Skip to content

fix: add _tunable_defines() to _hipcc_device_bc for Triton/FlyDSL - #17

Merged
itej89 merged 1 commit into
feat/ep-rdma-sharingfrom
feat/ep-rdma-sharing-fix-jit-defines
Aug 27, 2026
Merged

fix: add _tunable_defines() to _hipcc_device_bc for Triton/FlyDSL#17
itej89 merged 1 commit into
feat/ep-rdma-sharingfrom
feat/ep-rdma-sharing-fix-jit-defines

Conversation

@itej89

@itej89 itej89 commented Aug 27, 2026

Copy link
Copy Markdown
Owner

Summary

Addresses PR ROCm#558 review comment #4.

_hipcc_device_bc omitted _tunable_defines(), so device bitcode compiled by Triton/FlyDSL was built without -DMORI_PROXY_ENABLED. This compiled out the proxy transport dispatch branches in shmem kernels (_PROXY_ELSE macro, ShmemQuietThread proxy drain), causing Triton/FlyDSL EP ops to silently skip the proxy path.

One-line fix: add *_tunable_defines(), to the _hipcc_device_bc command list, matching _hipcc_genco which already has it.

Test plan

  • EP benchmark on DO AINIC: Dispatch Pass, Combine Pass
  • Dispatch avg 32.16 GB/s, Combine avg 54.08 GB/s (consistent)

🤖 Generated with Claude Code

_hipcc_device_bc omitted _tunable_defines(), so device bitcode was
built without -DMORI_PROXY_ENABLED. This compiled out the proxy
transport dispatch branches in shmem kernels, causing Triton/FlyDSL
EP ops to silently skip the proxy path.

Co-Authored-By: Claude <noreply@anthropic.com>
@itej89
itej89 merged commit 1efe5e9 into feat/ep-rdma-sharing Aug 27, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant