1. Introduction  
  1.1 Heterogeneous Multi-Core Architectures  
  1.2 Need for Adaptivity  
    1.2.1 Adaptivity at the Resource Management Level  
    1.2.2 Adaptivity at the Application Level  
  1.3 Contributions of This Thesis  
  1.4 Synopsis and Outline  
2. Foundations of Application Mapping onto HMAs  
  2.1 Preliminaries  
    2.1.1 Mapping, Scheduling, and Spatio-Temporal Mapping  
    2.1.2 Performance and Energy Estimation  
    2.1.3 Static and Dynamic Power Consumption  
    2.1.4 Notation  
  2.2 System Model  
    2.2.1 Architecture  
    2.2.2 Application  
    2.2.3 Mapping  
  2.3 Mapping of Dataflow Applications  
    2.3.1 Dataflow Models of Computation  
    2.3.2 Design-time and Runtime Mapping Approaches  
    2.3.3 Trace-Based Simulation  
  2.4 Hybrid Application Mapping  
    2.4.1 Pareto-Optimal Operating Points  
    2.4.2 Spatial Mapping Optimization  
    2.4.3 Spatio-Temporal Mapping Optimization  
    2.4.4 Addressing Adaptivity Challenges  
  2.5 Mocasin Framework  
    2.5.1 Overview of Mocasin  
    2.5.2 Contributions to Mocasin  
  2.6 Synopsis  
3. Related Work  
  3.1 Design-Time Application Mapping  
  3.2 Runtime Application Mapping  
    3.2.1 Runtime Mapping of Embedded Software  
    3.2.2 Runtime Mapping within OS Schedulers  
  3.3 Hybrid Application Mapping  
  3.4 Application Adaptivity  
4. Efficient Spatio-Temporal Mapping Generation  
  4.1 Motivational Example  
  4.2 Spatio-Temporal Mapping Strategies  
  4.3 Fixed-Point Spatio-Temporal Mapping  
    4.3.1 MMKP-based Algorithm  
    4.3.2 Evaluation  
  4.4 Flexible Spatio-Temporal Mapping  
    4.4.1 STEM: Spatio-Temporal Evolutionary Mapping  
    4.4.2 FFEMS: Fast Flexible Energy-Minimizing Scheduler  
    4.4.3 Evaluation  
  4.5 Synopsis  
5. Domain-Specific Hybrid Mapping for Baseband Processing  
  5.1 Approaches to Baseband Processing  
  5.2 Baseband Processing Architecture and Parameterization  
  5.3 Task Graph with Phase-Sequential Structure  
  5.4 Efficient Mapping Algorithm for Phased Task Graphs  
  5.5 Spatio-Temporal Mapping Reusing Previous Solutions  
  5.6 Evaluation  
    5.6.1 Platform Setup  
    5.6.2 Workload Model  
    5.6.3 Generation and Estimation of Operating Points  
    5.6.4 Energy-Efficient Runtime Mapping  
  5.7 Synopsis  
6. Extending Kahn Process Networks with Adaptivity  
  6.1 Limitations of KPNs: A Motivational Example  
  6.2 Adaptive Process Network  
    6.2.1 Parallel Regions  
    6.2.2 Parallel Channels and Workload Distribution  
    6.2.3 Malleability  
  6.3 Dynamic Process Manager (DPM) Library  
    6.3.1 Programmer Interface  
    6.3.2 Runtime Topology  
    6.3.3 Configuration Management  
  6.4 Evaluation  
    6.4.1 Experimental Setup  
    6.4.2 Performance Scalability with Parallelization  
    6.4.3 Runtime Adaptivity  
  6.5 Synopsis  
7. Coordinating Adaptivity in General-Purpose Environments  
  7.1 Need for Two-Way Communication  
  7.2 Adapting HAM Methodologies  
  7.3 HARP Design  
    7.3.1 Application Support via libharp  
    7.3.2 Resource Allocation  
  7.4 Runtime Exploration of Operating Points  
    7.4.1 Runtime Performance and Power Monitoring  
    7.4.2 Selection of the Regression Model  
    7.4.3 Runtime Exploration Algorithm  
  7.5 Evaluation  
    7.5.1 Experimental Setup  
    7.5.2 Intel Raptor Lake Evaluation  
    7.5.3 Odroid-XU3 Evaluation  
    7.5.4 Evaluation of the Learning Process  
    7.5.5 Performance Overhead of HARP  
  7.6 Synopsis  
8. Conclusions and Outlook  
Glossary  
List of Figures  
List of Tables  
List of Algorithms  
List of Listings  
Bibliography