Debugging...

I will probably find why that happens tomorrow, but here it is:

(gdb) n
631 __m256i C = _mm256_mul_epu32(A, B);
(gdb) p/x A.m256i_i32
$1 = {0x1,  0x1, 0x1, 0x1,  0x1, 0x1, 0x1, 0x1}
(gdb) p/x B.m256i_i32
$2 = {0xff, 0x0, 0x0, 0xff, 0x0, 0x0, 0x0, 0xff}
(gdb) n
(gdb) p/x C.m256i_i32
$3 = {0xff, 0x0, 0x0, 0x0,  0x0, 0x0, 0x0,  0x0}

WTF!!!!!

While posting, I RTFM'd, and found what is wrong. Very counter-intuitive, but above is expected!