Code decomposition and recomposition for enhancing embedded software performance

Multitasking of concurrent processes implements the concurrency inherited from applications, increasing the utilization of limited resources. It requires an operating system and imposes significant runtime overhead. Serializing multitasking codes removes the need of operating system and the overhead as well. In this paper, we propose a software synthesis method to transform multitasking codes into a single process code. For this, we decompose multitasking codes into a set of code fractions and then recompose the code fractions into a single process code, preserving the functionality of the original codes. We present two different techniques for the transformation - code partitioning and code covering - and propose a hybrid technique that combines the two techniques.