[Bug tree-optimization/113271] New: [Regression] Wrong code at -O1/2/3 since GCC-9 (could be due to wrong optimization)

["652023330028 at smail dot nju.edu.cn" <gcc-bugzilla@gcc.gnu.org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113271

            Bug ID: 113271
           Summary: [Regression] Wrong code at -O1/2/3 since GCC-9 (could
                    be due to wrong optimization)
           Product: gcc
           Version: 14.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: 652023330028 at smail dot nju.edu.cn
  Target Milestone: ---

Hello, we note that gcc may introduce erroneous code due to optimization:

https://godbolt.org/z/8vbn3KdqP

#include <iostream>
int a, b, c;
void func() {
    a=-b;
    //std::cout<<-a<<std::endl;    //Line 5
    c=(a)/(-2147483647-1);
}

int main(){
    b= -2147483647 - 1;
    func();
    std::cout<<c;
}

The output of GCC-O0 is:
1

The output of GCC-O3 is:
0

If we uncomment the fifth line:
#include <iostream>
int a, b, c;
void func() {
    a=-b;
    std::cout<<-a<<std::endl;    //Line 5
    c=(a)/(-2147483647-1);
}

int main(){
    b= -2147483647 - 1;
    func();
    std::cout<<c;
}

we get the correct output (GCC-O3):
-2147483648
1

This issue is probably due to a bad optimization of "func" by gcc -O3:
func():
        mov     DWORD PTR c[rip], 0
        mov     eax, DWORD PTR b[rip]
        neg     eax
        mov     DWORD PTR a[rip], eax
        ret

Thank you very much for your time and effort! We look forward to hearing from
you.