OpenAccess is one of the most successful and impactful standards in the history of the EDA industry. By providing a C++ API, data model, and reference database implementation, OpenAccess has brought unprecedented levels of integration to analog and digital IC implementation. This year OpenAccess is celebrating its 10th anniversary starting with a co-located EDA standards event at the Design Automation Conference (DAC) June 4, 2012.
While the OpenAccess database software was created by, and is maintained by, Cadence, OpenAccess is managed by the Silicon Integration Initiative (Si2) OpenAccess Coalition, which currently has 43 member companies. In this interview Steve Schulz (right), president and CEO of Si2, talks about the history of OpenAccess, its impact on IC design, the role of the OpenAccess Coalition, what's coming up with the standard, and the upcoming DAC event. (For a first-person perspective on OpenAccess, see Steve Schulz' March 20 blog post at Chip Design Magazine).
Q: Steve, how and why did OpenAccess get started and who originated it?
A: It started in the mid-90's with the Sematech Design FTAB [Focus Technical Advisory Board], which had the vision of a common data model and interoperable C++ API that would enable fully integrated design systems. The FTAB included AT&T, AMD, IBM, Hewlett-Packard, Intel, Texas Instruments, Digital Equipment Corp., Motorola, Rockwell, and National Semiconductor. Several of these companies already had their own object-oriented database with an API for their internal tools. The companies wanted to transition that architecture out to the competitive EDA marketplace, so they could shift more to buying commercial tools.
The Sematech Design FTAB ran its course in about 5 years. Members acknowledged that the approach they had taken - with thousands of pages of specifications - just isn't the way EDA works. Sematech decided to end the FTAB and took the effort over to Si2, where it became a group called the Design API Council [DAPIC]. They put out an RFT [request for technology] to the EDA vendors and only Cadence came back with an affirmative answer. Cadence offered its Genesis database, and although it was written in C, this became a starting point.
The OpenAccess Coalition was announced at DAC in 2001, but how it was going to work had not yet been solidified. That happened in January 2002.
Q: Since a variety of events led up to OpenAccess, on what do you base the tenth anniversary?
A: I think what really matters for OpenAccess is a very practical thing - the point at which you could actually use it. That was around December 2002. Cadence shared Genesis 1.0 in the middle of 2002, but that was still C code. They got to work quickly doing a C++ re-implementation, and in December 2002 they released Genesis 2.0 to Si2, which Si2 distributed as OpenAccess 2.0.
Q: How was OpenAccess different from previous attempts at EDA tool integration?
A: The vision of the DAPIC members was that there would be a common run-time data model. Previously the industry had put a lot of emphasis on file format exchanges, but these are not as scalable or amenable to incremental data sharing between tools.
Q: Fast forward 10 years. How has OpenAccess impacted EDA and semiconductor design?
A: It has had a fundamental positive impact on the entire EDA industry. In the custom/analog and processor design world, clearly it is the default environment that everyone from the high-level designer to the foundry assumes is there for them. Of course Cadence played a large role in making that happen. I would say we achieved the level of adoption and success we had anticipated in the analog/custom world.
The digital world is a little more complex. There is a large amount of digital use of OpenAccess, but largely by companies that have a mix of analog and digital. Not all information used on the digital side is standardized in OpenAccess yet. We look forward to more work and adoption on the digital side in the next 5 years.
Q: What's the role and structure of the OpenAccess Coalition today?
A: At its origin, the coalition was really there to put in place a roadmap of requirements and features to be added, and to provide a formal, fair, structured process for determining which features and changes were needed in what priority order. Then it was to guide the development of these changes in the reference implementation and documentation.
The role today is largely the same, but we have recently expanded its scope. Today we have an Extension Steering Group, ESG, which works in parallel to the Change Team. The core of OpenAccess and the reference implementation are still managed by the Change Team, but the ESG focuses on peripheral things that add value to the standard. One example is scripting languages. They're not part of the C++ software but they're part of the OpenAccess ecosystem.
Q: Data model, API, reference implementation - what's available to whom and how?
A: We haven't changed the OpenAccess license since 2004. It allows a company to download all three [data model, API, reference database]. We have made both compiled and source versions of the reference implementation available to members and non-members alike. Going forward, source code for the newest OpenAccess stream will be available only to OpenAccess Coalition members, but non-members can still get the compiled reference implementation [for the latest stream].
OpenAccess has never been open source, but we call it "open community" in order to emphasize its non-biased, non-discriminatory nature where anyone can be part of the community.
Q: Who implements changes to the reference database?
A: For all these years the industry has counted on the goodwill of Cadence to implement changes to the reference implementation. Going forward, there may be additional contributions from other companies, and the ESG provides a way to support that. But the core of OpenAccess remains something that Cadence continues to maintain at no charge to the OpenAccess Coalition, guided by changes approved by the Change Team.
Q: What's next for OpenAccess?
A: With the continued advancements in performance and capacity, and with new process nodes including 20nm and 14nm, there will always be a need for enhancements. For example, multi-threading has been released in its first phase and this support will expand. Other future items include scratch designs, multiple track patterns per layer, partial loads, enhanced performance for extension objects, and an improved "occurrence model" that covers an entire chip with fully flattened data.
An exciting area to watch in the future is photonics, which is designing with particles of light with 6 orders of magnitude smaller feature sizes than electronics. This is something that OpenAccess can uniquely enable.
Q: Finally, what are you planning for the Si2 DAC OpenAccess event?
A: We want to have fun and we want the industry to have fun with us. We'll have some reflections from the visionaries who made OpenAccess possible and also some testimonials from leaders across the supply chain about the value OpenAccess has offered all these years. Then we're going to offer special recognition awards to some of the technical leaders. We'll also have giveaways and prizes, and a special reception that includes a free lunch for all people who sign up.
(Note: To learn more about the Si2 DAC event, which includes an Si2 standards update as well as the OpenAccess anniversary celebration, click on http://www.si2.org/?page=1525).