[llvm-commits] [llvm] r130551 - in /llvm/trunk: include/llvm/Transforms/Utils/BasicBlockUtils.h lib/Transforms/Utils/BasicBlockUtils.cpp lib/Transforms/Utils/SSAUpdater.cpp

[Devang Patel]

On May 2, 2011, at 2:44 PM, Evan Cheng wrote:

> Hi Devang,
> 
> I share Duncan's concern. This may impact compile time. Is the scanning the only solution?

I understand the concern. I have not come up with a better alternative. But for now, I'll update it to use first instruction's location.

-
Devang

> 
> Evan
> 
> On Apr 30, 2011, at 5:23 AM, Duncan Sands wrote:
> 
>> Hi Devang,
>> 
>>> +/// GetFirstDebugLocInBasicBlock - Return first valid DebugLoc entry in a
>>> +/// given basic block.
>>> +DebugLoc llvm::GetFirstDebugLocInBasicBlock(const BasicBlock *BB) {
>>> +  for (BasicBlock::const_iterator BI = BB->begin(), BE = BB->end();
>>> +       BI != BE; ++BI) {
>>> +    DebugLoc DL = BI->getDebugLoc();
>>> +    if (!DL.isUnknown())
>>> +      return DL;
>>> +  }
>>> +  return DebugLoc();
>>> +}
>> 
>> this potentially scans through the entire basic block which might be expensive
>> if it is big.  Also, it's not clear to me that this makes any sense.  Suppose
>> the first instructions in the basic block have no line number info.  Why should
>> the phi node get the same line number as some random instruction deep down in
>> the basic block?  How about just giving the phi node the same line info as the
>> first instruction in the basic block?  If it doesn't have any then too bad.
>> 
>> Ciao, Duncan.
>> _______________________________________________
>> llvm-commits mailing list
>> llvm-commits at cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>