Usage of sets is causing performance issues

[Steveplays28]

Tidal/src/main/java/net/superkat/tidal/wave/TidalWaveHandler.java

Line 111 in d6434aa

updatedCovered.addAll(wave.getCoveredBlocks());

Tidal/src/main/java/net/superkat/tidal/wave/Wave.java

Line 120 in d6434aa

public Set<BlockPos> getCoveredBlocks() {

Adding sets to other sets is slow, see this Spark profile.
I'm not sure how this should be optimised yet.

[Superkat32]

Yeah, that whole method was a pretty janky, last minute and last hope solution.

What's worse is I have no clue how to fix it either. I was hoping Minecraft's Box class(used for a Wave's hitbox) had a way to find all intersecting BlockPos's but I couldn't find one.

I don't think a List would work great either because then the wet block overlay renderer would render on the same blocks multiple times, which isn't what I want(there's a chance that'll look nice, but I'm afraid it might also be too much at times).

If you happen to know, would replacing the addAll call with an iterator (maybe even forEach ? ) to add each block individually be any faster?

[Steveplays28]

Iterators are slow. foreach is also slower than fori because it uses iterators.
Iterators are slow because they allocate a lot of objects, sadly Java does not have zero-cost abstractions and good compiler optimisations like native-compiled languages, afaik.

I'll look into how to optimise this

[Superkat32]

Another idea I've just had - if I used a List (even better, a fastutil List) then cast it to a Set, would that remove the duplicates while still being faster? Possibly something we could experiment with if it works.

[Steveplays28]

I don't think that'll work. Sets probably check for duplicates in any case (which is why it's currently slow), that's how they work.
If we want to optimise this, we'll need to find a way to check for duplicates more efficiently (removing the need for sets).