Why is matrix multiplication faster with numpy than with ctypes in Python? Why is matrix multiplication faster with numpy than with ctypes in Python? c c