-
Notifications
You must be signed in to change notification settings - Fork 367
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
Upstream BLIS patches for ARM SVE? #807
Comments
IIUC these |
Thanks for the response. I think that's basically the idea. |
OK, I think I maybe see where this started and what the intent was. It looks like @xrq-phys's patch added However, the part about adding a64fx to the arm64 family could be upstreamed. |
This is, assuming run-time support for checking for A64fx exists which I'm not sure about. I think A64fx also has some other optimizations relying on specific Fujitsu compiler support which could make it difficult to co-exist with other arm64 configurations. |
OK, looks like you understand the reasons (definitely more than I do.) Appreciate you looking into it more deeply. |
We have a multi-platform, cross-compiling build system for binaries in the Julia ecosystem ( BinaryBuilder.jl ), with the Yggdrasil repo hosting the recipes.
While the latest recipe for the shiny BLIS v1.0 is here: JuliaPackaging/Yggdrasil#8626 ,
the original recipe was kindly written by BLIS contributor @xrq-phys who wrote patches for optimisations on ARM SVE architectures. The patches essentially reduce to
in a number of locations (see https://github.com/JuliaPackaging/Yggdrasil/pull/8626/files for the full picture).
So our question is: do we need to keep carrying these patches, or can these be upstreamed to BLIS in any form?
Thanks for any clarification or suggestions here.
The text was updated successfully, but these errors were encountered: