SpringSoft brings more smarts to Certitude

One of the more interesting tools in the verification space is SpringSoft’s Certitude—what you might call a testbench qualification tool. In brief, Certitude cleverly and automatically injects bugs into your RTL based on an analysis of the structure of your code, runs your regression test sequence, and checks to see if all of the injected bugs resulted in errors. It sounds like a simple idea, and it probably is, once you figure out an algorithm for inserting the permutations at the right spots in the RTL.

One problem with this approach is too much information (also known as TMI, if you have kids of a certain age.) If there’s a significant gap in your test coverage, Certitude can generate an undetected-error report that begins to look more like War and Peace than a tool output. That leaves you with some unanswered questions, such as where the heck to start, and why didn’t I study finance? Are you looking at 50 unrelated gaps in functional coverage, or one critical test that is coded incorrectly? That’s not an easy question to answer when you are graced with a report that deserves to be issued in a hard-cover edition.

SpringSoft has come up with a solution: a set of heuristics that can prioritize the faults Certitude injects, so that the tool leads you to the most significant coverage problems first. For example, Certitude will now give priority to faults inserted close to an output, or faults in resets to synchronous processes, according to SpringSoft director of product marketing George Bakewell. If faults at these key points are going unnoticed, you pretty much don’t have to worry about faults upstream from them until after you’ve amended the testbench. By using this prioritization tool iteratively as the design progresses, you can focus on the largest gaps in coverage first, and your testbench can evolve along with the design, under the guidance of Certitude.

SpringSoft has introduced another welcome improvement, Bakewell said. There is now a direct bi-directional link between Certitude and Verdi. You can automatically generate Certitude set-up files from Verdi’s Knowledge Database. And once you find an undetected fault in Certitude, you can push a button and be looking at the fault in Verdi’s source-code and waveform windows, so you can track down exactly why the testbench is missing it.

Altogether, these two changes are incremental, but valuable, steps in the increasingly critical task of knowing just how good your testbench really is. With verification gradually becoming the largest single threat to design schedules, and verification escapes becoming a major source of dreaded respins, this is no minor achievement.