Implementing finite difference ocean circulation models on MIMD, distributed memory computers

This paper considers the use of parallel computers for ocean modelling. After a brief review of the topic, the authors describe the experience of porting a simplified ocean model onto the Computing Surface. The parallel implementation is based on a straightforward domain decomposition. The use of CSTools in the code is briefly discussed. The performance of the parallel code, when run on Inmos T800 transputers and Intel 1860's, is compared with the performance of a serial implementation when run on a range of commonly used serial computers (including a CRAY X-MP and an AMDAHL VP1200).