Repository navigation
perf: avoid redundant copies when batching a single multimodal tensor. - #2745
Open
liutongxuan wants to merge 1 commit into
Open
liutongxuan wants to merge 1 commit into
liutongxuan wants to merge 1 commit into
Conversation
Multimodal batching concatenates even a single tensor, copying the full visual payload before transport and when rebuilding the received batch. Retain a contiguous singleton through tensor ownership while preserving concatenation for multiple inputs and tensor lists for unsupported layouts. Keep serialization interfaces and wire schemas intact. Cover source lifetime, protobuf round trips, noncontiguous inputs, multiple images, and packed and shared-memory visual transport.
liutongxuan
requested review from
DongheJin,
DragonFive,
JimHsiung,
Kang-Meng,
XuZhang99,
xiao-yu-chen,
yingxudeng and
zhang-minchao
as code owners
November 15, 2026 09:49
Codex Review Summary: ✅ CompletedThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
This branch has not been deployed
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Description
Reuse the storage of a single contiguous tensor during multimodal batching instead of concatenating it into a new allocation. Preserve existing behavior for multiple tensors and noncontiguous inputs, along with serialization interfaces and wire formats for shared-memory and multi-node transport.
For a 56 MB visual input, an isolated batching benchmark reduced median CPU aggregation time from approximately 11.14 ms to 0.012 ms. End-to-end measurements have not yet established a consistent latency improvement due to run-to-run variation.
The full NPU build and six targeted tests passed, covering batching correctness, tensor lifetime, and packed/shared-memory transport. Full regression validation is still in progress.
Related Issues
Change Type
Pull Request Checklist
Thank you for contributing to xLLM. Before requesting review, please make sure the following items are complete.
PR Title and Commit Messages
<type>: <subject>.Pre-commit Checks
pre-commitby runningpip install pre-commitor an equivalent command.pre-commit install.pre-commit run --all-filesand fixed any reported issues.Self Review
.agents/skills/code-review/references/custom-code-style.md, especially code written or assisted by AI.mainbranch.Build and Test Coverage
python setup.py build testhas passed on a CUDA machine.python setup.py build testhas passed on an NPU machine.python setup.py build testhas passed on an MLU machine.Reviewer Notes