[Bug middle-end/128307] New: wrong code with -fopenmp: omp for schedule(simd:static, 1) runs only the first iteration of each chunk

[kyuwoncho18 at gmail dot com via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=128307

            Bug ID: 128307
           Summary: wrong code with -fopenmp: omp for
                    schedule(simd:static, 1) runs only the first iteration
                    of each chunk
           Product: gcc
           Version: 17.0
            Status: UNCONFIRMED
          Keywords: openmp, wrong-code
          Severity: normal
          Priority: P3
         Component: middle-end
          Assignee: unassigned at gcc dot gnu.org
          Reporter: kyuwoncho18 at gmail dot com
  Target Milestone: ---
            Target: x86_64-pc-linux-gnu

#include <stdio.h>

int main(void)
{
  int a[8] = { 0 };
#pragma omp parallel num_threads(1)
#pragma omp for schedule(simd:static, 1)
  for (int i = 0; i < 8; ++i)
    a[i] = 1;
  int sum = 0;
  for (int i = 0; i < 8; ++i)
    sum += a[i];
  printf("%d\n", sum);
  return 0;
}

godbolt: https://godbolt.org/z/xzG77Yzvf

Each of the 8 iterations stores 1, so the program must print 8, as it does at
-O0 -fopenmp and with -O2 -fopenmp -fno-tree-loop-vectorize. With -fopenmp at
-O1, -O2, -O3 and -Os it prints 1; with schedule(simd:static, 2) or
schedule(static, 1) it prints 8.