If power is the ultimate aphrodisiac, then the emerging Don Juan of the computer software world may well be Canadian startup RapidMind Inc.

The new addition to the Waterloo, Ont., high-tech region is turning the heads of software company executives around the globe, according to Ray DePaul, RapidMind's president, and chief executive officer. The reason?

"It's incredibly tough to program to multicore [chips]; there are not that many people in the world that can even do it," says RapidMind president and chief executive officer Ray DePaul.

"There are 20,000 software companies in the world and thousands more in-house development groups in places like banks and insurance services companies," DePaul says. "And they all need to deal with the challenge of [adapting their software for] multicore processors if they want to survive."

If you're not an investor or tech industry insider, you probably haven't heard of RapidMind. Still, the company has those in the know excited about its potential business prospects.

Steve Conway, research vice-president of technical computing systems at market analysis firm IDC, says high-performance computing is the number three growth industry among the high-tech sectors his company monitors. "Only gaming and flat panel televisions are better performing over the last couple of years. It's a $10 or $15 billion market place just on the hardware side … and RapidMind is sitting in a very prominent place."

Performance issues

Consumers demand that today's software performance be faster and smoother than ever. Canada's RapidMind is currently the only independent player in the world helping software companies build products that really take advantage of the capabilities of new multicore computer processors.

Traditionally, it has been the clock speed of a computer's processor that determines how powerful it is, and big chip manufacturers like Intel have generally upgraded their processor speeds every year or so. Until recently, running old software on computers with new chips simply meant the programs would run faster.

"Software developers got an easy ride using the same old tools and tricks for years to be successful," DePaul says.

But the latest chip designs take a more complex approach to improving performance than simply boosting clock speeds, and it's giving old-school programmers headaches.

"They [chip designers] kept figuring out how to add more transistors on one chip," Conway says. "But eventually problems started showing up. The chips got very hot or there were different, spiky, energy draws."

So chipmakers like Intel and AMD have gotten fickle about the idea of creating faster chips, and instead have started adding more processor cores to a single computer chip. In other words, instead of one fast processor doing all the work to run programs on a computer, multicore systems break computing tasks into pieces that can be handled simultaneously by a number of processors built into the same chip.

"So they [chips] are not going faster the way they used to go faster," DePaul explains. "The software industry is struggling to stay up-to-date with that."

The problem is that software developers out there have fallen into a coding rut; they've gotten used to building software in certain ways, and your computer's multicore processor requires new programming techniques that streamline software to run on multicore machines.

"It's incredibly tough to program to multicore; there are not that many people in the world that can even do it," DePaul says. "And a lot of them work for us."

New techniques

For the past 20 years, the average software developer has been coding in C++, which is sort of the international programming language of software developers. It's a reliable and effective method for making software that runs on computers with single-core processors. But, like most programming languages, C++ doesn't do the trick when it comes to squeezing performance out of multicore chips.

"Most software that the average person or business has on their desktop only runs on a single-core," says DePaul, which means it is meant to run in its entirety on a single chip. "So, if your computer has a dual or quad-core processor in it — and nearly every new computer with an Intel or AMD chip in it does — that means your software is only running at 25 or 50 per cent of its potential capacity."

Another way to put it is that it doesn't matter how fast your computer's multicore processor is, because your old-style software is probably too dumb to take advantage of it. So companies that make software have to rethink the way they do things.

"It's really a paradigm shift for the software development industry," Conway says.

RapidMind helps software companies address the challenge of creating smarter applications by providing an introduction service of sorts between old-school software and those sexy, young multicore chips. In simple terms, RapidMind lets a developer translate code written for single-core machines into something that can take advantage of the design of a multicore chip.

"Our platform sits between the application [software] and the hardware [a multicore processor]," DePaul says. "So, basically, software developers can use tools they're comfortable with like C++, and the RapidMind platform does all the heavy lifting," to optimize the software's performance for your multicore computer.

The end result for the consumer is faster loading videos, smoother game play, and the demise of that little spinning hourglass you get when your computer is coquettishly stalling for time.

A value proposition

Every software development company in the world knows it has to master multicore soon if it wants its products to remain attractive to businesses and consumers, and at the moment RapidMind is the only independent shop where developers can go to get the multicore makeover.

RapidMind's only competitor, Peakstream Inc., was snatched up by technology titan Google a few months ago.

"I think it says something that Google, which arguably has one of the biggest computer operations centres in the world, went out and bought an entire company to deal with their own multicore issues," DePaul says. "It certainly makes our platform more attractive to software companies that need it."

That leaves RapidMind as the belle-of-the-multicore-ball, with potential partners like IBM and Hewlett-Packard clamouring for attention. "It's a fast growing market and what Rapidmind does addresses a big piece of it," Conway says.

So how come you've never heard of them?

"It reminds me of the early days when I was at RIM [Waterloo's Research In Motion] and we were trying to create a market, and nobody really knew what wireless e-mail was all about," DePaul says. "But this time it's not consumers, it's big software companies that are eyeballing multicore; and they already know what we have to offer."

It doesn't hurt RapidMind's attractiveness that the company was recently infused with a $10-million dowry from investors, either.

"Maybe there's an IPO in the future or maybe not," DePaul says. "Our funding is going to carry us for a long time."

One way or another, Rapidmind is likely to garner a lot of attention from technology companies.

"I'd be foolish to try to compare RapidMind to RIM, but we solve as big a problem," DePaul says. "Multicore is one of those technology disruptions that only happens every couple of decades."

[an error occurred while processing this directive]