On the growing up problem for semilinear heat equations