Implicit conversion from float -3.5980551053484646 to int loses precision