Some questions...

[numerobis]

I just startet testing the Skatter demo and i get some extremely high processing times (more than 100 min !?!) . I have 16 different scatter objects which are simple round, 12 sided proxy elements for grass patches (Maxwell mxs-references); . The host object (terrain) has ~3300 faces in this case.
Maybe there is something wrong with my settings?

And second question: Is there a way to abort the processing?

[Rich O Brien]

As Thomas mention to me. The Random distribution method is not suited to large surfaces.

Use uniform and adjust the XY spacing and jitter settings.

[numerobis]

I have done some tests now with a smaller object (5x5m) and it looks more like it's the opposite

Random - 3.18s for 976 objects

Uniform - 11.79s for 1141 objects

And there are some other things that i noticed doing these tests.
The probability seems to be completely off in both versions. The percentages are listed on the left. And you can clearly see that there are far too many small objects in both distributions. There should only 4% of them.
And the collision control is also not working as it should. In some cases there is clearly too much space between the objects and some spaces are not filled where the smaller objects would fit in.

[numerobis]

Some more testing... better, but still far from good.

I adjusted the probability and the overlapping. But even with less than 0.03% i still get many small objects, even in places where bigger objects would fit. There is much space at the boundaries and still holes and double overlappings. And the processing time was more than 5 mins for the uniform version.
How is the object percentage calcutated? Is it based on the number of objects or the covered area? From the look i would say it can't be the number of objects.

[numerobis]

How is the density parameter working? I can't see a relation to the entered numbers. Doubling the value does not double the density - but the calculation time is skyrocketing. And even with such high values there are still blank spaces.

[jiminy-billy-bob]

1 - The whole thing is slow because of the collisions. This is a slow algorithm because each object has to be compared with every other objects.
That's why you should avoid using it if possible.

2 - Skatter is not intended for this kind of accurate circle packing. I actually plan to add this type of distribution in the future, but it requires a very complex algorithm.

[jiminy-billy-bob]

How is the density parameter working? I can't see a relation to the entered numbers. Doubling the value does not double the density - but the calculation time is skyrocketing.

This is because you have collisions enabled. The calculation time is skyrocketing because you are generating more and more objects, which have to be checked individually for collisions with each others. Since they were already overlapping at the first stage, in the end you don't see more objects.

[numerobis]

I don't need "accurate circle packing", but what i would expect is to see an increasing coverage until all holes have been filled. And when i can enter percentages for different objects i would expect to see a result that is at least in the closer range of the input values.
There are many areas filled up with far too much small objects where bigger objects would fit in. And free spaces where smaller objects would fit, especially at the border. How can i get a cleaner border?

[jiminy-billy-bob]

I understand.
The percentage should be correct if you don't activate collisions. But as the percentages get screwed up by the collisions as you have bigger objects than others. For now, Skatter doesn't compensate for this.
I'm not even sure how to do it actually, I'll think about it.