I’m unfamiliar with C++ and CUDA but trying to learn them both. I tried to use cudaMalloc and cudaMemcpy but failed.

Header file:

... // Normal include
namespace calcu{
namespace cpu{
int var;
}
namespace gpu{
__device__ int* var;
}
void cuda_test_print(int useless); 
}

Source file:

...
namespace calcu{
void cuda_test_print(int useless){
  cuda_test_print_kernel<<<1,1>>>(useless);
}
__global__ void cuda_test_print_kernel(int useless)
{
  if(gpu::var == nullptr){
    printf("wtf... n");
  }else{
    printf("ftw... n");
  }
}
}

Main source file:

int main(){
...
using namespace calcu;
cpu::var = 1;

cugo(cudaMalloc((void**)&gpu::var, sizeof(int))); // cugo is a function that checks whether cuda API works.
cugo(cudaMemcpy(gpu::var, &cpu::var, sizeof(int),
     cudaMemcpyHostToDevice));

cuda_test_print(0);
...
}

When I ran these codes, wtf is printed which means that gpu::var is a nullptr, and cudaMemcpy seems not to work. Could anyone help me out?