Adaptive resource scheduling for energy efficient QRD processor with DVFS

Abstract

This paper presents an energy efficient adaptive QR decomposition scheme for Long Term Evolution Advance (LTE-A) downlink system. The proposed scheme provides a performance robustness to fluctuating wireless channels while maintaining lower workload on a reconfigurable hardware. A statistic based algorithm-switching strategy is employed in the scheme to achieve workload reduction and stable computing resource requirement for QR decomposition. With run time resource allocation, computing resources are assigned to highest performance gain segments to reduce performance loss. By utilizing the dynamic voltage and frequency scaling (DVFS) technique, we further exploit the potential of power saving in various workload situation while maintaining fixed throughput. The proposed technique brings power reduction upto 57.8% in EVA-5 scenario and 24.4% with a maximum SNR loss of 1 dB in EVA-70 scenario, when mapped on a coarse grain reconfigurable vector-based platform.