eda platform

An Industry at the Crossroads As DAC approaches, Steve takes a broad look at recent EDA industry trends, and how these address the rapidly changing needs of designers in response to emerging design and semiconductor technologies. by Steven E. Schulz

Well, it's that DAC time of year again, and I feel the need to be profound. Yes, that's it--profound and wise. So let's take a broad look at this EDA industry and where things seem to be headed.

Ten years ago, the commercial EDA industry was in its infancy. Five years ago, it merely acted like it. Today, however, things are quite different. From having its own market indices (thanks to the EDA Consortium) and the attention of Wall Street analysts; to new business-oriented thrusts in consulting and design; to more attention on understanding customers, things are certainly different. I characterize the industry as having reached puberty: it's grown up physically, but doesn't yet know what to do with itself. In an awkward, teenager-like manner, we see EDA companies trying to form tentative relationships with their customers in areas where they have no prior experience. Rather than stay within their roots, many choose to expand in search of higher growth and profit margins.

Let's face it--the EDA business is becoming difficult. Just when the EDA vendors finally understood the timing-driven­everything problem, the semiconductor industry began another shift: from deep submicron to very deep submicron. Now there are many more challenges to completing a working design, and any one of them can be a new barrier to entry when selling tools. The intricacy of new EDA tool development requirements are rising along with design complexity. I can almost hear EDA professionals saying, "Please, I just wanted to write software, not get a Ph.D. in device physics!" Of course, many users are still buying today's tools, but the emphasis on IC physical design tools is merely a sign of much more ominous things to come.

Lately, we've seen a mass migration from tool development to design consulting services. Despite those who stood at the Wailing Wall predicting doom, gloom, and general despair for all designers, the services thrust has generally been a good idea. A look at EDAC's growth numbers shows clearly that EDA suppliers found opportunities to fill genuine customer needs, so it's surely been good for them and their customers. Fortunately, the scarcity of engineering resources and high electronics growth rates have kept EDA consultants employed. Contrary to popular myth, the user companies' intellectual property is still quite safe, as most of the consulting is actually in physical layout and tool methodology deployment. Consultants won't replace valued designers; they just add flexibility to corporate staffs in certain areas with variable demand. Yet as the vendors are beginning to find out, consulting expertise doesn't possess the same economies of scale that products do: it's not cost-effective to mass-produce consultants with average skill levels.

The real problem is that the EDA industry has lost much needed momentum for its own retooling. The emphasis on services has, in effect, made everyone a user in that no one is leading the future of the EDA tools business. On many occasions, eager sales forces have pushed through consulting contracts, only to find that the staffing of qualified personnel in that region is a problem. In what amounts to a human shell game, a "parts is parts" mentality takes over. Rumors of entire tool development groups becoming cannibalized to fulfill services contracts have spread, forcing tool technology to wait. Tool developers suddenly become high-dollar consultants to user companies. As part of their "service," they push tools available only when bundled with consulting contracts, which gives rise to poor excuses for hard-to-use, buggy software technology.

Having several years' experience under their belts, the EDA companies that emphasize service realize all too well that they can't achieve companywide growth goals without new, high-value products. Rather than scale the high wall of silicon-based challenges for new design automation tool flows, some have instead chosen to focus on the design product itself. Why not? IP is by its very nature a high-margin concept, and it fits well with current market trends. So the new focus of some EDA companies is to design, broker, or integrate IP. In this way, the tool suppliers­cum­consultants are again migrating to yet another area: the fabless design house. It is indeed a very rocky road out there, as many are finding. The design house arena is a risky investment for EDA vendors--as their core competency is neither novel designs nor the intricacies of manufacturing processes and product design. They find themselves sandwiched between two areas where the minimum knowledge requirements are rapidly rising. For their sake, let's hope their expertise can catch up (and keep up).

Meanwhile, it isn't difficult to see that the user companies, particularly those depending on VDSM technologies, are becoming very worried. For the last three years, many of them have been trying to explain why it's a flow problem, not just a tool problem (if that isn't your thing, I'll explain it in a future column). Perhaps you've noticed an increase in the level of semiconductor and user participation driving various EDA industry standards efforts--it's no coincidence. It's also no coincidence that more of the recent standards involve APIs (application programming interfaces), a more sophisticated and complicated means of communication between tools. The root of the problem is more than just the volume of data exchanged between tools or the increasing accuracy requirements. The fundamental issue is the need for managing greater numbers of design requirements and constraints simultaneously in order to achieve an acceptable design result.

The need for agreement between interacting tool algorithms and data structures will require a new approach to architecting, integrating, packaging, and selling EDA tools. More than that, it will also change the roles of responsibility and ownership for advanced calculations and library characterization in the design flow. Managing all of that data will require more shared views of a design, taking advantage of object-oriented database technologies. If you think I'm going to use the word "framework," don't worry; what we need goes much farther and deeper. Fortunately, the needed technology already exists. We are awaiting its adoption by EDA vendors and its interoperability for users.

Will EDA tool development return to the in-house model from which the commercial industry was born? Or will the tool developers rise up to the challenges of designing very complex functions in a very complex world? Why shouldn't we expect that our old, familiar approach to connecting EDA tools will change in response to significant changes in design needs?

One positive aspect of having EDA companies directly involved in product design is that they, too, should be experiencing the same problems and frustrations as traditional users. But if no one in this industry takes leadership to solve the problem, it's unimaginable that the entire electronics industry should be held hostage in this design bottleneck forever. The entire EDA industry is several orders of magnitude smaller than its customer base, so its best interests will be served by investing in the next-generation tool flow architecture. Even if user companies make the investments in-house, though, no company could even hope to supply a full line of tools for the new system-on-a-chip era. Furthermore, the long lead time required for such a solution is estimated to restrict design growth opportunities for the next three to five years.

Surely the situation can't really be this bad, or people would already be solving the problem, right? Well, in fact, the problem is worse because not everyone will immediately require the latest technology. Just as there's still a relatively healthy market for 0.5-µm designs, even as 0.18 µm enters the fray, so too will a sizable number of designers not require the new tool architecture--yet. Those who should be most worried are at the "bleeding edge" of design needs, at 0.18 µm and below. The system-on-a-chip designer may find that by achieving the density, cost savings, or performance required, he becomes one of those at risk. Although the electronics industry should see enormous growth from system-on-a-chip and design reuse techniques, realize that in terms of EDA design seats sold, this type of design still represents a minority. Thus you may find the EDA industry isn't committed to solving your problem in your time frame.

Plenty of EDA vendors still care about advancing automation technology and are dedicated to novel solutions to the problem. Mostly, they're the niche and start-up suppliers. Unfortunately, they're the ones who cannot integrate their advanced capabilities until the large, broad-line suppliers change the architecture. The existing integration options are often ASCII file­based and are at best loosely coupled to other algorithms. So vendors know the problem all too well, but they can't solve it by themselves.

Unless this year's DAC contains real surprises, one can expect that these problems will remain unanswered for yet another year. Meanwhile, I expect more noise from the EDA industry about its design IP and services business than about preparing advanced commercial design flow capabilities to take advantage of them in the coming era. If you see me at DAC this year, let me know whether my prediction was right or wrong--I'd rather be wrong.

Contributing Editor Steven E. Schulz is a senior member of the technical staff in Texas Instruments, Inc.'s Worldwide ASIC group in Dallas, where he's responsible for defining the ASIC backplane roadmap for TI's system-level integration strategy. He also serves as the chairman of the EDA Industry Council PTAB, a technical advisory board for EDA standards directions and coordination, and is the president of the board of directors of VHDL International.

To voice an opinion on this or any Integrated System Design article, please email your message to miker@isdmag.com.

integrated system design  July 1998