Hi,

For bedpostx, yes, I think is better to use the CUDA version.

The binaries are only included in the CentOS 6 distribution, 
but you can compile  the CUDA source code and it should work with Ubuntu and CUDA 7.5.

First you need to set some variables:
export FSLDEVDIR=$FSLDIR
export COMPILE_GPU=1
export CUDA='path to your CUDA directory' (/usr/local/cuda for instance)

You also need to set $FSLCONFDIR and $FSLMACHTYPE:
export FSLCONFDIR=$FSLDIR/config
export FSLMACHTYPE=`$FSLDIR/etc/fslconf/fslmachtype.sh`

And to compile it:
cd $FSLDIR/src/fdt
make install

Let me know if you find problems doing it.

Moises.


On 12 December 2015 at 20:50, d m <[log in to unmask]> wrote:
Hello,

We are using linux Ubuntu 14.04 and are setting up SGE to run fsl parallel processing.

However, when typing the following command:



[%] sudo -u sgeadmin qconf -am myusername

error: commlib error: access denied (client IP resolved to host name "localhost". This is not identical to clients host name "server.domain.name")
ERROR: unable to contact qmaster using port 6444 on host "server.domain.name"


I've installed it the same way we previously installed gridengine with Ubuntu 10.04.

I found the following link:


Where it mentions to edit the /etc/hosts file. I'm hesitant to do this since I fear it may cause harm to our system. Additionally, this step was not needed in Ubuntu 10.04.

Are we better off just using alternative methods instead, such as condor? We also have a GPU with Cuda 7.5 installed. Does this only work for CentOS 6 or CentOS 6.5? Can it work with Ubuntu?  I searched for the bedpostx_gpu command but could not find it. We are currently running fsl 5.0.9.

Any help would greatly be appreciated.


Sincerely,


D.M.