Moore Open Source Coming

The sunsetting of Moore’s Law is creating some interesting ripples throughout the EDA and IP industries. No longer is the low-risk path defined by a migration to the next node. Most companies cannot afford it and don’t need it. Neither can their competitors. Suddenly, they have to do more with less, or at least the same amount.

Consider just a few things that are changing today:

• Sticking to an older node greatly reduces cost and risk, especially for startups. Instead of being a rush to get the next generation of product out as soon as the new node becomes quasi-stable, the rush is one of innovation. How can I rethink the problem and its solution? Can I come up with something that outperforms existing approaches? Many times, this approach yields results significantly in excess of the advantages provided by a new node.
• Many companies purchased IP because they could not afford, or did not want, the expense associated with migrating that IP to the next node. While that may not have been a huge concern for a CPU, it was for a SerDes or for analog components. Now a company can reasonably expect that the lifetime of their analog block may well extend for several generations of product.
• The change also affects digital blocks such as CPUs. If you cannot expect the processor to become more capable with a new node, then why continue to pay the verification tax every time you lay that core down when nothing much changes? If a hundred other designs contain the same core, then you are collectively paying for verification once and then everyone is reaping the benefits.
• EDA tools no longer have to handle new physical effects brought on by each new node and thus the capabilities provided by an old tool will continue to be perfectly adequate for the next design and the one after that. If you are not buying EDA tools for the new features, how do you assess value?

These are just a few of the questions that a growing number of people in the industry are beginning to ask. Recently, I attended Latch-up, a small open source conference held in Portland and organized by FOSSi. Held on a weekend, it was decently attended by academics, designers, makers, EDA tool developers, IP developers and curious onlookers. Most of the time, when I did a rough count, there were 70 or so people in the room. They all had one thing on their mind – how can we do more when we turn our attention towards architectural innovation rather than running the Moore’s law rat race?

We all have seen the success that RISC-V is having in the industry. Momentum continues to build, and many are now asking for open source EDA tools to go along with it. This is where the community perhaps has a little to learn from the established EDA companies.

There are quite a few open source EDA tools available today, such as simulators and generators of various kinds. However, they are not well unified. This problem is not unique to the open source community – almost every EDA company had to deal with these issues whenever they purchased a new company. Many times, the reason why they did it was because the tools became more valuable when they were tightly integrated. They have also learned that attempting something completely new becomes a tough sell – even if it provides a significant benefit. That is a lesson that the open source community needs to learn and, until they do, will result in many failed attempts.

Continuity, integration and flows are important. Coming up with a new language or methodology for a point tool produces a very small overall benefit. This makes it difficult to rely totally on the academic community to do this – they need to work out of the box at times, they need to find pint sized problems that one student or small group of students can work on and produce something that justifies their PhDs. It is not their role to provide tools for commercial companies, even though there have been quite a few successes in that area.

An interesting dynamic is that many of the conference attendees appear to be quite happy with tools that are less capable than the ones supplied by EDA companies. These are the individuals who believe that innovation comes from the architecture and design, not from the implementation tools and flows. They are ready and willing to give up a little to save a lot. And those savings can be measured in different ways – if a designer does not want to get deeply involved in optimizing the design to the Nth degree, then they probably want more push button operation.

We are seeing many changes in the industry, and it is quite likely that open source will become a reality in the near future. We are already seeing companies that are willing to commercialize open source IP and companies that want to buy from them. It remains to be seen if the business model is sustainable.

Will an existing EDA company embrace open source and make some of their technology openly available and switch to a maintenance model? Will a new EDA company emerge that commercializes existing open source tools? Will larger companies be willing to acquire and support open source EDA tools?

The answer to all of those questions is yes.