I have a really good plan for this feature. But this sort of approach would make this plugin a bit too complicated.Ī feature request was made for Grouped Vertices. If you want to lock something along an arc, then you would need to assign a some sort of thickness (or edge cylinder radius) to a pin, so it knows how much each edge should capture. If you treat all edges of groups as cylinders you can also, in reasonable time, determine which vertices overlap the cylinders. But of course, a simple variant can be created. Doing it other way, for complex shapes it's hard. The thing with this, the reason I went with bounding boxes is that it is easy to determine if a vertex is inside it. I would be interested in finding out if SU could optimize this, say avoid updating edge lengths (in case they do) or other things that occur.īox wrote:A feature request of sorts, more of a think about this and see if it is possible.Ĭan you make the shape of the geometry of the pin the thing that does the pinning rather than the bounding box of the pin/pin object. Thus at the moment there are no bottlenecks but as physics computation performance is improved, transform_by_vectors (and rendering) might become the bottleneck. At the moment, the part of physics that consumes most performance is collision detection and contact processing. The physics computation aspect can and will be improved, if not so much algorithm and optimization-wise, then with multithreading. Assuming transform_by_vectors is off, with physics enabled, the performance is 32 FPS with physics disabled, the performance 60 FPS. However, the physics computation also takes its chunk out of performance. On a draping cylinder sample (with no textures), with transform_by_vectors enabled, the performance is 17 FPS with transform_by_vectors disabled, the performance is 32 FPS. The transform_by_vectors takes quite a reasonable amount of performance out of this (and I do wrap every frame in its own operation with second parameter set to true). Dynamic texture UV updating for faces with texture (which can be disabled).Vertex transformations via transform_by_vectors.There are three components to simulation: Jiminy-billy-bob wrote:Hey Anton, I was wondering today: what's the performance bottleneck? Is it the physics simulation itself, or moving the whole geometry at each frame?
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |