Announcement of C ++ AMP - an open platform for building heterogeneous (GPU + CPU) C ++ applications

Hello, Habra users! Now I am in the US, Bellevue, at the AMD Fusion 11 Developer Summit conference . I will write about my impressions of the city and the conference later, when I summarize all the accumulated materials. In the meantime, I would like to tell you about the important announcement that was made today during the Microsoft plenary report.

Microsoft presented its vision of developing applications for working in a heterogeneous environment using all the capabilities of multi-core processors and GPUs. A new platform was introduced, which will be part of the next version of Visual Studio.

The new platform is called C ++ Accelerated Massive Parallelism or C ++ AMP. The goals of this platform are as follows: to make it possible to take advantage of parallel computing and GPU at the same time in C ++, just as today these advantages are used in C (or its similarity) in different development platforms: OpenCL, CUDA, DirectCompute.

During the plenary report, the use of a new platform for developing applications that require high computing power was clearly demonstrated. So, with the capabilities that C ++ AMP has, excellent results have been achieved on the use of existing GPUs and multi-core CPUs, even on a netbook, which was able to produce dozens of gigaflops for solving the demonstration task of calculating the interaction of many bodies in space. A more powerful system with two GPUs allowed hundreds of gigaflops of resources to be used in the same code (a result of 800+ gigaflops was demonstrated).

Representatives of Microsoft especially emphasize the prospects of this platform for the development of cloud computing in the future. The platform was created with the expectation that soon we will be able to use hundreds of millions of GPU computing modules in our applications through cloud services. C ++ AMP will allow you to write such code today so that it works at similar capacities tomorrow.

An important part of the C ++ AMP announcement was Microsoft's announcement that the new platform will be released under an open specification, so that any company or developer for any of the platforms can create its implementation! Microsoft also announced that it intends to help third-party implementations of the new platform and cited cooperation with AMD in this matter as an example.

Some details about the new platform can be obtained right now in the developers blog at www.danielmoth.com/Blog/C-Accelerated-Massive-Parallelism.aspx .

In addition, Microsoft today announced that in the next version of Visual Studio, in addition to C ++ AMP support, support for debugging and profiling code on the GPU will be added!

Today it’s all the news from the AMD Fusion 11 conference, later I will write a detailed review of the event with company announcements and my impressions.