基于CPU-GPU异构并行计算的刀具路径规划方法研究

发布时间：2018-03-30 20:06

本文选题：CPU-GPU异构　切入点：OpenCL　出处：《杭州电子科技大学》2017年硕士论文

【摘要】：21世纪以来,数控技术取得飞速进步的同时新挑战也随之出现。新型智能数控系统需要在线生成或编辑修改刀具路径,实时或准实时的进行刀具路径的计算任务,因此对软硬件系统提出了很高的性能要求。同时,开放式数控系统搭建于不同架构处理器和控制器组成的异构硬件平台上,由于不同架构的处理器和控制器互不兼容,造成这些硬件资源得不到充分利用。论文在全面分析了新型智能数控系统和软件架构的研究现状之后,针对传统串行刀具路径规划算法效率低下和在异构硬件平台上的不兼容问题,提出一种基于CPU-GPU异构系统对于刀具路径并行化实现的方法。论文主要工作如下:1.根据新型智能数控系统的开放性和兼容性基本要求,提出将异构系统通用并行计算技术引入数控领域的新思路,最后在CPU-GPU异构环境下实现通用并行计算;2.通过在数控系统中应用OpenCL技术,针对数控系统中CPU与GPU不同体系架构的处理器,按照软件设计方案和OpenCL编程规范,提出并实现一种基于CPU-GPU异构系统并行刀具路径规划算法的详细设计方案;3.针对CPU-GPU异构系统搭建计算平台和OpenCL实现的环境,然后依据等参数线法对不同采样密度的刀具路径进行并行化重构;4.在Microsoft Visual Studio 2010开发环境下,使用C++和OpenCL编程语言分别对B样条曲面和NURBS曲面的并行刀具路径规划算法进行了实现,最后完成了新的并行刀具路径规划算法与传统串行刀具路径规划算法对比实验,并结合Matlab进行了刀具路径规划的仿真应用实验,结果显示新的并行算法效率较传统串行算法有1.5~16倍的大幅提升。
[Abstract]:Since the 21st century, with the rapid progress of numerical control technology, new challenges have emerged. The new intelligent NC system needs to generate or edit the tool path online, real-time or quasi-real-time tool path calculation task. At the same time, the open CNC system is built on heterogeneous hardware platform composed of different architecture processors and controllers, because the processors and controllers with different architectures are incompatible. As a result, these hardware resources are not fully utilized. After a comprehensive analysis of the research status of new intelligent numerical control system and software architecture, The traditional serial tool path planning algorithm is inefficient and incompatible on heterogeneous hardware platforms. This paper presents a method of tool path parallelization based on CPU-GPU heterogeneous system. The main work of this paper is as follows: 1. According to the basic requirements of openness and compatibility of the new intelligent NC system, This paper puts forward a new idea of introducing the general parallel computing technology of heterogeneous systems into the field of numerical control. Finally, the general parallel computing is realized in the heterogeneous environment of CPU-GPU. By applying OpenCL technology in numerical control system, a new method is proposed. According to the software design scheme and OpenCL programming specification, according to the processor of different architecture of CPU and GPU in NC system, A detailed design scheme of parallel tool path planning algorithm based on CPU-GPU isomerism system is proposed and implemented. 3. For CPU-GPU heterogeneous system, a computing platform and a OpenCL implementation environment are built. Then parallel reconstruction of tool paths with different sampling densities is carried out according to isoparametric line method. 4. In the Microsoft Visual Studio 2010 development environment, The parallel tool path planning algorithms of B-spline surface and NURBS surface are implemented with C and OpenCL programming languages respectively. Finally, the comparison experiment between the new parallel tool path planning algorithm and the traditional serial tool path planning algorithm is completed. The simulation results of tool path planning with Matlab show that the efficiency of the new parallel algorithm is 1.516 times higher than that of the traditional serial algorithm.
【学位授予单位】：杭州电子科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TG659;TP332

【参考文献】