Multi-objectiveabstract data type refinementfor mapping tables in telecom network applications

Abstract

We present a new multi-objective exploration method at the system level to select customized implementations for mapping tables, dynamically allocated, as encountered in telecom network, database, and multimedia applications. Our method fits in the context of embedded system synthesis for such applications, and it enables the optimization of the system-level memory management of these applications. To this end it mainly aims at trading off the average memory footprint, number of memory accesses, and memory power. Compared with existing methods, for large mapping tables, 90% (resp. 80%) of the average memory footprint (resp. power) can be saved, without decreasing the performance.