[Pcgen-trackers] [ pcgen-Bugs-1036184 ] Performance degradation with 5.7.x?

[SourceForge.net <no...@so...>]

Bugs item #1036184, was opened at 2004-09-28 10:04
Message generated for change (Comment added) made by kingpaul
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=384719&aid=1036184&group_id=25576

Category: None
Group: None
Status: Open
Resolution: None
Priority: 9
Submitted By: Paul W. King (kingpaul)
Assigned to: Devon Jones (soulcatcher)
Summary: Performance degradation with 5.7.x?

Initial Comment:
Per 
http://groups.yahoo.com/group/pcgen/message/79925

> > Greetings!
> >
> > I recently moved from PcGen 5.6.1 to 5.7.5, then
> > 5.7.6. I've noticed a really impressive drop in
> > performance ... :-( I've tried a handful of machines
> > from a 900MHz, 512 MB RAM Windows 2000 system
> > to a 3.0GHz, 1GB RAM Windows XP machine. All of
> > them exhibited very slow performance. Loading
> > characters, saving characters, even switching
> > between tabs for the same character all result in
> > unresponsive behavior. Any ideas? Is there any data
> > I can collect that might prove useful?
>
> There are 2 places that are currently known to cause
> slow performance:
>
> 1 - If your character has lots of spells
>
> 2 - If your character has lots of equipment
>
> I have also seen that the new Preview tab seems to
> slow things down as well, but I can not prove that it is
> the Preview tab.
>
> --
> regards,
> Frugal
> -OS Chimp

Both from observation (using PCGen), other users' 
comments, and what I can infer from developers' 
comments over the past few months, it seems that 
PCGen is generating *lots* and *LOTS* of objects 
related to spells and equipment. (Skills, too, I think; but 
that list is almost always shorter, so the problem isn't 
quite as pronounced.) It seems that these objects are 
being created and destroyed faster than the memory 
manager (trash) can consume them and return the 
memory to the pool. So PCGen starts swapping like mad, 
and grinds to a crawl.

Because it seems to be related to *all* lists, I infer that 
the problem is either in a) a common piece of code used 
by all lists, or b) a piece of similar code which was 
copy/pasted into each of the list handlers. At the time 
the list handler was written, it might have seemed easier 
to create all these objects (and then immediately 
destroy them), but in reality it's creating a *huge* 
performance problem.

For several versions now, the response to "It's slow!" 
has been "Tell your machine to give more RAM to 
PCGen." I'm sorry, but that's not a solution. At best, it's 
a work-around, and should be seen as such. 64 Mb (the 
default for the JRE) should be plenty for a "relatively 
small" database application like PCGen. The fact that 
we're having to give it 1/4 Gb in order to run smoothly 
should be *screaming* "problem!" to someone.

Question: is someone *currently* looking into 
this "runaway object creation / memory consumption" 
problem? If so, what have you found? If not, is there 
someone on the team who could? Oh benevolent code 
monkeys, is it possible to make this a priority for the 
next target release (instead of more features)?

===

Paul W. King
TM SB, OGL/PL Chimp, Data Tamarin, BoD

----------------------------------------------------------------------

>Comment By: Paul W. King (kingpaul)
Date: 2004-09-28 10:55

Message:
Logged In: YES 
user_id=277877

Per http://groups.yahoo.com/group/pcgen/message/79931

Also, as soon as a character is opened, the console starts 
spitting out messages like:

java.awt.event.InvocationEvent
[INVOCATION_DEFAULT,runnable=pcgen.util.
SwingWorker$1@1385ce,notifier=null,catchExceptions=false,
when=1096381642205] on sun.awt.windows.WToolkit@43caa4

whenever you change focus to another window and the come 
back to PCGen.

This is with PCGen 5.7.5. on XP SP2, with Java(TM) 2 Runtime
Environment, Standard Edition (build 1.4.2_05-b04) as well as 
on Windows 2000 SP4 with same Java version.

Thomas/Chipoulou

===

Paul W. King
TM SB, OGL/PL Chimp, Data Tamarin, BoD

----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=384719&aid=1036184&group_id=25576