https://gcc.gnu.org/bugzilla/show_bug.cgi?id=120457
--- Comment #2 from Hongtao Liu <liuhongt at gcc dot gnu.org> ---
(In reply to Hongtao Liu from comment #1)
> double __attribute__((noinline,noclone))
> compute_integral (double w_1[18])
> {
> double A = 0;
> double t33[2][6] = {{0.0, 0.0, 0.0, 0.0, 0.0, 0.0},
> {0.0, 0.0, 0.0, 0.0, 0.0, 0.0}};
> double t43[2] = {0.0, 0.0};
> double t31[2][2] = {{1.0, 1.0}, {1.0, 1.0}};
> double t32[2][3] = {{0.0, 0.0, 1.0}, {0.0, 0.0, 1.0}};
>
> for (int ip_1 = 0; ip_1 < 2; ++ip_1)
> {
> #pragma GCC unroll 0
> for (int i_0 = 0; i_0 < 6; ++i_0)
> t33[ip_1][i_0] = ((w_1[i_0*3] * t32[ip_1][0])
> + (w_1[i_0*3+2] * t32[ip_1][2])); --- the loop is
> not vectorized anymore for power
W/o -fno-vect-cost-model, power doesn't vectorize the loop, because it's
suboptimal.
