-
Notifications
You must be signed in to change notification settings - Fork 275
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
avx512gfni and avx512vaes feature names might be misleading #1325
Comments
so as vpclmulqdq |
Intel's documentation has these intrinsics marked as requiring AVX512-VL. But if there really is a CPU that has GFNI without AVX512 then you can send a PR similar to #1348 to fix this. |
Tremont supports GFNI with smaller vectors only. |
Thanks! I think it would still make sense to rename the target features and detection names to not include "avx512", since they do not imply AVX512F. I'm not sure how that would work, though, since it would need to be synchronized between this repository and the main rust repo. |
I feel that issue should be discussed in rust-lang/rust#100752 since feature names are defined in rustc, not stdarch. Once a decision has been made on that issue then we can make the necessary changes to stdarch. I expect the most likely resolution would be to add feature name aliases, since we need to keep the old names for backward compatibility. |
Sounds good to me. |
I already opened rust-lang/rust#100752, but this might be the more relevant location. More details are in the attached issue, but there exist CPUs that have GFNI and VAES without AVX-512, and Intel doesn't seem to classify them as part of AVX-512. It might be better to name them simply
gfni
andvaes
.The text was updated successfully, but these errors were encountered: