GPU-based fast error recovery for high speed data communication in media technology

Abstract

The heterogeneous wireless network is an essential constituent of the emerging advanced active media technology (AMT). However, reliable wireless data communication is a challenge since a packet is easily corrupted by noise or interference. As a remedy, error control mechanisms are adopted and ever-growing networks demand a high speed solution. In this paper, we present a faster implementation of the extended Hamming code, a widely-used error correction algorithm, using graphics processing units (GPU). A GPU performs parallel computations by employing a cluster of processors, and can operate on both single bit and multiple bit errors. We compare the performance of the GPU-based approach with the equivalent sequential algorithm that runs on traditional CPUs with regards to error strength, t, such that 1≤t≤7. Experimental results demonstrate that the GPU-based approach significantly outperforms the CPU-based approach in terms of execution time and buffer memory requirement. Furthermore, the proposed approach reduces the computational complexity from O(n) for CPUs to O(1) for the GPU-based approach, yielding significant increases in speed.