Last week, Intel made a strategic acquisition in the shape of Rapidmind, a company that grew out of the University of Waterloo in Canada. The company was formed to commercialise Sh, a programming system designed to optimise the use of the GPU.
From a programming point of view, the important thing is that Rapidmind enables programmers to focus on developing code as they always have. They can write serial programs in C++ as they did before, and Rapidmind’s technology takes care of parallelising execution. Not only does that save the time and complexity associated with parallel programming, but it also avoids any of the bugs specific to parallel programs, including data races, deadlock and livelock.
This is a good fit with Intel’s own work on Intel Ct technology, which brings data parallelism to C++ programming and is designed to abstract the programming away from the hardware, which is what all the popular single core languages have done.
There is already a roadmap that includes Rapidmind. Intel is planning a beta release of Intel Ct technology later this year, and next year will release a new version incorporating technology and intellectual property from Rapidmind and Cilk++ (which was acquired earlier this month).
For more details, read this blog post by James Reinders of Intel.