clockworklabs
Allocate fewer vecs in `eval_updates`
Description of Changes
Instead of collecting to a single vec, we collect to a linked list of vecs, which is rayon's recommended way of collecting a parallel iterator into a single data structure. This means there will be fewer allocations - for N worker threads and M query plans, it's now N vecs and N linked list nodes with standard amortizable growth, instead of M vecs that get concatenated together one at a time.
API and ABI breaking changes
Expected complexity level and risk
2 - I'm pretty confident this doesn't have a behavior change and that the changes are only optimizations, but it'd probably be worth running benchmarks to make sure
Testing
- [ ] benchmark run
Criterion benchmark results
Error when comparing benchmarks: Couldn't find AWS credentials in environment, credentials file, or IAM role.
Caused by: Couldn't find AWS credentials in environment, credentials file, or IAM role.
![github-actions[bot] avatar](https://avatars.githubusercontent.com/in/15368?v=4 "Published by a pub.dev verified publisher"){kind=small}
Criterion benchmark results
Error when comparing benchmarks: Couldn't find AWS credentials in environment, credentials file, or IAM role.
Caused by: Couldn't find AWS credentials in environment, credentials file, or IAM role.
This optimization makes sense, just bear in mind that we may decide to remove parallel iteration from eval_updates entirely.
