[llvm-bugs] [Bug 51700] New: opt -passes='loop-bound-split' failing with "PHINode should have one entry for each predecessor of its parent basic block!"
via llvm-bugs
llvm-bugs at lists.llvm.org
Wed Sep 1 00:10:43 PDT 2021
https://bugs.llvm.org/show_bug.cgi?id=51700
Bug ID: 51700
Summary: opt -passes='loop-bound-split' failing with "PHINode
should have one entry for each predecessor of its
parent basic block!"
Product: libraries
Version: trunk
Hardware: PC
OS: Linux
Status: NEW
Severity: enhancement
Priority: P
Component: Scalar Optimizations
Assignee: unassignedbugs at nondot.org
Reporter: mikael.holmen at ericsson.com
CC: llvm-bugs at lists.llvm.org
Created attachment 25210
--> https://bugs.llvm.org/attachment.cgi?id=25210&action=edit
bbi-59713.ll reproducer
llvm commit: d396d09fc2278bfe34b9db2ee27ecfd470149da8
Reproduce with:
opt -passes='loop-bound-split' -S -o - bbi-59713.ll
Result:
PHINode should have one entry for each predecessor of its parent basic block!
%call.lcssa.ph = phi i16
PHI node entries do not match predecessors!
%call.lcssa = phi i16 [ %call, %cond.end ], [ %call.lcssa.ph, %end.loopexit ]
label %cond.end
label %entry.split.split
LLVM ERROR: Broken module found, compilation aborted!
Looking att -print-after-all we see that loop-bound-split has created this
broken phi instruction:
end.loopexit: ; preds = %cond.end.split
%call.lcssa.ph = phi i16
Note:
If we add e.g. -verify-each we actually get less information, because then we
only get
LLVM ERROR: Broken module found, compilation aborted!
and not the error message about the broken PHI.
It was also a mess reducing the input because bugpoint kept crashing.
--
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/20210901/36e007cd/attachment.html>
More information about the llvm-bugs
mailing list