Implicit conversion from float -0.3092783505154557 to int loses precision