Efficient utilization of scratch-pad memory for embedded systems

Abstract

Hierarchy memory units are used in embedded systems for their different performance. The scratch-pad memory has been used to meet the real-time constraints in embedded systems. This paper presents an efficient compiler-assisted approach based on scratch-pad memory for heap and stack management of embedded systems. The stack and heap will be mapped to scratch-pad memory and some special code will be inserted into programs. Thus the heap and stack of the same program will be allocated to scratch-pad memory together. Results of the benchmarks show a 17% reduction in runtime from our strategy vs. using the strategy that the heap and stack are allocated to off-chip memory.