I want to start using CUDA. My programming environment is VS 2010, C#.Net 4.0. Already downloaded CUDA SDK and Latest Drivers. I spend few hours looking for resources to learn how to make the connection; but there are lots of options and little information about each one of them: GPU.NET, Cloo, OpenTK, Cudafy, etc. and couldn’t figure out which one is more reliable or popular.
Eventually I’ll be using this in a commercial project, so I try to stay away from opensource projects that die after a while and there is not good support and reliability assurance around them.
I would appreciate it if someone could provide some information about which tool is more mature and reliable, and direct me to some resources.
Cudafy seems to cover the bases fairly well. I’m not familiar with the other projects you mentioned. I too initially wanted to start CUDA development from the comfort of C#, but eventually gave that up to work with CUDA in C++, using the NVidia CUDA toolkit.
You have enough new concepts to get your head around learning core CUDA itself. Adding a third party layer on top of that only adds to the conceptual load. I’d suggest learning CUDA on its own terms first (in C++), then once you are proficient in core CUDA you can take a look at .NET layers and evaluate how well each exposes core CUDA for the application you have in mind.