Spring 2010 Event ...

EDAC Panel: Embedded Software & EDA

by Peggy Aycinena

The EDA Consortium hosted their Spring Event on the Cadence campus in San Jose on Thursday, April 15th. Approximately 75 people enjoyed an artistic array of hors d'oeuvres and an equally appealing array of wines prior to the evening's panel discussion. EDAC's not-a-drop-to-drink after 7 pm policy helped shoe everyone into the auditorium at the top of the hour, however, so the evening's program started right on time.

[Please note: The entire April 15th panel was recorded for posterity, and will be available in perpetuity on the EDAC website.]

The presentation began with a welcome from Cadence and a somewhat ill-advised infomercial from ChipEstimate, followed by EDAC Executive Director Bob Gardner describing the Consortium's Market Statistics Services [MSS] and quarterly EDA revenue report.

Gardner said the Q4 2009 numbers suggest something of an economic recovery is en route, even if not yet quite arrived at the station. He then turned things back over to the evening's host, Cadence Chief Marketing Officer John Bruggeman.

Bruggeman welcomed the audience and celebrated the fact that the evening's topic – Solving the Software & Silicon Complexity Challenge – allowed him to orchestrate an overdue conversation between EDA and the embedded software industry.

He quipped that Mentor Graphics CEO Wally Rhines, sitting in the audience, had asked why Mentor was not represented on the panel, even though it's the single largest provider of embedded software in the industry. Bruggeman then introduced the panelists:

Tomas Evensen, CTO at Wind River
Jack Greenbaum, Director of Engineering for Advanced Products at Green Hills
Jim Ready, Founder and CTO at MontaVista

At Bruggeman's request, Jack Greenbaum began by listing current embedded software challenges. His list included: the move from 16-bit controllers to 32-bit controllers, just as the traditional market for 32-bit is moving to 64-bit; the shift to full-feature operating systems with increases in RAM and Flash memory capacity; multiple processors on separate die moving to multiple cores on a single die; and virtualization, which requires running more than one operating system, while still protecting all resources on a single die.

Tomas Evensen followed with his concerns: “Ten years ago, we had much more control over the software. Now it takes millions of line of code, and you can't have just one guy who understand the whole system. [In addition], you have to get all this software from other places – you need more software and more features than you can possibly create yourself – hence the drive to open source and Linux, which means complexity.

“The more stuff you're putting together, the more opportunities for failure. Plus, multiple cores means you have more things running concurrently, [including] multiple operating systems.”

Jim Ready added: “Remember this is the 4th decade of this panel, starting with the 1980s, 1990s, 2000's, and now it's 2010. This [exponential] increase in complexity has been a persistent problem [throughout this period]. In a sense, however, the gap in software has been a great opportunity; it's had to come from somewhere.

“It used to all be proprietary software. Then we came to the realization that to help solve the problem, the industry had to figure out how to create and cooperate on open source. It's been very interesting – now, using the software you don't write is the most efficient way you can make [systems].”

Back on message ...

Bruggeman, an employee of a company that produces proprietary software, was undoubtedly a little uncomfortable with his panelists' ringing endorsements for open source. He thanked the speakers, but noted they'd offered little by way of a primer on embedded software, too much discussion of proprietary versus open source software, and still needed to relate their conversation to EDA.

To do that, Bruggeman began again: “I see consolidation in the market, with semiconductor companies buying embedded software companies. Why is that?”

Tomas Evensen said everything's related: “[At one time], semi's sucked up software companies and then used them in a proprietary fashion on their own chips. But that wasn't really good for the semi or software guys in the long run. They didn't want to be locked in. They wanted to be able to switch from software architecture to hardware architecture. [Yes], both MontaVista and Wind River have been bought, but we're being left alone so we can pursue [excellence].”

[Editor's Note: Wind River was bought by Intel in 2009 for $884 million. MontaVista was bought by Cavium Networks in 2009, purportedly for $50 million.]

Bruggeman responded, “If I listen to Tomas, [semi's buying up software companies] is fraught with danger for everyone. Why should the semi's continue to let these companies deliver value?”

Jim Ready answered, “Lock-in and flexibility are both part of this. Large customers have a very large influence, and folks just like the model of independent software providers.”

Following this exchanges, the panelists went on to speak at length about: a) the “sad state” of affairs where people don't want to pay for software, b) which semi bought which software company and why, c) the benefits of Linux, and d) whether or not a viable business can be built on top of Linux – if money can be made, how much.

Along the way, Bruggeman offered poignantly, “Your customers – the systems companies and the device manufacturers – place tremendous value on the software that you deliver, but they refuse to pay like it's something of great value. And, as all of my EDA brethren know well, we also provide tremendous value, but do not get paid for it.”

Bruggeman also commented on an evolution: “The system companies and the device manufacturers have had a model in their minds for the past 50 years. They used to start with the hardware, then add the software infrastructure, then add the middleware, and then more software. They would deliver it all in a product to their user customers, but hat model has now been turned upside down because of two things.

“The economic blip – no longer can they afford to do the whole thing, demanding that more of it be done by someone else. The other thing is the iPhone, which is not about the hardware or the infrastructure software. It's about the applications!

“If you're going to be a great and successful device manufacturer today, you have to make great applications. So, systems companies like Lucent, Nokia, Alcatel – the iPhone showed them that it's about the applications. And now, the device companies have said that they're out of the vertical integration business. They want the semi's not to provide a chip, but to provide an application platform high enough up the stack to get this done.”

In hoping to sort out where to go from here Bruggeman asked, “How do you think the relationship between EDA and the embedded software companies should change in order to better serve your customers, these device manufacturers and system companies?”

Jim Ready suggested that improve contact was the solution – the intersection between the various EDA flows – because the big EDA vendors “sell to the same guys.”

Jack Greenbaum offered: “Do system level design, so you can shake out the problems before you go to silicon. Use virtual platforms.”

He also acknowledged, “Chip designers and embedded software are like oil and water. Hardware designers use VHDL and Verilog, but software engineers are completely different. They can't talk to each other. It's C versus VHDL/Verilog. Design engineers know they're going to pay the EDA guy more than they want to, while the software engineer will squeeze a chip vendor until they get the tools for free.”

Advise and consent ...

And so the conversation continued on through to 8:30 pm with a mix of optimism, pessimism, and fatalism. In the closing minutes, Bruggeman asked the panelists to offer advice that might be useful to their EDA-centric audience.

Tomas Evensen said: “Realize and understand the software problems that our customers have. When you change one bit in the hardware, it has a ripple effect.”

Jack Greenbaum said: “Make your tools more open to software developers.”

Jim Ready said: “Before venturing into the software world, have the deepest possible understanding of the dynamics. Because EDA has been so successful with efficiency around hardware design, certainly there must be opportunities in all of this morass of software.

"There must be gold in there!”

April 25, 2010


Print Version

Peggy Aycinena owns and operates EDA Confidential:

Copyright (c) 2010, Peggy Aycinena. All rights reserved.