DATE with Destiny ...
By the time Thursday dawned at DATE'09, like many people I was beginning to run out of steam and/or attention span. Nonetheless, Thursday included a full day's treasure trove of discussion about Everything Multicore.
I was involved in industry conversations for a portion of Thursday (see above), but did manage to late-morning presentations from ARM's John Goodacre talking about multicore technologies, and NXP's Peter Kollig talking about heterogeneous multicore platforms for consumer multimedia applications. I have to say, despite all the doom and gloom from many speakers at DATE about how tough multicore is, or will be, both NXP and ARM talk as if multicore is already a reality in life. True, there are a lot of additional problems to resolve, but these guys are shipping products that are seriously multicore, or helping their customers to do it.
I actually started out Thursday attending the second half of Grant Martin's tutorial on multicore technology. Just as with the folks from ARM and NXP, Martin carefully articulated the real issues, and possible solutions, for those who are working to add new cores to legacy cores in increasingly multicore devices.
The lunchtime keynote at DATE was also in the spirit of the Special Day on Multicore. The CTO of STMicro, Eric Flamand, presented a detailed, almost academic talk on the complexities at hand. His argument: replace heterogeneous, mixed hardware and software solutions to a design problem with a multicore "fabric," the cores being activated as needed, the others left dormant. Flamand's implication, if I understood him, was that there is so much real estate on-chip today, why not lay it out in advance with multicore in mind, deciding after the fact how to utilize the capabilities there.
The afternoon sessions at DATE on Thursday included a panel discussion evaluating the role startups will play in driving multicore technologies. I only stayed for part of the session, long enough to hear Grant Martin say, to get funding "startups must offer significant differentiation over what is already out there. The last thing I want to hear is another me-too [offering], with only a 15 or 20 percent improvement over existing products."
As Thursday drew to a close, the Exhibition Hall also began to bring down the curtain on DATE. As booths came down, I went upstairs at the Convention Center to attend part of a technical session on mixed-signal and RF testing.
Tima Labs' H-G Stratigopoulos described a proposed strategy using machine learning to attack the problem of testing RF chips. Stratigopoulos said the Tima Labs technique required only 30 minutes to sample 1 million on-chip devices, and only 2 minutes to generate the required guardbands necessitated by the findings.
After his presentation, we spoke about the commercial applications of the technology. He said full implementation would require shutting down testing facilities to "re-tool" if this type of strategy is to be explored. I predict, given the promise of machine learning, that it won't be long before the strategy is fully utilized across the width and breadth of electronic device manufacturing and test. At that point, the topic will no longer be relegated to the last session on the last day at DATE.
Thursday ended on a high note for the Editorial Community attending DATE'09. Thanks to a gracious invite from the ARTEMISIA Association, we were treated to dinner and a concert on the top floor of the convention center. The 'concert' was actually a robotics contest, pitting student design teams from Australia, The Netherlands, and Sweden. The task was to build robots that could play instruments.
There were a variety of rules for the contest, a tough panel of judges, and big prizes - 3000 euros, 4000 euros, and 8000 euros for the third, second, and first place teams, respectively. The team from Eindhoven won, not only because their instruments were spot-on musically and could follow a human conductor, but because the electronics for one of the other teams failed during the performance.
The instruments in the various bands included a guitar, a drum set, a flute, an accordion, and a piano. The Eindhoven team had participated in the contest last year and was able to learn from last year's errors. The losers in this year's contest were new to the effort, but the prizes are so large, surely they'll be back to try again next year.
What a spectacular way to end DATE'09. Enthusiastic engineering students, robotics - mostly built on programmable devices, fascinating music, a kind of Robots Have Talent pre-dinner extravaganza, and a fabulous meal. It doesn't get much better than that!
Next year DATE'10 will be in Dresden. Much may change between now and March 2010, but no matter if you have to be pushed or pulled, you should plan to attend.