Cadence Executives on Computational Software

CadenceLIVE starts today, Tuesday, August 11, and runs through Thursday. One thing that I know some of the keynotes will cover will be computational software, a name for many of the sorts of algorithms that underlie EDA  and system design tools. It is not the algorithms that are unique to EDA but scale on which they have to operate.

Chin-Chi Teng

The domain that operates at the biggest scale of all is the digital full flow. Chin-Chi Teng wrote about this recently in a LinkedIn blog post titled Computational Software: Foundation for an Optimal Digital Design and Signoff Full Flow. Here's the opening paragraphs:

Cadence has started to use the term, "computational software", to collectively refer to a superset of algorithms and technologies of those powering electronic design automation (EDA) tools. Providing the most powerful computational software for EDA is probably most applicable in the business group I’m responsible for: Digital Design and Signoff. We have to enable our customers to handle the biggest designs with the most objects whether we are measuring in transistors, gates or placeable objects, often navigating through the longest runtimes. If you work on these leading-edge designs, you already know that when we say that “physical design might take a week”, this is not a figure of speech. Not only that, but it may also require use of the most powerful servers in existence, too.

In the last couple of years as Cadence has gotten more heavily involved with the commercial cloud companies, I’ve discovered that they are surprised when they discover that we mean it when we say, “run for a week”. Nothing other than EDA in general, and especially digital physical design and signoff, comes close. For example, the ImageNet Large Scale Visual Recognition Challenge (ILSVRC) contains 1,281,167 images for training, 50,000 for validation, and 100,000 for testing. By EDA standards, these are small numbers.

Anirudh Devgan

As it happens, Cadence's president, Anirudh Devgan also wrote a few weeks ago about the topic in a bylined article at EE Times, Computational Software Meets Generational Industry Drivers. Here are a couple of paragraphs from the middle of the piece:

The five generational drivers are well-known: 5G, artificial intelligence/machine learning (AI/ML), hyperscale computing, the industrial internet of things (IIoT), and autonomous vehicles. The common denominator among them, at a macro level, is perhaps less well known. Computational software is the enabling technology upon which all these drivers rely, and it is innovation in computational software that is helping shape the future.
...
Computational software innovation is occurring in three key areas. First, we’re seeing fully integrated EDA tool flows, including the introduction of ML capabilities with the goal of producing the best possible silicon. Second, the growing prevalence of distributed computing and multi-core computation on a broader scale is resulting in gains in both scalability and throughput while utilizing less expensive servers. Finally, system complexities make it necessary to perform multi-physics analysis of silicon, packages, PCBs, and connectors. Co-design and co-optimization have become essential to meet escalating system performance requirements.

Over on Gloria Nichols' website, Anirudh has another computational software article 4 Ways Computational Software Is Transforming System Design & Hardware Design. Here's Anirudh's definition of computational software:

Computational software refers to a software classification comprised of complex algorithms and sophisticated numerical analysis for heuristics and pattern recognition processes. Today’s computational software applications span numerous industries, including semiconductors, systems, weather prediction, scientific software, and financial, medical, and business analytics. It is also used heavily in today’s AI and machine learning algorithms. Below are a few examples of computational software:

In a LinkedIn reply, Anirudh sums up the story:

Gloria, thanks for posting the article. EDA has mastered the underlying algorithms, heuristics, and pattern recognition aspects for decades now—including computations that enable chip design with billions of elements. It's exciting to see how we have been able to successfully apply this computational software expertise to adjacent areas like system design and machine learning applications. And this is only the beginning.

CadenceLIVE Americas

Anirudh is giving his keynote on Wednesday. Come along and hear about all the latest development in computational software. Then, on Thursday, come and listen to Alberto Sangiovanni-Vincentelli's presentation on how we got from EDA algorithms developed in academia, to the EDA industry of today, and the computational software required for the far more complex systems of the future, involving not just electronics but mechanical and biological interfaces.

If you are not already registered, then here's the event page.

Sign up for Sunday Brunch, the weekly Breakfast Bytes email