Zigzag-HVP: A Cost-effective Technique to Mitigate Soft Errors in Caches with Word-based Access

Abstract

Error Correction Code (ECC) is widely used to detect and correct soft errors in VLSI caches. Maintaining ECC on a per-word basis, which is preferable in caches with word-based access, is expensive. This paper proposes Zigzag-HVP, a cost-effective technique to detect and correct soft errors in such caches. Zigzag-HVP utilizes horizontal-vertical parity (HVP). HVP maintains the parity of a data array both horizontally and vertically. Basic HVP can detect and correct a single bit error (SBE), but not a multi-bit error (MBE). By dividing the data array into multiple HVP domains and interleaving bits of different domains, a spatial MBE can be converted to multiple SBEs, each of which can be detected and corrected by the corresponding parity domain. Vertical parity updates and error recovery in Zigzag-HVP can be efficiently executed through modifications to the cache data paths, write-buffer, and Built-In Self Test. The evaluation results indicate that the area and power overheads of Zigzag-HVP caches are lower than those of ECC-based ones.