www.neuron.yale.edu

Posted: **Mon Jul 16, 2007 4:51 am**

Hi,

We had neuron 5-9.9 and just upgraded to 6.0.3, compiled with:

./configure --prefix=`pwd`  --without-x --with-paranrn --with-mpi MPICC=/opt/mpich2/bin/mpicc MPICXX=/opt/mpich2/bin/mpicxx

I had to edit mpispike.c according to
https://www.neuron.yale.edu/phpBB2/viewtopic.php?t=946

the test in $NEURONHOME/src/parallel/ run but every nrniv process outputs:

Code: Select all

NEURON -- Release 6.0.3 (1793) 2007-06-21
Duke, Yale, and the BlueBrain Project -- Copyright 1984-2007
See http://www.neuron.yale.edu/credits.html

for example, when running test0.hoc:

Code: Select all

mpiexec -n 6 nrniv test0.hoc

NEURON -- Release 6.0.3 (1793) 2007-06-21
Duke, Yale, and the BlueBrain Project -- Copyright 1984-2007
See http://www.neuron.yale.edu/credits.html

NEURON -- Release 6.0.3 (1793) 2007-06-21
Duke, Yale, and the BlueBrain Project -- Copyright 1984-2007
See http://www.neuron.yale.edu/credits.html

There are 1 processes. My rank is 0 and I am on ocean
NEURON -- Release 6.0.3 (1793) 2007-06-21
Duke, Yale, and the BlueBrain Project -- Copyright 1984-2007
See http://www.neuron.yale.edu/credits.html

There are 1 processes. My rank is 0 and I am on brain21
There are 1 processes. My rank is 0 and I am on brain3
NEURON -- Release 6.0.3 (1793) 2007-06-21
Duke, Yale, and the BlueBrain Project -- Copyright 1984-2007
See http://www.neuron.yale.edu/credits.html

NEURON -- Release 6.0.3 (1793) 2007-06-21
Duke, Yale, and the BlueBrain Project -- Copyright 1984-2007
See http://www.neuron.yale.edu/credits.html

There are 1 processes. My rank is 0 and I am on brain3
There are 1 processes. My rank is 0 and I am on brain3
NEURON -- Release 6.0.3 (1793) 2007-06-21
Duke, Yale, and the BlueBrain Project -- Copyright 1984-2007
See http://www.neuron.yale.edu/credits.html

There are 1 processes. My rank is 0 and I am on brain3

We didn't see this behavior in the previous release, is it optional to disable this output on every process?

Thanks.

Posted: **Mon Jul 16, 2007 9:55 am**

You want to get rid of the banner?
nrngui -h
reveals a list of command line options that include

Code: Select all

-nobanner        do not print startup banner

Posted: **Mon Jul 16, 2007 11:33 am**

ted wrote:You want to get rid of the banner?
nrngui -h
reveals a list of command line options that include
Code: Select all
-nobanner        do not print startup banner

Indeed, the -nobanner disables the banner.
Apart from UNIX/Linux solutions like making an alias, Is there a way to tell nrniv to disable the banner when running parallel?

Thanks a lot.

Posted: **Mon Jul 16, 2007 12:54 pm**

Not sure what you're asking. Isn't it enough to include -nobanner on the command line,
whether you're typing the command line yourself or using a shell script that "packages"
the name of your hoc file with whatever other command line options are needed?

Posted: **Mon Jul 16, 2007 1:15 pm**

in the previous release we had, nrn-5.9, there was no need to include the -nobanner, more convenient.. the banner is a new behavior(?). it does make sense that only the master process print the banner. running 'mpiexec -n 500 ...', for example, will result in 500 printed banners if the -nobanner wasn't provided!

either way, it's not a problem to include -nobanner

thanks.

Posted: **Mon Jul 16, 2007 2:33 pm**

The multiple printing of banners signifies that the program is running not as a parallel program but as n independent non-communicating programs. What has been left out when you launch is the -mpi argument after the nrniv.

Posted: **Mon Jul 16, 2007 2:38 pm**

hines wrote:The multiple printing of banners signifies that the program is running not as a parallel program but as n independent non-communicating programs. What has been left out when you launch is the -mpi argument after the nrniv.

nrn-5.9 was compiled with 'always_call_mpi_init=yes' , hence there was no need to '-mpi'.

thanks.

Posted: **Mon Jul 16, 2007 3:14 pm**

mpiexec -n 6 nrniv test0.hoc
nrn-5.9 was compiled with 'always_call_mpi_init=yes'

You can continue to configure in that way and hence avoid the necessity of -mpi
when running in parallel. But then you would not be able to run that installation in single processor mode with a GUI.

www.neuron.yale.edu

mpi neuron output

mpi neuron output