Acceleration of High-Fidelity Wireless Network Simulations

摘要

Network simulation with bit-accurate modeling of modulation, coding and channel properties is typically computationally intensive. Simple link-layer models that are frequently used in network simulations sacrifice accuracy to decrease simulation time. We investigate the performance and simulation time of link models that use analytical bounds on link performance and bit-accurate link models executed in Graphical Processing Units (GPUs). We show that properly chosen analytical bounds on link performance can result in simulation results close to those using bit-level simulation while providing a significant reduction in simulation time. We also show that bit-accurate decoding in link models can be expedited using parallel processing in GPUs without compromising accuracy and decreasing the overall simulation time.