Skip to content

open source MLFMA (multi level fast multipole algorithm) or open source MLFMM (multi level fast multipole method)

License

Notifications You must be signed in to change notification settings

Gjacquenot/Puma-EM

Repository files navigation

Puma-EM

Parallel Unified Multipole Algorithm for Electromagnetics

Travis

Puma-EM provides surface Method of Moments for Electromagnetics, enhanced by using the Multilevel Fast Multipole Method. Code is parallelized and runs on Desktops and clusters.

Theory and user manuals can be downloaded on the release page.

Introduction

This is Puma-EM, a Parallel Unified Multipole Algorithm for Electromagnetics.

The aim of Puma-EM is to solve surface integral equations that arise in Computational Electromagnetics, by using boundary elements methods, namely the Method of Moments. The method is enhanced by the use of the Multilevel Fast Multipole Method, which expedites the matrix-vector multiplication required by the iterative algorithm.

The term "Unified" refers to the fact that the Multipole Engine could be used for acoustics or mechanics. "Unified" should therefore be seen as a goal to reach. Hopefully the Open-Sourcing of the code is a step towards that goal.

Puma-EM is distributed under the terms of the GNU General Public License v3. See COPYING for more information.

For any questions/bugs/requirements, mail vandenbosch.idesbald@gmail.com.

Capabilities

This code is currently capable of computing the bistatic Radar Cross Section (RCS), monostatic RCS and monostatic (Sythetic Aperture Radar) SAR of PEC targets. These targets can include junctions (i.e. plates that intersect volumes for example), therefore the code can solve complex geometries.

Puma-EM has already solved problems containing more than 40 million of variables on a "modest" cluster (2 HP proliant ML350 machines, quad core and 16 GB RAM each), but it is able to solve around 1-2 million on a Desktop/laptop having 2 GBytes of RAM.

Please note that Puma-EM has been developed and used only on Linux machines. Windows (R) is not supported (yet). For running Puma-EM on Windows machines, you should consider installing a virtual linux machine through VMware or any other virtualization solution.

Puma-EM should work on *NIX/*BSD systems, including MacOS (R) 10.x, although it has not been tested yet.

Install Puma-EM

First, extract the *.tar.gz file that you downloaded in your home directory.

Second, cd into Puma-EM, then open the guide.pdf file. You can install the necessary files locally or, if you have root access to the machine, you can install it for all the users. The needed libraries and programs are:

  • gmsh, an open-source mesh generator
  • g++ and gfortran compilers (works with intel compilers too)
  • blitz++, a C++ container library
  • OpenMPI for the parallel code
  • scipy and numpy, Python scientific libraries
  • mpi4py, which is mpi for Python
  • matplotlib, a Python plotting library.

There are automated scripts for installing Puma for some distributions. Due to a lack of time, not all of them are fully tested, but they should work. Type ./install.sh in a terminal and follow the instructions. If unsure, install as per the guide.

Ready to Go

You are now ready to rock'n'roll!!

Choose a master node, open a terminal, and simply type:

$ ./run.sh

See the doc/EXAMPLES file for other examples, then play around with the simulation_parameters.py file.

Code documentation

If you want to generate the documentation of the code itself, type (doxygen and LaTeX required):

$ make documentation

About

open source MLFMA (multi level fast multipole algorithm) or open source MLFMM (multi level fast multipole method)

Topics

Resources

License

Stars

Watchers

Forks

Packages

No packages published