Enable OMP multithreading in lookup_table_v2 #45249
Merged
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
This reverts commit f4e4f8d.
|
你的PR提交成功，感谢你对开源项目的贡献! |
jczaja
reviewed
Aug 22, 2022
| @@ -48,6 +48,11 @@ struct EmbeddingCPUFunctor { | |||
| dev_ctx_.template Alloc<T>(out_); | |||
| auto* output = out_->data<T>(); | |||
|
|
|||
| #if defined _OPENMP | |||
| #ifndef PADDLE_WITH_CUDA | |||
There was a problem hiding this comment.
Maybe you could reformulate this conditions like: #if defined(_OPENMP) && !defined(PADDLE_WITH_CUDA)..... perhaps it would look more readable?
tsocha
suggested changes
Aug 22, 2022
jczaja
approved these changes
Aug 25, 2022
tsocha
approved these changes
Aug 25, 2022
|
@jiangjiajun Please review |
|
你的PR已合入Paddle库，请关注后续测试结果。 |
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Outdated suggestions cannot be applied.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.
PR types
Performance optimization
PR changes
OPs
Describe
This PR enables OMP multithreading on the lookup_table_v2 operator which makes the operator take less time when running the ernie 3.0 model with more than 1 thread,
The results of the speedup on SPR with FP32 datatype are as follows (Total execution time measured by the Paddle profiler. Before all the times were similar to the 1 thread time):
1 thread: 727.331
2 threads: 483.426
5 threads: 374.056
The change allows for a 2x speedup of the operator after increasing the number of threads.