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

[LLVMGPU] Enable scf.forall distr. on vectorDistribute Pipeline #19420

Open
wants to merge 1 commit into
base: main
Choose a base branch
from

Conversation

pashu123
Copy link
Contributor

@pashu123 pashu123 commented Dec 9, 2024

Enables scf.forall distribution on the vector distribute pipeline.

@pashu123 pashu123 force-pushed the scfforallgpu branch 3 times, most recently from f6c10c5 to 6719bb3 Compare December 16, 2024 06:39
@pashu123 pashu123 force-pushed the scfforallgpu branch 3 times, most recently from f96b601 to f442fdf Compare December 17, 2024 12:30
@pashu123 pashu123 force-pushed the scfforallgpu branch 10 times, most recently from f06be62 to 08aa578 Compare January 9, 2025 13:14
@pashu123 pashu123 marked this pull request as ready for review January 9, 2025 13:16
@pashu123 pashu123 requested a review from Max191 January 9, 2025 13:59
@pashu123 pashu123 force-pushed the scfforallgpu branch 4 times, most recently from a42b243 to 2a25ddd Compare January 10, 2025 16:13
@pashu123 pashu123 force-pushed the scfforallgpu branch 6 times, most recently from 44709eb to d5f856a Compare January 10, 2025 19:02
Copy link
Contributor

@Max191 Max191 left a comment

Choose a reason for hiding this comment

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

It seems a bit strange that the vector sizes have changed in vector distribute and some of the read/writes from global to shared memory have disappeared in tile and fuse. Do you know what caused these differences?

Copy link
Contributor

@Max191 Max191 left a comment

Choose a reason for hiding this comment

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

Also, there are 3 separate things happening here. Can you split out the 3 commit into separate PRs (The workgroup reordering, slice optimization, and forall distribution enablement)?

@pashu123 pashu123 force-pushed the scfforallgpu branch 3 times, most recently from a079c3b to afa5601 Compare January 17, 2025 06:41
@pashu123 pashu123 requested a review from Max191 January 17, 2025 07:35
Copy link
Contributor

@Max191 Max191 left a comment

Choose a reason for hiding this comment

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

Looks good, just one nit. Please wait for @Groverkss's review too.

Enables scf.forall distribution on the vector distribute pipeline.
@Max191 Max191 self-requested a review January 17, 2025 16:15
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants