Brown Bag Lunch: Sanguinetti & Sandler
John – Pretty much everybody these days uses a high-level model of their specification. Not everybody uses a means [or a tool] of moving to a lower model. The two-level model is a problem we’ve had for a long time. Basically, you have to have two different models to compare them. It’s very real that the testbench and a set of tests that you run is the model, typically regression suites. You’re comparing the two. People have been trying to figure out how we automatically derive the testbench and vectors from the design.
Peggy – Do we call that innovation or implementation?
Scott – Innovation comes in creating, in building the automation in a practical way. Remember the example of simulation. When Verilog/VHDL came on the scene, there were a whole variety of simulators in use, and in various stages of market acceptance. For a variety of reasons at Gateway, we were able to effectively wipe them off the map. There were 8 different multi-mixed level simulators, RTL mixed with gate level, and most of them had warts. They weren’t as fun or as effective, and they withered on the vine. The analogy today – there are Hyundai’s and there are BMWs.
Peggy – But are the differences there real or psychological? They’ll both get you where you’re going.
Scott – If you measure the speed to go from 0 to 60, the differences are very real. In our world, there was Cadabra, but Verilog was so much more satisfying.
John – It was easier to do the job I wanted to do. The first job I had in simulation, I had to choose between N-dot and Verilog. I looked at both simulators and decided I could write a system model in Verilog, and couldn’t do it in N-dot. Verlog addressed the problem better.
Scott – It was more satisfying. In EDA, some things work well. Some things don’t.
Peggy – Does the best technology always win?
John – That’s the conventional wisdom in Silicon Valley. There was BetaMax versus VHS, but the more technical the field, the less likely that some marketing organization can push an inferior product on to the customers. Although, we know of examples in our industry of products that didn’t make it because of a comparable product from a more powerful vendor.
Scott – John’s right. The more technical the stuff is, the less impact of marketing. Keep in mind the premise of crossing the chasm [involves] the satisfaction of buying and using a product. Overall, how well is it satisfying the needs of the purchasing organization? The answer has more facets that just being the fastest thingy or the most accurate. It also has to do with the infrastructure around the product. A really fast simulator that no one knows how to create is useless. There may be one team who can create it, but other guys who say it’s a waste of time.
The infrastructure around VHS was there, but it wasn’t there for BetaMax. So, it’s not just about the technology, it’s also about creating a complete product. If you get too focused on the technology, you lose sight of what it really takes to satisfy the customer. That’s when the best technology doesn’t win. Even Verilog might now have succeeded if we hadn’t asked the ASIC vendors to build models for their libraries. We had to enhance the language to succeed.
John – That was ultimately the differentiator between Verilog and VHDL. VHDL would have succeeded if it had wiped out Verilog. That was the intention, but it couldn’t do gate-level simulation.
Peggy – So, are the Big Guys in EDA setting the pace?
Scott – Hell no. They compete tooth and nail around pricing and supplying multiple pieces [of the flow]. But we provide the better mousetrap.
John – We do have constraints in the general design flow – the small companies can’t do everything – but often we can fit better into the design flows of the larger companies than their own [point tools].
Peggy – Isn’t there an established infrastructure that must be met?
Scott – It’s not enough to have great technology. One of the aspects that we’ve mastered in creating complete products, is to make sure that our great technology fits into the existing infrastructure.
Peggy – Can you maximally innovate?
John – Maybe not maximally, but you can still innovate. When I started Chronologic, Redwood DA started out at the same time. They were making a compiled simulator for their own language. It was supposed to be 10x faster than the existing simulators, and they built a whole collection of stuff around it. They went out [into the market] at the same time that we did our Verilog simulator. There were benchmarks where their simulator would go faster than VCS, but my analysis and characterizations were better. They were saying to their customers, You have got to come to our Brave New World and leave your old baggage behind.’ But, VCS was successful because we said, You don’t have to leave your baggage behind.’
Scott – We’ll carry your baggage for you.’
John – Redwood had good technology, but they didn’t make it.
Peggy – Can additional investment make a difference, particularly in a small company?
Scott – It’s [often] simply a matter of how much you’re willing to invest. Is the idea worthy of that investment? It takes a lot of appetite for risk to go after the whole enchilada. Over the last 20 years, the bar has been raised and the threshold is much higher now because of the depth and breadth of the 3 major companies supplying virtually the same stuff. So [you need to provide] a very sharp arrow that fits within the existing flow. If you’re going in broadside, taking on the big companies, you need an enormous amount of energy.
John – You have to have a compelling proposition. We see that with ESL. The value proposition with ESL is simply compelling, assuming all the ESL tools live up to the promise. [Still] there’s a lot of inertia out there that always prevents people from saying we’ll switch to ESL.
Peggy – Differentiate between your two worlds? Between working at the current level of abstraction versus moving up?
Scott – It take just as much innovation to improve an existing practice as it does to move that practice to a new place. In comparison between us, at Novas we’re focusing on improving existing practices.
John – We’re focused on moving the practice to a new place. A great deal of our effort over the history of the company has been trying to make the move to higher levels of abstraction in a seamless way.
Peggy – Why do it in a seamless way?
John – Because we think it can be done. We’re just layering a level of abstraction [on existing levels], but at the level of the details, it does change the way people do design. It’s actually much more disruptive than I would have thought back in 1998 when we started the company.
Peggy – I’m not trying to describe your work as dull, Scott, but where’s the innovation?
Scott – Well, your notion of innovation is quite generalized. In some ways, you’re equating innovation with a sea change within the whole sales channel. And I would say, yes sometimes innovation requires a sea change. At other times, it takes just as much innovation and engineering to make important incremental changes in present practices. It’s easier to throw things out and start [from scratch], than to make improvements in existing practices.
Peggy – Isn’t it about engineering versus science?
Be the first to review this article