[bert/RoBERTa] Optimize LayerNorm with explicit vectorization using Vec256 (2/2) #29154
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
…ec256 (2/2) We would like to optimize LayerNorm with explicit vectorization using Vec256. This PR handles the special part of using fmadd with AVX256. Differential Revision: [D18307639](https://our.internmc.facebook.com/intern/diff/D18307639/) [ghstack-poisoned]
This was referenced Nov 4, 2019
jianyuh
added a commit
that referenced
this pull request
Nov 4, 2019
…ec256 (2/2) We would like to optimize LayerNorm with explicit vectorization using Vec256. This PR handles the special part of using fmadd with AVX256. Differential Revision: [D18307639](https://our.internmc.facebook.com/intern/diff/D18307639/) ghstack-source-id: 93220825 Pull Request resolved: #29154
…ion using Vec256 (2/2)" We would like to optimize LayerNorm with explicit vectorization using Vec256. This PR handles the special part of using fmadd with AVX256. Differential Revision: [D18307639](https://our.internmc.facebook.com/intern/diff/D18307639/) [ghstack-poisoned]
jianyuh
added a commit
that referenced
this pull request
Nov 10, 2019
…ec256 (2/2) Pull Request resolved: #29154 We would like to optimize LayerNorm with explicit vectorization using Vec256. This PR handles the special part of using fmadd with AVX256. ghstack-source-id: 93608764 Differential Revision: [D18307639](https://our.internmc.facebook.com/intern/diff/D18307639/)
jamesr66a
approved these changes
Nov 20, 2019
| const T bias = -rstd_val * mean_val; | ||
| for (int64_t j = 0; j < N; ++j) { | ||
| for (j = 0; j < N / kVecSize * kVecSize; j += kVecSize) { | ||
| const vec256::Vec256<T> gamma_vec = gamma_null |
Collaborator
There was a problem hiding this comment.
What does the emitted code look like with these conditionals?
Collaborator
|
Looks like this PR hasn't been updated in a while so we're going to go ahead and mark this as |
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.
Stack from ghstack:
We would like to optimize LayerNorm with explicit vectorization using Vec256. This PR handles the special part of using fmadd with AVX256.
Differential Revision: D18307639