Print

Print


Dear Moises,

The Queue system for the cluster i use is : qsub  , https://wiki.scinet.utoronto.ca/wiki/index.php/GPC_Quickstart#Managing_jobs_on_the_Queuing_system

Thanks
vasudev


On 22 June 2017 at 17:33, Moises Hernandez <[log in to unmask]> wrote:
Hi Vasudev,

I guess you are having problems with the script (bedpostx_gpu) that submits the jobs to the cluster.
Probably your cluster has a different queue system. The script was written for a cluster with SGE and using fsl_sub.
It also assumes that some environment variables are set.

What queue system does the cluster have?

You can adapt the script,
Maybe someone has already adapted it for the same queue system.

Moises.

On 22 June 2017 at 13:20, Dev vasu <vasudevamurthy.devulapally@gmail.com> wrote:
Dear moises,

I am using cuda 6.0 and i am finding it hard to run bedpostx on the cluster, Could you please let me know how i could adapt the code.

Thanks
vasudev

On 22 June 2017 at 13:09, Moises Hernandez <[log in to unmask]> wrote:
Thanks Andras,

Yes, that is right. 
The kernels that bedpostx runs take the control of the entire GPU. 

Moises.

On 22 June 2017 at 08:39, Dr. András Jakab <[log in to unmask]> wrote:
Hi Moises,

Thanks a lot, it solved the problem.

There was another glitch, may be worth mentioning on the FSL page: the same GPU cannot do calculations and render X11 at the same time, at least for me, X11 recurrently crashed and I had to run commands after closing X (Ubuntu 16.04).

Cheers,
Andras

On Wed, Jun 21, 2017 at 2:17 PM, Moises Hernandez <[log in to unmask]> wrote:

On 21 June 2017 at 12:32, Dr. András Jakab <[log in to unmask]> wrote:
Dear FSL team

We have installed CUDA8 on our server, and I was wondering if there is a build (e.g. bedpostx_gpu / xfibres_gpu / probtrackx_gpu) for this version.

These links don't work at the moment:
https://users.fmrib.ox.ac.uk/~moisesf/
https://users.fmrib.ox.ac.uk/~moisesf/bedpostx/CUDA-7_5.zip

I am actually getting error messages. ldd xfibres_gpu could locate the dependencies properly. However bedpostx_gpu quits with error "what(): basic_string::_S_construct null not valid  ... "

Thank you.
Best Regards,
Andràs


--
Dr. med. Andras Jakab, PhD




--
Dr. med. Andras Jakab, PhD
Marie Curie Post Doctoral Researcher
Computational Image Analysis and Radiology Lab
Medizinische Universität Wien
Universitätsklinik für Radiologie und Nuklearmedizin
Währinger Gürtel 18-20
A-1090 Wien
Cellular: +36305095083
----------------------------------