Add worley noise optional feature #74
Conversation
|
what are u gonna use this for? |
jungle |
|
biomes divison or something? |
perlin noise handles biome divisions, gonna use this bad boy for dense foliage generation as this noise provides much more 'forest like' pattern |
|
Is this based off of an implementation of Worley noise you've seen somewhere else? I can't find any that look similar and I was hoping to compare them. |
i went off 100% of what i read on multiple sites that documented what worley noise actually is and how it is achieved, I didnt actually read code snippets from a pre-existing implementation. Chances are high there is a faster implementation somewhere that im not aware of. The noise outputs a gradient that is later crunched into a true/false vector given a threshold, so i guess if you hooked it up before it gets crunched to like some kind of visualization program it'd start to look simmiliar? |
|
Well... I was mostly concerned by the vector that gets sorted once for every single map tile. If you've tested that it's fast enough for what you want to do... I guess I can ignore the performance. |
I mean i dont really know any other solution as worley noise requires me to find nth (in this case 1st and 2nd) closest node in every polling point, so unless there is some rust thing i dont know about that'd speed it up i'd gladly implement it |
|
tho there could possibly be performance increase from parallelizing some parts |
|
It's fairly slow at larger resolutions (especially over 128x128 and assuming the density isn't set super low, at least.) What sort of map sizes are you planning on passing into this and with what density? Other implementations seem to work quite differently, and the sorting is definitely the majority of the processing time here. If you can just confirm that it's actually going to work for what you need I'll merge it. |
im working on parallelizing it, will look into diffrent solutions |
Generation lost a tiny bit of it's fidelity but oh well, looks fantastic rn
i absolutely reworked how this noise works so it is very very fast compared to the previous implementation |
|
As discussed in coderbus, it's not the most idiomatic Rust code but it's self contained and there's no real reason not to let it in. |
* Updates the byond testing version to tgstation ver (tgstation#66) * runs cargo fmt (tgstation#63) * Sets the debug and test linux build to use all features (tgstation#67) Problem: PRs and all features aren't properly running created unit tests. Nor are they actually being fully compiled (only checked) on Linux. Solution: Only have the release builds that are distributed as artifacts run with the default features. This way, tests can cover all features. * Base64 encoding (tgstation#65) * Adds base64 encoding * Adds tests * Commented out import * Fixes the naming of variables * ...of course it was a typo * Update Cargo.toml Co-authored-by: William Wallace <[email protected]> * Clippy tweaks (tgstation#62) * this `else { if .. }` block can be collapsed * question mark operator is useless here * this call to `from_str_radix` can be replaced with a call to `str::parse` * equality checks against true are unnecessary * Revert "question mark operator is useless here" This reverts commit 01272c5. * Cargo update (tgstation#64) * updates cargo dependencies versions to latest * Fixes perlin noise generation to use an inclusive range * Update src/cellularnoise.rs Co-authored-by: William Wallace <[email protected]> * Updates cargo.lock Co-authored-by: William Wallace <[email protected]> * Basic version getter (tgstation#68) * Bump Version sto v0.4.8 * Setup cross compiling using the cross project (tgstation#70) * Make the url feature default (tgstation#71) Co-authored-by: Jared-Fogle <[email protected]> * 0.4.9 * Runs cargo update (tgstation#72) most importantly git2, which made it error when trying to build on linux with the latest rust version * Stringify url_encode and decode input before sending it to Rust (tgstation#73) Co-authored-by: Jordan Brown <[email protected]> * 0.4.10 * Add worley noise optional feature (tgstation#74) * worley_noise * a simple fix, forgot to ad dmsort as a necessary package * replaces .unwrap() with ? * forgot to save :cryingintoheavens: * IM SORRY I CANT USE ? IN CLOSURES * Completely reworks how the code works, massively optimizing it. Generation lost a tiny bit of it's fidelity but oh well, looks fantastic rn * one last change * tfw errors * makes the noise generate pretties noise * cargo fmt * Add TOML feature (tgstation#75) Adds rustg_read_toml_file, which takes a filepath and spits out the output after decoding (a list, probably). Uses the rust-toml lib * 0.5.0 Cargo updates * Adds TOTP generator to the hash feature (tgstation#76) * Adds TOTP generator to the hash feature * Updated TOTP based on requested changes, added a function that allow specification of a tolerance, also updated the hash.dm file to reflect the new functions * Remove debug print, convert offset output to be JSON, added a test * Added comments * Added some error handling * Improves error handling again * cargo fmt hash.rs * separate mod block for hash tests * Functions for more precise time measurement (tgstation#77) * fix linux ci by removing pkg-config requirement This isnt needed * Aho-Corasick string replacements to help clean up replaceText spam (tgstation#82) * Cleanups and additions for the Aho-Corasick replacement stuff (tgstation#83) * Fix clippy lints, Rustfmt, put both on CI, and update Cargo.lock (tgstation#85) Co-authored-by: Mothblocks <[email protected]> * Adds Redis Pub/Sub integration (tgstation#80) dds the ability to connect to Redis, subscribe to messages and publish them as well. See here for example usage. The API, prefixed with rustg_redis_: connect(addr) - Connects to a Redis instance using the given address, for example redis://127.0.0.1/. Returns an empty string on success, returns the error otherwise. disconnect() - Closes the connection to Redis and stops the thread managing it. Call this before restarting or attempting to reconnect after an error. subscribe(channel) - Subscribes to a given channel and starts receiving messages from it. get_messages() - Returns all received messages as a JSON string, in the format of {"channel_1": ["msg1", "msg2", "msg3", ...], "channel_2": ["msg1", "msg2", "msg3", ...]}. Also includes errors, which appear on the channel "RUSTG_REDIS_ERROR_CHANNEL". publish(channel, message) - Publishes a message on the given channel. Remember to check the error channel every time you call get_messages(). If any occur, you need to call disconnect(), then connect() again, then resubscribe to desired channels. * fixes warning in redis (tgstation#88) * Fix clippy CI, format byond_fn (tgstation#89) * Adds alphabetical order tests for README.md, lib.rs and Cargo.toml (tgstation#84) Adds alphabetical order tests for README.md, lib.rs and Cargo.toml tgstation#84 Also renamed "non-default features" to "additional features" in Cargo.toml for consistency with readme updated the cargo.lock and ran rustfmt * Bump to 0.6.0 (tgstation#90) * Fixes redis dmsrc (tgstation#91) * Cleaner and saner http request parsing. (tgstation#93) * Add functions for seeking lines and getting line counts in functions (tgstation#95) dds rustg_file_get_line_count for getting the line count of a given filename, and rustg_file_seek_line for getting a specific line. These are useful as to not load the entire file into memory in DM, which is useful for very large files like dictionaries. * 0.7.0 (tgstation#96) * Adds a new type of noise: Discrete Batched Perlin-like Noise (DBP) (tgstation#99) * Updates WorleyNoise to be multi-threaded and faster (tgstation#102) Another noise algorithm updated in my streak, this time it's worleynoise, since i didn't know jack shit and i heavily abused Rc<> to acomplish my goals it was really shit. Now it is multi-threaded, blazing fast and really sexy lookin. NOTE: this update changes ABI of worley noise slightly. * Updates CellularNoise to be faster and multi-threaded (tgstation#101) Hello, I guess i'm on a streak. I reimplemented CAnoise to be multi-threaded and much faster as i removed a considerable amount of branches. No change to behaviour or ABI noted, it's just faster now. * toml2json now returns errors to byond (tgstation#98) Co-authored-by: ZeWaka <[email protected]> * Updates cargo packages (tgstation#104) * 0.8.0 * Tiny README update * Small improvements for before 1.0 (tgstation#105) * 1.0.0 (tgstation#106) * Pallette Encoding Image Fix (tgstation#107) * Uploads rust_g.dm as an Actions Artifact (tgstation#108) * 1.0.1 * add note to README about not using the .so * Returns the TOML library to previous functionality (tgstation#112) * 1.0.2 * Clippy - unneeded returns (tgstation#114) * Add an a* pathfinder (tgstation#113) Returns the shortest path in a static node map. That is made for TGMC wich uses manually placed nodes for pathfinding. Benchmark : image That's the average number of path computed in one second, out of 30 runs with random nodes. Size of the node map is roughly 800 On average 32 times faster than tgmc a* implementation. Another possible comparison is with TG's JPS system, and according to the benchmark done here it's 7000 times faster (tgstation/tgstation#56780). Not exactly the same use cases though * Add rustg_toml_encode (tgstation#116) Add rustg_toml_encode * 1.1.0 * Fixes cellular noise not working for diff height and width values (tgstation#117) * Update README.md update readme with `pathfinder` feature and rust-analyzer note, re: tgstation#115 * Fix the `pathfinder` feature's loc in the README * Package debug symbols with rust-g (tgstation#119) * 515 Support (tgstation#121) * add a unix_timestamp extern (tgstation#122) Co-authored-by: Mothblocks <[email protected]> * 1.2.0 * Remove useless features from the image crate (tgstation#128) * lmao we don't need jpeg dependencies for image * update `noise` to remove the rest of the junk deps * small clippy fixes (cherry picked from commit 65cf48a) * Update rust.yml * adds all-features flag to linux and windows build * reverts actions to tg's current * Update rust.yml * Update rust.yml * Update rust.yml * Update rust.yml --------- Co-authored-by: ZeWaka <[email protected]> Co-authored-by: William Wallace <[email protected]> Co-authored-by: oranges <[email protected]> Co-authored-by: Mothblocks <[email protected]> Co-authored-by: Jared-Fogle <[email protected]> Co-authored-by: Mark Suckerberg <[email protected]> Co-authored-by: Gamer025 <[email protected]> Co-authored-by: Jordan Brown <[email protected]> Co-authored-by: EdgeLordExe <[email protected]> Co-authored-by: adamsong <[email protected]> Co-authored-by: pali <[email protected]> Co-authored-by: AffectedArc07 <[email protected]> Co-authored-by: vuonojenmustaturska <[email protected]> Co-authored-by: MCHSL <[email protected]> Co-authored-by: AnturK <[email protected]> Co-authored-by: tralezab <[email protected]> Co-authored-by: san7890 <[email protected]> Co-authored-by: BraveMole <[email protected]> Co-authored-by: Zephyr <[email protected]>
Doing worley noise in DM is a performance suicide rust-g is the only option.
Worley Noise!
This noise works by creating regions with nodes that are later mapped onto a set grid.
Each cell on the grid finds the nearest node in neighbouring and it's own region, it then finds the second closest and subtracts the distance one from another.
If it passes a threshold then congratulations, it is set to true.
Samples