GitLab

update OpenVolumeMesh to newest commit



See merge request !151

Fix for meshcompiler

fix issue #76

See merge request !149

Fix crash for meshes with large faces. If the input adjacency is faulty, it gets recomputed internally (fixes #76)

Render object mods

Scenegraph nodes can attach/detach mods for RenderObjects to a renderer when the traversal enters/leaves the node. Mods are applied to any object emitted within the node subtree.

ClippingNode makes use of it to add slicing code to the shaders.

fix issue #65 

See merge request !130

FloatingSubtreeNode, GLMatrixT

Enhancement in FloatingSubtreeNode.
Enhancement in GLMatrixT.

See merge request !147

Unnecessarily, the constructor required the inherited type GLMatrixT only
to initialize its parent type Matrix4x4T. Thus, we can loosen the type
requirement and simply require Matrix4x4T as the initializer.

# Conflicts:
#	ACG/GL/IRenderer.cc
#	ACG/GL/IRenderer.hh
#	ACG/Scenegraph/BaseNode.hh

Floating subtree node



See merge request !146

Scenegraph traversal

Added getCollectedSubtreeObjects() function, mad traversal of scene graph non-recursive.

See merge request !145

Principal axis rendering

- use instancing when hardware support available
- added 3d arrow mode with instancing
- optimize vertex format of 2d line mode (non instanced branch),  see issue #71

See merge request !141

Performance Fixes

These changes fix some serious performance problems and maintenance nightmares. The commit messages explain in detail what has been fixed and why it was slow and and poorly maintainable.

See merge request !144

The additional enter and leave methods have an IRenderer* parameter
that allow the implementation of additional functionality in nodes
for modern renderers implementing the IRenderer interface.

std::sort is a template function so the compiler can inline the comparator
avoiding unnecessary calls. In contrast, qsort is a non-template function
taking a function pointer which results in one function call per comparison
operation.

Someone apparently had the misconception that clearing a vector
deallocates the reserved memory of that vector and went through great
lengths, writing hideous code in order to avoid ever decreasing the size
of the three vectors sortedObjects_, overlayObjects_, and lineGL42Objects_.
As a result, unnecessary variables were used for unsafe and hard to
maintain bookkeeping and the size of these vectors was clumsily computed
with an O(N) linear search instead of a constant time size lookup.

Each of the getNum*Objects iterates over the entire renderObjects_ vector.