Implicit conversion from float -7.4683544303797476 to int loses precision