In wireless sensor networks, each sensor node has very small memory space compared with any other embedded computing systems. For this reason, operating systems running on the sensor nodes cannot allocate sufficient fixed-size stack space for all threads. In the previous work, SESAME was proposed to allocate stack space more space-efficiently, but there is a problem of time overhead. In this paper, we present SESAME-P, which is a dynamic stack allocation scheme based on memory pool. The size of memory pool is predetermined by using static analysis of each function's stack usage information. Using the determined memory pool, SESAME-P reduces the dynamic stack allocation cost. Our experimental results show that SESAME-P significantly reduces the time overhead compared with the existing SESAME.
展开▼
