Speed up leb128 encoding and decoding for unsigned values.

Author: michaelwoerister

src/libserialize/leb128.rs

@@ -8,6 +8,8 @@
 // option. This file may not be copied, modified, or distributed
 // except according to those terms.
 
+use std::cmp;
+
 #[inline]
 fn write_to_vec(vec: &mut Vec<u8>, position: usize, byte: u8) {
     if position == vec.len() {
@@ -18,56 +20,95 @@ fn write_to_vec(vec: &mut Vec<u8>, position: usize, byte: u8) {
 }
 
 #[inline]
-/// encodes an integer using unsigned leb128 encoding and stores
-/// the result using a callback function.
-///
-/// The callback `write` is called once for each position
-/// that is to be written to with the byte to be encoded
-/// at that position.
-pub fn write_unsigned_leb128_to<W>(mut value: u128, mut write: W) -> usize
-    where W: FnMut(usize, u8)
-{
-    let mut position = 0;
-    loop {
-        let mut byte = (value & 0x7F) as u8;
-        value >>= 7;
-        if value != 0 {
-            byte |= 0x80;
-        }
-
-        write(position, byte);
-        position += 1;
+fn write_slice_to_vec(output: &mut Vec<u8>, start_position: usize, input: &[u8]) {
+    let input_len = input.len();
+    let capacity = output.len() - start_position;
+    let first_half = cmp::min(capacity, input_len);
 
-        if value == 0 {
-            break;
-        }
+    if first_half > 0 {
+        (&mut output[start_position..]).copy_from_slice(&input[.. first_half]);
     }
 
-    position
+    if first_half < input_len {
+        output.extend_from_slice(&input[first_half..]);
+    }
 }
 
-pub fn write_unsigned_leb128(out: &mut Vec<u8>, start_position: usize, value: u128) -> usize {
-    write_unsigned_leb128_to(value, |i, v| write_to_vec(out, start_position+i, v))
+#[cfg(target_pointer_width = "32")]
+const USIZE_LEB128_SIZE: usize = 5;
+#[cfg(target_pointer_width = "64")]
+const USIZE_LEB128_SIZE: usize = 10;
+
+macro_rules! leb128_size {
+    (u16) => (3);
+    (u32) => (5);
+    (u64) => (10);
+    (u128) => (19);
+    (usize) => (USIZE_LEB128_SIZE);
 }
 
-#[inline]
-pub fn read_unsigned_leb128(data: &[u8], start_position: usize) -> (u128, usize) {
-    let mut result = 0;
-    let mut shift = 0;
-    let mut position = start_position;
-    loop {
-        let byte = data[position];
-        position += 1;
-        result |= ((byte & 0x7F) as u128) << shift;
-        if (byte & 0x80) == 0 {
-            break;
+macro_rules! impl_write_unsigned_leb128 {
+    ($fn_name:ident, $int_ty:ident) => (
+        #[inline]
+        pub fn $fn_name(out: &mut Vec<u8>, start_position: usize, mut value: $int_ty) -> usize {
+            let mut encoded = [0u8; leb128_size!($int_ty)];
+
+            for i in 0 .. leb128_size!($int_ty) {
+                encoded[i] = (value as u8) & 0b0111_1111;
+                value = value >> 7;
+
+                if value == 0 {
+                    let bytes_written = i + 1;
+                    write_slice_to_vec(out, start_position, &encoded[0 .. bytes_written]);
+                    return bytes_written
+                } else {
+                    encoded[i] |= 0b1000_0000;
+                }
+            }
+
+            unreachable!()
         }
-        shift += 7;
-    }
+    )
+}
 
-    (result, position - start_position)
+impl_write_unsigned_leb128!(write_u16_leb128, u16);
+impl_write_unsigned_leb128!(write_u32_leb128, u32);
+impl_write_unsigned_leb128!(write_u64_leb128, u64);
+impl_write_unsigned_leb128!(write_u128_leb128, u128);
+impl_write_unsigned_leb128!(write_usize_leb128, usize);
+
+macro_rules! impl_read_unsigned_leb128 {
+    ($fn_name:ident, $int_ty:ident) => (
+        #[inline]
+        pub fn $fn_name(data: &[u8], start_position: usize) -> ($int_ty, usize) {
+            // Copy things into a fixed size buffer so we can skip the bounds check
+            let mut bytes = [0u8; leb128_size!($int_ty)];
+            let copy_len = cmp::min(data.len() - start_position, leb128_size!($int_ty));
+            (&mut bytes[0 .. copy_len])
+                .copy_from_slice(&data[start_position .. start_position + copy_len]);
+
+            let mut result = 0;
+
+            for i in 0 .. leb128_size!($int_ty) {
+                let byte = bytes[i];
+                result |= ((byte & 0b0111_1111) as $int_ty) << i * 7;
+
+                if (byte & 0b1000_0000) == 0 {
+                    return (result, i + 1);
+                }
+            }
+
+            unreachable!()
+        }
+    )
 }
 
+impl_read_unsigned_leb128!(read_u16_leb128, u16);
+impl_read_unsigned_leb128!(read_u32_leb128, u32);
+impl_read_unsigned_leb128!(read_u64_leb128, u64);
+impl_read_unsigned_leb128!(read_u128_leb128, u128);
+impl_read_unsigned_leb128!(read_usize_leb128, usize);
+
 #[inline]
 /// encodes an integer using signed leb128 encoding and stores
 /// the result using a callback function.
@@ -130,26 +171,36 @@ pub fn read_signed_leb128(data: &[u8], start_position: usize) -> (i128, usize) {
     (result, position - start_position)
 }
 
-#[test]
-fn test_unsigned_leb128() {
-    let mut stream = Vec::with_capacity(10000);
-
-    for x in 0..62 {
-        let pos = stream.len();
-        let bytes_written = write_unsigned_leb128(&mut stream, pos, 3 << x);
-        assert_eq!(stream.len(), pos + bytes_written);
-    }
-
-    let mut position = 0;
-    for x in 0..62 {
-        let expected = 3 << x;
-        let (actual, bytes_read) = read_unsigned_leb128(&stream, position);
-        assert_eq!(expected, actual);
-        position += bytes_read;
-    }
-    assert_eq!(stream.len(), position);
+macro_rules! impl_test_unsigned_leb128 {
+    ($test_name:ident, $write_fn_name:ident, $read_fn_name:ident, $int_ty:ident) => (
+        #[test]
+        fn $test_name() {
+            let mut stream = Vec::new();
+
+            for x in 0..62 {
+                let pos = stream.len();
+                let bytes_written = $write_fn_name(&mut stream, pos, (3u64 << x) as $int_ty);
+                assert_eq!(stream.len(), pos + bytes_written);
+            }
+
+            let mut position = 0;
+            for x in 0..62 {
+                let expected = (3u64 << x) as $int_ty;
+                let (actual, bytes_read) = $read_fn_name(&stream, position);
+                assert_eq!(expected, actual);
+                position += bytes_read;
+            }
+            assert_eq!(stream.len(), position);
+        }
+    )
 }
 
+impl_test_unsigned_leb128!(test_u16_leb128, write_u16_leb128, read_u16_leb128, u16);
+impl_test_unsigned_leb128!(test_u32_leb128, write_u32_leb128, read_u32_leb128, u32);
+impl_test_unsigned_leb128!(test_u64_leb128, write_u64_leb128, read_u64_leb128, u64);
+impl_test_unsigned_leb128!(test_u128_leb128, write_u128_leb128, read_u128_leb128, u128);
+impl_test_unsigned_leb128!(test_usize_leb128, write_usize_leb128, read_usize_leb128, usize);
+
 #[test]
 fn test_signed_leb128() {
     let values: Vec<_> = (-500..500).map(|i| i * 0x12345789ABCDEF).collect();

src/libserialize/opaque.rs

@@ -8,7 +8,7 @@
 // option. This file may not be copied, modified, or distributed
 // except according to those terms.
 
-use leb128::{read_signed_leb128, read_unsigned_leb128, write_signed_leb128, write_unsigned_leb128};
+use leb128::{read_signed_leb128, write_signed_leb128};
 use std::borrow::Cow;
 use std::io::{self, Write};
 use serialize;
@@ -31,9 +31,9 @@ impl<'a> Encoder<'a> {
 
 
 macro_rules! write_uleb128 {
-    ($enc:expr, $value:expr) => {{
+    ($enc:expr, $value:expr, $fn:ident) => {{
         let pos = $enc.cursor.position() as usize;
-        let bytes_written = write_unsigned_leb128($enc.cursor.get_mut(), pos, $value as u128);
+        let bytes_written = ::leb128::$fn($enc.cursor.get_mut(), pos, $value);
         $enc.cursor.set_position((pos + bytes_written) as u64);
         Ok(())
     }}
@@ -56,23 +56,23 @@ impl<'a> serialize::Encoder for Encoder<'a> {
     }
 
     fn emit_usize(&mut self, v: usize) -> EncodeResult {
-        write_uleb128!(self, v)
+        write_uleb128!(self, v, write_usize_leb128)
     }
 
     fn emit_u128(&mut self, v: u128) -> EncodeResult {
-        write_uleb128!(self, v)
+        write_uleb128!(self, v, write_u128_leb128)
     }
 
     fn emit_u64(&mut self, v: u64) -> EncodeResult {
-        write_uleb128!(self, v)
+        write_uleb128!(self, v, write_u64_leb128)
     }
 
     fn emit_u32(&mut self, v: u32) -> EncodeResult {
-        write_uleb128!(self, v)
+        write_uleb128!(self, v, write_u32_leb128)
     }
 
     fn emit_u16(&mut self, v: u16) -> EncodeResult {
-        write_uleb128!(self, v)
+        write_uleb128!(self, v, write_u16_leb128)
     }
 
     fn emit_u8(&mut self, v: u8) -> EncodeResult {
@@ -172,10 +172,10 @@ impl<'a> Decoder<'a> {
 }
 
 macro_rules! read_uleb128 {
-    ($dec:expr, $t:ty) => ({
-        let (value, bytes_read) = read_unsigned_leb128($dec.data, $dec.position);
+    ($dec:expr, $fn:ident) => ({
+        let (value, bytes_read) = ::leb128::$fn($dec.data, $dec.position);
         $dec.position += bytes_read;
-        Ok(value as $t)
+        Ok(value)
     })
 }
 
@@ -198,22 +198,22 @@ impl<'a> serialize::Decoder for Decoder<'a> {
 
     #[inline]
     fn read_u128(&mut self) -> Result<u128, Self::Error> {
-        read_uleb128!(self, u128)
+        read_uleb128!(self, read_u128_leb128)
     }
 
     #[inline]
     fn read_u64(&mut self) -> Result<u64, Self::Error> {
-        read_uleb128!(self, u64)
+        read_uleb128!(self, read_u64_leb128)
     }
 
     #[inline]
     fn read_u32(&mut self) -> Result<u32, Self::Error> {
-        read_uleb128!(self, u32)
+        read_uleb128!(self, read_u32_leb128)
     }
 
     #[inline]
     fn read_u16(&mut self) -> Result<u16, Self::Error> {
-        read_uleb128!(self, u16)
+        read_uleb128!(self, read_u16_leb128)
     }
 
     #[inline]
@@ -225,7 +225,7 @@ impl<'a> serialize::Decoder for Decoder<'a> {
 
     #[inline]
     fn read_usize(&mut self) -> Result<usize, Self::Error> {
-        read_uleb128!(self, usize)
+        read_uleb128!(self, read_usize_leb128)
     }
 
     #[inline]