Implicit conversion from float -1.5048543689320297 to int loses precision