David May, parallel processing pioneer
We salute the architect of the Transputer
Unsung Heroes of Tech "It's very distressing - I'm watching almost with disbelief. The Americans cannot get it out of their heads that if you're trying to build machines with lots of processors, you don't assume that they all share a common memory. The world doesn't have a common database. We pass messages to one another."
David May, professor of computer science at the University of Bristol, is talking about the current trend in chip design that proliferates cores - Intel's 'Knights Corner' currently runs to 50 processors on a single chip - but has them all dipping into the same memory pool.
David May today
"The memory is a bottleneck even in a well-designed single-processor system," he says. May should know. He was the chief architect of the UK's famous foray into parallel computing way back in the 1980s - the Inmos Transputer.
Inmos was a government-funded IT company founded in 1978. Its mission was the implementation of a radically new microprocessor architecture: a complete computer-on-a-chip, comprising a fast, low-complexity CPU, substantial local memory and agile input-output circuitry to enable it to communicate efficiently with other transputers.
The key idea was to create a component that could be scaled from use as a single embedded chip in dedicated devices like a TV set-top box, all the way up to a vast supercomputer built from a huge array of interconnected Transputers.
Connect them up and you had, what was, for its era, a hugely powerful system, able to render Mandelbrot Set images and even do ray tracing in real time - a complex computing task only now coming into the reach of the latest GPUs, but solved by British boffins 30-odd years ago.
Cutting the cord
To understand where May is coming from, we need to wind back to the Warwick University robotics labs of the early 1970s. A machine with articulated arms and a Cyclops eye trundles across the floor towards a plastic cup on the lab bench. In the corner of the room, a lineprinter plugged into a DEC PDP-11 raps out the word "CUP". The robot reaches out towards the cup and tries to pick it up.
But May isn't looking at the robot or the cup. He's focused on the thick umbilical that connects the robot to the minicomputer. It's a big, big bundle of cables, a pair for each of the sensors in the robot feeding data back to the PDP-11, and a pair for each of the robot's actuators. All those wires are telling May the computer's in the wrong place - it should be inside the robot.
On-die memory: the Transputer had it 30-odd years before it became fashionable
Instantly, it strikes him that that's wrong too. You don't want one big PDP-11 in the robot, you want lots of tiny PDP-11s, one for each of the sensors and actuators. And a way of networking them all together.
Shrinking the minicomputer
Tiny computers were already available to May. Intel, founded in the late 1960s by Robert Noyce and Gordon Moore as a memory manufacturer, had already introduced the 8-bit 8008, the chip credited as the seed for the PC revolution.
But the processor May eventually chose came from Texas Instruments. "The TM9900 was a PDP-11-like machine on a chip," he says. "It had fast serial interfaces, so I was sticking these things on boards, each with their own memory, and then networking them together for message-passing."
The system needed a programming language, so May wrote something he called EPL - Experimental Programming Language. At Cambridge he'd been taught compiler writing by Martin Richards, whose own BCPL programming language laid down the foundations for what eventually became C. BCPL was the language that drove the Warwick departmental minicomputer, the Modular One, built by a company founded by British engineer and entrepreneur Iann Barron.
"The BCPL compiler was faulty," says May, "because the port to the Modular One was never finished." Fixing BCPL was May's nursery slope to developing EPL.
One board - the B000 - 42 Transputers
From Parallel Theory to Practice
But EPL was very different. BCPL instructed a single processor to do one thing, and then the next thing, and then the next. EPL sent out its instructions to a hive of processors, each with its own memory. This made them effectively autonomous computers, and EPL's job was to distribute tasks among them and co-ordinate the results.
The far-from-trivial logic behind this kind of parallel computing had been formulated in the mid-1970s in a paper by Tony, now Sir Charles Antony Richard, Hoare. Already famous for Quicksort, the widely used sorting algorithm, Hoare had devised a theoretical language, CSP (Communicating Sequential Processes) that might enable parallel computing. It was this logic that May was now trying to implement in the real world.
May wasn't alone in his quest for workable parallelism. "In 1975, 76 there were three groups in the UK," says May. "Myself and a colleague, a couple of guys in Edinburgh, and George Coulouris and his group at Queen Mary College. We'd all, in one way or another, made proposals to the then Science Research Council (SRC) to do research on distributed computing when nobody knew what it was."
Dramatis Personnae: Iann Barron (left) and Tony Hoare
Source: Mark Jones/Inmos.com (IB) and Wikipedia (TH)
Nobody in the Labour government knew either. There was no Minister for Information Technology, although there had been a vociferous advocate for one for several years: Modular One maker Iann Barron. And it was to Barron that the SRC turned for advice.
Inmos: the rise and fall
Barron already had a scheme, hatched in conjunction with IT entrepreneur Richard Petritz and chip designer Paul Schroeder, both from the US. Together they ended up with £50m funding for an almost insanely bold new venture Barron sold to the government as a job-creation programme, the seed for a modern UK electronics industry and a bridge to the US market.
Inmos was born in 1978 with a five-year plan: spend that time - and a lot of money - developing and bringing to market a dedicated parallel processing chip, the Transputer, guided by May. But the company would start making money almost immediately by selling memory from its SRAM and, soon, DRAM factory established in Colorado Springs.
Promoting the Transputer: an Inmos brochure from 1986
Source: Dean Allum/Inmos.com
"The theory was that the static and dynamic Ram chips would be a very quick entry into the market," says May. "We had top-class Ram designers stolen out of Mostek. The memory story was right. If you can make a cheaper, faster products with the same interface as the competition the customers will buy from you. But the microprocessor has all the opposite dynamics. It takes a long time to design, a long time to establish, but once it's established the customer's locked in."
The multilingual - human and computing - polymath B Lee Jones remembers working as a systems programmer for the Colorado Springs facility. Jones gave the first US presentation of the Occam parallel programming language - May's EPL successor - to a DEC Symposium in 1981.
"David, Tony Hoare and the gang changed the whole computing paradigm," he says. "But it was too early, because a lot of people didn't get it."
Occam source code
Source: Dr Daniel Hyde
His fellow workers at Colorado Springs certainly didn't understand what the Brits were up to. "There was this focus on getting from the 16K SRams to 64K DRams. Yeah, Occam and the transputer, OK, that's interesting... But we're in the Ram business."
And then the Colorado Springs memory cash cow created to feed the Transputer's long-term development, dried up when the Japanese started dumping 64K Drams on the US market in the early 1980s. Rival memory manufacturer and neighbour Mostek was the first to be hit. "I had a nice corner office overlooking the Mostek parking lot," says Jones. "I came in one Monday morning and there were no cars there. They'd shut down Mostek over the weekend."
Footnote or stepping stone?
When Inmos was sold to Thorn EMI under the privatising Thatcher government in the mid-80s, and later to SGS Thomson, Transputer development continued, but was eventually abandoned.
The end of an era? It's not so easy to kill the Hydra-headed Transputer. The site of the original Inmos design centre in Bristol is now owned by STMicroelectronics. It is ablaze with engineering talent nurtured under Inmos.
Source: Ram Meenakshisundaram/Inmos.com
"They're developing products that are turning over more than a billion dollars a year," says May. "And have been for the last decade. It's the biggest Silicon Valley in Europe in terms of design capability. It's all there because of this huge expertise in processor technology."
Other companies like Infineon, Element 14 and Icera have mushroomed from the same Bristol talent base, as has May's own university spin-off, XMOS, that perpetuates the Transputer idea in low-cost microcontrollers.
So Inmos paid off in UK job creation. But in these days of ever faster Intel-style processors, who remembers the technology?
Roger Shepherd was one of graduates who worked with May in the Inmos days. Now director of processor design at STmicroelectronics, he sees the ideas around the Transputer coming back into relevance. "Most PCs today have two or four cores," he says. "How to program for multicore systems is one of today's key challenges. People are now asking the same sorts of questions we were trying to answer at Inmos 30 years ago. They are trying to solve real problems that Occam overcame."
Iann Barron, now in his mid-seventies, looks back on Inmos as a venture deferred rather than defeated. He likens the Transputer to Babbage's mechanical computing engines, devised in the 19th Century, their significance not appreciated for more than 50 years after his death. "People will eventually see what we were talking about, and they'll say, 'yes, well, they did it'." ®