From Time Moore, "This submission implements 3 optimizations for meshes. INDEX_MESH turns DrawArrays style geometry into DrawElements, uniquifying the vertices in the process. This is useful for certain loaders, like ac3d, which just spit out DrawArrays. VERTEX_POSTTRANSFORM and VERTEX_PRETRANSFORM optimize mesh triangle and vertex order for the caches on a modern GPU, using Tom Forsyth's algorithm. I describe this and the big difference it makes (38% improvement on a very large mesh) in my blog,

http://shiny-dynamics.blogspot.com/2010/03/vertex-cache-optimization-for-osg.html."

include/osgUtil/MeshOptimizers

@@ -0,0 +1,101 @@
+/* -*-c++-*- OpenSceneGraph - Copyright (C) 1998-2006 Robert Osfield 
+ *
+ * This library is open source and may be redistributed and/or modified under  
+ * the terms of the OpenSceneGraph Public License (OSGPL) version 0.0 or 
+ * (at your option) any later version.  The full license is in LICENSE file
+ * included with this distribution, and on the openscenegraph.org website.
+ * 
+ * This library is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the 
+ * OpenSceneGraph Public License for more details.
+*/
+
+#ifndef OSGUTIL_MESHOPTIMIZERS
+#define OSGUTIL_MESHOPTIMIZERS 1
+
+#include <set>
+#include <vector>
+
+#include <osg/Geode>
+#include <osg/Geometry>
+#include <osg/NodeVisitor>
+
+#include <osgUtil/Optimizer>
+
+namespace osgUtil
+{
+// Helper that collects all the unique Geometry objects in a subgraph.
+class OSGUTIL_EXPORT GeometryCollector : public BaseOptimizerVisitor
+{
+public:
+    GeometryCollector(Optimizer* optimizer,
+                      Optimizer::OptimizationOptions options)
+        : BaseOptimizerVisitor(optimizer, options) {}
+    void reset();
+    void apply(osg::Geode& geode);
+    typedef std::set<osg::Geometry*> GeometryList;
+    GeometryList& getGeometryList() { return _geometryList; };
+protected:
+    GeometryList _geometryList;
+};
+
+// Convert geometry that uses DrawArrays to DrawElements i.e.,
+// construct a real mesh. This removes duplicate vertices.
+class OSGUTIL_EXPORT IndexMeshVisitor : public GeometryCollector
+{
+public:
+    IndexMeshVisitor(Optimizer* optimizer = 0)
+        : GeometryCollector(optimizer, Optimizer::INDEX_MESH)
+    {
+    }
+    void makeMesh(osg::Geometry& geom);
+    void makeMesh();
+};
+
+// Optimize the triangle order in a mesh for best use of the GPU's
+// post-transform cache. This uses Tom Forsyth's algorithm described
+// at http://home.comcast.net/~tom_forsyth/papers/fast_vert_cache_opt.html
+class OSGUTIL_EXPORT VertexCacheVisitor : public GeometryCollector
+{
+public:
+    VertexCacheVisitor(Optimizer* optimizer = 0)
+        : GeometryCollector(optimizer, Optimizer::VERTEX_POSTTRANSFORM)
+    {
+    }
+
+    void optimizeVertices(osg::Geometry& geom);
+    void optimizeVertices();
+private:
+    void doVertexOptimization(osg::Geometry& geom,
+                              std::vector<unsigned>& vertDrawList);
+};
+
+// Gather statistics on post-transform cache misses for geometry
+class OSGUTIL_EXPORT VertexCacheMissVisitor : public osg::NodeVisitor
+{
+public:
+    VertexCacheMissVisitor(unsigned cacheSize = 16);
+    void reset();
+    virtual void apply(osg::Geode& geode);
+    void doGeometry(osg::Geometry& geom);
+    unsigned misses;
+    unsigned triangles;
+protected:
+    const unsigned _cacheSize;
+};
+
+// Optimize the use of the GPU pre-transform cache by arranging vertex
+// attributes in the order they are used.
+class OSGUTIL_EXPORT VertexAccessOrderVisitor : public GeometryCollector
+{
+public:
+    VertexAccessOrderVisitor(Optimizer* optimizer = 0)
+        : GeometryCollector(optimizer, Optimizer::VERTEX_PRETRANSFORM)
+    {
+    }
+    void optimizeOrder();
+    void optimizeOrder(osg::Geometry& geom);
+};
+}
+#endif

include/osgUtil/Optimizer

@@ -85,6 +85,9 @@ class OSGUTIL_EXPORT Optimizer
             TEXTURE_ATLAS_BUILDER =     (1 << 15),
             STATIC_OBJECT_DETECTION =   (1 << 16),
             FLATTEN_STATIC_TRANSFORMS_DUPLICATING_SHARED_SUBGRAPHS = (1 << 17),
+            INDEX_MESH =                (1 << 18),
+            VERTEX_POSTTRANSFORM =      (1 << 19),
+            VERTEX_PRETRANSFORM =       (1 << 20),
             DEFAULT_OPTIMIZATIONS = FLATTEN_STATIC_TRANSFORMS |
                                 REMOVE_REDUNDANT_NODES |
                                 REMOVE_LOADED_PROXY_NODES |