ON METHODS FOR DESIGN OF AUTOMATICALLY REPAIRING VLSI COMPUTERS.

In computer designs that will use very large scale integration (VLSI), prospects for manual repair of failures seem remote. Indeed, it may be physically impossible, having only remote sensing. Economics might also preclude human intervention. A schema is given for automatically repairing VLSI computers (ARCs). An ARC is an interconnection, basically, of repairable modules (RMs) which consist of acyclic interconnections of repairable computing devices (RCDs); it contains a diagnoser based on the linear algorithm TestDetect (hardware version) which analyzes errors and, together with the controls for replacement with the RM, pinpoints a failure; TestDetect contains a description of the logic of the RMs it diagnoses. The method of design permits exact diagnosis--never before achieved--and repair by replacement.