REHOSTING A MINICOMPUTER MODEL ON A SUPERCOMPUTER

Many ecosystem simulation computer codes have been developed in the la st twenty-five years. This development took place initially on main-fr ame computers, then mini-computers, and more recently, on micro-comput ers and workstations. Supercomputing platforms (both parallel and dist ributed systems) have been largely unused, however, because of the per ceived difficulty in accessing and using the machines. Also, significa nt differences in the system architectures of sequential, scalar compu ters and parallel and/or vector supercomputers must be considered. We have transferred a grassland simulation model (developed on a VAX) to a Cray Y-MP C90. We describe porting the model to the Cray and the cha nges we made to exploit the parallelism in the application and improve code execution. The Cray executed the model 30 times faster than the VAX and 10 times faster than a Unix workstation. We achieved an additi onal speedup of 30 percent by using the compiler's vectorizing and ''i n-line'' capabilities. The code runs at only about 5 percent of the Cr ay's peak speed because it ineffectively uses the vector and parallel processing capabilities of the Cray. We expect that by restructuring t he code, it could execute and additional six to ten times faster. Our goal was not just to increase the speed of code execution, but to enab le the restructured and ported code to access and manipulate large dat a matrices holding intermediate and state variables, to increase the s ize of the geographical areas that can be simulated, and to be able to use large remote sensing data sets to drive the model or use as valid ation.