-
Notifications
You must be signed in to change notification settings - Fork 244
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
allow libm_intrinsics fallback #985
base: main
Are you sure you want to change the base?
Conversation
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I feel like it would be less effort to not have entries in the libm_intrinsics
map for any function we don't want to hook at all.
Looking a bit closer at it, several LibmCustomIntrinsic
s are implemented by expanding them to GLSL ops, so I think the reason the remaining ones panic is that they were meant to be implemented the same way.
Do you have a specific subset in mind that you need?
} | ||
*/ | ||
_ => return None | ||
}) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Commenting this stuff out doesn't seem the best. I think I understand what's going on, but maybe each of these should be evaluated to see if there is a GLSL fallback and removed from the LimCustomIntrinsic
type entirely, if not.
It would be nice if the only software operations kept would be the ones that are relatively cheap (and mostly just operate on the binary representation).
i need erf* functions and it doesn't seem to be on GLOps(i'm not sure) |
allow using libm software implementation as fallback