Code

Rebase xsmm to main#177

Code

Draft

Devjiu wants to merge 17 commits intotriton-lang:maintriton-lang/triton-cpu:mainfrom Devjiu:rebase_xsmm_to_mainDevjiu/triton-cpu:rebase_xsmm_to_mainCopy head branch name to clipboard

+4,982-2

Commits

Commits on Jan 20, 2025

• Matrix multiplication tutorial block pointer variant (triton-lang#1)

  Show description for 7d75054
  

  rolfmorel

  authored and

  Devjiu

  committed
  7d75054

  Copy full SHA for 7d75054

• Vector to XSMM (triton-lang#3)

  Show description for 1a3d5a9
  

  adam-smnk

  authored and

  Devjiu

  committed
  1a3d5a9

  Copy full SHA for 1a3d5a9

• [XSMM] Place contraction accumulation buffer outside reduction loop (triton-lang#5)

  Show description for 8f6be0d
  

  adam-smnk

  authored and

  Devjiu

  committed
  8f6be0d

  Copy full SHA for 8f6be0d

• Triton to XSMM (triton-lang#4)

  Show description for 552f38d
  

  adam-smnk

  authored and

  Devjiu

  committed
  552f38d

  Copy full SHA for 552f38d

• Lift -triton-raise-block-pointer pass from intel-xpu-backend-for-triton (triton-lang#7)

  Show description for 01436d2
  

  rolfmorel

  authored and

  Devjiu

  committed
  01436d2

  Copy full SHA for 01436d2

• [triton][XSMM] Hoist accumulation buffer (triton-lang#8)

  Show description for 737d210
  

  adam-smnk

  authored and

  Devjiu

  committed
  737d210

  Copy full SHA for 737d210

• Bump libxsmm (triton-lang#9)

  Show description for 64650be
  

  adam-smnk

  authored and

  Devjiu

  committed
  64650be

  Copy full SHA for 64650be

• Enable bfloat16 support (triton-lang#10)

  Show description for e658655
  

  adam-smnk

  authored and

  Devjiu

  committed
  e658655

  Copy full SHA for e658655

• Dynamic shape/stride/offset support by way of memref.extract_strided_metadata (triton-lang#11)

  

  rolfmorel

  authored and

  Devjiu

  committed
  b3554c6

  Copy full SHA for b3554c6

• Matmul tutorial - K padding and dynamic K block size (triton-lang#12)

  Show description for 4070364
  

  adam-smnk

  authored and

  Devjiu

  committed
  4070364

  Copy full SHA for 4070364

• Matmul tutorial - cache padding (triton-lang#14)

  Show description for 4283581
  

  adam-smnk

  authored and

  Devjiu

  committed
  4283581

  Copy full SHA for 4283581

• Lower GEMM to BRGEMM kernel (triton-lang#13)

  Show description for 2930ef7
  

  adam-smnk

  authored and

  Devjiu

  committed
  2930ef7

  Copy full SHA for 2930ef7

• Matmul tutorial - external preprocessing (triton-lang#15)

  Show description for ad183bb
  

  adam-smnk

  authored and

  Devjiu

  committed
  ad183bb

  Copy full SHA for ad183bb

• Utility libxsmm Python extension (triton-lang#17)

  Show description for b0ae885
  

  adam-smnk

  authored and

  Devjiu

  committed
  b0ae885

  Copy full SHA for b0ae885

• Reduction loop GEMM to XSMM BRGEMM (triton-lang#18)

  Show description for 0a2717f
  

  adam-smnk

  authored and

  Devjiu

  committed
  0a2717f

  Copy full SHA for 0a2717f

• Matmul tutorial - pad weights only (triton-lang#19)

  Show description for 946f255
  

  adam-smnk

  authored and

  Devjiu

  committed
  946f255

  Copy full SHA for 946f255

• remove changes in tutorial

  Show description for c60aad6
  Devjiu

  committed
  c60aad6

  Copy full SHA for c60aad6