Sid,
None of the default processing (that, recon-all -all) uses matlab. Only the local gyrification index (the -lgi flag, which is not part of the default stream) requires matlab. FS-FAST requires matlab, but that is a tool outside of the recon-all stream.
We use a cluster of some 300 Linux nodes here at the NMR Center, managed by the Portable Batch System job scheduling software. It seems to work pretty well given the number of people at the center who use it. We have a system support team to keep this going.
This page lists the lgi process: https://surfer.nmr.mgh.harvard.edu/fswiki/LGI
it takes the ?h.pial surface as input. to get that surface, you'll need to run recon-all -all, although if you really wanted to trim some time, you can follow this chart: https://surfer.nmr.mgh.harvard.edu/fswiki/ReconAllDevTable and you could do this:
recon-all -autorecon1 -autorecon2 -noaseg
Nick
On Sun, 2008-12-07 at 13:17 -0800, Siddharth Srivastava wrote:
Hi all, 1) which steps of the freesurfer processing pipeline depend on/ use matlab for execution? I am trying to find out if we have to have matlab installed on all the computing nodes on our cluster here, or, if there is just one step, we can process it locally, and submit the downstream steps back to the cluster. 2) if anyone would like to share their experiences with running freesurfer on a cluster/grid computing environment, I would be happy to know, and maybe optimize my computing environment based on your recommendations 3) If there a canonical list of dependencies for the output generated by the -localGI flag. I was wondering if, if LGI is the only thing i need at the end, do i have to let freesurfer run through the entire processing steps? I mean, i do need other outputs as well, but can i chose a subset of processing steps based on what i would ultimately be using? :-) best regards, sid.
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer