couldn't compile default application

I’m new to CUDA toolchain, and this question may be silly, but… Newly created project in Nsight (Eclipse) fail to compile.

What I’m doing:
Create new C++ Project with type “CUDA Runtime Project” and Toolchain “CUDA Toolkit 5.5”
Separate compilation with PTX and GPU codegen 3.0 (where I can read more about this 2 things?).
And thatn just “click on hammer”.

Dosens of erros arrived. 90% of them dissapears after addition of cuda_runtime_api.h. But 3 of them still here.

Here is make output (sorry for russian):

05:26:03 **** Incremental Build of configuration Debug for project Test ****
make all 
Building file: ../src/Test.cu
Invoking: NVCC Compiler
/usr/local/cuda-5.5/bin/nvcc -G -g -O0 -gencode arch=compute_30,code=sm_30 -odir "src" -M -o "src/Test.d" "../src/Test.cu"
/usr/local/cuda-5.5/bin/nvcc -G -g -O0 --compile  -x c++ -o  "src/Test.o" "../src/Test.cu"
../src/Test.cu: В функции «void bitreverse(void*)»:
../src/Test.cu:41:8: ошибка: нет декларации «threadIdx» в этой области видимости
../src/Test.cu: В функции «int main()»:
../src/Test.cu:59:14: ошибка: expected primary-expression before «<» token
../src/Test.cu:59:54: ошибка: expected primary-expression before «>» token
make: *** [src/Test.o] Ошибка 1

05:26:03 Build Finished (took 209ms)

What I’m doing wrong?

PS:

OS: Ubuntu 12.04 LTS
Device: GTX 690

Your file language is set to “c++” (see the “-x c++” switches passed to the nvcc). It should be “CUDA C”.

This can be caused by a user setting on the file (check “General / Language Mappings” page), file parent folder, project - or on the “C/C++ / File Types” preference page.

Note that CUDA C language is a C++ extension, meaning you may use most C++ (except for some C+±1x features) in .cu files.