Sync-prepared batches account for the vast majority of outbound packets. Avoiding these useless objects further reduces the overhead of zero-compressed packets, as the creation of these objects is a significant part of the overhead for these cases.
closes#6157
this is supported since libdeflate 1.15 and ext-libdeflate 0.2.0.
Everyone should be using these versions by now anyway, and if they aren't, they should update.
libdeflate's level 0 compression is over 20 times faster than zlib, so this is a nice performance improvement.
this came up during 1.20.50 testing, where minecraft:stone[stone_type=stone] continues to have an ID minecraft:stone without the stone_type property.
This could have appeared in other ways, such as with an invalid legacy blockstate.
we can safely assume this for blocks (though the type info doesn't reflect it) but this is not safe to assume for random APIs that might be used by plugins.
if there's no state data to encode, we can avoid useless calls and object allocations.
For the best cases (blocks which don't use state data at all) this improves the performance of getStateId() by more than 10x.
Blocks which use one or the other benefit by a smaller but still significant margin.
closes#6070
there are some unresolved questions about the growth speed of beetroots, pitcher plants and torchflower crops, but that's a topic for another commit.
this change also doesn't account for the light levels.
during unset(), the destructors for other objects with cyclic references can get triggered, resulting in the functions being reentered before the count() call. This leads to a crash because the offset no longer exists.
Instead, we check if only the given PermissibleInternal is present, and clean everything up with a single unset instead of two.
This could also have been solved by adding extra isset() checks before checking the counts, but this way seemed more elegant.
This is similar to an issue with AsyncTask thread-local storage a few months ago, which was also caused by GC reentrancy.
closes#6119
