[PATCH] D14477: Remove unnecessary call to getAllocatableRegClass

[escha via llvm-commits]

Reverting breaks a bunch of AMDGPU tests; are you sure you want me to do that?

—escha

Failing Tests (12):
    LLVM :: CodeGen/AMDGPU/bswap.ll
    LLVM :: CodeGen/AMDGPU/commute-compares.ll
    LLVM :: CodeGen/AMDGPU/fconst64.ll
    LLVM :: CodeGen/AMDGPU/fcopysign.f32.ll
    LLVM :: CodeGen/AMDGPU/fcopysign.f64.ll
    LLVM :: CodeGen/AMDGPU/fneg-fabs.f64.ll
    LLVM :: CodeGen/AMDGPU/fp_to_sint.f64.ll
    LLVM :: CodeGen/AMDGPU/llvm.AMDGPU.class.ll
    LLVM :: CodeGen/AMDGPU/llvm.AMDGPU.rsq.clamped.f64.ll
    LLVM :: CodeGen/AMDGPU/llvm.round.f64.ll
    LLVM :: CodeGen/AMDGPU/llvm.round.ll
    LLVM :: CodeGen/AMDGPU/use-sgpr-multiple-times.ll

> On Nov 11, 2015, at 5:12 PM, Quentin Colombet <qcolombet at apple.com> wrote:
> 
> 
>> On Nov 11, 2015, at 5:10 PM, escha <escha at apple.com> wrote:
>> 
>> Yup, the assert does trigger.
> 
>> It’s a rather weird target, so I wouldn’t be surprised if it does nonintuitive things that don’t make any good sense.
> 
> Heh, feel free to revert, I’ll get back to you to try to understand why we end up in this situation.
> 
> Sorry for the breakage!
> 
> Q.
>> 
>> —escha
>> 
>>> On Nov 11, 2015, at 5:08 PM, Quentin Colombet <qcolombet at apple.com> wrote:
>>> 
>>> Hi Fiona,
>>> 
>>> Does the assert trigger?
>>> 
>>> Cheers,
>>> -Quentin
>>>> On Nov 11, 2015, at 4:01 PM, escha <escha at apple.com> wrote:
>>>> 
>>>> escha added a subscriber: escha.
>>>> escha added a comment.
>>>> 
>>>> This broke one of our out-of-tree targets.
>>>> 
>>>> —escha
>>>> 
>>>> 
>>>> http://reviews.llvm.org/D14477
>>>> 
>>>> 
>>>> 
>>> 
>> 
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20151112/fae11e1d/attachment.html>