Implicit conversion from float -21.446629213483135 to int loses precision