double_as_int64 as defined in the linked source code does take care of the sign so that double_as_int64(x) == -double_as_int64(-x). This is also important for t

double_as_int64 as defined in the linked source code does take care of the sign so that double_as_int64(x) == -double_as_int64(-x). This is also important for the negative zero case.