OpenEuphoria

Re: Core Euphoria Functions

• Posted by George Orr <gorr at w?h.rr.?om> Jan 24, 2008
• 683 views

CChris wrote:
> 
> Robert Craig wrote:
> > 
> > George Orr wrote:
> > > To Rob and company - 
> > > 
> > > Is there a reason that we have the natural logarithm log(x) as a core
> > > Euphoria
> > > function but do not have exp(x)?  I know I can get the equivalent using
> > > power(x,y)
> > > provided I remember the value of e to enough significant places.
> > 
> > I didn't like to add lots of very simple global functions that
> > the programmer could create for himself, probably in 
> > less time than it would take him to browse through the manual.
> > 
> > We have "PI" in misc.e.
> > We could add "e" as well,
> > but I bet that would cause a lot of naming conflicts.
> > 
> > Regards,
> >    Rob Craig
> >    Rapid Deployment Software
> >    <a href="http://www.RapidEuphoria.com">http://www.RapidEuphoria.com</a>
> 
> For largish values of x (around 12 iirc, I don't have my test files at hand),
> power(E,x) starts drifting from exp(x), even with the most accurate value of
> E as double..
> Introducing exp() would have the side benefit of not needing to define E -
> because
> of the likely name clashes -, since exp(1.0) would work well.
> 
> CChris

I tried this and didn't really see any difference until x is larger than 20 and
insignificant differences afterwards.
For x = 30 difference between the power(x,y) and libm exp(x) solution was less
than 0.02 in ten trillion.  Of course, libm could be unstable in this range as
well!

I don't really see a problem in accuracy, although I would prefer to have exp(x)
as a core Euphoria function.

George