Is there a problem in this example code from the CUDA C Programming Guide?
__device__ float devData;
float value = 3.14f;
cudaMemcpyToSymbol(devData, &value, sizeof(float));
I can't understand how it could write to devData without having the address of devData