lilypond-devel
[Top][All Lists]
Advanced

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

Re: Gets vertical skylines from grob stencils (issue 5626052)


From: address@hidden
Subject: Re: Gets vertical skylines from grob stencils (issue 5626052)
Date: Mon, 6 Feb 2012 22:42:57 +0100

On Feb 6, 2012, at 9:41 PM, address@hidden wrote:

> On Feb 6, 2012, at 9:24 PM, Werner LEMBERG wrote:
> 
>>> To all interested parties, this work is now up on:
>>> 
>>> dev/skylines
>> 
>> Thanks.  A minor thing which causes me grief each time I see it:
>> Please configure your editor so that it automatically inserts a
>> newline character at EOF.
>> 
>> 
>>   Werner
> 
> OK.  I've cleaned up all of my cryptic commit messages and pushed a single 
> commit to dev/staging.
> I tried to rebase to remove the whitespace errors but to no avail.  If 
> someone knows how to do cleanup on this, please do!
> 
> Cheers,
> MS
> 

Some stats for Mozart's 2nd horn concerto (all averages).

Without the patch: 47s
With the patch: 1m
of which
16s are added on by the Scheme code
11s are gobbled up alone by (make-curves-for-glyph-and-cache d font name-style 
name)

So the skyline problems are minimal to non-existant.  It's the Scheme code 
that's slowing things down.
This can (should) be solved in two ways.

1) Do all of the calculations in make-curves-for-glyph-and-cache during the 
build of the fonts.  This will likely slow the font build down significantly as 
it'll need to be done on all of emmentaler, but it won't need to be redone 
unless someone changes a glyph.

2) Port a chunk of this code (if not all of this code) to C++.  I have no clue 
what time of time gains this'll bring.  In my naive mind, all of the scm_to_X 
and scm_from_X feels like it adds time because it means more lines of code and 
morphing things in and out of C++, but I'm guessing this has no basis and is 
still in fact faster.

I'll look into doing 1.  Meanwhile, if someone can tell me how to push changes 
to a remote branch (after having created the branch), I'd be much obliged!

Cheers,
MS


reply via email to

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