C++ AMP - Accelerated Massive Parallelism with Microsoft Visual C++ is the first book on the market to address the C++ AMP API. I preordered it months ago and when it finally arrived, I devoured it in just a few days.
The book starts with an overview of General Purpose GPU computing and how C++ AMP provides the developer the ability to write naturally parallel general purpose solutions (normally executed on the CPU) in C++ that will execute on the GPU.
Immediately following, in chapter 2, is a taste of the power of using Accelerated Massive Parallelism on the GPU, an NBody program that simulates the gravitational interaction between thousands of stars in 2 star clusters. Implemented first on the CPU and then with C++ AMP. The results are startling.
Then the learning begins with the C++ AMP API, which is implemented entirely with C++ templates and just 2 new keywords added to the C++ language. The authors explain the fundamentals of C++ AMP and provide enough code to demonstrate the handful of templates that make up the API. This is followed by more advanced techniques (Tiling) that speed up your C++ AMP programs even more.
As the book progresses, you will learn how to debug your C++ AMP code with the support for C++ AMP that was added to Visual Studio 2012: IntelliSence, GPU debugging and profiling with the concurrency visualizer. A few case studies, with plenty of code, are used to demonstrate what you have learned.
Other chapters include: Optimization, Working with Multiple Accelerators, Graphics Interop with Direct3D and the books finale with Tips, Tricks, and Best Practices.
If you work on projects that use large datasets that can be processed in parallel, you can move that work to the GPU and multiple accelerators without learning a new language. This book does a wonderful job of showing you how!
From the end of Ch. 1,
"With the cloud, using Infrastructure as a Service (IaaS) or Hardware as a Service (HaaS) offerings, you could conceivably leverage tens of thousands of hardware threads. But imagine being able to combine the two and reach the GPU cores on those cloud machines, reaching tens of millions of hardware threads. What could that enable?"
On a personal note, I submitted an issue I had to Microsoft Press through the books web site. In less than 3 hours I received a response from Ade Miller himself! That is a great demonstration of the author's commitment to the quality of this book.