Could it be bug in code? about interactive-erosion

[Scrawk]

It should be 0. I use the read/write when ‘ping ponging’ between two textures. In this case I just need two textures for the advention stage.

You can ask some questions but I wrote this a long time ago so might not remember.

[Nashet]

Thanks to your great work I get almost all code.

1. What is going on at the edges of map? I understand that ApplyFreeSlip(RenderTexture[] field) (line 244) clears some noise at the edge of texture. But where is that noise coming from?
2. Did you faced memory/processor time shortage in simulation?
3. Were you satisfied by simulation realism?

[Scrawk]

1. The edges need to be handled by the free slip (a type of boundary condition) so the simulation is correct there. Clamp can not be used. You get the water slowly 'leak' away over time otherwise. Not sure what you mean by noise but precision issues can cause some artifacts at the edges over time.

2. Yes, memory is a issue. You cant just increase the texture sizes to scale up as memory would be a issue and the GPU may time out. You would have to split it into smaller chunks.

3. The erosion effect is not that good. The sediment advection has a precision issue so a lot of the sediment just disappears. Advection is a hard problem so its to be expected but maybe it can be improved. Not sure if the original had this issue. The 4 terrain layers (I think I only use 3) is also over kill. I would just have 2. The base height and the sand. It does make good rivers though.

[Nashet]

Thanks for your answers. If you still have time:

1. I meant that my problem was that when I removed ApplyFreeSlip() (replaced it with "Oceans" mechanics at the edges of map) water started to multiply wildly at the edges. Turns out problem was that shader called pixels with negative coordinates at the edges trying to get "left" pixel. I added check for boundaries and it fixed it. No need for ApplyFreeSlip() anymore.
2. Is that why you used RenderTextureFormat.RHalf instead of RenderTextureFormat.RFloat for m_sedimentField and some others? Do you think that rising it to 32 bits would improve simulation?
3. With rain I can't get delta formation so far neither meanders. I'll try to improve it. BTW regolith flow looks bad for me. It looks as it just polish everything at bottom to one level.