Jason Goh wrote:
>
> Hi again,
>
> I am wondering if the framerate of an OpenGL scene in a java applet is
> comparatively slower than one using C++?
> This is a perception i have while doing the programming for a web-enabled
> OpenGL scene. :(
>
> Jason.
Well - it depends :)
On MS-Windows machines:
MS-IE's JVM is much faster than Netscapes JVM :-(
The resulting speed of MS-IE looks fine for me
(play with the "use repaint" and "sleep" buttons on
e.g. morph3D applet - remind, the default
(using repaint & sleep) is safe for most JVM's ...
but you can play with and customize it for diff. JVM/OS).
Unix:Netscape's JVM is also quiet fast
May be the Java2 PlugIn for Netscape is faster ?!
I tested:
Linux 2.3.99 pre6, XFree86 4.0 DRI (Vodoo3, 16bpp)
gl-demo: gears
native : 75-79 fps
java2 & javacomp (inprise) (no repaint, no sleep -> max-fps): 78 fps
Netscape (no repaint, no sleep -> max-fps): 58 fps
So - the results looks acceptable ? YES !
AND the responsivenes of the native application
(cursor keys to rotate) are not that nice, the java one is !
(Remember that the rotation-speed, is set slower in the java app.
This is not the fps render speed)
It depends on your OpenGL hardware acceleration also, of course.
And - of course - Java IS slower than native code !
Well, all demos we have for GL4Java runs well with hardware
acceleration,
even the GLLandscapeWorld stuff, etc.
Also: The usage of display lists is nearly as fast as the native usage !
Just the data generation/manipulation in java is, of course, slower
than the native counterpart !
But you are right, we should create a fps measurement
- and a kind of performance user guide and performance comparison
between native and all the different java vms (incl. OS).
So we can make a statement what we can expect !
Yours, Sven
--
mailto:sgo...@ja...
www : http://www.jausoft.com
voice : +49-521-2399440, +49-170-2115963; fax: +49-521-2399442
|
