"populate noise" step of chunk generation

[kralverde]

This PR implements (most of) the "populate noise" stage of chunk generation

The blender (for smoothing (i think)) and the beardifier (for leaving space for world structures) are not implemented in this PR

This PR changes the default world gen to this noise generation

Checklist

• Re-implement chunk noise
• Optimize chunk noise references
• Look into hot code if chunks take too long to load
• Add tests for actual chunk noise
• Code is well-formatted and adheres to project style guidelines: cargo fmt
• Code does not produce any clippy warnings cargo clippy

[kralverde]

The chunk noise generation and population each take like 0.5 seconds, so 1 second per chunk. I'll now start a deep dive into finding optimizations

[kralverde]

Chunk gen times are now ~0.1sec which i think is good enough for now

[OfficialKris]

This is a very large PR. I'm just curious, what was the methodology of this implementation?

[kralverde]

What do you mean by methodology?

[kralverde]

But basically I would look at the source code, write an equivalent in rust, and then write a bunch of unit tests for correctness. It was not fun lol

[kralverde]

If this PR is merged as-is, the default world generation would become that of the populate noise

[kralverde]

Final bench is ~80msec per chunk on my machine. Can likely be improved but I think its good enough for now

[Snowiiii]

What the hell is this class and why is this in pumpkin-world ?

[kralverde]

Its code to generate the no_bench_no_blend_0_0.chunk file, a fabric mod

[Snowiiii]

Why you don't put it into the extractor/ ?

[kralverde]

I didn't think about that, I'll do that

[kralverde]

~60msec/chunk on my machine with the unsafe code

[Q2297045667]

Report BUG
"populate noise" step of chunk generation

[kralverde]

Report BUG

"populate noise" step of chunk generation

I don't think this is necessarily relevant to this PR, as this PR does not change any spawning code. It might illustrate an upstream bug though

[kralverde]

There's a regression somewhere in here for terrain gen, don't merge this PR yet.

Isolated the issue to 68949ec likely line 539. I'll make a fix and add a test later today.

[kralverde]

currently benching at ~55msec/chunk now, I think thats adequate for this PR

[kralverde]

@Snowiiii should be good to go now I think

[OfficialKris]

Runs great on my computer, very impressive. But why does it have more loc than Minecraft itself?

[kralverde]

Runs great on my computer, very impressive. But why does it have more loc than Minecraft itself?

Loc?

[user622628252416]

Runs great on my computer, very impressive. But why does it have more loc than Minecraft itself?

Loc?

lines of code

[OfficialKris]

How are the asset files generated? Are they from minecraft? Is it possible to generate them at compile time? each of them are 3MB and that could be a problem to the git repo if they change often (like version updates).

[kralverde]

How are the asset files generated? Are they from minecraft? Is it possible to generate them at compile time? each of them are 3MB and that could be a problem to the git repo if they change often (like version updates).

They're generated from Java, not really able to generate them thru rust... maybe I could write a script to generate and run the tests? But then they couldn't be run with the PR tests.

[Snowiiii]

Mostly looks good to me. Only few NIT's but i can improve the code later, i would like to have this merged.
Big Big Thanks @kralverde ❤️