Issues

• Warp-specialized persistent fp8 matmul fails to compile on sm_100 (two distinct crashes: 2-CTA AccelerateMatmul assertion + auto-WS PartitionLoops failure)

  meta-pytorch/tritonbench#1164 · Filter by author yushangdi opened on Jul 2, 2026
  1

  

• Add comm and distributed kernels

  meta-pytorch/tritonbench#636 · Filter by author xuzhao9 opened on Nov 11, 2025

• [AMD][AITER] Add more AMD kernels

  meta-pytorch/tritonbench#604 · Filter by author xuzhao9 opened on Oct 30, 2025

• Increase operator coverage

  help wanted
  meta-pytorch/tritonbench#479 · Filter by author xuzhao9 opened on Sep 26, 2025
  1

• profiler + cudagraph hangs forever for rope

  meta-pytorch/tritonbench#413 · Filter by author BoyuanFeng opened on Sep 10, 2025
  8

• [BE] Enable CUDAGraph by default

  behelp wanted
  meta-pytorch/tritonbench#349 · Filter by author xuzhao9 opened on Aug 21, 2025

• Request for branches in repo

  meta-pytorch/tritonbench#306 · Filter by author bdbowyer opened on Jul 28, 2025
  3

• [Help wanted][aten] Fix the broken PT2 ATen operators

  help wantedpt2
  meta-pytorch/tritonbench#270 · Filter by author xuzhao9 opened on Jul 9, 2025
  1