Implicit conversion from float -32.50632911392404 to int loses precision