Here in Cadence Product Marketing, we're still recovering from our very busy annual sales conference last week. Of course, I can't say much about what transpired there but I do want to comment that functional verification was a hot topic. In my many conversations with our field team, two trends were clear.
The first is that customers just can't afford to re-spin chips in the current economic climate. With the cost of deep submicron mask sets running so high, even a single re-spin can be a project/product killer, or truly a company-killer for a startup. Naturally, this puts even more emphasis on all forms of verification to achieve first-pass silicon success.
My second observation is that the ongoing trend of having to do more with fewer resources is accelerating. It's no secret that most of our customers are not growing right now, and many have in fact reduced their staffs. A lot of the functional verification advances in the last ten years involve greater automation and play directly to this situation. Customers simply can't hire armies of verification engineers to hand-write tests anymore. Instead, they write assertions and let formal tools do most of the work at the block and subsystem levels, then leverage OVM-based constrained-random testbenches to do most of the work at the chip and system levels. Of course, these approaches rely on high-quality metrics that expose verification holes and measure verification completeness.
So I returned from the event with the strong feeling that verification will be more important than ever this year. What do you think? Am I suffering from tunnel vision since I spend most of my time in this space, or are you also seeing functional verification looming large in your near future? Any comments are most welcome!