[llvm-commits] [llvm] r73291 - in /llvm/trunk: lib/CodeGen/LLVMTargetMachine.cpp lib/Target/ARM/ARM.h lib/Target/ARM/ARMLoadStoreOptimizer.cpp lib/Target/ARM/ARMTargetMachine.cpp lib/Target/ARM/ARMTargetMachine.h test/CodeGen/ARM/stm.ll

Evan Cheng evan.cheng at apple.com
Mon Jul 13 16:26:14 PDT 2009


On Jul 12, 2009, at 8:07 AM, Jakob Stoklund Olesen wrote:

> Sorry I am so late in commenting.
> Please see comments below.
>
>>
>> =
>> =
>> =
>> =====================================================================
>> --- llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp (original)
>> +++ llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp Sat Jun 13 04:12:55
>> 2009
>> @@ -240,7 +240,7 @@
>>  if (OptLevel != CodeGenOpt::None) {
>>    PM.add(createMachineLICMPass());
>>    PM.add(createMachineSinkingPass());
>> -    printAndVerify(PM, /* allowDoubleDefs= */ true);
>> +    printAndVerify(PM, /* allowDoubleDefs= */ false);
>>  }
>
> Was this change intended? It seems unrelated to the others.
>
> The intention is that the machine code verifier allows double
> definitions until LiveVariables has run, inserting <kill> in all the
> right places. I don't think MachineLICM or MachineSinking require the
> LiveVariables analysis, so there could still legally be double
> definitions in the code at this point.

It's an accidental commit. I'll fix.

Evan

>
> Regards,
> /jakob
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits




More information about the llvm-commits mailing list