Executive Insight: Simon Davidmann

The serial entrepreneur, investor and long-time industry insider sounds off on EDA, open source, and the importance of playing the cello.

popularity

Every industry has some colorful characters and within the EDA industry, Simon Davidmann is certainly one that comes to mind. For the past 30-something years, Davidmann has provided guidance to the industry, stood up for what he believes in, been an inspiration to many entrepreneurs, and had some fun along the way. Simon is a serial entrepreneur, angel investor and he has been a key person involved in the development of the Verilog and SystemVerilog languages, along with many other accomplishments. What follows are excerpts from an interview.

SE: You have been both an entrepreneur and an investor. Do you approach them in the same way?

Davidmann: I am a technology enthusiast. My brain never stops working. When I am asleep and wake up, I have ideas and I write stuff down. This is a constant process. I often hear of a good idea and I want to get involved and I may invest in it. I put some thought into it and then I stop. There are other cases where I get more involved. In a way I am delegating something – my money – to them.

When it is my own company, it is a constant thought process. I continue to think about it and consider alternatives. There are always alternative paths. These are conscious decisions based on lots of data. With investing you never have enough information. You can’t set a course and blindly go down it; you are constantly making small changes based on what you see and here.

I have been involved in several companies that have been wildly successful and I think it is because of the amount of time and effort that I have put into them.

SE: Imperas is your current company and it has changed course several times.

Davidmann: Well twice. After Co-Design Automation—which developed Superlog (foundational technology for SystemVerilog)—was bought by Synopsys, I had a non-compete agreement that prevented me from working in the area of simulation for two years. I saw some big challenges in software. Everyone I talked to had issues with processors and software.

I put together a business plan related to multi-processor software and this was the starting point for Imperas. The original goal was to target software to different processor architectures, in just the same way that you write RTL and can target it towards different ASIC libraries. This is not the same way in which a C compiler targets code to a single processor, but to be able to target a chip with any number of processors that could be homogeneous or heterogeneous.

There was a lot of interest but we failed to find a customer. We had a great idea but we were too early.

What people liked was the simulation architecture. We were ready for another funding round and I wanted to change the tack. The VCs wanted us to continue down the original path so the funding didn’t happen. Instead we did a management buyout, restructured the company, and went down the new path. The focus became the verification of software on virtual platforms.

We had three pieces, the retargeting software, which we shelved; the modeling capabilities; and then the simulation and tools. We decided to take the modeling capabilities and made them open-source. This was a somewhat radical idea. I had read a book called Wikinomics (Tapscott, Williams. Wikinomics: How Mass Collaboration Changes Everything) and concluded that the problem with the previous generation of virtual prototypes was that they were based on a proprietary modeling technology. Their business model was not growing. Customers hated it. We wanted to create an open modeling structure on which people could build their own models and we could focus on the tools. We didn’t want to be a service company, and so we created Open Virtual Platforms for the modeling side of the business and concentrated on developing tools for simulation and debug technologies for software.

We do develop CPU models because these are so core. We work with third parties for modeling the other pieces of the system. As an example, for Xilinx we provide the processor models for the Zynq platform but work with Cadence to provide the other models.

Doing open source and giving stuff away has been interesting. I am not trying to be benevolent in any way. I get a lot of feedback from the Universities about what they can do with our technology. I don’t get a penny from them, but it has worked well because we provided them with so much documentation that leads them through everything without needing a lot of support.

SE: How different is the software space compared to hardware development?

Davidmann: We have seen how the methodology to improve hardware quality has evolved and yet in the software space they still have a bunch of boards lying around. They might be trying to move an OS to a new platform. We showed one customer how to do this using a virtual platform and this would have given them full visibility. They decided to stick with their old ways and a year later they were still having problems because only one of the processors in a four processor system had debug logic built into it. They were operating in the dark. It frustrates me at how slow these new technologies are being adopted.

SE: Do you consider Imperas to be an EDA company or an embedded software company?

Davidmann: A while back I would say I was in the embedded software space. We started going to the Embedded Systems Conference, but we have been drawn back into the EDA space. I have always believed that nobody can develop a complex SoC without providing their customers with a complete simulation and development environment. Initially we worked with the people buying the chips, but now we find ourselves working with the people developing the chips. Today, the semiconductor vendors have to provide simulation platforms. While we inherently knew that, we went off and talked to the embedded software guys. We couldn’t figure out how to make a business selling to these guys. So now we find ourselves back working with the semiconductor companies. The market came and found us rather than us really finding the market.

SE: As the boundary of EDA grows, will it help to bring VC money back into the industry?

Davidmann: I am not the best person to answer this, and I am not sure I want people to know my opinions. I have raised quite a lot of money and returned a fair bit to them. EDA has grown because people have ideas, they are funded by their spouses, companies or VCs, it gets adopted by customers and they finish up being part of the big three EDA companies. There are also many examples of companies that are the living dead. They grow to a certain size and then stop. They are cash flow neutral and many of them solve a good problem but eventually someone gets tired and leaves and the company fades away. EDA is a tough business for VCs. If you grow a company to $10 million in revenue, you may be able to sell it for 3X to 5X. To get to $10 million you probably have to spend $5 million. If you sell for $30 million and the VC wants a 5X return, there is nothing left for anyone else. An EDA company is never going to come in at more than 10X.

SE: Should the EDA companies be the ones providing the VC money or funding the research?

Davidmann: Cisco does this thing called a spin-in. If people are going to leave and start a new company, they give them the ability to do it in house and then if they are successful they can be spun into the company. I haven’t seen this happen with EDA. They don’t even have investment funds any more. The funding should come from strategic investors. In many cases the semiconductor companies need solutions and if you pitch an idea to them and you have credibility, they have the incentive. EDA is just beginning to wake up to the need for software verification, but the semiconductor companies need solutions now.

SE: What do you do when you are not working?

Davidmann: I have always tried to find things in life that distract me from work. I like to race. When you are sitting in a small racing car, you do not think about work. Consider that you are in a 1930 race car with no seatbelts, no roll cage, running on methanol and supercharged. You are traveling at 140 mph toward a corner, there is a brick wall in front of you. Directly in front of you there is another car and he is not braking. You don’t brake either. It is a fantastic release.

In the past I have played drums and guitar. I recently started to learn to play the cello. The cello has no frets and I have never read music before. If you are not careful the bow goes off the bridge and you finish up playing your knee. There is so much going on that you have to be fully immersed.

Davidmann