What Would Joe Do?
Peggy Aycinena is a freelance journalist and Editor of EDA Confidential at www.aycinena.com. She can be reached at peggy at aycinena dot com.
Canada’s Space Codesign: Calm, Cool and Collected
October 25th, 2012 by Peggy Aycinena
Montreal is not a place that normally comes to mind when you think of EDA. Space Codesign Systems, however, is on a fast track to change that in a classically Canadian way – calm, cool, and collected.
When I spoke with General Manager Dr. Gary Dare on a beautiful afternoon in Southern France at SAME Forum in early October, he explained how the company started in Canada, and the road map they have set out for themselves: “We’re an EDA company, an EDA startup, and we are definitely based in Montreal. If you doubt that EDA has a place in Canada, we will soon convince you otherwise.
“Space Codesign comes from the acronym, SystemC Partition of ACE, which was the 2004 research project at the Ecole Polytechnique [University of Montreal] that our technology is based on. In 2008, Professor Guy Bois and various graduate students associated with the project decided to do a spin-out, and in 2010 Space Codesign Systems went into operation.”
He laughed and added, “Our company has nothing to do with space, however. But it has everything to do with hardware/software co-design – doing it simultaneously, rather than the usual way of ESL hardware design followed by software design. The audience we are targeting is the systems architects who are looking at the algorithmic level and need a route to design exploration and implementation. Our tools give them that route.
“At this point, Space Codesign is looking for its first commercial customer, talking with a number of different entities – early adopters interested in innovative solutions, including several universities in North America and Europe, companies in aerospace and automotive on both continents, and other categories of companies as well.
“We are looking initially to customers targeting FPGA implementation for their designs, or at least for prototyping, and also want to partner with one of the big ASIC players, as well as smaller fabless companies with connections to TSMC or Global Foundries.
“To all of these organizations, we are offering the ability to come up with high-quality hardware/software architectures for embedded systems and a way to realize improvements in their product cycle time. We are offering a way for customers to implement high-quality systems where the hardware and the software work together well.”
I asked if the solution space for hardware/software co-design isn’t already overfull.
Dare responded, “Yes, but the state of the art today in co-design remains mostly a cyclical or back-and-forth process, not a process where the hardware and software development is happening in lock-step. The traditional approach has been hardware-centric, while then attempting to extend and add-on software development.
“The technology behind our product, however, is based on a different idea. Guy Bois’s background and research was in the back-end layout and partitioning of the design. His first focus was on functionality and getting it right. He realized, however, that once the entire system was put together, only then should developers start worrying about what went into the hardware versus what went into the software.
“He knew that co-design existed, which included developing a hardware virtual platform and then putting to the software on that platform, and found an improvement on the old process. But, as you suggest, that was still sequential design. Guy’s solution was to adopt a systems perspective instead – the traditionally less popular top-down approach, which requires abstracting up to the system level and going from there.”
Is that point of view unique to Space Codesign, I asked.
Dare said, “We are compatible with existing vendors who sell tools that generate C/C++ code and other modules from a system-level abstraction – in fact, they can provide our tool with the models. What is unique to the Space Codesign solution is that we take the C/C++ and drop it into a SystemC wrapper, whereupon they can proceed to architectural design exploration.
“If someone sketches out their design in Matlab, we could import the C/C++ generated there, and also C/C++ code that might be generated by UML-based tools such as Mentor’s BridgePoint. Ours is an innovative solution to a long-standing problem, which we expect the industry to embrace as they become aware of the power our tool offers.”
If it is so innovative, what is to stop companies like Synopsys and Mentor from attempting to develop similar tools?
Dare answered, “Space Codesign’s technology is based on original research, so there is a lot of intellectual property and know-how behind it that has to be matched.
“Nonetheless, we are aware that people in those companies – particularly those from CoWare and Vista, acquired by Synopsys and Mentor, respectively – would argue that they have a point of view which competes with ours. I respect their opinions, as they are pioneers in our field, but I would disagree. Our technology is a next step in the evolution of ESL, and time will prove we are right.”
Could Space Codesign potentially partner with companies like MathWorks? And, is the company at a disadvantage being based in Montreal?
Per Dare: “We would love to partner with companies like MathWorks, and we already do since our implementation options include sending ESL designs to Forte’s Cynthesizer and Calypto’s Catapult C.
We’re also part of the ARM Connected Community, which has helped greatly to raise our profile. All of this helps us in working with commercial customers to validate our business model. As you know, securing the first customer is always the most difficult and critical milestone for a startup. That is going to happen, however, and then we are confident a number of companies will be looking to partner with us.
“As far as our being based in Montreal is concerned, in fact there are a number of advantages of being ‘isolated’ from Silicon Valley or other tech centers around the world. In Montreal, we are situated in a surprisingly vibrant start-up community with a wealth of experience in various areas of technology, including a host of mobile app developers and investors who provide us with different perspectives from outside of the EDA industry.
“Also in Canada, the pitch to the VC community is different. The public sector there provides a number of industry innovation grants – a sort of bridge where Angels and VCs may be missing in a particular technology niche and generally in shorter supply than the US, although that is changing.
“These grants lean towards research and development, so a number of embryonic ideas can see the light of day more easily than in traditional VC environments such as you find in Silicon Valley. Yes, productizing ideas is always an issue – in Canada and everywhere in the world – but the environment for technology startups in Canada is excellent.”
Dare concluded, with Canadian calm, “Twelve months from now, we hope to have our first customer engagements, and to be talking with additional potential customers. And we expect to be extending into the ASIC world at that point, having moved beyond our current focus on the FPGA implementation process. We see the ASIC NoC as a crucial emerging issue, and actually have on our staff various people with R&D backgrounds in this area as well.
“Overall, we have complete confidence that our in-house capability in both FPGAs and ASICs will guarantee our success. You may not think of Montreal when you think of EDA, but Space Codesign will be changing that for the better.”
Tags: ARM Connected Community, ASIC, C/C++, Calypto, Canadian VCs, CoWare, ESL, Forte, FPGA, Gary Dare, Global Foundries, Guy Bois, hardware/software codesign, MATLAB, Mentor Graphics, Montreal, NoC, Space Codesign Systems, Synopsys, SystemC, TSMC, UML, University of Montreal, Vista