0
I am working on a Windows machine, and try to compile a cross platform application, that uses CUDA. I have installed version 10 of the Cuda toolkit. I am using CLion as IDE. As this is cross platform application, I am using the CygWin Toolchain. This works fine, until I try to enable CUDA as a language in my CMakeLists.txt. CygWin tries to build a simple .cu file to test the NVCC compiler. However, it specifies the input path in Linux style absolute paths
-c "/cygdrive/d/..."
Obviously, nvcc is unable to use this path. It fails with the error:
fatal error C1083: File (Source) can not be opened: "D:/cygdrive/d/..."
The output path for the object file is specified as relative path
-o CMakeFiles/cmTC_cf6a2.dir/main.cu.o
How can I tell the CygWin toolchain to use a relative path also for the input or solve the issue of the wrong mapped path?