Translate python hamming distance algorithm code to low level C using popcount register to get 1000x+ speed improvement

We are large sets of fingerprints that we are performing hamming distance calculations on (xor and sum bits). The python implementation using numpy is not fast enough. Each CPU has a dedicated register called popcount that performs the XOR and sum of 2 64 bit numbers in 1 CPU cycle. So if your clock speed is 1 ghz, 1 E9 hamming calc's per second in its purest form. Our fingerprints are 10 bytes each so it may need to be done in two clock cycles. In the attached example we will process two sets of data in python. You can run the code easily and it is well commented. We want the Cpython replacement that is matrix/vectorized like our example and utilizes popcount. It must be a drop in replacement. Please benchmark speeds and provide the end result speeds and samples. I have attached simple bitcount C code as an example as well (just how to use popcount). From what I can calculate we could get a 1000x improvement in speed using popcount and writing it in C.