I compiled some kernels to PTX so that I could experiment with embedding them as resource strings in my DLL’s; I know the kernels work just fine (they’re just SDK examples), but whenever I call cuLoadModuleData, I get a ‘no binary for gpu’ error.
All the other posts on the forum about this said to call cuLoadModuleDataEx() to execute the PTX string, but the CUDA reference manual says that cuLoadModuleData will work as well. I don’t really need the extra options (I don’t think…) so what could the problem be?