I recently attended an excellent new course offered by Acceleware covering the C++ Accelerated Massive Parallelism (AMP) library and tools. AMP extends the C++ language to enable high-scale data-parallel computational algorithms and targets devices such as Graphics Processing Units (GPUs). Significant advantages include device independence (i.e. via a DirectX abstraction layer) and integration with Visual Studio for development, debugging, and profiling.
See the course information with schedules here. Learn more about C++ AMP via the MSDN Parallel Programming in Native Code blog.