Because if we define it, then gcc compilation fails because it doesn't support --with-float for aarch64.
e316672dcb