lilypond-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: LilyPond speed. Idea for Schikkers list


From: Janek Warchoł
Subject: Re: LilyPond speed. Idea for Schikkers list
Date: Fri, 11 Mar 2011 23:24:27 +0100

2011/3/10 Graham Percival <address@hidden>:
> On Thu, Mar 10, 2011 at 12:42:11AM +0100, Janek Warchoł wrote:
>> 2011/3/9 Han-Wen Nienhuys <address@hidden>
>> > You can easily try this out yourself. Just set ragged-right=##t and
>> > set linelength to 100 meters (or something).  You may need to remove
>> > some internal sanity checks on distances.
>>
>> Maybe i should've explained why i ask this question. My idea for
>> shikkers list is this: it would show a user only a rough preview of
>> music,
>
> This has been proposed a few times in the past, but nobody has yet
> seriously investigated it.
>
> I think the first step would be to do exactly what Han-Wen
> suggests -- hey, it'll only take 5 minutes!

I did it.

> (assuming that you
> choose a very large score to test with?)

I tried one horn concerto from mutopia (45 pages, ~300 measures).
Results are interesting: compiling it with regular settings took 2
minutes. After adding

\paper
{
  ragged-last-bottom = ##t
  ragged-bottom = ##t
  ragged-right = ##t
  paper-width = 1000 \cm
}

It compiles for 15 minutes now and still didn't finish.
2 days ago I tried much smaller score (4 pages) and normally it
compiled in 12 seconds, while using this 'infinite' paper size it took
about twice as long.
Perhaps something in line breaking algorythm isn't working as we expect at all?

> If you see a
> significant increase, then clearly it's worth investigating more.
> But that other investigation will probably require modifying the
> C++ code.

I would see a significant increase only if calculating line-breaks was
a significant part of score compiling (computational-wise), but right
from the start I didn't suppose that was the case.
I hoped that some experienced developer could give an estimate based
on intuition, but maybe that's impossible.
As for me this idea is postponed indefinately then.

cheers,
Janek



reply via email to

[Prev in Thread] Current Thread [Next in Thread]