The massive integration of cores in multi-core system has enabled chip designer to design systems while meeting the power-performance demands of the applications. However, full-system simulations traditionally used to evaluate the speedup with these systems are computationally expensive and time consuming. On the other hand, analytical speedup models such as Amdahl’s law are powerful and fast ways to calculate the achievable speedup of these systems. However, Amdahl’s Law disregards the communication among the cores that play a vital role in defining the achievable speedup with the multi-core systems. To bridge this gap, in this work, we present PaSE a parallel speedup estimation framework for multi-core systems that considers the latency of the Network-on-Chip (NoC). To accurately capture the latency of the NoC, we propose a queuing theory based analytical model.
Using our proposed PaSE framework, We conduct a speedup analysis for multicore system with real application based traffic i.g. Matrix Multiplication. the multiplication of two [32x32] matrices is considered in NoC based multi-core system with respect to three different cases ideal-core case, integer case (i.g. matrix elements are integer number), and denormal case (i.g. matrix elements are denormal number, NaNs, or infinity). From this analysis, we show how the system size, Network-on-Chip NoC architecture, and the computation to communication (C-to-C) ratio effect the achievable speedup.
To sum up, instead of the simulation based performance estimation, our PaSE framework can be utilized as a design guideline i.g. it is possible to use it to understand the optimal multi-core system-size for certain applications. Thus, this model can reduce the design time and effort of such NoC based multi-core systems.
Electrical Engineering (MS)
Department, Program, or Center
Electrical Engineering (KGCOE)
Dharb, Ghassan Bachay, "PaSE : Parallel Speedup Estimation Framework for Network-on-Chip Based Multi-core Systems" (2017). Thesis. Rochester Institute of Technology. Accessed from
RIT – Main Campus