Computational Software for Cyber-Physical System Design

The recent 34th International Conference on VLSI Design, also known as VLSID, was a virtual event, of course. But it is India-based and the conference ran on India time. The theme for this year was From the Transistor to Cyber-Physical Systems, for Solving Societal Challenges.

One of the keynotes was by Paul Cunningham and Jaswinder Ahuja titled Computational Software for Cyber-Physical System Design. Jaswinder led off with an overview of the megatrends that are driving innovation in...well, pretty much everything, but electronics in particular. Then Paul made a deeper dive into the verification of these systems, not just the electronics but how the electronics fit into the physical environment.

VLSID

First, a word about VLSID which has been running since 1985, so this is the 34th. It is combined with the conference on embedded systems, which has been running since 2002, so this is the 20th. Technically the conference is thus called The 34th International Conference on VLSI Design and the 20th International Conference on Embedded Systems, so you can see why everyone just calls it VLSID. Both conferences have moved all over India during their history. I was brought up in Britain with more Indian restaurants than fish and chip shops (most of them actually run by Bangladeshis). As a result, the list of conference locations sounds like a mouthwatering tour of Indian cuisine: Chennai/Madras, Goa, Hyderabad, Mumbai/Bombay, Kolkata/Calcutta, Pune. And, of course, Delhi and Bengaluru/Bangalore.

The conference got started when:

In December 1985, a group of about eighty-five delegates assembled on the campus of the Indian Institute of Technology, Chennai to attend the First International Workshop on VLSI Design. The sole purpose of that meeting was to sense the level of VLSI activities in India with a focus on engineering education and research. It started with only 193 Proceedings along with 29 presented papers and just one tutorial. Since then, the conference has come a long way and has grown equilaterally with the VLSI community and has seen the likes of multinational Industries and academic contributors, from all around the globe. The numbers speak for themselves and have amplified to tenfold the initial attendees, proceedings, presented papers and tutorials.

I don't think "equilaterally" and "amplified" are quite the right words. But the point is that the VLSI community in India has grown a lot in the last decades, and the conference has gone along with it. Until relatively recently, most of the "VLSI community" in India consisted of subsidiaries of companies established elsewhere in the world. Not just the US, but also Japan, Korea, and Europe. Including Cadence, of course, with big offices in Noida (Delhi) and Bangalore, and smaller ones in Pune and Ahmedabad. In tech in general, the biggest Indian companies specialized in what The Economist calls "dull business-process-outsourcing and back-office management". That has changed a lot in the last decade, with a thriving startup ecosystem. Like everywhere else in the world, a lot of that is software-based, but there is a healthy semiconductor ecosystem, too. In just the last five years, Reliance Jio has changed the game in terms of wireless internet, but that's a story for another day. For more on all this, see my post The Economist on RISC-V and Indian Semiconductors.

Jaswinder Ahuja

Jaswinder started by sharing some high-level industry trends. In the last decade, the big driver was mobile (and before that the PC). But now we are in the unprecedented situation of having five big drivers simultaneously. These are 5G, hyperscale computing, artificial intelligence (AI) and deep learning, autonomous vehicles, and industrial IoT.

All these trends are based on processing large quantities of data. I'm sure you've seen those graphs showing the exponential growth to some enormous number. Jaswinder had one, too, along with the usual amazing statistics that, for example, 90% of all the data was generated just in the last two years. 80% of data is unstructured (think video, but there are other domains) and this means that we need to use AI techniques to analyze it.

This opportunity is all driving the growth of hyperscale computing, and in turn, this is driving a lot of M&A activity and redefining the industry structure.

Paul Cunningham

This data explosion and the associated processing is driving an explosion in semiconductor growth.

There is a change in how Cadence's customers are approaching these challenges. They are looking for a partner to automate their design, a partner to help differentiate their products. The challenges are not only at the chip level, but they have to worry about the performance of the overall system that the chip goes in, and the performance of the end-applications.

Stepping back from Cadence's products themselves, and look at what Cadence has built over the last thirty years, it is clear that we have built the world's best computational software team. Cadence has over $2.5bn in revenue each year, and spends 40% of that on R&D engineers who know how to write very complex computational software. We have many of the world's experts in numerical solvers, SAT solvers, graph algorithms, computational geometry, and more. This expertise in computational software is enabling us to expand into multi-physics analysis and simulation with the Clarity 3D Solver, Celsius Thermal Solver, and Clarity 3D Transient Solver, and we recently announced the planned acquisition of Numeca for computational fluid dynamics.

Paul then went over the Cadence product line. But I've covered that in so many blog posts, I won't repeat everything here. Cadence can't do everything on its own, and so another strength is our ecosystem: cloud providers, foundries, academia, system modeling, and more.

Paul then thanked everyone for coming to hear about Cadence and our Intelligent System Design vision. as a key contributor to the ecosystem to deliver a cyber-physical future for us all with real societal changes for the better.

Sign up for Sunday Brunch, the weekly Breakfast Bytes email.