About Closing this Question
Just the comments to my question already helped me a lot. This question is helpful to programmers. It should therefor remain open. It is also not a duplicate, because the other CUDA with Boost questions mostly concern using Boost in host code, not in device code. This is a significant difference, because these programming environments are different, the resulting machine code runs on different hardware and the languages have different compilers.
--
Question
There are numerous questions on SO regarding the use of boost with CUDA. Many of them give the impression that using boost in CUDA is a hassle. Some use boost only for host code while meticulously keeping an eye on not including any boost dependencies in the CUDA part or even in device code.
One example is Cuda with Boost, which claims that
NVCC cannot handle the certain constructs used in the boost code
However, the link in this question to the nvidia forum is dead.
Is it still a hassle to use boost in CUDA device code?