You're absolutely right! Shouldn't rely on memory for this. As usual, Julia gets it right, as I should have known.
--Peter On Friday, November 7, 2014 3:12:01 PM UTC-8, Alex wrote: > > Hi Peter, > > Seems you mixed this up, Excel is the one using atan2(x,y) while the > others use atan2(y,x) [1]. > > Best, > > Alex. > > [1] http://en.m.wikipedia.org/wiki/Atan2 > > > On Friday, 7 November 2014 22:50:30 UTC+1, Peter Simon wrote: > > Matlab and Fortran both use the convention atan2(x,y) to mean the polar > angle of the Cartesian point (x,y). Excel uses the convention atan2(y,x). > I think it would be better for Julia to be consistent with Matlab and > Fortran, rather than Excel. > > > > > > My 2cents, > > --Peter > > > > On Friday, November 7, 2014 1:11:35 PM UTC-8, Ivar Nesje wrote: > > > > > > kl. 22:01:23 UTC+1 fredag 7. november 2014 skrev Patrick O'Leary > følgende: > > On Friday, November 7, 2014 2:53:28 PM UTC-6, Ivar Nesje wrote: > > If you are referring to the a and b in the Float16 method, that is an > inconsistency, not a error. > > > > And you can do the same thing to swap x and y in the definitions in > math.jl. > > > > > > > > How could I not see that. That's definitely a more important issue, but > it seems like the documentation is correct about the argument order. > Swapping them in the libm wrapper would fix that. >