Cure for the common code gives apps more flexibility

Alasdair Rawsthorne

Alasdair Rawsthorne may have started out slowly, but he’s making up for lost time.

He experienced his “aha!” moment in the early ‘90s. A computer science professor at the University of Manchester in England, Rawsthorne was spending his sabbatical working at the Advanced Computer Research Institute, a supercomputer manufacturer in Paris.

“Our CEO decided the supercomputer needed to be compatible with an existing processor’s instruction set, but he kept changing his mind about which processor,” says Rawsthorne, CTO and founder of Transitive. “One week it was PowerPC, the next week Alpha, then it was MIPS. Each time he changed his mind, we had to throw away all our work. And then he’d say, ‘This isn’t going to change the schedule, is it?’”

Rawsthorne believed there had to be a way to run software across different hardware platforms without losing functionality, killing performance, or recompiling the code for each CPU. From this idea, QuickTransit Hardware Virtualization Technology was born.

But the development of QuickTransit was anything but quick. Rawsthorne began the project in 1995 with a university researcher. Three years later they had a version they were willing to show in public. In 2000 Rawsthorne formed Transitive to market the technology to hardware manufacturers, but it wasn’t until 2004 that the company announced its first customer, Silicon Graphics.

SGI had developed a new graphics visualization system based on Intel’s Itanium processor and Linux. But SGI’s customer base used software built for a MIPS processor and a proprietary version of Unix. The solution? Using QuickTransit, the new SGI machines run “intermediate representations” of the old software, passing instructions to and from the processor and optimizing them in real time. Rawsthorne says SGI’s old software runs faster on the Itanium than it did on the MIPS.

Without the freedom and unhurried pace of academic research, however, QuickTransit would not have been possible. “We had the luxury to try things that might conceivably fail.”

Now his biggest challenge is convincing skeptical customers the technology really works: “Everyone’s first reaction is that it’s clearly too good to be true.”

One true believer is Steve Jobs, who plans to use Transitive technology to run existing Mac apps on Intel CPUs. Rawsthorne foresees QuickTransit running on cell phones, PDAs, or any other hardware that evolves more rapidly than the apps that run on them.