“Blue Ocean Strategy” + OVP (Open Virtual Platform)

[ Back ] [ More News ] [ Home ]

Introduction

Last October I wrote an article entitled “ Can a Firm Prosper or Even Survive, If It Gives Away Its Product? about Ciranova. Recently I found another company, namely Imperas, that just gave away its technology. I was tempted to chalk it up to naiveté but Imperas CEO Simon Davidmann is a successful serial entrepreneur involved with 8 successful EDA startups. He is a believer in “Blue Ocean Strategy” which calls for going after new markets rather than fighting for market share in existing markets. I had an opportunity to interview Simon at DAC.

Would you give us a brief biography?
I have a checkered history in that I have been sort of in and out of EDA several times. I was a researcher in the UK in simulation. I got lured away from research to build something which became known as HILO which was a logic simulator back in the early 80s. It was really the first logic simulator that could handle timing and behavioral modeling. We got acquired by Genrad. They used it for board testing. One of the guys who worked on that was Phil Morbi. Our first customer was in the states. This evolved and turned into Verilog. Phil went and built Verilog. I went off into the music business and built embedded systems for musical instruments, electronic percussion instruments.

I had been in universities (Essex and Brunel) and done simulation research in EDA. I thought that I would like to build things, something practical. I played guitar and got to know guys who played drums. We built this electronic percussion company. It became pretty famous and made several million pounds a year. About $10 million at the peak but in the end it went bust. We spent money on airplanes and racing cars. We had those kinds of things. We had a good time and worked very hard. When I look back at it we had some wild technology. The last product I built had five processors in it. It had a graphics controller and CRT. This was the mid-eighties. It was just like a Macintosh. It had all this DSP audio stuff, a complete synthesizer, an audio recorder, sampler and tracker. It was a drum kit. These guys just hit it with sticks. It was a very interesting experience. I went to a lot of good gigs with a load of world class musicians. We had a lot of fun. But sadly the Japanese came and built better porducts, no not better but cheaper products that were easier to use and more reliable. We could not compete. The English engineering just could not do it. We did not have the scale to do it better.

So I went back into EDA and became the first European employee for Gateway (Design Automation) with Verilog. Gateway was in Boston. I spent a lot of time going to Boston. In the end Cadence acquired it. I went and lived in Boston for six months. I just had my first child. I actually lived there early on with Gateway, again a simulation company with Verilog. Then I left Cadence and joined a small Silicon Valley company (Chronoligic’s Simulation, Inc) as the seventh employee and built something called VCS which was next generation Verilog, another simulator that ran 10x faster. I set them up in Europe and moved into sales. At Gateway I was in technical development. At Cadence I was in applications. I moved into sales and did local marketing in Europe because Europe was very much VHDL. We were trying to sell Verilog with VCS. Ultimately we were very successful and were acquired by Viewlogic and ended up at Synopsys.

I started my own silicon IP company called Virtual Chips. It ended up being acquired by Phoenix (Technology Ltd.), a bios company, so a bit out of simulation. I thought verified blocks would be a good way to save simulation. It got acquired. Then I went to Ambit (Design Systems, Inc.) and built digital synthesis technology. Cadence acquired that ($280 million).

Then I set up my own company, Co-Design Automation. We ended up building something called SUPER which became known as SystemVerilog. We got to 27 people and were acquired by Synopsys, a second time around.

We were acquired twice by Synopsys and twice by Cadence. Quite interesting! I stayed there a while. Then basically I decided not to do anything for a while. Luckily, I didn’t have to go to work. I spent a bit of time with my family.

I got intrigued by the idea of multicore. I talked to people. I said “What is the problem? Is it down in the analog stuff or this or that; someone trying to do behavioral synthesis?” What I thffought of these challenges was that everything is going to be software. Multicore makes it a real challenge. I started having ideas. I was an advisor to a venture capital company. In the end they talked me into writing a business plan proposal. They funded it and we put together the company Imperas. It has been going for several years now. Last year we had a change in focus. We moved to open source footing for some of the technology. That is what I’ve come to DAC to talk about. What we call our open virtual platform. This is the elevator pitch (see slide). I like to carry it around.

Let me try to explain about OVP (Open Virtual Platforms) and then what we did recently with Tensilica.

I recall reading where Imperas (pronounced 'imp-ear-as') donated something to OVP.
What happened at Imperas was that at the end of last year we started rethinking what we were doing. What we saw was what I call four steps to epiphany, nirvana for IP.

If I look back at the EDA industry history, especially RTL design, it went through similar sorts of things. The first thing, it moved from doing it by hand to simulation. The first phase of adoption was to get more efficient by moving to a simulation platform, to a simulation methodology for digital chip design. That has been absolutely done for 15 years. I was involved in the early phase of that. We see that from a software development point of view, it has got to be a move to virtual platforms, especially for multicore because if you are trying to build something with 8 cores on it is really hard to see with software what is going on. If you wait until you get a chip, you are six months or a year behind where you want to be and when you get the chip, you probably can’t control it properly. You can’t start and stop things, get visibility; all those things. The first thing people have to move to is the understanding that you need simulation. When they realize that, it does not take them long to realize that their existing analysis and debugging tools are not good enough. If you are writing software running on four processors, you don’t want four windows with GDB interacting with it trying to debug it. There needs to be a better solution there. Once you have something there, you have so much data and files. If you have four different processors, you need some sort of workbench to look after all the processors and manage the methodology of that. Where we see the ultimate vision going is that there is a sort of fourth phase where we program in a better way for concurrency. This is all about multicore and that is what we are concerned with. So we see that in four steps. The conclusion we came to was that it would be a long time before there were markets to which you could sell products. We demonstrated some technology in this programming space at DAC two years ago. We had 30 or 40 analysts under non-disclosure for demos; lots of people from large and small companies and from academia were very interested. Almost every one of them said “This is good and we can see the need for it in the future but the problem we have is getting our exsiting stuff on these new platforms with two or four processors. We are not going to rewrite everything yet. We know we want correct by construction in the future but today we just have to get everything up and running. So what we need is better things in the verification and debug area.” They virtually brought us down to earth. Rather than these great ideas we had for a better universe in programming, they said that they now had more immediate problems to solve. So we began thinking “How are we going to do this?” We have all of this technology in simulation. That was our most advanced technology. We were building quite a lot of stuff in verification and debug areas. The conclusion we came to was that the industry needed something better than lots of proprietary solutions. There already were lots of companies out there: VAST, CoWare, Versatech, Virtuoso. There were a lot of people out there building virtual platforms because the need is there. But they are very proprietary and most of them have 10 year old technology. They have been around for a while. If we want an ecosystem to come which will build multicore software development environments, it needs to have a standard base to build upon. That’s the problem we saw. We thought we could make money here by selling to a few companies. We would be a startup growing quite slowly, competing with these established players with a good technology. We came to the conclusion that that was not the right business model for us. I have been involved in 8 startups. They have all been successfully acquired. The last thing I want to do is spend 5 years in a company making small progress. It will either be widely successful or fail quickly. I do not want the limbo of 10 years in my company making a poor salary. I am just not interested in that. I want tremendous excitement and a fire fight; swish. That’s how I am. I like hard work. I like change and I like progress.

I read this book called “Blue Ocean Strategy”. I fly a lot. I tend to browse books. I tend to buy a lot of junk books. With most of them I get 20 pages in and leave them on the plane. But this one captured me. Basically the “Red Ocean Strategy” is where you are in a competitive situation. You are fighting with existing players for market share. No one is making much progress but you grow with the market. It is hard work but you can make money with it. Another approach is to find an area where no one is there and move some people to that area and do something different. The book has all these examples. I sat there thinking that that seems interesting but I can’t see how that would fit. Then we realized that we are doing an awful lot of stuff with open source. I said “if we do this in open source, we would change the whole game.” I thought about it. There was a company called Artisan that ARM acquired. What happened was that Artisan was in a tough business competing with big companies building libraries. What they said was let’s change the model ala Blue Ocean. They basically destroyed the competition. That’s what we want to do. We want to do it in a different way. That is basically what OVP is all about. We think things should be free and freely available. We basically put our technology and said lets give away our whole simulation technology which we calculated was about $4M of effort which is a lot for a startup that had only raised $7 million. Let’s put it out there and give it away to try to establish an open standard and to do so in such a way that an industry could build. We toyed with the idea that we would give just a little bit of it to tease people. Everywhere we went, people said “No. It has got to really work very well and really be of use, and then people will use it and a community will grow and maybe then there will be an opportunity to upsell professional things on top.” That’s where we went. We created this website for OVP ( http://www.ovpworld2.org). We support and manage it. We accept input from others. It is not just through us. Initially, it was just us. When I was having these ideas, I talked to a lot of people. We came up with some requirements of what it should be. It has to be easy to create these platforms and to create these models. There should be a library in there. It should not want too much detail. It has got to be very fast. There were lots of requirements that these people came up with. We ended up with about 17 to 20 people who went public, who said we think this is a good idea and agreed to support us. These were not people who said that we will buy this thing. They were not committing. They were just saying the industry needs something and this looks like it could be in the right direction. The Denali guys are pretty excited as is Critical Blue. IP providers MIPS and Tensilica are saying yeah, we can see there is a need for this sort of thing. Some people who are our customers and already using the tools, academics and consultant. Brian Bailey just last week published a white paper. I think it is phenomenal in what it does in explaining the issues and why he thinks what we are doing is a good idea. I’ve known Brian a long time. I think that he is dry and very critical sometimes. You never know what you are going to get sometimes when people write things. Actually it is very good. A lot of different companies in different places. It seems a good idea to the tools vendors, academics, consultants, end users. A lot of different things. Some people not knowing exactly how it might fit. But I remember when we did SUPERLOG which became SystemVerilog. I remember talking with Denali. They were not sure where it would fit, this idea about extending Verilog. It makes sense when I talked to them six months or a year ago. I said “How did SystemVerilog work out?” They are making tons of money out of it building all these models for people. It has brought a lot of light to different areas. They are actually capitalizing on the vision that we had. When they look at what we have now, they are not sure how it might fit. But they are building chips where they could use it.

So people are not quite sure. They know something has got to change and this seems a good idea. They are nervous how much effort we will put into it and how much control we will exert. That will play out over time. One of the things I have leaned is that there are right ways and wrong ways. If you put in too much control, it will fail. If we let it all go, it will fail too. There has to be a balance. I spent a lot of my time in EDA was trying to understand the marketing to be successful with different things. I think what we are trying to do with OVP is trying to pitch it at the right level. The thing that will win it is the API to model. There is this open source library but then there is the simulator that is free but it is not open source. One of the feedbacks we got is that some companies will not touch open source. For whatever reason they just do want it to infiltrate what they do in case it ends up related to their product. What they want is executables only. They want close source. The simulator is completely free. There are only two restrictions: You can not reverse engineer and you can not give it away. Everyone has to come back to the website for their own copy and sign the license agreement. In that way it is very open. It is completely full speed. That is what we have got out there. At the end of April I did a quick check of where we were. There were some 70 unique companies that have downloaded. It goes from Schlumberger, Epson, Freescale, Honeywell, Infineon, NXP, .. ; some academics as well. An interesting collection. Some OS guys, some programming guys. We haven’t advertised this to anyone out of the EDA industry. At the moment all these companies are semiconductor guys.

At the moment we are building the model base with the likes of Tensilica. We are starting to put stuff up that is useful to the software guys. That is when we think it will start to take off. They are really the guys who make use of the open source. Hardware people do not so much. There are some EDA companies that have gotten involved. OVP is focusing on that to hopefully become the foundation for software development environments. There is a lot of good stuff on how to build processor models. You can put it together on platforms. You can work in SystemC or in Spirit and plug it into existing methodologies.

Editor: The announcement was

Tensilica Inc. has signed a partnership agreement with Imperas to allow fast functional, instruction accurate models of its popular Xtensa and Diamond Standard processors to run on Open Virtual Platform (OVP) based virtual platforms. Specifically, wrapper files enabling integration of the Tensilica processor models are now available for free download from the www.OVPworld.org website. These models will run with Tensilica's TurboXim fast functional simulator, which simulates at speeds 40 to 80 times faster than a traditional instruction set simulator.

Actually this is from the first page of the website. These things run unbelievable fast in terms of the speed people need. If they are doing RTL, everybody needs faster. They will live with where they are. To develop applications you need hundreds of millions of instructions per second. To boot Linux is I don’t know about 500 billion instructions. You can’t do that at 1K instructions per second. You need a simulator faster than that. That is what our processor does. The average speed is probably 200 million or 300 million instructions per second. The peak is 1.3 billion instructions per seconds and that is running on a 3GHz machine.

The recent announcement with Tensilica is that they have been working with us and they have customers that want to use this approach for multicore. They have done work to ensure that their models are encapsulated inside this and are available. So we put a page as part of the website. This is sort of a Wiki type thing. Companies can put their own pages and link to it. It grows as they build stuff. For Tensilica there is a distribution and licensing agreement. The Tensilica dongle works on these platforms. It’s all free. You instance it just like any other native model.

Adding to the momentum of what is happening we are working with several other silicon IP providers to go for the certification of models which will be our next announcement.

On the website we have many demonstrations which go from simple processors. You basically just click on it. This is our simulator running. It runs on Windows. It has been downloaded to this machine. This is running 700 million instructions per second.

The great thing abut this is if you look at the C code which is the platform, you instance one processor, instance another and instance the shared resources, all very simple. That’s connecting the bus up. You end up with a structure. The stuff just runs and runs extremely fast. It runs at speeds that people can actually use it on laptops to get real work done. It is not this humongous thing where you need a big workstation. With embedded systems you typically run 100 MHZ to 200 MHz. That was running at 400 MIPS. It is faster than real time. Our first customer was a company called Azul. They had a 64 bit processor, 24 processors on a chip. Actually three chips in total for 96 processors. They said that it boots the OS faster than the real hardware. I could not believe it. They are very happy with it which is great. Here is an example of 24 hours running a Drystone benchmark. With OVP (free to use, free to download, usable with existing processor models), it ran 270 million instruction per seconds.

The interesting thing about Tensilica is that it talks through a FIFO not through shared memory. It is point to point. The example here is C code. You have a string, part of the US constitution. You get the next character and write it to the FIFO. A very simple use model. We are trying to make it very easy to use. It works with GDB, the standard industry debugger. It works inside Eclipse which is a big open source from IBM. You can boot Linux.

This is a MIPS executable running on a 686. It is going to boot Linux. It runs 3 to 4 times faster than real time. The platform thinks it is an MIPS processor. You can write your own programs and compile them like the “Hello, World” here. You can mount file systems. A 2GHz platform is 5x faster then the real system. You can do real software development work without the hardware. We do heterogeneous things like 3 ARMS and a MIPS

The platform is a model of all the components. You get VGA, keyboard, PCI, real time clock, memory, … You can add your own behavioral components. Someone in India is putting in a USB peripheral. They have a behavioral model of that. They are writing a driver to run on Linux. The great thing is that they develop the whole thing running on simulation. This is what OVP is.

Where Imperas comes in is that we sell a good development environment on top of this to do debug and analysis. OVP is fast, free, easy to use, gives simulation running at 500 MIPS. It is all about virtual platforms, multiprocessors, shared local memory, memory mapping, caching, debugging, .. This is what people do. We have website where we have the library of components. There is a forum there, downloadable pages and videos of demonstrations so that people can see how it works. With the OVP there are full instructions. People can scan the website. There are screenshots of what to do. For example this is how you install.

We are trying to make it very easy for people all around the world. It is very interesting. Universities are starting to do this. Some big companies have come back to us and said “We really like this. We did not like the idea that if we built a model in your environment, you would then charge us $10,000 every time we shipped it. We didn’t want that but this has changed the whole dynamic. We can use this. We can download it. We can build a model and give it to our customers”. From the Imperas point of view that’s exactly what we want because as they go multicore, they will have to buy tools. For a single processor they can use GDP. It is very easy GDP in that case but that doesn’t do the logical stuff at all. We believe that to do multicore fundamentally you have to have a simulator underneath. These tools like GDP are not designed for simulation. Better tools are needed.

Your basic business model is to sell follow on tools for debugging and verification.

Yes. I showed you the four steps. Imperas tools are filling the next step. We have technology in here. We have not really announced the tools. We have all this debug stuff. We have some very clever technology called “interception technology” which allows you to track what is going on. We have the technology and we are building something sort of like Verisity Specman but for embedded software. Verisity Specman is built on top of Verilog. What this does is build on top of simulation.

So the technology is not currently available for sale.
We have early adopters of it. We have not announced it. Our first customer actually started using it two years ago. We have been evolving it, improving it. We are trying to understand what it is that people are doing when doing software development of kernels and drivers; what they truly need. We have to package things for a specific processor like MIPS Linux development kit. They do not want a bunch of tools. They want solutions. They know that they have got a MIPS processor and they know that they are putting Linux on it. Lots of people do that.

What we are trying to do is to focus on the customers rather than publicizing it. We are focusing on OVP and making sure that people can use this technology. Because we have this community of people using OVP, there are some channels to talk to. They are asking us how to do this and how to do that. It is almost one on one. It is growing. It is very soon.

At the moment we are out fund raising. As a startup we need to keep doing this until you get enough revenue to pay for everything. My first job is to close the next round of financing (You can’t run out of money.) My second job is to make sure that OVP is useful. Continue to add real stuff, more processors, more certified models.

We are working hard and making good progress but we are trying to do things in a different way from the traditional way where people try to sell everything. In the software world so much of what people use is open source. We are a great believer in open source. We use a lot of it internally. We are trying to build a business on that.

Editor: Under products on Imperas website it simply says: “Imperas will announce its products in early 2008. Imperas is engaging with select customers today. If you are facing the challenges of MPSoC - then contact us.”

Editor: “Blue Ocean Strategy” by W. Chan Kim and Renee Mauborqne was published by Harvard University Press in 2005. There is a website www.blueoceanstrategy.com On the website there is information about the book, the authors, teaching tools related to the strategy and INSEAD Blue Ocean Strategy Institute (IBOSI) which is located in the forest of Fontainebleau just opposite INSEAD, the world’s second largest business school

An excerpt from that website appears below:

“Blue oceans, in contrast, denote all the industries not in existence today -- the unknown market space, untainted by competition. In blue oceans, demand is created rather than fought over. There is ample opportunity for growth that is both profitable and rapid. In blue oceans, competition is irrelevant because the rules of the game are waiting to be set. Blue ocean is an analogy to describe the wider, deeper potential of market space that is not yet explored. Like the “blue” ocean, it is untouched, vast and deep in terms of profitable growth.

Blue ocean strategy provides a systematic approach to break out of the red ocean of bloody competition and make the competition irrelevant by reconstructing market boundaries to create a leap in value for both the company and its buyers. Instead of competing in existing industries, blue ocean strategy equips companies with frameworks and analytic tools to create their own blue ocean of uncontested market space. The book, however, tackles not only the challenge of how to create blue oceans, but also the equally important challenge of how to execute these ideas in action in any organization.”

Examples of products utilizing Blue Ocean Strategies include Nintendo’s Wii, Apple’s Ipod and Cirque du Soleil.

The top articles over the last two weeks as determined by the number of readers were:

Fending off Cadence, Mentor tries again for British company Portland Business Journal reports that even as it decided to reject a $1.6 billion offer from rival Cadence Design Systems Inc., the board of Mentor Graphics Corp. has been pushing forward its offer for British company Flomerics Group PLC -- an offer that Flomerics' board says is unacceptably low.
Flomerics' board has urged shareholders to reject the offer, which values the company at 1.2 times its historic revenue. The board pointed out that Mentor had rejected an offer equal to 1.8 times historic revenue from San Jose, Calif.-based Cadence.

Mentor itself owns almost 30 percent of Flomerics. The company purchased just over 20 percent of Flomerics' outstanding shares in mid-March, sending the stock soaring from less than 60 pence per share, where it had been trading since last summer, to more than 80 pence per share. The stock has continued to climb since.

Mentor first made its all-cash offer for Flomerics on May 9. The price of 104 pence per share represented a 19 percent premium to the previous day's closing price.

The offer represented an even higher premium -- 109 percent -- over the price of Flomerics' stock prior to Mentor's March 14 announcement that it was buying 20.1 percent of the company from a venture capital firm.

Flomerics has revealed it is also in talks with San Rafael, Calif. based Autodesk Inc. the maker of AutoCAD design software. Flomerics announced on June 12 that Autodesk has informed Flomerics' board that any offer would be all cash. A firm offer has not yet been made.

EVE Names Vice President of Worldwide Operations EVE appointed Christophe Ballan to the position of vice president of worldwide operations working in EVE’s corporate headquarters in Palaiseau, France. He reports to Dr. Luc Burgun, CEO and president.

Prior to joining EVE, he worked for ENEA, provider of embedded software and services, where he most recently served as vice president of business development and previously as sales and operations manager for Europe and Asia. Previously, he worked as an organizational consultant and as the Central European sales manager for Wind River Systems, helping grow the business in Central Europe to a significant portion of the company’s revenues. He was part of Wind River’s executive management during its IPO.

Dassault Systemes to Acquire Engineous Software Dassault Systèmes and Engineous Software announced an agreement in which DS would acquire Engineous Software. This acquisition will extend SIMULIA’s leadership in providing Simulation Lifecycle Management solutions on the V6 IP collaboration platform. The proposed acquisition, for an estimated price of 40 million USD, should be completed before the end of July subject to specific closing conditions. The transaction is expected to be non dilutive on DS non-GAAP 2008 earnings.

Aldec Delivers Clock Domain Crossing (CDC) Solution Aldec, Inc. announced ALINT 2008.06. The software compares Verilog source code against 195 design rules based on STARC 2006. ALINT gives engineers instant feedback on structural, coding and consistency problems early in the design verification cycle. New enhancements to ALINT 2008.06 include the addition of 15 new STARC rules, a new GUI including an advanced violation viewer, configuration manager and dedicated Clock Domain Crossing (CDC) rule-sets.

STMicroelectronics and NXP Unveil Management Team for Joint Venture NXP, the independent semiconductor company founded by Philips, and STMicroelectronics announced that their new joint venture name will be ST-NXP Wireless.

Created from the mobile and wireless businesses of STMicroelectronics and NXP, which together generated $3 billion USD in revenue in 2007, the new company will begin operations in a strong position to meet customer needs in 2G, 2.5G, 3G, multimedia, connectivity and all future wireless technologies. ST-NXP Wireless is expected to enter the market in number three position once full closure of the joint venture takes place, as scheduled, in the third quarter of 2008.

It was also announced today that the management team of ST-NXP Wireless will comprise experienced industry leaders from both parent companies. Alain Dutheil, presently chief operating officer (COO) of STMicroelectronics, has been assigned to lead ST-NXP Wireless as Chief Executive Officer (CEO) focusing exclusively on the new joint venture.

ST-NXP Wireless will be headquartered in Switzerland. Aiming to close in Q3 2008, the deal is subject to regulatory approvals and labor council consultation.

Other EDA News

Other IP & SoC News