Implicit conversion from float -91.75182481751824 to int loses precision