Adaptive quadrature on a message-passing multiprocessor (Q1200129)

An algorithm based on an adaptive Simpson's rule modified to efficiently use a message-passing multiprocessor network is presented to study the problem of numerically integrating a continuous function over a closed interval, i.e. to compute \(\int^ b_ a f(x) dx\approx \sum^ n_{i=1} a_ if(x_ i)\). Such a network consists of a set of independent, asynchronous processors (nodes). Data and results are sent from node to node. All nodes are connected to a central host processor which directs the startup, loading of code, problem distribution and shutdown. Four redistribution strategies are considered and three of them (called Avail, Skip, Noskip) are discussed in more details. The serial quadrature algorithm and its parallel version are listed. Future work is mentioned. In the Appendix 20 integrals are given to test the algorithm.