You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Merge doubles currently uses an internal
Method where it Merges doubles per vertex and then tries to read the vertex changes per merge.
The issue is that even with the optimizations it has received, it is extremely slow on big models with lots of vertices.
To fix this, meshes should be turned into BMesh structures with 2 custom data layers that store both the vertex position and vertex number. These will be per vertex and the original stored in a dictionary.
The vertices after merging all moving vertices will have their custom layers checked, and if a vertex number or a vertex position in the layer doesn't match its original, then mark the vertex as missing. (Save for some extra logic to count properly)
This will allow the system to do all vertices at once without duplicating the mesh per moving vertex. This will greatly accelerate the advanced merge doubles process, and make it possibly just as fast as simple.
The text was updated successfully, but these errors were encountered:
Merge doubles currently uses an internal
Method where it Merges doubles per vertex and then tries to read the vertex changes per merge.
The issue is that even with the optimizations it has received, it is extremely slow on big models with lots of vertices.
To fix this, meshes should be turned into BMesh structures with 2 custom data layers that store both the vertex position and vertex number. These will be per vertex and the original stored in a dictionary.
The vertices after merging all moving vertices will have their custom layers checked, and if a vertex number or a vertex position in the layer doesn't match its original, then mark the vertex as missing. (Save for some extra logic to count properly)
This will allow the system to do all vertices at once without duplicating the mesh per moving vertex. This will greatly accelerate the advanced merge doubles process, and make it possibly just as fast as simple.
The text was updated successfully, but these errors were encountered: