Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]


Groups > comp.lang.javascript > #29232

Re: Improve speed calculations?

Newsgroups comp.lang.javascript
Date 2016-01-11 07:56 -0800
References <86bb6f49-d74d-426d-bc2a-5b4cd7199bf4@googlegroups.com> <69c96cdd-1020-470c-b0b2-66e2898468bb@googlegroups.com> <87si24w8id.fsf@bsb.me.uk> <5693bf11$0$23861$e4fe514c@news.xs4all.nl> <87fuy4w27e.fsf@bsb.me.uk>
Message-ID <58985c2c-1555-448d-9180-e513da33a0fc@googlegroups.com> (permalink)
Subject Re: Improve speed calculations?
From jonas.thornvall@gmail.com

Show all headers | View raw


Den måndag 11 januari 2016 kl. 16:44:32 UTC+1 skrev Ben Bacarisse:
> Silvio <silvio@internet.com> writes:
> 
> > On 01/11/2016 02:28 PM, Ben Bacarisse wrote:
> >> jonas.thornvall@gmail.com writes:
> >>
> >>> Den söndag 10 januari 2016 kl. 14:33:27 UTC+1 skrev jonas.t...@gmail.com:
> >>>> <script language="Javascript">
> >> <snip code>
> >>> http://jt.node365.se/BAUTABASE.html
> >>>
> >>> Using the biggest base possible adding "4503599627370495" encoding
> >>> fibonacci in javascript without overflow.
> >>>
> >>> Also fixed ambiguity reading out the numbers in bases over ten by
> >>> using separator for digitplaces. The add can not cause overflow but to
> >>> get rid of base overflow would require some real deep thinking.
> >>>
> >>> The 100000th Fibonacci took 34 seconds i do not think i can get it
> >>> much faster in javascript if your a wizard please prove me wrong.
> >>
> >> I have not tried your code, but I've just written a quick bignum add
> >> function and I can get fib(100000) in about 3.5 seconds.  If I have time
> >> to unravel yours, I'll do a comparative test.  Note that a language with
> >> a good bignum type (Haskell, for example) will do it in less than 100th
> >> of a second.
> >>
> >> Note that your 'challenge' makes it easy to optimise the code.  For
> >> example, using a power of 10 as the base makes the conversion for
> >> printing very simple.
> >>
> >
> > That is well neigh impossible if you used a 10-based decimal since
> > Jonas uses a huge base and is convinced that using a larger base will
> > make computations much faster. Your result would invalidate that
> > claim...
> 
> Using base 10 is quite slow (my code needs 23s for fib(100000) using
> base 10), but I use a higher power of 10 as the base for speed whilst
> keeping printing simple.  With higher powers, you do get faster results
> (up to a limit, of course!).
> 
> -- 
> Ben.

What timing for the 1 million th Ben?

Back to comp.lang.javascript | Previous | NextPrevious in thread | Next in thread | Find similar | Unroll thread


Thread

Improve speed calculations? jonas.thornvall@gmail.com - 2016-01-10 05:33 -0800
  Re: Improve speed calculations? jonas.thornvall@gmail.com - 2016-01-11 03:19 -0800
    Re: Improve speed calculations? jonas.thornvall@gmail.com - 2016-01-11 04:41 -0800
    Re: Improve speed calculations? Ben Bacarisse <ben.usenet@bsb.me.uk> - 2016-01-11 13:28 +0000
      Re: Improve speed calculations? Silvio <silvio@internet.com> - 2016-01-11 15:41 +0100
        Re: Improve speed calculations? Ben Bacarisse <ben.usenet@bsb.me.uk> - 2016-01-11 15:44 +0000
          Re: Improve speed calculations? jonas.thornvall@gmail.com - 2016-01-11 07:53 -0800
          Re: Improve speed calculations? jonas.thornvall@gmail.com - 2016-01-11 07:56 -0800
          Re: Improve speed calculations? jonas.thornvall@gmail.com - 2016-01-11 08:28 -0800
      Re: Improve speed calculations? jonas.thornvall@gmail.com - 2016-01-11 06:48 -0800

csiph-web