Where the Rubber Meets the Road
“Global synthesis implies an efficiency in the run time - you consider the final physical effects, not just the placement, but the wires must be predicted as well. It's funny that inside of Cadence, we have the notion of the silicon virtual prototype. That's a realistically modeled physical design, which includes the wires at a very early stage in the process. We have to go to the end of the process, and bring it up to the beginning. But that can take too long, which is the advantage [offered] by silicon virtual prototyping. You make the best trade off between timing and accuracy - in favor of timing, mostly. Then, if you're able to be statistically within 5 percent of the final accuracy, and can do that 100 times faster, then this gives you a fast forward.”
“Silicon virtual prototyping is the notion of gaining very accurate physical data early on, based on the trade-off of getting to the physical model quickly while still including the wires. We can calibrate and control that trade-off by going through silicon virtual prototypes.”
“You go internally, and literally, through all of the process of synthesis, placement, and routing - except that it's fast-forward accurate. It's not sufficient to just have a first pass. We believe that most of the time you spend should be on the closing part [of the design process]. Prototyping [assists in the process] and helps with accurate models, but it's not something you can tape-out with. Synthesis still does all of the optimization, but now the technology allows you to model the physics and to optimize to the right thing.”
“We acquired an RTL compiler when we acquired Get2Chip, and subsequently integrated it with other technologies. Now many of our customers are using that in conjunction with our Encounter platform to get to tape-out. The RTL compiler basically has this notion of global synthesis and optimization. What I just described is actually at different stages of refinement right now. Ultimately, when the technologies do integrate, you'll have one integrated flow that shares all the internal models and engines, but we're not completely there yet.”
“Future demands, and markets, in technology are difficult to predict. I don't know the specific time frame that will apply for each particular technology node ahead, even for today's trend to 90 nanometers. I do know there's renewed interest in a couple of areas. Certainly co-verification of hardware and software has become quite interesting. I was involved at QuickTurn a number of years ago, but that technology is very different today.”
“Also it used to be the graphics guys [who were on the cutting edge]. Now it's the consumer and automotive guys who are doing interesting things. I was amazed recently to learn that the chip in a toy can have up to several hundred thousand lines of codes. For those guys, they say their tape-out schedule is primarily dependent on the bottleneck of whether they can verify their software. They need software sign-off before they can commit, so now the trend is towards high-level and system-level verification.”
“Cadence is pretty committed to SystemC. I'm watching [developments there] and it's pretty interesting. It's only now starting to show some momentum, although it's not quite there yet. But it seems that it's definitely going to happen.”
“We support both SystemC and SystemVerilog, but languages are moving towards system-level design and modeling verification - both of which are starting to happen now, driven by the nanometer process technologies. I believe it's going to be mark-up languages for system verification, which is going to be a challenge, as we'll have to double, almost triple, our work to support it. But that's how it's shaping up right now. If we need to address mark-up languages, we will - and we'll see it as a very interesting phenomenon. [Meanwhile], we know that SystemVerilog and VHDL are not going to go away anytime soon. Unfortunately, the final choice [in languages] is not something we can dictate. It's a hassle, but we'll deal with it as it happens.”
“I can't see into my crystal ball to predict anything in the future. In fact, I don't even have a crystal ball. That's my philosophy, actually - not to have a crystal ball. I'm not a visionary type of guy. I try to see what's going on in the marketplace, and to adjust to that. The important thing is for us to stay very close to what's going on, and to have a sense of pragmatism. Oftentimes, you can have a strategy and start to think that's how things will go. I think it's important for us to have a strategy, but I also think it's important to have an attitude of knowing what's going on in reality and then adjusting to that reality.”
“How do I keep track of what's going on? I read the [San Francisco] Chronicle and the [San Jose] Mercury, and Time, and other magazines. Also, I travel a lot. - although, I'm traveling more and enjoying it less. I enjoy most everything that I do, but not the [increased] travel.”
“When I took on this job, people asked me, 'Why did you take the job?' I thought about it and I would have asked instead, 'Why did I even stick with EDA?' After all, it's not my first love.”
“[Originally], I wanted to be in computer architecture. But now I like EDA a great deal. There are a lot of very sharp people in this industry and I'll bet this is more difficult technology compared to anything in software. In fact, it's really too hard - and maybe that's why I like it. The problems over the last two years have gotten even harder, just in the time since I joined Cadence.”
“We've been talking about the productivity gap in this industry for years and years. In fact, I often wonder - how do you actually measure engineering productivity? Today, it's even more difficult to measure, because it's not just how much you can design [in a certain amount of time], now it needs to be defined in terms of production as that technology has become so very, very hard.”
“[Clearly], we need all of the interplay between the different disciplines to solve these problems. We need hardware, and software, and manufacturability in design, and lots and lots more. Over the last two years, nanometer design has made it all even that much harder - and made for an even greater separation between the winners and the losers in the industry, now and going forward. It's going to be harder and harder to be a winner. But if you do win, you'll win big!”
Where the Rubber Meets the Road
John Isaac is Director of Market Development for the Systems Design Division at Mentor Graphics Corp., based in Longmont, CO. As an avid sports enthusiast - skiing, hiking, off-road biking, windsurfing - working adjacent to the Colorado Rockies is a natural fit for him. He can leave the office and be out on the slopes in less than hour.
Unfortunately, that rarely happens. John's way too busy attending to the system and board-level design issues for Mentor and their customers. He comes by his expertise in these areas through long years spent at IBM managing the development of their internal EDA systems and with Mentor Graphics since 1984 (except for a 3 year non-EDA start-up experience) where he has grown with their systems design solutions since they were first introduced in 1985.
On the day that I spoke by phone with John, he was hard at work. Never mind that his family was off skiing because, undoubtedly given the choice, he would surely choose the work over the sport. He could always hear later from those who did hit the slopes how things went. But being away from the office for even a day might mean that he missed out on a development or two in board or systems area. Clearly, John's first love is his work. The skiing is only secondary.
I asked John to expound on the state of affairs in the system/board area and to start by distinguishing between system and board-level issues. As he talked, I typed, and here's the outcome:
“I think systems design is rather analogous to board design. Basically, it's where the printed circuit board and the components come together to create a system. Digital designers may feel they're creating a system on a single chip, but there's normally lots of other stuff required on the board with that chip for the final product to function. We tend to call it system design because of that - it's a little broader than just board layout in that there's verification technology and tool design included [under the umbrella] of system design. That's where our division fits into Mentor.”
Be the first to review this article