But exams start soon, so this may be the last update for about two months 😞.

Reminder / Info for those who haven't seen it: I'm coding the mod as practice AND as a reference so I can improve the datapack's precision by knowing exactly how close each intermediate result is to the "real deal".

I'm not abandoning the datapack!

New GitHub Repository: github.com/SilicatYT/Ph...

(The mod is in the "floating-point-reference" branch, the datapack will be in the "main" branch)

Because of the 3 month break, semester exams are already almost here. So while I'd love to continue, I have to take a break from this for a while again.
Rest assured though: I'll pick up fast from there.

I already formulated a rough pseudocode for "Face culling".

(6/6)

So the end result will be an equally optimized datapack, but getting there will be easier for me. I won't have to redo 15 functions from scratch because I changed a small thing in a single formula. And maybe some optimizations I made make almost no difference. I'll see.

(5/6)

I'll hold off on hardcoding every calculation 15 times from the start (in the datapack). Instead, I'll write "readable" code that I'll be able to optimize once it works. I'll note down every instance of optimizable code, but overall that should make it much easier for me.

(4/6)

Once the base functionality is done for the mod, I'll restart development on the datapack (Don't worry, most will remain the same. I'll just change some conventions).

(3/6)

Basically I'll write this mod as a basis to test my formulas and theoretical approaches much more efficiently than implementing it straight into the datapack (which takes super long because of optimization, and might be imprecise due to scaling issues).

(2/6)

Overall, I'm confident that I can improve the stability of my datapack quite a bit. Without a reference, calculating it by hand to test my maths was too painful and error-prone. I couldn't tell where I should increase the scaling for best results. Now it should be easier.

(4/4)

Benefits:
- Test formulas & approaches more easily (less datapack maths pain)
- If my approach jitters even at full floating point range, I gotta change my formulas
- I can compare the datapack's calculated values directly with the perfect-accuracy values from the mod

(3/4)

I'll start writing a Fabric mod with the same formulas that I use for the datapack, and then "rewrite" my datapack based on it. The datapack will end up being mostly the same as before, but there are some things I want to clean up.

(2/4)

frogue

Hyped

Fake colour fan

Finde den Patch sehr gut (abgesehen von den von dir erwähnten Punkten und der Tatsache, dass immer noch so viel Customization fehlt), aber die erhöhte ?-Block Regenerationsgeschw. hätte nicht sein müssen...

I'm talking about Galaxy 1. If you bought 3D Allstars for Mario Galaxy, you have to buy it again ($40) if you want the update (better textures or 4k), even though you already have that game for the same console. People who trusted Nintendo that it's *the* Switch version basically got a worse game.

It's probably the expectation from 3D Allstars & the fact that SMG was $20 all the way back then in Nintendo Selects.
And if you bought 3D Allstars mostly to have a definitive Switch version for SMG, you have to buy it again for the same console for minor improvements and Switch 2 4K...

Objects can be cuboids with varied x, y and z dimensions (currently capped at 4 blocks per axis at most, but I'll likely increase it at some point), and vanilla blocks have a system to support multiple cuboid shapes (for stairs etc), but only the full block hitbox is coded atm. System's in place tho