What you say is true, but what the benchmark tests is SIN, COS, an other
math functions that are implemented in native code. Aparently, Sun has
switched to a much slower math library (much stricter?) for 1.4. It
would seem that any Java3D program that used these types of functions
would suffer like the benchmark did. The IBM 1.4 is also significatly
slower than 1.3.1. Hotspot won't have any impact on these math
functions because they are implemented as native code.
The only thing that may be good is that other parts of 1.4 may indeed be
faster than 1.3.1, but it is very disturbing to see such the huge
performance degredation exhibited by the benchmark.
Even so, I'm running 1.4 because it's got all the XML stuff built in :)
And it's less of a headache getting https to run with it.
Mike
Alessandro Borges wrote:
As far I know, the greatest impact in performance in Java3D is the video
card, not the JVM.
You have near equal fps when you run your java3D apps using HotSpot JVM or
the not optimized-interpreted-only Classic JVM. And sometimes the Classic
runs faster (!), becouse it doesnot loose time in useless
compilations/optimizations. But be carefull, it is not a rule.
To check it yourself, edit you JAVA1.3's JVM.CFG, changing the order of JVM
and run your Java3D app with a FPS meter, as Fraps.
The file JVM.cfg is at your JRE/lib folder.
And benchs about JVM are trick. The JVM needs a REAL hotspot to optimize it,
and then it will run at lightspeed. Otherwise it will run just as a ordinary
code, not fully optimized, becouse it is not needed optimizations ;-).
Alessandro
From: "Michael P. McCutcheon"
Subject: [JAVA3D] Alma bench and it effects on Java3D performance
If you goto:
http://www.coyotegulch.com/reviews/almabench.html
You will see a math benchmark where the latest offering from Sun,
1.4.1_01, gets absolutly slaughtered by 1.3.1_06. Runtimes for 1.4 are
TWICE that of 1.3.
Surely this has got to effect the performance of Java3D, since it does
so much math.
Can anyone validate this? Is 1.3.1_06 still the preferred VM for Java3D?
I find it very disturbing that the latest version of Java takes twice as
long to execute a simple benchmark. Will 1.4.2 change this?
Mike
_______________________________________________________________________
Busca Yahoo!
O melhor lugar para encontrar tudo o que voc� procura na Internet
http://br.busca.yahoo.com/
===========================================================================
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".
===========================================================================
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".
