Adaptive deferred shading

Abstract

The primary advantage of deferred shading is eliminating wasted shading operations for fragments that are occluded by others during rendering. Deferred shading starts by rendering the entire scene into a series of temporary buffers (a geometry buffer or G-buffer). Then, it performs the shading operations using the contents of these buffers. Hence, deferred shading executes the (potentially expensive) fragment shader exactly once for a given pixel sample, avoiding additional fragment shading invocations due to overdrawing. Although deferred shading requires an extra pass and significant memory overhead to store the G-buffer, it can still be beneficial to performance when fragment shading is sufficiently expensive.