| 摘 要: 排列熵算法随着嵌入维数的增大,运算规模将会呈平方级数增大,计算时效性问题突出,亟待解决。为此,提出一种基于任务并行编程模型的线程级并行方法,通过任务并行运行系统(StarPU)将密集型计算划分为多个独立的任务,再由调度器将任务调度到不同的CPU上执行,实现排列熵算法的并行化。基于StarPU的排列熵并行算法与串行程序相比较,加速比为23.79倍,相较于OpenMP(一种用于共享内存并行系统的并行计算方案),在分配28个线程时,加速比为1.17倍,结果表明该方法能够有效实现排列熵算法的加速执行。 |
| 关键词: 排列熵算法;任务并行编程模型;OpenMP;StarPU |
|
中图分类号: TP311
文献标识码: A
|
| 基金项目: 西兴技术智能语音系统(02030061003) |
|
| Parallel Implementation of Permutation Entropy Algorithm in Task Parallel Programming Mode |
|
LI Weiquan
|
(College of Inf ormation and Electrical Engineering, Qingdao Harbour Vocational & Technical College, Qingdao 266404, China)
2760794118@qq.com
|
| Abstract: With the increase of embedding dimension, the operation scale of permutation entropy algorithm will increase in a square series, and the problem of computational efficiency is prominent and needs to be solved. Therefore, a thread-level parallel method based on task parallel programming model is proposed. The task parallel running system (StarPU) divides intensive computing into multiple independent tasks, and then the scheduler schedules the tasks to be executed on different CPUs, achieving the parallelization of the permutation entropy algorithm. The parallel algorithm based on StarPU achieves 23.79 times speedup over serial programs, and 1.17 times speedup over OpenMP (a parallel computing scheme for shared memory parallel systems) when allocating 28 threads. The results show that this method can effectively accelerate the execution of the permutation entropy algorithm. |
| Keywords: permutation entropy algorithm; task parallel programming model; OpenMP; StarPU |
