VM_pRAY an efficient ray tracing algorithm on a distributed memory parallel computer
Résumé
The production of realistic image generated by computer requires a huge amount of computation and a large memory capacity. The use of highly parallel machines allows this process to be performed faster. Distributed memory parallel computers (DMPCs), such as hypercubes or transputer-based machines, offer an interesting performance/cost ratio when assuming that a load balancing and a partition of the data domain have been performed. This paper presents a software running on an iPSC/2 whose name is VM-pRAY. The aim of this experimental software is to show that using a virtual shared memory is an efficient strategy for parallelizing algorithms which, like ray tracing, use large read-only databases with no obvious data domain decomposition. The source code of VM-pRAY for an hypercube iPSC/2 is given in the appendix.