基于模型的异构多核嵌入式系统开发研究

发布时间：2018-06-20 08:37

本文选题：异构多核 + 嵌入式软件设计　；参考：《上海大学》2013年博士论文

【摘要】：伴随集成电路工艺的不断进步以及处理器体系的发展，异构多核处理器在嵌入式领域广泛应用并成为设计的主流体系结构，与单核平台上软件开发不同，应用任务需要合理划分和并行性调度到恰当的核上执行，实现资源的最佳化配置。异构多核软件面临巨大挑战，迫切需要相应开发方法的指导和开发工具的支持。传统的以编程为中心的方法难以满足市场对开发效率和质量的要求，基于模型的开发体现了嵌入式领域的现代设计思想和方法，是当前在异构环境中设计复杂软件的有效途径，对其深入研究具有重要的理论和实际意义。论文针对领域应用的异构多核处理器，，紧密围绕基于模型的嵌入式系统开发这一主题进行各阶段的探索、研究和开发，同时完成必要的创新性工作。所开展的工作和取得的成果主要体现在以下四个方面： 1提出基于模型的异构多核嵌入式系统开发方法。结合异构多核特性和嵌入式软件开发模式的特点，提出从任务划分、目标平台与应用建模、任务映射、模型编译到最终代码生成的开发方法。克服了多核并行开发的复杂性，提高了软件开发效率和软件可靠性。 2构建基于模型的异构多核开发环境。设计实现了模型化、可视化集成开发环境MV12-ModelIDE，其中模型编辑器MV_Modeling实现应用需求基于嵌入式实时操作系统的抽象建模和架构建模描述，任务分配器MV-Mapping实现DAG任务模型到异构多核平台的最优分配和调度仿真，模型编译器MV-Compilation则根据既定转换规则实现模型到代码的自动生成，汇编器MV-Assembler针对MV10从核实现了包括定制指令的汇编源程序准确和快速编译。 3提出一种求解异构多核DAG任务分配与调度问题的改进离散粒子群优化算法。以任务高度和粒子的运动位置作为优先级构造调度队列，采用基于插入的原则选择调度列表中首个任务具有最小累计EFT的处理器核作为任务分配对象，使任务集有序并行运行达到最小完成时间和最大加速比，充分发挥异构多核高性能特性。 4将上述方法和技术在MV12异构多核处理器及汽车车身控制网络实验平台实施和验证后，结果证明所研究的方法与技术正确有效。课题研究所取得的成果可推广应用于其它嵌入式多核处理器上。
[Abstract]:With the continuous progress of integrated circuit technology and the development of processor architecture, heterogeneous multi-core processors are widely used in the embedded field and become the mainstream architecture of design, which is different from software development on single core platform. Application tasks need to be reasonably partitioned and parallelized to perform on the appropriate kernel to achieve optimal allocation of resources. Isomeric multicore software is facing a great challenge, which urgently needs the guidance of corresponding development methods and the support of development tools. The traditional program-centered method can not meet the requirements of the market for development efficiency and quality. The model-based development embodies the modern design ideas and methods in the embedded field and is an effective way to design complex software in the heterogeneous environment. It has important theoretical and practical significance to further study it. Aiming at heterogeneous multi-core processors in domain applications, this paper focuses on the topic of model-based embedded system development in various stages of exploration, research and development, and completes the necessary innovative work at the same time. The work and achievements are mainly reflected in the following four aspects: 1. A model-based development method for heterogeneous multi-core embedded systems is proposed. According to the characteristics of heterogeneous multi-core and embedded software development mode, the development method from task partition, target platform and application modeling, task mapping, model compilation to final code generation is proposed. It overcomes the complexity of multi-core parallel development and improves the efficiency and reliability of software development. The modeling and visualization integrated development environment MV12-Model IDE is designed and implemented, in which the model editor MVStamp Modeling realizes the abstract modeling and architecture modeling description based on embedded real-time operating system. The task allocator MV-Mapping realizes the optimal allocation and scheduling simulation of DAG task model to heterogeneous multi-core platform, and the model compiler MV-Compilation realizes the automatic generation of model to code according to the established transformation rules. The assembler MV-assembler implements accurate and fast compilation of assembly source programs including custom instructions for MV10 slave kernels. 3 an improved discrete particle swarm optimization algorithm for heterogeneous multi-core DAG task assignment and scheduling problems is proposed. The task height and the moving position of particles are taken as priority to construct the scheduling queue. The processor core with minimum cumulative EFT in the first task in the scheduling list is selected as the task assignment object based on the principle of insertion. Make the task set run in sequence parallel to the minimum completion time and maximum speedup, The methods and techniques mentioned above have been implemented and verified in MV12 heterogeneous multi-core processor and vehicle body control network experimental platform, the results show that the methods and techniques studied are correct and effective. The research results can be extended to other embedded multi-core processors.
【学位授予单位】：上海大学
【学位级别】：博士
【学位授予年份】：2013
【分类号】：TP368.1;TP332

【引证文献】