Python JIT performances

Various JIT compilers available for python that I know of are:

  1. numba
  2. jax but is very specifically made to boost numpy-like operations
  3. Alternative interpreters like pypy

Performance between approaches

Note: todo Compare performance between

  1. pure CPython
  2. Cython C-extension + CPython
  3. numba + CPython
  4. pypy

Goal of the benchmarking will be to do minimal change to the original pure python function, and see the performance difference.

For reference codes:

https://www.cardinalpeak.com/blog/faster-python-with-cython-and-pypy-part-2 , code is at https://bitbucket.org/tom_craven/cython_test/src/master/?filedownload=https://bitbucket.org/tom_craven/

https://neurohackweek.github.io/cython-tutorial/04-numba/