consistent rounding #9
Closed
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
Summary: The vectorized code was rounding to even in halfway cases with _mm256_round_ps + (_MM_FROUND_TO_NEAREST_INT |_MM_FROUND_NO_EXC) (see more details in https://software.intel.com/en-us/node/523819), but we were still using std::round in a couple of places which does rounding away from zero in halfway cases. With this diff, we use std::nearbyint in all scalar code (except a few cases where we don't care exact rounding mode and uses rint which is the fastest in general) to be more consistent. nearbyint is the same as what the vectorized code does only when the current rounding mode is FE_TONEAREST but in practice this is OK because we almost always use the default rounding mode FE_TONEAREST. This is inspired by Marat's diff for mobile quantization. Reviewed By: dskhudia Differential Revision: D13017719 fbshipit-source-id: 3f5f4677b597663e1a34461feaacce71ac0e6c29
facebook-github-bot
pushed a commit
that referenced
this pull request
Nov 14, 2018
Summary: Pull Request resolved: #9 Pull Request resolved: pytorch/pytorch#13960 The vectorized code was rounding to even in halfway cases with _mm256_round_ps + (_MM_FROUND_TO_NEAREST_INT |_MM_FROUND_NO_EXC) (see more details in https://software.intel.com/en-us/node/523819), but we were still using std::round in a couple of places which does rounding away from zero in halfway cases. With this diff, we use std::nearbyint in all scalar code (except a few cases where we don't care exact rounding mode and uses rint which is the fastest in general) to be more consistent. nearbyint is the same as what the vectorized code does only when the current rounding mode is FE_TONEAREST but in practice this is OK because we almost always use the default rounding mode FE_TONEAREST. This is inspired by Marat's diff for mobile quantization. Reviewed By: dskhudia Differential Revision: D13017719 fbshipit-source-id: 6b8f99db7ea2e233aa2e3bd2adf622e03ed6258e
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.
Summary:
The vectorized code was rounding to even in halfway cases with _mm256_round_ps + (_MM_FROUND_TO_NEAREST_INT |_MM_FROUND_NO_EXC) (see more details in https://software.intel.com/en-us/node/523819), but we were still using std::round in a couple of places which does rounding away from zero in halfway cases.
With this diff, we use std::nearbyint in all scalar code (except a few cases where we don't care exact rounding mode and uses rint which is the fastest in general) to be more consistent. nearbyint is the same as what the vectorized code does only when the current rounding mode is FE_TONEAREST but in practice this is OK because we almost always use the default rounding mode FE_TONEAREST.
This is inspired by Marat's diff for mobile quantization.
Reviewed By: dskhudia
Differential Revision: D13017719