Fabricae Ex Nihilo

Fabricae Ex Nihilo

Mods
152k Downloads

Strainers cause a lot of network congestion

quat1024 opened this issue · 2 comments

commented

Strainers, if they are waterlogged and ready to work, always call markDirty from tick:

FabricaeExNihilo/src/main/java/wraith/fabricaeexnihilo/modules/strainer/StrainerBlockEntity.java

Line 94 in f5e0050

strainer.markDirty();

BaseBlockEntity always calls markForUpdate from markDirty:

FabricaeExNihilo/src/main/java/wraith/fabricaeexnihilo/modules/base/BaseBlockEntity.java

Lines 20 to 25 in f5e0050

public void markDirty() {
super.markDirty();
if (this.world != null && this.world instanceof ServerWorld serverWorld) {
var chunkManager = serverWorld.getChunkManager();
if (chunkManager != null) {
chunkManager.markForUpdate(pos);

And markForUpdate causes a packet to be dispatched to nearby players, so just like that, 20 packets per second per strainer :(

It looks like barrels also sync in both branches of an if/else:

FabricaeExNihilo/src/main/java/wraith/fabricaeexnihilo/modules/barrels/BarrelBlockEntity.java

Lines 189 to 193 in f5e0050

markDirty();
tickRecipe();
} else {
--tickCounter;
markDirty();

I understand that markDirty should be called because timers are being ticked down earlier in these tickers, but it should probably not result in a full sync of the entire NBT tag to clients !

I'm writing a packet logger and strainer syncs are one of the most commonly received packets across all received packets at Blanketcon. Fireblanket includes an optimization that prevents a block entity from syncing the same NBT tag twice in a row, but since these tags include the ticking-down timer, that optimization can't kick in :(

commented

Fixed in v1.0.0

  • 🎉1
commented

Yeah. Overall the sync code just lasily syncs everything (lazy as in lazy dev). It would be pretty easy to only send necessary data and only when it updates.