Just as an experiment, I used org.j3d.ui.ImageCaptureCanvas3D to
"capture" TextureTest from the java3d demos at 1024x768. I actually
had two versions, one where I registered a CapturedImageObserver,
and one where I didn't. Comparing the profiles of the two showed
that the cpu cost of the postSwap() processing for image capture
was a bottleneck, with about 18.6s out of 26.3s spent in postSwap
and its underlying routines. Of this 18.6s, 11.4s were in
glreadPixels() i.e. inside OpenGL.
The environment I used was 2 x 1015MHz UltraSPARC-IIICu, j2sdk 1.4.2,
java3d 1.3.1, and jvm options "-server -Xms256m -Xmx256m". I'd have to
think about how I could do the same test on my Linux box at home since
my favorite profiling tool isn't available for it, but worst case I
guess I could use interpose routines to timestamp the OpenGL layer.
--
============================================================================
,-_|\ Richard Smith - SE Melbourne
/ \ Sun Microsystems Australia Phone : +61 3 9869 6200
[EMAIL PROTECTED] Direct : +61 3 9869 6224
\_,-._/ 476 St Kilda Road Fax : +61 3 9869 6290
v Melbourne Vic 3004 Australia
===========================================================================
===========================================================================
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".
