Verification Technology Transfer

• Authors: R. P. Kurshan
• Venue: 25 Years of Model Checking
• Year: 2008

In the last quarter century computer-aided verification --- especially in the form of model checking --- has evolved from a research concept to a commercial product. While the pace of this technology transfer was anything but rapid, the new technology had almost insuperable hurdles to jump on its way to the market place. Hurdle number one was a required significant change in methodology. On account of its limited capacity, model checking must be applied only to design components (RTL blocks in the case of hardware) instead of the whole design as with simulation test. Thus, the functional behavior of these design components must be specified. Since component level functionality is often revealed at best obscurely in the design's functional specification, either designers must convey component functionality to those doing the testing or else testers must somehow fathom it on their own. The former was considered an unacceptable diversion of vaunted designer resources while the latter was often undoable. A second hurdle was uncertainty surrounding the quality of the new tools. Initially the tools were incomparable and required the user to create considerable tool-specific infrastructure to specify properties before a tool could be evaluated. Recreating the required infrastructure for several tools was infeasible. This meant choosing a tool without a head-to-head evaluation against other tools. With the high cost and uncertain outcome afforded by these hurdles, no circuit manufacturer was willing even to consider seriously this new technology. Not, that is, until the cost of testing-as-usual became higher than the cost of jumping these formidable hurdles. This essay is the saga of the transfer of computer-aided verification technology from research to the market place.