[PATCH] D74935: [LangRef][AliasAnalysis] Clarify `noalias` affects only modified objects

Johannes Doerfert via Phabricator via cfe-commits cfe-commits at lists.llvm.org
Sat Feb 22 15:19:25 PST 2020


jdoerfert added a comment.

In D74935#1887960 <https://reviews.llvm.org/D74935#1887960>, @nhaehnle wrote:

> In D74935#1887245 <https://reviews.llvm.org/D74935#1887245>, @efriedma wrote:
>
> > In D74935#1886020 <https://reviews.llvm.org/D74935#1886020>, @nhaehnle wrote:
> >
> > > I find this phrasing pretty confusing. How about the following:
> > >
> > > > This indicates that objects accessed via pointer values based on the argument or return value are not **modified**, during the execution of the function, via pointer values not based on the argument or return value. [...]
> >
> >
> > This isn't equivalent.  If the memory location is modified, we want to forbid both reads and writes that aren't based on the noalias pointer.
>
>
> Oh I see. So another way to put this is:
>
> - If an object is accessed via a pointer based on the noalias argument, then it cannot be modified via other pointer values.
> - If an object is modified via a pointer based on the noalias argument, then it cannot be accessed via other pointer values in any way.
>
>   Correct? I think it's useful to spell it out like this, because it makes the alias implications clearer (at least to me).


I think it might make sense to actually add a page with examples and more details. I don't have a strong opinion on the wording for this short paragraph though.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D74935/new/

https://reviews.llvm.org/D74935





More information about the cfe-commits mailing list