There are some issues with integration of NVCC cuda with QtCreator. In a CMake project, one can define the NVCC compiler in the CMAKE_CUDA_COMPILER or CUDA_NVCC_EXECUTABLE, as well as the cuda-gdb debugger in the Kit. This works fine, and debugging sessions work great. However:

• Output errors are not parsed correctly QTCREATORBUG-11315
• Cuda-memcheck cannot be run within QtCreator. Ideally one should be able to select it from the dropout menu where "callgrind, memcheck, etc. are shown
• A CMake project where both .CPP and .CU files are mixed (cmake automatically uses gcc or nvcc depending on the extension) build fine, but is not parsed correctly in the editor. The 'live' code analyzer thinks always it is a GCC project and reports a lot of warnings on the left of undefined variables, which makes it hard to digest if you are looking for real errors.
• In Manage Kits, one has to assign a CUDA compiler to either C or C+, which does not make it usable in case the project is mixed with CUDA and C++. An extra row for CUDA could be helpful?
• Clang-Tidy should recognize .cu files also, at least it could run in --host-mode http://clang-developers.42468.n3.nabble.com/clang-tidy-and-CUDA-tp4061048p4061098.html

Thanks in advance.