Concurrent Depth Surface Rendering Across GPU Pipelines

The patent divides the depth-only or Z-only workload into multiple segments and distributes them across different graphics pipelines (e.g., binning render pipeline and binning visibility pipeline). This segmentation allows parallel processing of depth surfaces, increasing GPU utilization while reducing the memory bandwidth demand on any single pipeline by sharing the load across multiple pipelines.

The patent makes multiple graphics pipelines (binning render pipeline and binning visibility pipeline) universally capable of handling depth-only or Z-only workloads. By enabling these pipelines to share the depth surface processing tasks, the system achieves better hardware resource utilization without significantly increasing complexity, as the pipelines already exist for other purposes.

The patent implements a self-service mechanism where the binning visibility pipeline automatically identifies and processes depth-only or Z-only surfaces that are not rendered by the binning render pipeline. This self-service approach enables concurrent execution across multiple pipelines without requiring complex centralized workload assignment management, as each pipeline independently determines its own tasks based on the workload characteristics.