Seeing how Cannolies is taking too long and I don’t want to build a map for my zombie game until after its ready for an actual map, I decided I’ll build this beast.
That’s what level of detail is for, you have a settings menu where users can choose between the quality of the tools and stuff in a game. With the new box physics on CSG, it shouldn’t be too bad though.
except by lower end graphics, he means integrated chips, and integrated chips weight heavily on the CPU just like physics does. The problem with low end graphics is that low end graphics people don’t even have a GPU. Reducing physics drain on the CPU frees up resources for graphics processing
That could really help me out ;o A lot of my unions are really complex, so I might actually do this on all the things I have. Thank you! Anything for performance
What are the differences between the CollisionFidelity types (the wiki hasn’t explained CF very well)? Is there one that uses the absolute least amount of computing possible, or do Box and Hull use roughly the same amount?
And last time I checked, which was last night, CSG is a monster with making unnecessary polygons. A normal part has 12 polygons, a union can have up to 2,500 polygons. So 35 parts times 12 polygons = 420 polygons. While in unions it can be a maximum of 2,500 per union, so lets say each union was 2,500, times that by 35 and that equals to 87,500 polygons. That is a lot for just a weapon in a game.
Rofl. Yes, the screenshots are just showing off the tutorial dialog on the bottom compared to the old one. (The old one just gave the player info to read, while the new one takes the player through the gameplay step by step.)
In short, it’s the hitbox’s level of detail. Box is the fastest; Hull is slower than Box but way faster than Default. Default stays true to the geometry, Hull is an approximation, and Box doesn’t even try to get it right.
In more words:
CollisionFidelity determines how accurately the shapes need to be represented to the physics engine and for raycasting.
Box lets roblox treat a CSG solid as an OBB (oriented bounding box). Geometry intersection tests with OBBs are very cheap; iirc a naive segment-OBB test is like 20 lines of code.
Hull means that the mesh is treated as a convex hull, which you can think about as the result of wrapping something in wrapping paper. Testing these for intersections isn’t that expensive but it really depends on how much accuracy you need. This is usually a good middleground for weapons, etc.
Default is the slowest because it uses close to the actual geometry of the solid, performance be damned. That shape isn’t necessarily convex, so relatively expensive tests need to be performed to determine whether or not an intersection exists and, if so, where that point is located.
Real-Time Collision Detection is worth its weight in gold if you’re interested in learning how the math behind this stuff works.
Also, studio has an option in render settings called something like ShowDecompositionalGeometry that lets you visualize the bounding surfaces.
Cool, thanks for the explanations, Mistertitanic44 and ScrapYard. At the very moment i’m working on an Alien architecture build and it uses Unions like nobody’s business. (Even the Semiotic Standard icons are Unions!)