一种针对SEU的同步纠错流水线设计

Designing a Synchronous Error Correction Pipeline for Dealing with SEU

单粒子效应是星载计算机工作异常和发生故障的重要诱因之一,国内外多颗卫星曾遭受了单粒子效应的危害,造成巨大的经济损失。文章提出了一种同步纠错的流水线结构,对错误的检测与纠正进行了任务分解;当存在可纠正错误时,将纠正后的数据写入寄存器堆之后重启流水线。采用直接纠错流水线技术的Longtium-FT2容错处理器的抗辐射总剂量能力在采用普通商用加工工艺实现时达到了30 krad(Si)。

Aim.The introduction of the full paper points out that SEU（single event upset） is one of the important causes of fault or even failure of satellite-borne computer.Using the Longtium-FT2 microprocessor developed by NWPU as an example,we explain our design in sections 1,2 and 3.Section 1 analyzes Ref.6 and points out that the direct error correction pipeline induces inherently significant penalty on the microprocessor＇s performance.Section 2 designs the synchronous error correction pipeline＇s architecture,whose block diagram is shown in Fig.2.Its core is that the error correction structure is divided into the detection part and the correction part;when a correctable error occurs,the corrected data are written into the register file,and then the pipeline is restarted.Section 3 presents Figs.3 through 6 to analyze respectively the performances of the pipeline when its instructions：（1） are normally executed,（2） generate a trap,（3） produce a correctable error,（4） produce an uncorrectable error.Section 4 implements the Longtium-FT2 microprocessor,whose chip is shown in Fig.7,and verifies it with some experiments;the experimental results show preliminarily that the Longtium-FT2 microprocessor we designed is reliable and can tolerate SEU up to 30 Krads（Si） for the case of the microprocessor＇s being implemented in commercial fabrication processes.