| 引用本文: | 邓丁,郭阳.面向RISC-V内核的标记指令复算与纠错机制的设计[J].国防科技大学学报,2020,42(6):90-97.[点击复制] |
| DENG Ding,GUO Yang.Recomputation and correction mechanism design for tagged instructions of the RISC-V core[J].Journal of National University of Defense Technology,2020,42(6):90-97[点击复制] |
|
| |
|
|
| 本文已被：浏览 357次 下载 286次 |
| 面向RISC-V内核的标记指令复算与纠错机制的设计 |
| 邓丁,郭阳 |
|
(国防科技大学 计算机学院, 湖南 长沙 410073)
|
| 摘要: |
| 由宇宙辐射等环境因素引起的硬件瞬时故障会极大降低计算机系统的可靠性。为了减小硬件瞬时故障对系统可靠性的影响,保证程序的正确运行,基于RISC-V开源内核“蜂鸟e203”提出一种标记指令复算与纠错机制。该机制为指令码额外增加了复算标记,以较小的硬件开销实现对指定指令的复算操作。此外,该机制可以在第一次复算结果与初始运算结果不等时自动进行二次复算,并由三次运算结果的多数表决结果来纠正大部分硬件瞬时故障所引起的数据流异常。实验表明,通过与中断处理程序相结合,在随机注入瞬时故障的情况下,程序的正确执行概率平均增加了86.67%。 |
| 关键词: RISC-V 蜂鸟 指令标记 复算 纠错 |
| DOI：10.11887/j.cn.202006011 |
| 投稿日期：2019-06-13 |
| 基金项目:国家自然科学基金资助项目(61832018) |
|
| Recomputation and correction mechanism design for tagged instructions of the RISC-V core |
| DENG Ding, GUO Yang |
|
(College of Computer Science and Technology, National University of Defense Technology, Changsha 410073, China)
|
| Abstract: |
| The reliability of the computer system is significantly compromised by the hardware transient faults which are mainly caused by the cosmic radiation and other environmental factors. To mitigate this undesirable impact and guarantee the correctness of the running programs, a recomputation and correction mechanism for tagged instructions for an open source core named “Humming bird e203”, which is based on the RISC-V instruction set architecture, was proposed. This mechanism adds extra flag bits for each instruction and thus enables flexible recomputation for any tagged instruction at low hardware cost. Besides, it can issue the tagged instruction again automatically if the result of the first recomputation is different from the original one. This majority voting scheme can efficiently rectify most data flow errors caused by transient hardware faults. The experimental results show that with our proposal and the interrupt handler, the average probability at which programs can operate correctly can be increased by 86.67% under the random transient fault insertion. |
| Key words: RISC-V Humming bird tagged instruction recomputation correction |
|
|
|
|
|
