Panelists debate role of design, verification teams

Richard Goering
(02/16/2005 10:30 PM EST)

  SANTA CRUZ, Calif. — There was no disagreement about the need for better quality designs at a Design and Verification Conference (DVCon) panel here Wednesday (Feb. 16). But panelists debated the extent to which design and verification should be done by separate teams.

The panel was entitled, "Design quality in — the better design paradigm." The basic idea behind the panel is that it's easier to avoid bugs than to find them, and that chip design practices must improve in order to reduce the need for functional verification.

Andrew Piziali, senior product engineer at Verisity Inc., said it's important to have separate design and verification teams. This, he said, will allow an "independent" team to interpret the design specification.

He also said specifications must be an integral part of the design process, as opposed to after-the-fact documentation. "We can improve first-time silicon quality two-fold by integrating the functional spec into the design process," he said. This, he said, will help expose "flawed ideas" earlier in the design cycle.

Verification engineers should review the earliest design specs, Piziali said. "Verification must be elevated to the top of the food chain," he noted. "Too often there are subtle messages that it's a second-class job."

Kevin Normoyle, architect at Azul Systems, said it's wrong to organize around design and verification teams. "We need to organize around skill sets," he said. "We need to change the design when it's lousy."

Normoyle said that "excess complexity" is the cause of many silicon problems. "In most places where I've seen a lot of bugs, it's in useless stuff that adds no value to the product," he said.

Limor Fix, principal engineer at Intel Research Labs, cited four areas that need attention in order for verification to improve. First is to move to a higher level of design abstraction and start with assertions, or design intent, rather than the design. "We do it in the wrong order," she said. "We design and then annotate that this is what we want."

Other needed advancements, Fix said, are design reuse without alteration, incremental design and validation, and a more automated approach to both dynamic and formal coverage. While the first of these is a management problem, there are technology challenges in the latter two areas, she noted.

The separation into separate design and verification teams did not help predictability, said Harry Foster, chief methodologist at Jasper Design Systems. "We need the separation, but there are cases where responsibility goes out the door, and you throw it over the wall to verification rather than have an integrated team."

Designers are part of the verification process, Foster emphasized. "We need to move from verification after the fact to a prove-as-you-design type of approach," he said.

Gary Smith, chief EDA analyst at Gartner Dataquest, said that "the concept of the verification team needs to be reconsidered." When there are separate design and verification teams, he said, verification productivity drops by 25 percent. Smith said that regression testing and emulation can be done by a separate team, "but the rest of the verification team needs to be moved down to the design team."

Smith also made a pitch for electronic system level (ESL) verification. "Both hardware and software verification must be integrated during the entire design flow," he said.

All material on this site Copyright © 2005 CMP Media LLC. All rights reserved.
Privacy Statement | Your California Privacy Rights | Terms of Service