nccl

Cuda ★ 0 updated 6y ago ⑂ fork

Optimized primitives for collective multi-GPU communication

No plain-English explanation yet — one is being written right now. Check back in a minute.