October 15, 2000 | |||||||||||||||||||
Home : Design | |||||||||||||||||||
|
|||||||||||||||||||
|
|||||||||||||||||||
|
Panel pits quality design versus EDA toolsBy Nicolas Mokhoff
As market windows close ever faster on system-on-a-chip designs, designers need to make frequent trade-offs between satisfying results from imperfect tools or awaiting "perfect" data from "perfect" tools. "Integrated EDA tools lag behind Moore's Law, and are the primary cause of the 'non-correctness' being experienced with many complex IC designs today, " said Robert N. Blair, chairman of RNB International, and the panel's organizer.
"You can live with mere results over quality results for checking out technical readiness and development, but you need quality over mere results for product development," said Ghassan Yacoub, advanced EDA tools program manager at Intel Corp. (Santa Clara, Calif.).
Most panelists supported Yacoub. Nancy Nettleton, ASIC design technology manager at Sun Microsystems Inc. (Mountain View, Calif.), said, " Quality becomes a barrier when it impacts productivity."
The notion of having a UL-style label for EDA tools was ridiculed, although the idea of having a third party ensure tool quality was not dismissed. "The UL [Underwriter Laboratories] approach used in the consumer industry is unworkable for evaluating software tools," said Dan Lincoln, chief technical officer of ASIC International Inc. (Oak Ridge, Tenn.), a design services firm. "Needed is statistical correlation of data that can be accepted by the industry. I believe designers will refrain from using UL-designated tools because these tools will never be at the bleeding-edge."
Some tool users in the audience were more adamant about getting EDA tool vendors to become responsive to their needs. "Cars have warranties and one can be assured that the vehicle you bought will work. The same cannot be said of a million-dollar EDA tool, and vendors should be held legally accountable for tools that don't deliver on their promises," said one attendee. Panelists quickly assured him that such accountability would retard industry creativity and place EDA tool producers into a self-defensive mode. "Remember that we stand on the collective shoulders of the tool developers who came before us, " said Lincoln. "The sophistication of today's tools reflects the inputs of a lot of intelligent minds."
"Rather than blame vendors for inadequate tools, we need to characterize the class of design problems we have and choose the tools that fit that class of designs," said Sun's Nettleton. "If they perform according to the original characterizations, then we have done our job properly."
Addressing the proper design flow, Nettleton said: "The primary vehicle for selection of a quality EDA flow is to review design results produced on design problems roughly equivalent to those planned for the flow."
Nettleton said two degrees of complexity come into play in such circumstances. First, for leading-edge design problems, an existing design collateral may not exist for a roughly equivalent design. And second, the quality of design results can vary as much or more with the skill set of the design team as with the design of the flow. "In general, I would never select a design flow without also considering the team that will run it, whether it is an in-house or contracted team," she said.
|
|