http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58756
--- Comment #2 from Uroš Bizjak <ubizjak at gmail dot com> ---
Some unscientific printf debugging yields following runtime difference:
foo (int a, int b)
{
int j, c = 0;
#pragma omp parallel for reduction(+: c)
for (j = 0; j < a; j += b)
{
int l;
#pragma omp simd reduction(+: c)
for (l = 0; l < b; ++l)
c += d[j + l];
printf ("%i\n", c);
}
printf ("tot %i\n", c);
return c;
}
alpha:
$ ./a.out
496
496
496
496
496
496
496
496
496
496
496
496
496
496
496
496
496
496
496
496
496
496
496
496
496
496
496
496
496
496
496
496
tot 1984
Aborted
Alpha spawned 3 new threads (as reported by gdb), so 4 * 496 = 1984.
x86_64:
$ ./pr58392.exe
496
992
1488
496
992
1488
496
992
496
992
1488
496
992
1488
496
992
1488
496
992
496
992
1488
496
992
496
992
1488
496
992
496
992
1488
tot 15872
