Improve element-wise tensor operation performance - #454
Merged
Conversation
fplus::transform_convert uses reserve() + back_insert_iterator, which prevents auto-vectorization because push_back updates the vector's internal end pointer on every write. Using resize() and writing via a plain random-access iterator allows the compiler to emit SIMD instructions, giving ~5% faster inference on VGG19.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
transform_tensorpreviously usedfplus::transform_convert, which callsreserve()followed bystd::transformwith aback_insert_iterator. Becausepush_backupdates the vector's internal end pointer on every write, the compiler cannot auto-vectorize the loop.This PR replaces it with
resize()and a plain random-access iterator, allowing GCC to emit SIMD instructions for element-wise operations. A fast path is also added to relu_layer for the common standard ReLU case (max(0, x)), using a simpler single-expression lambda that is easier for the compiler to optimize than the general three-branch version.Result: ~5% faster forward pass on VGG19 (measured on a single thread, release build, -march=native).