[llvm] r180573 - This patch adds the X86FixupLEAs pass, which will reduce instruction
Nadav Rotem
nrotem at apple.com
Thu Apr 25 13:51:29 PDT 2013
> +
> + virtual const char *getPassName() const { return "X86 Atom LEA Fixup";}
> + void seekLEAFixup(MachineOperand& p, MachineBasicBlock::iterator& I,
> + MachineFunction::iterator MFI);
> + void processInstruction(MachineBasicBlock::iterator& I,
> + MachineFunction::iterator MFI);
> + RegUsageState usesRegister(MachineOperand& p,
> + MachineBasicBlock::iterator I);
> + MachineBasicBlock::iterator searchBackwards(MachineOperand& p,
> + MachineBasicBlock::iterator& I,
> + MachineFunction::iterator MFI);
> + MachineInstr* postRAConvertToLEA(MachineFunction::iterator &MFI,
> + MachineBasicBlock::iterator &MBBI,
> + LiveVariables *LV) const;
> +
Can you please document the declarations of these functions ?
( see http://llvm.org/docs/CodingStandards.html#doxygen-use-in-documentation-comments )
>
> +
> +/// postRAConvertToLEA - if an instruction can be converted to an
> +/// equivalent LEA, insert the new instruction into the basic block
> +/// and return a pointer to it. Otherwise, return zero.
> +MachineInstr *
> +FixupLEAPass::postRAConvertToLEA(MachineFunction::iterator &MFI,
> + MachineBasicBlock::iterator &MBBI,
> + LiveVariables *LV) const {
Preston, LiveVariables are not available in this phase and it is only confusing. Please remove them from the code.
+ LV = getAnalysisIfAvailable<LiveVariables>(); <----- this will always return 0. Did you ever get a LiveVariable instance ?
> + // Process all basic blocks.
> + for (MachineFunction::iterator I = Func.begin(), E = Func.end(); I != E; ++I)
> + processBasicBlock(Func, I);
> + DEBUG(dbgs() << "End X86FixupLEAs\n";);
> +
> + return true;
> +}
This function always returns true. You probably only want to return true of the code changed.
Can you reduce the size of the test cases ? Ideally I would like to see a single-basic block function. You only need 5-instructions or so. Why did you include such long tests ?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20130425/940925aa/attachment.html>
More information about the llvm-commits
mailing list