Implicit conversion from float -2.0027816411682977 to int loses precision