项目名称: 嵌入式存储器容错设计关键技术研究
项目编号: No.61474001
项目类型: 面上项目
立项/批准年度: 2015
项目学科: 无线电电子学、电信技术
项目作者: 李正平
作者单位: 安徽大学
项目金额: 75万元
中文摘要: 嵌入式存储器是集成电路的重要基础部件。随着密度和复杂度的急剧上升,其发生故障的概率大大增加,严重影响了芯片的成品率和可靠性。针对上述问题,本项目针对存储器的容错设计开展多层次,多角度的研究工作。?通过对主流March自测试算法的优化改进,?融合散列技术与基于CAM、地址分割和纠错码的自修复技术,研究高覆盖率低开销的BISR技术,提高存储器的成品率。针对老化和软错误,研究统一的检测算法;采用进化硬件方法研究存储器关键部件的抗老化容软错设计,提高存储器的可靠性。研究存储器容错设计的快速进化理论,通过在遗传算法中引入新的变异算子,在进化硬件设计中融合粒子群算法和神经网络算法,减少进化计算的搜索空间,克服进化硬件的收敛性问题。
中文关键词: 容错设计;内建自修复;嵌入式存储器;进化硬件
英文摘要: Embedded memory is a key component of IC.??As its density and complexity have risen sharply, it is more prone to fault, and seriously affects the yield and reliability of the chip. To solve above problem, this program will take multi-level and multi-angle research on fault-tolerant design for embedded memory.Through the optimization of mainstream March algorithm and the fusion of hash, CAM-based BISR and ECC-based BISR technology, the program will proposes a high coverage, low overhead BISR method to improve the yield of memory.For aging and soft-error, the program will study unified detection algorithm; design anti-aging and soft-error memory key components using evolvable hardware technology. Finally, the program will study rapid convergence algorithm for memory fault-tolerant evolution design. Our propsed approach introduces novel mutation operators to genetic algorithm,merges particle swarm algorithm, neural network algorithm and genetic algorithm to reduce search space of evolution algorithm and accelerate the convergence of evolution.
英文关键词: fault-tolerant design;BISR;embedded memory;Evolvable hardware