Implicit conversion from float -32.51082251082251 to int loses precision