Accellera Updates at DVCon 2026

Lu Dai (chair of Accellera) and I had our regular chat at DVCon U.S. 2026. Accellera also hosted a reception in the exhibits hall, with free snacks and drinks, very well attended. We talked about what’s new in Accellera, with a particular emphasis on the recently released standard for CDC and RDC tool interoperability, also Lu’s thoughts on trust.

Accellera updates

The members list is largely unchanged: semiconductor and EDA of course but also multiple systems companies including Cariad (VW), Microsoft, Google and Apple. One new member is ChipAgents.

Conferences are expanding, geographically and in size. India, Taiwan and Japan conferences are leading in growth. The China conference is not growing as fast, though Lu sees opportunity to accelerate growth in China in a couple of ways. First the current event is hosted only in Shanghai and primarily attracts multinationals. This may limit attendance since their involvement in Accellera is already represented in other geographies. Growing into other areas outside Shanghai should attract more local interest.

Second, Lu suggested that an academic track would have significant appeal in the Chinese technical community. They want, as much as we do, an opportunity to be recognized for their research in a credible international forum. Adding this feature would extend the success DVCon Europe has already seen in this respect.

The SystemC group has scheduled a first SystemC code sprint in April, as a shift from traditional bi-weekly meetings to an approach more common in the open-source community. A bunch of developers will write code (libraries and reference examples) together and in-person. They are also continuing an online event in April called a Fika (a Swedish concept). These are short meetings to chat and catch up. Shout out to DVCon Europe for leading the way in multiple areas!

Accellera hosted a Summer of Code project directed at interns in 2025. This may be rebranded to target other standards such as CDC and UVM. Ambitious interns should note this year’s program. Lu is optimistic that with committed mentors this project could be a very effective method to grow the visibility and adoption of Accellera standards.

CDC and RDC standard – users need to step up!

The 1.0 version of this standard was released in March 2025. Lu says this is pretty complete for CDC, though he expects some iteration for RDC given respective levels of interest in those domains. Just as a reminder, IP and subsystem suppliers use their approved vendor checkers with those vendor’s constraints format to validate CDC and RDC with an output in that vendor’s format. Today if you as an integrator have a different approved vendor for CDC/RDC, you must translate those constraints and outputs to your vendor format. Which is doable but requires maintenance and, as always, is a potential source of errors.  The only other option is to run flat, an expensive approach and increasingly not feasible due to design complexity.

Now the community needs to push the compatibility burden back on the tool vendors. The standard is ready, vendors are aware of the standard, but they are not as motivated to prioritize support in their tools if customers aren’t demanding support. Time to express your expectations more vigorously!

Risk

I’m always interested in Lu’s viewpoints independent of Accellera; he also chairs RISC-V International and is VP of Technical Standards at Qualcomm. He has a broader view of risk than I have been thinking about recently (for AI in verification). I’ll just highlight a couple of points here.

One point he raised in the context of standards is possible exposure to IPR (intellectual property rights) conflicts in AI based contributions to the standard. Did the contributor use a bot to create or format their contribution? If so, does that create a legal hazard for the standards body or more widely? Lu said they have examined the legal question for Accellera and are comfortable that they are OK. However, he suggests that contributors (and innovators in general) should not use AI to document their final contributions because they don’t know how created material might innocently tread on other IPR.

Lu also mentioned a trust-related risk, illustrated by a hack introduced into the open-source Linux kernel in early 2024 and only caught by chance in unrelated beta testing. The hacker was a maintainer who had done good work for two years and therefore was trusted by the open-source community. Once trust was established, this sleeper agent introduced the hack. Here the problem was human generated but could equally have been AI generated.

Good chat. You can learn more about the CDC/RDC release HERE.