Few people have been as influential in the development of EDA as Alberto Sangiovanni-Vincentelli, professor at the University of California at Berkeley and Cadence board member. At the International Conference on Computer-Aided Design (ICCAD) Nov. 6, he delivered a presentation that ranged from the beginnings of chip design in the 1960s to a future in which a "sensory swarm" will bring trillions of intelligent devices into our environment.
The presentation was titled "ICCAD at Thirty Years: Where We Have Been, Where We Are Going, From Transistors to Small Planets." It was sponsored by the IEEE Council on Design Automation (CEDA) as part of its Distinguished Lecture Series.
To describe the history of EDA, Sangiovanni-Vincentelli referred to three repeating stages of human history described by the Italian philosopher Giambattista Vico (1668-1744). The first is the Age of the Gods (senses), the second is the Age of the Heroes (imagination), and the third is the Age of the Men (reason). Here's how Sangiovanni-Vincentelli mapped these ages into the history of semiconductor design, followed by his vision of a future immersed in intelligent devices.
The Age of the Gods (1964-1978)
This is the age when people started using computers to do design work. Many of the foundations of modern EDA were developed during this time. For example, IBM did key work in logic simulation and test; Bell Labs came up with the idea of standard cells to regularize IC layouts; the SPICE circuit simulator came into the market; the first synthesis algorithms were developed; and maze routing and channel routing were invented.
EDA in this period was mostly contained in large companies, but the first-generation commercial CAD companies also appeared. These included Applicon, Calma, and Computer Vision. All are gone. They were "dinosaurs," Sangiovanni-Vincentelli said, who were unable to keep up with software complexity and hardware evolution. "They were selling workstations, and software was an add-on nobody wanted to pay for," he noted. "Technology was low level and there was very little customer loyalty."
The Age of the Heroes (1978-1992)
This is the age "where everything happened," Sangiovanni-Vincentelli said - a vibrant period of creativity and expansion in both R&D and industrial development. This age of "imagination" led to key results in all areas of EDA, including verification and testing, layout, physical verification, logic synthesis, hardware description languages, hardware accelerators, FPGAs, and system-level design. It also led to a few dead ends, like neural networks for CAD.
Significant research developments included:
- Relaxation-based techniques for circuit simulation
- Binary decision diagrams (BDDs), used in logic optimization and formal verification
- Formal equivalence and model checking
- The Mead-Conway approach to VLSI design
- Logic synthesis approaches including Espresso, MIS, SIS, Socrates
- High-level synthesis
- Simulated annealing for placement and routing
- Quadratic placement
During this period, Sangiovanni-Vincentelli said, "all the best people in algorithms, in computer science, were playing in EDA." As late as 2005, he noted, the three most-cited computer science papers were about EDA related topics.
On the business side, three second-generation companies - Mentor Graphics, Daisy Systems, and Valid Logic - appeared and became the "big three" of EDA. All sold workstations as well as software. Then, software-only companies began to emerge, including SDA Systems and ECAD, which merged to form Cadence. Synopsys also appeared during this period. Mentor got out of the workstation business, and Cadence, Mentor and Synopsys became the "big three" EDA vendors.
The Age of the Men (1993-2012)
The current age, Sangiovanni-Vincentelli said, has "resulted in really nice work and scientific advantages, but this work is incremental and not revolutionary." An example is model order reduction, which reduces the number of equations needed to solve a large circuit. Another new development is fast SPICE simulation. Finally, formal verification has become more practical thanks to automatic abstraction and the decreasing complexity of SAT solvers.
On the business side, staying alive and prospering is a major challenge. That's because the EDA industry has remained fairly flat and has been a "zero sum game" for many years. However, Sangiovanni-Vincentelli noted that EDA growth in 2011 was 12%, "unbelievable" given the external economic conditions. The last EDA IPO was in 2002, and the startup exit strategy now is acquisition, but "in my opinion the acquisition model cannot last forever." Startups, he noted, cannot solve problems that require substantial resources.
The Near Future: Some Assembly Required
There's been much talk about the "next level of abstraction" in EDA, and Sangiovanni-Vincentelli believes he's found it - it's the megacell-based composition of systems-on-chip (SoC) allowed by IP blocks. This is what will enable companies like Microsoft, Apple, Amazon, and Google to design chips.
"You want to assemble components from a parameterized library, including configurable processor cores, memories, and glue logic, and then integrate everything with some standard interconnect," Sangiovanni-Vincentelli said. "We have to have a way to design chips out of available components, quickly and reliably."
Further into the Future: The World is the Device
Today, there's a market battle between providers of mobile devices and tablets. Will this continue? "My answer is no - there is something more exciting coming up. It's what we call the sensory swarm. There will be lots of devices everywhere, so we will not need tablets or cell phones. They will be embedded in our environment. So when people keep investing in smartphones and tablets, they are not investing in the future."
This new world, Sangiovanni-Vincentelli said, will be instrumented (with sensors that measure and sense), interconnected (people, systems and objects can interact) and intelligent (we can respond to changes quickly and accurately). He talked of three out of many potential applications - smart water, smart traffic, and smart energy. For example, much of the water that flows through pipes is wasted because of leaks - what could we save if smart sensors could detect those leaks?
By 2025, Sangiovanni-Vincentelli said, there will be 7 trillion devices, or around 1,000 devices per person. "You are not connected to the device, you are enmeshed in the device, because the device is the world. The entire world has become a device." Going further, we could plug chips into our brains and interact directly with the surrounding environment. Going even further, we could create new living organisms from components called "bio bricks," much as we now create ICs from standard cells.
But Sangiovanni-Vincentelli brought things more down to earth for his concluding remarks. "EDA is a unique, wonderful field where research, innovation, and business come together, so it is a gift to have been associated with this field for so many years," he said. "Can EDA innovate with its present structure? We need new paradigms to go beyond what we know."
Richard Goering
Related Blog Post
Jan Rabaey Keynote: For Lower Power, Re-Think Computing