I'm trying to find the documentation for all of the functions available for the CUDA kernels.
The CUDA Reference manual seem to include only the host functions and the CUDA programming guide only includes some details such as the accuracy of these functions but not their documentation.
Am I missing something or does this piece of documentation simply doesn't exist?
Appendices B and C of the CUDA Programming Guide contain documentation for functions that can be executed within kernels. Mostly they are for things like texture fetches, atomic instructions, warp voting, synchronization, and C standard library mathematical operations.
The documentation is very sparse because the meaning of these functions is described elsewhere, in the C standard library documentation. Some of these functions are quite esoteric. The meaning of others can be guessed at from the name. For the ones that you do not recognize, you can google for the man page with a query such as "site:opengroup.org nextafter". If you are on a *nix machine and have the C library documentation installed, typing man nextafter will also work.