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.

Sign up for GitHub

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

Jump to bottom

Change FH -> HF #2045

Closed
wants to merge 1 commit into from
Closed

Change FH -> HF #2045

wants to merge 1 commit into from

Conversation

charleskawczynski
Copy link
Member

WIP to explore HF datalayouts.

@charleskawczynski charleskawczynski force-pushed the ck/move_field_dim_to_end branch 2 times, most recently from ef81982 to 96e3e4d Compare October 16, 2024 20:41
@charleskawczynski charleskawczynski mentioned this pull request Oct 17, 2024
Closed
@sriharshakandala
Copy link
Member

sriharshakandala commented Oct 17, 2024 •
edited
Loading

I think this can happen naturally, with existing data layouts, by separating the storage for prognostic variables. This naturally restricts the NF to at most 3 (for example for velocity), except for passive scalars! The F dimension can be used for tuning most expensive kernels occuring in our workload rather than for collecting storage for all (cell-centered or face-centered) variables

@charleskawczynski charleskawczynski force-pushed the ck/move_field_dim_to_end branch 6 times, most recently from c7687f0 to 82f145e Compare October 21, 2024 13:35
@charleskawczynski charleskawczynski mentioned this pull request Oct 21, 2024
Merged
@charleskawczynski charleskawczynski force-pushed the ck/move_field_dim_to_end branch 2 times, most recently from dbd418b to 6c15d08 Compare October 21, 2024 16:33
@charleskawczynski
Change FH -> HF
b7d55db 
Swap F and H inds
@charleskawczynski
Copy link
Member Author

Superseded by #2055

@charleskawczynski
Copy link
Member Author

I think this can happen naturally, with existing data layouts, by separating the storage for prognostic variables. This naturally restricts the NF to at most 3 (for example for velocity), except for passive scalars! The F dimension can be used for tuning most expensive kernels occuring in our workload rather than for collecting storage for all (cell-centered or face-centered) variables

I'm not sure I follow this comment. This PR was changing the field dimension to the end in order to allow us to specialize on linear indexing for field variables without requiring users to change their code.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers
No reviews
Assignees
No one assigned
Labels
breaking 💥 performance
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
2 participants
@charleskawczynski @sriharshakandala