SAME forum continues to showcase innovative start up companies

Selected as this year's 'Most Promising Start up' was Adacsys, which specialises in fpga based verification and debug of high performance embedded systems. It says its approach complements traditional emulation approaches, providing a simpler means of verification before going into full hardware emulation. It also provides a low cost solution for smaller firms. Adacsys' target market is high peformance computing in aeronautics, which ceo and founder Erik Hochapfel describes as 'niche, but well defined'. The company expects to sell 30 systems by the end of 2013. Despite having innovative technology, the right people and a target market, not everything has run smoothly. Hochapfel left Mentor Graphics in 2007 and launched Adacsys in 2008 with a €300,000 grant from the French R&D ministry. But, due to misreading the market, the product was only finalised this year. "We didn't talk with potential clients early enough," Hochapfel admitted. "We should have made more effort much sooner to focus on the commercial aspects." The company has had to adapt the product so it fits better into customers' design flow. "We are largely competing with in house solutions and, because they don't want to lose their return on investment, we have to ensure our product is modular and integrates into their process easily." Adacsys has also adapted its sales techniques, with greater focus on discovering the potential customer's 'pain point'. Although he has learned some hard lessons, Hochapfel does not regret his decision to start his own company. "It's great to work with highly motivated people. We are all in collaboration mode at present and teamworking is critical." He emphasised that, to maintain motivation, it is important to ensure the people are happy. "If they want to stay in engineering, fine. If they want to expand their horizons to technical marketing or project management, then there are opportunities." Making a false start Space Codesign, although in a different market sector, had a similar false start, primarily due to underestimating the sales and marketing effort required. The company has developed SpaceStudio, an electronic systems level (esl) eda tool suite focused on automating hardware/software codesign. "ESL design is a 10 year old technology," commented chief scientist, Professor Guy Bois, "but now is the right time to be getting into it." SpaceStudio, the culmination of more than 30 person years of research at the Ecole Polytechnique in Montreal, is claimed to be the first tool suite that automates the partitioning process and, through the use of C/C++ and System C, to be capable of simultaneous hardware and software development. "This has been the perpetual challenge of our postgraduate students year on year," said Prof Bois. Once the product's kernel was formulated, Prof Bois saw the commercial opportunity. "We thought the challenges would be technical, but they were not," he admitted. In fact, the technical and academia oriented founders discovered swiftly that finding customers and selling licences was tough. "We needed business skills and a marketing strategy in order to find more funding," he said. Gary Dare joined as general manager in 2010, spearheading the company's relaunch, bringing business and technical marketing experience. "I was attracted by the technology and the freedom of a start up definitely appealed," Dare said. Dare has first hand experience in customer facing roles. "It's important to understand the politics of your customers," he remarked. Yet, despite Dare's market knowledge and business acumen, he chose to start work with Space Codesign as a developer 'to really get to know the technology'. "I firmly believe the time has come for hardware/software codesign automation," Prof Bois reiterated. Unlike many long time researchers and academics, Prof Bois has always wanted to get involved with start ups. "Over the years, I have been a technical consultant in other start ups, but at a distance. This time I wanted to be more involved." Three professors from the Royal Institute of Technology, Stockholm, made a less successful move from research to the commercial world when they founded Elsip in 2011. The trio wanted to exploit the IP they had designed to manage memory and interprocess communications in multiprocessor SoC based systems. The professors launched DME (Data Management Engine). This programmable and configurable synthesisable IP block is designed to solve a number of issues, including cache coherency, memory consistency, virtual address translation and dynamic memory allocation in heterogeneous and homogeneous architectures. The IP addresses architecture exploration, allowing 'what if' analysis, and helps designers to determine how much parallelism can be exploited in their multicore designs. But the founders had no idea how to sell the product. Technology editor Adam Edstrom, who had knowledge of the market and the people involved, joined as CEO in September 2012, along with Bengt Edlund from National Semiconductor as sales director. "For professors, the founders are very good in front of customers!" Edstrom noted. "They are outgoing and can present well." Although he said they largely understood the commercial issues, they quickly realised they needed people with more entrepreneurial flair. Plus, they realised they preferred the research side of the technology. Edstrom is finding the work rewarding. "It's a steep learning curve, but you get to see things from a different angle." Edlund and Edstrom agreed from day one that the company had to be global. "To some extent it doesn't really matter where the company is based," Edlund said, but one Elsip founder, Prof Zhonglai Lu, has Chinese roots and the company has set up an office in Shanghai. Taking the DIY approach is Alexandru Plesco, founder of start up to be ZettICE, which is perfecting ICEbuilder, described as a reconfigurable IP framework. Engineer turning technical marketeer Plesco has just completed a dedicated course for start ups at the Lyon Business School. "It's a good move at the right time for me. I understand most of the issues now," he said, "although I'm not sure I have all the answers." Explaining the technology The first challenge is finding the best way to explain the technology. A library of configurable IP, including linear algebra and video/imaging processing elements, is complemented by the software platform, which uses sequential algorithms, mathematical models and highly accurate floating point kernels to generate fine grain parallelism. The result is support for multitarget solutions in applications such as motion detection, video tracking, facial recognition and eye motion. Plesco declines to use the term 'optimising silicon compiler' due to preconceived ideas on past eda tools – but, in essence, that is what ICEbuilder is. Although the concepts have been proven, the developers have a lot more work to do to generate marketable products. Currently hosted by France's research and technology transfer agency Inria and incubator Créalys, three engineers, one researcher and five trainees are working to complete the arithmetic elements of the ICEbuilder IP portfolio.