Get email delivery of the Cadence blog featured here
On the last day of CadenceLIVE 2020, there was a keynote by Alberto Sangiovanni-Vincentelli. Just in case it wasn't obvious from his name that he's Italian, he delivered the keynote from his villa on the Mediterranean coast near Rome. He titled the keynote Cadence and Academia: From Transistors to Systems with Computational Software.
He started by likening innovation to truffles. These are tasty fungi (especially the white ones that grow mainly in Piedmont in Northern Italy) but nobody knows how to cultivate them. You need some local knowledge of where truffles have been found before (and I suspect if you weren't born in the area that is not knowledge you will get) and a trained dog who can sniff them out since they grow underground among the roots of certain trees. I think I can safely say that this the first time in over 1,000 posts on Breakfast Bytes that a truffle dog has featured. I have read that they used to use pigs, but the pigs like truffles and so will do everything they can to eat them, whereas dogs will take their reward as some other treat.
The second image shows where to look for the truffles of innovation in the US, based on some measure of patent references. Silicon Valley is the highest peak, another big one at Boston, and smaller ones scattered around. The big thing that they have in common is good engineering universities: Standford and Berkeley, where Alberto works, for Silicon Valley. These centers of excellence tend to change slowly since you need research professors, doctoral students, undergraduates (usually), and so on. If you read my post "Lick" Licklider, Unsung Hero of US Computer Science then you'd know he set up the first doctoral programs in CS. He funded four at Stanford, Berkeley, CMU, and MIT...in 1965. I would say they are still the top schools for computer science today over 50 years later. It's not just computer science: Yale Law School was founded in 1824 and is still so significant that most supreme court justices went there. Like I said, these things change slowly.
By the way, talking of innovation, I highly recommend Matt Ridley's wonderful book on the same topic: How Innovation Happens.
Alberto had a taxonomy of research. Whether it was a fundamental quest for understanding (yes/no), and whether it was concerned about whether the results might be used. As you can see, his archetypes are Neils Bohr, Louis Pasteur, and Thomas Edison. You can put your favorite useless subject or enemy in the bottom left box where there is no quest for understanding and no concern about using the results. Alberto likes the Pasteur corner in the top right.
Alberto went on to give a detailed history of how EDA research got started at Berkeley, was then picked up by Intel, and led to the formation of SDA, the forerunner of Cadence. But I covered that story when I interviewed Alberto last year in my post Alberto and the Origins of the EDA Industry. Also, for more specifically on the founding of SDA and the reverse-merger with ECAD to form Cadence, see my post A Brief History of Cadence: The Solomon-Costello Era. I won't repeat any of that here.
Next Alberto turned to the future and cyber-physical systems (CPS from now on). These are systems that are not pure electronics but involve mechanical aspects that need to be designed and analyzed together. I guess modern cars with lots of ADAS would be a perfect example. You can take it further to Bio-Cyber Systems, where there are brain interfaces and automated prostheses. Alberto thinks that computers and phones will disappear by 2025 (which is only five years out) and there will be 7 trillion devices for 7 billion people (1,000 each). That seems way too aggressive. It took many years to get to a billion smartphones, at least partly because they all needed to be manufactured. If my math is right, 7 trillion devices in 5 years is nearly 4 billion a day...from a standing start.
As you might expect, the modeling challenges for cyber-physical systems are analogous to what we've done in electronics. Made it possible to model things, and assemble them correctly. In this case, not just electronics and computer science, by physical objects such as sensors and actuators. This goes under the more general title of multiphysics.
Alberto outlined Cadence's multiphysics strategy, with various powerful solvers than can be used to attack different domains. You can see some of this in the above diagram but I'm not going to go through it in detail here. Maybe it will make a good Breakfast Bytes post one day all on its own.
Alberto wrapped up talking about artificial intelligence and security. I think these are two topics that nobody is allowed to leave out of a Cadence keynote—see my post from last Friday on what Lip-Bu, Anirudh, and our guest keynotes talked about!
But I will wrap up with Alberto's conclusion when he went back to his theme:
Sign up for Sunday Brunch, the weekly Breakfast Bytes email