[llvm] r241027 - RegisterCoalescer: Cleanup empty subranges after shrinkToUses()

[Quentin Colombet]

Ping!

On the highlighted points, look for ##

> On Jul 1, 2015, at 4:04 PM, Quentin Colombet <qcolombet at apple.com> wrote:
> 
>> 
>> On Jul 1, 2015, at 3:36 PM, Matthias Braun <matze at braunis.de <mailto:matze at braunis.de>> wrote:
>> 
>> 
>>> On Jul 1, 2015, at 3:12 PM, Quentin Colombet <qcolombet at apple.com <mailto:qcolombet at apple.com>> wrote:
>>> 
>>> Hi Matthias,
>>> 
>>> Is there a way we can do that as part of shrinkToUses?
>>> My concern is that future users of that API may have similar problems and would need a similar fix.
>> The problem is that these shrinkToUses variants work on a single SubRange, but in the context of a subrange you can't just remove yourself from the list (or you'll very likely invalidate the iterators of people calling the function).
> 
> That’s right. I was thinking that we could return a boolean in that case, but that would be inconsistent with the meaning of the boolean for the non-specialized version of shrinkToUses.
> 
> I am wondering if we should return via an output parameter (like the dead definition) the empty subrange.

## Anyway, we could add a comment on the method to point out that it might generate empty subranges and that those must be cleaned up.

> 
>> I can change some verifiers/check functions to catch those cases earlier though.
> 

## A check in the verifier would be nice indeed!

Thanks,
-Quentin
> 
>> It's probably also possible to loosen the requirements and allow empty subranges but that needs a careful code review.
>> 
>>> 
>>> Also, could you check that we generate something meaningful, not just don’t crash (I know we already have this discussion, I still don’t like test case with FileCheck :)).
>> I'm pretty sure you meant "without" here :)
> 
> Indeed :).
> 
>> I've added some FileCheck there now, but I can't help you with the "meaningful" part because I don't know enough about AMDGPU to judge that, it's just testing for a fixed output now.
> 
> That’s good enough for me.
> 
> Thanks,
> -Quentin
>> 
>> - Matthias
>> 
>>> 
>>> Thanks,
>>> -Quentin
>>> 
>>>> On Jun 29, 2015, at 5:33 PM, Matthias Braun <matze at braunis.de> wrote:
>>>> 
>>>> Author: matze
>>>> Date: Mon Jun 29 19:33:44 2015
>>>> New Revision: 241027
>>>> 
>>>> URL: http://llvm.org/viewvc/llvm-project?rev=241027&view=rev
>>>> Log:
>>>> RegisterCoalescer: Cleanup empty subranges after shrinkToUses()
>>>> 
>>>> A call to removeEmptySubranges() is necessary after every operation that
>>>> potentially removes all segments from a subregister range; this case in
>>>> the register coalescer was missing.
>>>> 
>>>> Added:
>>>> llvm/trunk/test/CodeGen/AMDGPU/subreg-coalescer-undef-use.ll
>>>> Modified:
>>>> llvm/trunk/lib/CodeGen/RegisterCoalescer.cpp
>>>> 
>>>> Modified: llvm/trunk/lib/CodeGen/RegisterCoalescer.cpp
>>>> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/RegisterCoalescer.cpp?rev=241027&r1=241026&r2=241027&view=diff
>>>> ==============================================================================
>>>> --- llvm/trunk/lib/CodeGen/RegisterCoalescer.cpp (original)
>>>> +++ llvm/trunk/lib/CodeGen/RegisterCoalescer.cpp Mon Jun 29 19:33:44 2015
>>>> @@ -1449,6 +1449,7 @@ bool RegisterCoalescer::joinCopy(Machine
>>>>                 << format("%04X", S.LaneMask) << ")\n");
>>>>    LIS->shrinkToUses(S, LI.reg);
>>>>  }
>>>> +    LI.removeEmptySubRanges();
>>>> }
>>>> if (ShrinkMainRange) {
>>>>  LiveInterval &LI = LIS->getInterval(CP.getDstReg());
>>>> 
>>>> Added: llvm/trunk/test/CodeGen/AMDGPU/subreg-coalescer-undef-use.ll
>>>> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/AMDGPU/subreg-coalescer-undef-use.ll?rev=241027&view=auto
>>>> ==============================================================================
>>>> --- llvm/trunk/test/CodeGen/AMDGPU/subreg-coalescer-undef-use.ll (added)
>>>> +++ llvm/trunk/test/CodeGen/AMDGPU/subreg-coalescer-undef-use.ll Mon Jun 29 19:33:44 2015
>>>> @@ -0,0 +1,20 @@
>>>> +; RUN: llc -march=amdgcn -mcpu=SI -o /dev/null %s
>>>> +; Don't crash when the use of an undefined value is only detected by the
>>>> +; register coalescer because it is hidden with subregister insert/extract.
>>>> +target triple="amdgcn--"
>>>> +
>>>> +define void @foobar(float %a0, float %a1, float addrspace(1)* %out) nounwind {
>>>> +entry:
>>>> +  %v0 = insertelement <4 x float> undef, float %a0, i32 0
>>>> +  br i1 undef, label %ift, label %ife
>>>> +
>>>> +ift:
>>>> +  %v1 = insertelement <4 x float> undef, float %a1, i32 0
>>>> +  br label %ife
>>>> +
>>>> +ife:
>>>> +  %val = phi <4 x float> [ %v1, %ift ], [ %v0, %entry ]
>>>> +  %v2 = extractelement <4 x float> %val, i32 1
>>>> +  store float %v2, float addrspace(1)* %out, align 4
>>>> +  ret void
>>>> +}
>>>> 
>>>> 
>>>> _______________________________________________
>>>> llvm-commits mailing list
>>>> llvm-commits at cs.uiuc.edu
>>>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>>> 
>>> 
>>> _______________________________________________
>>> llvm-commits mailing list
>>> llvm-commits at cs.uiuc.edu
>>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>> 
> 
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu <mailto:llvm-commits at cs.uiuc.edu>
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits <http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20150716/2cc149fa/attachment.html>