Natural simple float on integer class. However, we should use another clang -msoft-float option first when practical use.
