[llvm-bugs] [Bug 24521] New: assert in GVN::propagateEquality assert(LHS->getType() == RHS->getType() && "Equality but unequal types!")
via llvm-bugs
llvm-bugs at lists.llvm.org
Thu Aug 20 05:22:24 PDT 2015
https://llvm.org/bugs/show_bug.cgi?id=24521
Bug ID: 24521
Summary: assert in GVN::propagateEquality assert(LHS->getType()
== RHS->getType() && "Equality but unequal types!")
Product: new-bugs
Version: 3.7
Hardware: PC
OS: Windows NT
Status: NEW
Severity: normal
Priority: P
Component: new bugs
Assignee: unassignedbugs at nondot.org
Reporter: gekkekoe at gmail.com
CC: llvm-bugs at lists.llvm.org
Classification: Unclassified
When passing the following IR to the passmanager I'm getting assert failures on
some test code.
The IR works fine on 3.6.2. but with 3.7rc2 i'm getting assertion errors.
define double @bugfunc1(double %y, i64 %z) #0 {
entry:
%"1_k" = alloca i64
store i64 %z, i64* %"1_k"
br label %loopcond
loop: ; preds = %loopcond
%load.ident = load i64, i64* %"1_k"
%addIntTmp = add i64 %load.ident, 1
store i64 %addIntTmp, i64* %"1_k"
br label %loopcond
loopcond: ; preds = %loop, %entry
%load.ident1 = load i64, i64* %"1_k"
%cmpIntTmp = icmp slt i64 %load.ident1, 3
br i1 %cmpIntTmp, label %loop, label %loopexit
loopexit: ; preds = %loopcond
ret double 1.000000e+00
}
If I replace the
%cmpIntTmp = icmp slt i64 %load.ident1, 3
with a constant true, then the error goes away. It seems that something is
going wrong while trying to optimize the icmp instr.
The pass manager is initialized as:
passmanager = LLVMCreateFunctionPassManagerForModule(module);
LLVMAddTargetData(LLVMGetExecutionEngineTargetData(engine), passmanager);
LLVMAddBasicAliasAnalysisPass(passmanager);
LLVMAddPromoteMemoryToRegisterPass(passmanager);
LLVMAddInstructionCombiningPass(passmanager);
LLVMAddReassociatePass(passmanager);
LLVMAddGVNPass(passmanager);
LLVMAddCFGSimplificationPass(passmanager);
LLVMInitializeFunctionPassManager(passmanager);
I'm using visual studio 2015.
--
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/20150820/d3a9e4ad/attachment.html>
More information about the llvm-bugs
mailing list