Conversation
Each worker owns one input list, and SingleProducerSingleConsumerList only lets a consumer take an item once another item has been produced behind it (or the producer has finished). Reader backpressure used a fixed PACK_IN_MEM_LIMIT (32) that is smaller than the worker count on machines with more than 32 cores. With 33+ workers the readers stop after 33 packs, before any worker list holds two items, so no pack is ever consumed and all reader and worker threads wait on the backpressure condition variable forever. The writer-backlog check has the same shape: the writer drains worker lists round-robin and waits on a list until its worker produces a second output, so a fixed limit below the worker count can also block the readers permanently. Allow at least two in-flight packs per worker in both the reader/processor and reader/writer backpressure checks (max(PACK_IN_MEM_LIMIT, 2 * threads)), for PE, interleaved PE and SE readers. The lock-free list semantics are unchanged, so this does not reintroduce OpenGene#695.
The reader/writer backlog check still gated on the fixed PACK_IN_MEM_LIMIT * PACK_SIZE period instead of the new mPackInMemLimit, so it fired far more often than the (now larger) buffer actually needs at high thread counts. Not a correctness bug — the comparisons inside already used mPackInMemLimit — just an inconsistency worth cleaning up alongside it. Also removed a stray extra blank line left in common.h.
fastp benchmark✅ No regressions over 10%.
stage timings
|
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Layer 1: the OpenGene#721 deadlock fix (upstream OpenGene#723).
Cumulative: includes every layer below it, so the benchmark comment compares the whole stack up to here against master (upstream + CI only). The per-layer effect is the difference from the previous PR's comment.
🤖 Generated with Claude Code