StackMMU: Dynamic stack sharing for embedded systems

Abstract

Real-time multi-tasking systems may require an individual stack for each task to fulfill all hard real-time requirements. However, these stacks may consume a huge memory space, even if not all stacks are simultaneously fully utilized. Thus, sharing currently unused stack space may improve memory utilization as possible with Memory Management Units (MMUs). However, an MMU introduces temporal jitter to memory accesses, influencing the real-time behavior. In this work, we propose a new concept to share dynamically the complete available stack space across tasks. Thereby, every stack operation executes in a deterministic time, by giving the Microcontroller Unit (MCU) Operating System (OS)-awareness.