For native building we have to download CUDA and HIP.
What happens when we cross compile sycl from Linux to Windows?
Since it is not a runtime(yet) I assume we have to cross compile the whole toolchain.
But how do we cross compile the toolchain when CUDA Toolkit and HIP are Native installs
Do we just download binaries for them, point them in llvm install dir with some env vars then just cross compile as normal? Or something else?
A lot of development tools don't work on non linux systems, so having one host is very important