3.6.1 patch nomination: r230564 - only propagate equality comparisons of FP values that we are certain are non-zero

[Sanjay Patel]

I don't have any code ownership / authority, but that patch fixes an
obvious (in hindsight of course) miscompile that I introduced.

So IMO, yes it should go into 3.6 branch.

On Thu, Apr 30, 2015 at 2:29 PM, Tom Stellard <tom at stellard.net> wrote:

> Sanjay, Hal,
>
> Do either of you think this is OK for the 3.6 branch?
>
> http://llvm.org/viewvc/llvm-project?view=revision&revision=230564
>
> Thanks,
> Tom
>
> On Tue, Apr 14, 2015 at 09:41:08AM +0100, Simon Byrne wrote:
> > I'm just wanted to ping this again before the RC. Is there any chance
> > this could make 3.6.1?
> >
> > Simon
> >
> > On 12 March 2015 at 23:56, Hans Wennborg <hans at chromium.org> wrote:
> > > +Tom for 3.6.1.
> > >
> > > On Sun, Mar 8, 2015 at 6:58 AM, Simon Byrne <simonbyrne at gmail.com>
> wrote:
> > >> I would like to nominate patch r230564 for backporting to the 3.6
> > >> branch. As it stands currently, the gvn pass in 3.6.0 can sometimes
> > >> result in invalid changes to the sign of floating point zeros: see,
> > >> for example,
> > >> http://llvm.org/bugs/show_bug.cgi?id=22823
> > >> and the related Julia thread,
> > >> https://github.com/JuliaLang/julia/issues/10377
> > >>
> > >> Thank you,
> > >> Simon
> > >> _______________________________________________
> > >> llvm-commits mailing list
> > >> llvm-commits at cs.uiuc.edu
> > >> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20150430/4fc9747f/attachment.html>