Smooth Node

Okay, I have finished the planned refactoring, and I think that I have to wait for the solution from above.

6 Likes

Was wondering what the status of this node is? Really hoping that it may have landed in 3.3 :grinning:

I got a review. And by the looks of it, I’ve deviated a lot from blender standards. I will need to do one more refactoring in order to standardize most of the things. Now, due to the end of my studies, I will have more time, maybe end of 3.3, 3.4 (or 5) …

8 Likes

Thanks for the reply, looking forward to it.

I didn’t know I needed something like this until yesterday when I was trying to make a fast thickness attribute generator yesterday but the result wasn’t smooth, this node will be pretty useful once finished =D
image

1 Like

Hi @genesis2303, I had to remove your post discussing details of other proprietary software for copyright reasons. See here:

非常期待早日进入每日主干
look forward to entering the daily master as soon as possible.

1 Like

From the development news, now I’m trying a new algorithm for working with the number of iterations field:
At the expense of avoiding the constant checking of the number, I group all the points by the group of their number, after working out a simple algorithm. And now I go through all the groups. It is also a plus that all groups can be processed in parallel.
image

For example, with this option for the number of smoothing:

image

3 Likes

3x speed boost for the mesh!

19 Likes


I can look at this forever

13 Likes

I abandoned the division into complex and simple stages.
The new variant of the difficult stage differed due to the grouping. And because of this, he became the most optimal in principle. Now I redo the code in 3 days

Edit:
Okay, it looks like the group system overlaps with the earlier optimization of the curve algorithm.
I’m not sure if I’m losing much because of the 1-2 if.
But the current variant is the most simple and standardized.
Although even it is quite cumbersome

6 Likes


New comparison

11 Likes

Wow, 5000 repeats is super slow in the modifier.

If I don’t confuse anything, if the application doesn’t stop for more than 5 seconds, it windows tries to close it. It seems that this is what, for a few seconds, turned off the program in the process.

After some googling, found this thread. I definitely have a useful application for “Attribute Smooth” as soon as it’s possible, and it isn’t based on the Position attribute. Any word on its progress to a second chance at a pull request / merge?

1 Like

You can already use the closest substitution in a form of a node group made by @higgsas:
Thou this is not the place to discuss how to use Blender, so you might wan to to check thread on BA Forum:

I got sick of covid this week and can’t quickly update the patch, but last 2nd reviews Hans wrote that I’m close enough

4 Likes

silex, modmoderVAAAA, thanks both for being so responsive.

And Higgsas, thanks for that excellent short-term workaround. It worked beautifully. I am sure the modmoder version will work so much faster, especially since my modifier is made to be used on multiple objects and 10-20 iterations seems common from early tests.

I don’t think I’ve written anything here for a long time. Basically, the node is now in a stable state. That is, from the point of view of implementation, at least I do not see new directions for development. That is, until one of the developers checks it, there will only be small fixes.
Although, for the future, I have ideas for outputting part of the code from the node to other files in order to increase its reusability. But these will be separate projects.


From the node deconstruction ideas, I managed to find the following things:

  • Repeat attribute input.
  • Domain interpolation (vertex → edge, edge → vertex).
  • Isolated calculation for group.
  • Repeat calculation.

3 of these nodes are not implemented, specific, and the run time is very huge. But this divides 1 node into 4 others that can be used elsewhere. I think when they can be added, but for now the node decomposition variant is dead.

7 Likes

As a result, it was still necessary to abandon the count fields. But I may be able to add this later, more generally.

11 Likes