Code

[tuning] gemm tuning script v3.3#606

Code

Merged

xiaohuguo2023 merged 12 commits intotriton-mlirROCm/triton:triton-mlirfrom pipeline_tune_gemmROCm/triton:pipeline_tune_gemmCopy head branch name to clipboardJul 23, 2024

+1,113-476

Commits

Commits on Jul 20, 2024

• gemm tuning script v3.3

  zhanglx13

  committed
  c28b2e8

  Copy full SHA for c28b2e8

• Add support for EVEN_K

  zhanglx13

  committed
  061900f

  Copy full SHA for 061900f

• Output amdgcn

  zhanglx13

  committed
  b3b7d34

  Copy full SHA for b3b7d34

Commits on Jul 21, 2024

• Extract matmul_kernel_configStr kernels into a separate file

  zhanglx13

  committed
  c387760

  Copy full SHA for c387760

• Format python code with yapf

  Show description for 344e6f2
  zhanglx13

  committed
  344e6f2

  Copy full SHA for 344e6f2

• Construct the tuning space, i.e. myKernels accumulatively

  zhanglx13

  committed
  cd736ce

  Copy full SHA for cd736ce

Commits on Jul 22, 2024

• Better readme

  zhanglx13

  committed
  6ee25cc

  Copy full SHA for 6ee25cc

• typeo fix and address review comments

  zhanglx13

  committed
  85f7d93

  Copy full SHA for 85f7d93

• Fix the issue with bias_size and bias.stride

  zhanglx13

  committed
  b9038cf

  Copy full SHA for b9038cf

• Don't allocate tensor in compilation stage

  zhanglx13

  committed
  7319adc

  Copy full SHA for 7319adc

• Add an option to hack the triton compiler to avoid backend query

  zhanglx13

  committed
  a61967b

  Copy full SHA for a61967b

Commits on Jul 23, 2024

• Better README again

  zhanglx13

  committed
  e979474

  Copy full SHA for e979474