Hardware, software design on collision course, panel finds

Industry insiders have been saying for some time that the traditional electronic system design model, where hardware is developed and then software development follows, is no longer adequate. For one thing, the semiconductor industry's shifting focus toward consumer electronics products has resulted in dramatically shorter market windows, placing a premium on time-to-market that the traditional model cannot support.

The desire for concurrent hardware-software codesign has created strong momentum for the development of tools and methodologies based on electronic-system level (ESL) design, which after years of promise seem to be finally coming online.

Ajoy Bose, president, chairman and CEO of Atrenta Inc. (San Jose, Calif.), said hardware design is becoming increasingly cost prohibitive, noting as evidence the shrinking number of total industry design starts and the trend toward platform-based design. This, he said, is spurring a movement toward creating libraries of reusable software models that could make software design more affordable and less time consuming.

Chris Rowen, president and CEO of Tensilica Inc. (Santa Clara, Calif.), pointed to a bifurcation in the market between mass-market products that need greater optimization and can afford to invest in more hardware development and niche-market products that for economic reasons need to rely more heavily on platform-based designs and greater software development.

Sanjay Srivastava, president and CEO of Denali Software Inc. (Palo Alto, Calif.), said part of the problem is that while hardware design productivity has steadily grown at a healthy annual rate, software design productivity is growing at a much slower rate — about 8 percent.

"A great deal of concurrency already exists today in the way that people develop hardware and software," said A.K. Kalekos, vice president of marketing and business development at ESL tool provider CoWare (San Jose). "But concurrent hardware-software design remains the holy grail."

James Ready, president and CEO of MonteVista Software Inc., said the hardware-software codesign problem is exacerbated by the dramatic growth in system software content over the years. Today's handheld consumer devices, he said, have significantly more software than a Army tank did 30 years ago.

Bose said one of the problems facing EDA companies, including Atrenta — which provides predictive development tools for chips, systems and embedded software developers — is that while chip makers are accustomed to paying high prices for design tools, system developers are not.

Bose said tools currently being used for embedded software development are "very primitive." He predicted that companies would push EDA vendors to provide stronger solutions for embedded software development.

Kalekos said the mistake mainstream EDA vendors are making is that that they no longer provide competitive differentiation. Time-to-market, he said, is no longer a competitive advantage, but a requirement to secure business. To be successful, he said, a tool vendor must provide value that cannot be obtained elsewhere, something he believes can be done with ESL.

Kalekos said the market for ESL tools will eventually be worth more than $1 billion annually — perhaps significantly more. In the future, he said, ESL and intellectual property would be the biggest growth areas for EDA.

Kalekos also warned that a potential crisis in software development is coming more rapidly than most people realize, with the advent of multiprocessor systems. "People don't know how to write architectures for embedded software development on multiprocessor systems," he said.