Used a faster method for checking for power of 2 on the image dimensions and

did a check on forcing 16bits textures in osg::Texture (which speeds things
up on the Umea model which is performance bound by texture memory.), the
later check is now commented out.

src/osg/Texture.cpp

@@ -416,6 +416,11 @@ void Texture::applyTexImage(GLenum target, Image* image, State& state) const
             break;
 
     }
+    
+    // an experiment to look at the changes in performance
+    // when use 16 bit textures rather than 24/32bit textures.
+    // internalFormat = GL_RGBA4;
+    
 
     if (_subloadMode == OFF) {
         if( _min_filter == LINEAR || _min_filter == NEAREST )