Sorry, I thought jn (3, 10.0) == 0.058379379305186812343 was testing double precision but in fact the test was for single precision. So take the data that I posted but ignore my interpretations of it.