Migration of vectorized iterative solvers to distributed-memory architectures. This paper describes how PILS – an existing package for the iterative solution of large unstructured sparse linear systems of equations on vector computers – has been ported to distributed-memory parallel processors, using the parallelizing Fortran compiler Oxygen. PILS implements a large number of iterative methods, preconditioners, and other variants for iterative solvers and provides a high degree of flexibility, like automatic adaptation to more robust preconditioned iterative methods. Two distributed-memory parallel processors, namely an Intel Paragon and a Fujitsu AP1000, have been used to evaluate the performance of the generated parallel program quantitatively. Some experimental results are given. The results indicate how an application should be designed to be portable among supercomputers of different architecture. It is shown that applications like PILS can only be parallelized efficiently on distributed-memory parallel processors if the underlying architecture supports performing computation and communication engines equally well.
Keywords for this software
References in zbMATH (referenced in 4 articles , 2 standard articles )
Showing results 1 to 4 of 4.
- Hassan, Ahmed M.; El-Shenawee, Magda: Parallel implementation of the diffusion-drift algorithm for modeling the electrophysiological activity of breast tumors (2011) ioport
- Fokkema, Diederik R.; Sleijpen, Gerard L.G.; Van der Vorst, Henk A.: Generalized conjugate gradient squared (1996)
- Pommerell, Claude; Rühl, Roland: Migration of vectorized iterative solvers to distributed-memory architectures (1996)
- Pommerell, Claude; Fichtner, Wolfgang: Memory aspects and performance of iterative solvers (1994)