2022-11-29 Geometry Nodes sub-module meeting

The meeting will be open for anyone interested in joining the video call (link below).

Meeting time: 15:00 CEST (Your local time: 2022-11-29T14:00:00Z).

Links

Present

  • Dalai Felinto
  • Hans Goudey
  • Iliya Katueshenock
  • Jacques Lucke
  • Miro Horváth
  • Wannes Malfait

Since the Last Meeting

  • Dalai finished and posted the Post-Conference Workshop blog post
  • The first builds of the WIP simulation branch are available. Bugs are expected currently.
  • Jacques looked into improving socket alignment.
    • The idea is to make nodes look simpler and to clarify the sample nodes, especially when multiple fields can be sampled at once.
  • There was a meeting to discuss procedural hair workflows.
    • Jacques has started the Interpolate Curves is the first new node/group to be implemented.
  • Iliya made some improvements to naming on the Collection Info and Transform nodes as discussed last week.
  • The Image Info node from Iliya has been committed.
  • The Named attribute node now has an “Exists” output.
  • Pablo Vazquez committed his improvements the node context menu.
  • Yann Doersam changed copy and paste allow pasting common nodes between editors.
  • The team also worked on various performance improvements.
    • Wannes moved the split edges node from BMesh to Mesh, making it 2-3x faster.
      • Hans improved the performance slightly after that.
    • Hans worked on making the node editor faster when editing large node trees.
    • Jaques improved evaluator performance when there were disconnected nodes.
    • The points to volume node is slightly faster with large inputs.

General Discussion

  • Simulation nodes
    • Not working at the moment:
      • Run/Stop
      • Anonymous attributes
      • Socket inspection of cached data
        • We may want to be able to see the cached values from the last frame data which was evaluated, whether or not it was simulated in the current frame.
      • And many other things.
    • Iliya mentioned the possibility of accessing arbitrary previous frames
      • Jacques mentioned that it probably doesn’t work so well, since we would always have to keep the data of previous frames around.
      • There are ways to work around this: multiple simulation inputs and outputs, instances, storing more attributes, etc.
    • Dalai asked about simulation substep evaluation
      • Jacques mentioned that a per-scene option could make sense.
    • Cache
      • The current “Use Cache” option is temporary.
      • Options as node properties don’t necessarily work on the simulation input or output nodes, since they will be inside groups.
      • One idea is to follow the design for “Geometry Object” and have a panel on the object (and/or scene?) level that lists all the simulations, allowing freeing/caching individually.

Patch Review & Decision Time

  • D16638: Geometry Nodes: Edges to Face Sets Node
    • The names could use more thought-- particularly plurals, including “Edge” in the name, and the naming of “Face Sets” vs. “Group Index” or other options.
      • Make a list of the different naming options like “Face Set” or “Group Index”
      • It might noe be necessary to include “Edges” when field inspection should show that in the future.
    • Besides the naming, the node looks useful for some situations.
  • D16471: Geometry Node: Edge Rings node
    • An example of how it would be used for procedural modeling in practice would be helpful.
    • There was discussion about whether it should work across polygons
    • This and the “Edges to Face Sets” node would be useful with a “Select Similar” node that used the output integer group index and an initial selection.
  • D16397: Image to Points geometry node
    • Now that the functional parts have been made availabel with the image info node and future index-based image sampling, the purpose of the patch is performance.
      • Operations on point clouds are already fast, but the patch may still be an improvement.
      • The description for make a case for the feature on a performance level.
  • D16076: Geometry Node: Stretch profiles around turns in Curve to Mesh Node
    • Hans mentioned that the logic is a bit tricky here and it seems there are a quite a few different solutions. Anyone interested in helping out would be welcomed here.
  • D16133: Geometry Nodes: New Relax Face Sets node
    • The meeting ran out of time before this could be discused.

Follow Up

  • There were no specific follow-ups discussed in the meeting.

Next Meeting

The next meeting will be on Tuesday December 13, 15:00-16 CEST (Your local time: 2022-12-13T14:00:00Z), which is 2 weeks from this meeting. The provisional meeting agenda will be linked in the #geometry-nodes channel before the meeting.

6 Likes