[llvm-commits] [llvm] r149070 - in /llvm/trunk: lib/CodeGen/MachineBasicBlock.cpp test/CodeGen/ARM/tail-dup.ll

Chad Rosier mcrosier at apple.com
Thu Jan 26 13:58:51 PST 2012


On Jan 26, 2012, at 1:54 PM, Jakob Stoklund Olesen wrote:

> 
> On Jan 26, 2012, at 10:24 AM, Chad Rosier wrote:
> 
>> --- llvm/trunk/lib/CodeGen/MachineBasicBlock.cpp (original)
>> +++ llvm/trunk/lib/CodeGen/MachineBasicBlock.cpp Thu Jan 26 12:24:25 2012
>> @@ -535,13 +535,12 @@
>>  if (TII->AnalyzeBranch(*this, TBB, FBB, Cond)) {
>>    // If we couldn't analyze the branch, examine the last instruction.
>>    // If the block doesn't end in a known control barrier, assume fallthrough
>> -    // is possible. The isPredicable check is needed because this code can be
>> +    // is possible. The isPredicated check is needed because this code can be
>>    // called during IfConversion, where an instruction which is normally a
>>    // Barrier is predicated and thus no longer an actual control barrier. This
>>    // is over-conservative though, because if an instruction isn't actually
>>    // predicated we could still treat it like a barrier.
>> -    return empty() || !back().isBarrier() ||
>> -           back().isPredicable();
>> +    return empty() || !back().isBarrier() || TII->isPredicated(&back());
>>  }
> 
> Is the comment still correct now? Is the check still over-conservative?

No, I don't believe we're still being conservative.  I've already update the comment in r149075.  :)

 Chad

> /jakob
> 




More information about the llvm-commits mailing list