Performance patches prerequisite

[Spottedleaf]

Original patches:
https://github.com/Spottedleaf/Concrete/blob/bleeding/patches/server/0004-Util-patch.patch
https://github.com/Spottedleaf/Concrete/blob/bleeding/patches/server/0022-Improve-autosave-performance.patch
https://github.com/Spottedleaf/Concrete/blob/bleeding/patches/server/0023-Avoid-immediately-autosaving-newly-generated-loaded-.patch

See patch notes for details
The patches aren't always 1:1 ports, they may have been changed to improve behavior or performance in some areas

They've all been dumped in this PR since they tend to have dependencies on each other (most notably the mc-util patch)

[electronicboy]

pre-looking at the patches comment: gfdi leaf

[Spottedleaf]

now it's just the patches that cause big diffs

[AJMFactsheets]

I built this branch for local testing and the only problem I found is that when a player dies, there is a high chance they will get stuck in a limbo mode where chunks won't load, entities jitter, and they player can't move.

[Spottedleaf]

can't produce locally, paste output of /version

[HexedHero]

#LeafForProjectDev

[electronicboy]

mostly looks like your server is overworked from mob spawning/player movement

[Spottedleaf]

yeah I don't see much of a difference between the two, and without a testing environment I can't do anything about it

[AJMFactsheets]

Output of /version: This server is running Paper version git-Paper-"59df758f" (MC: 1.15.1) (Implementing API version 1.15.1-R0.1-SNAPSHOT)

After further testing, I have found that the death bug only happens after a player has slept in a bed.

[Spottedleaf]

unable to produce after sleeping in a bed

[AJMFactsheets]

I built a new jar based off of the new commit and the death glitch still persists. I realized beds have nothing to do with the glitch, but it has something to do with spawn chunks.

Start a new world with default settings, and /kill yourself at the world spawn. Upon respawning, you should be unable to move, unless you relog or teleport far away.

The glitch reminds me of MC-138539
I also tested with 1.15.1 Paper-31 and didn't have any problems.

[Spottedleaf]

Yeah I see what's going on here now - a typo. The reason it only happens at spawn is because once chunks get loaded they are sent to the player - at spawn they only load once.

Fixed in latest commit.

Thanks for your patience and detailed report.

[electronicboy]

• Spottedleaf/Concrete:patches/server/0004-Util-patch.patch@bleeding

• Spottedleaf/Concrete:patches/server/0022-Improve-autosave-performance.patch@bleeding

• Spottedleaf/Concrete:patches/server/0023-Avoid-immediately-autosaving-newly-generated-loaded-.patch@bleeding

• Spottedleaf/Concrete:patches/server/0013-Per-player-view-distance-implementation.patch@bleeding

• Spottedleaf/Concrete:patches/server/0015-Optimize-isOutsideRange-to-use-a-square-radius-check.patch@bleeding

• Spigot-Server-Patches/0420-Optimise-getChunkAt-calls-for-loaded-chunks.patch

[Spottedleaf]

updated to 1.15.2, fixed build

[electronicboy]

https://www.youtube.com/watch?v=woySeSNBL3o

[Misio12320]

How much % the game has been upgraded with performance and when release of view distance for 1.15.2?

[perseus5]

How much % the game has been upgraded with performance and when release of view distance for 1.15.2?

yes

[Spottedleaf]

it's a prereq patch so like maybe 2% at most

[jonttm]

Are you going to bring the per-player view distance patch back in the future?

[Spottedleaf]

yes, some other stuff needs to go in first though

[NicoNekoDev]

it's a prereq patch so like maybe 2% at most

that's a lot :O gonna test it now on my 20gb server, i'll report back the performance.
Edit: here's is the performance report: https://timings.aikar.co/?id=8fe23d8d5ecf4518a4c5d132e73e213a (i've maxed to 32 gb)
The only thing I've noticed was the chunks loading in a V form when I travel fast.
My only problem is the new chunks beeing generated very slow, also it looks like I do have somewhere a memory leak, as the server ram start to increase over time.