arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1506.01446cs.DC
收藏

基于CUDA的双调排序算法的实现与优化

The implementation and optimization of Bitonic sort algorithm based on CUDA

Qi Mu, Liqing Cui, Yufei Song

摘要:

本文详细介绍了双调排序算法,并基于cuda架构实现了双调排序算法,同时针对GPU的特点对实现细节进行了两次有效的优化,大大提高了效率。最后,我们通过比较快速排序算法在CPU上的加速比,对GPU上优化的Bitonic排序算法排序算法进行了研究,由于快速排序算法不适合并行实现,但在一定程度上比CPU上的其他排序算法更有效,因此,我们对GPU上的Bitonic排序算法和CPU上的快速排序算法进行了加速比和性能比较。对一系列32位的随机整数,实验结果表明,我们的工作的加速近20倍,当数组大小为216左右时,加速比甚至高达30。

英文摘要:

This paper describes in detail the bitonic sort algorithm,and implements the bitonic sort algorithm based on cuda architecture.At the same time,we conduct two effective optimization of implementation details according to the characteristics of the GPU,which greatly improve the efficiency. Finally,we survey the optimized Bitonic sort algorithm on the GPU with the speedup of quick sort algorithm on the CPU.Since Quick Sort is not suitable to be implemented in parallel,but it is more efficient than other sorting algorithms on CPU to some extend.Hence,to see the speedup and performance,we compare bitonic sort on GPU with quick Sort on CPU. For a series of 32-bit random integer,the experimental results show that the acceleration of our work is nearly 20 times.When array size is about 216,the speedup ratio is even up to 30.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑