Abstract
Graphics rendering is a complex, multi-step process whose data demands typically dominate memory system design in SoCs. GPUs create images by merging many, simpler scenes for each frame. For performance, scenes are tiled into parallel tasks, each of which produces different parts of the final output. This execution model results in complex memory behavior, whose bandwidth demands, reuse and sharing characteristics depend heavily on the structure and complexity of each application, frame, scene, and task, and vary over time. To design systems that can efficiently accommodate and schedule these workloads, we need to understand their behavior and diversity. In this work, we explore the data demands of modern graphics rendering quantitatively, using an architecturally-independent analysis that identifies the different types of data sharing present in the applications.