Systematic generation of diagnostic software-based self-test routines for processor components

Abstract

Recently some fine-grained self-repair techniques for processors have been published that can handle permanent faults in particular components of a processor in-the-field. Unfortunately, the generation of diagnostic tests that can be used in-the-field for fault localization in these components is not solved satisfactorily. A few papers paid attention on improving the diagnostic capabilities of software-based self-test programs, but with emphasis on manufacturing test. This paper presents a systematic approach for the generation of test programs for diagnostic tests in-the-field. Moreover, the test program is generated in such a way that it specifically targets for faults that can be handled with an available self-repair method. The functional test programs are constructed from test patterns that can be generated with standard ATPG tools. The results show that diagnostic test programs will have an overhead in test program length ranging from 0% to 391% compared with non-diagnostic test programs.