A Genetic Programming Approach to Shader Simplification

Report
Authors:SitthiAmorn, Pitchaya, Department of Computer ScienceUniversity of Virginia Modly, Nicholas, Department of Computer ScienceUniversity of Virginia Wiemer, Wesley, Department of Computer ScienceUniversity of Virginia Lawrence, Jason, Department of Computer ScienceUniversity of Virginia
Abstract:

The programmability of modern graphics hardware has led to an enormous increase in pixel shader complexity and a commensu- rate increase in the difficulty and effort needed to optimize them. We present a framework based on Genetic Programming (GP) for automatically simplifying pixel shaders. Our approach computes a series of increasingly simplified shaders that expose the inher- ent trade-off between speed and accuracy. Compared to existing automatic methods for shader simplification [Olano et al. 2003; Pellacini 2005], our approach is powerful, considering a far wider space of code transformations and producing faster and more faith- ful shaders. We further demonstrate how our cost function can be rapidly evaluated using graphics hardware, which allows tens of thousands of shader variants to be considered during the opti- mization process. Our approach is also general: unlike previous work, we demonstrate that our technique is applicable to multi-pass shaders and perceptual-based error metrics.

Rights:
All rights reserved (no additional license for public reuse)
Language:
English
Source Citation:

SitthiAmorn, Pitchaya, Nicholas Modly, Wesley Wiemer, and Jason Lawrence. "A Genetic Programming Approach to Shader Simplification." University of Virginia Dept. of Computer Science Tech Report (2011).

Publisher:
University of Virginia, Department of Computer Science
Published Date:
2011