Computational Complexity Analysis of PTS Technique under Graphics Processing Unit

Abstract

Peak-to-Average Power Ratio (PAPR) reduction is a important signal processing aspect of Orthogonal Frequency Division Multiplexing (OFDM). Minimization of PAPR can be achieved efficiently using PTS Technique. The main drawback of PTS lies in its computational complexity due to large order of multiple IFFT and phase factor computation. Typically simulation to tune 4 sub-blocks with 4 phase factors in PTS is achievable using standard computing hardware. In this paper we present performance comparison of PTS technique using different number of sub-blocks (V=2, 4, 6 and 8) with different phase factors (W=2, 4 and 8) under Central Processing Unit (CPU) and Graphics Processing Unit (GPU) environments. The performance gain of GPU over CPU in terms of speedup is presented and the computational complexity involved is analyzed. The GPU is approx 3.7×, 7×, and 9× faster than CPU in case of 2, 4, and 8 phase factors respectively.