The advent of the Intel Delta machine, a large-scale parallel computer with powerful i860 nodes and high speed communications between the nodes, combined with fast algorithms in computational fluid mechanics, will allow unprecedented explorations, via numerical simulations, of complex flows with and without solid boundaries. Our computational method is based in the Lagrangian description of the vorticity field. Computational elements of vector-valued vorticity are convected with the local fluid velocity and the vorticity vector is continually strained by the local velocity gradient. The classical vortex method is enhanced to account for viscous effects. An adaptive, fast algorithm has been efficiently implemented in two-dimensions, in both vector (CRAYS) and parallel (Mark III hypercube) architectures allowing unprecedented simulations using up to O(10 6) computational elements. We propose to extend the viscous technique to three-dimensional flows in the presence of arbitrarily-shaped solid bodies and implement a fast three-dimensional algorithm on the Delta. We expect to obtain computing speeds that are much higher than those of a serial supercomputer speed at a fraction of the cost, allowing simulations using up to O(10 6) computational elements for three and two-dimensional simulations. This capability will allow large scale extended numerical simulations, in two and three dimensions, for laboratory Reynolds numbers and should lead to significant breakthroughs in the understanding and control of complex flow phenomena that are fundamental to unsteady bluff- body flows.
