Dear fsl user
I would like to understand the required HW/SW specs to run bedpost_gpu on linux.
the installation instruction just report the possibility to use CentOS 6 & 6.5 with different type of cuda sdk according to the fsl version.
Linux CentOS 6 or CentOS 6.5
(FSL version 5.0.6 and CUDA Toolkit 5.0) or (FSL version 5.0.7/5.0.8 and CUDA Toolkit 5.5) or (FSL version 5.0.9 and CUDA Toolkit 6.5)
Ubuntu seems to not be supported at all.
in a previous post, Moises (hola moises !!!) redirected users who wanted alternative cuda configs to the following page https://users.fmrib.ox.ac.uk/~moisesf/Bedpostx_GPU
where different compiled version of bedpost_gpu are available, according to the cuda sdk version.
do all this versions work with every kind of centOS 6+ & 7 ???
and what about the supported fsl version?. applying the above patch, can I use whichever fsl version (5+) I prefer ???
I simply have to overwrite files in bin/lib folders and everything will work?
concerning ubuntu.
I downloaded the centos version 5.0.9, applied the patch, installed cuda 8, downloaded the cuda8 compiled files and run bedpost....the PC got frozen, rebooting after one hour, I realized that nothing was performed and, among the logs, I notices a bedpost_gpu.oXXXXXX saying
cuda_error at CUDA/diffmodels.cu:253. the launch timed out and was terminated
ERROR: fit_PVM_multi_kernel: the launch timed out and was terminated
is it a normal error message for ubuntu?
Is there a way to use ubuntu, or do I have to setup a centOS workstation?
moreover.....
will you ever release a compiled FSL version including bedpost_gpu for ubuntu?
It may be worth updating the installations instruction according to these question....
centos 7 for example is already quite old, and cuda is now at version 8.
thanks in advance...
Alberto
|