[llvm] r258847 - Allow X86::COND_NE_OR_P and X86::COND_NP_OR_E to be reversed.

[Xinliang David Li via llvm-commits]

Do you have a test case?

David

On Tue, Jan 26, 2016 at 8:11 PM, James Knight via llvm-commits <
llvm-commits at lists.llvm.org> wrote:

> This change is broken in some fashion; it causes llvm to rewrite branches
> incorrectly producing an infinite loop.
>
> > On Jan 26, 2016, at 3:08 PM, Cong Hou via llvm-commits <
> llvm-commits at lists.llvm.org> wrote:
> >
> > Author: conghou
> > Date: Tue Jan 26 14:08:01 2016
> > New Revision: 258847
> >
> > URL: http://llvm.org/viewvc/llvm-project?rev=258847&view=rev
> > Log:
> > Allow X86::COND_NE_OR_P and X86::COND_NP_OR_E to be reversed.
> >
> > Currently, AnalyzeBranch() fails non-equality comparison between
> floating points
> > on X86 (see https://llvm.org/bugs/show_bug.cgi?id=23875). This is
> because this
> > function can modify the branch by reversing the conditional jump and
> removing
> > unconditional jump if there is a proper fall-through. However, in the
> case of
> > non-equality comparison between floating points, this can turn the branch
> > "unanalyzable". Consider the following case:
> >
> > jne.BB1
> > jp.BB1
> > jmp.BB2
> > .BB1:
> > ...
> > .BB2:
> > ...
> >
> > AnalyzeBranch() will reverse "jp .BB1" to "jnp .BB2" and then "jmp .BB2"
> will be
> > removed:
> >
> > jne.BB1
> > jnp.BB2
> > .BB1:
> > ...
> > .BB2:
> > ...
> >
> > However, AnalyzeBranch() cannot analyze this branch anymore as there are
> two
> > conditional jumps with different targets. This may disable some
> optimizations
> > like block-placement: in this case the fall-through behavior is enforced
> even if
> > the fall-through block is very cold, which is suboptimal.
> >
> > Actually this optimization is also done in block-placement pass, which
> means we
> > can remove this optimization from AnalyzeBranch(). However, currently
> > X86::COND_NE_OR_P and X86::COND_NP_OR_E are not reversible: there is no
> defined
> > negation conditions for them.
> >
> > In order to reverse them, this patch defines two new CondCode
> X86::COND_E_AND_NP
> > and X86::COND_P_AND_NE. It also defines how to synthesize instructions
> for them.
> > Here only the second conditional jump is reversed. This is valid as we
> only need
> > them to do this "unconditional jump removal" optimization.
> >
> >
> > Differential Revision: http://reviews.llvm.org/D11393
> >
> >
> >
> > Modified:
> >    llvm/trunk/lib/Target/X86/X86InstrInfo.cpp
> >    llvm/trunk/lib/Target/X86/X86InstrInfo.h
> >    llvm/trunk/test/CodeGen/X86/block-placement.ll
> >    llvm/trunk/test/CodeGen/X86/fast-isel-cmp-branch2.ll
> >    llvm/trunk/test/CodeGen/X86/fast-isel-cmp-branch3.ll
> >    llvm/trunk/test/CodeGen/X86/fp-une-cmp.ll
> >
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20160126/7a468201/attachment.html>