Error detection and correction in VLSI systems by online testing and retrying

The conventional dual-module redundant (DMR) structure with comparison is a kind of self-testing structure used in some commercial general-purpose computers, such as IBM 4341 processor and IBM S/390 G3 and G4. One of its drawbacks is it does not have error recovery ability. Retrying finds applications because of its low hardware overhead. Micro-rollback and instruction retry as in VAX 8600 are effective methods for rapid error recovery that the error detection latency is only a few clock cycles.