Re: [Getfem-users] Python interface: memory leak?

[Yves Renard]

Dear Magnus,

A fix has been commited by Kostas. Many thanks to him.
The problem was a missing dereference for numpy arrays in Getfem interface.

Yves.


----- Mail original -----
De: "yves renard" <address@hidden>
À: "getfem-users" <address@hidden>
Envoyé: Vendredi 11 Mars 2016 13:24:09
Objet: Re: [Getfem-users] Python interface: memory leak?

Dear Magnus,

Quite strange memory leak. When I put "b= N.random.rand(50000)" out of
the loop, as this :


b= N.random.rand(50000)
for ii in range(1001):
    c=GFspmat.mult(b)
    if ii%100==0:
        print "SPmat: Mem use after %i mult %i
kb"%(ii,resource.getrusage(resource.RUSAGE_SELF).ru_maxrss)


Then there is no more leak ... May be it comes from how the python
interface deal with numpy arrays. I will have a look.


Yves.






Le 10/03/2016 13:11, Magnus Paulsson a écrit :
> Seems that there is a memory leak in SPmat.mult(). The code below
> shows increasing memory use for getfem 5.0 and 4.? on my linux boxes.
> (don’t know if the code works on OSX or windows but runs on windows)
>
> SciPy: Mem use after 0 mult 29588 kb
> SciPy: Mem use after 100 mult 29588 kb
> SciPy: Mem use after 200 mult 29588 kb
> SciPy: Mem use after 300 mult 29588 kb
> SciPy: Mem use after 400 mult 29588 kb
> SciPy: Mem use after 500 mult 29588 kb
> SciPy: Mem use after 600 mult 29588 kb
> SciPy: Mem use after 700 mult 29588 kb
> SciPy: Mem use after 800 mult 29588 kb
> SciPy: Mem use after 900 mult 29588 kb
> SciPy: Mem use after 1000 mult 29588 kb
> SPmat: Mem use after 0 mult 29588 kb
> SPmat: Mem use after 100 mult 65840 kb
> SPmat: Mem use after 200 mult 104912 kb
> SPmat: Mem use after 300 mult 143984 kb
> SPmat: Mem use after 400 mult 183056 kb
> SPmat: Mem use after 500 mult 222128 kb
> SPmat: Mem use after 600 mult 261200 kb
> SPmat: Mem use after 700 mult 300272 kb
> SPmat: Mem use after 800 mult 339344 kb
> SPmat: Mem use after 900 mult 378416 kb
> SPmat: Mem use after 1000 mult 417488 kb
>
>
>
>
> import numpy as N
> import getfem as GF
> import resource
> import scipy.sparse as SS
>
> # Random sparse diagonal 
> GFspmat = GF.Spmat('diag', N.random.rand(50000))
> GFspmat.to_csc()
>
> # Scipy version
> [indptr, indices] =  GFspmat.csc_ind()
> data = GFspmat.csc_val()
> SciPyspmat = SS.csc_matrix((data,indices,indptr))
>
> for ii in range(1001):
>     b=N.random.rand(50000)
>     c=SciPyspmat.dot(b)
>     if ii%100==0:
>         print "SciPy: Mem use after %i mult %i
> kb"%(ii,resource.getrusage(resource.RUSAGE_SELF).ru_maxrss)
>
> for ii in range(1001):
>     b= N.random.rand(50000)
>     c=GFspmat.mult(b)
>     if ii%100==0:
>         print "SPmat: Mem use after %i mult %i
> kb"%(ii,resource.getrusage(resource.RUSAGE_SELF).ru_maxrss)
>
>
>
> -----------------------------------------------
> Magnus Paulsson
> Associate Professor
> Dept. of Physics and Electrical Engineering
> Linnaeus University
> Phone: +46-480-446308
> Mobile: +46-70-6942987
>
>
>
> _______________________________________________
> Getfem-users mailing list
> address@hidden
> https://mail.gna.org/listinfo/getfem-users


-- 

  Yves Renard (address@hidden)       tel : (33) 04.72.43.87.08
  Pole de Mathematiques, INSA-Lyon             fax : (33) 04.72.43.85.29
  20, rue Albert Einstein
  69621 Villeurbanne Cedex, FRANCE
  http://math.univ-lyon1.fr/~renard

---------