Dear FSL experts,
I recently started using FSL for VBM on a total sample of 552 subjects of which 184 subjects are from the experimental group and the remaining 368 age-matched subjects are representing the control group.
I have access to a computer with 32 cores and 128 gb of memory. Using a modified fsl_sub (obtained here: https://rcbi.rochester.edu/users/vanooste/weblog/f657b/fsl_sub_implemented_for_local_multiprocessor_systems.html) I think I will be able to run fslvbm_2_template parallel.
I tested the modified fsl_sub script on my dual core macbook (2,2 ghz / 4 gb).
For this little experiment I used 3 T1 images (100 mb each) of which one belonged to the experimental group and the other two belonged to the control group. In the root of the directory I created the file 'template_list' that included the filenames of the scan from the experimental group and only one scan of the control group. The command I entered was the following:
fsl_sub fslvbm_2_template -n
I was happy to find out that indeed two 'fast' processes showed up in the proces list, but I still have some questions:
1) I noticed that it took the macbook a long time to complete the tasks (several hours) and I was wondering whether this is normal.
2) In the proces list I noticed that the 2 fast processes did not use the maximum capacity of the 2 cores, but were merely using it up to 10 percent (there was some peak activity though). Is this normal?
3) What do you experienced FSL users would think regarding the time needed for the local-32-core-cluster to complete the job for 552 scans?
I'm looking forward to your answers.
Kind regards,
Vincent
|