Interactive Physical Simulation on Multicore Architectures

Abstract

In this paper we propose a parallelization of interactive physical simulations. Our approach relies on a task parallelism where the code is instrumented to mark tasks and shared data between tasks, as well as parallel loops even if they have dynamic conditions. Prior to running a simulation step, we extract a task dependency graph that is partitioned to define the task distribution between processors. To limit the overhead of graph partitioning and favor memory locality, we intend to limit the partitioning changes from one iteration to the other. This approach has a low impact on physics algorithms as parallelism is mainly extracted from the coordination code. It makes it non parallel programmer friendly. Results show we can obtain good performance gains.