This page contains information on GCC's implementation of the OpenACC specification and related functionality.
OpenACC support has been merged into GCC 5, but a handful of patches are still pending that are needed for nvptx offloading, so that's not yet functional. See also known issues with offloading
See also Offloading documentation for OpenMP and OpenACC.
Note that the combined directives kernels loop and parallel loop are not yet supported; use the noncombined ones, i.e. kernels or parallel followed by loop, instead.