Dave Patterson and John Hennessy, two San Francisco Bay Area professors now associated with Google, have won the computing industry's top prize for revolutionizing processors with a technology called RISC.
The pair won the 2017 A.M. Turing Award from the Association for Computing Machinery, a $1 million prize named after , the British researcher who famously helped in World War II and lay the foundations of computer science. The high-prestige award is considered the Nobel Prize for computer science.
RISC, short for reduced instruction set computing, gave processors a major performance boost, fueled the growth of upstarts such as Silicon Graphics and Sun Microsystems, and paved the way for today's smartphone industry. It's been a remarkable success over the last three decades, and one emblematic of the Silicon Valley ethos of throwing out the old way of doing things.
Back in the early 1980s, when Hennessy at Stanford University and Patterson at the University of California at Berkeley were coming up with the idea, it wasn't so obvious the RISC technology would dominate.
"The initial reception was 'These two academics are crazy, they don't know what they're talking about, this technology will never work in the real world,'" Hennessy recounted.
But work it did. "After all the debate, RISC did win the day from a technology standpoint," said said Linley Gwennap, an analyst with the Linley Group.
Hennessy co-founded a startup called MIPS to commercialize his ideas, with Silicon Graphics the most notable customer, while the work by Patterson's team at Berkeley emerged as Sun Microsystems' SPARC processor. They helped shift the computing industry from East Coast giants like IBM and Digital Equipment Corp. to Silicon Valley.
Moving to accelerators
It's RISC's turn to feel the pressure now. With the steady progress charted by Moore's Law now faltering, even RISC chips have lost luster. Much of today's excitement focuses instead on that can speed up particular computing tasks like artificial intelligence. Patterson is now exploring that work at Google, which already is on its second-generation tensor processing unit for AI chores.
Even so, RISC chips are still crucial where general-purpose processors are needed -- which is just about everywhere, including devices that also benefit from accelerators. Apple's A series of processors for iPhones are RISC designs, as are Samsung's Exynos models and the Qualcomm chips that power countless Android phones. Even Intel, whose chips don't look like RISC chips from the outside, tuck RISC technology deeper inside for a speed boost to its x86 family of PC and server chips.
Patterson, now a distinguished engineer at Google, was a computer science professor at UC Berkeley from 1976 to 2016. Hennessy is chairman of Google parent company Alphabet and was Stanford University's president from 2000 to 2016. The pair have worked together since before Google, though, as coauthors of "Computer Architecture: a Quantitative Approach," a textbook nearly two decades old now and in its sixth edition. They're also members of the National Academy of Engineering and the National Academy of Sciences.
The Turing Award has previously gone to luminaries like World Wide Web inventor Tim Berners-Lee, internet technology inventors Vint Cerf and Robert Kahn, Unix creators Ken Thompson and Dennis Ritchie, and cryptography pioneers Whitfield Diffie, Martin Hellman, Len Adleman, Ron Rivest and Adi Shamir.
A simpler vocabulary
RISC swept away the prevailing approach, called complex instruction set computing (CISC), in which told chips what to do with the computing equivalent of "large, polysyllabic words," Patterson said. "With the RISC idea, we wanted designs to be quicker, with a simpler vocabulary of monosyllabic words."
Their work was directly commercially relevant.
Programmers use tools called compilers to translate high-level languages better-suited to human comprehension into the low-level machine code that the chips in computers understand. CISC designs dated to an era when programmers wrote much lower-level language called assembly language, and chip designs catered to that approach, but Patterson and Hennessy designed RISC for the newer high-level programming era.
"Microprocessor companies would have wars over who had the most instructions. Intel would come up with three new ones, then Zilog would add two more," Hennessy said. "No compiler in the world was ever going to use these instructions. It was a reflection of era when people wrote a lot of assembly."
Doing more at once
RISC's more nimble design paved the way for technology that boosted processor performance, Gwennap said.
One concept is pipelining, which feeds new instructions to the processor before old ones are finished executing for faster performance. Think of having a separate washer and dryer so you can wash one load of laundry while drying another instead of a combination appliance that can only do one thing at a time.
Another is superscalar chip design, which lets a chip use multiple pipelines in parallel.
"As soon as you start pipelining, you're overlapping instructions, and going superscalar, doing two, three or four instructions at the same time, that whole philosophy of the CISC process didn't work very well," Gwennap said. "The insight Hennessy and Patterson had was if we make these instructions simple ... we can build much more complicated processors that can overlap instructions and become a lot more powerful."
Processors are historically proprietary technology. Companies like ARM -- whose technology is used in essentially all mobile phones -- have open licensing policies, but it'll still cost money.
But there's a new design changing those rules, RISC-V (pronounced "risk five"). It's entirely open, permitting anyone to use and adapt its design. The project began at Berkeley, and Patterson is vice chairman of the RISC-V Foundation board.
"You can grab an open-source CPU off the internet, use the instruction set for free, do whatever you want -- change it or build your own processor. These are things you haven't been able to do in the past," Gwennap said. "It's opened the door to creativity. People can build a CPU who couldn't before."
Experimentation is great, but Hennessy doesn't expect any breakthroughs for chip design that will advance general-purpose processor performance the way RISC did.
"We would have found it by now," Hennessy said. Patterson agrees, but he and Hennessy aren't entirely pessimistic.
For Hennessy, the key is to look beyond the old ways of improving processors. "It's like the Renaissance. If you were in the new age of artists -- if you were Leonardo or Michelangelo or Botticelli, it was a great time. If you were still stuck in the old flat gothic images that didn't have perspective or look like human figures, it was a terrible time," he said.
Patterson recommends rethinking chip architectures, too. And the good news is that RISC-V, affordable manufacturing and new design tools are liberating processor experimentation.
"Anybody can prototype," Patterson said. "It's a brand-new golden age."
Blockchain Decoded: CNET looks at the tech powering bitcoin -- and soon, too, a myriad of services that will change your life.
The Smartest Stuff: Innovators are thinking up new ways to make you, and the things around you, smarter.