2005 Emerging Trends: When one thing leads to another …
It started at the Embedded Systems Conference this past spring in San Francisco, when I heard Wind River's Jerry Fiddler give the keynote address. In his talk, Fiddler made frequent, broad references to the EDA industry and it occurred to me that it might be interesting to explore the growing interface between the embedded software world (ESW) and the world of electronic design automation (EDA). Although he gave me his business card just after his talk, Fiddler was surrounded by well-wishers that morning and somehow I failed to impress upon him in that crowd that if he could just respond to my upcoming e-mail, I would be able to assemble an article based on his comments. Meanwhile, at DAC 2004 in San Diego in June, I had a brief conversation with a seasoned EDA editor who insisted that any article based on Fiddler's comments on EDA & ESW would be ... Been there But, I still wasn't sure. So last month, I attended the Software Developer's Forum in Palo Alto because a) the speaker promised to be really interesting, and b) Jerry Fiddler was going to be introducing the speaker. I thought to try (once again) to have a chance to ask him in greater detail about the growing EDA/ESW interface. Unfortunately, although he was polite enough to give me his contact info, again he did not respond to my follow-on e-mails. Meanwhile and quite coincidentally, Leslie Cumming of Skye Communications offered to help pull together an article on exactly this subject – 2005 Emerging Trends in EDA: EDA and Embedded Software I'm grateful to Leslie for her work on this article, and to Dataquest's Daya Nadamuni for her help in reviewing the questions. And although I'm disappointed in my inability to garner comment from Jerry Fiddler himself, I'm delighted to be able to publish the following responses to my questions on the topic. Comments were sent via e-mail, and include responses from the following companies: CriticalBlue Despite the occasional marketing pitch in here, please read on and see if you agree with the situation as laid out by these 4 companies – who, by the way, distinctly do not see eye-to-eye on the topic at hand. ***************************** Here's the e-mail I sent: All – Over the last several years, I've been struck by the number of times I've heard keynotes at EDA conferences mention embedded systems & software. Similarly, leaders in the embedded software space frequently reference EDA in their keynotes. So, my questions are: 1) Why isn't there a closer alignment between these two industries? 2) Do you see a future synergy between Embedded and EDA? 3) If so, what do you foresee this looking like? If not, why? 4) The EDA industry is looking for a way to increase their market. Is involvement through M&A with Embedded system tool enterprises one-way to accomplish that? Equally, could embedded software tool vendors broaden the market by acquiring EDA companies? 5) Can you envision a unified environment? Where are the main/key bridges that need to be built to two industries/sectors together efficiently? 6) Ultimately, it's a question of the technology. Why don't EDA vendors provide embedded SW development solutions and vice versa? 7) What is the customer’s current pain point that is being left unsolved because EDA and Embedded system tool vendors do not provide a unified solution?' In other words, what business opportunities are being overlooked now and in the future? Regards, Peggy Aycinena ***************************** David Stewart 1) Why isn't there a closer alignment between these two industries? David Stewart – That’s a very good question. I think it’s primarily a throw back to when the hardware and software disciplines were very separate. Originally, semiconductor vendors did the vast majority of custom silicon design and systems companies did the vast majority of software development and as a result the two tool industries grew up to address their respective markets. This dynamic has now changed considerably, particular for semiconductor companies who do so much software development these days, but the industry "silos" have remained. 2) Do you see a future synergy between Embedded and EDA? David Stewart – I think it’s mandatory. It’s actually astonishing to me that they operate so independently. If you ask an embedded software tool vendor what they would suggest if their customer’s software execution needs to be improved, they will suggest that the customer use a faster CPU; if you ask an EDA vendor the same question they will propose the manual partitioning and design of appropriate hardware accelerators. Both are valid answers, but in neither case does the vendor take ownership of providing the customer with an immediate and viable solution. 3) If so, what do you foresee this looking like? If not, why? David Stewart – The customer is clearly suffering pain and that ought to be sufficient incentive to develop an appropriate solution. The increasing amount of embedded software development in semiconductor companies has only increased the urgency but both vendors' camps appear to be pointing at the other to deliver a solution. Start-up companies, who could be classified either as embedded software tool providers or as EDA tool vendors, are most likely to emerge with the initial solutions. The larger vendors have shown no appetite for addressing the customer’s need in this regard, which is unfortunate for the customers, but is an opportunity for start-ups and venture capitalists. 4) The EDA industry is looking for a way to increase their market. Is involvement through M&A with Embedded system tool enterprises one-way to accomplish that? Equally, could embedded software tool vendors broaden the market by acquiring EDA companies? David Stewart – It could happen either way round although, as stated above, neither industry seems willing to make the necessary investment to address the customer’s need. The most likely scenario is that one or more start-ups will emerge to address the requirement and the larger vendors will acquire them once they have some market traction. Alternatively, the opportunity associated with migrating pure embedded software onto different silicon platforms is large and diverse enough that a new standalone company could grow up around it. 5) Can you envision a unified environment? Where are the main/key bridges that need to be built to two industries/sectors together efficiently? David Stewart – Yes, I can envisage such an environment. There needs to be a flow and methodology directly from embedded software running on a general purpose processor onto a processor + coprocessor platform, which meets the product goals in terms of performance, power consumption, silicon area, reprogrammablity and development time. At the moment, each customer has to come up with their own home-grown solution. The flow needs to include predictor technology that enables software developers to know early in the design cycle that suitable programmable hardware accelerators can be developed for their software. There is a lot of emphasis these days on various C-to-RTL solutions, primarily behavioral synthesis technology, but it is very important to understand that this does not solve the problem of migrating software onto flexible silicon platforms. C-to-RTL solutions are targeted at making traditional manual RTL design more efficient and productive; their input is a C syntax description of the desired hardware which is entirely different from embedded software. The fact that this is not obvious to everyone, is just one more indicator of the gap that needs to be bridged. 6) Ultimately, it's a question of the technology. Why don't EDA vendors provide embedded SW development solutions and vice versa? David Stewart – Both vendor types would typically say that the other solutions are purchased by different departments than they are used to dealing with, and that they would need to build complete new sales channels. EDA vendors may also feel the ASPs of embedded software tools are too low to make it worthwhile. In my opinion, these are just excuses. When the degree of pain is as great as it is now in implementing software/hardware platforms, then there is a significant business opportunity. In my opinion, it’s less a question of technology and more a question of vision. 7) What is the customer’s current pain point that is being left unsolved because EDA and Embedded system tool vendors do not provide a unified solution?' In other words, what business opportunities are being overlooked now and in the future? David Stewart – Customers are being offered solutions based on an historical market segmentation, rather than a practical assessment of their needs. Systems companies do very little custom silicon design any more; semiconductor companies do much more embedded software development now. The natural way to deliver key differentiating functionality from a systems company to a semiconductor company, is as a functional piece of embedded software. This new reality dictates that something needs to change, and that both EDA and embedded software tool vendors must concentrate on the customer’s pain rather than on their perception of the market they are in. ***************************** Sven Brehmer
***************************** Mark Milligan 1) Why isn’t there a closer alignment between the EDA and embedded software industries? Mark Milligan – Historically, the users are separate. The industry provides tools separately and the market has been separate. Often, the assumption has been made in the embedded software world that the hardware has already been designed. The embedded software industry has been focused on using hardware rather than helping design it. But the world is changing and these assumptions are no longer true! 2) Do you see a future synergy between the embedded and EDA industries? Mark Milligan – From CoWare’s point of view, the answer is No. The synergy already exists! From its inception, CoWare has always had the vision of linking these two worlds together and we are now seeing a dramatic acceleration of our business opportunities due to the merging of these two industries. 3) What do you foresee this looking like? Mark Milligan – What’s changing now, is the move toward putting a system – formally comprised of discrete components (existing hardware) – onto a chip. With today’s time-to-market pressures, you can’t afford to develop the software after the SoC has come back from the fab. And, more importantly, if the design wasn’t verified with software prior to tapeout, it’s pretty much guaranteed to fail. Optimizing a device for cost, performance and functionality is not just a matter of software anymore, but both hardware and software. Traditionally, software content has been less complex and could be optimized for a given device. Now, embedded software content has moved towards a mix of open-source, legacy, proprietary differentiating code, and open platforms (third-party code is not known at the time of development). At this level of complexity, it’s no longer possible to optimize the software after the hardware is available. What’s causing the two worlds to collide, is the need for software reuse while simultaneously optimizing for each successive hardware design and being able to leverage that design across different devices. 4) Could embedded software tool vendors broaden the market by acquiring EDA companies? Mark Milligan – Software vendors are driving towards enterprise integration technologies such as .Net and Java. EDA is driving towards the foundry to address the challenges in deep-submicron design for manufacturing and yield. Customers cannot get the tools that bridge the gap between embedded software and EDA from an IP vendor, as they need tools that work for more than one supplier. There is clearly a need for a new, independent category ***************************** Serge Leef 1) Why isn't there a closer alignment between these two industries? Serge Leef – There are basically two different markets within the embedded space. By far, the largest section incorporates standard hardware platforms, focusing their design resources on embedded software development. The other, much smaller but important sector, is focused on the demanding design that require custom hardware/software development and integration. Thus far, growth in system size and complexity for the mainstream embedded market has not resulted in a compelling event that would bring the hardware and software teams together throughout the entire system development process. With few exceptions, there are only two institutionalized contact points between the teams – one at the specification phase and one at the prototype bring-up phase. The teams are not sufficiently motivated to interact at other times. Co-verification tools have been able to bridge the two processes, but have been adopted largely by design teams at the extreme high end of system size and complexity, where such collaboration is viewed as required rather than optional. 2) Do you see a future synergy between Embedded and EDA? Serge Leef – Absolutely. The synergy will be driven by growth in the number of large, complex system projects that cannot be designed economically using traditional approaches for design and verification. At that point, convergence will cease to be optional. This is also true where configurable hardware is used. Either configurable processor architectures or configurable hardware (FPGA) with processors in them demand a closer relationship between the hardware and software designers. 3) If so, what do you foresee this looking like? If not, why? Serge Leef – In the near future for complex embedded design, hardware design teams will have to accept that software is a key component in their design. In IP-dominant systems, every module comes with software that is, subsequently, stitched into different layers of software stacks. These teams will, at the very least, have to incorporate these software elements into their verification strategies. In the more distant future, embedded design teams will focus increasingly on the "system" rather than hardware and software. In that context, the design activities would focus on IP (hardware and software) evaluation and integration. The focus will shift to testing interoperability of the building blocks and validation to make sure that the entire "system" behaves correctly. The remaining, original functionality may be designed algorithmically (i.e. in C/C++) and automatically implemented as either hardware or software to meet performance, power, and cost constraints. Of course, suppliers that can provide both the hardware IP and software IP in a pre-validated form will help meet this system approach. 4) The EDA industry is looking for a way to increase their market. Is involvement through M&A with Embedded system tool enterprises one-way to accomplish that? Equally, could embedded software tool vendors broaden the market by acquiring EDA companies? Serge Leef – These are credible ideas, but the full benefits of such convergence will not be realized instantly and cannot be driven by the EDA or the embedded software companies alone. The macroscopic trends discussed above will determine when the real payoff for embedded teams will occur. Currently, the two industries sell to distinct customers and have vastly different pricing/packaging strategies and business models. Thus, there is not a natural and instant synergy. So until true disciplinary convergence takes hold in the customer communities, both businesses need to be viable on stand-alone basis. 5) Can you envision a unified environment? Where are the main/key bridges that need to be built to two industries/sectors together efficiently? Serge Leef – The bridging technology that currently exists, and is being used effectively, is hardware/software co-verification. It enables hardware teams to subject their designs to realistic scenarios by running real software on them. At the same time, it gives software teams early access to the virtual prototype. Another bridging technology being explored is system prototyping with SystemC. While it is expected to be applied relatively early in the design process, benefits are similar to co-verification. Because of its higher performance – resulting from transaction-level modeling techniques – however, this approach is much more appealing for architectural exploration and software development. An exciting bridging technology that can be used at an even earlier phase of the design process is C-based synthesis. Functionality can be described in C and automatically implemented as either hardware or software. As the technology matures, it could have revolutionary impact on system design practices. Another interesting bridging technology developed in the software world is the Unified Modeling Language (UML). This language is not specific to software, and could be used as a single unified design tool, allowing the hardware and software parts to be designed together and leaving the implementation details to synthesis tools. 6) Ultimately, it's a question of the technology. Why don't EDA vendors provide embedded SW development solutions and vice versa? Serge Leef – Some, like Mentor Graphics, do. We recognize that a system design team consists of a spectrum of engineers working on various facets of the system: application code, RTOS, protocol stacks, firmware, board hardware, multi-chip integration, ASIC/SoC design. Today, Mentor sells tools to all of these people. 7) What is the customer’s current pain point that is being left unsolved because EDA and Embedded system tool vendors do not provide a unified solution?' In other words, what business opportunities are being overlooked now and in the future? Serge Leef – As mentioned earlier, there are bridging solutions on the market today that address the needs for embedded teams building the most complex systems. But for the next level of complexity in the embedded design market, only a disruptive change in their underlying assumptions will cause them to "re-adapt" to new realities. The factors most likely to cause such disruption are: * Dramatic growth in system size and complexity,
My conclusions: a) Who needs Jerry Fiddler, when all these guys have so much to say? b) The EDA/ESW Interface is far from: Been There c) On the contrary and depending on who you listen to, the EDA/ESW Interface: Is Coming d) And most importantly, the EDA/ESW Interface story is the natural segue for the other half of the emerging trends conversation. So, please stay tuned for next week's exciting conclusion ... 2005 Emerging Trends – The EDA/ESL Interface
Peggy Aycinena owns and operates EDA Confidential. She can be reached at peggy@aycinena.com
|