At 08:28 25/06/99 -0700, you wrote:
>See John Ashburner's recent message about this issue:
>
>http://www.mailbase.ac.uk/lists/spm/1999-06/0058.html
>
>basically the story is that images are being written by matlab routines
>in SPM99b rather than compiled MEX files, which slows things down. This
>was done to allow flexibility for writing to different image formats.
>
Thank you for the information. Indeed, it should make a big difference.
>My 2 cents: I am willing to sacrifice flexibility for speed (especially
>given how big the differential seems to be), and I would suggest that
>these routines be replaced with MEX files in the final release of SPM99.
>
I totally agree. I am currently starting to analyse data from an experiment
which includes 10 subjects with 8 series of about 200 images. I estimated
the time to realign everything : around 240 hours of processing!!, which is
not reasonable, except running spm99 on Crays!! As a result I will use
SPM96. Processing time becomes a real issue since fMRI experiments tend to
produce more and more data.
>best,
>russ
Thanks Russ for your response. We should suggest SPM team to propose at
least a fast version with key routines as MEX files.
>----------------------------------------------------------------------
>Russ Poldrack Phone: (650) 725-0797
>Dept. of Psychology Fax: (650) 725-5699
>Stanford University
>Jordan Hall Email: [log in to unmask]
>Stanford, CA 94305 Web Page: http://www-psych.stanford.edu/~poldrack/
>
>After August 1, 1999:
>MGH-NMR Center
>Building 149, 13th St.
>Charlestown, MA 02129
>Email: [log in to unmask]
>----------------------------------------------------------------------
-------
Etienne Koechlin, PhD.
Cognitive Neuroscience section
NINDS/Bldg 10 Room 5C205
National Institutes of Health
Bethesda, MD 20892
Phone: (301) 402 6390
Fax: (301) 480 2909
Email: [log in to unmask]
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|