Not following the IEEE 754 spec to the letter means "D has a broken floating point implementation." That's not going to happen. People write very carefully crafted floating point code relying on exact adherence to the spec.