Inventor forges fresh approach to writing software

McGraw spent nearly a decade starting about 1985 developing the Sisal functional language. Prototypes made extensive use of math formulas to solve large problems, but ultimately the effort collapsed from a lack of funding.

"I'm not happy with any of the parallel programming languages available today, but the cost and time it takes to develop a new language is enormous. It's a pretty depressing situation, frankly," McGraw said.

Aynur Unal took a more upbeat view of COSA. Unal is an electronics executive who has worked for a variety of startups in areas such as supply chain management as well as large companies such as Flextronics.

"Gordon's work represents a very generalized way of writing programs that could be very useful in software engineering," Unal said.

"I have introduced him to several people in Silicon Valley, and I think it would be great to get him in a university post where he could teach students about COSA," she said. "I have also recommended him to several large-scale private universities forming in India, and they are very interested. India is really the software engine now," she added.

"This is the right time for new parallel programming models, but there will be no silver bullet," said Anant Agarwal, an MIT researcher in the field and chief technologist at Tilera Corp. which is launching its own parallel CPU. Agarwal was not familiar with COSA.

Morrison said he still needs to work out a number of implementation details in COSA, and estimated it could require $15 million to realize the concept in production code.

Morrison started his career working on advanced radar systems for an arm of the government. Later he spent three years at Data General and then had a number of short stints at high-performance computing companies. After filing a handful of key patents related to parallel computing he was able to fund work as a sole inventor based on license royalties from a number of companies including IBM, Motorola, Texas Instruments and Intel.

He started work on COSA after filing a patent on some of its core ideas in 2002. "I just want to develop technology," he said.