fix merge conflict

10 files changed, 222 insertions, 26 deletions

diff --git a/Cargo.lock b/Cargo.lock
index 3ba21574..493cc6a1 100644
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -345,6 +345,7 @@ dependencies = [
  "flate2",
  "graphite_binary",
  "log",
+ "packed_simd_2",
  "serde",
  "valence_nbt",
 ]
@@ -1515,6 +1516,12 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "99227334921fae1a979cf0bfdfcc6b3e5ce376ef57e16fb6fb3ea2ed6095f80c"
 
 [[package]]
+name = "libm"
+version = "0.1.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7fc7aa29613bd6a620df431842069224d8bc9011086b1db4c0e0cd47fa03ec9a"
+
+[[package]]
 name = "linked-hash-map"
 version = "0.5.6"
 source = "registry+https://github.com/rust-lang/crates.io-index"
@@ -1741,6 +1748,16 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "b15813163c1d831bf4a13c3610c05c0d03b39feb07f7e09fa234dac9b15aaf39"
 
 [[package]]
+name = "packed_simd_2"
+version = "0.3.8"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a1914cd452d8fccd6f9db48147b29fd4ae05bea9dc5d9ad578509f72415de282"
+dependencies = [
+ "cfg-if",
+ "libm",
+]
+
+[[package]]
 name = "parking"
 version = "2.0.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
diff --git a/azalea-nbt/Cargo.toml b/azalea-nbt/Cargo.toml
index 2c35c3da..bead5ba3 100644
--- a/azalea-nbt/Cargo.toml
+++ b/azalea-nbt/Cargo.toml
@@ -15,6 +15,7 @@ compact_str = { version = "0.7.0", features = ["serde"] }
 enum-as-inner = "0.5.1"
 flate2 = "^1.0.25"
 log = "0.4.17"
+packed_simd_2 = "0.3.8"
 serde = { version = "1.0.152", features = ["derive"], optional = true }
 
 [dev-dependencies]
diff --git a/azalea-nbt/benches/compare.rs b/azalea-nbt/benches/compare.rs
index f0fa17c0..23e133da 100755
--- a/azalea-nbt/benches/compare.rs
+++ b/azalea-nbt/benches/compare.rs
@@ -79,7 +79,7 @@ pub fn bench_read_file(filename: &str, c: &mut Criterion) {
 }
 
 fn bench(c: &mut Criterion) {
-    bench_read_file("tests/bigtest.nbt", c);
+    // bench_read_file("tests/bigtest.nbt", c);
     // bench_read_file("tests/simple_player.dat", c);
     // bench_read_file("tests/complex_player.dat", c);
     // bench_read_file("tests/level.dat", c);
diff --git a/azalea-nbt/benches/nbt.rs b/azalea-nbt/benches/nbt.rs
index 0d413103..a219a46d 100755
--- a/azalea-nbt/benches/nbt.rs
+++ b/azalea-nbt/benches/nbt.rs
@@ -15,7 +15,10 @@ fn bench_file(filename: &str, c: &mut Criterion) {
     // decode the original src so most of the time isn't spent on unzipping
     let mut decoded_src_decoder = GzDecoder::new(&mut src);
     let mut decoded_src = Vec::new();
-    decoded_src_decoder.read_to_end(&mut decoded_src).unwrap();
+    if decoded_src_decoder.read_to_end(&mut decoded_src).is_err() {
+        // oh probably wasn't gzipped then
+        decoded_src = contents;
+    }
 
     let mut decoded_src_stream = Cursor::new(&decoded_src[..]);
 
@@ -26,12 +29,12 @@ fn bench_file(filename: &str, c: &mut Criterion) {
 
     group.throughput(Throughput::Bytes(decoded_src.len() as u64));
 
-    group.bench_function("Decode", |b| {
-        b.iter(|| {
-            black_box(Nbt::read(&mut decoded_src_stream).unwrap());
-            decoded_src_stream.set_position(0);
-        })
-    });
+    // group.bench_function("Decode", |b| {
+    //     b.iter(|| {
+    //         black_box(Nbt::read(&mut decoded_src_stream).unwrap());
+    //         decoded_src_stream.set_position(0);
+    //     })
+    // });
 
     // group.bench_function("Encode", |b| {
     //     b.iter(|| {
@@ -41,7 +44,16 @@ fn bench_file(filename: &str, c: &mut Criterion) {
 
     group.bench_function("Get", |b| {
         b.iter(|| {
-            black_box(nbt.as_compound().unwrap().get(""));
+            let level = nbt
+                .as_compound()
+                .unwrap()
+                .get("Level")
+                .unwrap()
+                .as_compound()
+                .unwrap();
+            for (k, _) in level.iter() {
+                black_box(level.get(black_box(k)));
+            }
         })
     });
     group.finish();
@@ -53,7 +65,10 @@ fn bench(c: &mut Criterion) {
     // bench_file("tests/complex_player.dat", c);
     // bench_file("tests/level.dat", c);
     // bench_file("tests/stringtest.nbt", c);
-    // bench_file("tests/inttest.nbt", c);
+    // bench_file("tests/inttest16.nbt", c);
+
+    // bench_file("tests/inttest1023.nbt", c);
+    // bench_file("tests/inttest3.nbt", c);
 }
 
 criterion_group!(benches, bench);
diff --git a/azalea-nbt/src/encode.rs b/azalea-nbt/src/encode.rs
index bf56c223..f5195327 100755
--- a/azalea-nbt/src/encode.rs
+++ b/azalea-nbt/src/encode.rs
@@ -2,6 +2,7 @@ use crate::tag::*;
 use azalea_buf::McBufWritable;
 use byteorder::{WriteBytesExt, BE};
 use flate2::write::{GzEncoder, ZlibEncoder};
+use packed_simd_2::{i32x16, i32x2, i32x4, i32x8, i64x2, i64x4, i64x8};
 use std::io::Write;
 
 #[inline(always)]
@@ -81,14 +82,78 @@ fn write_list(writer: &mut impl Write, value: &NbtList) {
         }
         NbtList::Int(l) => {
             writer.write_i32::<BE>(l.len() as i32).unwrap();
-            for &v in l {
-                writer.write_i32::<BE>(v).unwrap();
+            // flip the bits to big endian with simd
+            let l = l.as_slice();
+            let mut position = 0;
+            // x16
+            while l.len() - position >= 16 {
+                let l = unsafe { i32x16::from_slice_unaligned_unchecked(&l[position..]) };
+                l.to_be();
+                let l = unsafe { std::mem::transmute::<i32x16, [u8; 64]>(l) };
+                writer.write_all(&l).unwrap();
+                position += 16;
+            }
+            // x8
+            if l.len() - position >= 8 {
+                let l = unsafe { i32x8::from_slice_unaligned_unchecked(&l[position..]) };
+                l.to_be();
+                let l = unsafe { std::mem::transmute::<i32x8, [u8; 32]>(l) };
+                writer.write_all(&l).unwrap();
+                position += 8;
+            }
+            // x4
+            if l.len() - position >= 4 {
+                let l = unsafe { i32x4::from_slice_unaligned_unchecked(&l[position..]) };
+                l.to_be();
+                let l = unsafe { std::mem::transmute::<i32x4, [u8; 16]>(l) };
+                writer.write_all(&l).unwrap();
+                position += 4;
+            }
+            // x2
+            if l.len() - position >= 2 {
+                let l = unsafe { i32x2::from_slice_unaligned_unchecked(&l[position..]) };
+                l.to_be();
+                let l = unsafe { std::mem::transmute::<i32x2, [u8; 8]>(l) };
+                writer.write_all(&l).unwrap();
+                position += 2;
+            }
+            // x1 ... just a normal write_i32
+            if l.len() - position >= 1 {
+                writer.write_i32::<BE>(l[position]).unwrap();
             }
         }
         NbtList::Long(l) => {
             writer.write_i32::<BE>(l.len() as i32).unwrap();
-            for &v in l {
-                writer.write_i64::<BE>(v).unwrap();
+            // flip the bits to big endian with simd
+            let l = l.as_slice();
+            let mut position = 0;
+            // x16
+            while l.len() - position >= 8 {
+                let l = unsafe { i64x8::from_slice_unaligned_unchecked(&l[position..]) };
+                l.to_be();
+                let l = unsafe { std::mem::transmute::<i64x8, [u8; 64]>(l) };
+                writer.write_all(&l).unwrap();
+                position += 8;
+            }
+            // x4
+            if l.len() - position >= 4 {
+                let l = unsafe { i64x4::from_slice_unaligned_unchecked(&l[position..]) };
+                l.to_be();
+                let l = unsafe { std::mem::transmute::<i64x4, [u8; 32]>(l) };
+                writer.write_all(&l).unwrap();
+                position += 4;
+            }
+            // x2
+            if l.len() - position >= 2 {
+                let l = unsafe { i64x2::from_slice_unaligned_unchecked(&l[position..]) };
+                l.to_be();
+                let l = unsafe { std::mem::transmute::<i64x2, [u8; 16]>(l) };
+                writer.write_all(&l).unwrap();
+                position += 2;
+            }
+            // x1 ... just a normal write_i32
+            if l.len() - position >= 1 {
+                writer.write_i64::<BE>(l[position]).unwrap();
             }
         }
         NbtList::Float(l) => {
diff --git a/azalea-nbt/src/tag.rs b/azalea-nbt/src/tag.rs
index 32d73a7c..efb6bdd2 100755
--- a/azalea-nbt/src/tag.rs
+++ b/azalea-nbt/src/tag.rs
@@ -48,6 +48,17 @@ pub enum Nbt {
     IntArray(NbtIntArray) = INT_ARRAY_ID,
     LongArray(NbtLongArray) = LONG_ARRAY_ID,
 }
+impl Nbt {
+    /// Get the numerical ID of the tag type.
+    #[inline]
+    pub fn id(&self) -> u8 {
+        // SAFETY: Because `Self` is marked `repr(u8)`, its layout is a `repr(C)`
+        // `union` between `repr(C)` structs, each of which has the `u8`
+        // discriminant as its first field, so we can read the discriminant
+        // without offsetting the pointer.
+        unsafe { *<*const _>::from(self).cast::<u8>() }
+    }
+}
 
 /// An NBT value.
 #[derive(Clone, Debug, PartialEq)]
@@ -69,7 +80,7 @@ pub enum NbtList {
     LongArray(Vec<NbtLongArray>) = LONG_ARRAY_ID,
 }
 
-impl Nbt {
+impl NbtList {
     /// Get the numerical ID of the tag type.
     #[inline]
     pub fn id(&self) -> u8 {
@@ -80,15 +91,64 @@ impl Nbt {
         unsafe { *<*const _>::from(self).cast::<u8>() }
     }
 }
-impl NbtList {
-    /// Get the numerical ID of the tag type.
-    #[inline]
-    pub fn id(&self) -> u8 {
-        // SAFETY: Because `Self` is marked `repr(u8)`, its layout is a `repr(C)`
-        // `union` between `repr(C)` structs, each of which has the `u8`
-        // discriminant as its first field, so we can read the discriminant
-        // without offsetting the pointer.
-        unsafe { *<*const _>::from(self).cast::<u8>() }
+impl From<Vec<NbtByte>> for NbtList {
+    fn from(v: Vec<NbtByte>) -> Self {
+        Self::Byte(v)
+    }
+}
+impl From<Vec<NbtShort>> for NbtList {
+    fn from(v: Vec<NbtShort>) -> Self {
+        Self::Short(v)
+    }
+}
+impl From<Vec<NbtInt>> for NbtList {
+    fn from(v: Vec<NbtInt>) -> Self {
+        Self::Int(v)
+    }
+}
+impl From<Vec<NbtLong>> for NbtList {
+    fn from(v: Vec<NbtLong>) -> Self {
+        Self::Long(v)
+    }
+}
+impl From<Vec<NbtFloat>> for NbtList {
+    fn from(v: Vec<NbtFloat>) -> Self {
+        Self::Float(v)
+    }
+}
+impl From<Vec<NbtDouble>> for NbtList {
+    fn from(v: Vec<NbtDouble>) -> Self {
+        Self::Double(v)
+    }
+}
+impl From<Vec<NbtByteArray>> for NbtList {
+    fn from(v: Vec<NbtByteArray>) -> Self {
+        Self::ByteArray(v)
+    }
+}
+impl From<Vec<NbtString>> for NbtList {
+    fn from(v: Vec<NbtString>) -> Self {
+        Self::String(v)
+    }
+}
+impl From<Vec<NbtList>> for NbtList {
+    fn from(v: Vec<NbtList>) -> Self {
+        Self::List(v)
+    }
+}
+impl From<Vec<NbtCompound>> for NbtList {
+    fn from(v: Vec<NbtCompound>) -> Self {
+        Self::Compound(v)
+    }
+}
+impl From<Vec<NbtIntArray>> for NbtList {
+    fn from(v: Vec<NbtIntArray>) -> Self {
+        Self::IntArray(v)
+    }
+}
+impl From<Vec<NbtLongArray>> for NbtList {
+    fn from(v: Vec<NbtLongArray>) -> Self {
+        Self::LongArray(v)
     }
 }
 
@@ -161,7 +221,9 @@ impl NbtCompound {
 
     #[inline]
     fn is_worth_sorting(&self) -> bool {
-        self.inner.len() >= 128
+        // i don't actually know when binary search starts being better, but it's at
+        // least more than 12
+        self.inner.len() >= 32
     }
 }
 #[cfg(feature = "serde")]
@@ -191,3 +253,9 @@ impl FromIterator<(NbtString, Nbt)> for NbtCompound {
         Self { inner }
     }
 }
+
+impl From<Vec<(NbtString, Nbt)>> for NbtCompound {
+    fn from(inner: Vec<(NbtString, Nbt)>) -> Self {
+        Self { inner }
+    }
+}
diff --git a/azalea-nbt/tests/inttest1023.nbt b/azalea-nbt/tests/inttest1023.nbt
new file mode 100644
index 00000000..481dde19
--- /dev/null
+++ b/azalea-nbt/tests/inttest1023.nbt
diff --git a/azalea-nbt/tests/inttest.nbt b/azalea-nbt/tests/inttest16.nbt
index ad9255f2..ad9255f2 100755
--- a/azalea-nbt/tests/inttest.nbt
+++ b/azalea-nbt/tests/inttest16.nbt
diff --git a/azalea-nbt/tests/inttest3.nbt b/azalea-nbt/tests/inttest3.nbt
new file mode 100644
index 00000000..2890f577
--- /dev/null
+++ b/azalea-nbt/tests/inttest3.nbt
diff --git a/azalea-nbt/tests/tests.rs b/azalea-nbt/tests/tests.rs
index a7f6912f..5c82c3c7 100755
--- a/azalea-nbt/tests/tests.rs
+++ b/azalea-nbt/tests/tests.rs
@@ -1,4 +1,4 @@
-use azalea_nbt::{NbtCompound, NbtList, Nbt};
+use azalea_nbt::{Nbt, NbtCompound, NbtList};
 use std::io::Cursor;
 
 #[test]
@@ -108,3 +108,33 @@ fn test_simple_player() {
 
     assert_eq!(decoded_tag, original_tag);
 }
+
+// #[test]
+// fn test_inttest() {
+//     let original = include_bytes!("inttest.nbt").to_vec();
+
+//     let mut original_stream = Cursor::new(original);
+//     let original_tag = Nbt::read_gzip(&mut original_stream).unwrap();
+
+//     let mut result = Vec::new();
+//     original_tag.write(&mut result);
+
+//     let decoded_tag = Nbt::read(&mut Cursor::new(&result)).unwrap();
+
+//     assert_eq!(decoded_tag, original_tag);
+// }
+
+#[test]
+fn test_inttest1023() {
+    let original = include_bytes!("inttest1023.nbt").to_vec();
+
+    let mut original_stream = Cursor::new(original.as_slice());
+    let original_tag = Nbt::read(&mut original_stream).unwrap();
+
+    let mut result = Vec::new();
+    original_tag.write(&mut result);
+
+    let decoded_tag = Nbt::read(&mut Cursor::new(&result)).unwrap();
+
+    assert_eq!(decoded_tag, original_tag);
+}