This large scale project is an implementation of the Jacobi method for solving steady-state partial differential equations. The implementation is based on OpenMP, openMPI and CUDA to test the performance gains of using parallel technology for finding the steady-state solution to a partial differential equation. In our implementation we have tested the program with a simple heat equation using a sinusodial function as the source term. The repository contain all source code for the implementation along with shell and matlab files which can be used to test it.

The repository contain two folders,

• Poisson
  • Contain all files needed for compiling the program including a makefile and a readme describing the functionality of the program in depth.
• Testing
  • Containg shell scripts and matlab code used for testing and visualizing the results of the program.
• Technical report describing the process of developing the code and the theory the behind problem and method.
• Additional files used by us for streamlining the development process of the project.

In order to compile and run the code the following libraries must be available: (the versions specified in paranthesis is the tested versions).

• GCC ( 6.4.0 built with OpenMP suport )
• MPI ( openMPI 3.1.1 )
• CUDA ( 9.2 )

The poisson folder should be readily usable by the make utility. Please examine the makefile before compiling to check for compliation flags and compilers.

For direct usage of the jacobiSolver in the commandline please see the readme located in the Poisson folder. This describes the functionality in depth. The test scripts located in Testing/shell is designed to run on the LSF based cluster at the Technical University of Denmark. Please modify these in order to use them elsewhere. No additional packages are needed in order to run the matlab code used in the visualization. Matlab version R2017a was used for testing.

• Mathias Lorenz - MathiasLorenz
• Tim Felle Olsen - TimFelle

• Bernd Dammann
• Hans Henrik Brandenborg Sørensen