Abstract | ||
---|---|---|
Particle-based models are widely spread in the field of computer graphics, and mainly used for real-time simulations of soft deformable bodies. One of the high-demanding computational components of such physically-based simulations is numerical time integration. As the size of the models increases, this component becomes a potential bottleneck in the simulation process, thus parallelism must be deployed in order to preserve the real-time attribute. This paper presents an approach to parallel implicit numeric time integration for graphics clusters. To deal with very large models, we employ a static domain decomposition approach and we fully utilize the massively data parallel capabilities of GPU clusters. We discuss the parallel conjugate gradient algorithm used for time integration and we present a parallel method for generating the very large sparse matrices and vectors involved. |
Year | DOI | Venue |
---|---|---|
2014 | 10.1109/MIPRO.2014.6859587 | Information and Communication Technology, Electronics and Microelectronics |
Keywords | Field | DocType |
computer graphics,conjugate gradient methods,GPU clusters,computer graphics,data parallel capabilities,graphics clusters,numerical time integration,parallel conjugate gradient algorithm,parallel implicit numeric time integration,parallel implicit time integration,parallel method,particle-based models,physically-based simulations,real-time simulations,soft deformable bodies,sparse matrices,static domain decomposition approach | Graphics,Cluster (physics),Computer graphics (images),Computer science,Computational science,Particle | Conference |
Citations | PageRank | References |
1 | 0.40 | 7 |
Authors | ||
2 |
Name | Order | Citations | PageRank |
---|---|---|---|
Adrian Sabou | 1 | 1 | 0.40 |
Dorian Gorgan | 2 | 1 | 0.73 |