SANTA CLARA, Calif. A top-to-bottom change in corporate culture is necessary for any company wishing to develop or use system-on-chip technology successfully, according to a panel discussion at this week's IP2000 system-on-chip (SoC) conference.
If semiconductor intellectual property (IP) is to become an important part of a company, attitudes from the executive level to the workbench will have to shift, incentives and accounting procedures will need to be revised, and the role of marketing as a forecasting organization will become critical.
"The design reuse culture needs to change," said Mark Birnbaum, director of strategic technology for Fujitsu Microelectronics Inc. and an executive with the Virtual Socket Interface Alliance (VSIA). " The VSIA benefits are unlikely to occur unless reuse from multiple sources is supported by SoC companies. Companies will have to change from design for one use to design for multiple use."
Barriers to this change exist at three levels, Birnbaum said: the executive level; in middle management; and among engineers.
At the executive level, Birnbaum said managers tend to focus on near-term revenues, "so investment for a midterm return is usually penalized." Accounting and incentive structures also penalize anything that shifts resources away from near-term revenue, he said.
"Upper management has to show that the company is serious about the SoC business," Birnbaum said. "They have to change the accounting and incentives to reward [IP] sharing that optimizes company profit rather than local profit, plus rewards near-term investment in reusable IP to achieve revenue gains in the medium term."
Birnbaum said middle management serves as a barrier to reuse by focusing on one-use products rather than on product families, by focusing on near-term revenues rather than time-to-market or reuse, and by running business units that optimize their own revenues rather than company revenues. This tends to penali ze inter-group cooperation and IP sharing, he said.
To remove middle-management barriers to IP sharing, Birnbaum suggested that marketing teams need to plan whole families of fast time-to-market products, rather than one-use products focused on vertical markets. Middle manager should also reward the sharing of IP for the company's benefit, and focus resources on key skills areas.
Birnbaum pointed out several obstacles at the engineering level: lack of expertise; a focus on near-term issues; the constant desire to tweak blocks; and the misapprehension that IP block reuse was a non-creative way to design. To change those attitudes, Birnbaum suggested that engineers work with marketers to define 'family product' variations. They should also try to learn from the systems approach to design, and change their attitudes, he said.
Danny Ryan, IP libraries manager for Cypress Semiconductor Corp., said his efforts to build a reuse structure within Cypress were initially bas ed on three dimensions the technical, business and strategic. But he found that he needed a forth dimension that of culture.
On the business front, Ryan said he had to prove the benefits of design reuse with measures such as design cycle times and first-pass yields, and by suggesting that an internal reuse structure could lead to a business wherein Cypress could sell IP externally. On the strategic front, he said he chose to form an IP-creation rather than an IP-collection culture within Cypress, believing that this would support the company's shift from a products-focus to a market-focus strategy, he said.
The technical issues of IP reuse were "probably the easiest to tackle," Ryan said, and revolved around good infrastructure, well-defined processes and clear standards.
But cultural issues were problematic. "Managing cultural change is very difficult you need to wield a baseball bat," he said. "But we have a CEO who can do that."
Tim Pontius, engineering fellow at Phil ips Semiconductors, said the key to creating a reuse culture within a company was to start by answering the question, "Do you believe that reuse is truly efficient?"
"Reuse is not new we've had it around forever, for example in the reuse of knowledge," Pontius said. "It's easier to build an AND gate for your own use rather than for multiple uses, but no one would question doing that. The same holds true for larger blocks."
Pontius was firm in his belief that design for reuse had to strike a balance between the perfectly reusable IP block and the one you need now and can afford to produce. "Reuse is not free. You need guidelines but you have to strike a balance.
"You need to support reuse by making it a goal," he said. "The goals of your design cannot be taping out the design or reaching production. If that's your goal you'll never get reuse."
Pontius argued that because the level of reusability is a delivered block, there has to be a balance between the ideal and the practical."IP developers have to be customer focused and IP users have to be creative," he said.
"If you create a reuse culture, you have to manage it or it will die," Pontius said. "You need to staff IP groups with the appropriate skill sets and personalities trusted designers who develop good products. If you have respected designers creating IP, your users won't have to test it [so much]."
The true test of a reuse culture, Pontius said, was getting "second- or third-generation products with reuse of IP and new IP. Then you know you have successfully created a successful reuse culture."
Thomas Harms, marketing and applications manager for Motorola's SoC design technology in Europe, elaborated on the issues of corporate backing for IP reuse. Companies need to provide reuse incentives on an organizational and an individual level, along with high-level management support, he said. Companies need education and awareness programs, design services and support, and the collaborative de finition and development of the reuse infrastructure. Companies would also need a dedicated organization to drive the methodology shift.
"You need to teach engineers more about reuse," Harms said. "What it is, why it's good for us, how it helps the business."
Dan Weed, who runs LSI Logic Corp.'s worldwide customer engineering efforts as well as the CoreWare development organization, said: "You've got to have a strategy for reuse built on a reuse philosophy."
Call for standards
"You have to go back to standards, a standard flow, and you have to go back to the best cores," he said. "Cores do not support themselves, so you have to make sure you have support for that."
As part of its effort to use standards to ease reuse, Weed announced that LSI Logic will port all of its CoreWare blocks to the AMBA on-chip bus structure specified by ARM Holdings. "We've chosen AMBA bus as a standard to glue all those things together," he said.
Weed identified some major challenges for those wh o take up reuse methodologies, including the management of in-context effects of blocks that interact with one another as they are integrated into a design. Maintenance is a big cost for library development, he said.
Luke Collins is editor of Electronics Times, EE Times' sister publication in the U.K.