[llvm] r227809 - Fix ARM peephole optimizeCompare to avoid optimizing unsigned cmp to 0.

[Hans Wennborg]

Great. Merged in r227864.

Thanks,
Hans

On Mon, Feb 2, 2015 at 1:23 PM, Renato Golin <renato.golin at linaro.org> wrote:
> Hi Hans, I remember this one. I think it should be safe to pull it in, since
> it's self contained and fixes a real bug.
>
> Cheers,
> Renato
>
> On 2 Feb 2015 21:12, "Hans Wennborg" <hans at chromium.org> wrote:
>>
>> On Mon, Feb 2, 2015 at 8:56 AM, Jan Wen Voung <jvoung at google.com> wrote:
>> > Author: jvoung
>> > Date: Mon Feb  2 10:56:50 2015
>> > New Revision: 227809
>> >
>> > URL: http://llvm.org/viewvc/llvm-project?rev=227809&view=rev
>> > Log:
>> > Fix ARM peephole optimizeCompare to avoid optimizing unsigned cmp to 0.
>> >
>> > Summary:
>> > Previously it only avoided optimizing signed comparisons to 0.
>> > Sometimes the DAGCombiner will optimize the unsigned comparisons
>> > to 0 before it gets to the peephole pass, but sometimes it doesn't.
>> >
>> > Fix for PR22373.
>>
>> Renato: I think we want this for 3.6. Does it look ok to you?
>>
>>  - Hans