Stephen,
I have been faced with a similar problem. From my reading of the SPM
archives, broadly speaking, I understand that SPM will process data in
chunks of variable sizes ('planks/planes'). The size of these is
determined by the value 'MaxMem' in spm_spm_ui.m On the one hand,
increasing the value will increase the chunk size, thus processing large
amounts of data at once. In this case, you would make full use of the
RAM that you have. On the other hand, you may already be using the RAM
that you have to its full extent, and may find that the reason things
are slow is because SPM is using lots of swap space (reading/writing to
disk). In this instance, perhaps decreasing the MaxMem value will allow
SPM to handle smaller chunks and process everything in RAM - which might
make things faster.
Why not check to see how much RAM and SWAP are being used during each
process and adjust the MaxMem setting accordingly?
Best wishes,
Narender
Stephen Fromm wrote:
>
> Fellow SPMers,
>
> Has anyone encountered difficulties processing large datasets via fixed
> effects analyses? I have 12 subjects, 4 sessions/subject, and 251
> volumes/session, which is about 12,000 images. Right now, spm99 is
> calculating the globals, but it's pretty slow. My impression from watching
> it is that it was fast at the beginning but now is slowing down. (I'm
> running spm99 under SunOS 5.5.1 and have a lot of RAM---2 GB.)
>
> Specifically, is there a practical bound on the number of volumes one can
> use? If so, does anyone have a workaround?
>
> Best wishes,
>
> Stephen Fromm, PhD
> NIDCD/NIH
--
********************************************************************
Dr Narender Ramnani
Sensorimotor Control Group
Department of Physiology
University of Oxford
Parks Road
Oxford OX1 3TP
Oxford University Centre for
Functional Magnetic Resonance Imaging of the Brain,
John Radcliffe Hospital,
Headington,
Oxford OX3 9DU
Tel. 01865 222704 (Direct)
01865 222729 (Admin)
mob. 0771 2632785
Fax. 01865 222717
email [log in to unmask]
*******************************************************************
|