-
-
Notifications
You must be signed in to change notification settings - Fork 8.7k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Support bitwise allreduce in NCCL communicator #9300
Conversation
Apologies for the oversight, the no_sync is a CTK-12 feature. |
Switched to use |
nccl_unique_id_ = GetUniqueId(); | ||
dh::safe_cuda(cudaSetDevice(device_ordinal_)); | ||
dh::safe_nccl(ncclCommInitRank(&nccl_comm_, world, nccl_unique_id_, rank)); | ||
dh::safe_cuda(cudaStreamCreate(&cuda_stream_)); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Just a side note, there's dh::CUDAStream
, which is a RAII version of async stream.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Switched to it.
Use
AllGather
to collect all the data first, then do the bitwise allreduce locally.