A parallel selection sorting algorithm on GPUs using binary search

Abstract

This paper describes a hybrid sorting which is the combination of radix sort and selection sort on graphic processing unit (GPU). The proposed algorithm is based on “Split and Concurrent Selection” (SCS) strategy. First, the data sequence is split in several pieces that are sorted in parallel using Radix sort. After that it applies parallel selection sort to obtain the final sorted sequence. Parallel selection sort finds the correct position of each elements of a data sequence and then copy the elements of a data sequence to corresponding position to obtain the final sorted data sequence. This paper analyses the computational complexity of proposed parallel sorting algorithm and compares it with other existing algorithms. It is implemented using CUDA 5.0 and results are evaluated on Tesla C2075 GPU. Experimental results of proposed algorithm are compared with results of best sequential sorting algorithm and odd- even merge sort based parallel sorting algorithm. Proposed algorithm shows up to 50 times speed up as compare to serial and two fold speedup as compare to parallel algorithm.