[llvm] r274786 - Include SelectionDAGISel in the opt-bisect process

Michael Kuperstein via llvm-commits llvm-commits at lists.llvm.org
Thu Jul 7 16:37:37 PDT 2016


Andrew, I think this broke the PPC bots. Could you please take a look?

Thanks,
  Michael

On Thu, Jul 7, 2016 at 11:55 AM, Andrew Kaylor via llvm-commits <
llvm-commits at lists.llvm.org> wrote:

> Author: akaylor
> Date: Thu Jul  7 13:55:02 2016
> New Revision: 274786
>
> URL: http://llvm.org/viewvc/llvm-project?rev=274786&view=rev
> Log:
> Include SelectionDAGISel in the opt-bisect process
>
> Differential Revision: http://reviews.llvm.org/D21143
>
>
> Added:
>     llvm/trunk/test/Other/X86/opt-bisect-isel.ll
> Modified:
>     llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp
>
> Modified: llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp
> URL:
> http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp?rev=274786&r1=274785&r2=274786&view=diff
>
> ==============================================================================
> --- llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp (original)
> +++ llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp Thu Jul  7
> 13:55:02 2016
> @@ -443,7 +443,7 @@ bool SelectionDAGISel::runOnMachineFunct
>    TM.resetTargetOptions(Fn);
>    // Reset OptLevel to None for optnone functions.
>    CodeGenOpt::Level NewOptLevel = OptLevel;
> -  if (Fn.hasFnAttribute(Attribute::OptimizeNone))
> +  if (OptLevel != CodeGenOpt::None && skipFunction(Fn))
>      NewOptLevel = CodeGenOpt::None;
>    OptLevelChanger OLC(*this, NewOptLevel);
>
>
> Added: llvm/trunk/test/Other/X86/opt-bisect-isel.ll
> URL:
> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Other/X86/opt-bisect-isel.ll?rev=274786&view=auto
>
> ==============================================================================
> --- llvm/trunk/test/Other/X86/opt-bisect-isel.ll (added)
> +++ llvm/trunk/test/Other/X86/opt-bisect-isel.ll Thu Jul  7 13:55:02 2016
> @@ -0,0 +1,22 @@
> +; This test verifies that no optimizations are performed on the @f
> function
> +; when the -opt-bisect-limit=0 option is used.  In particular, the X86
> +; instruction selector will optimize the cmp instruction to a sub
> instruction
> +; if it is not run in -O0 mode.
> +
> +; RUN: llc -O3 -opt-bisect-limit=0 -o - %s | FileCheck %s
> +
> +target triple = "x86_64-unknown-linux-gnu"
> +
> +define void @f() {
> +entry:
> +  %cmp = icmp slt i32 undef, 8
> +  br i1 %cmp, label %middle, label %end
> +
> +middle:
> +  br label %end
> +
> +end:
> +  ret void
> +}
> +
> +; CHECK: cmpl $8, %eax
>
>
> _______________________________________________
> 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/20160707/0c248fc2/attachment.html>


More information about the llvm-commits mailing list