A software-based self-test methodology for in-system testing of processor cache tag arrays

Abstract

Software-Based Self-Test (SBST) has emerged as an effective alternative for processor manufacturing and in-system testing. For small memory arrays that lack BIST circuitry such as cache tag arrays, SBST can be a flexible and low-cost solution for March test application and thus a viable supplement to hardware approaches. In this paper, a generic SBST program development methodology is proposed for periodic in-system (on-line) testing of L1 data and instruction cache memory tag arrays (both for direct mapped and set associative organization) based on contemporary March test algorithms. The proposed SBST methodology utilizes existing special performance instructions and performance monitoring mechanisms of modern processors to overcome cache tag testability challenges. Experimental results on OpenRISC 1200 processor core demonstrate that high test quality of contemporary March test algorithms is preserved while low-cost in-system testing in terms of test duration and test code size is achieved.