Math Forum Discussions

Steven Lord

Posts: 17,627
Registered: 12/7/04

Re: Reducing memory requirements for tiled matrices
Posted: Nov 14, 2012 9:53 AM

Plain Text

"Harry Commin" <hmc05_remove_this_@ic.ac.uk> wrote in message
news:k8010m$98f$1@newscl01ah.mathworks.com...
>> The following code uses the for-loop method I mentioned earlier ... it
>> will save you memory space, but at the cost of slower execution than with
>> the use of the 'kron' function.
>
> Yes, using multiple loops was actually my starting point, but then I
> removed the loops in an attempt to speed up execution. This worked pretty
> well for modestly-sized matrices (allowing execution to finish within a
> day or two). However, I've hit a wall with this slightly larger data size
> in that I now run out of memory.
>
> I suppose a brute-force approach would be to get my hands on a 64-bit
> architecture. Or I could try keeping "some" loops and not others... but
> it's difficult to identify a good trade-off.

Those are two options, but first I recommend that you answer Roger's "real
question":

"Roger Stafford" wrote:
> > Perhaps the real question to ask is what you intend to do with these
> > final matrices if you have a number of them to generate. You won't have
> > enough memory to store more than a very few at any one time. Perhaps
> > there is some shortcut in achieving this purpose that can be
> > accomplished without creating these huge matrices.

This group has readers and posters with experience in many different
application areas. One of them may recognize your description of your goal
as a common problem in their application area and suggest an algorithm or
function that avoids having to create such matrices altogether.

If not, switching to a 64-bit OS and 64-bit MATLAB or even to a parallel
computing cluster using Parallel Computing Toolbox and MATLAB Distributed
Computing Server are other potential options.

--
Steve Lord
slord@mathworks.com
To contact Technical Support use the Contact Us link on
http://www.mathworks.com