Skip to content

Conversation

@llvm-beanz
Copy link
Collaborator

This just removes the comment about being tied to wave size since the hardware sizes do not match wave sizes.

This just removes the comment about being tied to wave size since the
hardware sizes do not match wave sizes.
@llvm-beanz llvm-beanz changed the title Small update to language to clarify dimension requirements [0034] Small update to clarify dimension requirements Nov 18, 2025
@llvm-beanz llvm-beanz changed the title [0034] Small update to clarify dimension requirements [0035] Small update to clarify dimension requirements Nov 18, 2025
@mjbedy
Copy link

mjbedy commented Nov 18, 2025

As I recall, wasn't there also a concern that the power of 2 limitation on K may also not map well to hardware? Multiples of a granularity might be a better limit; although unclear if that detail belongs here or elsewhere. (I'm looking at the table just below this change.)

@llvm-beanz
Copy link
Collaborator Author

As I recall, wasn't there also a concern that the power of 2 limitation on K may also not map well to hardware? Multiples of a granularity might be a better limit; although unclear if that detail belongs here or elsewhere. (I'm looking at the table just below this change.)

Right! I'll just provide the range of valid values.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

Status: No status

Development

Successfully merging this pull request may close these issues.

4 participants