June 16, 2008
DAC 2008 – Trepidation to Triumph
Please note that contributed articles, blog entries, and comments posted on EDACafe.com are the views and opinion of the author and do not necessarily represent the views and opinions of the management and staff of Internet Business Systems and its subsidiary web-sites.
Q – If ARM was a no-show, why were John Goodenough and Rob Aitken everywhere, as well?
ARM has been working so hard in recent years on their presence at DAC, some were starting to believe
DAC really stood for
Design with ARM Conference. Hence, it was a mystery when they pulled out as an exhibitor this year. Maybe they figured they’ve got so much market share at this point, there’s no need to exhibit at DAC any more; they’re already so well known to the designers. Besides, they’ve got their own Developer’s Conference in Silicon Valley, so why spend dollars on DAC?
Nonetheless, John Goodenough appeared on the EDAC Game Show and
Rob Aitken appeared on the Thursday afternoon panel on DFM chaired by
Andrew Kahng. Along with panelists
Luigi Capodieci from AMD,
Cliff Hou from TSMC,
Vivek Singh from Intel,
Lars Liebmann from IBM, and
Riko Radojcic from Qualcomm, the discussion was well underway when I arrived.
Rob Aitken had generated a lot of agitated back-and-forth with his slide listing the “Most Flagrant DFM Hypes” of the recent past: The Yield Disaster Curve; Yield-Driven Synthesis; and Secret Sauce DFM (this last, only useful as an engine for generating VC funding). Meanwhile, Radojcic said all of the stuff described as DFM today is part of a magic soup developed to make things manufacturable; Hou hinted more DFM tools are coming; and Liebmann said grad students should work on models for stressed silicon now that good models for etch and litho have been produced.
At session’s end, I asked Kahng about my impression that EDA didn’t appear to play a role in DFM as laid out by his panelists. He disagreed: “Point tools are not DFM, but EDA does have a role.” Aitken responded to the same question: Consolidation in the EDA industry will increase the opportunity for the tool vendors to contribute to DFM.
Upstairs, Downstairs at DAC …
Q – Is DAC a technical conference or a commercial conference?
Depending on your point of view, DAC may be only one or only the other. And it’s this overt schizophrenia that oftentimes leads to
controversy and confusion about DAC.
If the population that’s Downstairs on the Exhibition Hall never attends the technical sessions upstairs, and those at the technical sessions Upstairs rarely make it down to the show floor Downstairs, there really are two different events going on. To add to the schism, there’s a certain ethereal cerebralism that characterizes the crowds mingling at the technical sessions Upstairs, while there’s a certain Cooley-esque grittiness to the “We’re the real engineers in EDA” that characterizes the crowds in and around the booths Downstairs.
Of course, you’re welcome to yell at me for pointing this out, but that would simply be shooting the messenger. We all pretty much know this is the case at DAC, even though some are reluctant to talk about it. Others, however, have offered some solutions, which is where the Pavilion Panels come in.
Q – What is a Pavilion Panel?
I attended 4 Pavilion Panels (out of 20) over the course of the week at DAC. One was about the pending crisis in 22-nanometer design, a conversation moderated by Mentor Graphics
Joe Sawicki, that included PFD Solutions’
John Kibarian, TSMC’s
S.T. Juang, and IBM’s
Lars Liebmann. This was a technical discussion, with spillovers into the business of manufacturing. Per Kibarian, “At the end of the day, it’s about variability.” End of discussion.
The second panel was about “the view from Sand Hill Road,” which presumed you knew the VCs of Silicon Valley aren’t very innovative about where they reside, all living side-by-side in Palo Alto. This conversation was moderated by EDA VC legend,
Lucio Lanza, and included Argon Venture’s
Juan Antonio Carballo, Nollenburger’s
Kent Shimasaki, and IRC’s
Erach Desai. These panelists talked business and explored the blame game about why there’s not more VC money available in EDA. Carballo also predicted Open Source software will knock the EDA industry on its keister. Others were not so convinced.
The third panel I attended was about behavioral synthesis, moderated by another VC guru, Vista Venture’s
Jim Hogan, and included UCLA’s
Jason Cong, TI’s
Loic Le Toumelin, and Calibra’s
Jan Willis. When I arrived, it had devolved into a discussion with members of the audience as to what (and who) defines behavioral synthesis – a technical conversation that also has huge economic implications for the industry.
The fourth panel was my one-on-one conversation on Thursday morning with DAC Chair
Limor Fix about career development. This topic was a book-end discussion to the keynote given on Monday at the Women’s Workshop by Magma’s
Mar Hershenson about the intricacies of networking and negotiation, and their impact on sales, hiring, and garnering venture capital for your next start-up.
My conversation with Fix was about PhD versus no PhD, and how to stay positive, professional, and pro-active about advancement throughout a technical career. It also was about how to manage your manager. Again, a conversation with an overtly business-slant.
So, the answer to the question of what defines a Pavilion Panel is, there is no answer. Any topic fits the venue, in my opinion, which may or may not help to resolve the controversy of what constitutes the fundamental motivation for DAC.
Q – Aren’t there panels Upstairs that could be Downstairs?
Absolutely! For instance, Upstairs on Tuesday you could have attended a panel entitled “What the industry needs from the incoming U.S. Administration.”
Moderated by Chapman University’s
Pete Weitzner – EVE’s
Luc Burgun, Agilent’s
Todd Cutler, SEMI’s
Vicki Hadfield, IEEE’s
Russell Lefevre, Magma’s
Clayton Parker, and Tensilica’s
Chris Rowen attempted to make heads or tails of a confused Federal policy on immigration, H-1B visas, R&D funding for science and engineering, and a bucket of other concerns about energy and environmental issues that seem obvious to the technology sector and not so obvious to the political sector. This group of panelists, highly educated and thoughtful, is clearly holding out hope for new policies, come next January. Only time will tell if their optimism is warranted. Meanwhile, if they had frustrations, those concerns were not so obviously showcased on the panel.
While this superb panel was underway in one room, another session was going on down the hall that definitely belonged Upstairs – if you think Upstairs is the provenance of technology.
Kurt Keutzer was giving a talk on “Parallelizing CAD” to an enraptured crowd of 200+ technology folks. He described the problems in CAD, and attempted to defend narrowing the problem set for research purposes to (somewhat) accessible graphing algorithms which might lend themselves to first-pass attempts at parallelization (try saying that 10 times fast).
After much detail (see the DAC Proceedings for additional info on many of the presentations this past week in Anaheim), Keutzer concluded: “The key to [all of this] is to find a way to partition [the CAD algorithms] into highly independent modules.” He added a disclaimer, saying his work in this area is only in its preliminary stages, so come back next year to hear more.
Keutzer’s talk was a segue to a panel later in the day in that same track about ManyCore processors. The panel included Magma’s
Anirudh Devgan, Intel’s
Desmond Kirkpatrick, Synopsys’
Steve Meier, Mentor’s
Duaine Pryor, and Cadence’s
Tom Spyrou, and was moderated by Keutzer. It’s difficult to do justice to the complex conversation here, but please note at this juncture that you’re out of step if you don’t know there’s a fork in the road developing over terminology.
You can find the full EDACafe event calendar here.
To read more news, click here.
-- Peggy Aycinena, EDACafe.com Contributing Editor.