A practical approach of GPU bubble sort with CUDA hardware

摘要

Presently GPU computing is in big call with the sphere of sorting. This paper presents the design of GPU (Graphic Processing Unit) bubble sort with CUDA (Compute Unified Device Architecture) hardware. The bubble sort is considered as comparison based sorting. The design of bubble sort using many core GPUs was previously unknown. The paper also presents the speedup acquired by the GPU bubble. The sorting benchmark is used to test bubble sort (parallel & sequential). The six types of test cases are considered by the sorting benchmark. On the basis of experimental analysis, parallel bubble sort achieved 37229 times faster execution time using zero test case and 6375 times faster using sorted test case at N = 2500000 and T = 512. The best case time complexity of the GPU bubble sort is compressed O(n) to O(1) because of the GPU.