A Novel Parallel Approach of Radix Sort with Bucket Partition Preprocess

Radix sort is an important sorting algorithm which is widely used in applications such as binary search and database. The most important advantage of radix sort is its time complexity is O(n), lower than other sorting algorithms based on comparison operation. However, a factor that hampers its application is long execution time of its loop body. In this paper, based on data level parallelism idea, we present a new parallel radix sort algorithm. In this algorithm, each iteration of loop can be fully parallelized via thounds of concurrent threads, eliminating the original performance bottleneck. Furthermore, it is a scalable algorithm, which means the performance of algorithm can be linearly improved with the increase of core number of modern many-core processors. Experiment results show the efficiency of the algorithm.