No abstract

No abstract

Abstract. This paper describes how the specific access structure of the Brusselator equation, a typical example for ordinary differential equations (ODEs) derived by the method of lines, can be exploited to obtain scalable distributed-memory implementations of explicit Runge-Kutta (RK) solvers. These implementations need less communication and therefore achieve better speed-ups than general explicit RK implementations. Particularly, we consider implementations based on a pipelining computation scheme leading to an improved locality behavior.

SUMMARYSince a static work distribution does not allow for satisfactory speed-ups of parallel irregular algorithms, there is a need for a dynamic distribution of work and data that can be adapted to the runtime behavior of the algorithm. Task pools are data structures which can distribute tasks dynamically to different processors where each task specifies computations to be performed and provides the data for these computations. This paper discusses the characteristics of task-based algorithms and describes the implementation of selected types of task pools for shared-memory multiprocessors. Several task pools have been implemented in C with POSIX threads and in Java. The task pools differ in the data structures to store the tasks, the mechanism to achieve load balance, and the memory manager used to store the tasks. Runtime experiments have been performed on three different shared-memory systems using a synthetic algorithm, the hierarchical radiosity method, and a volume rendering algorithm.

The estimation and evaluation of the energy consumption of computers is becoming an important issue. In this article, we address the question how the energy consumption for computations can be captured by an analytical energy consumption model. In particular, we address the possibility to reduce the energy consumption by dynamic frequency scaling and model this energy reduction in the context of task execution models. We demonstrate the use of the model by simulating task executions and their energy consumption.