July 31, 2006
DAC 2006: As the world turns, the pendulum swings
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.
of this information is a) old news, b) new news, c) daunting news, or d) just plain daunting.
5 - Good Night and Good Luck. Why is journalism such a hotly debated topic these days? Is it possible to seek out the truth, report the truth, and bear up with the consequences? Within the EDA Press Corps, the answers are looking to be, more and more - no, no, and no.
4 - You can't have your cake and eat it, too, yet the big EDA vendors talk as if they can. At one moment, they're celebrating the conviviality of EDAC and the next moment they're telling you that all customers really want is to have a single vendor supplying them with all of the tools they need to do design. I'm sure it's not my imagination that this mixed message continues to be served up with wine and cheese.
3 - The SEC will soon be announcing a number of additions to their alleged Rogues Gallery of The Indicted who have been naughty with regards to stock-option backdating. The rumor mill says some of our friends within the EDA industry may be on the list. Either spend time whispering about the rumors, or find other things to occupy yourself with in these brief moments we have on this earth. The SEC will let us know, in the fullness of time, what they've got on whom.
2 - Analog and mixed-signal design continue to be the toughest tasks in the industry, but at times seem to get the least amount of attention or buzz. Eventually, that's got to change, but how to make that happen still seems to elude the industry.
1 - Encyclopedic TI Keynote. On Tuesday morning, during the opening session at DAC, Senior Vice President and CTO at Texas Instruments, Hans Stork, presented a definitive talk on all of the technology behind the current generation of 65-nanometer single-chip cell phones. Why, you may ask, is this the Number One item on my DAC list of things to be jaded about? Well, Stork's talk was so encyclopedic, and involved so many aspects of the technology, it drove home the point that mastering it all, and solving the universe of complex design, manufacturing, and business issues related to this type of sophisticated product development, is almost beyond imagining.
Top 10 Reasons why traditional multi-vendor design, development, and verification flows are inefficient
10 - People can't agree on definitions. It used to be, what's an SOC? Now it's, what's ESL or DFM?
9 - Standard take forever, and that's not just a knock against IEEE or any other standards body. It just takes time, pure and simple.
8 - People can't agree on standards. It takes time and patience, with a heavy dollop of maturity, to build consensus. Blessed are the Peace Makers. Their work is the business of building standards.
7 - Companies think their technology should be the de-facto standard. Why not, if it's got dominant market share and gives the company a competitive edge?
6 - IP integration problems. From NIH, to NREs and ECOs, IP is the toughest solution you'll ever love to the quickest TTM. However, collecting, cataloging, updating, and verifying IP - not to mention establishing standardized wrappers - might consume a few minutes of your time here and there.
5 - Lack of interoperability across flows. All the nice words and efforts notwithstanding, this is still a thornier issue than anybody wants to acknowledge.
4 - Incompatible database configurations - with apologies to OpenAccess.
3 - Competition is often counter-productive to cooperation. This is not news. Trying to work together with customers who are also sometimes competitors is tough going. Whatever happened to co-opetition? Well, whatever happened to World Peace?
2 - People can't agree on what beer is best. Cultural differences across corporations, cultures, and time zones can be real buzz killers when it comes to multi-vendor flows.
1 - But ego remains the ultimate stumbling block in any attempt to work across vendors to provide methodologies and/or end products. It's up to management, and management's management, to resolve this most painful of human impediments.
Top 10 Reasons why ESL must succeed
10 - Chips are too big to not go at them from higher levels of abstraction.
9 - Teams are too small not to take advantage of the efficiencies offered by going after big designs from higher levels of abstraction.
8 - There's too much software in the mix in sophisticated systems, big or small, and ESL tools, technologies, and methodologies are a must if you're going to do hardware/software co-design efficiently and in a thoroughly modern/parallel way.
7 - FPGAs are overtaking the world. You don't agree? How about, boards are taking over the world. Still, don't agree? How about, chip & package co-design will be a must going forward. Don't like that one either? How about, complex multi-core, multi-chip, multi-package products are everywhere. Oh, come on. Surely, you can't argue with that. So, choose your battle and choose your weapon. I'll bet that weapon's got something to do with ESL.
6 - The compute power is there to make ESL happen. Meanwhile, abstraction, models, and virtual platforms require a helluva lot of computer power. Like the proverbial chicken and the egg, ESL will be providing the tools to design the platforms that are needed to design the tools that are needed to design the platforms, etc., etc., etc.
5 - You want to stay competitive in this global economy? Think ESL.
4 - Define ESL for me. Can't exactly do that? Then, it's still not commoditized and so there's still lots and lots of room for innovation, and even more room for commercialization.
3 - There were two absolutes expressed at DAC in various settings and by various parties. A) It's all moving toward software differentiation on a commoditized hardware platform. B) It's all about minimizing the software and capturing product intent/differentiation in the hardware. As convincing as these diametrically opposed visions can sound, and no matter who's ultimately right, both visions need to start out at the system level.
2 - Convergence of features, markets, and technologies are best facilitated at the system level.
1 - Configurability is cool; reconfigurability is really cool. For systems to be nimble and responsive to varying needs over the product lifecycle, you're going to need to understand the implications and inferences of change. Do that within the ESL paradigm. Pass Go. Collect $100. Possibly more.
Top 10 Reasons why SystemVerilog is, and isn't, a slam-dunk
10 - SystemVerilog promotes vendor neutrality.
9 - But, companies have millions of lines of verification code written in legacy languages.
8 - And, it takes time and corporate commitment to translate that code to SystemVerilog.
You can find the full EDACafe event calendar here.
To read more news, click here.
-- Peggy Aycinena, EDACafe.com Contributing Editor.