[llvm-bugs] [Bug 27696] New: phi node arguments oscillate with passes loop-vectorize and simplifycfg

via llvm-bugs llvm-bugs at lists.llvm.org
Mon May 9 19:31:06 PDT 2016


https://llvm.org/bugs/show_bug.cgi?id=27696

            Bug ID: 27696
           Summary: phi node arguments oscillate with passes
                    loop-vectorize and simplifycfg
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: Linux
            Status: NEW
          Severity: normal
          Priority: P
         Component: Scalar Optimizations
          Assignee: unassignedbugs at nondot.org
          Reporter: nwilson20 at gmail.com
                CC: llvm-bugs at lists.llvm.org
    Classification: Unclassified

Created attachment 16345
  --> https://llvm.org/bugs/attachment.cgi?id=16345&action=edit
loop-vectorize_simplifycfg.ll

When using the flags -loop-vectorize and -simplifycfg on the attached IR
(loop-vectorize_simplifycfg.ll) the phi node arguments will keep flipping. This
was produced by using the attached IR and using the output as input back to
opt.

For example:
$ ./opt -S -loop-vectorize -simplifycfg loop-vectorize_simplifycfg.ll > out1.ll
$ ./opt -S -loop-vectorize -simplifycfg out1.ll > out2.ll
$ ./opt -S -loop-vectorize -simplifycfg out2.ll > out3.ll

$ diff out2.ll out1.ll
1c1
< ; ModuleID = 'out1.ll'
---
> ; ModuleID = 'loop-vectorize_simplifycfg.ll'
20c20
<   %2 = phi i64 [ %__holeIndex, %entry ], [ 0, %while.body.i ], [ undef,
%land.rhs.i ]
---
>   %2 = phi i64 [ %__holeIndex, %entry ], [ undef, %land.rhs.i ], [ 0, %while.body.i ]


$ diff out3.ll out2.ll
1c1
< ; ModuleID = 'out2.ll'
---
> ; ModuleID = 'out1.ll'
20c20
<   %2 = phi i64 [ %__holeIndex, %entry ], [ undef, %land.rhs.i ], [ 0,
%while.body.i ]
---
>   %2 = phi i64 [ %__holeIndex, %entry ], [ 0, %while.body.i ], [ undef, %land.rhs.i ]

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20160510/1b02354e/attachment.html>


More information about the llvm-bugs mailing list