Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[Speed] Refine elementwise sub,div,min,max gradient functor #8820

Merged

Conversation

chengduoZH
Copy link
Contributor

@chengduoZH chengduoZH commented Mar 7, 2018

fix #8821
related issue #8811

@chengduoZH chengduoZH changed the title [Speed]Refine elementwise sub,div,min,max [Speed]Optimize elementwise sub,div,min,max Mar 7, 2018
@jacquesqiao
Copy link
Member

maybe the title should be refine elementwise sub,div,min,max gradient functor

@chengduoZH chengduoZH changed the title [Speed]Optimize elementwise sub,div,min,max [Speed] Refine elementwise sub,div,min,max gradient functor Mar 7, 2018
@chengduoZH
Copy link
Contributor Author

chengduoZH commented Mar 7, 2018

@jacquesqiao Thanks

@QiJune
Copy link
Member

QiJune commented Mar 7, 2018

Could you please show the performance improvement between these two versions code?

@chengduoZH
Copy link
Contributor Author

@QiJune I did not do some experiment but from the results of optimized elementwise_add_grad and elementwise_mul_grad, it will improve nearly ten times.
elementwise_mul_grad: #8810
elementwise_add_grad: #8402 (review)
elementwise_add_grad: #8539

Copy link
Member

@jacquesqiao jacquesqiao left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM!

@chengduoZH chengduoZH merged commit f3cdeb9 into PaddlePaddle:develop Mar 8, 2018
@chengduoZH chengduoZH added this to Done in Performance Tuning Mar 8, 2018
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
Development

Successfully merging this pull request may close these issues.

Optimize elementwise sub,div,min,max gradient functor
3 participants