Performance: Optimize Mel filterbank application with sparse matrix indices#9
Open
Performance: Optimize Mel filterbank application with sparse matrix indices#9
Conversation
…ndices - Precompute `_fbStart` and `_fbEnd` indices in `MelSpectrogram` constructor to identify non-zero ranges in the filterbank. - Update `computeRawMel` to iterate only over non-zero elements, skipping unnecessary multiplications by zero. - Yields a ~3.5x speedup (RTF from 0.014 to 0.004) in benchmarks. - Verified with existing tests and ONNX reference cross-validation.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Performance Improvement
Optimized the Mel filterbank matrix multiplication step in
computeRawMelby leveraging the sparsity of the filterbank matrix.Changes:
MelSpectrogramconstructor, we now scan themelFilterbankmatrix (which is ~98.5% sparse) and store the start and end indices of non-zero values for each mel bin in_fbStartand_fbEnd(Int32Arrays).computeRawMel, we now iterate only fromstarttoendinstead of the fullnFreqBinsrange.Impact:
process()calls.npm test, passing all 52 tests including numerical cross-validation against ONNX reference implementation.Verification:
Run
npm testto verify correctness.(Optional) Create a benchmark script (like the one used during development) to verify speedup.
PR created automatically by Jules for task 10792195913597466588 started by @ysdede