[PATCH] D63046: [Attributor] Deduce "willreturn" function attribute

Hideto Ueno via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Fri Jul 12 14:56:30 PDT 2019


uenoku added a comment.

In D63046#1583582 <https://reviews.llvm.org/D63046#1583582>, @jdoerfert wrote:

> In D63046#1583471 <https://reviews.llvm.org/D63046#1583471>, @uenoku wrote:
>
> > For some intrinsics, I can not judge whether it is allowed to mark `willreturn`. I want to hear an opinion.
> >
> > - Standard C libray: especially `memcpy, memmove, memset`
>
>
> Can we do the following:
>
> - Commit this with conservative intrinsic handling. That is, without all the intrinsics marking stuff.
> - Open another patch for the marking stuff as you have it now in here. Basically, split the patch in two and commit the first part assuming the comments in (https://reviews.llvm.org/D63046#1572140) have been addressed.
>
>   To answer your question: `memcpy, memmove, memset` are `willreturn`


OK, I'll split a patch for `mem*`or other stuff. 
In my opinion, the annotation in this patch is enough conservative (I annotated to arithmetic function and debugging function). Is there any problem with this annotation?


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D63046/new/

https://reviews.llvm.org/D63046





More information about the llvm-commits mailing list