Implicit conversion from float -33.364485981308405 to int loses precision