A Heterogeneous Graphics Procedure for Visualization of Massively Parallel Solutions *

P. J. Crossno, et. al. * Accepted for presentation at ASME Fluids Engineering Conference, June 20-24, 1993 Washington, D.C. ABSTRACT Scientific visualization is playing an increasingly important role in the analysis and interpretation of massively parallel CFD simulations due to the enormous volume of data that can be generated on these machines. In this paper we will describe the development of a visualization technique based on a parallel analogue to the Marching Cubes algorithm. The algorithm has been developed for Multiple-Instruction, Multiple-Data (MIMD) massively parallel computers and is designed to take advantage of the heterogeneous programming capabilities of the MIMD architecture. We examine several different configurations and conclude that for producing animations the best one, in terms of both frame generation time and disk usage, is to run the two applications heterogeneously and send the resulting geometry description directly to a workstation for rendering, thereby totally eliminating the use of files from the animation process.