Lee,
Yes, most of our images have been resized to powers of two before being
imported. It isn't a "distortion" corruption the Voodoo 3 is doing, the
Voodoo is chewing the images up - some look like repeat tiling. It's as
if the Voodoo 3 sees that the image doesn't fill the space and tries to
tile it to fit the space.
I imagine we might be able to play around with the filters on either the
Voodoo driver or in Java 3D. But I'm not attempting to make it look
right on my machine. In fact I only built up that machine as a testbed
specifically to test with the Voodoo 3 since it was a popular video card
(and performance is ok). My concern and my reporting is so that the
community is aware of these issues. To produce a mass distributed
application I don't believe we should expect end users to tweak their
video settings!
And as a software developer I try to avoid coding specifically around an
odd compatibility problem (after all the other video cards handle it
just fine). I don't intend to waste my time finding "fixes" when the
next version of Java 3D or Voodoo drivers might change the behavior
again. I'd rather see 3dfx fix their product.
- John
"J. Lee Dixon" wrote:
>
> John,
> It just seems odd because by the time it gets to the voodoo it is power
> of 2. Have you considered resizing the image to power of 2 by hand (ala
> Photoshop or other), and then running it? If you have a (176x256)
> texture, from looking at the "TextureLoader" code it would be resized to
> the "nearest" power of 2: 128x256. Then you could see what the texture
> looks like before the voodoo gets hold of it.
>
> I sincerely would like to see better results from the Voodoo3 (I have
> one at home), but the 256x256 limit on the chipset is killing me anyway.
> Perhaps the problem is just a default minification/magnification filter
> selection for the chipset. Don't they have some interface for changing
> those in the display properties?
>
> -Lee
>
> -----Original Message-----
> From: John Wright [mailto:[EMAIL PROTECTED]]
> Sent: Monday, November 20, 2000 8:24 AM
> To: [EMAIL PROTECTED]
> Subject: Re: [JAVA3D] Performance Report 1.2.1 beta 1 and Voodoo 3 v1.06
> drivers
>
> Yes Lee, Java 3D by specs requires images that are powers of two. As it
> was explained to me the TextureLoader does this for us (assuming you use
> it and don't load the textures yourself). None of our odd shaped
> textures give a GeForce, TNT, Matrox or software rendering any trouble -
> it's only the Voodoo 3 that corrupts the textures (much better now with
> the 1.06 drivers).
>
> - John
>
> "J. Lee Dixon" wrote:
> >
> > I assume that you realize of course that textures MUST be power of 2
> for
> > Java3D, and that somewhere in the texture loading the image is being
> > resized to power of 2 (or is there something new in the current Beta
> > release I don't know about?)
> >
> > -Lee
> >
> > -----Original Message-----
> > From: John Wright [mailto:[EMAIL PROTECTED]]
> > Sent: Sunday, November 19, 2000 10:52 AM
> > To: [EMAIL PROTECTED]
> > Subject: [JAVA3D] Performance Report 1.2.1 beta 1 and Voodoo 3 v1.06
> > drivers
> >
> > Our application is achieving almost a 50% performance increase using
> > Java 3D 1.2.1 beta 1 (OpenGL) (on a GeForce video card system). Frame
> > rates in different areas of our scene went from 22 fps to 29 fps and
> > from 55 fps to 88 fps.
> >
> > On one low end machine (AMD K6-200, 32 Meg RAM, 3dfx Voodoo 3, Win98
> SE,
> > DirectX 8.0) the new Java 3D 1.2.1 beta 1 SDK installer is failing,
> > otherwise the new installation routines are very nice!
> >
> > Also I've tested the v1.06 drivers for the Voodoo 3 and it appears
> 3dfx
> > has improved the texture issues. Textures that are powers of two
> > (256x256) now appear to display perfectly. Textures with odd sizes
> > (176x256) are still displaying corruption.
> >
> > - John Wright
> > Starfire Research
> >
> >
> ========================================================================
> > ===
> > 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".
>
> ========================================================================
> ===
> 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".
===========================================================================
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".
