Predictive Lazy Amplification: Synthesis and Rendering of Massive Procedural Scenes in Real Time

In this paper we propose a new paradigm for procedural modeling that enables the real time visualization of massive procedural scenes. For this, we use a combination of memory and task management with two well known procedural modeling paradigms: data amplification and lazy evaluation. Experimental results show that, in addition to obtaining performance gains through parallelism, the implemented system can generate and visualize a procedural scene far greater than the available memory in real time using only a single PC equipped with a GPU and a multicore processor.