ABSTRACT
As seen in the previous chapter, iterative methods can be used on a large class of numerical problems. However, in numerous scientific applications, the size of the problem and/or the amount of required computations implies the use of a parallel system. Unfortunately, or fortunately, there is neither a single kind of parallel system nor a single kind of parallel iterative algorithm. Hence, the subject of this chapter is twofold: the first goal is to present the most common kinds of parallel architectures which can be encountered throughout the world and the second goal is to provide a classification of the parallel iterative algorithms. A brief review of the evolution of parallel systems is given in Section 3.1.
Then, Section 3.2 presents the classical parallel architectures and the main features which differentiate them. In Section 3.3, the trends of used configurations are discussed. Finally, a classification of parallel iterative algorithms is proposed in Section 3.4 with a focus on their respective advantages and drawbacks.
