Pythran – a compiler for Python scientific kernels

mratsim wrote: post #3

I know Pythran from the Julia challenge [1] but you should link to some docs / examples / tutorials in this thread because the release notes doesn't do justice of what Pythran can do. [1] https://github.com/SimonDanisch/julia-challenge/pull/4

If you have some time would you mind sharing some of your thoughts about Pythran? I'm curious but wouldn't really know where to start looking.

Re: Pythran – a compiler for Python scientific kernels – release

#5

Post by mattip » Thu, Nov 07, 2019, 11:05 AM UTC

chenzhekl wrote: post #2

How does it compare to Numba and Cython?

Here is a nice comparison https://flothesof.github.io/optimizing-python-code-numpy-cyt...

Tl;dr: pythran is very similar to numba but blazingly fast on cpu

Re: Pythran – a compiler for Python scientific kernels – release

#6

Post by mlthoughts2018 » Thu, Nov 07, 2019, 11:54 AM UTC

mattip wrote: post #5

chenzhekl wrote: post #2

How does it compare to Numba and Cython?

Here is a nice comparison https://flothesof.github.io/optimizing-python-code-numpy-cyt... Tl;dr: pythran is very similar to numba but blazingly fast on cpu

The Cython example in that link is actually not a fair comparison, since it still forces the numpy ndarray type in the signature.

Instead it should use typed memoryviews [0], which are faster and can avoid more cases that will rely on the GIL accidentally (such as when an ndarray has to be treated as a Python object).

[0]: https://cython.readthedocs.io/en/latest/src/userguide/memory...

Re: Pythran – a compiler for Python scientific kernels – release

#7

Post by serge-ss-paille » Thu, Nov 07, 2019, 3:43 PM UTC

More materials for the curious:

Some benchmarks here: http://serge-sans-paille.github.io/pythran-stories/testing-p...

Some more benchmark you can run on you own: https://github.com/serge-sans-paille/numpy-benchmarks/

A comparison with Julia and native code: http://serge-sans-paille.github.io/pythran-stories/micro-ben...