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

Refine the semantic of DPAS OpsPerChannel and Dot Layout kWidth #3152

Open
LiyangLingIntel opened this issue Jan 14, 2025 · 0 comments
Open

Refine the semantic of DPAS OpsPerChannel and Dot Layout kWidth #3152

LiyangLingIntel opened this issue Jan 14, 2025 · 0 comments
Assignees
@LiyangLingIntel

Comments

@LiyangLingIntel
Copy link
Contributor

This issue is derived from #3140.

In upstream, kWidth in DotOperandLayout control the minimum number of continuous elements in a thread, which is decoupled with MMA layout.
While for Intel GPU, it is controled by OpsPerChannel in DPAS Layout, kWidth is not actually used.

To reduce the changes in common file and maximum reuse common functions, we may need to refine the semantic of DPAS Layout and DotOperand Layout for Intel GPU to align with the upstream definition.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@LiyangLingIntel LiyangLingIntel

Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
1 participant
@LiyangLingIntel