In the case of dx as an infinitesimal it makes sense. You are making a sum of all the values of the function in the integral range and multiplying with a constant dx.