Nitin.Jain wrote:
> Hi Doug,
>
> I'm attaching the source code where I'm creating 1000 sphere and it
takes
> around 7 secs to render it. If I increase this number to 15000 it
takes 100+
> secs, however in both the cases initial number of spheres visible on the
> screen are same.
First - render is maybe misleading here. You take 7 seconds from
starting up to displaying window. After that it refreshes in realtime
(30+fps), at least on my Geforce 256.
On my system, it is not 7 second, but around 3.5. About 1 second is
spent in BranchGroup creation (which is possible to understand, given
number of objects there). Around 0.5 in compilation. Another 1s in
adding stuff to universe. Around 0.5 in two runs of full gc after that
(I suppose that Sphere stuff creates a lot of temporary objects - you
might want to use on of j3d.org spheres to reduce usage). And then
another 0.5 second to get window to appear (OS stuff I suppose, plus
some class loading).
There is no clear slowdown here. You could optimize some parts and gain
maybe 10-50%, but you will not go under certain time - awt,opengl and
java3d take some time to initialize. After first initialization,
rendering is real time.
With 5000 spheres, entire init took 10s.
Part of problem is live memory - 5000 shapes, transform groups etc have
to be set up, transformed etc, but this is clear. But when looking at
traces I stumbled at something really worrying:
percent live alloc'ed stack class
rank self accum bytes objs bytes objs trace name
94 0.11% 87.51% 20016 1 50080000 5000 10593
javax.media.j3d.CachedTargets
95 0.11% 87.62% 20016 1 50079984 4999 10601 <Unknown_class>
96 0.11% 87.73% 20016 1 50079984 4999 10600
javax.media.j3d.HashKey
97 0.11% 87.85% 20016 1 50079984 4999 10602
javax.media.j3d.CachedTargets
98 0.11% 87.96% 20016 1 50079984 4999 10599 <Unknown_class>
Problem is not amount of live data, but amount of data which got
allocated and collected. 50MB !!! for each trace. This gives 250MB of
garbage to be collected. But looking at verbose gc output, about 2-4s of
these 10s where spend in gc. This is much. But something looks wrong
here - is it possible for non-array object to take up 10-20k ? Anyway,
output from verbose gc confirms that there is 200+MB of garbage
collected, with about 20MB of live data staying.
All traces point around
javax.media.j3d.SharedGroupRetained.setNodeData(SharedGroupRetained.java:167)
javax.media.j3d.NodeRetained.doSetLive(NodeRetained.java:621)
javax.media.j3d.GroupRetained.doSetLive(GroupRetained.java:2101)
javax.media.j3d.GroupRetained.setLive(GroupRetained.java:2075)
javax.media.j3d.SharedGroupRetained.setLive(SharedGroupRetained.java:296)
javax.media.j3d.LinkRetained.setLive(LinkRetained.java:159)
javax.media.j3d.GroupRetained.childDoSetLive(GroupRetained.java:2083)
javax.media.j3d.GroupRetained.doSetLive(GroupRetained.java:2132)
javax.media.j3d.TransformGroupRetained.setLive(TransformGroupRetained.java:543)
javax.media.j3d.GroupRetained.childDoSetLive(GroupRetained.java:2083)
javax.media.j3d.GroupRetained.doSetLive(GroupRetained.java:2132)
javax.media.j3d.TransformGroupRetained.setLive(TransformGroupRetained.java:543)
javax.media.j3d.GroupRetained.childDoSetLive(GroupRetained.java:2083)
javax.media.j3d.GroupRetained.doSetLive(GroupRetained.java:2132)
javax.media.j3d.BranchGroupRetained.setLive(BranchGroupRetained.java:161)
javax.media.j3d.Locale.doAddBranchGraph(Locale.java:206)
javax.media.j3d.Locale.addBranchGraph(Locale.java:163)
com.sun.j3d.utils.universe.SimpleUniverse.addBranchGraph(SimpleUniverse.java:356)
TestApp.<init>(TestApp.java:86)
TestApp.main(TestApp.java:121)
Is this normal ?
Artur
===========================================================================
To unsubscribe, send email to [EMAIL PROTECTED] and include in the body
of the message "signoff JAVA3D-INTEREST". For general help, send email to
[EMAIL PROTECTED] and include in the body of the message "help".
