Gerris2D & Gerris3D

Overview

Gerris is a program for the solution of the partial differential equations describing fluid flow.

Version snapshot 13.01.12 is installed on the CSF, compiled for AMD Bulldozer. The compilation includes

• GTS snapshot 12.11.30
• gfsview-batch2D and gfsview-batch3D
• OSMesa 7.8.2
• FFMPEG 0.6.5
• Hypre 2.0.0 has NOT been compiled in to this version

Restrictions on use

There are no restrictions on user access. However, this software has only been compiled for the AMD Bulldozer nodes and so you must have access to these nodes.

Set up procedure

To access the software you must first load the modulefile:

module load apps/amd-4.5.2/Gerris/13.01.12

This will set up your path to run the tools in Gerris, including gfsview-batch2D and gfsview-batch3D. It will also set the environment variable GERRIS_HOME which can be used in your own Makefiles if compiling code against the Gerris libraries. See the page on compiling OpenMPI applications on AMD Bulldozer nodes for more information.

Running the application

Please do not run Gerris on the login node. Jobs should be submitted to the compute nodes via batch and should use the orte-64bd-ib.pe parallel environment.

Gerris will not run on the Intel nodes and will not run serially (i.e., without mpirun)

It is highly recommended that you run your jobs from the scratch file system (especially when generating large .mpg files).

Parallel batch job submission

Make sure you have the modulefile loaded then create a batch submission script, for example:

#!/bin/bash
#$ -S /bin/bash
#$ -cwd
#$ -V
#$ -pe orte-64bd-ib.pe 64      # Multiples of 64 must be used

mpirun -n $NSLOTS gerris2D input.gfs
  #
  # Can also run gerris3D for example
  #

Submit the jobscript using:

qsub scriptname

All Gerris command must be run via mpirun. Hence If you need to run gerris2D or gerris3D to do pre-processing (e.g., to split your domains) before running the solver, add the following to the jobscript above (check the docs for your required flags)

# Example commands to split and parallelise a domain for use with 4 processors.
# Preprocessing is done serially (i.e., using 1 MPI process)
mpirun -n 1 gerris2D -s 2 -m input.gfs > input-s2.gfs
mpirun -n 1 gerris2D -p 2 -m input-s2.gfs > input-p2.gfs
# Now run the simulation in parallel with 4 processors (set the PE to this many processors)
mpirun -n $NSLOTS gerris2D input-p2.gfs

In the above example the input.gfs file contains the commands to process the results through gfsview-batch2D. For example, it may contain the following lines:

ppm2mpeg -s 800x100 > pid.mpg 
gfsview-batch2D pid.gfv end.gfs

See the parallel example at http://gerris.dalembert.upmc.fr/gerris/examples/examples/cylinder.html#htoc5 for a complete parallel example.

Further info

• Gerris website.
• See the section on preparing your input files for running Gerris in parallel.
• After loading the Gerris modulefile you can use man gerris2D or man gerris3D to access the manual pages.

Updates

None.