Re: Lilypond performance

[Urs Liska]

Am 07.08.2010 09:14, schrieb David Kastrup:
> Graham Percival<address@hidden>  writes:
>
>    
> > On Thu, Aug 05, 2010 at 05:26:15AM -0700, ornello wrote:
> >      
> > > In my installation, Lilypond runtime seems to increase exponentially (at
> > > least not linearly) with the number of pages to engrave. Is there any option
> > > to speed up Lilypond, e.g. by removing time-consuming engravers, such that
> > > the performance only increases (almost) linearly with the number of pages?
> > >        
> >
> > There's a section in the Learning manual called "speeding up
> > lilypond", or something like that.  This sounds like a good place
> > to look.
> >      
> Essentially exponential behavior can't really be cured with anything
> except fixing the algorithm.
>
>    
Can one expect a program like LilyPond to work in a linear fashion?
IIUC a linear increase could only be expected if lily would just "walk" 
through the score from beginning to end.
But if it has to process the score as a whole (i.e. "still know at the 
end what was at the beginning" or "know the end before typesetting the 
first barline") I find it quite natural that its usage of processing 
time and probably memory increases exponentially.

While practically any software may be improved and optimized I think 
that if the scores become too complex to be handled within an acceptable 
amount of time the only real solution is new hardware.

Best
Urs