A Theoritical Review on Static and Dynamic Scratchpad Memory

Abstract

Cache memory has been broadly used to improve the s. However, due to some weaknesses of caches, especially the unpredictability memory access latency, many embedded systems today use a small and fast on- chip memory, known as scratchpad memory (SPM). The allocation of data in scratchpad lies in the hand of software and is performed during the compilation times, which makes the memory access latency predictable. In any real-time systems, this is the major cause that will reduce the worst-case execution time (WCET) of each task. Hence, by knowing the most beneficial data to be the member of scratchpad residents, more chances there are to reduce the execution time of a task. The SPM allocation strategies can be classified into two categories static and dynamic. Both techniques aim for better timing constraint and increase performance of the execution time. Static and dynamic allocation techniques can be implemented on different types of system real-time and non real-time systems. In this paper, we present the comparison for the two types of scratchpad allocation techniques on two different types of systems.