Change Details

This implements a merge by distance operation for point clouds. Besides the geometry input, there are two others-- a selection input to limit the operation to certain points, and the merge distance. While it would be a reasonable future improvement, the distance does not support a field currently, since that would make the algorithm significantly more complex. All attributes are merged to the merged points, with the values mixed together. This same generic method is used for all attributes, including the position. When many points are merged to a single point, floating point innacuracy had an effect. Possibly averaging floats with doubles would resolve that, but it would have a performance cost, and I don't expect it to be a common use case. For the implementation, I fiddled for quite a while trying to give the algorithm an optimal structure. Some parts are inherently single threaded, like finding the final indices accounting for the merged points. Merging the attributes is multi-threaded, though it might be possible to localize memory access there more. I expect most of the time to be spent in the KD tree anyway. This is a first step towards implementing T86913. I'm adding a patch for the point cloud part separately, because this is new code. The weld modifier code will be adapted for the mesh case.

This implements a merge by distance operation for point clouds. Besides the geometry input, there are two others-- a selection input to limit the operation to certain points, and the merge distance. While it would be a reasonable future improvement, the distance does not support a field currently, since that would make the algorithm significantly more complex. All attributes are merged to the merged points, with the values mixed together. This same generic method is used for all attributes, including `position`. When many points are merged to a single point, floating point inaccuracy has an effect. Averaging floats as doubles temporarily would help resolve that, but it would have a performance cost, and I don't expect it to be a common use case. For the implementation, I fiddled for quite a trying to give the algorithm an optimal structure. Some parts are inherently single- threaded, like finding the final indices accounting for the merged points. Merging the attributes is multi-threaded, though it might be possible to localize memory access there more. I expect most of the time to be spent in the KD tree anyway. This is a first step towards implementing T86913. I'm adding a patch for point cloud support separately, because this is new code. The weld modifier code will be adapted for the mesh case.

This implements a merge by distance operation for point clouds. Besides the geometry input, there are two others-- a selection input to limit the operation to certain points, and the merge distance. While it would be a reasonable future improvement, the distance does not support a field currently, since that would make the algorithm significantly more complex. All attributes are merged to the merged points, with the values mixed together. This same generic method is used for all attributes, including the `position`. When many points are merged to a single point, floating point innaaccuracy hads an effect. Possibly aAveraging floatss as doubles with doublestemporarily would help resolve that, but it would have a performance cost, cost, and I don't expect it to be a common use case. For the implementation, I fiddled for quite a while trying to givee the the algorithm an optimal structure. Some parts are inherently single- single threaded, like finding the final indices accounting for the merged merged points. Merging the attributes is multi-threaded, though it might might be possible to localize memory access there more. I expect most most of the time to be spent in the KD tree anyway. This is a first step towards implementing T86913. I'm adding a patch for the point cloud pasupport separately, because this is new code. The weld modifier code will be adapted for the mesh case.