LGTM complains if this calculation is done from inside the sqrt() call
because it sees the explicit cast to float and assumes that sqrt() is
intended to take a double. By adding an intermediate step it should be
clear to LGTM that the float version of sqrt is intended.