Excellent! Thanks for the answer. I am very impressed with the amount of support given here.
Jacob Marner ----- Original Message ----- From: "Chien Yang" <[EMAIL PROTECTED]> To: <[EMAIL PROTECTED]> Sent: Wednesday, February 13, 2002 20:00 Subject: Re: [JAVA3D] Three performance questions > Jacob, > > > > > Hi, > > I have a few performance questions > > > > (I am fairly new to Java3D but has used OpenGL for quite some time) > > > > 1. > > --------------- > > There seem to be some people that think that java3D immediate > > mode is faster than compiled retained mode. However, in the > > Sun java3d faq it says the opposite. So which is fastest? (Disregarding > > ease of use) > > > > In general, compiled-retained mode is the fastest. User should only > use immediate mode if needed, such as have a better control of > the rendering. Such flexibility usually comes at the expense of > performance. > > > 2. > > --------------- > > > > It seems that everybody believes that the placement of viewer > > should be controlled in the transform just above the view platform. > > > > However, doing so to look in a certain direction using lookAt > > requires a matrix inversion (v is the Transform3D class > > above the Viewplatform): > > > > v.lookAt(cameraPos, endPoint, upVector); > > v.invert(); > > > > because of the way the modelview matrix is calculated. > > (See the Java3D 1.3 beta 1 spec pages 294 and 295). > > > > If I placed the same transform at the top of the scene graph > > and dropped the transform above the viewplatform I > > could avoid this invertion. > > > > Is there any way I can avoid the matrix invert while still > > placing the view transform as recommended? I would > > rather like to avoid compatibility mode since it is not > > recommended for java3D applications but just > > included to make porting easier. > > The invertion to the view's transform, is a once per frame cost. > The transform change above the scene is a per object cost (including > TG, SG, S3D ... ), and such operation can be very expensive. > > > > > 3. > > ----------------- > > Does the scene graph compiler in Java3D 1.3 beta 1 merge > > Shape3D objects assuming: > > 1. They have the same Appearance object > > 2. They are transformed the same way, i.e. are > > in the same branch of the scene graph (there is some transforms > > but these are eliminated because no capability bits are set in them > > nor below them in the tree) > > Yes. > > > > > > I think I read somewhere that Shape3D objects have an > > overhead each - if they are not combined automatically, > > does this mean I should merge them manually for speed? > > Unless that's the bottleneck. It's best to do some benchmark/ > analysis first. > > - Chien Yang > Java 3D Team. > > =========================================================================== > 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".
