ofi: merge the default ctx and target endpoints #1200
Merged
Conversation
This file contains hidden or 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
|
Polite ping! Any thoughts on this change? FWIW, I'm seeing great performance improvement and vastly improved support for FI_MANUAL_PROGRESS with this change. |
markbrown314
approved these changes
Apr 7, 2025
bcmIntc
approved these changes
Apr 8, 2025
davidozog
force-pushed
the
pr/ofi_single_endpoint
branch
from
75808b5 to
2eab81c
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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Outdated suggestions cannot be applied.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.
SOS currently opens 2 libfabric endpoints (EPs) per process, which can lead to Rx/TX resource exhaustion in scale-up scenarios (like with 1 process per core, when there are more cores on the CPU than there are contexts on the NIC - see #1127). This PR merges those 2 endpoints, which benefits scaling (i.e., it improves the maximum PPN by 2x in typical cases), and seems to improve overall performance due to better runtime progress.
In this PR I propose we make single-EP the default behavior for SOS over OFI transport. The optimization can be disabled by setting the new env var
SHMEM_OFI_DISABLE_SINGLE_EP.This also has the benefit of enabling
FI_MANUAL_PROGRESSfor a few providers, like OPX, TCP, and sockets - when SOS is configured with--enable-ofi-manual-progressand--enable-manual-progress.