Home
  • Products
  • Solutions
  • Support
  • Company
  • Products
  • Solutions
  • Support
  • Company
Community Breakfast Bytes ESD Alliance Evening with Paul Cunningham

Author

Paul McLellan
Paul McLellan

Community Member

Blog Activity
Options
  • Subscriptions

    Never miss a story from Breakfast Bytes. Subscribe for in-depth analysis and articles.

    Subscribe by email
  • More
  • Cancel
semi
Jim Hogan
esd alliance
verification

ESD Alliance Evening with Paul Cunningham

23 Apr 2019 • 8 minute read

 breakfast bytes logo Paul Cunningham was interviewed by Jim Hogan at the latest ESD Alliance "fireside chat". (Ever since Lip-Bu's fireside chat at DAC was interrupted by the fire-alarms due to a real fire, I've been a little wary of that name.)

Jim said that he'd wanted to find someone who'd "crossed the chasm". Paul was the founder of Azuro, which Cadence acquired, and so he clearly qualifies.

One thing I discovered in the dinner and networking beforehand was the Paul had a "blue" in rowing. That's Cambridge-speak for rowing for the university. I presume there are other races, but the one everyone knows is the annual "boat race" on the Thames in London, against Oxford. This year, the 165th annual race took place on April 7, three days before the fireside chat (Cambridge won, yeah!).

Jim started by asking Paul how he got into EDA in the first place.

I’ve had an unusual journey. I loved programming at school and all that but got into EDA through self-timed chips. I spent way to much time on sport as an undergraduate so I chose to stick around for asynchronous design. It is a hard problem, really difficult, and great benefits. I drank the KoolAid. But it’s wrong. If you take away the clock, each component takes as long as it needs. I bought that and was inspired. I was in the tool side. I came out here and interned with Ivan Sutherland at Sun Microsystems. He was running a research group pushing self-timed circuits. At the end of the internship, he said, “why not come and get a job with us”. It was 2000. The tech bubble hadn’t quite burst. He also told me, “if you want to start your own company, let me tell you about Angel investors. I would invest." I told him, “I haven’t got an idea yet." He said "It's about people and passion."

Paul then came to the US for an opportunity to get funded. He got a little bit of money "like $100,000", and they rented a room in Cambridge. Paul said "Steve [Wilcox] was the smartest guy I knew and he was designing modems". Paul called him and Steve said it was no fun since he was so much brighter than his boss, so he joined Azuro to solve the asynchronous timing problem.

But once they started talking to people like CSR (Cambridge Silicon Radio), it became clear that something wasn't right. Asynchronous circuits wouldn't change the world.

Whether you have a clock or not, it is the same underlying transistors. Self-timed is event-driven and more and more components become edge-sensitive. Synchronous is great since you can gate and shut down the clock, and you can have slow and fast clocks. You can borrow time with useful skew. But nobody was actually doing that.

So we decided to build a new clock useful skew optimizer. Benchmark had opened in London and somehow we got their attention and they put in $4M.

Once we had Broadcom as an announced customer, and then ST, Benchmark told us "it's EDA...you have to move to California." We kept the team in UK, but the ecosystem is here.

Jim pointed out the pros and cons. It's so expensive and hard to get talent, so it's hard to do anything here today. But all the customers have research groups within an hour of here.

Paul agreed:

This is to tech what Milan is to fashion. Cambridge calls itself the Silicon Fen but its not the same.  Here, if a startup doesn't work, do another one. That doesn't exist there. There's this zone where nobody wants to let it die.

"So what did you learn as CEO?" Jim asked.

You have to completely embrace your customer. They define if something is interesting or compelling. Nothing is statc, and thinking is constantly evolving.

"So you're rolling out," Jim said. "You've got interest. What do you do about getting the brand up?"

I don’t think we did a great job on marketing, to be honest. But in some ways you get the brand with the users. Certain customer wins self-market since customers talk among themselves. The big thing that gave us escape velocity was the right customer wins. Our website probably sucked. Given my time again, I'd be a lot more vocal, keep saying the same thing. I think I spent way to much time over-optimizing it. I could have had a really detailed white paper but the real goal was just to make sure people know the name Azuro.

Jim moved on. "You've got an office over at Techmart and along comes Charlie Huang [Cadence's VP sales at the time]."

There were lots of steps and things started to explode. Apple launched the iPhone. At heart was one version of Arm, A9 I think, that caught on and was going into a lot of phones. Our useful clock skew technology meant we could get an extra 10% on those designs. There was no plan to build a full flow or go public, so we decided it would be a good idea to be exclusive to one of the flows. We parntered with Cadence and never felt it was the wrong decision.

Jim: So then Cadence acquired you, what was it like?

When you sell your company, from the acquirer's point of view is day #1, but a huge number of things that have to happen on that first day. However, you spent a few weeks on intensive due diligence. I never slept more than 1 or 2 hours per night. It was very emotional. Shareholders, investors, employees. It was the most intensive few weeks of my life. Then you go into security and they want to take a picture for your badge. Worst picture ever. My first time in a big company. Azurro was straight out of school. We were all Cambridge grads and thought the same way. Cadence was very mixed people.

Jim: You need to focus on what you do best. There are people that do all the other stuff. Paul [me—Paul McLellan, not Paul Cunningham] and I did CadMOs and we ended up sending Charlie [then CEO of Cadmos] to hospital for a couple of days.

"Yes," Paul said. "All those boring things you think are irrelevant like signed contracts and NDAs. Someone is going to go over them with a fine-tooth comb."

Jim: So Anirudh is Cadence president and calls you to ask you to run verification. CDNLive was last week. What was his message?

Lip-Bu and Anirudh’s leadership is why I’m still there. Relentless focus on putting technology and customer first is great. We have the most comprehensive technology across the whole thing. We execute a business that is very efficient, best margins, great growth. At CDNLive, Anirudh is looking at what is the next area we can expand into. Two big calls to action.

Where do we have is right to win? We are not going to be a social networking company. But we have a deep understanding of complex technical software, matrix solvers, advanced graph traversal, and so on. At CDNLive, we announced finite element analysis and Clarity to move into a new adjacent space. Super exciting. Potential to take Cadence into new areas.

Jim moved Paul onto his specific area of responsibility, verification.

I went to kickoff in 2011 and different GMs got on stage and the verification group showed a trend of linked to the total spend of semiconductor and as a clock guy, the implementation was basically not growing. But chip validation and software validation was exploding. I thought “I’m in the wrong business”. So when Anirudh called me up 15 months ago, whatever, and offered me this lead, it was an easy decision.

Moore’s Law doubled the gate count, so all 2n so how do we do twice as much. But in verification, doubling squares the state space so it’s on a different trajectory. You are never done in verification. In software, the ability to verify the software before you get the chip back is huge. So the opportunity is infinite. The way I describe it is you have “infinite problem and finite resources”, so it is bug/$/day (I call that “verification throughput”), so how do we enable our customers to maximize that?

We just made an investment in Green Hills. They may be the most knowledgeable about building secure systems. Back to “right to win”, we are not the leaders in security. But we are already starting projects with them and learning a lot together.

Jim's final question was how Cadence's strong mixed-signal offering plays into verification.

It’s a big vector. Think of how many analog devices are in a cellphone, let alone a car. Advanced mixed-signal design is exploding, so we need rich and sophisticated MS verification environments. There's still a long way to go. If you do everything with SPICE, it’s too slow. So we use real-number models to bring floating-point continuous voltage into a digital simulation without doing matrix solves like you would in circuit simulation.

Finally, it was time for audience questions.

Q: Why has there never been an analog hardware acceleration business? Everyone who has tried it has failed.

I don’t know about the past…but it seems like it is solvable. Figuring out how to massively parallelize the problem like with Clarity is a first step. It's not directly in my product line responsibility at Cadence but I doubt it will remain unsolved.

Q: Why Azuro? The name.

It was a pun on Italian with blue. Icy-blue. But we were determined we wanted a dot com and short. So we got it since we didn't spell it correctly.

Q: What’s going to happen in the next two to three years in verification?

I'm repeating myself: strong automotive solution across the industry, need to do some foundational things for the security of chips and software, need to make sure we enable a full spectrum of bare metal from FPGA to emulation to x6 and Arm servers. One understated things is that if you have this 2n, you have to constantly invest in driving performance and capacity. Another thing is bug hunting. Lots of things that are there to make us smart about doing something intelligent with the engines. A Ferrari is amazing but you need a great race-car driver. Portable stimulus and formal lint, protocols, all growing fast.

CEO Outlook

The next ESD Alliance event is the CEO Outlook, which will take place at SEMI in Milpitas on May 23rd. Drinks, dinner, and networking from 6:00pm. The panel itself from 7:00pm to 8:30pm. It's not yet been announced which CEOs will participate this year. I'll see you there.

 

Sign up for Sunday Brunch, the weekly Breakfast Bytes email.


© 2023 Cadence Design Systems, Inc. All Rights Reserved.

  • Terms of Use
  • Privacy
  • Cookie Policy
  • US Trademarks
  • Do Not Sell or Share My Personal Information