[llvm] r266260 - ARM: override cost function to re-enable ConstantHoisting (& fix it).

[Arnaud Allard de Grandmaison via llvm-commits]

Thanks for looking into the issue. I'll keep an eye on the bots and let you
know if the issue is fixed.

Cheers,
Arnaud

On Fri, Apr 15, 2016 at 8:23 PM, Tim Northover via llvm-commits <
llvm-commits at lists.llvm.org> wrote:

> Hi Arnaud,
>
> > On 15 Apr 2016, at 06:27, Arnaud De Grandmaison <
> Arnaud.DeGrandmaison at arm.com> wrote:
> > Would you mind having a look at those ?
>
> Sure. It looks like hoisting constants out of div/rem instructions
> (particularly in loops) prevents them being converted into cheaper
> multiplies. This only causes a slowdown for hardware I have access to on
> bigfib, but I see the same pattern in fasta's assembly.
>
> Long term, it's probably another issue for GlobalISel. For now, it doesn't
> seem horribly unreasonable to special-case it in the cost model. I've
> committed r266464 which should fix it.
>
> Thanks for bisecting the issue.
>
> Tim.
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20160415/100f09bd/attachment.html>