"Phil Holmes" <address@hidden> writes:
From: "Knut Petersen" <address@hidden>
About 4 times slower than your "Ivy Bridge" system.
Half of that can be accounted to the doubled clock rate.
I suspect that a lot of the rest is caused by general
improvements in cpu architecture. So lilypond seems
not to scale well with the number of available cores.
Generally, it does. Read
http://lilypond.org/doc/v2.17/Documentation/contributor/saving-time-with-the-_002dj-option
Not really. That just helps when doing _multiple_ files at once. If
you use -j4, LilyPond initializes itself, splits the file list into 4
parts, then splits itself into 5 copies. 4 of those copies process
their file list independently, and the fifth copy reports when all them
have finished.
You save the time of initialization for three of those jobs. But that's
a rather small amount. Other than that, there is no advantage to
splitting the job list yourself and calling parallel copies of LilyPond.
It would be somewhat better if the fifth copy of LilyPond performed the
actual task of a job server, telling each running copy what file to
process next and thus keeping all of them active until the end. In
practice, I doubt it would make much of a difference. You can try
experimenting with GNU parallel if you want to try this.
--
David Kastrup