[pythran] serge.guelton@xxxxxxxxxxxxxxxx: Re: mesure

  • From: serge Guelton <serge.guelton@xxxxxxxxxxxxxxxx>
  • To: pythran@xxxxxxxxxxxxx
  • Date: Mon, 17 Feb 2014 20:10:35 +0100

On Sun, Feb 16, 2014 at 06:55:02PM +0100, Pierrick Brunet wrote:
> Salut,
> 
> Voila deux code qui sont quasi les mêmes. J'ai des temps meilleur
> avec Pythran quand je découpe le calcul que quand je le regroupe.
> Vous avez pareil?
> 
> Pierrick

(answering in englisg and fwd to pythran@freelists)

Hi Pierrick,

I have tested your two codes using the fix-complex branch and the
following snippet:

>>> import numpy as np
>>> r0, r1, r2, r3 = np.random.random(1000000), np.random.random(1000000),
>>> np.random.random(1000000), np.random.random(1000000)
>>> import arc2
>>> %timeit arc2.arc_distance(r0, r1, r2, r3)
10 loops, best of 3: 147 ms per loop
>>> import arc_distance
>>> %timeit arc_distance.arc_distance(r0, r1, r2, r3)
10 loops, best of 3: 148 ms per loop

Both were compiled with pythran + g++ 4.8.2

Other related posts:

  • » [pythran] serge.guelton@xxxxxxxxxxxxxxxx: Re: mesure - serge Guelton