Use _mm_testz_si128() in check_indices()

[nkurz]

Looks like a nice project!

Untested, but you could probably get a boost by using _mm_testz_si128() in check_indices(). It combines the operations you are doing with _mm_and_si128, _mm_movemask_epi8, and the comparison into a single instruction (PTEST). It requires SSE4.1, but this is pretty universally supported at this point.

Good documentation for it here: http://software.intel.com/sites/landingpage/IntrinsicsGuide/
(search for PTEST or testz)