Game Development Reference
designed for stochastic rasterization. Unfortunately on current GPUs we need
to implement stochastic rasterization and the shading cache using shaders, to
overcome the limitations of the hardware pipeline. We consider our results ben-
eficial for interactive applications, where shading cost dominates the rendering,
however, the overhead of the global cache implementation is generally too high
for real-time rendering.
We expect that the major synchronization bottleneck will disappear in future
rendering architectures. While we cannot predict whether future GPUs would
have a hardware-accelerated version of the memoization cache, some way of lo-
cal synchronization among fragement shaders would already remove most of the
overhead. Using a tile-based rendering architecture instead of sort-last-fragment
would allow us to use a more ecient, per-tile on-chip shading cache.
In our examples we have assumed that the visible color of surfaces remains
constant in a single frame, and shading can be prefiltered. This might cause ar-
tifacts on fast-moving surfaces, therefore we could extend our method to support
interpolation among temporal shading samples. In the future it will be interesting
to separate the frequency content of shading itself: a hard shadow edge in fact
cannot be prefiltered, but there are low-frequency components of shading, e.g.,
diffuse indirect illumination, where sparse shading can bring relevant speedup.
We would like to thank Anton Kaplanyan and Balazs Toth for the helpful discus-
sions during the development of this project. Gabor Liktor is funded by Crytek
[Chajdas et al. 11] Matthaus G. Chajdas, Morgan McGuire, and David Luebke.
“Subpixel Reconstruction Antialiasing for Deferred Shading.” In Proceedings
of Symposium on Interactive 3D Graphics and Games , pp. 15-22. New York:
[Dachsbacher and Stamminger 05] Carsten Dachsbacher and Marc Stamminger.
“Reflective Shadow Maps.” In Proceedings of the 2005 Symposium on Inter-
active 3D Graphics and Games , pp. 203-231. New York: ACM, 2005.
[Laine and Karras 11] Samuli Laine and Tero Karras. “High-Performance Soft-
ware Rasterization on GPUs.” In Proceedings of the ACM SIGGRAPH Sym-
posium on High Performance Graphics , pp. 79-88. New York: ACM, 2011.
[Liktor and Dachsbacher 12] Gabor Liktor and Carsten Dachsbacher.
pled Deferred Shading for Hardware Rasterization.”
In Proceedings of the