[llvm-dev] analyzePhysReg question
JF Bastien via llvm-dev
llvm-dev at lists.llvm.org
Fri Dec 4 14:32:43 PST 2015
On Fri, Dec 4, 2015 at 2:22 PM, Smith, Kevin B <kevin.b.smith at intel.com>
> Ø It would be good to check that this maps correctly onto
> computeRegisterLiveness: there's a bug in analyzePhysReg and I think other
> parts of the code base are slightly wrong or will become slightly wrong as
> well :-(
> Yes, I agree. I will also have to look into all other users of
> analyzePhysReg as well. There are surprisingly few users of either
> or analyzePhysReg. The other thing that I am trying to think about would
> be a way to comprehensively test the correctness of these, both as they
> and after changes.
Great, thanks! I'd be happy to review this, I just don't have much time to
tackle the problem myself.
This code should also be fixed at the same time:
I am curious, does anyone know how do get this kind of Machine IR to be
> created from LLVM:
> xor eax, eax
> movw ax, mem
> read eax
> where the actual intent is for the xor, and the sub-register def in the
> movw to create form of zero-extension, and then to have full uses of eax? I
> know this isn't very desirable
> code for X86, but these kinds of unusual machine IR are what would stress
> the implementations of these routines.
I'm not sure, sorry. You could play tricks with asm and clobbers?
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the llvm-dev