Patent attributes
One embodiment of the present invention sets forth a technique for efficiently creating and accessing an A-Buffer with a GPU. The A-Buffer is organized in arrays of uniformly-sized tiles. Each array represents a group of pixels, and each tile within an array includes the set of fragments at a specific depth complexity that are associated with the pixels in the pixel group represented by the array. The size of the tiles may be selected to be the minimum necessary for efficient memory access. The GPU determines the number of tiles in each array by calculating the maximum of the depth complexity associated with the pixels in the pixel group represented by the array and creates a corresponding prefix sum image to allow the GPU to efficiently locate the array associated with a given pixel group in the A-Buffer for addressing purposes. Advantageously, structuring the A-Buffer by both pixel proximity and depth complexity improves memory locality, thereby improving the overall rendering performance of the graphics pipeline.
