https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93871
--- Comment #22 from Steve Kargl <sgk at troutmask dot apl.washington.edu> --- On Tue, Feb 25, 2020 at 07:32:31AM +0000, thenlich at gcc dot gnu.org wrote: > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93871 > > --- Comment #21 from Thomas Henlich <thenlich at gcc dot gnu.org> --- > One should also ask: What is the least surprising way to implement the > cotangent function? > > If someone uses a (non-standard) function bearing a name similar to "tangent", > they probably expect a function similar in precision and runtime to the > standard tangent function, and nothing which is more accurate in certain > ranges, but which is also slower. > > For being able to use a construct like "y = a * cotan(x)" instead of "y = a / > tan(x)" there should be no additional punishment aside from standard > non-compliance. > Having spent considerable time implementing tanl() that resides in FreeBSD's libm, I suspect no one is going to do the work to ensure the max ULP of cotan is no worse than max ULP of tan.