#endif
constexpr IndexType InputSimdWidth = sizeof(vec_t) / sizeof(std::int32_t);
// Inputs are processed InputSimdWidth at a time and outputs are processed 8 at a time so we process in chunks of max(InputSimdWidth, 8)
#endif
constexpr IndexType InputSimdWidth = sizeof(vec_t) / sizeof(std::int32_t);
// Inputs are processed InputSimdWidth at a time and outputs are processed 8 at a time so we process in chunks of max(InputSimdWidth, 8)