[PATCH] D112326: [AArch64] Async unwind - do not schedule frame setup/destroy
Momchil Velikov via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Fri Jan 21 03:14:35 PST 2022
chill added a comment.
In D112326#3256195 <https://reviews.llvm.org/D112326#3256195>, @efriedma wrote:
> Oh, I understand the issue; we need to ensure that any instruction annotated by a CFI directive is not rescheduled away from the directive. (This isn't a problem for Windows EH because there's never more than one instruction between two SEH markings.)
>
> I don't see how this is specific to instructions that modify fp, though.
> We end up with inaccurate CFI if we put any instruction between an instruction and its associated CFI directive, whether or not that instruction modifies fp.
No, not necessarily, but, that said ...
> Can we explicitly check for instructions followed by a CFI directive?
... I think this would work. I'll give it a try.
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D112326/new/
https://reviews.llvm.org/D112326
More information about the llvm-commits
mailing list