[llvm] r258184 - [SCEV] Fix PR26207

[Sanjoy Das via llvm-commits]

+CC Andy

I'd wait for Andy to chime in, I think this is okay to merge into 3.8
provided no problems are seen over the next 1-2 days.



On Tue, Jan 19, 2016 at 2:34 PM, Hans Wennborg <hans at chromium.org> wrote:
> On Tue, Jan 19, 2016 at 12:53 PM, Sanjoy Das via llvm-commits
> <llvm-commits at lists.llvm.org> wrote:
>> Author: sanjoy
>> Date: Tue Jan 19 14:53:51 2016
>> New Revision: 258184
>>
>> URL: http://llvm.org/viewvc/llvm-project?rev=258184&view=rev
>> Log:
>> [SCEV] Fix PR26207
>>
>> In some cases, the max backedge taken count can be more conservative
>> than the exact backedge taken count (for instance, because
>> ScalarEvolution::getRange is not control-flow sensitive whereas
>> computeExitLimitFromICmp can be).  In these cases,
>> computeExitLimitFromCond (specifically the bit that deals with `and` and
>> `or` instructions) can create an ExitLimit instance with a
>> `SCEVCouldNotCompute` max backedge count expression, but a computable
>> exact backedge count expression.  This violates an implicit SCEV
>> assumption: a computable exact BE count should imply a computable max BE
>> count.
>>
>> This change
>>
>>  - Makes the above implicit invariant explicit by adding an assert to
>>    ExitLimit's constructor
>>
>>  - Changes `computeExitLimitFromCond` to be more robust around
>>    conservative max backedge counts
>
> Is this something that should be merged to 3.8?



-- 
Sanjoy Das
http://playingwithpointers.com