Great idea! I just noticed however that their IR language has no "real" (long double) float representation? L.