Hi everyone,
I am wondering by what rule is one among many cudart{bits}{toolkit.version}{build.number}.dll chosen as a dependency when compiling cuda code in Visual Studio 2008. I have installed CUDA toolkit a while ago and I have cudart32_32_7.dll and cudart32_32_12.dll in my bin path. The 7 one is used. My colleague downloaded toolkit few weeks after me and he has only cudart32_32_16.dll in his bin.
My questions:
- How to dynamically detect a dependency? I need to include a cudart dll in deployment package.
- Should I be afraid about the difference that we code against slightly different version of the same toolkit 3.2 ?
Many thanks!
Jozef