alex-breslow-amd cbb648505a Cheaper threadfence for gfx942 in postPeer [1/N]: enable for single node allreduce (#1766) ...

Boosts single node bfloat16 allreduce performance by up to 20% for some data sizes and provides gating with the RCCL_GFX942_CHEAP_FENCE_OFF environment variable

[ROCm/rccl commit: 11fabf1de1]

2025-07-22 07:15:15 -07:00

projects/rccl

Cheaper threadfence for gfx942 in postPeer [1/N]: enable for single node allreduce (#1766)

2025-07-22 07:15:15 -07:00

S

説明

説明が提供されていません

282 MiB

言語

C++ 67.5%

C 20.6%

Python 6.6%

CMake 3.4%

Shell 0.6%

その他 1.1%