Program compaction for real-time applications

This paper presents a method of program compaction based on the frequency characteristics of program behavior. The proposed method keeps in the compiled form only frequently executed portions of programs and stores infrequently executed portions of programs in a compacted interpreted form, and dynamically unpacks and loads them into the memory for execution only when they are requested. Determination of infrequently executed code is done by estimating the execution frequency of the program's basic blocks based on the known distribution functions of its input parameters. It allows to control the growth of the compacted program execution time. The theoretical and experimental results of the research prove the possibility of using the proposed program compaction method in real-time systems