Between The Lines

Cadence is definitively not going to open up its proprietary SKILL Pcell language, Magma and Synopsys seem to be making up after a long legal dispute, and Cadence’s DRC/LVS tool isn’t ready for primetime but Magma’s DRC/LVS is and its CEO claims it is gaining share.

These are some of the more provocative takeaways from yesterday’s “EDA Troublemaker’s panel” (Formerly “EDA Bigwigs panel”) moderated by industry gadfly John Cooley.

Every year, Cooley solicits biting questions from EDA users and industry watchers and selects a subset of those questions to ask a panel of EDA executives and analysts.

This year’s panel consisted of Rajeev Madhavan, CEO of Magma Design Automation; John Chilton, senior vice president of marketing at Synopsys; Jue-Hsien Chern, vice president and general manager of the deep submicron division at Mentor Graphics; Ted Vucurevich, CTO of Cadence Design Systems; Vic Kulkarni, CEO of Sequence Design; Atul Sharan, CEO of Clear Shape; Brett Cline, vice president of sales and marketing at Forte; and EDA industry analyst Gary Smith (garysmitheda.com).

Over the last year, Cadence Design Systems has been clashing with the other big vendors in the industry over low power standards and its failure to give other vendors access to its SKILL pcell analog library description language, which many vendors consider to be key to making the OpenAccess common database, which Cadence started years ago, a success. Cadence’s SKILL has helped the company maintain a dominant marketshare lead in analog/custom digital layout for nearly two decades.

During the panel, Cadence’s Vucurevich fielded several questions regarding CPF, SKILL and its part in OpenAccess.

When asked why Cadence has not given its competitors access to SKILL, Vucerevich said that Cadence has done its part to be open and has established industry standards by noting the company’s many years of work on OpenAccess. “Over 200 engineering years of R&D have been made available to the industry at no charge,” said Vucurevich, referring to OpenAccess. That’s fourth generation of database technology put out in the open marketplace to be used.”

But Madhavan and Chilton took exception to that.

“The question is free engineering to what?” said Madhavan. “To me, releasing OpenAccess but holding back pcells and holding back a number of pieces that are associated with keeping the analog monopoly franchise going is not open,” said Madhavan, noting that if Cadence wanted OpenAccess to really be a success, it would need to provide access to SKILL and an in-memory data model, not just a database.

Chilton said he believes the OpenAccess database is “reasonably open” but is still pretty much controlled by Cadence. “It would be much more interesting if SKILL and pcells were open, then the really last closed citadel in EDA would be open for interoperability.”

Vucurevich was really quick to dispel any notions that Cadence will open SKILL any time soon. “Bluntly, that’s not going to happen,” said Vucurevich. “…we are not going to open up from a competitive point of view and say ‘here’s a free ride for coming in and trying to take accounts.’”

Failing Cadence opening up access to Skill, EDA startup Ciranova has devised a backdoor approach that allows third party tools to access skill via OpenAccess. Vucurevich noted that Cadence is “neither pleased nor displeased” with Ciranova’s efforts, saying Ciranova effort “is not a problem at all.”

Madhavan noted that to grow its value, the EDA industry needs to stop competing on formats and compete on technologies that run on common formats.

Panelists also took Cadence to task for its lack of openness regarding CPF. Cadence did not allow large competitors to have a say in the development of the format. Thus many competitors have rallied around Accellera’s UPF (United Power Format), which coincidentally was approved as an Accellera standard at DVCon.

Cadence announced that its tools already to support the format, but non-CPI members have not even seen the format. However, in an earlier DVCon panel, LSI’s said the two formats are indeed very similar, which holds promise for a future melding of the two formats.