Prev: Pixelated postscript text
Next: Set same variables and/or constants for MANY SUBSYSTEMS
From: Claire Lee on 10 Feb 2010 16:23
I encounter this situation very often. You thought your code doesn't use all the memory, but actually it does. When an LU factorization is almost finished, a spike in memory usage will occur. This is because UMFPACK and MATLAB have different data structures. It takes memory to create L and U matrices in MATLAB from UMFPACK's LU factors. In my experience, this operation can double the memory usage. For example, you may need about 8 GB memory even though UMFPACK only consumes 4 GB. The spike in memory usage usually lasts for a short period. Therefore you probably didn't notice that.

In my opinion, this is a SERIOUS problem that limits MATLAB's application in large-scale technical computing involving differential equations.



"Cristina " <cri21(a)libero.it> wrote in message <he4bl3$8es$1(a)fred.mathworks.com>...
> Thank you for your reply. I have a 64-bit operating system (ubuntu). When I run the matlab code it takes something like 4 GB of memory but I have much more memory than that. Is there a way to allow matlab to use more memory?
>
>
> "Matt " <xys(a)whatever.com> wrote in message <he423k$dis$1(a)fred.mathworks.com>...
> > "Cristina " <cri21(a)libero.it> wrote in message <hdveej$19c$1(a)fred.mathworks.com>...
> >
> > > I have seen some posts that were associating this problem to lack of memory. This is not my case since I have enough memory and matlab is only using a fraction of it.
> > =====
> >
> > That doesn't rule this out as a problem. MATLAB is only using a fraction of your memory, because it is only allowed to use a fraction of it (2GB under 32-bit Windows and only a portion of this is available to the MATLAB workspace). How much memory does the matrix consume?
 | 
Pages: 1
Prev: Pixelated postscript text
Next: Set same variables and/or constants for MANY SUBSYSTEMS