Parallel computing or the cooperative utilisation of a set of processors for scientific application
programming has become current nowadays. For irregular application a load
balancing system is necessary in order to optimise performance. This master's thesis is
focusing on this particular issue of dynamic load balancing. It presents a synthetic view
of some parallel programming models and runtime supports.
A specific system, AMPI (Adaptiv MPI), carries out intelligent adaptive optimisation for
traditional parallel MPI programs. This system is specially studied. Results are given for a
test application and an application experience involving molecular dynamics. These results
show that AMPI can be useful for irregular or dynamic applications on an heterogeneous
architecture, like a non dedicated cluster.
Équilibrage dynamique de charge pour des calculs parallèles sur cluster Linux: une évaluation de l'environnement AMPI
Teller, F. (Author). 2004
Student thesis: Master types › Master in Computer science