Prev: MEX Authentication file read Failed -- SOLVED
Next: matlab code for quantization
From: H SM on 1 Feb 2008 10:09
Hello,

I am running Matlab 2007b on Linux with 8 (2 quad-core)
processors. When I run matlabpool, I get the following message:

Submitted parallel job to the scheduler, waiting for it to
start.
Performing parallel job cleanup...
Done.
??? Error using ==>
distcomp.interactiveclient.pGetSockets>iThrowIfBadParallelJobStatus
at 137
The interactive parallel job finished without any messages.

Error in ==> matlabpool at 88
client.start('matlabpool', numlabs, config, 'nogui');

I have tried limiting the number of threads to 4, but the
message still persists.

Thanks.
From: Narfi on 1 Feb 2008 12:08
Could you try to start a new matlab session, run the
following commands:

setenv('MPICH_INTERFACE_HOSTNAME', '127.0.0.1')
dctconfig('hostname', 'localhost')

and then try to start the matlab pool.

Please let me know whether this solves your problem. If it
does, you can add those commands into your startup.m file so
that they are applied on every MATLAB startup.

Best,

Narfi
"H SM" <hm.nospam(a)gmail.com> wrote in message
<fnsoae$rgt$1(a)fred.mathworks.com>...
> Hello,
>
> I am running Matlab 2007b on Linux with 8 (2 quad-core)
> processors. When I run matlabpool, I get the following
message:
>
> Submitted parallel job to the scheduler, waiting for it to
> start.
> Performing parallel job cleanup...
> Done.
> ??? Error using ==>
>
distcomp.interactiveclient.pGetSockets>iThrowIfBadParallelJobStatus
> at 137
> The interactive parallel job finished without any messages.
>
> Error in ==> matlabpool at 88
> client.start('matlabpool', numlabs, config, 'nogui');
>
> I have tried limiting the number of threads to 4, but the
> message still persists.
>
> Thanks.

From: H SM on 1 Feb 2008 12:46
Thanks Narfi, that worked! One other question- what is the
easiest way to use all 8 cores on my desktop? If is the
Configuration Manager, what do I put for the Job Manager
Hostname? Thanks for your help.
From: Narfi on 1 Feb 2008 13:07
When you get the toolbox, you get the local scheduler and
the ability to run 4 local workers with it. If you want to
scale higher than that, or want to use remote machines, you
have to get the MATLAB Distributed Computing Engine. The
job manager comes with the Engine.

Best,

Narfi
"H SM" <hm.nospam(a)gmail.com> wrote in message
<fnt1gq$bmt$1(a)fred.mathworks.com>...
> Thanks Narfi, that worked! One other question- what is the
> easiest way to use all 8 cores on my desktop? If is the
> Configuration Manager, what do I put for the Job Manager
> Hostname? Thanks for your help.

 | 
Pages: 1
Prev: MEX Authentication file read Failed -- SOLVED
Next: matlab code for quantization