I’m trying to create an exportable C library DLL so that I can use the Cuda engine to accelerate algorithms that I can use from C# apps. I used the regular C syntax to make an exportable C funtion from a DLL and created a C# app to use it. I can run in debug mode and see it step into the C function just fine, but it crashes on the first Cuda function it hits. Does anyone know of any tricks to get this to work? The demo that I created worked fine when it was an executable, and the export code seems to work fine because I can step into the library from the C# app. The problem seems to be that Cuda isn’t being initialized properly when used inside a dll (?). Any help would be appriciated. Thanks
I’m running into the same problems. If I compile as an .exe, my code runs fine, but if I compile as a .dll, the same code crashes. If I comment out the cuda-specific code and just call simple dummy functions, then my DLL runs fine. With the cuda code in place, sometimes the DLL crashes my calling application when I first load the DLL, before explicitly calling any exported functions, while other times it doesn’t crash until the host application invokes a function in the DLL which calls cuda code. But in either case, I’m never able to run cuda code which is inside of a dll. Has anyone made this work – putting all the CUDA / GPU code in a .dll, with the host application unaware of the GPU’s presence?
It is a known bug that has been fixed.
You will need to wait for the next release.
When is the next release and what will be the release number?
Thanks in advance