pub fn _mm_cvtsh_u32(a: __m128h) -> u32
Convert the lower half-precision (16-bit) floating-point element in a to a 32-bit unsigned integer, and store the result in dst.
Intel’s documentation