AVX2 code cannot be faster than gcc base optmization

Question

I am studying AVX by writing AVX code with inline assembly. In this case, I tried to implement AVX in a simple function. The function name I made is lower_all_chars_ba