[PATCH] D11393: [X86] Allow X86::COND_NE_OR_P and X86::COND_NP_OR_E to be reversed.

[Cong Hou via llvm-commits]

On Mon, Jan 11, 2016 at 5:35 PM, Xinliang David Li <davidxl at google.com>
wrote:

> On Mon, Jan 11, 2016 at 4:59 PM, Cong Hou <congh at google.com> wrote:
> > congh added a comment.
> >
> > In http://reviews.llvm.org/D11393#323944, @davidxl wrote:
> >
> >> Restart discussion on this thread:
> >>
> >> For the motivating example where two conditional branches have
> different targets,
> >>
> >> jne   .BB1
> >>  jnp  .BB2
> >>  .BB1:
> >>  ...
> >>  .BB2:
> >>  ...
> >>
> >> Is it possible to teach AnalyzeBranch to recognize the pattern -- with
> opcode COND_NE_OR_P ?
> >
> >
> > Yes, I think this is done in this patch. Or do I misunderstand what you
> mean?
>
> The patch removes the branch condition reversing optimization in
> AnalyzeBranch -- is that needed? For the motivating example, no new
> opcode seems to be needed either ..
>

I think branch condition reversing optimization in AnalyzeBranch is
unnecessary: we will do it in block-placement anyway. And just as you once
said, it is a bad idea to modify the branch in a function called
AnalyzeBranch(), which seems a readonly function.

Cong


>
> David
>
> >
> >
> > http://reviews.llvm.org/D11393
> >
> >
> >
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20160112/42c5c31a/attachment.html>