[LLVMbugs] [Bug 23757] New: incorrect comparison result between INT_MAX and INT_MIN with optimizations
bugzilla-daemon at llvm.org
bugzilla-daemon at llvm.org
Thu Jun 4 09:08:16 PDT 2015
https://llvm.org/bugs/show_bug.cgi?id=23757
Bug ID: 23757
Summary: incorrect comparison result between INT_MAX and
INT_MIN with optimizations
Product: libraries
Version: trunk
Hardware: PC
OS: Linux
Status: NEW
Severity: normal
Priority: P
Component: Scalar Optimizations
Assignee: unassignedbugs at nondot.org
Reporter: todd at cloudera.com
CC: llvmbugs at cs.uiuc.edu
Classification: Unclassified
The following program gives different results with -O1 vs -O0:
#define __STDC_LIMIT_MACROS
#include <stdint.h>
#include <iostream>
#include <string.h>
using namespace std;
bool Increment32(int32_t* cell_ptr) {
int32_t orig = *cell_ptr;
int32_t inc;
// Signed overflow is undefined in C. So, we'll use a branch here
// instead of counting on undefined behavior.
if (orig == INT32_MAX) {
inc = INT32_MIN;
} else {
inc = orig + 1;
}
*cell_ptr = inc;
return inc > orig;
}
int main(int argc, char** argv) {
int32_t input = INT32_MAX;
cout << "input: " << input << endl;
bool ret = Increment32(&input);
cout << "output: " << input << endl;
if (ret) {
cout << "BAD ANSWER" << endl;
}
return 0;
}
I narrowed it down to the following set of passes on llvm 3.4:
todd at todd-ThinkPad-T540p:/tmp$ $I/clang++ -O0 -c -emit-llvm test.cc
todd at todd-ThinkPad-T540p:/tmp$ $I/opt -sroa -simplifycfg -instcombine -o
test-opt.bc test.bc ; $I/clang++ -o test test-opt.bc ; ./test
input: 2147483647
output: -2147483648
WRONG ANSWER
I also verified that a 3.7 build (from the chromium toolchain) has the same
bug.
Compiling with -fwrapv or -fsanitize=undefined makes the error go away. gcc
doesn't have the same issue.
--
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/20150604/9667738e/attachment.html>
More information about the llvm-bugs
mailing list