Date of this Version
In addition to real-time requirements, the program code size is a critical design factor for real-time embedded systems. To take advantage of the code size vs. execution time tradeoff provided by reduced bit-width instructions, we propose a design framework that transforms the system constraints into task parameters guaranteeing a set of requirements. The goal of our design framework is to derive the temporal parameters and the code size parameter of each task in such a way that they collectively guarantee the system end-to-end timing requirements while the system code size is minimized. Our design framework is based on asynchronous periodic tasks with pre-period deadlines under EDF scheduling. For schedulability analysis, we present a new feasibility condition that can be more efficiently evaluated than existing ones. When the code size vs. execution time tradeoff can be safely approximated as linear functions, the minimization problem becomes a linear programming problem. However, when the tradeoff is given by a table of possible (code size, execution time) pairs, the problem becomes NP-hard. We provide three heuristic algorithms that can find sub-optimal solutions and evaluate their performance with simulation results.
Date Posted: 09 November 2004
This document has been peer reviewed.