Thanks Tim. FWIW it might add flexibility to make the # of cores a command line option to the script.
-M On Mon, Jun 22, 2015 at 4:31 PM, Timothy B. Brown <[email protected]> wrote: > Hi again Michael, > > Thanks again for your input. > > I have made the following changes to the FreeSurferPipeline.sh script. > They are not quite the same as your changes, because for the time being I > want to maintain a level of backward compatibility. That is, if NSLOTS is > set, we'll use it to determine the input value for the -openmp option. If > it is not set, I want to continue to behave just as before (use 8 as the > value for the -openmp option.) > > Using the current value of 8 as the default for when NSLOTS is not set > might cause problems if specifying more cores than are actually available > to the recon-all command causes problems with the recon-all run. I have > not been able to find any documentation that indicates what would happen in > that case. So I'm being a bit conservative here and not changing the > default from the current value of 8 to a new value of 1. > > My changes look like the following. They have been checked into the > repository, are not yet in a released version, and should end up (unless > someone convinces me to set the default to 1 instead of 8) in the next > release. > > cp "$SubjectDIR"/"$SubjectID"/mri/brainmask.auto.mgz > "$SubjectDIR"/"$SubjectID"/mri/brainmask.mgz > > # Both the SGE and PBS cluster schedulers use the environment variable > NSLOTS to indicate the number of cores > # a job will use. If this environment variable is set, we will use it to > determine the number of cores to > # tell recon-all to use. > > if [[ -z ${NSLOTS} ]] > then > num_cores=8 > else > num_cores="${NSLOTS}" > fi > > recon-all -subjid $SubjectID -sd $SubjectDIR -autorecon2 -nosmooth2 > -noinflate2 -nocurvstats -nosegstats -openmp ${num_cores} > > > Best Regards, > > Tim > > On Tue, Jun 16, 2015, at 11:24, m s wrote: > > Version 3.4.0 > > Hi, > > The Pipelines-3.4.0//FreeSurfer/FreeSurferPipeline.sh script has a > hard-coded value of 8 for the -openmp argument in the call to recon-all. > This doesn't seem like a good idea. Seems it should be an option and > default to 1. > > We've modified it to work with NSLOTS for SGE on our cluster. Here's the > diff if you're interested: > > < cp "$SubjectDIR"/"$SubjectID"/mri/brainmask.auto.mgz > "$SubjectDIR"/"$SubjectID"/mri/brainmask.mgz > < > < # CFN mod - use $NSLOTS to define multithreading > < if [[ -z "$NSLOTS" ]]; then > < NSLOTS=1 > < fi > < > < recon-all -subjid $SubjectID -sd $SubjectDIR -autorecon2 -nosmooth2 > -noinflate2 -nocurvstats -nosegstats -openmp $NSLOTS > --- > > cp "$SubjectDIR"/"$SubjectID"/mri/brainmask.auto.mgz > "$SubjectDIR"/"$SubjectID"/mri/brainmask.mgz > > recon-all -subjid $SubjectID -sd $SubjectDIR -autorecon2 -nosmooth2 > -noinflate2 -nocurvstats -nosegstats -openmp 8 > > _______________________________________________ > HCP-Users mailing list > [email protected] > http://lists.humanconnectome.org/mailman/listinfo/hcp-users > > -- > Timothy B. Brown > Business & Technology Application Analyst III > Pipeline Developer (Human Connectome Project) > tbbrown(at)wustl.edu > ________________________________________ > The material in this message is private and may contain Protected > Healthcare Information (PHI). > If you are not the intended recipient, be advised that any unauthorized > use, disclosure, copying > or the taking of any action in reliance on the contents of this > information is strictly prohibited. > If you have received this email in error, please immediately notify the > sender via telephone or > return mail. > _______________________________________________ HCP-Users mailing list [email protected] http://lists.humanconnectome.org/mailman/listinfo/hcp-users
