Implicit conversion from float -100.63745019920319 to int loses precision