On Sunday 11 December 2005 23:26, Chris Cannam wrote:
> On Sunday 11 Dec 2005 21:50, Pedro Lopez-Cabanillas wrote:
> > HEAD - No Jack
> > 1: 8.17
> > [...]
> > HEAD - With Jack
> > 1: 21.22
>
> Wow.  These figures are something of an eye-opener.

You asked only for startup time comparisions between HEAD and experiments 
branch, but not for measurements near the typical user conditions.

I've compiled both branches with debug=full, and all optional libraries and  
features were present at compile and run time (LIRC, DSSI, LADSPA, lilypond, 
project packager...). No optimizations were made by GCC because debug=full, 
and lots of debugging information going to the standard output. I had 
installed lots of LADSPA plugins (161 of them) and some DSSI synths, which 
made the startup time even worse. And finally, 14 MIDI devices (7 ALSA seq. 
clients, duplex) also contributing a bit to the startup time. 

This is not a race, but I can improve things a lot in both branches:

branch=experiments; no Jack; debug=full; 161 ladspa plugins
startup times: 5.83, 4.54, 3.58, 3.68, 4.39

branch=experiments; with jack; debug=full; 161 ladspa plugins
startup times: 6.19, 6.47, 6.45, 6.70, 6.41

branch=experiments; no Jack; no debug; 0 ladspa plugins
startup times: 4.10, 3.70, 2.99, 3.69, 3.13

branch=experiments; with Jack; no debug; 0 ladspa plugins
startup times: 4.21, 4.24, 4.19, 4.73, 4.57



branch=HEAD; no Jack; debug=full; 0 ladspa plugins
startup times: 5.56, 5.04, 5.37, 5.17, 5.47

branch=HEAD; with Jack; debug=full; 0 ladspa plugins
startup times: 10.10, 12.52, 12.97, 10.89, 12.14

branch=HEAD; no Jack; no debug; 0 ladspa plugins
startup times: 3.65, 3.86, 4.07, 3.96, 4.10

branch=HEAD; with Jack; no debug; 0 ladspa plugins
startup times: 4.36; 5.06; 5.04; 4.94; 4.67


> > experiments - With Jack
> > 1: 7.14
>
> I wonder what on earth could have made that 21 -> 7 second difference?
> That _really_ wasn't in the script.

Tou mean this changeset, right?

Modified Files:
      Tag: experiments
        SConscript audiopluginmanager.cpp audiopluginmanager.h 
        audiosynthmanager.cpp rosegardengui.cpp rosegardengui.h 
        sequencemanager.cpp sequencemanager.h 
Added Files:
      Tag: experiments
        startuptester.cpp startuptester.h 
Log Message:
 * various changes intended to speed up (perceived) startup time, mostly
   by shunting off some of the startup tests into a separate thread to be
   run after the UI proper is visible.

I think that the key change between jack and non-jack startup times is in 
AudioPuginManager::Enumerator, working in a separate thread. Without LADSPA 
plugins the average startup time for HEAD/debug fall from 16.9s to 11.7s, but 
compiling with optimization gives the better results (avg=4.8s) in this 
branch. The experiments branch gives better startup times, specially  when 
running in the worse conditions (lots of plugins, busy CPU, lots of debug 
output).

Regards,
Pedro


-------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc. Do you grep through log files
for problems?  Stop!  Download the new AJAX search engine that makes
searching your log files as easy as surfing the  web.  DOWNLOAD SPLUNK!
http://ads.osdn.com/?ad_idv37&alloc_id865&op=click
_______________________________________________
Rosegarden-devel mailing list
[email protected] - use the link below to unsubscribe
https://lists.sourceforge.net/lists/listinfo/rosegarden-devel

Reply via email to