Re: Why myself building lilypond much slower than official release?

[Jonas Hahnfeld]

Am Montag, den 05.10.2020, 20:25 +0800 schrieb K.L.:
> Greetings.
> 
> I tried building the lilypond source code from the master branch latest 
> revision (18 Sep 2020) from this gitlab repository, and that's accomplished 
> successfully.
> But when I try to engrave a medium size ly file with myself built lilypond, 
> it costs 35s, as compared, the official lilypond of 2.20.0 costs only about 
> 11s.
> 
> I noticed the latest code depends guile 2.2 and the old version depends guile 
> 1.8, can that cause the time cost difference?

A slight correction: The latest code doesn't depend on Guile 2.2, but
may use it if it cannot find Guile 1.8. That said, yes this incurs a
possibly high penalty during runtime and probably explains the
difference.

To make the configure script find your Guile 1.8 interpreter, consider
setting PKG_CONFIG_PATH to the directory containing guile-1.8.pc or use
the old style of pointing GUILE_CONFIG to the right guile-config
(discouraged).

> I followed every step in here, and I got a huge lilypond binary file 
> (104.3MB), is it a debug building? Did I miss something?

The default configuration produces a release binary (with respect to
compiler optimizations), but includes debugging symbols. You can turn
that off (don't remember the option right now) or run strip on the
binary. In any case, this shouldn't be a problem.

> For my work, the runtime efficiency is critical. Thanks for any help.

If 2.21 is still slower than 2.20, please let us know - it might be a
regression.

Hope this helps!Jonas

signature.asc
Description: This is a digitally signed message part