An Architecture for Distributed Behavioral Models with GPUs.

We describe an architecture for massive simulation of a distributed behavioral model using graphics hardware. By leveraging on the recent programmable capabilities of GPUs we implemented the model capable of managing a large aggregate motion of birds in a virtual environment that can avoid both static and dynamic obstacles. We demonstrate the effectiveness of our GPU implementation by comparing the results to a CPU implementation and, emphasize the modularity of the proposed architecture that favors reusability in several context.