My rule of thumb is that if I need to worry about the IEEE rounding mode making a difference in my results, then my problem is really ill-conditioned and rounding the least of my worries. So I usually just ignore it.
Same applies to rounding for data displays: if I am worried about 3.5 being rounded to 3 or 4, maybe I should just display that extra significant digit. Best, Tamas On Fri, Dec 26 2014, Stefan Karpinski <ste...@karpinski.org> wrote: > Regarding this particular change, I'm personally not really in favor of the > default being the IEEE ties-to-even behavior – I don't think the numerical > argument in its favor is really compelling and I think the current C-like > round-ties-from-zero behavior is less surprising and should continue to be > the default. But I'm not the one doing the work on this and I'm not the > most informed person to make such a decision. So let's try the IEEE > behavior and see how it goes. If this turns out to be problematic or > annoying (I kind of suspect it will), we can still change our minds before > Julia 1.0, so now is the time to give it a shot. > > On Fri, Dec 26, 2014 at 3:08 PM, Simon Byrne <simonby...@gmail.com> wrote: > >> On Friday, 26 December 2014 06:14:34 UTC-6, Hans W Borchers wrote: >>> >>> I started this thread long time ago with a question about rounding rules >>> and the IEEE floating point standard. I felt like being criticized for even >>> thinking Julia could follow the "round-to-even" rule. Now I learn that >>> Julia version 0.4 will apply this rule (as default?). >>> >> >> I apologise if you felt I was being critical, that was not at all my >> intention. These are complicated issues, and I don't claim to have all the >> solutions. In fact, we still have yet to resolve the problem with the >> digits argument: I have just opened an issue explaining the problem here: >> >> https://github.com/JuliaLang/julia/issues/9464 >> >> Contributions to the discussion are certainly welcome. >> >> -simon >>
