How can European projects develop microprocessor core IP without unnecessary overlap? And how can we avoid falling behind the rest of the world in open-source hardware? These are just two of the questions posed in the new report ‘Recommendations on Open-Source HW & SW, incl. RISC-V, for European Sovereignty’, which sets out a roadmap to deliver foundational RISC-V IP. Working group members Patrick Pype (NXP), Marc Duranton (CEA) and Luca Benini (ETH Zurich/University of Bologna) share their key takeaways.
Once bitten, twice shy
As the title suggests, sovereignty is at the heart of the report. In some ways, this is about learning from past mistakes: Europe may have been a pioneer in open-source software, but a lot of the value from this has ultimately been exploited elsewhere. “In open hardware,” begins Luca, “I see the risk of this happening again. There’s huge momentum around open hardware in academia but we’re being hesitant in Europe. The main goal for me in drafting this document was to gather consensus around this initiative, which is the first step to raising awareness in industry. European industry is less proactive in this field than in other parts of the world, so we need to make a case.”
A major risk is fragmentation in the form of many small projects which inadvertently compete, thereby diluting European efforts. While open hardware allows a great deal of freedom, organisations must avoid needlessly reinventing the wheel. “For me,” Patrick says, “the challenge is to get a European programme in place with a clear strategy on RISC-V and open-source development and how to apply this in system-on-chip design in different application areas. A lot of players are doing good work, but a unified European approach is still missing.”
“I totally agree with Patrick and Luca,” adds Marc. “In Europe, we tend to have good initiatives but are all going in different directions as though global momentum is not very important. If this initiative can streamline all players – industrial, academic and so on – into one common direction, we could really make the difference.”
Making the case
Of course, learning from past mistakes also means recognising that past successes cannot simply be repeated. Unlike open-source software, open-source hardware requires largely proprietary tools for translation, mapping and verification, as well as access to libraries and foundries for transforming IP into real silicon. “RISC-V is like an engine but you need everything else to build a working car,” notes Marc. “Big companies know how to make value from open source but a lot of small or medium players are scared. How can I really create business with open-source hardware? What are the problems of licensing? Start-ups and academia need to have more access to tools and IP and helping them to understand the business aspect of open source could also help them to engage.”
In other words, the challenges are both technological and commercial. From an industrial perspective, the business case is being made in what the working group calls ‘the Airbus of RISC-V’, a proposal in which companies that view RISC-V and open-source environments as strategically important can focus on the productisation of both hardware and software results. “In the world of RISC microprocessors, we see RISC-V and open-source hardware as providing a compelling opportunity that allows for both innovation and differentiation in our products,” explains Patrick. “Of course, we are interested in how to industrialise open source because there is a clear need for professional support and maintenance. It is not merely as simple as duplicating existing IP licensing business models or trying to clone the open-source software movement; we have to find other ways and this proposal is on the table from our side.”
On a technical level, another difficulty for open-source hardware is the lack of clear interfaces between the open and closed aspects of the system. This is in sharp contrast to the clear hand-off point in open-source software, the hardware’s instruction set architecture. “As an academic, I believe that open hardware is extremely valuable, but I also think that business model creation will require some form of agility in moving in and out of open hardware,” says Luca. “The idealistic view is that everything will be open. In practice, there will be a lot of mixed business models with open and closed hardware or software. This is what we need to work on in this programme.”
Focus on a realistic future
With all this in mind, the report makes a number of concise recommendations for both the short and long term, ranging from input for the upcoming KDT calls to incentives by the European Commission to make open hardware the default paradigm for all hardware development in publicly-financed institutions. This calls for pragmatism regarding the current computing spectrum, parts of which are closer to the market than others. Luca: “The programme should make advancements with respect to the status quo in all areas. It’s not realistic to think that we’ll have high-performance processors based on open-source hardware on the market in two years, but it’s much more realistic to say that we’ll have many IoT processors because we already have some today. In my view, it’s important not to be over-optimistic.”
“This is a report that summarises what the community has been thinking for a long time but did not express in one document,” continues Patrick. “Public authorities now need to recognise the importance of these types of activities in order to make sure that they are coordinated and embedded in both national and European funding programmes. We should also look at domain-specific requirements like accelerators, safety, security, automotive quality and scalability, as well as the perspective of end-user markets such as automotive, telecommunications and IoT in order to make sure that whatever is developed meets the requirements at the end. This is my main message: we should cover the entire value chain and all aspects of open source, not just one particular item.”
“It’s the same for me,” agrees Marc. “We should have a holistic view not only of RISC-V but of the complete ecosystem. Having a closed link with software is very important. I will also stress what Luca said earlier: we need to be fast or else the US and China will be in front of us. We will be the initiator but will not benefit.”
Patrick smiles. “We use the word ‘fast’ in our report actually. It means Firmly Act Strongly Today.”