MLAS: optimize QuantizeLinear - #2660
Merged
Merged
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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 adds MLAS support to quantize a buffer using vector instructions for x86/x64/ARM64. This helper is now used for the QuantizeLinear and DynamicQuantizeLinear ops.
Motivation and Context
The old implementation had a significant cost from using the C runtime to do rounding and also didn't benefit from SIMD access. The new version should be bit identical to the old version. On x64 Windows, a sample quantized resnet50 model goes from 420ms->130ms.
The ARM32 build continues to the use the C runtime. A vectorized implementation is possible here too, but stick in ARMv7A NEON, such an implementation would require dropping to assembly and that's not worth it at this point.
An AVX+ implementation is also possible, but also not worth it at this point as the test quantized models are now dominated by GEMM and MlasQuantizeLinear is a small % of the runtime.