A GPU-accelerated physical layer for simulating wireless networks

In recent years, graphics processing units (GPUs) have been leveraged to speed-up massively parallel computations. Knowing the path loss between nodes in a wireless network is crucial in accurately simulating physical layer effects in wireless network simulators and emulators. Because path loss and interference calculations are often repeated for every transmitter and receiver pair, leveraging GPU computing to parallelize these calculations can lead to significant reduction in processing time. In this paper, we present an implementation of a high-fidelity GPU-accelerated PHY that calculates path loss and interference over time for every receiver/transmitter pair using realistically defined node antenna patterns. Performance results are compared against traditional CPU calculations and we demonstrate that by offloading parallel computations to the GPU, significant gains can be had for wireless network simulation and emulation. Additionally, GPU limitations and design considerations are presented to aid in future GPU-based wireless simulation implementations.