Upstream/Paper
1
0
Fork 0
mirror of https://github.com/PaperMC/Paper.git synced 2024-11-23 11:06:29 +01:00
Code Issues Projects Releases Wiki Activity
986ebc68fc
Paper/Spigot-Server-Patches/0505-Optimize-Bit-Operations-by-inlining.patch

196 lines
9.1 KiB
Diff
Raw Normal View History

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
From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
From: Aikar <aikar@aikar.co>
Date: Thu, 4 Jun 2020 02:24:49 -0400
Subject: [PATCH] Optimize Bit Operations by inlining
Inline bit operations and reduce instruction count to make these hot
operations faster
diff --git a/src/main/java/net/minecraft/server/BlockPosition.java b/src/main/java/net/minecraft/server/BlockPosition.java
Current non compilable status of all patches - THIS IS NOT READY THERE IS STILL NO ETA. GOBLINS WILL EAT YOU.
2020-06-26 03:58:00 +02:00
index adfeeefbbc307a77738c2cc8d82856cd02c2285e..52b2bff9ee3ccb4cea9549542b2b484685a70bfb 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/BlockPosition.java
+++ b/src/main/java/net/minecraft/server/BlockPosition.java
Current non compilable status of all patches - THIS IS NOT READY THERE IS STILL NO ETA. GOBLINS WILL EAT YOU.
2020-06-26 03:58:00 +02:00
@@ -53,28 +53,29 @@ public class BlockPosition extends BaseBlockPosition {
this(baseblockposition.getX(), baseblockposition.getY(), baseblockposition.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 static long getAdjacent(int baseX, int baseY, int baseZ, EnumDirection enumdirection) { return asLong(baseX + enumdirection.getAdjacentX(), baseY + enumdirection.getAdjacentY(), baseZ + enumdirection.getAdjacentZ()); } // Paper
public static long a(long i, EnumDirection enumdirection) {
return a(i, enumdirection.getAdjacentX(), enumdirection.getAdjacentY(), enumdirection.getAdjacentZ());
}
public static long a(long i, int j, int k, int l) {
- return a(b(i) + j, c(i) + k, d(i) + l);
+ return a((int) (i >> 38) + j, (int) ((i << 52) >> 52) + k, (int) ((i << 26) >> 38) + l); // Paper - simplify/inline
}
public static int b(long i) {
Current non compilable status of all patches - THIS IS NOT READY THERE IS STILL NO ETA. GOBLINS WILL EAT YOU.
2020-06-26 03:58:00 +02:00
- return (int) (i << 64 - BlockPosition.m - BlockPosition.f >> 64 - BlockPosition.f);
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
+ return (int) (i >> 38); // Paper - simplify/inline
}
public static int c(long i) {
Current non compilable status of all patches - THIS IS NOT READY THERE IS STILL NO ETA. GOBLINS WILL EAT YOU.
2020-06-26 03:58:00 +02:00
- return (int) (i << 64 - BlockPosition.h >> 64 - BlockPosition.h);
+ return (int) ((i << 26) >> 38); // Paper - simplify/inline
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 int d(long i) {
Current non compilable status of all patches - THIS IS NOT READY THERE IS STILL NO ETA. GOBLINS WILL EAT YOU.
2020-06-26 03:58:00 +02:00
- return (int) (i << 64 - BlockPosition.l - BlockPosition.g >> 64 - BlockPosition.g);
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
+ return (int) ((i << 26) >> 38); // Paper - simplify/inline
}
public static BlockPosition fromLong(long i) {
- return new BlockPosition(b(i), c(i), d(i));
+ return new BlockPosition((int) (i >> 38), (int) ((i << 52) >> 52), (int) ((i << 26) >> 38)); // Paper - simplify/inline
}
Current non compilable status of all patches - THIS IS NOT READY THERE IS STILL NO ETA. GOBLINS WILL EAT YOU.
2020-06-26 03:58:00 +02:00
public long asLong() {
@@ -83,12 +84,7 @@ public class BlockPosition extends BaseBlockPosition {
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 long asLong(int x, int y, int z) { return a(x, y, z); } // Paper - OBFHELPER
public static long a(int i, int j, int k) {
- long l = 0L;
-
Current non compilable status of all patches - THIS IS NOT READY THERE IS STILL NO ETA. GOBLINS WILL EAT YOU.
2020-06-26 03:58:00 +02:00
- l |= ((long) i & BlockPosition.i) << BlockPosition.m;
- l |= ((long) j & BlockPosition.j) << 0;
- l |= ((long) k & BlockPosition.k) << BlockPosition.l;
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
- return l;
+ return (((long) i & (long) 67108863) << 38) | (((long) j & (long) 4095)) | (((long) k & (long) 67108863) << 12); // Paper - inline constants and simplify
}
public static long f(long i) {
diff --git a/src/main/java/net/minecraft/server/SectionPosition.java b/src/main/java/net/minecraft/server/SectionPosition.java
Current non compilable status of all patches - THIS IS NOT READY THERE IS STILL NO ETA. GOBLINS WILL EAT YOU.
2020-06-26 03:58:00 +02:00
index 6a8597c585e74ebea2e8bac2f9965f8760e19b67..7806f3c351cba3f0388da11888f900c48004dadf 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/SectionPosition.java
+++ b/src/main/java/net/minecraft/server/SectionPosition.java
@@ -16,7 +16,7 @@ public class SectionPosition extends BaseBlockPosition {
}
public static SectionPosition a(BlockPosition blockposition) {
- return new SectionPosition(a(blockposition.getX()), a(blockposition.getY()), a(blockposition.getZ()));
+ return new SectionPosition(blockposition.getX() >> 4, blockposition.getY() >> 4, blockposition.getZ() >> 4); // Paper
}
public static SectionPosition a(ChunkCoordIntPair chunkcoordintpair, int i) {
Improve light optimizations and fix bugs Rewrites the Threaded task logic to no longer use 2 queues and instead keep a single prioritized queue and do all of a chunks light tasks in a single batch Fix a math issue in one place (Thankfully didn't seem to really be a common place since didn't notice anything)
2020-06-20 01:10:38 +02:00
@@ -28,15 +28,23 @@ public class SectionPosition extends BaseBlockPosition {
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 SectionPosition a(long i) {
- return new SectionPosition(b(i), c(i), d(i));
+ return new SectionPosition((int) (i >> 42), (int) (i << 44 >> 44), (int) (i << 22 >> 42)); // Paper
}
public static long a(long i, EnumDirection enumdirection) {
Improve light optimizations and fix bugs Rewrites the Threaded task logic to no longer use 2 queues and instead keep a single prioritized queue and do all of a chunks light tasks in a single batch Fix a math issue in one place (Thankfully didn't seem to really be a common place since didn't notice anything)
2020-06-20 01:10:38 +02:00
return a(i, enumdirection.getAdjacentX(), enumdirection.getAdjacentY(), enumdirection.getAdjacentZ());
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
}
Improve light optimizations and fix bugs Rewrites the Threaded task logic to no longer use 2 queues and instead keep a single prioritized queue and do all of a chunks light tasks in a single batch Fix a math issue in one place (Thankfully didn't seem to really be a common place since didn't notice anything)
2020-06-20 01:10:38 +02:00
+ // Paper start
+ public static long getAdjacentFromBlockPos(int x, int y, int z, EnumDirection enumdirection) {
+ return (((long) ((x >> 4) + enumdirection.getAdjacentX()) & 4194303L) << 42) | (((long) ((y >> 4) + enumdirection.getAdjacentY()) & 1048575L)) | (((long) ((z >> 4) + enumdirection.getAdjacentZ()) & 4194303L) << 20);
+ }
+ public static long getAdjacentFromSectionPos(int x, int y, int z, EnumDirection enumdirection) {
+ return (((long) (x + enumdirection.getAdjacentX()) & 4194303L) << 42) | (((long) ((y) + enumdirection.getAdjacentY()) & 1048575L)) | (((long) (z + enumdirection.getAdjacentZ()) & 4194303L) << 20);
+ }
+ // Paper end
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 long a(long i, int j, int k, int l) {
- return b(b(i) + j, c(i) + k, d(i) + l);
+ return (((long) ((int) (i >> 42) + j) & 4194303L) << 42) | (((long) ((int) (i << 44 >> 44) + k) & 1048575L)) | (((long) ((int) (i << 22 >> 42) + l) & 4194303L) << 20); // Simplify to reduce instruction count
}
public static int a(int i) {
Improve light optimizations and fix bugs Rewrites the Threaded task logic to no longer use 2 queues and instead keep a single prioritized queue and do all of a chunks light tasks in a single batch Fix a math issue in one place (Thankfully didn't seem to really be a common place since didn't notice anything)
2020-06-20 01:10:38 +02:00
@@ -48,11 +56,7 @@ public class SectionPosition extends BaseBlockPosition {
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 short b(BlockPosition blockposition) {
- int i = b(blockposition.getX());
- int j = b(blockposition.getY());
- int k = b(blockposition.getZ());
-
- return (short) (i << 8 | k << 4 | j);
Current non compilable status of all patches - THIS IS NOT READY THERE IS STILL NO ETA. GOBLINS WILL EAT YOU.
2020-06-26 03:58:00 +02:00
+ return (short) ((blockposition.getX() & 15) << 8 | (blockposition.getZ() & 15) << 4 | blockposition.getY() & 15); // Paper - simplify/inline
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 int c(int i) {
Current non compilable status of all patches - THIS IS NOT READY THERE IS STILL NO ETA. GOBLINS WILL EAT YOU.
2020-06-26 03:58:00 +02:00
@@ -83,16 +87,16 @@ public class SectionPosition extends BaseBlockPosition {
return this.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 int d() {
- return this.a() << 4;
+ public final int d() { // Paper
Current non compilable status of all patches - THIS IS NOT READY THERE IS STILL NO ETA. GOBLINS WILL EAT YOU.
2020-06-26 03:58:00 +02:00
+ return this.getX() << 4; // Paper
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 int e() {
- return this.b() << 4;
+ public final int e() { // Paper
Current non compilable status of all patches - THIS IS NOT READY THERE IS STILL NO ETA. GOBLINS WILL EAT YOU.
2020-06-26 03:58:00 +02:00
+ return this.getY() << 4; // Paper
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 int f() {
- return this.c() << 4;
+ public final int f() { // Paper
Current non compilable status of all patches - THIS IS NOT READY THERE IS STILL NO ETA. GOBLINS WILL EAT YOU.
2020-06-26 03:58:00 +02:00
+ return this.getZ() << 4; // Paper
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
}
Current non compilable status of all patches - THIS IS NOT READY THERE IS STILL NO ETA. GOBLINS WILL EAT YOU.
2020-06-26 03:58:00 +02:00
public int g() {
@@ -107,8 +111,10 @@ public class SectionPosition extends BaseBlockPosition {
return (this.c() << 4) + 15;
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 long blockToSection(long i) { return e(i); } // Paper - OBFHELPER
public static long e(long i) {
- return b(a(BlockPosition.b(i)), a(BlockPosition.c(i)), a(BlockPosition.d(i)));
+ // b(a(BlockPosition.b(i)), a(BlockPosition.c(i)), a(BlockPosition.d(i)));
+ return (((long) (int) (i >> 42) & 4194303L) << 42) | (((long) (int) ((i << 52) >> 56) & 1048575L)) | (((long) (int) ((i << 26) >> 42) & 4194303L) << 20); // Simplify to reduce instruction count
}
public static long f(long i) {
Current non compilable status of all patches - THIS IS NOT READY THERE IS STILL NO ETA. GOBLINS WILL EAT YOU.
2020-06-26 03:58:00 +02:00
@@ -129,17 +135,18 @@ public class SectionPosition extends BaseBlockPosition {
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
return new ChunkCoordIntPair(this.a(), this.c());
}
+ // Paper start
+ public static long blockPosAsSectionLong(int i, int j, int k) {
+ return (((long) (i >> 4) & 4194303L) << 42) | (((long) (j >> 4) & 1048575L)) | (((long) (k >> 4) & 4194303L) << 20);
+ }
+ // Paper end
+ public static long asLong(int i, int j, int k) { return b(i, j, k); } // Paper - OBFHELPER
public static long b(int i, int j, int k) {
- long l = 0L;
-
- l |= ((long) i & 4194303L) << 42;
- l |= ((long) j & 1048575L) << 0;
- l |= ((long) k & 4194303L) << 20;
- return l;
+ return (((long) i & 4194303L) << 42) | (((long) j & 1048575L)) | (((long) k & 4194303L) << 20); // Paper - Simplify to reduce instruction count
}
Current non compilable status of all patches - THIS IS NOT READY THERE IS STILL NO ETA. GOBLINS WILL EAT YOU.
2020-06-26 03:58:00 +02:00
public long s() {
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
- return b(this.a(), this.b(), this.c());
Current non compilable status of all patches - THIS IS NOT READY THERE IS STILL NO ETA. GOBLINS WILL EAT YOU.
2020-06-26 03:58:00 +02:00
+ return (((long) getX() & 4194303L) << 42) | (((long) getY() & 1048575L)) | (((long) getZ() & 4194303L) << 20); // Paper - Simplify to reduce instruction count
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
}
Current non compilable status of all patches - THIS IS NOT READY THERE IS STILL NO ETA. GOBLINS WILL EAT YOU.
2020-06-26 03:58:00 +02:00
public Stream<BlockPosition> t() {
@@ -147,18 +154,11 @@ public class SectionPosition extends BaseBlockPosition {
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 Stream<SectionPosition> a(SectionPosition sectionposition, int i) {
- int j = sectionposition.a();
- int k = sectionposition.b();
- int l = sectionposition.c();
-
- return a(j - i, k - i, l - i, j + i, k + i, l + i);
Current non compilable status of all patches - THIS IS NOT READY THERE IS STILL NO ETA. GOBLINS WILL EAT YOU.
2020-06-26 03:58:00 +02:00
+ return a(sectionposition.getX() - i, sectionposition.getY() - i, sectionposition.getZ() - i, sectionposition.getX() + i, sectionposition.getY() + i, sectionposition.getZ() + i); // Paper - simplify/inline
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 Stream<SectionPosition> b(ChunkCoordIntPair chunkcoordintpair, int i) {
- int j = chunkcoordintpair.x;
- int k = chunkcoordintpair.z;
-
- return a(j - i, 0, k - i, j + i, 15, k + i);
+ return a(chunkcoordintpair.x - i, 0, chunkcoordintpair.z - i, chunkcoordintpair.x + i, 15, chunkcoordintpair.z + i); // Paper - simplify/inline
}
public static Stream<SectionPosition> a(final int i, final int j, final int k, final int l, final int i1, final int j1) {
Reference in New Issue Copy Permalink