Doghouse – RISC-V Summit

About a year ago, I wrote an article on the RISC-V architecture and how it seemed to be interesting. I also pointed out that it seemed to be moving along very fast, but there were naysayers who were questioning whether it would get off the ground. I advised the naysayers to wait a year before offering their opinions.

On December 9-12, 2019, I attended the second RISC-V Summit in San Jose, California. There were a little under 3,000 attendees, a few of whom I had worked with over the years at Digital Equipment Corporation, on the Linux kernel project, or other similar projects and companies.

Quite frankly, I was blown away.

While there is much that still has to be done, there was a lot finished including some basic chips produced that were running Linux. There was much additional work being done on software emulators and FPGAs, which created excitement for me. You had chip designers designing extensions and passing those off to operating system and compiler engineers, who would then try out the design with the compilers and operating systems to see if the extensions were useful. This was collaboration that I had never experienced on this level, at this speed, and with this much openness before.

Everyone was talking about being "open," and stressing the sharing of technical information back and forth. It was exciting … electrifying.

For those of you who are not familiar with RISC-V, it is a movement to have an "open" Reduced Instruction Set Computer (RISC) that was started at the University of California, Berkeley, the same place that the Berkeley Software Distribution (BSD) of Unix started. Trying to learn from the previous decades of instruction set design, they wanted a clean design that would be efficient, compact, extendable, and free of licensing fees to allow both research and development of new processors. Many companies, universities, and even individuals have joined this organization.

The first pleasant thing that happened was their CEO, Calista Redmond, gave the opening address. Not only was her address great, I was also happy to meet Redmond several times, and she was both welcoming and fun to talk with. I think she was an excellent choice for a CEO.

Krste Asanovic, one of the professors at UC Berkeley who started the RISC-V project, gave the "State of the Union" address after Redmond, and was followed by an old friend of mine, Martin Fink of Western Digital.

I met Martin years ago when he was working for Hewlett Packard. Martin, like myself, saw in GNU/Linux more than a "hobbyist toy," more than a "geek thing," but a fundamental building block of software development that would change the computing industry forever.

Martin wrote the first book on Open Source business plans, The Business and Economics of Linux and Open Source, which was originally published in 2002 and is still considered a classic in the field.

Today Martin is the advisor to the CEO of Western Digital, a very large data storage company, and is focused on "data driven computing." According to Martin, we put too much focus on "processing" and CPUs and not enough focus on the data. Because of this, Western Digital has created a new way of giving access to many different types of computational power that they call OmniXtend, a cache-coherent fabric [1]. Martin almost shook with glee as he unveiled a system based on this technology.

With that as a start, the conference only got better, as more and more people stood up to talk about how RISC-V technologies and collaboration were going to help solve some of the biggest problems of computing today:

• Secure systems
• Verifiable code
• Efficient power systems to help with energy and cooling savings

I was also happy to see some old friends, such as Keith Packard  [2], who I first met when he was working on the X Window System, and, who later worked near me as we wove our way through various companies and GNU/Linux. Keith has been a long-time contributor to Debian, and, he now works for SiFive [3] on their contributions to the RISC-V community.

RISC-V is an Instruction Set Architecture (ISA) that allows a very small "core" set of instructions, purposely kept simple and small so operating systems like GNU/Linux will always be able to support it, but allowing extensions so special-purpose distributions can take advantage of those extensions if they so desire.

I look forward to the next RISC-V Summit, probably next December. In the mean time, if you would like to get started learning about RISC-V, the slides and videos from the RISC-V 2019 Summit are available online [4].

Carpe Diem!