[PATCH] D30161: [LoopUnrolling] Peel loops with invariant backedge Phi input

[Maxim Kazantsev via llvm-commits]

Thanks for finding this problem! I've uploaded a follow-up fix https://reviews.llvm.org/D30632 

-----Original Message-----
From: Sanjoy Das [mailto:sanjoy at playingwithpointers.com] 
Sent: Saturday, March 4, 2017 9:26 AM
To: Michael Kuperstein <mkuper at google.com>
Cc: reviews+D30161+public+7d0eaf13f8916a85 at reviews.llvm.org; Maxim Kazantsev <max.kazantsev at azul.com>; Artur Pilipenko <apilipenko at azul.com>; Igor Laevsky <igor at azul.com>; Anna Thomas <anna at azul.com>; Philip Reames <listmail at philipreames.com>; Friedman, Eli <efriedma at codeaurora.org>; Mikhail Zolotukhin <mzolotukhin at apple.com>; llvm-commits <llvm-commits at lists.llvm.org>
Subject: Re: [PATCH] D30161: [LoopUnrolling] Peel loops with invariant backedge Phi input

On Fri, Mar 3, 2017 at 11:55 AM, Michael Kuperstein <mkuper at google.com> wrote:
> Hm, right, thanks Eli!
>
> This ought to have the same threshold checks as the rest of loop 
> unrolling - since we're only unrolling one iteration here, I guess it 
> should just be "LoopSize <= UP.Threshold".

Should I revert the change for now?

>
> Michael
>
> On Fri, Mar 3, 2017 at 11:48 AM, Eli Friedman via Phabricator 
> <reviews at reviews.llvm.org> wrote:
>>
>> efriedma added inline comments.
>>
>>
>> ================
>> Comment at: llvm/trunk/lib/Transforms/Utils/LoopUnrollPeel.cpp:95
>> +      UP.PeelCount = 1;
>> +      return;
>> +    }
>> ----------------
>> Do we need to do some sort of threshold check here?  At first glance, 
>> it looks like this will peel a loop of any size.
>>
>>
>> Repository:
>>   rL LLVM
>>
>> https://reviews.llvm.org/D30161
>>
>>
>>
>