SEVILLE, Spain – While experts often extrapolate current growth figures and developments to paint futures abundant in technology Professor Alberto Sangiovanni-Vincentelli, of the University of California, Berkeley electronic engineering and computer science department, has cautioned that we must also look at the design challenges that will enable or prevent us getting there.
In a talk entitled "1000 electronic devices per person, dream or nightmare?" Sangiovanni-Vincentelli, known as ASV, looked forward to a future where the internet-of-things is wireless and ubiquitous but the possibilities of design and unintended interactions are much more complex than today.
ASV was presenting to the International Electronics Forum, organized by market analysis company Future Horizons. "Swarm computing is where the cyber world meets the physical world," ASV said arguing that future machine-to-machine computing will combine both centralized, cloud computing, and distributed computing in a way rarely demonstrated before. And deciding where computation should be done should be done on the basis of energy consumption.
In 2030 ICs will be approaching molecular limits this progress will be accompanied by the birth of societal IT systems including fail-safe objects such as cars do not crash and immersive home entertainment and living systems. Going beyond tire pressure monitoring systems (TPMS) automobile companies are already making strides towards making tires into intelligent adaptive sensors to control and prevent collisions.
"We will have cars that drive themselves in the market within five years. It is going to happen," said ASV albeit without touching on the human aspect of resistance to such changes which must include the perception of safety, responsibility for safety and contractual obligations such as insurance.
But ASV did highlight that for this to happen the discipline of design must step up. Complex objects such as automobiles at the latest generations of airplane have millions of lines of software. For cars recalls are frequent and usual for a software fix. "We are on the brink of disaster because of unconsidered dependencies," he said.
A traditional, highly detailed bottom-up approach is destined to fail, he said. "The functionality is the collective behavior. It is time to abandon the component-oriented vision. The swarm of bees or the school of fish; that is the approach we have to take."
The challenges therefore include learning to cope with heterogeneity and how to use resources on a dynamically changing basis.
Nonetheless, there are lessons to be learned from the VSLI era. "The classical approach to design was decomposition and abstraction. We have to have freedom FROM choice. Restrain choice and the design can become manageable. There must be a new meaning to the phrase system design with power consumption and latency as the driving metrics."
And who will make this design breakthrough? ASV said might not be one of the current EDA leaders and would more likely come from a software leader such as IBM. "The challenge to the use of the swarm is to have an OS that can make efficient use of distributed resources.
When asked if this challenge was essentially the same as that which has faced developers of multicore processors for many years, ASV acknowledged it was.
David Patterson, known for his pioneering research that led to RAID, clusters and more, is part of a team at UC Berkeley that recently made its RISC-V processor architecture an open source hardware offering. We talk with Patterson and one of his colleagues behind the effort about the opportunities they see, what new kinds of designs they hope to enable and what it means for today’s commercial processor giants such as Intel, ARM and Imagination Technologies.