Fast Correction of Multiple Soft Errors in Highly Associative Cache with CAM-Based Tag

Content addressable memory (CAM) is a key component to build the tag memory of a highly associative cache. As CMOS process technology scales,soft error rates (SER) in CAM cells increase significantly. Bit flipping in CAM cell leads to a false miss upon a cache access, which could be fatal from a system point of view. Previous schemes either focused on reducing the probability of soft errors or had an unbounded single soft error correction time. Compared with previous schemes, our approach completely detects and corrects multiple soft errors. The detection and correction in our scheme happens as a background process, does not interfere with concurrent cache accesses, and does not affect the performance of time-critical cache operations. In addition, we enhance the cache miss holding register structure of a non-blocking cache to avoid data corruption due to any false cache miss happening between occurrence and correction of errors.

[1]  N. Jouppi,et al.  Complexity/performance tradeoffs with non-blocking loads , 1994, Proceedings of 21 International Symposium on Computer Architecture.

[2]  Tryggve Fossum,et al.  Cache scrubbing in microprocessors: myth or necessity? , 2004, 10th IEEE Pacific Rim International Symposium on Dependable Computing, 2004. Proceedings..

[3]  Hyuk-Jun Lee Immediate soft error detection using pass gate logic for content addressable memory , 2008 .

[4]  Shuichi Sakai,et al.  Mitigating soft errors in highly associative cache with CAM-based tag , 2005, 2005 International Conference on Computer Design.

[5]  Matthias S. Müller,et al.  Memory Performance and Cache Coherency Effects on an Intel Nehalem Multiprocessor System , 2009, 2009 18th International Conference on Parallel Architectures and Compilation Techniques.