On Feb 17, 8:59 am, Johan Oudinet <johan.oudi...@gmail.com> wrote:
> Hi,
Hi Johan,
> When I using the following simple script to get a square dxd
> inversible matrix (T) from a dxr matrix (T0), I got a memory overflow:
> #######################
> T=T0;rt=r;d=A.ncols();i=0
> while rt != d:
> while rt == rank(T.augment(matrix(d,1,{(i,0):1}))):
> i+=1
> T=T.augment(matrix(d,1,{(i,0):1}))
> rt+=1
> #######################
>
> Since the matrix A is a dxd - with d equals to 1183 - that easily fits
> into the memory (4GB), it seems strange to me that this script needs
> more and more memory until reach a memory overflow. Maybe there is a
> memory leak in the function rank or augment? Or, more likely, I did
> something wrong when writing this script (since I'm a beginner in both
> Sage and Python)?
>
> The entire script (with the 1183x1183 matrix) is available
> here:http://www.lri.fr/~oudinet/pub/script.sage
What Sage release are you running? It sounds very much like this could
be a memory leak.
The matrix is sparse and pre Sage 3.3 rank was computed using pari
which tends to blow up since it uses *a lot* of memory. We now switch
back to computing the rank of such a matrix by using the much faster
dense representation, but John Palmieri as the author of that code
should fill you in on the details there.
> I'd appreciate any help.
I am running the computation right now on a box with 128 GB, so we
will see how far I get :) Right now we are in the 30th iteration of
while rt != d:
while rt == rank(T.augment(matrix(d,1,{(i,0):1}))):
i+=1
T=T.augment(matrix(d,1,{(i,0):1}))
rt+=1
and we are already consuming about 2.5GB RAM. There are some known
problem with LA in Sage that are leaky, but I suspect those are
reference count issues in Cython. Cython 0.11 out soon should help
there with the new reference count nanny.
> Regards,
Cheers,
Michael
--~--~---------~--~----~------------~-------~--~----~
To post to this group, send email to sage-support@googlegroups.com
To unsubscribe from this group, send email to
sage-support-unsubscr...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/sage-support
URLs: http://www.sagemath.org
-~----------~----~----~----~------~----~------~--~---
