I still quite strongly suggest you to write down the mixture logp directly. That way you can directly verify the result, avoid the loops and stacking etc. Since the density is one function as other densities, it would be faster as well. As I mentioned earlier:
You can also post your code with (simulation) data here.