The performance of the data parallel model is directly related to how large a part of the game engine can be parallelized by data. As the amount of processor cores goes up, the data parallel parts of the engine take less time to run. Fortunately these are usually also the performance heavy parts of a game engine. If an engine can use data parallelization for most parts of the game loop, then the data parallel model will give the best performance of the three models.

The biggest drawback of the model is the need to have components that support data parallelism. For example, a physics component would need to be able to run several physics updates in parallel, and be able to correctly calculate collisions with objects that are in these separate threads. A potential solution is to leave the physics calculation completely out of the data parallel part of the game loop, and use internal parallelization in the physics component. Fortunately many other components won't need extensive changes. For example a component calculating the animations of graphical objects has no interaction between concurrent threads, and won't need any special support for parallelism.

Conclusion

Because the synchronous function parallel model does not require special changes to engine components, and is really just an enhancement of a regular game loop, it is well suited for adding some amount of parallelism to an existing game engine. The model is not suited for future use because of it's weak scaling support and low amount of parallelism.

The asynchronous function parallel model can be recommended for new game engines because of the high amount of possible parallelism and the fact that existing components need only few changes. This model is a good choice for game engines aimed at the generation of multicore processors that have a relatively small number of cores. The only drawback is the need to tune thread running times to minimize the impact of worst case thread timings. More research is needed to know how this timing fluctuation actually affects game play.

The data parallel model is definitely something to think about for the future. It will be needed when the amount of processor cores increases beyond the number of tasks available for a function parallel model. For current use the increased scalability doesn't offer enough benefits compared to the trouble of coding custom components to support this type of data parallelism.

The current trend seems to be towards creating engine components that use some internal form of parallelization. While this allows engine developers to not worry about threading issues, it may leave large parts of the program sequential, which results in poor performance. The view presented in this article has been that whole game loops could be parallelized, not just some parts of them. The models presented here can be a good starting point for developing specialized game engine architectures.

Suggested Reading

[1] Gabb H., Lake A. 2005. Threading 3D Game Engine Basics. http://www.gamasutra.com/features/20051117/gabb_01.shtml

[2] Costa S. 2004. Game Engineering for a Multiprocessor architecture. MSc Dissertation Computer Games Technology. Liverpool John Moores University.

Wilson K. 2006. Managing Concurrency: Latent Futures, Parallel Lives. http://www.gamearchitect.net/Articles/ManagingConcurrency1.html