PARALLELIZATION STRATEGIES AND EXPERIENCES WITH THE DYNAMICALLY DEFINED REACTION-PATH (DDRP) METHOD

The possibilities and advantages of parallel realizations of the ''Dyn amically Defined Reaction Path'' (DDRP) method on large (Single Instru ction Multiple Data - SIMD) and small (Multiple Instruction Multiple D ata - MIMD) computer architectures, together with some runtime estimat es and simulations, are discussed, preceded by a short theoretical int roduction referring to the basic mathematical concepts, a description of the algorithm and a numerical realization of the general DDRP proce dure. The main difficulty in getting optimal runtimes when using the m ethod of small steps was found to be in the storage strategy of SCF da ta. (C) 1997 Elsevier Science B.V.