[llvm] r235673 - RegisterCoalescer: implicit phsreg uses are fine when rematerializing

[Quentin Colombet]

> On Apr 23, 2015, at 5:22 PM, Quentin Colombet <qcolombet at apple.com> wrote:
> 
> 
>> On Apr 23, 2015, at 5:01 PM, Matthias Braun <matze at braunis.de> wrote:
>> 
>> Author: matze
>> Date: Thu Apr 23 19:01:37 2015
>> New Revision: 235673
>> 
>> URL: http://llvm.org/viewvc/llvm-project?rev=235673&view=rev
>> Log:
>> RegisterCoalescer: implicit phsreg uses are fine when rematerializing
>> 
>> The target hooks should have already checked them. This change is
>> necessary to enable the remateriailzation on R600.
>> 
>> 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=235673&r1=235672&r2=235673&view=diff
>> ==============================================================================
>> --- llvm/trunk/lib/CodeGen/RegisterCoalescer.cpp (original)
>> +++ llvm/trunk/lib/CodeGen/RegisterCoalescer.cpp Thu Apr 23 19:01:37 2015
>> @@ -963,8 +963,8 @@ bool RegisterCoalescer::reMaterializeTri
>>  for (unsigned i = NewMI->getDesc().getNumOperands(),
>>         e = NewMI->getNumOperands(); i != e; ++i) {
>>    MachineOperand &MO = NewMI->getOperand(i);
>> -    if (MO.isReg()) {
>> -      assert(MO.isDef() && MO.isImplicit() && MO.isDead() &&
>> +    if (MO.isReg() && MO.isDef()) {
>> +      assert(MO.isImplicit() && MO.isDead() &&
>>             TargetRegisterInfo::isPhysicalRegister(MO.getReg()));
> 
> Wait!
> This is not good. The simple rematerialization shouldn’t create register uses.

For more context, here is why I explained it is wrong:
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20140421/213946.html

> 
> What do you remat here?
> 
> Q.
>>      NewMIImplDefs.push_back(MO.getReg());
>>    }
>> 
>> 
>> _______________________________________________
>> llvm-commits mailing list
>> llvm-commits at cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>