| As the technology of Graphics Processing Unit (GPU) developing，GPU waswidely used because of its capability of floating-point calculations and parallelprocessing．GPU for general purpose computation (GPGPU) become a hot topic forresearch.Compute Unified Device Architecture (CUDA),which was proposed by NVIDIACorporation,is a development environment based on GPU general purposecomputation.In cognizance of the features of multi core of GPU,it could increase thespeed of the massive data computing through parallel computation.Adaptive equalization based on self-adaptive filter has been widely used in the fieldof telecommunication and radar.The digital filter is synthesized by means ofself-adaptive filtering principle,solving amplitude variation and phase shiftingmismatching among receiver channels effectively.In many cases, system calibration is expected to complete as soon as possible.It isvery necessary to improve the speed of calibration and reduce the time of calculation.Inthe process of system calibration,calculation using a lot of time. Reduce the time ofcalculation has great significance for calibration.The development of GPGPU and CUDA provides an effective method for solvingthis problem. The calculation of calibration is accelerated by using the capability offloating-point calculations and parallel processing．In this paper,the development and characteristics of GPU are described,and thesoftware and hardware patterns of Compute Unified Device Architecture (CUDA) fromNVIDIA are analyzed．An algorithm of broadband multichannel calibration based onadaptive equalization is studied，and the computation are implemented on GPU andCPU．The experimental results show that the computing speed of the proposed methods inmost cases is much faster than that of the CPU based methods, but in some casesslower.It indicates that the CUDA technology has both merits and demerits.A propermethod is applied to the broadband multichannel calibration,which improved theperformance of the system．... |
PDF Full Text Request