Dear Bennet,
I'm afraid that the majority of FSL is not multi-threaded, although a couple of special cases are.
Hence using OMP_NUM_THREADS does not affect the majority of FSL.
Our submission tool, fsl_sub, is only configured to run with queuing systems, such as SGE.
One option is to install a queuing system on your machine, or otherwise just run multiple jobs simultaneously (which is most of what fsl_sub does).
We are working on a version of fsl_sub to make it easier to use multi-core machines, but it is still in development. If you'd like a pre-release copy for testing then I'd be happy to send it to you.
All the best,
Mark
> On 23 Jul 2017, at 20:35, Bennet Fauber <[log in to unmask]> wrote:
>
> I hope I have not missed something that is stunningly obvious.
>
> I would like to be able to use more than one processor on the machine
> on which I run FSL.
>
> The fsl_sub script seems to be set up only for SGE and Condor, not for
> multicore. I saw in it references to OMP_NUM_THREADS, but in search
> the FSL source code and looking in the binary distribution, I do not
> find anything to indicate that the FSL utilities are multithreaded.
>
> I have access to machines with 4-56 cores, some in a cluster, some in
> labs. I was hoping there was a way to specify manually, either by
> command option or environment variable, the number of processors that
> could be used on the current machine so that the same script would be
> most portable amongst the various installations and Linux versions.
>
> I thought at one point I saw a version of fsl_sub that used GNU
> Parallel, but I cannot find it now.
>
> I am not an imager, I just help out with the Linux part when able, so
> pardon my ignorance.
>
> Thanks, -- bennet
|