Skip to content
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.

Sign up for GitHub

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

Jump to bottom

UCX plugin: Add multi-receive posting to avoid excessive flush #125

Merged
merged 11 commits into from
Oct 18, 2023
Merged

UCX plugin: Add multi-receive posting to avoid excessive flush #125

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
11 commits
Select commit Hold shift + click to select a range
6fcabab
Advertize multiple operations per call for UCX
tvegas1 Oct 3, 2023
1f237ef
Multiple receive for UCX plugin
tvegas1 Oct 3, 2023
abf2795
Only use one comm
tvegas1 Oct 4, 2023
3259585
Code cleanup
tvegas1 Oct 5, 2023
b5c3e3a
Code cleanup (2)
tvegas1 Oct 5, 2023
bb76a10
Remove UCX <1.10 support
tvegas1 Oct 5, 2023
9cd0407
Rename ucx_nccl_request to ucx_request
tvegas1 Oct 5, 2023
5650fbd
Fix codestyle, and cleanup code
tvegas1 Oct 5, 2023
90a623b
Address review: remove datatype parameter
tvegas1 Oct 11, 2023
d6ff08f
Address review: add freelist and various code fixes
tvegas1 Oct 11, 2023
264e50b
Add missing ucp_request_free
Oct 17, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 6 additions & 2 deletions src/p2p_plugin.c
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,5 +1,5 @@
/*************************************************************************
* Copyright (c) 2016-2020, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
* Copyright (c) 2016-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
*
* See LICENSE.txt for license information
************************************************************************/
Expand Down Expand Up @@ -186,7 +186,11 @@ ncclResult_t nccl_p2p_ib_get_properties(nccl_ib_dev_t *devs, int dev, ncclNetPro
props->latency = 0; // Not set
props->port = devs[dev].port + devs[dev].realPort;
props->maxComms = devs[dev].maxQp;
props->maxRecvs = (p2p_plugin == NCCL_P2P_IB) ? NCCL_NET_IB_MAX_RECVS : 1;
if (p2p_plugin == NCCL_P2P_IB || p2p_plugin == NCCL_P2P_UCX) {
props->maxRecvs = NCCL_NET_IB_MAX_RECVS;
} else {
props->maxRecvs = 1;
}
props->netDeviceType = NCCL_NET_DEVICE_HOST;
props->netDeviceVersion = NCCL_NET_DEVICE_INVALID_VERSION;

Expand Down
Loading