Up to today, I was using CUDA tk+sdk 1.0 under visual studio 2008 beta 2. Everything was working perfectly fine. CUDA 1.1 breaks this compatibility and says nvcc supports only msvc 7.1 and 8.0.
we are planning on moving to 9.0 as soon as it is available on the market. Is there any reason why the compatibility was broken?
Visual Studio 2008 was never a supported platform/environment with CUDA. We are considering additional platforms in the future, however the 1.1 release will only support MSVC 7.1 & 8.0.
We’re planning to support Vista in a release next year. Visual C++ 2008 support is currently under consideration, however has not been slated for any specific release.
If you have both Visual C++ 2008 and 2005 installed, you can tell nvcc to use VC8 while compiling the rest of your application with VC9 by adding the following to C:\CUDA\bin\nvcc.profile:
compiler-bindir = C:\Program Files\Microsoft Visual Studio 8\VC\bin
Choosing a different compiler for CUDA is also very useful in Linux to use gcc 4.1 for nvcc but gcc 4.2 or 4.3 for everything else.
Currently I am only using windows in a virtual machine to test that my code compiles correctly, and the program also started successfully, without however executing any CUDA kernels as it wasn’t able to access the real graphics card. However I used a similar trick with the first version of CUDA which only supported VC7.1 and it worked very well.
I’m looking to head up a commercial development using CUDA. The nature of the project is related to the acceleration of video decoding (many simuntaneous streams, several codecs) with possibly some analysis.
We develop using Visual Studio 2008 and personally I develop on a Vista platform. The target platforms are both Vista and XP.
The compromise of having portions of a project built using a different compiler tool (i.e. vs2005) is unacceptable (because the vs2008 compiler is superior for many reasons plus it’s a risk).
Therefore, support for vs2008 is highly desirable. We use Visual Studio 2008 team edition for developers specifically. Could you confirm that you will support this development platform eventually and provide an approximate date for such support?
We also use a Team Foundation Server 2008 build server to build releases, as this is a server it has no video hardware to mention, will the SDK and CUDA build tools work on such a platform?
We are currently tentatively planning to add VC9 support in the next CUDA release after 2.0. The timeframe of that release is currently not solid enough to provide an estimate.
we would need VC9 support as well. I am working for a company doing video analysis real time, and we are looking whether we could speed up video analysis using CUDA.
VC 2008 integration would be needed for us in order to proceed with CUDA. (Or some solid instructions on how to do batch compilation in a mixed environment).