llvm dev - Sep 2007 - [LLVMdev] Prolog/Epilog Insertion Question

I've been looking through the code for pologue/epilogoue generation and
noticed this oddity:

void PEI::replaceFrameIndices(MachineFunction &Fn) {
  [...]
    for (MachineBasicBlock::iterator I = BB->begin(); I != BB->end(); ) {
        [...]
      if (I->getOpcode() == FrameSetupOpcode ||
          I->getOpcode() == FrameDestroyOpcode) {
        [...]
      } else {
        [...]
        for (unsigned i = 0, e = MI->getNumOperands(); i != e; ++i)
          if (MI->getOperand(i).isFrameIndex()) {
            // If this instruction has a FrameIndex operand, we need to use 
            // that
            // target machine register info object to eliminate it.
            MRI.eliminateFrameIndex(MI, SPAdj, RS);

            // Revisit the instruction in full.  Some instructions (e.g. 
            // inline
            // asm instructions) can have multiple frame indices.
            --I;
            MI = 0;
            break;
          }
      }
      // Update register states.
      if (RS && MI) RS->forward(MI);
    }

The break causes the instruction to be reprocessed but forward() is called
on the register scavenger.  Doesn't this create an inconsistency?

                                               -Dave

On Sep 6, 2007, at 12:52 PM, David Greene wrote:
> I've been looking through the code for pologue/epilogoue generation  
> and
> noticed this oddity:
>
> void PEI::replaceFrameIndices(MachineFunction &Fn) {
>   [...]
>     for (MachineBasicBlock::iterator I = BB->begin(); I != BB->end 
> (); ) {
>         [...]
>       if (I->getOpcode() == FrameSetupOpcode ||
>           I->getOpcode() == FrameDestroyOpcode) {
>         [...]
>       } else {
>         [...]
>         for (unsigned i = 0, e = MI->getNumOperands(); i != e; ++i)
>           if (MI->getOperand(i).isFrameIndex()) {
>             // If this instruction has a FrameIndex operand, we  
> need to use
>             // that
>             // target machine register info object to eliminate it.
>             MRI.eliminateFrameIndex(MI, SPAdj, RS);
>
>             // Revisit the instruction in full.  Some instructions  
> (e.g.
>             // inline
>             // asm instructions) can have multiple frame indices.
>             --I;
>             MI = 0;
>             break;
>           }
>       }
>       // Update register states.
>       if (RS && MI) RS->forward(MI);
>     }
>
> The break causes the instruction to be reprocessed but forward() is  
> called
> on the register scavenger.  Doesn't this create an inconsistency?
It's ok because MI is set to NULL so RS->forward() is skipped.

Evan
>
>                                                -Dave
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev