Hi Vasudev,

 

in case you use v6, you can use the -parallel flag to process the hemispheres (and a few other things if I remember correctly) in parallel. Note that doing so will parse the -openmp flag to all instances. E.g.

 

recon-all -all -openmp 4 -parallel -s subjID -i inputFile

 

Also the reduction in processing time using openmp is limited – I don’t recall exactly if it was bound to the number of physical cores your CPU has or if it was an upper limit to due the way the parallelization works (I have something like in mind like not much of an increase beyond openmp 8).

 

Best,

Falk

 

Von: freesurfer-bounces@nmr.mgh.harvard.edu <freesurfer-bounces@nmr.mgh.harvard.edu> Im Auftrag von Dev vasu
Gesendet: Mittwoch, 18. September 2019 15:08
An: freesurfer@nmr.mgh.harvard.edu
Betreff: [Freesurfer] How to fasten recon-all execution time without using CUDA

 

        External Email - Use Caution