We hope that you
took as much pleasure as us to work on this subject.
To sum up, results obtained with 2D code and 3D code coincide with physic and numerical mathematics. However, it's a pity we don't have more time to make use of 3D code (which was unmanageable) in order to show all the vorticy instabilities in the domain.
Parallelization seems very efficient with the FFTW library. Indeed, scaling graphs show that we decrease computation time when we increase processors number (time = 11371*exp(-0.5827*processors number)).
The main part of our job has been to understand how to use FFTW library (very capricious) and to manage its use. We have realized that our formation helped us during all this placement although all compilers, system and computers are different and require a permanent adaptation.
Thank particularly to Mr. Nicholas KEVLAHAN (our supervisor).
We would also like to thank Mr. James WADSLEY (HPC Consultant), Mr. Mark HAHN (System administrator of McMaster Sharcnet) and all the other people of the Mathematics and Statistics Department who helped us during this work.