yes, but -openmp N should give significant imrpovements up to about N=4

On Sat, 9 Feb 2019, Tim Schäfer wrote:

       External Email - Use Caution

I do not have experience with running on Google Cloud, but from what I know, 
the common way is to run FreeSurfer in parallel over many subjects instead of 
using openmp to speed up the process for a single subject.

So your desktop should run 4 subjects in parallel, and the cloud instance 16. 
Not that bad.

Tim

On February 9, 2019 at 4:33 PM Yi-Fang Chuang <chua...@ym.edu.tw> wrote:


        External Email - Use Caution

Dear Freesurfer developers and users,

Has anyone had experience running Freesurfer on Linux VM on Google Cloud
Platform?
I have a question about runtime of recon-all -all.
The specs of VM on GCP as follows,
Linux CentOS 6, 16 CPU(Intel(R) Xeon(R) @ 2.50GHz) with 60 G memory
I installed Freesurfer v 6.0.0 and run recon-all -all -parallel -openmp num
No matter what number I put (6,8,12,16 after openmp), the recon-all runtime
of a single subject ranges 6-7 hours.
It's just disappointing to compare to my own desktop.
My own Desktop is Mac Pro (3.7 GHz quad-core intel xeon E5, 16G memory).
The recon-all runtime (Freesurfer version 5.3.0, no parallel) for a single
subject is usually 7~8 hours.
Is this a reasonable runtime given the setting of VM?  Only 1-2 hour
shorter than my Mac Pro??
Or is there any configuration of VM I should pay attention to set up?

Thanks,
Yi-Fang Chuang
ᐧ
ᐧ
_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer

_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer


_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer

Reply via email to