GPU acceleration for fixed complexity sphere decoder in large MIMO uplink systems

Large MIMO constitutes a principal technology for next generation wireless communication systems. Detection algorithms for large MIMO schemes tend to require long simulation times on conventional computer systems. In order to improve simulation efficiency of decoders in large MIMO systems, we propose to use General Purpose Graphic Processing Units (GPGPU). This paper considers the implementation of a Fixed Complexity Sphere Decoder (FCSD) for large scale MIMO uplink systems using the framework of GPGPU-Computed Unified Device Architecture (CUDA). Using GPGPU, significant speedups were found for large MIMO systems and constellation sizes, over conventional CPU implementation, while providing the same Bit Error Rate (BER) performance. Hence the proposed approach shows great potential of GPGPU to accelerate computer simulations of large MIMO systems.