Implicit conversion from float -20.063694267515928 to int loses precision