[PATCH] D89246: [SVE] Remove aarch64_sve_vector_pcs attribute

Cullen Rhodes via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Mon Oct 12 10:23:56 PDT 2020


c-rhodes added a comment.

In D89246#2325266 <https://reviews.llvm.org/D89246#2325266>, @efriedma wrote:

> I thought we kept it around intentionally because it has a different set of callee-save registers?  If it turns out we don't care about that, sure, I guess we can drop it.

The callee-saves are the same I think

In D89246#2325301 <https://reviews.llvm.org/D89246#2325301>, @paulwalker-arm wrote:

> I think we should keep this because it gives us the possibility to attach it to functions where the function signature wouldn't use it otherwise.
>
> Not exactly sure of the validity of this example but perhaps somebody will want to implement a printf like function (or some other debug/validation/timing function) that's only called by an existing SVE function and doesn't want the whole SVE state being preserved by the caller.

Fair enough, there may well be uses I wasn't aware of and if so we should keep it around, although I don't think in C there's anyway of telling Clang to emit the attribute.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D89246/new/

https://reviews.llvm.org/D89246



More information about the llvm-commits mailing list