13
0
geforkt von Mirrors/Paper
Paper/Spigot-Server-Patches/0071-Optimize-isValidLocation-getType-and-getBlockData-fo.patch

207 Zeilen
9.9 KiB
Diff

From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
2019-04-27 05:05:36 +02:00
From: Aikar <aikar@aikar.co>
Date: Thu, 3 Mar 2016 02:07:55 -0600
Optimize Light Engine Massive update to light to improve performance and chunk loading/generation. 1) Massive bit packing/unpacking optimizations and inlining. A lot of performance has to do with constant packing and unpacking of bits. We now inline a most bit operations, and re-use base x/y/z bits in many places. This helps with cpu level processing to just do all the math at once instead of having to jump in and out of function calls. This much logic also is likely over the JVM Inline limit for JIT too. 2) Applied a few of JellySquid's Phosphor mod optimizations such as - ensuring we don't notify neighbor chunks when neighbor chunk doesn't need to be notified - reduce hasLight checks in initializing light, and prob some more, they are tagged JellySquid where phosphor influence was used. 3) Optimize hot path accesses to getting updating chunk to have less branching 4) Optimize getBlock accesses to have less branching, and less unpacking 5) Have a separate urgent bucket for chunk light tasks. These tasks will always cut in line over non blocking light tasks. 6) Retain chunk priority while light tasks are enqueued. So if a task comes in at high priority but the queue is full of tasks already at a lower priority, before the task was simply added to the end. Now it can cut in line to the front. this applies for both urgent and non urgent tasks. 7) Buffer non urgent tasks even if queueUpdate is called multiple times to improve efficiency. 8) Fix NPE risk that crashes server in getting nibble data Fixes #3489 Fixes #3363
2020-06-05 07:25:11 +02:00
Subject: [PATCH] Optimize isValidLocation, getType and getBlockData for
inlining
2019-04-27 05:05:36 +02:00
Hot methods, so reduce # of instructions for the method.
Move is valid location test to the BlockPosition class so that it can access local variables.
Replace all calls to the new place to the unnecessary forward.
Optimize getType and getBlockData to manually inline and optimize the calls
diff --git a/src/main/java/net/minecraft/server/BaseBlockPosition.java b/src/main/java/net/minecraft/server/BaseBlockPosition.java
index 6b6ea0b33b10a9517b7af55fb8292fe245e3ca1e..e126f9d1d03de085f2d09d53fe8bcc1972aedeff 100644
2019-04-27 05:05:36 +02:00
--- a/src/main/java/net/minecraft/server/BaseBlockPosition.java
+++ b/src/main/java/net/minecraft/server/BaseBlockPosition.java
2020-06-25 13:00:35 +02:00
@@ -20,6 +20,15 @@ public class BaseBlockPosition implements Comparable<BaseBlockPosition> {
private int b;public final void setY(final int y) { this.b = y; } // Paper - OBFHELPER
private int e;public final void setZ(final int z) { this.e = z; } // Paper - OBFHELPER
2020-06-25 13:00:35 +02:00
2019-04-27 05:05:36 +02:00
+ // Paper start
+ public boolean isValidLocation() {
2020-06-25 13:00:35 +02:00
+ return getX() >= -30000000 && getZ() >= -30000000 && getX() < 30000000 && getZ() < 30000000 && getY() >= 0 && getY() < 256;
2019-04-27 05:05:36 +02:00
+ }
+ public boolean isInvalidYLocation() {
+ return b < 0 || b >= 256;
+ }
+ // Paper end
2020-06-25 13:00:35 +02:00
+
2019-04-27 05:05:36 +02:00
public BaseBlockPosition(int i, int j, int k) {
this.a = i;
2020-06-25 13:00:35 +02:00
this.b = j;
2019-04-27 05:05:36 +02:00
diff --git a/src/main/java/net/minecraft/server/Chunk.java b/src/main/java/net/minecraft/server/Chunk.java
index 330492e918efa42717ef2208947e8be927df5398..8d13c2fab572428d61777699d5463f0678d9ad7c 100644
2019-04-27 05:05:36 +02:00
--- a/src/main/java/net/minecraft/server/Chunk.java
+++ b/src/main/java/net/minecraft/server/Chunk.java
2020-06-25 13:00:35 +02:00
@@ -308,12 +308,27 @@ public class Chunk implements IChunkAccess {
2019-04-27 05:05:36 +02:00
return this.sections;
}
- @Override
2019-05-14 04:20:58 +02:00
+ // Paper start - Optimize getBlockData to reduce instructions
+ public final IBlockData getBlockData(BlockPosition pos) { return getBlockData(pos.getX(), pos.getY(), pos.getZ()); } // Paper
public IBlockData getType(BlockPosition blockposition) {
2019-04-27 05:05:36 +02:00
- int i = blockposition.getX();
- int j = blockposition.getY();
- int k = blockposition.getZ();
+ return this.getBlockData(blockposition.getX(), blockposition.getY(), blockposition.getZ());
+ }
Optimize Light Engine Massive update to light to improve performance and chunk loading/generation. 1) Massive bit packing/unpacking optimizations and inlining. A lot of performance has to do with constant packing and unpacking of bits. We now inline a most bit operations, and re-use base x/y/z bits in many places. This helps with cpu level processing to just do all the math at once instead of having to jump in and out of function calls. This much logic also is likely over the JVM Inline limit for JIT too. 2) Applied a few of JellySquid's Phosphor mod optimizations such as - ensuring we don't notify neighbor chunks when neighbor chunk doesn't need to be notified - reduce hasLight checks in initializing light, and prob some more, they are tagged JellySquid where phosphor influence was used. 3) Optimize hot path accesses to getting updating chunk to have less branching 4) Optimize getBlock accesses to have less branching, and less unpacking 5) Have a separate urgent bucket for chunk light tasks. These tasks will always cut in line over non blocking light tasks. 6) Retain chunk priority while light tasks are enqueued. So if a task comes in at high priority but the queue is full of tasks already at a lower priority, before the task was simply added to the end. Now it can cut in line to the front. this applies for both urgent and non urgent tasks. 7) Buffer non urgent tasks even if queueUpdate is called multiple times to improve efficiency. 8) Fix NPE risk that crashes server in getting nibble data Fixes #3489 Fixes #3363
2020-06-05 07:25:11 +02:00
+
+ public IBlockData getType(final int x, final int y, final int z) {
+ return getBlockData(x, y, z);
+ }
2019-04-27 05:05:36 +02:00
+ public final IBlockData getBlockData(final int x, final int y, final int z) {
+ // Method body / logic copied from below
+ final int i = y >> 4;
Optimize Light Engine Massive update to light to improve performance and chunk loading/generation. 1) Massive bit packing/unpacking optimizations and inlining. A lot of performance has to do with constant packing and unpacking of bits. We now inline a most bit operations, and re-use base x/y/z bits in many places. This helps with cpu level processing to just do all the math at once instead of having to jump in and out of function calls. This much logic also is likely over the JVM Inline limit for JIT too. 2) Applied a few of JellySquid's Phosphor mod optimizations such as - ensuring we don't notify neighbor chunks when neighbor chunk doesn't need to be notified - reduce hasLight checks in initializing light, and prob some more, they are tagged JellySquid where phosphor influence was used. 3) Optimize hot path accesses to getting updating chunk to have less branching 4) Optimize getBlock accesses to have less branching, and less unpacking 5) Have a separate urgent bucket for chunk light tasks. These tasks will always cut in line over non blocking light tasks. 6) Retain chunk priority while light tasks are enqueued. So if a task comes in at high priority but the queue is full of tasks already at a lower priority, before the task was simply added to the end. Now it can cut in line to the front. this applies for both urgent and non urgent tasks. 7) Buffer non urgent tasks even if queueUpdate is called multiple times to improve efficiency. 8) Fix NPE risk that crashes server in getting nibble data Fixes #3489 Fixes #3363
2020-06-05 07:25:11 +02:00
+ if (y < 0 || i >= this.sections.length || this.sections[i] == null || this.sections[i].nonEmptyBlockCount == 0) {
+ return Blocks.AIR.getBlockData();
2019-04-27 05:05:36 +02:00
+ }
Optimize Light Engine Massive update to light to improve performance and chunk loading/generation. 1) Massive bit packing/unpacking optimizations and inlining. A lot of performance has to do with constant packing and unpacking of bits. We now inline a most bit operations, and re-use base x/y/z bits in many places. This helps with cpu level processing to just do all the math at once instead of having to jump in and out of function calls. This much logic also is likely over the JVM Inline limit for JIT too. 2) Applied a few of JellySquid's Phosphor mod optimizations such as - ensuring we don't notify neighbor chunks when neighbor chunk doesn't need to be notified - reduce hasLight checks in initializing light, and prob some more, they are tagged JellySquid where phosphor influence was used. 3) Optimize hot path accesses to getting updating chunk to have less branching 4) Optimize getBlock accesses to have less branching, and less unpacking 5) Have a separate urgent bucket for chunk light tasks. These tasks will always cut in line over non blocking light tasks. 6) Retain chunk priority while light tasks are enqueued. So if a task comes in at high priority but the queue is full of tasks already at a lower priority, before the task was simply added to the end. Now it can cut in line to the front. this applies for both urgent and non urgent tasks. 7) Buffer non urgent tasks even if queueUpdate is called multiple times to improve efficiency. 8) Fix NPE risk that crashes server in getting nibble data Fixes #3489 Fixes #3363
2020-06-05 07:25:11 +02:00
+ // Inlined ChunkSection.getType() and DataPaletteBlock.a(int,int,int)
+ return this.sections[i].blockIds.a((y & 15) << 8 | (z & 15) << 4 | x & 15);
2019-04-27 05:05:36 +02:00
+ }
Optimize Light Engine Massive update to light to improve performance and chunk loading/generation. 1) Massive bit packing/unpacking optimizations and inlining. A lot of performance has to do with constant packing and unpacking of bits. We now inline a most bit operations, and re-use base x/y/z bits in many places. This helps with cpu level processing to just do all the math at once instead of having to jump in and out of function calls. This much logic also is likely over the JVM Inline limit for JIT too. 2) Applied a few of JellySquid's Phosphor mod optimizations such as - ensuring we don't notify neighbor chunks when neighbor chunk doesn't need to be notified - reduce hasLight checks in initializing light, and prob some more, they are tagged JellySquid where phosphor influence was used. 3) Optimize hot path accesses to getting updating chunk to have less branching 4) Optimize getBlock accesses to have less branching, and less unpacking 5) Have a separate urgent bucket for chunk light tasks. These tasks will always cut in line over non blocking light tasks. 6) Retain chunk priority while light tasks are enqueued. So if a task comes in at high priority but the queue is full of tasks already at a lower priority, before the task was simply added to the end. Now it can cut in line to the front. this applies for both urgent and non urgent tasks. 7) Buffer non urgent tasks even if queueUpdate is called multiple times to improve efficiency. 8) Fix NPE risk that crashes server in getting nibble data Fixes #3489 Fixes #3363
2020-06-05 07:25:11 +02:00
2019-04-27 05:05:36 +02:00
+ public IBlockData getBlockData_unused(int i, int j, int k) {
+ // Paper end
2020-06-25 13:00:35 +02:00
if (this.world.isDebugWorld()) {
2019-04-27 05:05:36 +02:00
IBlockData iblockdata = null;
Optimize Light Engine Massive update to light to improve performance and chunk loading/generation. 1) Massive bit packing/unpacking optimizations and inlining. A lot of performance has to do with constant packing and unpacking of bits. We now inline a most bit operations, and re-use base x/y/z bits in many places. This helps with cpu level processing to just do all the math at once instead of having to jump in and out of function calls. This much logic also is likely over the JVM Inline limit for JIT too. 2) Applied a few of JellySquid's Phosphor mod optimizations such as - ensuring we don't notify neighbor chunks when neighbor chunk doesn't need to be notified - reduce hasLight checks in initializing light, and prob some more, they are tagged JellySquid where phosphor influence was used. 3) Optimize hot path accesses to getting updating chunk to have less branching 4) Optimize getBlock accesses to have less branching, and less unpacking 5) Have a separate urgent bucket for chunk light tasks. These tasks will always cut in line over non blocking light tasks. 6) Retain chunk priority while light tasks are enqueued. So if a task comes in at high priority but the queue is full of tasks already at a lower priority, before the task was simply added to the end. Now it can cut in line to the front. this applies for both urgent and non urgent tasks. 7) Buffer non urgent tasks even if queueUpdate is called multiple times to improve efficiency. 8) Fix NPE risk that crashes server in getting nibble data Fixes #3489 Fixes #3363
2020-06-05 07:25:11 +02:00
diff --git a/src/main/java/net/minecraft/server/ChunkEmpty.java b/src/main/java/net/minecraft/server/ChunkEmpty.java
index 84207e7da6df97246f756940142a2af9eb9ca815..ed22ff28ea6c0978ec0d9d1ecf7baa3f422ed677 100644
Optimize Light Engine Massive update to light to improve performance and chunk loading/generation. 1) Massive bit packing/unpacking optimizations and inlining. A lot of performance has to do with constant packing and unpacking of bits. We now inline a most bit operations, and re-use base x/y/z bits in many places. This helps with cpu level processing to just do all the math at once instead of having to jump in and out of function calls. This much logic also is likely over the JVM Inline limit for JIT too. 2) Applied a few of JellySquid's Phosphor mod optimizations such as - ensuring we don't notify neighbor chunks when neighbor chunk doesn't need to be notified - reduce hasLight checks in initializing light, and prob some more, they are tagged JellySquid where phosphor influence was used. 3) Optimize hot path accesses to getting updating chunk to have less branching 4) Optimize getBlock accesses to have less branching, and less unpacking 5) Have a separate urgent bucket for chunk light tasks. These tasks will always cut in line over non blocking light tasks. 6) Retain chunk priority while light tasks are enqueued. So if a task comes in at high priority but the queue is full of tasks already at a lower priority, before the task was simply added to the end. Now it can cut in line to the front. this applies for both urgent and non urgent tasks. 7) Buffer non urgent tasks even if queueUpdate is called multiple times to improve efficiency. 8) Fix NPE risk that crashes server in getting nibble data Fixes #3489 Fixes #3363
2020-06-05 07:25:11 +02:00
--- a/src/main/java/net/minecraft/server/ChunkEmpty.java
+++ b/src/main/java/net/minecraft/server/ChunkEmpty.java
@@ -7,7 +7,7 @@ import javax.annotation.Nullable;
public class ChunkEmpty extends Chunk {
- private static final BiomeBase[] b = (BiomeBase[]) SystemUtils.a((Object) (new BiomeBase[BiomeStorage.a]), (abiomebase) -> {
+ private static final BiomeBase[] b = SystemUtils.a((new BiomeBase[BiomeStorage.a]), (abiomebase) -> { // Paper - decompile error
Arrays.fill(abiomebase, BiomeRegistry.a);
Optimize Light Engine Massive update to light to improve performance and chunk loading/generation. 1) Massive bit packing/unpacking optimizations and inlining. A lot of performance has to do with constant packing and unpacking of bits. We now inline a most bit operations, and re-use base x/y/z bits in many places. This helps with cpu level processing to just do all the math at once instead of having to jump in and out of function calls. This much logic also is likely over the JVM Inline limit for JIT too. 2) Applied a few of JellySquid's Phosphor mod optimizations such as - ensuring we don't notify neighbor chunks when neighbor chunk doesn't need to be notified - reduce hasLight checks in initializing light, and prob some more, they are tagged JellySquid where phosphor influence was used. 3) Optimize hot path accesses to getting updating chunk to have less branching 4) Optimize getBlock accesses to have less branching, and less unpacking 5) Have a separate urgent bucket for chunk light tasks. These tasks will always cut in line over non blocking light tasks. 6) Retain chunk priority while light tasks are enqueued. So if a task comes in at high priority but the queue is full of tasks already at a lower priority, before the task was simply added to the end. Now it can cut in line to the front. this applies for both urgent and non urgent tasks. 7) Buffer non urgent tasks even if queueUpdate is called multiple times to improve efficiency. 8) Fix NPE risk that crashes server in getting nibble data Fixes #3489 Fixes #3363
2020-06-05 07:25:11 +02:00
});
@@ -15,6 +15,11 @@ public class ChunkEmpty extends Chunk {
super(world, chunkcoordintpair, new BiomeStorage(world.r().b(IRegistry.ay), ChunkEmpty.b));
Optimize Light Engine Massive update to light to improve performance and chunk loading/generation. 1) Massive bit packing/unpacking optimizations and inlining. A lot of performance has to do with constant packing and unpacking of bits. We now inline a most bit operations, and re-use base x/y/z bits in many places. This helps with cpu level processing to just do all the math at once instead of having to jump in and out of function calls. This much logic also is likely over the JVM Inline limit for JIT too. 2) Applied a few of JellySquid's Phosphor mod optimizations such as - ensuring we don't notify neighbor chunks when neighbor chunk doesn't need to be notified - reduce hasLight checks in initializing light, and prob some more, they are tagged JellySquid where phosphor influence was used. 3) Optimize hot path accesses to getting updating chunk to have less branching 4) Optimize getBlock accesses to have less branching, and less unpacking 5) Have a separate urgent bucket for chunk light tasks. These tasks will always cut in line over non blocking light tasks. 6) Retain chunk priority while light tasks are enqueued. So if a task comes in at high priority but the queue is full of tasks already at a lower priority, before the task was simply added to the end. Now it can cut in line to the front. this applies for both urgent and non urgent tasks. 7) Buffer non urgent tasks even if queueUpdate is called multiple times to improve efficiency. 8) Fix NPE risk that crashes server in getting nibble data Fixes #3489 Fixes #3363
2020-06-05 07:25:11 +02:00
}
+ // Paper start
+ @Override public IBlockData getType(int x, int y, int z) {
+ return Blocks.VOID_AIR.getBlockData();
+ }
+ // Paper end
@Override
public IBlockData getType(BlockPosition blockposition) {
return Blocks.VOID_AIR.getBlockData();
2019-04-27 05:05:36 +02:00
diff --git a/src/main/java/net/minecraft/server/ChunkSection.java b/src/main/java/net/minecraft/server/ChunkSection.java
index 882c2733beaff1df68b892d44fc77cacf4364ff4..5c7068cd93806d67c643ed6aabdfcab8888ed94e 100644
2019-04-27 05:05:36 +02:00
--- a/src/main/java/net/minecraft/server/ChunkSection.java
+++ b/src/main/java/net/minecraft/server/ChunkSection.java
2020-06-25 13:00:35 +02:00
@@ -7,10 +7,10 @@ public class ChunkSection {
Optimize Light Engine Massive update to light to improve performance and chunk loading/generation. 1) Massive bit packing/unpacking optimizations and inlining. A lot of performance has to do with constant packing and unpacking of bits. We now inline a most bit operations, and re-use base x/y/z bits in many places. This helps with cpu level processing to just do all the math at once instead of having to jump in and out of function calls. This much logic also is likely over the JVM Inline limit for JIT too. 2) Applied a few of JellySquid's Phosphor mod optimizations such as - ensuring we don't notify neighbor chunks when neighbor chunk doesn't need to be notified - reduce hasLight checks in initializing light, and prob some more, they are tagged JellySquid where phosphor influence was used. 3) Optimize hot path accesses to getting updating chunk to have less branching 4) Optimize getBlock accesses to have less branching, and less unpacking 5) Have a separate urgent bucket for chunk light tasks. These tasks will always cut in line over non blocking light tasks. 6) Retain chunk priority while light tasks are enqueued. So if a task comes in at high priority but the queue is full of tasks already at a lower priority, before the task was simply added to the end. Now it can cut in line to the front. this applies for both urgent and non urgent tasks. 7) Buffer non urgent tasks even if queueUpdate is called multiple times to improve efficiency. 8) Fix NPE risk that crashes server in getting nibble data Fixes #3489 Fixes #3363
2020-06-05 07:25:11 +02:00
public static final DataPalette<IBlockData> GLOBAL_PALETTE = new DataPaletteGlobal<>(Block.REGISTRY_ID, Blocks.AIR.getBlockData());
private final int yPos;
- private short nonEmptyBlockCount;
+ short nonEmptyBlockCount; // Paper - package-private
2019-04-27 05:05:36 +02:00
private short tickingBlockCount;
private short e;
- private final DataPaletteBlock<IBlockData> blockIds;
+ final DataPaletteBlock<IBlockData> blockIds; // Paper - package-private
2019-04-27 05:05:36 +02:00
public ChunkSection(int i) {
this(i, (short) 0, (short) 0, (short) 0);
2020-06-25 13:00:35 +02:00
@@ -24,8 +24,8 @@ public class ChunkSection {
this.blockIds = new DataPaletteBlock<>(ChunkSection.GLOBAL_PALETTE, Block.REGISTRY_ID, GameProfileSerializer::c, GameProfileSerializer::a, Blocks.AIR.getBlockData());
Optimize Light Engine Massive update to light to improve performance and chunk loading/generation. 1) Massive bit packing/unpacking optimizations and inlining. A lot of performance has to do with constant packing and unpacking of bits. We now inline a most bit operations, and re-use base x/y/z bits in many places. This helps with cpu level processing to just do all the math at once instead of having to jump in and out of function calls. This much logic also is likely over the JVM Inline limit for JIT too. 2) Applied a few of JellySquid's Phosphor mod optimizations such as - ensuring we don't notify neighbor chunks when neighbor chunk doesn't need to be notified - reduce hasLight checks in initializing light, and prob some more, they are tagged JellySquid where phosphor influence was used. 3) Optimize hot path accesses to getting updating chunk to have less branching 4) Optimize getBlock accesses to have less branching, and less unpacking 5) Have a separate urgent bucket for chunk light tasks. These tasks will always cut in line over non blocking light tasks. 6) Retain chunk priority while light tasks are enqueued. So if a task comes in at high priority but the queue is full of tasks already at a lower priority, before the task was simply added to the end. Now it can cut in line to the front. this applies for both urgent and non urgent tasks. 7) Buffer non urgent tasks even if queueUpdate is called multiple times to improve efficiency. 8) Fix NPE risk that crashes server in getting nibble data Fixes #3489 Fixes #3363
2020-06-05 07:25:11 +02:00
}
- public IBlockData getType(int i, int j, int k) {
- return (IBlockData) this.blockIds.a(i, j, k);
+ public final IBlockData getType(int i, int j, int k) { // Paper
+ return this.blockIds.a(j << 8 | k << 4 | i); // Paper - inline
}
public Fluid b(int i, int j, int k) {
diff --git a/src/main/java/net/minecraft/server/DataPaletteBlock.java b/src/main/java/net/minecraft/server/DataPaletteBlock.java
index 32849e360a396128bd228db269ad1a8f7c6583a8..eabc9d7b934f27c823e012f3f10fffc23b461292 100644
Optimize Light Engine Massive update to light to improve performance and chunk loading/generation. 1) Massive bit packing/unpacking optimizations and inlining. A lot of performance has to do with constant packing and unpacking of bits. We now inline a most bit operations, and re-use base x/y/z bits in many places. This helps with cpu level processing to just do all the math at once instead of having to jump in and out of function calls. This much logic also is likely over the JVM Inline limit for JIT too. 2) Applied a few of JellySquid's Phosphor mod optimizations such as - ensuring we don't notify neighbor chunks when neighbor chunk doesn't need to be notified - reduce hasLight checks in initializing light, and prob some more, they are tagged JellySquid where phosphor influence was used. 3) Optimize hot path accesses to getting updating chunk to have less branching 4) Optimize getBlock accesses to have less branching, and less unpacking 5) Have a separate urgent bucket for chunk light tasks. These tasks will always cut in line over non blocking light tasks. 6) Retain chunk priority while light tasks are enqueued. So if a task comes in at high priority but the queue is full of tasks already at a lower priority, before the task was simply added to the end. Now it can cut in line to the front. this applies for both urgent and non urgent tasks. 7) Buffer non urgent tasks even if queueUpdate is called multiple times to improve efficiency. 8) Fix NPE risk that crashes server in getting nibble data Fixes #3489 Fixes #3363
2020-06-05 07:25:11 +02:00
--- a/src/main/java/net/minecraft/server/DataPaletteBlock.java
+++ b/src/main/java/net/minecraft/server/DataPaletteBlock.java
2020-06-25 13:00:35 +02:00
@@ -124,7 +124,7 @@ public class DataPaletteBlock<T> implements DataPaletteExpandable<T> {
Optimize Light Engine Massive update to light to improve performance and chunk loading/generation. 1) Massive bit packing/unpacking optimizations and inlining. A lot of performance has to do with constant packing and unpacking of bits. We now inline a most bit operations, and re-use base x/y/z bits in many places. This helps with cpu level processing to just do all the math at once instead of having to jump in and out of function calls. This much logic also is likely over the JVM Inline limit for JIT too. 2) Applied a few of JellySquid's Phosphor mod optimizations such as - ensuring we don't notify neighbor chunks when neighbor chunk doesn't need to be notified - reduce hasLight checks in initializing light, and prob some more, they are tagged JellySquid where phosphor influence was used. 3) Optimize hot path accesses to getting updating chunk to have less branching 4) Optimize getBlock accesses to have less branching, and less unpacking 5) Have a separate urgent bucket for chunk light tasks. These tasks will always cut in line over non blocking light tasks. 6) Retain chunk priority while light tasks are enqueued. So if a task comes in at high priority but the queue is full of tasks already at a lower priority, before the task was simply added to the end. Now it can cut in line to the front. this applies for both urgent and non urgent tasks. 7) Buffer non urgent tasks even if queueUpdate is called multiple times to improve efficiency. 8) Fix NPE risk that crashes server in getting nibble data Fixes #3489 Fixes #3363
2020-06-05 07:25:11 +02:00
}
public T a(int i, int j, int k) {
- return this.a(b(i, j, k));
+ return this.a(j << 8 | k << 4 | i); // Paper - inline
}
protected T a(int i) {
diff --git a/src/main/java/net/minecraft/server/IChunkAccess.java b/src/main/java/net/minecraft/server/IChunkAccess.java
index e0672c3c0bf30d88ea1cc609459050499a7271f3..887366f4c2ab608974113e75760b58c47f2afa00 100644
Optimize Light Engine Massive update to light to improve performance and chunk loading/generation. 1) Massive bit packing/unpacking optimizations and inlining. A lot of performance has to do with constant packing and unpacking of bits. We now inline a most bit operations, and re-use base x/y/z bits in many places. This helps with cpu level processing to just do all the math at once instead of having to jump in and out of function calls. This much logic also is likely over the JVM Inline limit for JIT too. 2) Applied a few of JellySquid's Phosphor mod optimizations such as - ensuring we don't notify neighbor chunks when neighbor chunk doesn't need to be notified - reduce hasLight checks in initializing light, and prob some more, they are tagged JellySquid where phosphor influence was used. 3) Optimize hot path accesses to getting updating chunk to have less branching 4) Optimize getBlock accesses to have less branching, and less unpacking 5) Have a separate urgent bucket for chunk light tasks. These tasks will always cut in line over non blocking light tasks. 6) Retain chunk priority while light tasks are enqueued. So if a task comes in at high priority but the queue is full of tasks already at a lower priority, before the task was simply added to the end. Now it can cut in line to the front. this applies for both urgent and non urgent tasks. 7) Buffer non urgent tasks even if queueUpdate is called multiple times to improve efficiency. 8) Fix NPE risk that crashes server in getting nibble data Fixes #3489 Fixes #3363
2020-06-05 07:25:11 +02:00
--- a/src/main/java/net/minecraft/server/IChunkAccess.java
+++ b/src/main/java/net/minecraft/server/IChunkAccess.java
2020-06-25 13:00:35 +02:00
@@ -12,6 +12,7 @@ import org.apache.logging.log4j.LogManager;
Optimize Light Engine Massive update to light to improve performance and chunk loading/generation. 1) Massive bit packing/unpacking optimizations and inlining. A lot of performance has to do with constant packing and unpacking of bits. We now inline a most bit operations, and re-use base x/y/z bits in many places. This helps with cpu level processing to just do all the math at once instead of having to jump in and out of function calls. This much logic also is likely over the JVM Inline limit for JIT too. 2) Applied a few of JellySquid's Phosphor mod optimizations such as - ensuring we don't notify neighbor chunks when neighbor chunk doesn't need to be notified - reduce hasLight checks in initializing light, and prob some more, they are tagged JellySquid where phosphor influence was used. 3) Optimize hot path accesses to getting updating chunk to have less branching 4) Optimize getBlock accesses to have less branching, and less unpacking 5) Have a separate urgent bucket for chunk light tasks. These tasks will always cut in line over non blocking light tasks. 6) Retain chunk priority while light tasks are enqueued. So if a task comes in at high priority but the queue is full of tasks already at a lower priority, before the task was simply added to the end. Now it can cut in line to the front. this applies for both urgent and non urgent tasks. 7) Buffer non urgent tasks even if queueUpdate is called multiple times to improve efficiency. 8) Fix NPE risk that crashes server in getting nibble data Fixes #3489 Fixes #3363
2020-06-05 07:25:11 +02:00
public interface IChunkAccess extends IBlockAccess, IStructureAccess {
+ IBlockData getType(final int x, final int y, final int z); // Paper
@Nullable
IBlockData setType(BlockPosition blockposition, IBlockData iblockdata, boolean flag);
diff --git a/src/main/java/net/minecraft/server/ProtoChunk.java b/src/main/java/net/minecraft/server/ProtoChunk.java
index e356bd73901ac7f230492e654af579d21c8fc086..2bcd26ccccf4503241c6b77600ed6ce1d94ccfcc 100644
Optimize Light Engine Massive update to light to improve performance and chunk loading/generation. 1) Massive bit packing/unpacking optimizations and inlining. A lot of performance has to do with constant packing and unpacking of bits. We now inline a most bit operations, and re-use base x/y/z bits in many places. This helps with cpu level processing to just do all the math at once instead of having to jump in and out of function calls. This much logic also is likely over the JVM Inline limit for JIT too. 2) Applied a few of JellySquid's Phosphor mod optimizations such as - ensuring we don't notify neighbor chunks when neighbor chunk doesn't need to be notified - reduce hasLight checks in initializing light, and prob some more, they are tagged JellySquid where phosphor influence was used. 3) Optimize hot path accesses to getting updating chunk to have less branching 4) Optimize getBlock accesses to have less branching, and less unpacking 5) Have a separate urgent bucket for chunk light tasks. These tasks will always cut in line over non blocking light tasks. 6) Retain chunk priority while light tasks are enqueued. So if a task comes in at high priority but the queue is full of tasks already at a lower priority, before the task was simply added to the end. Now it can cut in line to the front. this applies for both urgent and non urgent tasks. 7) Buffer non urgent tasks even if queueUpdate is called multiple times to improve efficiency. 8) Fix NPE risk that crashes server in getting nibble data Fixes #3489 Fixes #3363
2020-06-05 07:25:11 +02:00
--- a/src/main/java/net/minecraft/server/ProtoChunk.java
+++ b/src/main/java/net/minecraft/server/ProtoChunk.java
2020-06-25 13:00:35 +02:00
@@ -95,16 +95,18 @@ public class ProtoChunk implements IChunkAccess {
Optimize Light Engine Massive update to light to improve performance and chunk loading/generation. 1) Massive bit packing/unpacking optimizations and inlining. A lot of performance has to do with constant packing and unpacking of bits. We now inline a most bit operations, and re-use base x/y/z bits in many places. This helps with cpu level processing to just do all the math at once instead of having to jump in and out of function calls. This much logic also is likely over the JVM Inline limit for JIT too. 2) Applied a few of JellySquid's Phosphor mod optimizations such as - ensuring we don't notify neighbor chunks when neighbor chunk doesn't need to be notified - reduce hasLight checks in initializing light, and prob some more, they are tagged JellySquid where phosphor influence was used. 3) Optimize hot path accesses to getting updating chunk to have less branching 4) Optimize getBlock accesses to have less branching, and less unpacking 5) Have a separate urgent bucket for chunk light tasks. These tasks will always cut in line over non blocking light tasks. 6) Retain chunk priority while light tasks are enqueued. So if a task comes in at high priority but the queue is full of tasks already at a lower priority, before the task was simply added to the end. Now it can cut in line to the front. this applies for both urgent and non urgent tasks. 7) Buffer non urgent tasks even if queueUpdate is called multiple times to improve efficiency. 8) Fix NPE risk that crashes server in getting nibble data Fixes #3489 Fixes #3363
2020-06-05 07:25:11 +02:00
@Override
public IBlockData getType(BlockPosition blockposition) {
- int i = blockposition.getY();
-
- if (World.b(i)) {
+ return getType(blockposition.getX(), blockposition.getY(), blockposition.getZ());
+ }
+ // Paper start
+ public IBlockData getType(final int x, final int y, final int z) {
+ if (y < 0 || y >= 256) {
return Blocks.VOID_AIR.getBlockData();
} else {
- ChunkSection chunksection = this.getSections()[i >> 4];
-
- return ChunkSection.a(chunksection) ? Blocks.AIR.getBlockData() : chunksection.getType(blockposition.getX() & 15, i & 15, blockposition.getZ() & 15);
+ ChunkSection chunksection = this.getSections()[y >> 4];
+ return chunksection == Chunk.EMPTY_CHUNK_SECTION || chunksection.c() ? Blocks.AIR.getBlockData() : chunksection.getType(x & 15, y & 15, z & 15);
}
}
+ // Paper end
@Override
public Fluid getFluid(BlockPosition blockposition) {
diff --git a/src/main/java/net/minecraft/server/ProtoChunkExtension.java b/src/main/java/net/minecraft/server/ProtoChunkExtension.java
index 065eeed71075bb8f4069cb172ea7fca54793ddc0..09da0cc8f02c504191dfec8be93e6cf67c6afb78 100644
Optimize Light Engine Massive update to light to improve performance and chunk loading/generation. 1) Massive bit packing/unpacking optimizations and inlining. A lot of performance has to do with constant packing and unpacking of bits. We now inline a most bit operations, and re-use base x/y/z bits in many places. This helps with cpu level processing to just do all the math at once instead of having to jump in and out of function calls. This much logic also is likely over the JVM Inline limit for JIT too. 2) Applied a few of JellySquid's Phosphor mod optimizations such as - ensuring we don't notify neighbor chunks when neighbor chunk doesn't need to be notified - reduce hasLight checks in initializing light, and prob some more, they are tagged JellySquid where phosphor influence was used. 3) Optimize hot path accesses to getting updating chunk to have less branching 4) Optimize getBlock accesses to have less branching, and less unpacking 5) Have a separate urgent bucket for chunk light tasks. These tasks will always cut in line over non blocking light tasks. 6) Retain chunk priority while light tasks are enqueued. So if a task comes in at high priority but the queue is full of tasks already at a lower priority, before the task was simply added to the end. Now it can cut in line to the front. this applies for both urgent and non urgent tasks. 7) Buffer non urgent tasks even if queueUpdate is called multiple times to improve efficiency. 8) Fix NPE risk that crashes server in getting nibble data Fixes #3489 Fixes #3363
2020-06-05 07:25:11 +02:00
--- a/src/main/java/net/minecraft/server/ProtoChunkExtension.java
+++ b/src/main/java/net/minecraft/server/ProtoChunkExtension.java
@@ -26,6 +26,11 @@ public class ProtoChunkExtension extends ProtoChunk {
public IBlockData getType(BlockPosition blockposition) {
return this.a.getType(blockposition);
}
+ // Paper start
+ public final IBlockData getType(final int x, final int y, final int z) {
+ return this.a.getBlockData(x, y, z);
+ }
+ // Paper end
@Override
public Fluid getFluid(BlockPosition blockposition) {
2019-04-27 05:05:36 +02:00
diff --git a/src/main/java/net/minecraft/server/World.java b/src/main/java/net/minecraft/server/World.java
index 1d0238566d8cef75c28389c5b7084019d25c8f72..b5a6e6152b020201de631a924c8fd077eb1b546a 100644
2019-04-27 05:05:36 +02:00
--- a/src/main/java/net/minecraft/server/World.java
+++ b/src/main/java/net/minecraft/server/World.java
@@ -187,7 +187,7 @@ public abstract class World implements GeneratorAccess, AutoCloseable {
2019-04-27 05:05:36 +02:00
}
public static boolean isValidLocation(BlockPosition blockposition) {
- return !isOutsideWorld(blockposition) && D(blockposition);
+ return blockposition.isValidLocation(); // Paper - use better/optimized check
2019-04-27 05:05:36 +02:00
}
public static boolean l(BlockPosition blockposition) {