Microsoft opens C++ extension for other compilers

Microsoft's C++ AMP specification allows mainstream programmers to harness GPUs for number crunching

Microsoft has opened its C++ AMP specification so it can also be used by non-Microsoft compilers to harness the power of GPUs (graphics processing units).

"What we see is more and more people are looking to take advantage of the GPU in their applications," said Tony Goodhew, a Microsoft product manager for Visual Studio.

[ Keep up with software development issues and trends with InfoWorld's Fatal Exception blog. | Master the latest in Java development with our JavaWorld Enterprise Java newsletter. ]

The purpose of C++ Accelerated Massive Parallelism, first developed for Visual Studio 11 and released along with a developer preview of that software in September, is to allow mainstream C++ developers to write their programs to run in parallel across heterogenous computing environments.

It is now free to be used by other C++ compiler makers, such as Embarcadero, Intel, and the Free Software Foundation.

Increasingly, GPU makers such as Nvidia have been offering their GPUs to accelerate program performance, which can help the CPU to complete number-crunching intensive tasks. Traditionally, programming for parallel computing has been a tricky task, one mastered by highly specialized coders who have usually worked for supercomputer labs and financial trading firms.

With C++ AMP, "I won't need to be a highly trained specialist to reach this hardware," Goodhew said.

Microsoft engineers worked to minimize the number of changes to the canonical version of C++. C++ AMP does have some additional restrictions that are necessary when running the code across multiple processors. It also has a number of additional features, such as multidimensional array types, as well as support for asynchronous memory transfer, shared memory and synchronization.

For its own implementation, Microsoft used Windows' DirectCompute library for DirectX to interact with the GPUs, though compiler engineers can use other GPU interfaces, such as Nvidia's Cuda. And while this version of AMP was designed specifically for GPUs, it also lays the foundation for eventually supporting other kinds of hardware accelerators, Goodhew said.

The specification is published under the Microsoft Community Promise license.

Microsoft first released C++ AMP last September, as part of the Visual Studio 11 Developer Preview, after announcing its intentions to open the specification for outside use in June. The upcoming release of Visual Studio 11 Beta will feature additional support of the specification. The company released the specification Friday at the GoingNative 2012 C++ conference, held last week in Redmond, Washington.

The company is working with the C++ standards committee in hopes that the group will use the ideas in C++ AMP as part of the core C++ specification. It may be a while before this happens, however, given the long periods of time between updates of the standard. The latest version of the C++ standard, informally called C++2011, was released last year, and was the first major update in four years.

Recommended
Join the discussion
Be the first to comment on this article. Our Commenting Policies