Architecture Design for Soft Errors

CHAPTER 7

Hardware Error Recovery

7.1 Overview

As discussed in Chapter 1, soft errors can be classified into two categories: SDC and DUE. Some of the error coding techniques, such as parity, discussed in Chapter 5 and the fault detection techniques outlined in Chapter 6 convert SDC to DUE. In some cases, these techniques may result in additional DUE rates arising from the mechanics of the fault detection technique. This chapter discusses additional hardware recovery techniques that will reduce both SDC and DUE rates arising from transient faults. The same recovery techniques may not be as useful for permanent hardware faults because fixing the permanent fault typically would involve a part replacement, which is not discussed in this chapter. ...