Who says standards are dull? The 15th Silicon Integration Initiative (Si2) OpenAccess Conference opened Oct. 20 with a rousing call to action by two speakers - a call to think beyond the traditional boundaries of EDA, and to include systems and software when it comes to standards. At the same time, the speakers emphasized the importance of supporting silicon at advanced process nodes.
The first speaker was Steve Schulz, Si2 president and CEO (right), who talked about the standards that will be necessary to support the semiconductor industry over the next five years. The second speaker was Charlie Huang, Cadence chief strategy officer (below, left), who presented a keynote on "EDA360 and the Electronics Ecosystem."
Setting the tone for the session, Schulz noted that IC technology is pushing standards efforts in two opposing directions. One direction is silicon complexity and the "problems of the small," where semiconductor physics and accuracy are vital. The other direction is system complexity and the "problems of the large," an area he described as less understood and more abstract, yet crucial to meet the potential of raw silicon.
Not Your Usual Standards Efforts
In my view as a long-time observer, most EDA standards efforts have focused on digital RTL-to-GDSII design, which is a small part of what's needed to get a product out the door. Schulz talked about six "five year challenges" for standardization that clearly step outside of this box. I see a close correspondence with the concepts of Silicon Realization, SoC Realization, and System Realization described in the EDA360 vision paper.
On the systems side, Schulz talked about a value shift to software integration. "What's really happening is that the quality of how the software and the silicon work together to solve a problem, or deliver an application, is where the market is," he said. "We need to be looking more broadly if we're going to have a healthy industry ahead. How do we expand our view of design so that embedded [software] code is included in the co-development process? We need to get serious about this."
How can standards help? Existing standards like OpenAccess and the Common Power Format (CPF) may need extensions, and new standards may be needed to bridge hardware/software development, Schulz said. "We need to figure out how to communicate data in a way that's interoperable."
Power was also on the list of five-year standards challenges, and here the focus was at the system level as well. "Power management started at the physical and RTL levels, but it's moving much more up to the architectural and behavioral level," Schulz said. Standardization possibilities include power modeling at multiple abstraction levels, a "bridge" with software development, an open power API, and advances in CPF.
SoC and Silicon Challenges
A third challenge, IP block modeling and integration, corresponds with what EDA360 would call SoC Realization. Schulz said that IP blocks need a more consistent use of standards at different levels of abstraction, and noted that OpenAccess may someday replace LEF/DEF as a delivery vehicle.
A fourth challenge, 3D ICs with through-silicon vias (TSVs), brings in both silicon complexity and systems complexity. "You have to optimize interacting physics. Your goal is the best fit of cost, performance, power, and yield," Schulz said. Standardization areas include a dictionary of terms, 3D process design kit (PDK) support, and data exchange between timing, power, thermal, and physical domains.
A silicon-based challenge is "variation-aware everything." Standardization possibilities include extensions to the new Si2 OpenDFM standard, a DFM library, power modeling, and additional OpenAccess data models. A sixth challenge is analog/RF automation. Standardization areas include the ongoing Si2 OpenPDK effort, possible extensions to CPF, and possible extensions OpenDFM standard.
EDA360 and Standards
Charlie Huang's talk focused on EDA360, which he said is based on a "new paradigm in which systems companies differentiate on applications." He began by noting that "in the old days we talked about what was in the box. Now when we pull out our iPhones we don't know what processor runs underneath it. The system and the user interface has become the focal point."
However, he also noted that "silicon is still the substrate. Without the steady progression of process nodes, we'd still be running Windows on 640K."
Huang noted that most EDA standards are silicon-based. There is a "pressing need" to tie software and hardware together through a standardized environment, he said. How do you model a processor, he asked? How do you do a multi-core cache coherency check? How do you debug hardware and software? These are all questions that need to be explored.
These two talks fit very well together. So I thought it was fitting that the session concluded as Schulz handed Huang an award - a Distinguished Service Award that Si2 gave to Cadence for its work in "developing, donating, maintaining, and improving" the OpenAccess database. Mark Mason of Texas Instruments also won a Distinguished Service Award for his work with OpenDFM. Congratulations, Mark -- and congratulations to the hard-working folks on the Cadence OpenAccess team.
Conference presentations are available for downloading here.