

  Copy full SHA for cc3c29a

Commits on Oct 14, 2024

• Handle MNK Sm90{Row, Col}Reduction problem shapes (NVIDIA#1803)

  saagarjha

  authored
  5366879

  Copy full SHA for 5366879

Commits on Oct 17, 2024

• add is_last_tile

  hwu36

  committed
  755194a

  Copy full SHA for 755194a

Commits on Oct 18, 2024

• Improve sm90 mixed dtype kernel (NVIDIA#1883)

  sklevtsov-nvidia

  authored
  08101d9

  Copy full SHA for 08101d9

Commits on Oct 22, 2024

• Add GMMA shape m64n40k16 (NVIDIA#1864)

  tridao

  authored
  5b50a8f

  Copy full SHA for 5b50a8f

• Add all supported GMMA shapes (NVIDIA#1890)

  sklevtsov-nvidia

  authored
  d65266a

  Copy full SHA for d65266a

Commits on Oct 23, 2024

• add maximum support (NVIDIA#1833)

  Xinyu302

  authored
  f3a3bfc

  Copy full SHA for f3a3bfc

• fix typo (NVIDIA#1853)

  sijialouintel

  authored
  ea69cc2

  Copy full SHA for ea69cc2

• fix by adding public (NVIDIA#1753)

  Xinyu302

  authored
  b0c09ed

  Copy full SHA for b0c09ed

• added mapping for bf16 to torch::kBFloat16 (NVIDIA#1843)

  Show description for 83ae20c
  

  Bogumil-Sapinski-Mobica

  and

  hwu36

  authored
  83ae20c

  Copy full SHA for 83ae20c

• Fix README (NVIDIA#1658)

  Show description for e5f3caf
  

  leimao

  and

  hwu36

  authored
  e5f3caf

  Copy full SHA for e5f3caf

• Adjusting code indentation (NVIDIA#1639)

  103yiran

  authored
  03e3bff

  Copy full SHA for 03e3bff

• Include of regular_tile_iterator.h fixed for NVRTC (NVIDIA#1765)

  Show description for f02913c
  MaxAkaAltmer

  authored
  f02913c

  Copy full SHA for f02913c

• Update gemm_f16n_f16t_f32t_tensor_op_f32_sm80.cu with include "cutlass/gemm/device/gemm_universal.h" (NVIDIA#1569)

  Show description for 12626bc
  houqi

  authored
  12626bc

  Copy full SHA for 12626bc

• remove redundant hardcoded packing configs in mixed dtype gemm (NVIDIA#1894)

  Show description for be692b4
  

  IwakuraRein

  and

  Siyuan Fu

  authored
  be692b4

  Copy full SHA for be692b4

Commits on Oct 24, 2024

• fix wrong A/BLayout in MMA_Traits for binary mma and append other MMA_Traits support (NVIDIA#1856)

  Show description for a424ca6
  CalebDu

  authored
  a424ca6

  Copy full SHA for a424ca6

• Add a print for the uint{x}b_t type. (NVIDIA#1871)

  luliyucoordinate

  authored
  08a4995

  Copy full SHA for 08a4995

Commits on Oct 26, 2024

• Refactor some GroupedGEMM logic (NVIDIA#1899)

  azhurkevich

  authored
  e8a8b69

  Copy full SHA for e8a8b69

Commits on Oct 29, 2024

• feat: support kFactor 8 used in mma tensor op tile iterator (NVIDIA#1512)

  gavinchen430

  authored
  19f5159

  Copy full SHA for 19f5159

Commits on Nov 6, 2024

• Update publications (NVIDIA#1912)

  wenlei-bao

  authored
  9004ed2

  Copy full SHA for 9004ed2

• remove restriction of stride == kernel in nhwc_pooling (NVIDIA#1896)

  thorneliu

  authored
  32e3c38

  Copy full SHA for 32e3c38

• fix undefined in device code error (NVIDIA#1880)

  luliyucoordinate

  authored
  d656afb

  Copy full SHA for d656afb

Commits on Nov 8, 2024

• Fix the racing condition of mixed-input gemm when writing the registers (NVIDIA#1931)

  Show description for 8aa95db
  

  IwakuraRein

  and

  Siyuan Fu

  authored
  8aa95db

  Copy full SHA for 8aa95db

Commits on Nov 18, 2024

• Fix cutlass python library with cuda 12.6.2.post1 (NVIDIA#1942)

  Show description for b0e09d7
  

  danthe3rd

  and

  jackkosaian

  authored
  b0e09d7

  Copy full SHA for b0e09d7

Commits on Nov 28, 2024

• Merge branch 'main' into aacosta/3.6

  Show description for 389d3bd
  aacostadiaz

  committed
  389d3bd

  Copy full SHA for 389d3bd

Commits on Dec 3, 2024

• Solve issues

  aacostadiaz

  committed
  a119855

  Copy full SHA for a119855

Commits on Dec 4, 2024

• Merge branch 'sycl-develop' into aacosta/3.6

  Show description for 2333de6
  aacostadiaz

  committed
  2333de6

  Copy full SHA for 2333de6

• Add dAlpha & dBeta to epilogue callback args

  joeatodd

  committed
  9ffe59b

  Copy full SHA for 9ffe59b

• Guard CUDA stuff

  joeatodd

  committed
  4a4b9a9

  Copy full SHA for 4a4b9a9

• Fix IsLegacyEpiloguePolicy & rename VectorBeta->VectorScale

  joeatodd

  committed
  684a5a8

  Copy full SHA for 684a5a8

Commits on Dec 5, 2024

• Pass EpiloguePolicy, not Epilogue

  joeatodd

  committed
  fd197d4

  Copy full SHA for fd197d4

• Merge branch 'sycl-develop' into aacosta/3.6

  aacostadiaz

  authored
  e322028

  Copy full SHA for e322028

• Use syclcompat::dp4a

  aacostadiaz

  committed
  ce8d683

  Copy full SHA for ce8d683