[feat] Add NUMA-aware CPU core split for vllm worker and store threads#854
Merged
Infinite666 merged 7 commits intoModelEngine-Group:developfrom Mar 24, 2026
Merged
Conversation
wangwenxin0312
force-pushed
the
dev_fuse_op
branch
from
f5324a0 to
8001c07
Compare
wangwenxin0312
requested review from
Infinite666,
harrisonyhq,
mag1c-h,
qyh111,
wuhuxiao and
ygwpz
as code owners
wangwenxin0312
force-pushed
the
dev_fuse_op
branch
from
8001c07 to
c7ef0ba
Compare
mag1c-h
reviewed
Mar 23, 2026
wangwenxin0312
force-pushed
the
dev_fuse_op
branch
3 times, most recently
from
958e621 to
64be838
Compare
Infinite666
reviewed
Mar 24, 2026
wangwenxin0312
force-pushed
the
dev_fuse_op
branch
from
64be838 to
040addf
Compare
There was a problem hiding this comment.
Pull request overview
Adds optional NUMA-aware CPU affinity logic to the vLLM UCM connector so worker and (intended) store operations can be pinned to CPU cores local to the device (CUDA / Ascend), enabled via VLLM_CPU_AFFINITY=1.
Changes:
- Add NUMA detection and core splitting logic (CUDA PCI → NUMA; Ascend visible-device mapping → NUMA fallback).
- Attempt to split cores into worker vs. store groups and pass store cores through the connector config.
- Bind the worker process/thread to the computed worker cores when enabled.
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
wangwenxin0312
force-pushed
the
dev_fuse_op
branch
from
24daf60 to
e2ce546
Compare
qyh111
reviewed
Mar 24, 2026
wangwenxin0312
force-pushed
the
dev_fuse_op
branch
4 times, most recently
from
c07b986 to
2c3a7ed
Compare
wangwenxin0312
force-pushed
the
dev_fuse_op
branch
from
e9ea313 to
caf8823
Compare
Infinite666
approved these changes
Mar 24, 2026
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
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.
Purpose
Improve CPU affinity by assigning worker and store cores based on NUMA locality for CUDA and Ascend devices.
Modifications
Test
Enable by setting the environment variable VLLM_CPU_AFFINITY.