- Add debug, benchmark code.
- Change linux/x86 FAST_MATH code to GCC/x86, and clear FP exceptions
before exiting the fast math block.
- Remove divide-by-zero test in x86 cliptest, and set clipped vertices
to [0,0,0,1] instead of leaving them uninitialized.
New array_cache module
Support 8 texture units in core mesa (now support 8 everywhere)
Rework core mesa statechange operations to avoid flushing on many
noop statechanges.
math: Provides basic matrix and vector functionality that
might be useful to multiple software t&l
implementations, and is used by core mesa to
manage the Model, Project, etc matrices.
tnl: The real transform & lighting code from core mesa,
including everything from glVertex3f through vertex
buffer handling, transformation, clipping, lighting
and handoff to a driver for rasterization.
The interfaces of these can be further tightened up, but the basic
splitting up of state and code move is done.
