Implicit conversion from float -91.17977528089887 to int loses precision