[llvm] r244676 - Fix PR24354.

[Chris Lattner via llvm-commits]

The patch doesn’t look scary to me, but I think it would be better to find someone who has worked in this area more recently.  Is anyone else on llvm-commits willing to do a review here?

We should really also get an instcombine maintainer, it seems like there are several people who could be qualified for this.

-Chris

> On Aug 11, 2015, at 3:01 PM, Hans Wennborg <hans at chromium.org> wrote:
> 
> Doesn't seem like we have a specific instcombine owner, so that makes
> Chris the owner. OK for merging?
> 
> On Tue, Aug 11, 2015 at 2:58 PM, David Majnemer <majnemer at google.com> wrote:
>> Seems fine to me but I'm not code owner of InstCombine.
>> 
>> On Tue, Aug 11, 2015 at 5:55 PM, Hans Wennborg <hans at chromium.org> wrote:
>>> 
>>> Should we merge this to 3.7? David, what do you think?
>>> 
>>> On Tue, Aug 11, 2015 at 2:33 PM, Sanjoy Das via llvm-commits
>>> <llvm-commits at lists.llvm.org> wrote:
>>>> Author: sanjoy
>>>> Date: Tue Aug 11 16:33:55 2015
>>>> New Revision: 244676
>>>> 
>>>> URL: http://llvm.org/viewvc/llvm-project?rev=244676&view=rev
>>>> Log:
>>>> Fix PR24354.
>>>> 
>>>> `InstCombiner::OptimizeOverflowCheck` was asserting an
>>>> invariant (operands to binary operations are ordered by decreasing
>>>> complexity) that wasn't really an invariant.  Fix this by instead having
>>>> `InstCombiner::OptimizeOverflowCheck` establish the invariant if it does
>>>> not hold.
>>>> 
>>>> Added:
>>>>    llvm/trunk/test/Transforms/InstCombine/pr24354.ll
>>>> Modified:
>>>>    llvm/trunk/lib/Transforms/InstCombine/InstCombineCompares.cpp
>> 
>>