Design of a reliable cache based on grouped checking and data reloading

Abstract

Reviewing some past research[1–4], we have come to believe that SEU (single event upset) is the main reason for fault or even failure of satellite-borne computer. This paper proposes a reliable cache structure which is based on grouped checking and data reloading. In this structure, data and tag domain are divided into several groups to check. If there have some errors during checking, the cache accessing is identified to cache missing and to reload the right data from memory. The experimental results show preliminarily that the Longteng-FT1 microprocessor we designed is reliable and can tolerate SEU up to 300 Krads (Si).