Use IB_MERGE_VFS argument when detecting PCI path #145
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
bureddy
approved these changes
Jan 31, 2024
|
Thanks @bureddy. I’ll fix CI in an hour or so. Question: how long does it take for plug-in updates to get included in HPC-X? |
When running in a cloud-hypervisor guest, IB VFs are exposed as a RCiEP. If the IB VFs are merged, NCCL does not correctly detect PCI topology.
thomasbarrett
force-pushed
the
nccl-merge-vfs
branch
from
e6f477e
to
2501bc4
Compare
it will be part of HPCX-2.18 (ETA: mid of Feb) |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
When running in a cloud-hypervisor virtual-machine, IB VFs are exposed as a RCiEP. If the IB VFs are merged, the nccl-rdma-sharp-plugin incorrectly assumes that all IB virtual-functions are part of the PCI host bridge. Disabling the IB_MERGE_VFS variable allows NCCL to correctly identify IB VFs as independent devices.
This is the same logic used by NCCL native IB transport, it should be used here as well.