Code

R2.13 rocm enhanced gfx942 cleanup#2767

Code

Open

fsx950223 wants to merge 9 commits intor2.13-rocm-enhanced-upate-llvmROCm/tensorflow-upstream:r2.13-rocm-enhanced-upate-llvmfrom r2.13-rocm-enhanced-gfx942-cleanupROCm/tensorflow-upstream:r2.13-rocm-enhanced-gfx942-cleanupCopy head branch name to clipboard

+631-112

Commits

Commits on Nov 21, 2024

• Update LLVM GPU backend for ROCm support, make rocm llvm compiler as a default option

  fsx950223

  committed
  4ef126b

  Copy full SHA for 4ef126b

• xla/gpu: Use full csv string as cache key

  fsx950223

  committed
  61f9c5b

  Copy full SHA for 61f9c5b

• xla/gpu: Use default algorithm as fallback when can't find cache algorithm

  fsx950223

  committed
  2875d46

  Copy full SHA for 2875d46

• Use full csv string

  fsx950223

  committed
  76f7976

  Copy full SHA for 76f7976

• update datalayout to avoid warnings

  fsx950223

  committed
  37b9a1a

  Copy full SHA for 37b9a1a

• add cs8 dockerfile

  fsx950223

  committed
  465dd62

  Copy full SHA for 465dd62

• Enable clang as a compiler option

  fsx950223

  committed
  88529bf

  Copy full SHA for 88529bf

• refactor build_rocm_python3

  fsx950223

  committed
  10ed244

  Copy full SHA for 10ed244

• add hlo benchmark

  fsx950223

  committed
  180c0cf

  Copy full SHA for 180c0cf