Optimize memory usage in eval_mle_at_point_blocking #4
Conversation
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
erabinov
pushed a commit
that referenced
this pull request
Dec 2, 2025
* init * init * params * merkle tree traits * d * poseidon2_lernels * compiles * fix indices * fix index * fix kernels * yay * refactor traits * checkpoint * config * default clone implementation * update paths * fix openings * ten * refs * checkpoint * checkpoint * init fixed * dimensions * indexing * indexing * index * view mut * tensor indexing * save * works * test init * mtree * fix dir * with warmup in merkle test * transpose * transpose works * gr * transpose * alloc mut * derive cuda send * runner * remote * try * try * action * setup * better * comment out toolchain * try it * try * ff * repo * gchange settung * with * rm runs-on * try * merkle tree * try test * try * runs-on * verify cuda * different runner * back to default * runs-on * check if needed * check if needed * echo * try with apt * from installer * install from repo * setup cuda via paths * try echo cuda * try * fix cuda version * add clippy action
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Outdated suggestions cannot be applied.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.
Description:
This PR improves the performance of the eval_mle_at_point_blocking function in multilinear/src/eval.rs by reducing memory allocation overhead.
Closes #3
What Was Optimized
The original implementation created temporary vectors repeatedly inside parallel blocks, which caused unnecessary allocations and impacted performance. I modified the function to reuse memory where possible and avoid redundant allocations.
Why This Matters
These changes make the function more efficient—particularly for large tensors—by improving cache locality and reducing allocation overhead. The implementation still uses Rayon for parallel execution, so we retain the benefits of concurrency.
Benchmarks
Running local benchmarks showed a performance improvement of 15–25% on large input sizes.
Compatibility and Testing