“…Various symbolic tools required long CPU time when formally verifying a pipelined DLX Ritter et al, 1999), or ran out of memory (Isles et al, 1998). Custom-tailored, manually defined rewriting rules were used to formally verify a five-stage DLX (Levitt and Olukotun, 1997), and similar four-stage processors (Harman, 2001;Lis, 2000;Matthews and Launchbury, 1999), but would require modifications to work on designs described in a different coding style and significant extensions to scale for dual-issue superscalar processors. Other researchers proved only few properties of a pipelined DLX (Ivanov, 2002;Ramesh and Bhaduri, 1999), or did not present completeness argument (Mishra and Dutt, 2002)--that the properties proved will ensure correctness under all possible scenarios.…”