I know this is not a CMAKE forum, but perhaps somebody can help me. Perhaps I might be totally lost.
When I enable CUDA_SEPARABLE_COMPILATION in CMake, the PTX are not being added to the dll file.
(cuobjdump file.dll -lptx is empty).
However, if I disable it (it is disabled by default), I see several PTX files generated inside the dll.
I know CMake uses the -dc flag instead of -rdc=true, but I guess this is equivalent.
Does anybody know how to generate PTXs using separable compilation?
Thanks in advance,