I don't disagree. But there are other, faster, ways of doing it using tables of coefficients (if I again remember correctly from my days of disassembling BASIC interpreters). It all depends what the ultimate goal is at the end of the day. In some instances I still code things from first principles that will work, and the algorithms are fully documented in plenty of textbooks. If I then require them to be optimised further, I perform the optimisation - but leave the original code in (either as commentary as to what the optimised code should still produce as answers, or conditionally compiled or assembled so I can switch around for test and comparison purposes). Dave On Fri, 30 Sept 2022, 13:04 Michal Pleban, <lists_at_michau.name> wrote: > David Roberts wrote on 29.09.2022 22:28: > > > From what I remember, any continuous function can be expressed as a > > power series. This should, therefore, also be possible for LN and LOG as > > well as the trigonometric and hyperbolic functions etc. > > That is true, but the problem is how many steps you may need to > approximate the function with desired accuracy. For SIN(X), which always > returns a value from the range [-1; 1] it's around a dozen steps; for > unbounded functions like LOG(X) you may need dozens or hundreds of steps > which would render this method unpractical. > Regards,Michau. > > > > >Received on 2022-09-30 18:00:07
Archive generated by hypermail 2.3.0.