High performance Python: Reference

Cython can speed up some computations dramatically

Cython uses the pyx extension and distutils to compile code to C

Another simpler mechanism is provided by pyximport

Additional performance gains can be won by using cdef and cpdef objects

Cython provides an annotation system to diagnose the level of interaction with Python

Improvements can often be made with this information

Additional performance gains can be made with simple compiler instructions

Python can be slow in some conditions

Cython and Numba can give substantial performance boosts in some of these

Use them judiciously, to make your code faster

If computations lend themselves to parallelization, they can be sped up substantially

Dask arrays can speed up things even more, for array data

Dask provides a whole universe of tools for parallelization

Python can be slow in some conditions

Cython an Numba can give substantial performance boosts in some of these

There are use-cases in which one of these is better than the other

Use them judiciously, to make your code faster