Parallelisation methodology for a turbine-blade design program

A design/production procedure for impeller turbochargers is used to demonstrate the benefits of parallel computing. The procedure is based on sequential fluid dynamics evaluation code. Parallel programs run many times faster than the sequential versions, and future engineering workstations based on multiprocessors will speed up the design/prototype cycle. A hybrid parallelisation technique is developed, including data partitioning and functional decomposition. The target parallel architecture is based on the distributed memory model. Transputers are powerful devices for parallel systems. The FAST9 multitransputer card was used for parallel Fortran implementation of the sequential 3D Navier-Stokes equations solver. A speed-up of five to six times on eight transputers was achieved (using fixed problem data size)