Electronic News, sister publication to Electronics Weekly, sat down at the Design Automation Conference in July to discuss what it takes to design and verify software-rich embedded systems with Alain Labat, president and CEO of VaST Systems; Dr. James Truchard, president, CEO and cofounder of National Instruments; Jeff Roane, VP of marketing at VaST Systems; Serge Leef, general manager of the system level engineering group at Mentor Graphics; and Eshel Haritan, VP of engineering for CoWare. What follows are excerpts of that conversation.

Electronic News: What does it take to design and verify software-rich embedded systems?
Leef: Today we are working with an intrinsically very conservative design community and a design community that is partitioned between software and hardware teams that don’t really communicate all that well and their design flows are completely disjointed. Hardware designers principally focus on RTL, software designers principally use $99 debuggers and at some point the problem is left to a firmware or integration engineer to bring it all together. The practices today I would characterize as fairly medieval.
Labat: Stepping back with some perspective on DAC, this is truly the first time Gary Smith from Gartner Dataquest said for the first time that the ability for traditional EDA to move out of from the traditional hardware/shrinking marketplace is to address the embedded software. It reminds me of the early 90’s when people were talking about moving to high-level synthesis. We are facing an inflection point and we have to realize that we have an enormous opportunity to actually address the needs of the embedded software design community.
Truchard: That’s precisely why National Instruments is here for its first time at DAC. Historically we’ve been a test and measurement company and we see a crying need for a new approach. The idea is to create a development environment that can do both: design and the test that goes with it. That’s what we are bringing Labview to do. Our approach includes the big, hairy audacious goal (BHAG) to do for embedded what the PC did for the desktop by creating standardized platforms with tremendous capability that all use the same software development tools so that we can essentially eliminate most of that very difficult integration that has historically been done with systems.
Roane: Especially in the backdrop of DAC, I think this whole industry is kind of wondering why. Clearly there is something going on in this space, and they are wondering why. DAC for the most part continues to be an IC-focused show. The why is: what’s different today, if you look at a lot of the chips that our customers are designing, the significant majority contain embedded processors. As soon as you take a processor and embed it in silicon, you create a time-to-market problem. If you look at wireless or things that go into the automotive dashboard or consumer electronics, the software development effort in many cases exceeds the hardware design effort. If you look at that from a schedule perspective, you simply cannot stomach serial design: spend a year designing the IC, wait until the hardware is available, and then begin to develop the software. The only solution is concurrent design so the challenge then becomes how do you do that? The ESL class of tools then becomes the answer.
Haritan: People have been talking about hardware/software codesign and concurrent design for years but what has changed is the huge increase in software content that is coming from the major force of hardware design becoming so expensive. The other part is the multiprocessor SoCs that we are starting to see appearing, mostly from a power problem, meaning, you have increased software content, you want to increase the clock speed but you cannot forever so you have this multiprocessors SoC. One of our customers told us the only way to verify the SoC today is to run the software on it, and see that the software runs. The SoC may still have bugs, but who cares? If the OS runs, if all the applications run, it is done. The ability to say when the SoC is done is no longer in the hands of the IC designer; it is in the software designers hands. That’s the inflection point that I think is changing the market.

Electronic News: How are these changes impacting how design teams work together?
Leef: I have been living in this world for at least 10 years now, and this problem has been articulated for a long time. The issue that does not get focused on is that today’s organizational footprint that our customers have does not lend itself to deployment of ESL. We can sell ESL to early adopters: visionaries that want to leapfrog their competitors.  At the end of the day, this will become a viable business only after the current methodology totally breaks down because we are serving very conservative customers and they are resistant to change and will only change when they can not do the next chip the way they did the previous one. As long as they can find a way to do the next chip incrementally they will keep doing it.
Roane: We are focusing on customers where things have broken down.

Electronic News: How fast is the uptake on ESL tools?
Labat: Gartner Dataquest says it is supposedly a 4 percent growth rate. What’s perhaps more relevant is an example about a mainstream company in the automotive industry that has decided to implement a full system-level methodology and are requiring their silicon suppliers to adopt this system-level methodology. We are in an early deployment phase with this automotive company that has plans from today through 2011/2012 for their high-end hybrid models.
Haritan: I think it is for real if you look at the benefits that people are getting from this. One of our customers – a cell phone manufacturer – says that by using this method, they save about $200,000 per cell phone, which sounds very little until you hear that they do about 60 versions of a cell phone a year. We think about a new cell phone coming about every six months, but for every country, for every new feature, they need to verify a cell phone a week. So they calculated about $35 million in annual savings by using these methods and about 30 days getting to the market ahead of what they could do without using these methods. The benefits are huge and there are early adopters that are starting to realize it.
Truchard: Our company has always had the view to do the software and then define the hardware that goes with it. And so, this is not a new view of the world as we look at creating a prototyping platform with Labview, where Labview is implemented and then once you see what you need, then you fill it out with the hardware. As a matter of fact, several years ago at one of our conferences, I made the statement, ‘We want to make the hardware disappear.’ Fundamentally you’re building a framework or platform for the software and then defining the hardware that matches that. In my mind, the hardware is what’s left over when you finish the software.