The software facet of supercomputing, and the promise of RISC-V and open hardware

Editor's note: last summer, I was asked to contribute to a book called 'Faster than the future', commissioned by the Digital Future Society, a programme supported by the Government of Spain in collaboration with Mobile World Capital Barcelona that seeks to build an inclusive, equitable and sustainable future in the digital era.

I wrote a chapter on 'supercomputing' or high-performance computing, and now that the book is out - download it for free here - I'm sharing it on tech.eu in three parts, slightly updated with the latest news/information since the initial write-up. Below, you can find the second part of the series. Check out part 1 as well.

In the first part, I covered the basics of what supercomputing is, why the race for global dominance in the high-performance computing field is extremely important (with Japan's Fugaku currently in the lead, but probably not for much longer), and how microprocessors initially designed primarily for smartphones have grown up to power some of the top systems on the planet today. Now, let's move on to the software side of things.

The software part of the high-performance computing equation

If the choice of a processor designed initially for mobile phones was a surprising decision, at least for people without Barcelona Supercomputing Center director Matteo Valero’s knowledge and experience on the subject, what was definitely not surprising about the creation of Fugaku is on the software side of things.

Its operating system, at least partially, banks on the Linux kernel - but then so do literally all 500 of the world’s speediest computers. That said, there are interesting connections to be made between the history and ongoing rise of Linux - and open source software programming in general - and the future of supercomputing, at least according to some.

One of the prime examples of open-source software development, the Linux kernel was initially released by Finnish software engineer Linus Torvalds back in 1991 (coincidentally a few years before the first edition of the TOP500 supercomputing speed ranking was even published) and was originally developed primarily for use in personal computers.

Today, Linux is by far the leading operating system on servers, mainframe computers and supercomputers, and also boasts an enormous install base on smartphones, courtesy to the dominance of Linux-based mobile OS Android.

The results are in, and they say open source has won, basically. Since its release now almost 30 years ago, Linux has undoubtedly and comprehensively altered the entire software industry beyond recognition.

Now, a relatively young organisation wants to do the same for processors and other hardware, which is highly relevant to the future of supercomputing, as well as to Europe’s chance to compete and perhaps even play a leading role in the further development of high-performance computing systems and infrastructures.

RISC-V rising

To ‘supercompute’, one of the things you need is a harmonious combination of highly specialised hardware, processor and software programs. As mentioned, the latter are quite easy to find in a free and open form in large part thanks to the maturation of the Linux kernel, Arduino and the open software communities that support them, but the same does not go for the more tangible elements in the stack equation.

Take chips, for instance.

Even though Fugaku has licensed microprocessor designs from Arm, most makers of high-performance computing devices still turn to Intel and AMD for their chips. There are simply not a lot of alternatives out there when it comes to processors, which comes with drawbacks for supercomputer makers: the relatively steep licensing costs for chips, but also the intellectual property that big semiconductor companies hold and charge for.

This is the current situation, and it’s one that RISC-V hopes to fix.

Born out of the University of California-Berkeley more than a decade ago, RISC-V (which is pronounced 'Risk-Five') is a nascent computer architecture that's available under open, free and non-restrictive licences.

“RISC-V enables members to design microprocessors and other types of chips that are compatible and interoperable with software designed for the architecture, which means its licensees do not have to pay royalties to large, entrenched semiconductor industry incumbents,” explained Valero.

When SoftBank acquired Arm in 2016, it did so in part to capitalize on an expected boom in the ‘Internet of Things’, in which everyday devices from traffic lights to refrigerators are expected to connect to the Internet. But in IoT, the chip company has already been reported to face competition from RISC-V, leading the company to spin off two IoT software businesses back into SoftBank to focus on its core chip technology.

Notably, no patents were filed related to the open RISC-V architecture to date, as its infrastructure itself does not necessarily represent any new technology; it is based on computer architecture ideas that date back at least 40 years. RISC processor implementations - including some based on other open instruction set architecture (ISA) standards - are widely available from various vendors across the globe.

The eponymous foundation behind RISC-V, which was founded in 2015, says it counts more than 1,000 member companies today, including the well-known likes of Google, Huawei, Samsung, IBM, Red Hat, Qualcomm, Alibaba Cloud, Western Digital, Raspberry Pi and NVIDIA.

Modular in nature

One of the reasons it boasts widespread industry support from both processor and device makers is that it is ‘modular’, as in designed to be freely extensible and customisable to fit any market niche.

Another is that the RISC-V movement comes from the world of academia and is billed as ‘politically neutral’, not an unimportant aspect considering the high stakes involved in the global supercomputing race.

In fact, the RISC-V Foundation announced in November 2019 that it would relocate to Switzerland, citing uneasiness over US trade regulations. As of March 2020, the organisation is now de facto called RISC-V International, a Swiss non-profit business association, meeting at least part of its growing community’s concerns.

Although it is still early days, and RISC-V has many mountains to climb in the face of large, established and aggressive competitors, many hope that the architecture will have much of the same effect on the world of high-performance computing as Linux has had on the software industry.

Today, it is arguably the strongest contender to truly become the open hardware platform of the future. And for some, it is considered Europe’s best chance to vye for global leadership when it comes to supercomputing - or at least to stay in the game.

And in the upcoming third and final chapter of this series, we'll look at Europe's present and future ability - or lack thereof - to play a role in the high-stakes supercomputing battle.