Implicit conversion from float -6.343873517786562 to int loses precision