it seems that there is SOME support for bitwise operations in NumPy:
https://docs.scipy.org/doc/numpy/reference/generated/numpy.bitwise_xor.html
Bitwise operations are also supported by CUDA:
https://docs.nvidia.com/cuda/cuda-c-programming-guide/#warp-shuffle-functions
yet, the implementation of binary operation xor in PyTorch is
return (self + other).eq(1)
which takes two ByteTensor. My undertsanding is that this is inefficient, because only one Bit within every Byte of this tensor is actually encoding something. Please correct me if I'm wrong!!

for the record neither TensorFlow has bitwise operations