Implicit conversion from float -12.047244094488192 to int loses precision