Support AVX-512 for SIMD unary operators, softmax #131
Conversation
This file contains 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
Support AVX-512 for Exp, Sigmoid, Tanh, Erf, Softmax and other operators which use SIMD-accelerated implementations in rten-vecmath. The performance benefit is small because these operations are limited by memory bandwidth, but doing this revealed some issues with supporting different instruction sets on the same architecture. Usage of `#[inline]` and `#[target_feature]` had to be corrected to ensure that SIMD intrinics are actually inlined reliably. Otherwise the function call overhead negates the benefits of SIMD entirely. The `SimdInt` and `SimdFloat` docs have been updated to explain the rules around inlining and target features.
robertknight
added a commit
that referenced
this pull request
Apr 29, 2024
Remove the `#[target_feature]` attributes from the generic erf, tanh functions. Previously they were being compiled with AVX2 instructions even when called from the fallback (non-AVX2) dispatch. This change follows the pattern established in #131, which fixed the issue for Exp, Sigmoid and Softmax operators.
This was referenced Apr 29, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
Support AVX-512 for Exp, Sigmoid, Tanh, Erf, Softmax and other operators which use SIMD-accelerated implementations in rten-vecmath.
The performance benefit is small because these operations are limited by memory bandwidth, but doing this revealed some issues with supporting different instruction sets on the same architecture. Usage of
#[inline]and#[target_feature]had to be corrected to ensure that SIMD intrinics are actually inlined reliably. Otherwise the function call overhead negates the benefits of SIMD entirely.The
SimdIntandSimdFloatdocs have been updated to explain the rules around inlining and target features.