Hotspot-aware task-resource co-allocation for heterogeneous many-core networks-on-chip

To fully exploit the massive parallelism of many-core on a chip, this work tackles the problem of mapping large-scale applications onto heterogeneous networks-on-chip (NoCs) while minimizing hotspots. A task-resource co-optimization framework is proposed which configures the on-chip communication infrastructure and maps the applications simultaneously and coherently, aiming to minimize the peak energy under the constraints of computation power, communication capacity, and total cost budget of on-chip resources. The problem is first formulated into a linear programming model to search for optimal solution. A heuristic is further developed for fast design space exploration at design-time and run-time in large-scale NoCs. Extensive simulations are carried out under real-world benchmarks and randomly generated task graphs to demonstrate the effectiveness and efficiency of the proposed schemes. Real system simulations show the significant improvement (30-200%) in NoCs latency and throughput compared to the state-of-the-art minimum-path approach because of the diminishing hotspots and balanced load distribution.