Re: octave memory

[jean francois sauvage]

hello Jaroslav,

You are right, the inverse of a sparse matrix is generally not sparse.
My idea is to use Cholesky factorisation. But I want to know how much
space is needed by this factorisation. I can check this on Windows
Task Manager, but without separating the contributions of the
different process.

but okay, it seems there is no built-in way to reach this information
from octave. maybe I can write an octave function interfaced with a
terminal command (like "top").


On Wed, Feb 11, 2009 at 6:44 PM, Jaroslav Hajek <address@hidden> wrote:
> On Wed, Feb 11, 2009 at 6:05 PM, jean francois sauvage
> <address@hidden> wrote:
>> hello
>>
>> I need these informations because I have to perform quite large
>> simulations, that may be limited by the total amount of free memory.
>> Basically, I want to know how much memory is used for the Cholesky
>> decomposition of a NxN matrix, compared to a basic inversion.
>>
>> I know that the inversion of a N-elements matrix needs N^(3/2) free
>> space in memory. but I'm using sparse matrix, so this rule is no more
>> pertinent.
>>
>> Jeff
>>
>
> The inverse of a sparse matrix is generally dense (unless the matrix
> is banded). Therefore, you should avoid it at nearly all costs.
> Working directly with Cholesky factorization will be better with
> probability almost 1.
>
> cheers
>
> --
> RNDr. Jaroslav Hajek
> computing expert
> Aeronautical Research and Test Institute (VZLU)
> Prague, Czech Republic
> url: www.highegg.matfyz.cz
>