Mirrors/Paper
3
0
Fork 1
Mirror von https://github.com/PaperMC/Paper.git synchronisiert 2024-12-16 19:40:07 +01:00
Code Issues Projekte Releases Wiki Aktivität
3acf9df3be
Paper/Spigot-Server-Patches/0090-Optimize-Chunk-Unload-Queue.patch

233 Zeilen
11 KiB
Diff
Originalformat Normale Ansicht Verlauf

Remove patch pulled upstream and rebuild patches
2016-03-19 23:52:09 +01:00
From 3f28611a24fa1d3b92132c6755774d90f0cce970 Mon Sep 17 00:00:00 2001
Optimize Chunk Unload Queue Removing chunks from the unload queue when performing chunk lookups is a costly activity. It drastically slows down server performance as many methods call getChunkAt, resulting in a bandaid to skip removing chunks from the unload queue. This patch optimizes the unload queue to instead use a boolean on the Chunk object itself to mark if the chunk is active, and then insert into a LinkedList queue. The benefits here is that all chunk unload queue actions are now O(1) constant time. A LinkedList will never need to resize, and can be removed from in constant time when used in a queue like method. We mark the chunk as active in many places that notify it is still being used, so that when the chunk unload queue reaches that chunk, and sees the chunk became active again, it will skip it and move to next.
2016-03-18 23:03:44 +01:00
From: Aikar <aikar@aikar.co>
Date: Fri, 18 Mar 2016 17:57:25 -0400
Subject: [PATCH] Optimize Chunk Unload Queue
Removing chunks from the unload queue when performing chunk lookups is a costly activity.
It drastically slows down server performance as many methods call getChunkAt, resulting in a bandaid
to skip removing chunks from the unload queue.
This patch optimizes the unload queue to instead use a boolean on the Chunk object itself to mark
if the chunk is active, and then insert into a LinkedList queue.
The benefits here is that all chunk unload queue actions are now O(1) constant time.
A LinkedList will never need to resize, and can be removed from in constant time when
used in a queue like method.
We mark the chunk as active in many places that notify it is still being used, so that
when the chunk unload queue reaches that chunk, and sees the chunk became active again,
it will skip it and move to next.
diff --git a/src/main/java/net/minecraft/server/Chunk.java b/src/main/java/net/minecraft/server/Chunk.java
index b6d84d7..a4c0b4e 100644
--- a/src/main/java/net/minecraft/server/Chunk.java
+++ b/src/main/java/net/minecraft/server/Chunk.java
@@ -55,6 +55,8 @@ public class Chunk {
// Keep this synced with entitySlices.add() and entitySlices.remove()
private final int[] itemCounts = new int[16];
private final int[] inventoryEntityCounts = new int[16];
+ public boolean isChunkActive = true;
+ public boolean isInUnloadQueue = false;
// Paper end
// CraftBukkit start - Neighbor loaded cache for chunk lighting and entity ticking
@@ -780,6 +782,7 @@ public class Chunk {
}
public void addEntities() {
+ isChunkActive = true; // Paper
this.i = true;
this.world.b(this.tileEntities.values());
@@ -798,6 +801,7 @@ public class Chunk {
}
public void removeEntities() {
+ isChunkActive = false; // Paper
this.i = false;
Iterator iterator = this.tileEntities.values().iterator();
diff --git a/src/main/java/net/minecraft/server/ChunkProviderServer.java b/src/main/java/net/minecraft/server/ChunkProviderServer.java
Fix chunk unload issues - Resolves #97
2016-03-19 16:29:46 +01:00
index 9ef6246..247a6dd 100644
Optimize Chunk Unload Queue Removing chunks from the unload queue when performing chunk lookups is a costly activity. It drastically slows down server performance as many methods call getChunkAt, resulting in a bandaid to skip removing chunks from the unload queue. This patch optimizes the unload queue to instead use a boolean on the Chunk object itself to mark if the chunk is active, and then insert into a LinkedList queue. The benefits here is that all chunk unload queue actions are now O(1) constant time. A LinkedList will never need to resize, and can be removed from in constant time when used in a queue like method. We mark the chunk as active in many places that notify it is still being used, so that when the chunk unload queue reaches that chunk, and sees the chunk became active again, it will skip it and move to next.
2016-03-18 23:03:44 +01:00
--- a/src/main/java/net/minecraft/server/ChunkProviderServer.java
+++ b/src/main/java/net/minecraft/server/ChunkProviderServer.java
@@ -21,7 +21,7 @@ import org.bukkit.event.world.ChunkUnloadEvent;
public class ChunkProviderServer implements IChunkProvider {
private static final Logger a = LogManager.getLogger();
- public final LongHashSet unloadQueue = new LongHashSet(); // CraftBukkit - LongHashSet
+ public final ChunkUnloadQueue unloadQueue = new ChunkUnloadQueue(); // CraftBukkit - LongHashSet // Paper -> ChunkUnloadQueue
public final ChunkGenerator chunkGenerator; // CraftBukkit - public
private final IChunkLoader chunkLoader;
public LongObjectHashMap<Chunk> chunks = new LongObjectHashMap<Chunk>(); // CraftBukkit
Fix chunk unload issues - Resolves #97
2016-03-19 16:29:46 +01:00
@@ -79,18 +79,24 @@ public class ChunkProviderServer implements IChunkProvider {
Optimize Chunk Unload Queue Removing chunks from the unload queue when performing chunk lookups is a costly activity. It drastically slows down server performance as many methods call getChunkAt, resulting in a bandaid to skip removing chunks from the unload queue. This patch optimizes the unload queue to instead use a boolean on the Chunk object itself to mark if the chunk is active, and then insert into a LinkedList queue. The benefits here is that all chunk unload queue actions are now O(1) constant time. A LinkedList will never need to resize, and can be removed from in constant time when used in a queue like method. We mark the chunk as active in many places that notify it is still being used, so that when the chunk unload queue reaches that chunk, and sees the chunk became active again, it will skip it and move to next.
2016-03-18 23:03:44 +01:00
// CraftBukkit start - Add async variant, provide compatibility
public Chunk getOrCreateChunkFast(int x, int z) {
- Chunk chunk = chunks.get(LongHash.toLong(x, z));
- return (chunk == null) ? getChunkAt(x, z) : chunk;
+ return getChunkAt(x, z); // Paper
}
Fix chunk unload issues - Resolves #97
2016-03-19 16:29:46 +01:00
+ // Paper start
+ public Chunk getLoadedChunkAtWithoutMarkingActive(int i, int j) {
+ return chunks.get(LongHash.toLong(i, j));
+ }
+ // Paper end
+
Optimize Chunk Unload Queue Removing chunks from the unload queue when performing chunk lookups is a costly activity. It drastically slows down server performance as many methods call getChunkAt, resulting in a bandaid to skip removing chunks from the unload queue. This patch optimizes the unload queue to instead use a boolean on the Chunk object itself to mark if the chunk is active, and then insert into a LinkedList queue. The benefits here is that all chunk unload queue actions are now O(1) constant time. A LinkedList will never need to resize, and can be removed from in constant time when used in a queue like method. We mark the chunk as active in many places that notify it is still being used, so that when the chunk unload queue reaches that chunk, and sees the chunk became active again, it will skip it and move to next.
2016-03-18 23:03:44 +01:00
public Chunk getChunkIfLoaded(int x, int z) {
- return chunks.get(LongHash.toLong(x, z));
+ return getLoadedChunkAt(x, z); // Paper - Bukkit has a duplicate method now.
}
public Chunk getLoadedChunkAt(int i, int j) {
Chunk chunk = chunks.get(LongHash.toLong(i, j)); // CraftBukkit
this.unloadQueue.remove(i, j); // CraftBukkit
+ if (chunk != null) { chunk.isChunkActive = true; }// Paper
return chunk;
}
Fix chunk unload issues - Resolves #97
2016-03-19 16:29:46 +01:00
@@ -151,6 +157,7 @@ public class ChunkProviderServer implements IChunkProvider {
Optimize Chunk Unload Queue Removing chunks from the unload queue when performing chunk lookups is a costly activity. It drastically slows down server performance as many methods call getChunkAt, resulting in a bandaid to skip removing chunks from the unload queue. This patch optimizes the unload queue to instead use a boolean on the Chunk object itself to mark if the chunk is active, and then insert into a LinkedList queue. The benefits here is that all chunk unload queue actions are now O(1) constant time. A LinkedList will never need to resize, and can be removed from in constant time when used in a queue like method. We mark the chunk as active in many places that notify it is still being used, so that when the chunk unload queue reaches that chunk, and sees the chunk became active again, it will skip it and move to next.
2016-03-18 23:03:44 +01:00
runnable.run();
}
+ chunk.isChunkActive = true; // Paper
return chunk;
}
Fix chunk unload issues - Resolves #97
2016-03-19 16:29:46 +01:00
@@ -201,7 +208,7 @@ public class ChunkProviderServer implements IChunkProvider {
continue;
}
- Chunk neighbor = this.getChunkIfLoaded(chunk.locX + x, chunk.locZ + z);
+ Chunk neighbor = this.getLoadedChunkAtWithoutMarkingActive(chunk.locX + x, chunk.locZ + z); // Paper
if (neighbor != null) {
neighbor.setNeighborLoaded(-x, -z);
chunk.setNeighborLoaded(x, z);
@@ -300,10 +307,17 @@ public class ChunkProviderServer implements IChunkProvider {
Optimize Chunk Unload Queue Removing chunks from the unload queue when performing chunk lookups is a costly activity. It drastically slows down server performance as many methods call getChunkAt, resulting in a bandaid to skip removing chunks from the unload queue. This patch optimizes the unload queue to instead use a boolean on the Chunk object itself to mark if the chunk is active, and then insert into a LinkedList queue. The benefits here is that all chunk unload queue actions are now O(1) constant time. A LinkedList will never need to resize, and can be removed from in constant time when used in a queue like method. We mark the chunk as active in many places that notify it is still being used, so that when the chunk unload queue reaches that chunk, and sees the chunk became active again, it will skip it and move to next.
2016-03-18 23:03:44 +01:00
if (!this.world.savingDisabled) {
// CraftBukkit start
Server server = this.world.getServer();
- for (int i = 0; i < 100 && !this.unloadQueue.isEmpty(); ++i) {
- long chunkcoordinates = this.unloadQueue.popFirst();
- Chunk chunk = this.chunks.get(chunkcoordinates);
- if (chunk == null) continue;
+ // Paper start
+ final Iterator<Chunk> iterator = this.unloadQueue.iterator();
+ for (int i = 0; i < 100 && iterator.hasNext(); ++i) {
+ Chunk chunk = iterator.next();
+ iterator.remove();
+ long chunkcoordinates = LongHash.toLong(chunk.locX, chunk.locZ);
+ chunk.isInUnloadQueue = false;
+ if (chunk.isChunkActive) {
+ continue;
+ }
+ // Paper end
ChunkUnloadEvent event = new ChunkUnloadEvent(chunk.bukkitChunk);
server.getPluginManager().callEvent(event);
Fix chunk unload issues - Resolves #97
2016-03-19 16:29:46 +01:00
@@ -325,7 +339,7 @@ public class ChunkProviderServer implements IChunkProvider {
continue;
}
- Chunk neighbor = this.getChunkIfLoaded(chunk.locX + x, chunk.locZ + z);
+ Chunk neighbor = this.getLoadedChunkAtWithoutMarkingActive(chunk.locX + x, chunk.locZ + z); // Paper
if (neighbor != null) {
neighbor.setNeighborUnloaded(-x, -z);
chunk.setNeighborUnloaded(x, z);
@@ -367,4 +381,22 @@ public class ChunkProviderServer implements IChunkProvider {
Optimize Chunk Unload Queue Removing chunks from the unload queue when performing chunk lookups is a costly activity. It drastically slows down server performance as many methods call getChunkAt, resulting in a bandaid to skip removing chunks from the unload queue. This patch optimizes the unload queue to instead use a boolean on the Chunk object itself to mark if the chunk is active, and then insert into a LinkedList queue. The benefits here is that all chunk unload queue actions are now O(1) constant time. A LinkedList will never need to resize, and can be removed from in constant time when used in a queue like method. We mark the chunk as active in many places that notify it is still being used, so that when the chunk unload queue reaches that chunk, and sees the chunk became active again, it will skip it and move to next.
2016-03-18 23:03:44 +01:00
public boolean e(int i, int j) {
return this.chunks.containsKey(LongHash.toLong(i, j)); // CraftBukkit
}
+
+ // Paper start
+ public class ChunkUnloadQueue extends java.util.LinkedList<Chunk> {
+ public void remove(int x, int z) {
+ // nothing! Just to reduce diff
+ }
+ public void add(int x, int z) {
+ final Chunk chunk = chunks.get(LongHash.toLong(x, z));
+ if (chunk != null) {
+ chunk.isChunkActive = false;
+ if (!chunk.isInUnloadQueue) {
+ chunk.isInUnloadQueue = true;
+ add(chunk);
+ }
+ }
+ }
+ }
+ // Paper end
}
diff --git a/src/main/java/net/minecraft/server/SpawnerCreature.java b/src/main/java/net/minecraft/server/SpawnerCreature.java
index 63e118d..721bcae 100644
--- a/src/main/java/net/minecraft/server/SpawnerCreature.java
+++ b/src/main/java/net/minecraft/server/SpawnerCreature.java
@@ -28,7 +28,7 @@ public final class SpawnerCreature {
Long coord = it.next();
int x = LongHash.msw( coord );
int z = LongHash.lsw( coord );
- if ( !((ChunkProviderServer)server.chunkProvider).unloadQueue.contains( coord ) && server.isChunkLoaded( x, z, true ) )
+ if ( server.isChunkLoaded( x, z, true ) ) // Paper
{
i += server.getChunkAt( x, z ).entityCount.get( oClass );
}
diff --git a/src/main/java/org/bukkit/craftbukkit/CraftWorld.java b/src/main/java/org/bukkit/craftbukkit/CraftWorld.java
Remove patch pulled upstream and rebuild patches
2016-03-19 23:52:09 +01:00
index be311cd..66d3e94 100644
Optimize Chunk Unload Queue Removing chunks from the unload queue when performing chunk lookups is a costly activity. It drastically slows down server performance as many methods call getChunkAt, resulting in a bandaid to skip removing chunks from the unload queue. This patch optimizes the unload queue to instead use a boolean on the Chunk object itself to mark if the chunk is active, and then insert into a LinkedList queue. The benefits here is that all chunk unload queue actions are now O(1) constant time. A LinkedList will never need to resize, and can be removed from in constant time when used in a queue like method. We mark the chunk as active in many places that notify it is still being used, so that when the chunk unload queue reaches that chunk, and sees the chunk became active again, it will skip it and move to next.
2016-03-18 23:03:44 +01:00
--- a/src/main/java/org/bukkit/craftbukkit/CraftWorld.java
+++ b/src/main/java/org/bukkit/craftbukkit/CraftWorld.java
Fix chunk unload issues - Resolves #97
2016-03-19 16:29:46 +01:00
@@ -232,7 +232,7 @@ public class CraftWorld implements World {
continue;
}
- net.minecraft.server.Chunk neighbor = world.getChunkProviderServer().getChunkIfLoaded(chunk.locX + x, chunk.locZ + z);
+ net.minecraft.server.Chunk neighbor = world.getChunkProviderServer().getLoadedChunkAtWithoutMarkingActive(chunk.locX + x, chunk.locZ + z); // Paper
if (neighbor != null) {
neighbor.setNeighborUnloaded(-xx, -zz);
chunk.setNeighborUnloaded(xx, zz);
Optimize Chunk Unload Queue Removing chunks from the unload queue when performing chunk lookups is a costly activity. It drastically slows down server performance as many methods call getChunkAt, resulting in a bandaid to skip removing chunks from the unload queue. This patch optimizes the unload queue to instead use a boolean on the Chunk object itself to mark if the chunk is active, and then insert into a LinkedList queue. The benefits here is that all chunk unload queue actions are now O(1) constant time. A LinkedList will never need to resize, and can be removed from in constant time when used in a queue like method. We mark the chunk as active in many places that notify it is still being used, so that when the chunk unload queue reaches that chunk, and sees the chunk became active again, it will skip it and move to next.
2016-03-18 23:03:44 +01:00
@@ -302,7 +302,7 @@ public class CraftWorld implements World {
world.timings.syncChunkLoadTimer.startTiming(); // Spigot
chunk = world.getChunkProviderServer().getOrLoadChunkAt(x, z);
world.timings.syncChunkLoadTimer.stopTiming(); // Spigot
- }
+ } else { chunk.isChunkActive = true; } // Paper
return chunk != null;
}
Fix chunk unload issues - Resolves #97
2016-03-19 16:29:46 +01:00
@@ -319,7 +319,7 @@ public class CraftWorld implements World {
continue;
}
- net.minecraft.server.Chunk neighbor = world.getChunkProviderServer().getChunkIfLoaded(chunk.locX + x, chunk.locZ + z);
+ net.minecraft.server.Chunk neighbor = world.getChunkProviderServer().getLoadedChunkAtWithoutMarkingActive(chunk.locX + x, chunk.locZ + z); // Paper
if (neighbor != null) {
neighbor.setNeighborLoaded(-x, -z);
chunk.setNeighborLoaded(x, z);
Optimize Chunk Unload Queue Removing chunks from the unload queue when performing chunk lookups is a costly activity. It drastically slows down server performance as many methods call getChunkAt, resulting in a bandaid to skip removing chunks from the unload queue. This patch optimizes the unload queue to instead use a boolean on the Chunk object itself to mark if the chunk is active, and then insert into a LinkedList queue. The benefits here is that all chunk unload queue actions are now O(1) constant time. A LinkedList will never need to resize, and can be removed from in constant time when used in a queue like method. We mark the chunk as active in many places that notify it is still being used, so that when the chunk unload queue reaches that chunk, and sees the chunk became active again, it will skip it and move to next.
2016-03-18 23:03:44 +01:00
@@ -1538,7 +1538,7 @@ public class CraftWorld implements World {
}
// Already unloading?
- if (cps.unloadQueue.contains(chunk.locX, chunk.locZ)) {
+ if (!chunk.isChunkActive) { // Paper
continue;
}
Fix chunk unload issues - Resolves #97
2016-03-19 16:29:46 +01:00
diff --git a/src/main/java/org/bukkit/craftbukkit/chunkio/ChunkIOProvider.java b/src/main/java/org/bukkit/craftbukkit/chunkio/ChunkIOProvider.java
index 482af17..a1a6d5a 100644
--- a/src/main/java/org/bukkit/craftbukkit/chunkio/ChunkIOProvider.java
+++ b/src/main/java/org/bukkit/craftbukkit/chunkio/ChunkIOProvider.java
@@ -62,7 +62,7 @@ class ChunkIOProvider implements AsynchronousExecutor.CallBackProvider<QueuedChu
continue;
}
- Chunk neighbor = queuedChunk.provider.getChunkIfLoaded(chunk.locX + x, chunk.locZ + z);
+ Chunk neighbor = queuedChunk.provider.getLoadedChunkAtWithoutMarkingActive(chunk.locX + x, chunk.locZ + z); // Paper
if (neighbor != null) {
neighbor.setNeighborLoaded(-x, -z);
chunk.setNeighborLoaded(x, z);
Optimize Chunk Unload Queue Removing chunks from the unload queue when performing chunk lookups is a costly activity. It drastically slows down server performance as many methods call getChunkAt, resulting in a bandaid to skip removing chunks from the unload queue. This patch optimizes the unload queue to instead use a boolean on the Chunk object itself to mark if the chunk is active, and then insert into a LinkedList queue. The benefits here is that all chunk unload queue actions are now O(1) constant time. A LinkedList will never need to resize, and can be removed from in constant time when used in a queue like method. We mark the chunk as active in many places that notify it is still being used, so that when the chunk unload queue reaches that chunk, and sees the chunk became active again, it will skip it and move to next.
2016-03-18 23:03:44 +01:00
--
Remove patch pulled upstream and rebuild patches
2016-03-19 23:52:09 +01:00
2.7.3
Optimize Chunk Unload Queue Removing chunks from the unload queue when performing chunk lookups is a costly activity. It drastically slows down server performance as many methods call getChunkAt, resulting in a bandaid to skip removing chunks from the unload queue. This patch optimizes the unload queue to instead use a boolean on the Chunk object itself to mark if the chunk is active, and then insert into a LinkedList queue. The benefits here is that all chunk unload queue actions are now O(1) constant time. A LinkedList will never need to resize, and can be removed from in constant time when used in a queue like method. We mark the chunk as active in many places that notify it is still being used, so that when the chunk unload queue reaches that chunk, and sees the chunk became active again, it will skip it and move to next.
2016-03-18 23:03:44 +01:00
In neuem Issue referenzieren Permalink kopieren