Let us consider an example demonstrating the concept of Pareto optimality. In this example, we study the results generated by task concurrency management (TCM) tools designed at the IMEC research center (Interuniversitair Micro-Electronica Centrum). TCM tools aim at establishing efficient mappings from applications to processors. Different multiprocessor systems are evaluated and represented as sets of Pareto optimal designs. Wong et al. [595] describe different options for the design of an MPEG-4-player. The authors assume that a combination of StrongARM processors and specialized accelerators should be used. Four designs meet the timing constraint of 30 ms (see Table 5.4). These different designs are shown in Fig. 5.32. For combinations 1 and 4, the authors report that only one mapping of tasks to processors meets the timing constraints. For combinations 2 and 3, different time budgets lead to different task to processor mappings and different energy consumptions.

Which area in the objective space is dominated by at least one design of configuration 3? Is there any design belonging to configuration 2 which is not dominated by at least one design of configuration 3? Which area in the objective space dominates at least one design of configuration 3?

Transcribed Image Text:

Configuration 2 490- 480- 470- 460- 450- 440- 430- Configuration 3 Energy [mJ] 420- Pareto-optimal design points 410- 5 400- 3 20 To 390- T 24 25 26 27 28 29 30 Time [ms] Fig. 5.32 Pareto points for multiprocessor systems 2 and 3 Table 5.4 Processor configurations Processor combination 1 2 3 4 Number of high-speed processors 6 5 4 3 Number of low-speed processors Total number of processors 03 57 6 8 9 10