Implicit conversion from float -0.4143646408839743 to int loses precision