extradoc / talk / sea2013 / abstract.rst

High performance Python

I would like to present an intermediate level talk on High Performance Python.
Python has seen quite a lot of adoption by the scientific community, however,
the main part of it has been as a glue language. Python is used a lot for
data processing, driving simulations and presenting, however the main
part of scientific processing is usually done with Fortran, C or C++.
Recent advancements in numeric tools available for Python make it feasible
to develop a prototype in Python and often even leave your entire model
in Python. I will give a brief overview on available tools, profilers and
literature how you can achieve that. This talk will briefly cover CPython,
PyPy, NumPy, Cython, Numba, Numexpr and other available performance-related

I am a core developer of the PyPy project as well as an implementor of
the numpy package for PyPy. I have extensive knowledge in profiling
python software including numeric packages. Furthermore I've implemented
various tools that simplify profiling.

The preferred talk length is one hour.