perf: run map_variations in reconsensus concurrently #107
Conversation
This file contains 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
This uses rayon's parallel iterator to run `map_variations()` (basically Nextclade) for individual sequences concurrently in `reconsensus()` step. We already run it concurrently in `solve_promise()`step, [for individual blocks](https://github.com/neherlab/pangraph/blob/2d2e7b046cbbbc24d1c364f5a0afe2176b662a99/packages/pangraph/src/pangraph/graph_merging.rs#L145-L150) and then concurrently [for alignment within each block](https://github.com/neherlab/pangraph/blob/2d2e7b046cbbbc24d1c364f5a0afe2176b662a99/packages/pangraph/src/pangraph/reweave.rs#L38-L49), so I thought we can repeat this success in the `reconsensus()` step as well. This change results in 64.4% speedup in my measurements on ecoli. Command: ```bash /usr/bin/time -qf 'Cmd : %C\nTime: %E\nMem : %M KB' pangraph build -b 20 -l 500 --circular data/ecoli.fa.gz -o tmp/ecoli.fa.gz.json -v ``` | Branch | Time | |-----------------------------------------|---------| | rust (commit 2d2e7b0) | 5m 51s | | perf/parallel-nextclade-reconsensus | 3m 33s | Things to watch out: * increased memory usage due to increased concurrency * potential reorder of the results
ivan-aksamentov
added a commit
that referenced
this pull request
Jan 15, 2025
Similar to #107 but less succesful. Here I tried to introduce more parallelism in random places where I found `.iter()` or a a plain loop and where rayon's `.par_iter()` could be used (technically; scientific correctness is to be verified). In my measurements this brings no speedup at all compared to base branch. I'll just leave this as an idea for the future optimization. Perhaps there's a smarter way to find places where we can squeeze some more parallelism, or we could restructure the algo such that new parallelization opportunities appear. At the time, this is low priority I think, so let's focus on other things.
Closed
mmolari
approved these changes
Jan 17, 2025
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.
This uses rayon's parallel iterator to run
map_variations()(basically Nextclade) for individual sequences concurrently inreconsensus()step.We already run it concurrently in
solve_promise()step, for individual blocks and then concurrently for alignment within each block, so I thought we can repeat this success in thereconsensus()step as well.This change results in 64.4% speedup in my measurements on ecoli.
Command:
/usr/bin/time -qf 'Cmd : %C\nTime: %E\nMem : %M KB' pangraph build -b 20 -l 500 --circular data/ecoli.fa.gz -o tmp/ecoli.fa.gz.json -vThings to watch out:
Both are also true for all our existing parallel loops.
Related: #108