Multiply the low signed 32-bit integers from each packed 64-bit element in a and b, and return the signed 64-bit results.