Re: possible memory leak in vintrp?

Hi Mary,

I would suggest using f2py (part of the numpy/scipy distribution) for
wrapping fortran, works very well in my experience

Rodrigo

On 8 Sep 2009, at 13:29, Mary Haley wrote:

> Hi Jim,
>
> It's been on our list to look into this memory leak, which was
> reported before.
>
> I believe the problem is that "swig" was used to wrap the Fortran
> code, and this code doesn't do any kind of memory management that I
> can tell.
>
> I'll check with the original developer and see if I can get some
> more information for a fix.
>
> --Mary
>
> On Fri, 4 Sep 2009, James Boyle wrote:
>
>> I am putting a lot of high frequency CAM hybrid data on pressure
>> levels using Ngl.vinth2p.
>> I have used this routine many times before.
>> But this time , the number of calls per day exceeds 1500.
>> One day goes OK, the data look reasonable but on the second day I get
>> the error that is at the end of this message.
>> It appears that Ngl has a memory leak that is exposed when the
>> routine
>> is called this many times.
>>
>> I am on OS X 10.5.8, using pyngl 1.3.0b1, python 2.6.1, numpy 1.3.
>>
>> python(98559) malloc: *** mmap(size=36864) failed (error code=12)
>> *** error: can't allocate region
>> *** set a breakpoint in malloc_error_break to debug
>> Bus error
>>
>> --Jim Boyle
>> _______________________________________________
>> pyngl-talk mailing list
>> List instructions, subscriber options, unsubscribe:
>> http://mailman.ucar.edu/mailman/listinfo/pyngl-talk
>>
> _______________________________________________
> pyngl-talk mailing list
> List instructions, subscriber options, unsubscribe:
> http://mailman.ucar.edu/mailman/listinfo/pyngl-talk

_______________________________________________
pyngl-talk mailing list
List instructions, subscriber options, unsubscribe:
http://mailman.ucar.edu/mailman/listinfo/pyngl-talk
Received on Tue Sep 08 2009 - 06:35:09 MDT