Testing of Hard Faults in Simultaneous Multithreaded Processors

With the increasing circuit complexity and aggressivetechnology scaling, faults such as dielectric, conductor,and metallization failures are becoming more common.Traditional stuck-at testing does not detect these types offaults because these faults may be dormant and need to bestressed to manifest as "fails" (during burn-in). Asvoltage scaling and power consumption reduce theeffectiveness of burn-in, these faults will cause failuresduring the useful life of the part. We propose the use of atest thread on Simultaneous Multi-Threaded processors toprovide a means of detection of lifetime failures. In thisinitial study the test thread is allowed to compete withexecuting programs for processor resources. When thesystem has only one workload thread, the test thread isable to execute with no significant impact on the workloadthread. When the system has many active workloadthreads, the test thread has somewhat larger impact onthe execution time of the workload threads.