From: Bryan Bush (bbush++at++sgi.com)
Date: 05/18/2005 12:22:17
performer cliptexture emulation for Prism
is it functional, especially for very deep cliptextures?
how do programmable vertex and pixel shaders impact their use? Will
these two features be mutually exclusive, or will it be possible to
have user defined vertex and fragment shaders that tap multiple
cliptextures and procedurally combine them to shade the fragment,
with some kind of encapsulation of whatever magic needs to happen to
the texture coordinates/texture lookups?
vertex shader displacement cliptextures
will there be support for floating point cliptextures tappable by
vertex shaders? (I would think this is free since cliptextures
dont's seem to care much about texel format/type.)
what is the status of OpenGL Shader (and hooks for using it in Performer)?
although installed in /usr/share/shader, there isn't mention of it on
SGI's web site and i was wondering if it's being deprecated. if
there are problems using emulated cliptextures with programmable
shaders (or multitexture with texture combiners) then multipass could
be an acceptable work around, and OpenGL shader would be convenient.
did OpenGL Shader or some scaled back version of it ever work on
InfiniteReality? IR doesn't support GL_ARB_multitexture, which seems
to be required to compile it, and i think ARB_imaging is required,
which IR doesn't support either.
pfRep future directions/intentions
i am curious about the future directions of pfRep's, which
seem to be somewhat still under development. a few thoughts:
there seems to be a memory leak in
pfSphereSurface::clearTessellation().
i retessellate my pfSphereSurface every frame in APP and only can run
about 15 seconds before i run out of memory. the program isn't doing
anything else. if i manually remove and pfDelete the geoset instead
of calling clearTessellation(), it runs fine.
it would be nice to be able to generate texcoords instead of just
setting them to (u,v). for now a scale and bias of (u,v) seems
sufficient for what i need, but the pfScalar machinery would be more
flexible. right now i just have APP go through and xform them after
the tesselation. taking this idea further: given the existence of
programmable vertex shaders taking arbitrary attributes does it make
any sense for the extra machinery a pfRep subclass adds to a pfGeode
to be replicated per vertex attribute?
are there any thoughts being given to tesselations that generate
multiple geosets (segregating geometry into cliprings for example),
or parallel tesselation in CULL with fluxed geosets (particularly in
multipipe mode)?
-Bryan
This archive was generated by hypermail 2b29 : Wed May 18 2005 - 12:31:35 PDT