Parallel Implementation Strategies for MIMO ID-BICM Systems

Abstract

[EN] One of the current techniques proposed for multiple transmit and receive antennas wireless communication systems is the use of error control coding and iterative detection and decoding at the receiver. These sophisticated techniques produce a significant increase of the computational cost and require large computational power. The use of modern computer facilities as multicore and multi-GPU (Graphics Processing Unit) processors can decrease the computational time required, representing a promising solution for the receiver implementation in these systems. In this paper we explain how iterative receivers can improve the performance of suboptimal detectors. We also introduce a novel parallel receiver scheme based on a hybrid computing model where CPUs and GPUs work together to accelerate the detection and decoding steps; this design comes to exploit the features of the GPU NVIDIA Kepler architecture respect to the previous one in order to optimize the communication system performance.