Michael Platings wrote:
Unless I'm misunderstanding something, this is not an OpenGL
issue. If my scene graph uses double precision, and you write an
FBX file, then I'd expect to get double precision when I load that
FBX file. The FBX plugin should not decide, on the app's behalf,
that it really doesn't need double precision, right?
/Every plugin/ decides, on the app's behalf, that it really doesn't
need double precision.
Just the way you phrased that sentence makes it sound wrong to me. An
API should never assume it knows more about what the app needs than the
app itself.
DAE, OBJ, DXF, 3DS etc all import floats, because that's what OpenGL
renders.
If an app does need double precision then I'm not quite sure what the
correct solution to that is, but an arbitrary option for one
particular plugin isn't the right way to go about it.
Sounds like a lot of plugins need fixing...
Unless I'm mistaken, the OpenGL API has handled double-precision
values for years. More importantly, the latest hardware from ATi
and NVIDIA supports double-precision values as well
The API has support, but all the /implementations/ I knew of last time
I checked converted the doubles to floats in the driver. This may have
changed in the latest greatest hardware although I got the impression
it was directed at OpenCL/CUDA/CTM?
OpenGL too...
http://www.opengl.org/registry/specs/ARB/gpu_shader_fp64.txt
If double precision is generally useful then it's definitely something
to be addressed at a higher level than a plugin option. A config
option like OSG_USE_FLOAT_VERTEX perhaps?
Isn't it possible you might have an app that needs double precision and
a different app that only requires single-precision? Personally, I
wouldn't want to have to keep two copies of OSG around just because it's
not configurable enough at runtime.
--"J"
_______________________________________________
osg-submissions mailing list
[email protected]
http://lists.openscenegraph.org/listinfo.cgi/osg-submissions-openscenegraph.org
