Hi,
I'm on Linux.
I call bedpostx_gpu in $FSL_DIR/src/fdt/CUDA directory and this error occured :
/tools/FSL/fsl-5.0.8/fsl/bin/xfibres_gpu: error while loading shared libraries: libcudart.so.5.5: cannot open shared object file: No such file or directory
May be it's a wrong installation of FSL/CUDA but I don't know fix that.
Danaele
From: FSL - FMRIB's Software Library [[log in to unmask]] on behalf of Moises Hernandez [[log in to unmask]]
Sent: Wednesday, September 16, 2015 5:42 PM
Hi Danaele,
What problem did you have using the CUDA version of bedpostx ?
What OS do you have ?
Moises.
On 16 September 2015 at 22:24, Puechmaille, Danaele <[log in to unmask]> wrote:
Hi Moises,
Yes I have a gpu and I tried to use the CUDA implementation of bedpostx but that didn't work. So I write a script which parallelized xfibres.
Thanks for your help
Regards,
Danaele
From: FSL - FMRIB's Software Library [[log in to unmask]] on behalf of Moises Hernandez [[log in to unmask]]
Sent: Friday, September 04, 2015 5:08 AM
To: [log in to unmask]
Subject: Re: [FSL] Parallelising bedpost
Hi Danaele,
Do you have a GPU ?It will be much faster if you use the CUDA implementation of bedpostx.
Moises.
On 3 September 2015 at 17:44, Danaele Puechmaille <[log in to unmask]> wrote:
Dear experts,
I try to parallelise bedpostx in order to obtain bedpostx results faster.
So I have followed the FSL Tutorial - Parallelising bedpost : http://fsl.fmrib.ox.ac.uk/fsl/fsl4.0/fdt/fdt_bedpostx_parallel.html
I set environment variables FSLMACHINELIST and FSLREMOTECALL like that :
setenv FSLMACHINELIST “mymachine mymachine mymachine”
setenv FSLREMOTECALL ssh
Moreover, I have a file called fsl.sh in this place : ${HOME}/.fslconf/
In this file I set the environment variable FSLDIR to my FSL directory.
Also I'm a tcsh user, so I have created a file fsl.csh in the same directory and I did the same thing : set the environment variable FSLDIR to my FSL directory.
At this point I have followed the tutorial and bedpost should be parallelised.
But when I execute the command bedpostx in my data directory, nothing is parallelised.
Furthermore, if I am connected to another machine (for example “myothermachine”) and I want to execute bedpost in mymachine with the environment variables set to the same values, the command bedpostx is executed in myothermachine and not parallelised too.
So, I think bedpostx never tried to be connected to another machine with shh.
If someone has already meet this problem and found a solution, please explain me. Otherwise if someone have a solution to parallelise bedpost, I am really interested and that could be very helpful for me!
Thank you,
Best regards,
Danaele Puechmaille