Implicit conversion from float -54.54545454545453 to int loses precision