| 摘要 |
Methods for vertex caching to decrease geometry bandwidth and to reduce bus traffic between a graphics subsystem and memory include a strip-growing technique and a local optimization technique. The strip-growing technique determines an ordering of the faces in memory for the purpose of maximizing the use of the cache. This technique minimizes the number of vertices that are retrieved from a vertex buffer, and maximizes the number of needed vertices that are obtained from a vertex cache. The local optimization technique improves the results of the strip-growing technique by exploring a set of perturbations to the face ordering. The order is perturbed semi-randomly to determine if the perturbation improves the caching behavior. Types of perturbations include reflection and insertion. Thus, data is preprocessed to optimize the use of the cache stored data so that when the data is rendered at a future time, the rendering speed is improved.
|
