Difference between revisions of "Espresso on CSC"

From ICPWiki
Jump to navigation Jump to search
Line 2: Line 2:
 
Here you can find a short introduction on how to compile and use ESPResSo on the CSC computer cluster.  
 
Here you can find a short introduction on how to compile and use ESPResSo on the CSC computer cluster.  
 
   
 
   
== Compiling ==
+
== CSC-I (login.csc.uni-frankfurt.de) ==
=== CSC-I (login.csc.uni-frankfurt.de) ===
+
=== Compiling ===
 +
To be done.
 +
=== Running ===
 
To be done.
 
To be done.
  
=== CSC-II (drei.csc.uni-frankfurt.de) ===
+
== CSC-II (drei.csc.uni-frankfurt.de) ===
 +
=== Compiling ===  
 
You have three different options of compiling Espresso on the CSC-II cluster.
 
You have three different options of compiling Espresso on the CSC-II cluster.
  
Line 47: Line 50:
 
                   LDFLAGS="-L/opt/fftw3/lib -L/opt/tcl8.4.9/lib"
 
                   LDFLAGS="-L/opt/fftw3/lib -L/opt/tcl8.4.9/lib"
  
=== CSC-III (quad.csc.uni-frankfurt.de) ===
+
=== Running ===
You have three different options of compiling Espresso on the CSC-II cluster.
+
'''All jobs (even single processor jobs) have to be submitted to the queueing system. Do not run time consuming applications (i.e. any kind of simulation) on the head node.'''
 +
The myrinet partion is only available for jobs that request more than 2 CPUs.
 +
 
 +
==== Environment Variables ====
 +
Make sure that you include
 +
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/opt/fftw3/lib:/opt/tcl8.4.9/lib
 +
in your profile (i.e. ~/.bashrc) in order to find the libraries during runtime.
 +
 
 +
==== Sample script for queueing system ====
 +
To be done
 +
 
 +
== CSC-III (quad.csc.uni-frankfurt.de) ==
 +
=== Compiling ===  
 +
You have three different options of compiling Espresso on the CSC-III cluster.
  
 
All three options assume that you want to use the recent way of configuring using object directories and have the Espresso sources in the following folder:
 
All three options assume that you want to use the recent way of configuring using object directories and have the Espresso sources in the following folder:
Line 81: Line 97:
 
  $SRCDIR/configure --with-mpi=mpich
 
  $SRCDIR/configure --with-mpi=mpich
  
== Running ==
+
=== Running ===
 
'''All jobs (even single processor jobs) have to be submitted to the queueing system. Do not run time consuming applications (i.e. any kind of simulation) on the head node.'''
 
'''All jobs (even single processor jobs) have to be submitted to the queueing system. Do not run time consuming applications (i.e. any kind of simulation) on the head node.'''
=== Environment Variables ===
+
The myrinet partion is only available for jobs that request more than 4 CPUs.
Make sure that you include
+
 
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/opt/fftw3/lib:/opt/tcl8.4.9/lib
+
==== Environment Variables ====
in your profile (i.e. ~/.bashrc) in order to find the libraries during runtime.
+
You don't have to add any specific environment variables to your profile in order to run Espresso, but it advisable to include the following line, which lets you use the complete functionality of the queueing system.
 +
 
 +
module add torque maui
 +
 
 +
==== Sample script for queueing system ====
 +
To be done

Revision as of 20:05, 2 December 2006

Espresso on CSC

Here you can find a short introduction on how to compile and use ESPResSo on the CSC computer cluster.

CSC-I (login.csc.uni-frankfurt.de)

Compiling

To be done.

Running

To be done.

CSC-II (drei.csc.uni-frankfurt.de) =

Compiling

You have three different options of compiling Espresso on the CSC-II cluster.

All three options assume that you want to use the recent way of configuring using object directories and have the Espresso sources in the following folder:

$SRCDIR = $HOME/src/Espresso

.

Single Processor

1. Create an object directory

mkdir ~/bin/Espresso/myobj-single
cd ~/bin/Espresso/myobj-single

2. Create and edit the configuation file 'myconfig.h'

cp $SRCDIR/myconfig-sample.h ./myconfig.h
vi myconfig.h

3. Execute configure

$SRCDIR/configure --without-mpi \
                  CPPFLAGS="-I/opt/fftw3/include -I/opt/tcl8.4.9/include" \
                  LDFLAGS="-L/opt/fftw3/lib -L/opt/tcl8.4.9/lib"

4. Build Espresso

make -j

Multi Processor with Ethernet Communication

As above but exchange step 3 by the following:

3. Execute configure

export PATH=/opt/mpich/gcc323/bin/:$PATH
$SRCDIR/configure --with-mpi=mpich \
                  CC="/opt/mpich/gcc323/bin/mpicc" \
                  CPPFLAGS="-I/opt/fftw3/include -I/opt/tcl8.4.9/include" \
                  LDFLAGS="-L/opt/fftw3/lib -L/opt/tcl8.4.9/lib"

Multi Processor with Myrinet Communication

As above but exchange step 3 with the following:

3. Execute configure

export PATH=/opt/mpich_gm/gcc323/bin/:$PATH
$SRCDIR/configure --with-mpi=mpich \
                  CC="/opt/mpich_gm/gcc323/bin/mpicc" \
                  CPPFLAGS="-I/opt/fftw3/include -I/opt/tcl8.4.9/include" \
                  LDFLAGS="-L/opt/fftw3/lib -L/opt/tcl8.4.9/lib"

Running

All jobs (even single processor jobs) have to be submitted to the queueing system. Do not run time consuming applications (i.e. any kind of simulation) on the head node. The myrinet partion is only available for jobs that request more than 2 CPUs.

Environment Variables

Make sure that you include

export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/opt/fftw3/lib:/opt/tcl8.4.9/lib

in your profile (i.e. ~/.bashrc) in order to find the libraries during runtime.

Sample script for queueing system

To be done

CSC-III (quad.csc.uni-frankfurt.de)

Compiling

You have three different options of compiling Espresso on the CSC-III cluster.

All three options assume that you want to use the recent way of configuring using object directories and have the Espresso sources in the following folder:

$SRCDIR = $HOME/src/Espresso

.

Single Processor

1. Create an object directory

mkdir ~/bin/Espresso/myobj-single
cd ~/bin/Espresso/myobj-single

2. Create and edit the configuation file 'myconfig.h'

cp $SRCDIR/myconfig-sample.h ./myconfig.h
vi myconfig.h

3. Load needed modules & execute configure

module add fftw/gcc/64
$SRCDIR/configure --without-mpi

4. Build Espresso

make -j

Multi Processor with Ethernet Communication

As above but exchange step 3 by the following:

3. Load needed modules & execute configure

module add fftw/gcc/64 mpich/ge/gcc/64
$SRCDIR/configure --with-mpi=mpich

Multi Processor with Myrinet Communication

As above but exchange step 3 with the following:

3. Load needed modules & execute configure

module add fftw/gcc/64 mpich/mx/gcc/64 mx/64
$SRCDIR/configure --with-mpi=mpich

Running

All jobs (even single processor jobs) have to be submitted to the queueing system. Do not run time consuming applications (i.e. any kind of simulation) on the head node. The myrinet partion is only available for jobs that request more than 4 CPUs.

Environment Variables

You don't have to add any specific environment variables to your profile in order to run Espresso, but it advisable to include the following line, which lets you use the complete functionality of the queueing system.

module add torque maui

Sample script for queueing system

To be done