Implicit conversion from float -12.434456928838955 to int loses precision