Fix pipeline race condition: rotate all buffers by pipeline depth#53
Merged
Merged
Conversation
…ple#46) With pipeline depth 3, the GPU sampler output and logits buffers were shared across in-flight stages, causing stale reads (repeated tokens) under CPU contention. Introduce a shared pipelineDepth constant and rotate decodeOutputBuffers, decodeLogitsBuffers, and cachePositionBuffers so no two concurrent stages alias the same memory.
stikves
force-pushed
the
sukru/fix-pipeline-race-condition
branch
from
0247a60 to
7e0bc95
Compare
stikves
requested review from
DawerG,
alejandro-isaza,
carinapeng,
kevchengcodes and
tjia1818
alejandro-isaza
approved these changes
Jun 18, 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
2 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
With pipeline depth 3, the GPU sampler output and logits buffers were shared across in-flight stages, causing stale reads (repeated tokens) under CPU contention.
Introduce a shared pipelineDepth constant and rotate decodeOutputBuffers, decodeLogitsBuffers, and cachePositionBuffers so no two concurrent stages alias the same memory.
Fixes #46