Hi Kelvin, now thats interesting. So in consequence i have to find my own way around this. Now my remaining question is, what would you suggest? Using locales (maybe generating appropriate ones) and switch between them, to guarantee not to exceed a certain distance to the origin? (I fear this solution, as i heard other people having huge problems with get this to work, and so maybe causing even worse problems then i have now.) Or could you think about another workaround/solution i could adapt?
-emanuel > Emanuel Brzostowski wrote: > > Hi Artur > > Ok, just found a statement in this group (forgot where), telling me, J3D > is > > using floats internally. So the question arises, why i can use doubles > nearly > > everywhere (except for the reason of decreasing performance)? > > Wouldn't it make sense, to let J3D use doubles internally (everywhere), > and > > converting them to floats only when nessecary (i.e. when passing the > data to > > DirectX or OpenGL)?Imho this would drastically increase the accuracy of > > transformations (and maybe, as well decreasing performance :( ). > > I admit that i don't know the internals of J3D as good as many other > people > > in this list, but i would like to know, the hows and whys (concerning > the > > float vs. double problem), or at least a pointer to some information > about this > > topic. > > > > -emanuel > > > > Under Solaris (and Sun's brand of graphics card) both the test > program from you and in bug report 476087 works fine for big float. > > Note that the precision pass down to the Opengl call > is the same (JDK use same precision under windows and solaris). > > This is driver/card floating point precision problem > which is not pixel exact when vertices of big float > pass down. Probably it is loss in the T&L engine of the > pipeline. > > That's why game card is not good for 3D CAD design in workstation. > Check out, > http://www.3dlabs.com/product/technology/wildcat_quality.htm#sub-pixel > > - Kelvin > ------------- > Java 3D Team > Sun Microsystems Inc. > > > > > > >>Emanuel Brzostowski wrote: > >> > >> > >>>I just can't believe, precision of double is'nt sufficient for the > >> > >>given > >> > >>>testcase (but obviously i'm wrong). > >> > >>Float, not double. There are many places internally where java3d uses > >>floats, converting from doubles if needed. And precision of float is not > >>sufficient for your testcase. > >> > >>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". > >> > > > > > > -- > > +++ GMX - Mail, Messaging & more http://www.gmx.net +++ > > NEU: Mit GMX ins Internet. Rund um die Uhr f�r 1 ct/ Min. surfen! > > > > > =========================================================================== > > 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". > -- +++ GMX - Mail, Messaging & more http://www.gmx.net +++ NEU: Mit GMX ins Internet. Rund um die Uhr f�r 1 ct/ Min. surfen! =========================================================================== 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".
