Table of Contents - 8 Acknowledgments - 6 1 Introduction 1.1 Related work in system-level design 1.2 Organization and contributions of this thesis 2 The Sesame environment 2.1 Trace-driven co-simulation 2.2 Application layer 2.3 Architecture layer 2.4 Mapping layer 2.5 Implementation aspects 2.5.1 Application simulator 2.5.2 Architecture simulator 2.6 Mapping decision support 2.7 Obtaining numbers for system-level simulation 2.8 Summary 3 Multiobjective application mapping 3.1 Related work on pruning and exploration 3.2 Problem and model definition 3.2.1 Application modeling 3.2.2 Architecture modeling 3.2.3 The mapping problem 3.2.4 Constraint linearizations 3.3 Multiobjective optimization 3.3.1 Preliminaries 3.3.2 Lexicographic weighted Tchebycheff method 3.3.3 Multiobjective evolutionary algorithms (MOEAs) 3.3.4 Metrics for comparing nondominated sets 3.4 Experiments 3.4.1 MOEA performance comparisons 3.4.2 Effect of crossover and mutation 3.4.3 Simulation results 3.5 Conclusion 4 Dataflow-based trace transformations 4.1 Traces and trace transformations 4.2 The new mapping strategy 4.3 Dataflow actors in Sesame 4.3.1 Firing rules for dataflow actors 4.3.2 SDF actors for architecture events 4.3.3 Token exchange mechanism in Sesame 4.3.4 IDF actors for conditional code and loops 4.4 Dataflow actors for event refinement 4.5 Trace refinement experiment 4.6 Conclusion 5 Motion-JPEG encoder case studies 5.1 Sesame: Pruning, exploration, and refinement 5.2 Artemis: Calibration and validation 5.3 Conclusion 6 Real-time issues 6.1 Problem definition 6.2 Recurring real-time task model 6.2.1 Demand bound and request bound functions 6.2.2 Computing request bound function 6.3 Schedulability under static priority scheduling 6.4 Dynamic priority scheduling 6.5 Simulated annealing framework 6.6 Experimental results 6.7 Conclusion 7 Conclusion A Performance metrics B Task systems References Nederlandse samenvatting Scientific output Biography