[Dygraph] Fix performance of pp+mp by using send/recv_calc_stream instead of send/recv #46116
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
|
你的PR提交成功，感谢你对开源项目的贡献! |
LiYuRio
reviewed
Sep 16, 2022
| if use_calc_stream: | ||
| task.wait() | ||
| return None | ||
| return group.process_group.recv_on_calc_stream(tensor, src) |
There was a problem hiding this comment.
不能在这个接口里用use_calc_stream这个语义，这个参数要被改成sync_op了，这样语义就错了
LiYuRio
previously approved these changes
Sep 22, 2022
FeixLiu
approved these changes
Oct 8, 2022
haohongxiang
added a commit
to haohongxiang/Paddle
that referenced
this pull request
Oct 17, 2022
fuyinno4
pushed a commit
that referenced
this pull request
Oct 18, 2022
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Outdated suggestions cannot be applied.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
Bug fixes
PR changes
Others
Describe
[Dygraph] Fix performance of pp+mp by using send/recv_calc_stream instead of send/recv
Done：
1、c_identity_op反向op - c_allreduce_sum用ProcessGroupStream下的AllReduce代替，性能提升5%左右；
2、ProcessGroupStream下新增allgather_partial算子
3、PP策略下用send/recv/allgather_partial_on_calc_stream算子代替collective下的API，避免切流引起的性能损耗；性能提升1%左右。
新老动态图精度对齐：
