VERFAHREN UND SYSTEM ZUR WIEDERGABE VON GRAFISCHEN OBJEKTEN DURCH TEILUNG IN BILDSTÜCKE UND ZUSAMMENSETZEN VON BILDLAGEN ZU EINEM WIEDERGABEBILD

摘要

To render a display image, the system serially renders chunks of the gsprites, and composites the gsprites to compute a display image. To reduce rendering overhead, the system can perform affine transformations on gsprites to simulate the motion of a 3D object rather that re-rendering the object for frames animation. Rendering geometry in chunks enables sophisticated fragment list anti-aliasing. The system stores fragments representing partially covered pixel locations or translucent pixels in a fragment buffer. After rasterizing primitives for a chunk, a fragment resolution subsystem resolves the fragments to compute output pixels. The rasterizing component of the system attempts to merge fragments to save fragment memory. If the fragment memory is exceeded, the system can subdivide chunks into smaller regions and render these smaller regions independently. The system supports texture accessed in environments with high latency such as in cases where texture data is compressed. The latency of texture accessing is reduced using either a texture reference or "pixel queue" to buffer partially rendered pixel data as texture data is fetched from memory, or using a pre-rasterizer to generate texture requests and a post-rasterizer to rasterize primitives completely using texture data fetched as a result of the texture requests generated by the pre-rasterizer. The system support anisotropic filtering of texture data by repetitively sampling texture data along a line of anisotropy computed for pixel location mapped into a texture or MIP mapped textures.