Hi Joel, I looked at your script. MAtake was removed after version 1.1. Are you using version 1.1? It may be worth updating to version 1.2.
I ran your script against version 1.1 of fipy and didn't have any problems. The memory usage stayed fairly constant with small fluctuations. This is normal as python allocates and deallocates on the fly. My configuration is as follows. 86 (wd15) benson.nist.gov# python Python 2.3.5 (#2, Oct 16 2006, 19:19:48) [GCC 3.3.5 (Debian 1:3.3.5-13)] on linux2 Type "help", "copyright", "credits" or "license" for more information. >>> import numpy >>> numpy.__version__ '1.0.1' >>> import fipy >>> fipy.__version__ '1.1' Which version of numpy are you using? Cheers On 11/22/07, joelvuillon <[EMAIL PROTECTED]> wrote: > Hi Daniel, > > enclosed you will find a little script that describes the leak of memory. > If you try to run several time this script, you will see the level of the > available memory that tends to decrease. > > Bests, > > Joël > > > ----- Original Message ----- > From: "Daniel Wheeler" <[EMAIL PROTECTED]> > To: "Multiple recipients of list" <[email protected]> > Sent: Wednesday, November 21, 2007 9:53 PM > Subject: Re: Loss of RAM > > > > > > Hi Joel, > > > > Answers below. > > > > On Nov 21, 2007 11:55 AM, joelvuillon <[EMAIL PROTECTED]> wrote: > >> > >> > >> Hi everybody, > >> > >> Using Fipy for solving elasticity problems, I have noticed a loss of RAM > >> using fipy/tools/numerix library and, in particular, the numerix.MAtake > >> function when indices is a masked array. In addition, MA.take does not > >> work > >> in such a case (cf documentation on fipy.tools). > >> > >> Do you have already experienced such loss of RAM problem ? > > > > I haven't experienced problems with MA.take specifically. We have had > > problems in > > the past with reference counting, but we haven't experienced these > > problems for > > quite a while (at least a year) and we thought we had them ironed out. > > I am aware that > > the gist viewer leaks memory, but this is not your issue. > > > >> Is there a convenient mean to avoid such a problem while keeping indices > >> as > >> a masked array ? > > > > If it is the case that MA.take has problems, then I would like to > > understand the issue before > > suggesting alternatives. Do you have a small script that highlights > > the problem? > > > > Thanks for your feedback. > > > > -- > > Daniel Wheeler > > > > > > -- Daniel Wheeler
