Cached benchmark file I/O - #1700
Draft
cmdupuis3 wants to merge 12 commits into
Draft
Conversation
cmdupuis3
added
scalability
ASV BenchmarkingBenchmark Comparison ResultsBenchmarks that have improved:
Benchmarks that have stayed the same:
|
Collaborator
Author
|
In timing the |
asv preimports the benchmark suite before it runs any setup_cache (asv/runner.py, spawner.preimport() ahead of the run loop), so on a cold cache bench_connectivity's import-time preload_topologies is what fills it -- serially, in the forkserver parent, before a single benchmark starts. CachedFixtures.setup_cache then finds everything already built, and prime(workers=...) never runs on the path it was written for. Filling it from the CLI first puts those reads back in the parallel prime. Worth a second or two on the GitHub runners, which only see the oQU grids; worth rather more on a machine that can reach the four dyamond grids on campaign storage. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
cmdupuis3
force-pushed
the
cmd/bench_io
branch
from
August 25, 2026 18:07
fd82317 to
4dad004
Compare
This was referenced Aug 25, 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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Closes #1698
Overview
This PR is a large-scale overhaul of the benchmarks to support single-loads of files per benchmark suite.
Currently, the ASV benchmark suite runs a new process per benchmark, requiring full-scale loads of files for every single benchmark. There are a limited number of files that are actually used to benchmark, so we end up loading many files dozens of times, running up wallclock time on the benchmark suite as a whole. This PR instead has a unified approach, where a parent process is launched for the whole benchmark suite, uncached files are retrieved into
/uxarray/benchmarks/_io_cache/, and benchmarks are spawned on forked processes that inherit visibility into the file cache and some other properties.Simultaneously, this reorganization of file I/O will help to expand the scope of what most non-I/O benchmarks are able to test, since some of them were artificially restricted to smaller grids. That raises another question about how many benchmarks really need to be run at every resolution, but I think that's out of scope for this PR, since it would alter the actual benchmark suite, rather than just changing how they are run.
General CI runs should see a wallclock improvement for whole benchmark suites. CI benchmark runs are showing improvements of about 19 minutes so far (38min -> 19min). Runs on HPC can expect larger improvements. Larger benchmark suites will benefit more from cached file I/O, while benchmarks run individually won't see much improvement.
Other general speed improvements:
PR Checklist
General
Testing & Benchmarking
Documentation
_)AI Disclosure
AI Usage: Claude Opus 5