We're running SPM ('96) on an Dec Alpha, version 4.0d of their Unix.
512K of REAL ram, plenty of proc & swap space, no other users.
Model Alpha Statsion 255 (233 Mega Hurts).
version: % @(#)spm_spm.m 1.23 Andrew Holmes, Karl Friston 96/09/24
This is an fMRI analysis to start with.
The analysis has worked for smaller numbers of cases (patients); this
attempt is 3024 (ish) files for 14 subjects * 216 images/subject.
We had problems getting through mapping and we've passed that hurdle
by kernal tuning. Now this.
The symptom is it runs (AnCova) for a few days then just goes to sleep.
The 'progress bar' gets stuck at 25% in this one case. The CPU is at 100%
and no errors, not much swapping. We can supply lots more info if you need it. :-)
It's working sorta OK (5 hrs out of a projected 10 thus far) to do 1852 files (more patients,
less images per subject.)
Is there any diagnostic messages that I could put in 'spm_spm' to get a beter idea of the
'progress'. I don't know MATLAB; but I can sorta read the spm code.
The comment:
%-Cycle over cubes to avoid memory problems
%-----------------------------------------------------------------------
leads me to conclude that there may be a memory problem lurking around.
Can anyone illuminate me on this comment?
As a test we running this on a modern Alpha with 1 GIG of memory and
having the same results. Both stuck at 25%
We are willing to dedicate time to trying to solve this problem. we have a
'spare machine' to test with; but I would like to get better instrumenation
that "xx% complete." if I can.
Thank you all in advance for your consideration and kind attention to this
pressing matter.
Bill Scheding
[log in to unmask]
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|