Search

We discuss the use of component adaptive methods for the solution of time dependent partial differential equations. After a brief discussion of the methods we identify several deficiencies of commonly used languages of scientific computing and introduce a graphical editor, VOUS, and a language, Vorpal, to solve these problems. We also discuss the interaction of the various components of our system and their interfaces.

This paper outlines a design for implementing a linear programming system on the ILLIAC IV computer. The central concern is to take advantage of the special features of the ILLIAC IV (64 parallel processing elements, large fast disk memory and relatively small fast core memory) and at the same time to take advantage of the sparsity of real large-scale linear programs and the (mostly serial) methodology which has been developed to exploit this sparsity. This requires both the adaption of existing techniques to a parallel environment and the development of a new parallel techniques for efficient sparse matrix processing. It appears that this can be done successfully and that ILLIAC IV should be able to solve problems considerably larger than those which can be attempted on serial computers. (Author).