On Sun, Dec 28, 2014 at 9:09 AM, Jason Riedy <ja...@lovesgoodfood.com> wrote:
> > The choices in 754 were made with a nearly revolting amount of > deliberation. Always makes me sad to see them shrugged away. > I can assure you, we have never shrugged away anything in IEEE. We've actually asked Kahan in person and by email about a number of things. And the fact that we've been discussing this so extensively can't really be considered a shrug. Round-to-even is meant to allay bias problems for the vast > majority of programmers, like those responsible for the Vancouver > stock exchange index in 1982. They used truncation which, over > time, roughly halved the index. Rounding upward in the same > scenario would have doubled it... I just don't see this particular argument as very compelling – that's one example where ties to even happens to fix a problem, accidentally as it happens. There seem to be just as many cases where other rounding modes are better. My main argument is two-fold: a) none of the choices of rounding modes for ties are obviously better across the board and b) following the C behavior seems the least surprising to the largest number of people. But that's just my opinion. We are going with IEEE rounding as a default. Not only are we not shrugging off IEEE rules, we're going with them.