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

Use busy waiting barrier in reduction to band #864

Merged
merged 3 commits into from
Jun 12, 2023

Conversation

msimberg
Copy link
Collaborator

@msimberg msimberg commented May 12, 2023

Fixes #833, using the busy wait parameter to barrier::wait in pika-org/pika#685. I've added configuration option for the busy wait time (in microseconds) specifically for reduction to band. #860 will probably also need to use the busy waiting barrier. We can see if we want to merge the options into a single BARRIER_BUSY_WAIT_US option, or if we need separate TRIDIAG_SOLVER_BARRIER_BUSY_WAIT_US, RED2BAND_BARRIER_BUSY_WAIT_US etc. The default value for the busy wait is 1 ms which seemed to show the best performance across daint and eiger, while still being relatively low (about the min task size that we anyway recommend).

@msimberg
Copy link
Collaborator Author

cscs-ci run

@msimberg
Copy link
Collaborator Author

This is high priority for performance in reduction to band, but on hold until the next pika release.

src/init.cpp Outdated Show resolved Hide resolved
src/init.cpp Outdated Show resolved Hide resolved
src/init.cpp Show resolved Hide resolved
@msimberg msimberg force-pushed the barrier-no-yield branch 2 times, most recently from 74c655f to ad057ec Compare June 1, 2023 08:38
@msimberg
Copy link
Collaborator Author

msimberg commented Jun 1, 2023

cscs-ci run

@msimberg
Copy link
Collaborator Author

msimberg commented Jun 1, 2023

I've added a config option to busy wait in the tridiagonal solver barriers as well but I have not yet tested if it has the same effect as in the reduction to band algorithm.

@msimberg msimberg marked this pull request as draft June 2, 2023 06:37
@msimberg
Copy link
Collaborator Author

msimberg commented Jun 5, 2023

cscs-ci run

@msimberg msimberg force-pushed the barrier-no-yield branch 2 times, most recently from 2e9941b to 1fe1643 Compare June 7, 2023 07:29
@msimberg
Copy link
Collaborator Author

msimberg commented Jun 7, 2023

cscs-ci run

@msimberg msimberg marked this pull request as ready for review June 8, 2023 11:52
@msimberg msimberg force-pushed the barrier-no-yield branch 3 times, most recently from add94c2 to a5ff63e Compare June 12, 2023 15:16
@msimberg
Copy link
Collaborator Author

cscs-ci run

@msimberg
Copy link
Collaborator Author

I rebased this and made the default busy wait time 0 for the rank 1 problem barrier.

@rasolca rasolca merged commit 97db8c8 into eth-cscs:master Jun 12, 2023
@msimberg msimberg deleted the barrier-no-yield branch June 12, 2023 17:23
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
Archived in project
Development

Successfully merging this pull request may close these issues.

Need of a "semi" blocking barrier for bulk
2 participants