Blender Git Loki

Merge branch 'master' into temp_display_optimization

Conflicts:
source/blender/blenkernel/intern/cdderivedmesh.c

Merge branch 'master' into temp_display_optimization

Conflicts:
source/blender/blenkernel/intern/cdderivedmesh.c

VBO offscreen selection drawing, cdderivedmesh

Get rid of legacy drawing, it's only used for selection,
in which case we can prepare a temporary color buffer and draw
at once. Code is not complete here because we still redundantly
set the draw color in the draw function and don't ommit hidden
faces automatically. Still it works 100% without immediate mode
now.

Merge branch 'master' into temp_display_optimization

Merge branch 'master' into temp_display_optimization

Conflicts:
source/blender/blenkernel/intern/cdderivedmesh.c
source/blender/blenkernel/intern/subsurf_ccg.c

Fast drawing for sculpting multires using VBOs (non VBO won't have
fast mode but will not spend time on that).

With this commit branch should be in full feature parity with master.

Subsurf drawing refactoring:

All textured drawing now works correctly using VBOs (with usual CPU
overhead still, but this will be taken care of separately).
Solid texture painting should also work fine.
Renamed ccg GPU functions so they are more consistent with cddm

Merge branch 'master' into temp_display_optimization

Bah, merge change that was left uncommitted

Merge branch 'master' into temp_display_optimization

Conflicts:
source/blender/gpu/intern/gpu_buffers.c

Merge branch 'master' into temp_display_optimization

Use shorts for uploading normals to GPU instead of floats.

The range of shorts is acceptable for good looking
normals.

We allocate a buffer with 4 shorts to keep data aligned to 32bit
boundaries. The total mesh cost for full triangle meshes is now
equal to master branch even with index buffers and memory gains
are even more for ngons.

Cleanup: Use GPU material counter instead of allocating new counter
array every time.

Refactor subsurf multi material textured drawing.

Use the same system as cdderivermesh - still result is not correct
though.

GLSL fix one more error - meant that attributes would not get uploaded
correctly.

Merge branch 'master' into temp_display_optimization

Temporarily fix crash until subsurf textured multi material is
completely supported.

GLSL drawing redesign:

Use new upload scheme. Idea here is that materials share a VBO where
the size of each element is the maximum size of all mesh materials.

While this will waste some size if material sizes differ it is the
simplest scheme to use and allows easy reuse of indices as opposed to
separating the materials in separate vertex buffers. In fact if we do
that, management gets quite complex and code much more error prone.

I may write an extra blog post to explain the choices here
at some point.

Redesign of textured drawing: Get rid of triangle_to_mface array

Instead we can be smarter here and add an mface array to the
GPUBufferMaterial instead.
This will help us do less iterations on CPU for quad meshes as well as
avoiding checking materials every face (faces are now always sorted per
material so this happens implicitly).

Merge branch 'master' into temp_display_optimization