[llvm-commits] InstCombine patch to improve combining of nested ShuffleVectorInst's

[Eli Friedman]

On Tue, Oct 11, 2011 at 3:54 PM, Guo, Xiaoyi <Xiaoyi.Guo at amd.com> wrote:
> Sorry. Try again...

We intentionally don't do generalized shuffle transformations in
instcombine at the moment because, at least at the time the
shufflevector instruction was introduced, we didn't really trust the
quality of the shuffle-matching code in the backends for shuffles
which don't have an obvious single-instruction mapping.  I'm not sure
that situation has improved...

-Eli

> -----Original Message-----
> From: Bruno Cardoso Lopes [mailto:bruno.cardoso at gmail.com]
> Sent: Tuesday, October 11, 2011 3:06 PM
> To: Guo, Xiaoyi
> Cc: llvm-commits at cs.uiuc.edu
> Subject: Re: [llvm-commits] InstCombine patch to improve combining of nested ShuffleVectorInst's
>
> The attached patch only contain testcases! You probably missed something! :)
>
> On Tue, Oct 11, 2011 at 6:58 PM, Guo, Xiaoyi <Xiaoyi.Guo at amd.com> wrote:
>> The attached patch is to improve InstCombine to handle more cases of
>> nested ShuffleVectorInst's. New test cases are added to vec_shuffle.ll.
>>
>>
>>
>> It would be much appreciated if someone can review it and if ok help
>> to commit it.
>>
>>
>>
>> Thanks,
>>
>> Xiaoyi
>>
>> _______________________________________________
>> llvm-commits mailing list
>> llvm-commits at cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>>
>>
>
>
>
> --
> Bruno Cardoso Lopes
> http://www.brunocardoso.cc
>
>