Internal sorting algorithm for large-scale data based on GPU-assisted

This paper presents an internal sorting algorithm by GPU assisted. It consists of two algorithms: a GPU-based internal sorting algorithm and a CPU-based multi-way merging algorithm. The algorithm divided the large-scale data into multiple chunks to fit GPU global memory. Then copy the chunks to the GPU´s global memory one by one, and sort them by GPU quicksort algorithm. Then we merge these sub-sequences to one sorted sequence by CPU. We use the loser tree algorithm to reduce the number of comparisons when merging. Finally, this algorithm is tested using a variety of data distribution. The experimental results show that our algorithm improves the efficiency of large-scale data sorting effectively.