This study focuses on parallelization strategies for a numerical code, modelling semiconductor devices, exploiting the parallel features of distributed shared memory architectures. A simple, message passing parallelization, that uses MPI to manage interprocessor communication, is suggested. A good scaling is obtained, moreover it gives the chance to introduce a hybrid multilevel parallelism based on the mix of MPI and a multithreaded version of the linear systems solver included in the code.
Hybrid parallelism for modelling semiconductor devices
P Pietra;S Rovida;G Sacchi
2004
Abstract
This study focuses on parallelization strategies for a numerical code, modelling semiconductor devices, exploiting the parallel features of distributed shared memory architectures. A simple, message passing parallelization, that uses MPI to manage interprocessor communication, is suggested. A good scaling is obtained, moreover it gives the chance to introduce a hybrid multilevel parallelism based on the mix of MPI and a multithreaded version of the linear systems solver included in the code.File in questo prodotto:
Non ci sono file associati a questo prodotto.
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.