Parallelization of the molecular dynamics code GROMOS87 for distributed memory parallel architectures

This paper describes the techniques and methodologies employed during parallelization of the Molecular Dynamics (MD) code GROMOS87, with the specific requirement that the program run efficiently on a range of distributed-memory parallel platforms. We discuss the preliminary results of our parallel port and indicate to what extent we are able to satisfy the basic requirements of industrial users with regard to code performance and scalability. We introduce a scheme which is used to dynamically allocate charge groups to processors reducing the time imbalance during the non-bonded force calculation phase from 30 - 40% to <1%. We propose a new method for parallelizing the constraint dynamics algorithm (SHAKE) where the emphasis is placed on reducing the number and size of communicated packets during the iterative SHAKEing process.