Slow Forest Edge

[Victor Vlasenko]

Hi, I just noticed that adding a forest edge map to opacity slows down rendering, even if there is no forest pack or areas not using edge mode.
I thought it might be corona, but seems like it forest. And seems like this problem existed before v9.

After some investigation and comparison with chaos scatter edge and forest 8.26:
                 Rays/t   Noise
scatter        8.9M    16.22
forest 8.26  7.88M  18.16
forest 9.01  7.6M    18.42

x2 density, x4 instances
scatter            5.75M    32.54
forest  8.26     5. 1M     34.56
forest  9.01     5M         34.87

1m render, max 2024, latest corona daily.

The difference in speed between 8.26 and 9.01 indicates something wrong with forest pack I guess. I did a few renders for each forest and scatter and got the same results each time.

Any thoughts on this?

Thanks.

[Victor Vlasenko]

Update

Seems like we have similar problem with forest color (((

Forest color with 3 maps/ID from element, no tint, no color correction
CoronaMultiMap per element.

Even if I use only one default map in forest color, so no randomization - it's still slower than CoronaMultiMap with randomization.

[iToo]

Hi, I just noticed that adding a forest edge map to opacity slows down rendering, even if there is no forest pack or areas not using edge mode.
I thought it might be corona, but seems like it forest. And seems like this problem existed before v9.

I'm afraid some performance hit is normal.
Even if you don't use Edge mode, just adding ForestEdge to the opacity slot, triggers this process for each raytrace hit:

1) ForestEdge gets the underlying object and raytrace data.
2) It calls to Corona, to identify if the hitted object is a Forest or not, and what sub-item is rendered.
3) With this information, ForestEdge returns the base color or applies the edge cutting.

The difference in speed between 8.26 and 9.01 indicates something wrong with forest pack I guess. I did a few renders for each forest and scatter and got the same results each time.

Please give us some time to make tests. We'll also check if there is some difference in the code of ForestColor/ForestEdge between FP8 and FP9.

[Victor Vlasenko]

1. Yep, I understand that it's not a free feature and there will be some impact, but right now it's kind of quite a big impact compared to ChaosScatterEdgeTrimming, especially in the case when there is no ForestPack object in the scene.
2. Same for Forest Color, right now it without any randomization it's slower than CoronaMultiMap with randomization.

So is it something that can be fixed/improved or is it a problem on the corona side and I should ask them about it?
3. Definitely a bug, map->forestcolor->material=grey color (affects material editor and but base(custom) objects on render, but not forestpack). As for now can be fixed by enabling tint override.

[iToo]

1. I checked ForestEdge code, and found a way to improve performance a bit (15% from my tests). It's included in FP 9.0.2.

2. Same for For ForestColor, but in this case i cannot see any way to improve it.

Please note that although both plugins share similar functionality, they may work internally in a complete different way.
Also, Corona team have direct access to their render engine, and may optimize it more. We must use an intermediate interface to comunicate with the renderer.

3. Agreed. Fixed in FP 9.0.2.