Multiply the packed 16-bit integers in a and b, producing intermediate 32-bit integers, and return the low 16 bits of the intermediate integers.