BioGrid

@baquast I had to put it on hold for a bit, I'm working on a different project. I'll be able to continue development soon enough, in spring.

@BioGridGame what's happening on this front? any updates?

Following a really hungry bear around for a while. Predators aren't all that good at hunting yet since everyone moves at the exact same speed.

@lisyarus Hmm, is it using a 2D noise then? Seems like it only does vertical displacement.

@BioGridGame It's already based on perlin+worley! It's not as billowy as it could be because I'm trying to keep the vertex count manageable

Made the cloud normals smooth based on the noise gradient (which was a bit tricky because the noise → height mapping has infinite derivative on the cloud edge), and now it looks pretty much exactly how I envisioned it! ☁️ #indiedev #gamedev #indiegames #screenshotsaturday

@lisyarus @epibates What about the noise field itself? You could maybe use central differences/tetrahedron technique to calculate analytical normals, iq has an article in that.

@epibates I'm generating triangles one at a time, so I don't have access to neighbouring triangles and adjacency info and stuff like that, so I can't smooth them easily

Fixed cloud normals & added cloud shadows. Still contemplating ways to make the normals smooth... #indiegames #indiedev #gamedev

@lisyarus Yeah, like this popping effect, right? It might still be worth it to test just using a lot more angles or using a per-instance jitter to hide the line where they snap from one view to another. Octahedral is great for middistance, but for faraway stuff it's kinda overkill imo.

@BioGridGame Yep, octahedral. With no blending there are very noticeable edges between areas using different sprites because they react to lighting very differently.

So, some numbers: ~28k trees, rendering them as models (~1.2k triangles) takes 29.5 ms, rendering them as impostors takes 3.2 ms, about 10x faster. I probably can push this further my using a simpler shader for distant impostors, and there's almost no visual difference!

@lisyarus There are a few options - either encoding a quad scale per view angle, or automatically pregenerating something like a 8 vert or so bounding hull, not just a simple quad.

@BioGridGame Tight packing is hard given that each tree model has a different overall shape. Though I might compute per-sprite texture-space bbox to make it tighter, we'll see!

@_wallfacer Yeah, rendering and the terrain sim are generic and just use compute shaders and indirect instacing. Entity sim on CPU side uses Unity ECS though, so it's a bit more coupled to the engine.

@BioGridGame I was trying to respond to an older post but it won't work. Is this still mostly compute shaders/engine agnostic? Unity really just serves as a render engine at that point right? Absolutely fascinated by compute shaders right now.

@koguGameDev 森林破壊の一因になっている可能性はあると思うが、ビーバーは真似できないほど賢い。

ビーバーは地形を変え得るんだろうか？

Beaver Simulator 2025

@AHappyDragon There's no special handling or animation for them falling yet, it's basically showing the raw sim results.

@BioGridGame Why does it move so linear-ly?

Added a harsh but necessary system to deal with occasional overconfident animals

@MrJTroyer No, the opposite. They all try to follow the same path while avoiding others.

@BioGridGame Do they follow each other (flocking/herd behavior)?

Improved animal pathing a bit - their main steering algorithm only has local knowledge of obstacles, and it doesn't deal well with complex terrain and distant goals. So I added an incremental A* solver so that stuck entities can "think" for a bit and get a smarter path.

@ben_throop Nope, against a regular grid. Every entity has one DynamicBuffer with nodes and another one that's a binary heap for open set. They also have a tiny bitmap for O(1) visited/duplicate check.

@BioGridGame Are they solving against a navmesh? What's the data structure?

@squirtBeetle I should probably avoid tweeting late at night I guess I did use it for quite a while, the lack of threads was no problem if you didnt know how to multithread anyway.

Today I upgraded from a Core 2 Duo (died of old age 🫡) to a mid-range Ryzen 5. ...It seems I might've overoptimized my codebase a bit.

@tophatsandgains I'll keep them in reserve, but it's nice to know I have some headroom on modern specs. Im working on improved animal behaviour right now and I feel it'll become a lot more compute heavy before I'm done.

@BioGridGame Any plans with all those newly found performance reserves? 😎

@tombombadeel Yeah, I misremembered, it was an old i5, so probably like 2014 not 2006.🤷 I still had to throw out the whole mobo and RAM though.

@BioGridGame I had to look up if there was a more recent core 2 duo because I thought there was no way 😂 did you upgrade basically the whole rig, or was the 2006 processor sitting in a modern mobo?