[pymvpa] Speed up loading fmri dataset in mvpa 2.3

Michael Hanke michael.hanke at gmail.com
Tue Dec 16 10:12:44 UTC 2014


Hi,

On Tue, Dec 16, 2014 at 9:33 AM, gal star <gal.star3051 at gmail.com> wrote:

> Hello all,
> Is there a way to speed up the loading of a Nifti file
> using fmri_dataset?
>
> I'm running mvpa script on a 16 core machine.
> Having a 3080 scans concanated into single Nifti file.
>
> According to time measurements, It could take up to 20 min
> (only fmri_dataset command). I've also noticed it runs on a single core.
>
> Is there a way, similar to searchlight method, to use more than one core
> for the mission of loading - hence, speeding the time it takes?
>

What fmri_dataset() is doing is  essentially IO bound. Adding multiple
cores doesn't make things faster.

Are you pulling data over the network (NFS?)? How large is it (the number
of volumes doesn't matter much).

Michael
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.alioth.debian.org/pipermail/pkg-exppsy-pymvpa/attachments/20141216/5809159b/attachment-0001.html>


More information about the Pkg-ExpPsy-PyMVPA mailing list