[llvm] r274642 - [TTI] The cost model should not assume vector casts get completely scalarized

[Jordy Potman via llvm-commits]

I have created http://reviews.llvm.org/D22126 with a fix for this.

Jordy

On Thu, Jul 7, 2016 at 7:17 PM, Michael Kuperstein <mkuper at google.com>
wrote:

> Yes, I think you're right, this looks like a bug.
>
> On Thu, Jul 7, 2016 at 1:43 AM, Jordy Potman <jordy.potman.lists at gmail.com
> > wrote:
>
>> Hi,
>>
>> I know this functionality wasn't changed in this commit but I noticed
>> that the arguments to the getScalarizationOverhead call appear to be
>> inconsistent with the comment in the following piece of code:
>>
>> @@ -480,8 +497,7 @@ public:
>>>            Opcode, ValTy->getScalarType(), CondTy);
>>>
>>>        // Return the cost of multiple scalar invocation plus the cost of
>>> -      // inserting
>>> -      // and extracting the values.
>>> +      // inserting and extracting the values.
>>>        return getScalarizationOverhead(ValTy, true, false) + Num * Cost;
>>>      }
>>>
>>
>> Shouldn't the getScalarizationOverhead call be:
>>
>> getScalarizationOverhead(ValTy, true, true)
>>
>> ?
>>
>> Jordy
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20160708/028ae3cd/attachment.html>