Computational modelling on high performance computers (HPC) is appearing as an efficient method in industrial product development. This ranges from seismic modelling of large scale geophysical properties for the oil industry to small scale simulation of polymer flow, such as injection molding of plastics or foods, coating of wires and mixing of dough. SAGA aims to reduce the development costs of computational modelling software by investigating and exploiting improved methodologies for computational model prototyping and software development, and for the porting of applications to high performance computer (HPC) architectures.
A selection of prototype tools supporting the methodologies has been developed. These include a compiler translating recursive mathematical equations to imperative code and a prototype object-oriented library for computational modelling. The methodologies should be of use for software companies in the computational modelling area, and also for companies having specific modelling problems they need to solve. Use of the methodologies requires skill and training, but the benefits can be significant. One case study indicates a tenfold increase in development productivity over conventional techniques using the SAGA methods. A seismic simulator has been fully developed using these methodologies, and is now being employed for geophysical studies.
Project URL : http://www.ii.uib.no/saga