The objective of this project is to provide a methodology for mapping existing finite-element, finite-volume, modelling software packages for computational fluid dynamics semi-conductor simulation and computational electromagnetism applications onto parallel architectures, such as Supernode and Intel's Hypercube, and onto massively parallel architectures such as Connection Machine.
BECAUSE will be application-driven, and will complement several ESPRIT architectural development projects, such as SUPERNODE (project 1085) and GENESIS (2447), software-oriented development projects such as SUPERNODE II (2528) and TROPICS (2427), and CIM projects such as ACCORD (1062).
The development will use existing application packages as its starting point and address:
- different scenarios for the distribution of computational load among the processing elements of a multiprocessor
- different scenarios for mapping the physical discretisation network onto the processor network
The requirements of very large meshes will be established to allow computer manufacturers to incorporate them when addressing the scientific computation market.
OX11 0QX Didcot
78153 Le Chesnay
OX5 1JE Kidlington