基于COTS DSP的软件抗辐照技术研究
发布时间：2018-11-14 17:35
【摘要】：高性能数字信号处理器(Digital Signal Processor，DSP)越来越多地应用于星载信号处理平台。然而宇宙射线辐照引发的系统可靠性问题，使得高性能DSP在我国星载信号处理中的应用受到制约。而我国目前在抗辐照DSP研制方面基础薄弱，制约了DSP在我国航天领域中的应用。相比较抗辐照DSP，商用(Commercial Off-The-Shelf, COTS) DSP在性能、价格、功耗、体积和重量方面都拥有巨大的优势。本项目基于COTS DSP平台，利用软件容错技术来解决可靠性问题，最终实现兼顾可靠性和性能的星载数字信号处理平台，对提升我国航天器的在轨处理能力，具有重要的理论意义和应用价值。 本文在分析已有软件容错技术的基础上，围绕DSP的软件加固技术，开展了以下几方面的研究： 1.提出了一种基于线性汇编的DSP程序软加固算法。首先分析C6000系列DSP的体系结构、功能部件、并行执行等特点以及软错误的传播规律，提出基于线性汇编的DSP程序容错加固算法。该算法结合了控制流检测和数据流检测，使得加固后的程序能够检测出寄存器、指令存储空间中的软错误。 2.提出了基于软加固算法的性能优化技术。首先对本文提出的软加固算法进行分析，发掘数据流检测方法与控制流检测方法重复检测的部分，提出分支检测优化；基于检查点优化的思想，提出控制流检测优化，以增加一定的错误延迟为代价，不降低错误覆盖率的同时，降低算法的性能开销。然后在分析软加固前后程序的执行时钟周期数的基础上，根据加固后程序流水线编排失效的原因，对加固后的程序循环体进行优化，提出循环优化算法，，在不降低错误检测率的同时，以一定的错误处理延迟为代价，降低了软加固后程序的执行时钟周期数。 3.设计实现了线性汇编到线性汇编的容错转换工具，可对程序进行不同容错算法的加固，具有良好的可扩展性。然后设计实现了线性汇编的故障注入工具，可向程序随机注入数据流错误和控制流错误。在此基础上，对DSP容错加固算法和循环优化后的DSP容错加固算法加固后的典型程序进行故障注入实验，验证容错算法的有效性。
[Abstract]:High performance digital signal processor (Digital Signal Processor,DSP) is more and more used in spaceborne signal processing platform. However, the application of high performance DSP in spaceborne signal processing in China is restricted due to the reliability of the system induced by cosmic ray irradiation. However, our country has a weak foundation in the development of radiation resistant DSP at present, which restricts the application of DSP in the field of spaceflight in our country. Compared with irradiated DSP, commercial (Commercial Off-The-Shelf, COTS) DSP) has great advantages in performance, price, power consumption, volume and weight. This project is based on COTS DSP platform, using software fault-tolerant technology to solve the reliability problem, and finally realizes the spaceborne digital signal processing platform, which takes into account both reliability and performance, and improves the in-orbit processing capability of Chinese spacecraft. It has important theoretical significance and application value. Based on the analysis of the existing software fault-tolerant technology, this paper studies the following aspects around the software reinforcement technology of DSP: 1. A soft reinforcement algorithm for DSP programs based on linear assembly is proposed. Firstly, the architecture, functional components, parallel execution of C6000 series DSP and the propagation rule of soft errors are analyzed, and a fault-tolerant reinforcement algorithm for DSP programs based on linear assembly is proposed. The algorithm combines control flow detection with data flow detection, which enables the strengthened program to detect soft errors in the register and instruction storage space. 2. A performance optimization technique based on soft reinforcement algorithm is proposed. Firstly, the soft reinforcement algorithm proposed in this paper is analyzed, and the repeated detection part of data flow detection method and control flow detection method is excavated, and the branch detection optimization is proposed. Based on the idea of checkpoint optimization, the control flow detection optimization is proposed, which increases the error delay at the cost of not reducing the error coverage and reducing the performance overhead of the algorithm. Then on the basis of analyzing the number of executing clock cycles of the program before and after the soft reinforcement, according to the reasons of the failure of the program pipeline arrangement after the reinforcement, the loop body of the strengthened program is optimized, and the circulatory optimization algorithm is put forward. At the same time, at the cost of a certain error processing delay, the number of execution clock cycles of the soft strengthened program is reduced. 3. A fault-tolerant conversion tool from linear assembly to linear assembly is designed and implemented. It can reinforce different fault-tolerant algorithms of programs and has good extensibility. Then, a linear assembly fault injection tool is designed and implemented, which can randomly inject data flow errors and control flow errors into programs. On this basis, fault injection experiments are carried out on typical programs of DSP fault-tolerant reinforcement algorithm and cyclic optimized DSP fault-tolerant reinforcement algorithm to verify the effectiveness of the fault-tolerant algorithm.
【学位授予单位】：国防科学技术大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP332;V446
本文编号：2331844
[Abstract]:High performance digital signal processor (Digital Signal Processor,DSP) is more and more used in spaceborne signal processing platform. However, the application of high performance DSP in spaceborne signal processing in China is restricted due to the reliability of the system induced by cosmic ray irradiation. However, our country has a weak foundation in the development of radiation resistant DSP at present, which restricts the application of DSP in the field of spaceflight in our country. Compared with irradiated DSP, commercial (Commercial Off-The-Shelf, COTS) DSP) has great advantages in performance, price, power consumption, volume and weight. This project is based on COTS DSP platform, using software fault-tolerant technology to solve the reliability problem, and finally realizes the spaceborne digital signal processing platform, which takes into account both reliability and performance, and improves the in-orbit processing capability of Chinese spacecraft. It has important theoretical significance and application value. Based on the analysis of the existing software fault-tolerant technology, this paper studies the following aspects around the software reinforcement technology of DSP: 1. A soft reinforcement algorithm for DSP programs based on linear assembly is proposed. Firstly, the architecture, functional components, parallel execution of C6000 series DSP and the propagation rule of soft errors are analyzed, and a fault-tolerant reinforcement algorithm for DSP programs based on linear assembly is proposed. The algorithm combines control flow detection with data flow detection, which enables the strengthened program to detect soft errors in the register and instruction storage space. 2. A performance optimization technique based on soft reinforcement algorithm is proposed. Firstly, the soft reinforcement algorithm proposed in this paper is analyzed, and the repeated detection part of data flow detection method and control flow detection method is excavated, and the branch detection optimization is proposed. Based on the idea of checkpoint optimization, the control flow detection optimization is proposed, which increases the error delay at the cost of not reducing the error coverage and reducing the performance overhead of the algorithm. Then on the basis of analyzing the number of executing clock cycles of the program before and after the soft reinforcement, according to the reasons of the failure of the program pipeline arrangement after the reinforcement, the loop body of the strengthened program is optimized, and the circulatory optimization algorithm is put forward. At the same time, at the cost of a certain error processing delay, the number of execution clock cycles of the soft strengthened program is reduced. 3. A fault-tolerant conversion tool from linear assembly to linear assembly is designed and implemented. It can reinforce different fault-tolerant algorithms of programs and has good extensibility. Then, a linear assembly fault injection tool is designed and implemented, which can randomly inject data flow errors and control flow errors into programs. On this basis, fault injection experiments are carried out on typical programs of DSP fault-tolerant reinforcement algorithm and cyclic optimized DSP fault-tolerant reinforcement algorithm to verify the effectiveness of the fault-tolerant algorithm.
【学位授予单位】：国防科学技术大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP332;V446
【参考文献】
中国期刊全文数据库 前4条
1 邢克飞;王跃科;扈啸;;银河飞腾DSP芯片总剂量辐照试验研究[J];半导体技术;2006年07期
2 邢克飞;杨俊;周永彬;季金明;;星载高性能DSP加固设计方法研究[J];电子器件;2007年01期
3 傅忠传;陈红松;崔刚;杨孝宗;;处理器容错技术研究与展望[J];计算机研究与发展;2007年01期
4 贺兴华;肖山竹;张路;张开锋;陶华敏;卢焕章;;空间DSP信息处理系统存储器SEU加固技术研究[J];宇航学报;2010年02期
本文编号：2331844
本文链接：https://www.wllwen.com/kejilunwen/jisuanjikexuelunwen/2331844.html
