Abstract
The use of many-core processors such as general purpose Graphic Processing Units (GPUs) has recently become attractive for the efficient implementation of signal processing algorithms for communication systems. This is due to the cost-effectiveness of GPUs together with their potential capability of parallel processing. This paper presents an implementation of the widely employed fixed-complexity sphere decoder on GPUs, which allows to considerably decrease the computational time required for the data detection stage in multiple-input multiple-output systems. Both, the hard-and soft-output versions of the method have been implemented. Speedup results show the proposed GPU implementation boosts the runtime of the parallel execution of the methods in a high performance multi-core CPU. In addition, the throughput of the algorithm is evaluated and is shown to outperform other recent implementations and to fulfill the real-time requirements of several LTE configurations. ©2012-IOS Press and the authors. All rights reserved.
