I develop code on Linux-x86/X11/OGL. I was recently
experimenting with FOG, and I do all the development
on a 333Mhz CPU with a Voodoo 3 GPU (which supports
OGL by the XFree86/Mesa condtributions), running
kernel 2.4.18 and XFree86 4.2. The slow CPU/GPU help
be write lightweight code…
When I compiled and tested by codes under other
2.4.x kernel machines with X 4.3/4.3 and other GPUs
(nVidia Quadro, FireGL, ATI Radeon) the FOG behaved
differently. I get the same visual results with MESA
(indirect rendering) and Quadro/FireGL/Radeon, but
different results with the Voodoo…
Has anyone necountered such behaviour? If it is
only a Voodoo driver glitch, should I just forget
about it?
The OpenGL spec leaves a fair amount of flexibility for implementors. The implementation can either interpolate the Z value and recalculate the fog factor per-pixel or it can interpolate the fog factor. The former looks better, but the later is simpler to implement and can be faster. Most hardware does the later, but some may not.
That said, I suspect the difference is just in the implementation of fog on the Voodoo. I seem to remember foging looking quite different on Voodoo chips than on other chips. It could also be bugs in the driver. Could you explain more what the differences were?
Apologies for the lateness of this response and thanks for your responses.
I think it is safe to conclude that it is an issue with the card. Not necessarily
a bug with the drivers or the card, but a difference in the implementation.
However, in my opinion the difference in the implementation is almost
unacceptable.
Here is what I see. The fog saturates a lot quicker for my settings on the
MESA indirect than the Voodoo. In fact, the Voodoo results seem more
realistic in my opinion. I render to parallel rectangles, one above and one
below the observer - they are both textured (MODULATE). The plane above
the observer saturates everywhere in the field of view with MESA. On the
Voodoo DRI, the plane shows a gradual change in fog intensity within
the field of view. I stopped experimenting once I saw the differences.