search

NVIDIA / nccl

Optimized primitives for collective multi-GPU communication
Other
2.97k stars 758 forks source link

Local user buffer registration for NVLink SHARP #1312

Open zhang662817 opened 1 month ago

zhang662817 commented 1 month ago

Hi, can you introduce Local user buffer registration algorithm for NVLink SHARP in detail?

Thanks.

zhang662817 commented 3 weeks ago

@sjeaugey