How to reset memory limits

asked 2012-11-21 04:29:17 +0100

GaryMak gravatar image

Hi everyone: I am trying to compute the determinant of an 11-by-11 complex matrix on SAGE v5.4 notebook() on a 2010 Mac 4G RAM etc running OS X v10.6.8. I would have thought that that would not be a big deal, but after 30 seconds or so I get a screed of traceback stuff ending with:

"... RuntimeError: ECL says: Memory limit reached. Please jump to an outer pointer, quit program and enlarge the memory limits before executing the program again."

There are no other programs open on the Mac, and the only other things in SAGE memory are 10 other matrices of the same size. In addition I have been having similarly limiting behaviour in all my matrix calcs, also on an HP laptop (using Sage v5.4 on Virtual Machine etc).

Could someone please tell me where I might be going wrong?

Many thanks

edit retag flag offensive close merge delete

Comments

Will a numerical answer suffice, or is your complex matrix symbolic?

DSM gravatar imageDSM ( 2012-11-21 07:39:32 +0100 )edit

Good point - I am actually using things of the form 2pi/3 etc so I don't know what that prompts internally ...

GaryMak gravatar imageGaryMak ( 2012-11-21 07:56:55 +0100 )edit

According to this (http://trac.sagemath.org/sage_trac/ticket/6772) there shouldn't be an memory limit, besides the physical one. Back of the envelope suggests that you'll have on the order of 40 million terms there.

benjaminfjones gravatar imagebenjaminfjones ( 2012-11-21 09:12:56 +0100 )edit

yes but presumably the algorithm does sums within the complex numbers as it goes (ie there is only one or a few terms in a sense at any given time) - because notwithstanding the fact I have kept pi symbolic it nevertheless is within the ring of matrices over the complex numbers .... or am I missing the point?

GaryMak gravatar imageGaryMak ( 2012-11-21 09:32:24 +0100 )edit
2

No, if you actually have a symbolic `pi` then you should not expect this to happen. Maxima also does not expand automatically. So I guess this is not that surprising. Maybe you should just make the whole matrix over CDF?

kcrisman gravatar imagekcrisman ( 2012-11-21 10:10:25 +0100 )edit