Implicit conversion from float -32.93388429752066 to int loses precision