Some things were not tested with all available vectorizations. maskStore was previously untested and only worked with AVX512 float/double and AVX double. maskLoad was clearly broken, unused, and is a pretty useless instruction anyway, so I removed it.