Implicit conversion from float -12.363112391930827 to int loses precision