SUBROUTINE cuda_nvtx_init()
#if defined( __CUDA_PROFILING )
CHARACTER(len=default_string_length) :: threadname
INTEGER :: ithread
!$OMP PARALLEL DEFAULT (NONE), PRIVATE (ithread,threadname)
ithread = 0
!$ ithread = OMP_GET_THREAD_NUM()
WRITE (threadname, "(I3,A,I2,A)") ithread
CALL cuda_nvtx_name_osthread_cu(TRIM(threadname)//c_null_char)
!$OMP END PARALLEL
#endif
END SUBROUTINE cuda_nvtx_init