I have integrated CUDA in some C++ applications over the last month and was rather annoyed with redoing similar task for every applications. Therefor I decided to write a framework that helps with the integration of CUDA in existing C++ applications, especially with memory management. It includes a stl vector like class, that supports some sort of automatic memory management – meaning the class automatically transfers data to the device if it is needed there and copies it back to the host if it was changed by the device. An interface to adopt this technique to different kind of data structures is also included.
I personally consider it to be very useful, so I decided to make the framework available as open source in case someone else wants to work with it. All available information can be found at http://cupp.gpuified.de/. Some simple example applications are included in the download, which are probably the best starting point to see how to use the framework. A detailed description can be found here in chapter 4.