slight optimization

c_unpacker/.gitignore

@@ -0,0 +1,5 @@
+unpack
+unpack_bitstream
+unpack_debug
+*.upk
+

c_unpacker/Makefile

@@ -0,0 +1,10 @@
+all: unpack unpack_bitstream
+
+unpack: main.c unpack.c
+	cc -O2 -o unpack main.c unpack.c
+
+unpack_bitstream: main.c unpack.c
+	cc -O2 -D UPKR_BITSTREAM -o unpack_bitstream main.c unpack.c
+	
+unpack_debug: main.c unpack.c
+	cc -g -o unpack_debug main.c unpack.c

c_unpacker/main.c

@@ -0,0 +1,25 @@
+#include <stdio.h>
+#include <stdlib.h>
+
+int upkr_unpack(void* destination, void* compressed_data);
+
+int main(int argn, char** argv) {
+  void* input_buffer = malloc(1024*1024);
+  void* output_buffer = malloc(1024*1024);
+  
+  FILE* in_file = fopen(argv[1], "rb");
+  int in_size = fread(input_buffer, 1, 1024*1024, in_file);
+  fclose(in_file);
+  
+  printf("Compressed size: %d\n", in_size);
+  
+  int out_size = upkr_unpack(output_buffer, input_buffer);
+  
+  printf("Uncompressed size: %d\n", out_size);
+  
+  FILE* out_file = fopen(argv[2], "wb");
+  fwrite(output_buffer, 1, out_size, out_file);
+  fclose(out_file);
+  
+  return 0;
+}

c_unpacker/readme.txt

@@ -0,0 +1,4 @@
+a very simple unpacker in c, as a reference for people wanting to implement their own unpacker.
+absolutely not production ready, it makes no effort to ensure the output buffer can actually
+hold the uncompressed data.
+!!! Never run on untrusted input !!!

c_unpacker/unpack.c

@@ -0,0 +1,95 @@
+typedef unsigned char u8;
+typedef unsigned short u16;
+typedef unsigned long u32;
+
+u8* upkr_data_ptr;
+u8 upkr_probs[1 + 255 + 1 + 2*32 + 2*32]; 
+#ifdef UPKR_BITSTREAM
+u16 upkr_state;
+u8 upkr_current_byte;
+int upkr_bits_left;
+#else
+u32 upkr_state;
+#endif
+
+int upkr_decode_bit(int context_index) {
+#ifdef UPKR_BITSTREAM
+    while(upkr_state < 32768) {
+        if(upkr_bits_left == 0) {
+            upkr_current_byte = *upkr_data_ptr++;
+            upkr_bits_left = 8;
+        }
+        upkr_state = (upkr_state << 1) + (upkr_current_byte & 1);
+        upkr_current_byte >>= 1;
+        --upkr_bits_left;
+    }
+#else
+    while(upkr_state < 4096) {
+        upkr_state = (upkr_state << 8) | *upkr_data_ptr++;
+    }
+#endif
+   
+    int prob = upkr_probs[context_index];
+    int bit = (upkr_state & 255) < prob ? 1 : 0;
+    
+    if(bit) {
+        upkr_state = prob * (upkr_state >> 8) + (upkr_state & 255);
+        upkr_probs[context_index] = prob + ((256 - prob + 8) >> 4);
+    } else {
+        upkr_state = (256 - prob) * (upkr_state >> 8) + (upkr_state & 255) - prob;
+        upkr_probs[context_index] = prob - ((prob + 8) >> 4);
+    }
+    
+    return bit;
+}
+
+int upkr_decode_length(int context_index) {
+    int length = 0;
+    int bit_pos = 0;
+    while(upkr_decode_bit(context_index)) {
+        length |= upkr_decode_bit(context_index + 1) << bit_pos++;
+        context_index += 2;
+    }
+    return length | (1 << bit_pos);
+}
+
+int upkr_unpack(void* destination, void* compressed_data) {
+    upkr_data_ptr = (u8*)compressed_data;
+    upkr_state = 0;
+#ifdef UPKR_BITSTREAM
+    upkr_bits_left = 0;
+#endif
+    for(int i = 0; i < sizeof(upkr_probs); ++i)
+        upkr_probs[i] = 128;
+    
+    u8* write_ptr = (u8*)destination;
+    
+    int prev_was_match = 0;
+    int offset = 0;
+    for(;;) {
+        if(upkr_decode_bit(0)) {
+            if(prev_was_match || upkr_decode_bit(256)) {
+                offset = upkr_decode_length(257) - 1;
+                if(offset == 0) {
+                    break;
+                }
+            }
+            int length = upkr_decode_length(257 + 64);
+            while(length--) {
+                *write_ptr = write_ptr[-offset];
+                ++write_ptr;
+            }
+            prev_was_match = 1;
+        } else {
+            int byte = 1;
+            while(byte < 256) {
+                int bit = upkr_decode_bit(byte);
+                byte = (byte << 1) + bit;
+            }
+            *write_ptr++ = byte;
+            prev_was_match = 0;
+        }
+    }
+    
+    return write_ptr - (u8*)destination;
+}

src/greedy_packer.rs

@@ -3,9 +3,13 @@ use crate::match_finder::MatchFinder;
 use crate::rans::RansCoder;
 use crate::ProgressCallback;
 
-pub fn pack(data: &[u8], mut progress_callback: Option<ProgressCallback>) -> Vec<u8> {
+pub fn pack(
+    data: &[u8],
+    use_bitstream: bool,
+    mut progress_callback: Option<ProgressCallback>,
+) -> Vec<u8> {
     let mut match_finder = MatchFinder::new(data);
-    let mut rans_coder = RansCoder::new();
+    let mut rans_coder = RansCoder::new(use_bitstream);
     let mut state = lz::CoderState::new();
 
     let mut pos = 0;

src/lib.rs

@@ -2,17 +2,31 @@ mod context_state;
 mod greedy_packer;
 mod lz;
 mod match_finder;
-mod rans;
 mod parsing_packer;
+mod rans;
 
 pub use lz::unpack;
 
 pub type ProgressCallback<'a> = &'a mut dyn FnMut(usize);
 
-pub fn pack(data: &[u8], level: u8, progress_callback: Option<ProgressCallback>) -> Vec<u8> {
+pub fn pack(
+    data: &[u8],
+    level: u8,
+    use_bitstream: bool,
+    progress_callback: Option<ProgressCallback>,
+) -> Vec<u8> {
     if level == 0 {
-        greedy_packer::pack(data, progress_callback)
+        greedy_packer::pack(data, use_bitstream, progress_callback)
     } else {
-        parsing_packer::pack(data, level, progress_callback)
+        parsing_packer::pack(data, level, use_bitstream, progress_callback)
     }
 }
+
+pub fn compressed_size(mut data: &[u8]) -> f32 {
+    let mut state = 0;
+    while state < 4096 {
+        state = (state << 8) | data[0] as u32;
+        data = &data[1..];
+    }
+    data.len() as f32 + (state as f32).log2() / 8.
+}

src/lz.rs

@@ -77,7 +77,7 @@ fn encode_length(
 pub struct CoderState {
     contexts: ContextState,
     last_offset: u32,
-    prev_was_match: bool
+    prev_was_match: bool,
 }
 
 impl CoderState {
@@ -85,7 +85,7 @@ impl CoderState {
         CoderState {
             contexts: ContextState::new(1 + 255 + 1 + 64 + 64),
             last_offset: 0,
-            prev_was_match: false
+            prev_was_match: false,
         }
     }
 
@@ -94,8 +94,8 @@ impl CoderState {
     }
 }
 
-pub fn unpack(packed_data: &[u8]) -> Vec<u8> {
-    let mut decoder = RansDecoder::new(packed_data);
+pub fn unpack(packed_data: &[u8], use_bitstream: bool) -> Vec<u8> {
+    let mut decoder = RansDecoder::new(packed_data, use_bitstream);
     let mut contexts = ContextState::new(1 + 255 + 1 + 64 + 64);
     let mut result = vec![];
     let mut offset = 0;

src/main.rs

@@ -9,6 +9,7 @@ fn main() -> Result<()> {
         None => print_help(),
         Some("pack") => {
             let level = args.opt_value_from_str(["-l", "--level"])?.unwrap_or(2u8);
+            let use_bitstream = args.contains(["-b", "--bitstream"]);
 
             let infile = args.free_from_os_str::<PathBuf, bool>(|s| Ok(s.into()))?;
             let outfile = args.free_from_os_str::<PathBuf, bool>(|s| Ok(s.into()))?;
@@ -21,6 +22,7 @@ fn main() -> Result<()> {
             let packed_data = upkr::pack(
                 &data,
                 level,
+                use_bitstream,
                 Some(&mut |pos| {
                     pb.set(pos as u64);
                 }),
@@ -36,12 +38,14 @@ fn main() -> Result<()> {
             File::create(outfile)?.write_all(&packed_data)?;
         }
         Some("unpack") => {
+            let use_bitstream = args.contains(["-b", "--bitstream"]);
+
             let infile = args.free_from_os_str::<PathBuf, bool>(|s| Ok(s.into()))?;
             let outfile = args.free_from_os_str::<PathBuf, bool>(|s| Ok(s.into()))?;
 
             let mut data = vec![];
             File::open(infile)?.read_to_end(&mut data)?;
-            let packed_data = upkr::unpack(&data);
+            let packed_data = upkr::unpack(&data, use_bitstream);
             File::create(outfile)?.write_all(&packed_data)?;
         }
         Some(other) => {

src/parsing_packer.rs

@@ -6,7 +6,7 @@ use crate::match_finder::MatchFinder;
 use crate::rans::{CostCounter, RansCoder};
 use crate::{lz, ProgressCallback};
 
-pub fn pack(data: &[u8], level: u8, progress_cb: Option<ProgressCallback>) -> Vec<u8> {
+pub fn pack(data: &[u8], level: u8, use_bitstream: bool, progress_cb: Option<ProgressCallback>) -> Vec<u8> {
     let mut parse = parse(data, Config::from_level(level), progress_cb);
     let mut ops = vec![];
     while let Some(link) = parse {
@@ -14,7 +14,7 @@ pub fn pack(data: &[u8], level: u8, progress_cb: Option<ProgressCallback>) -> Ve
         parse = link.prev.clone();
     }
     let mut state = lz::CoderState::new();
-    let mut coder = RansCoder::new();
+    let mut coder = RansCoder::new(use_bitstream);
     for op in ops.into_iter().rev() {
         op.encode(&mut coder, &mut state);
     }

src/rans.rs

@@ -1,6 +1,5 @@
 use crate::context_state::Context;
 
-const L_BITS: u32 = 12;
 pub const PROB_BITS: u32 = 8;
 pub const ONE_PROB: u32 = 1 << PROB_BITS;
 
@@ -13,43 +12,75 @@ pub trait EntropyCoder {
     }
 }
 
-pub struct RansCoder(Vec<u16>);
+pub struct RansCoder {
+    bits: Vec<u16>,
+    use_bitstream: bool,
+}
 
 impl EntropyCoder for RansCoder {
     fn encode_bit(&mut self, bit: bool, prob: u16) {
         assert!(prob < 32768);
-        self.0.push(prob | ((bit as u16) << 15));
+        self.bits.push(prob | ((bit as u16) << 15));
     }
 }
 
 impl RansCoder {
-    pub fn new() -> RansCoder {
-        RansCoder(Vec::new())
+    pub fn new(use_bitstream: bool) -> RansCoder {
+        RansCoder {
+            bits: Vec::new(),
+            use_bitstream,
+        }
     }
 
     pub fn finish(self) -> Vec<u8> {
         let mut buffer = vec![];
-        let mut state = 1 << L_BITS;
+        let l_bits: u32 = if self.use_bitstream { 15 } else { 12 };
+        let mut state = 1 << l_bits;
 
-        const MAX_STATE_FACTOR: u32 = 1 << (L_BITS + 8 - PROB_BITS);
-        for step in self.0.into_iter().rev() {
+        let mut byte = 0u8;
+        let mut bit = 8;
+        let mut flush_state: Box<dyn FnMut(&mut u32)> = if self.use_bitstream {
+            Box::new(|state: &mut u32| {
+                bit -= 1;
+                byte |= ((*state & 1) as u8) << bit;
+                if bit == 0 {
+                    buffer.push(byte);
+                    byte = 0;
+                    bit = 8;
+                }
+                *state >>= 1;
+            })
+        } else {
+            Box::new(|state: &mut u32| {
+                buffer.push(*state as u8);
+                *state >>= 8;
+            })
+        };
+
+        let num_flush_bits = if self.use_bitstream { 1 } else { 8 };
+        let max_state_factor: u32 = 1 << (l_bits + num_flush_bits - PROB_BITS);
+        for step in self.bits.into_iter().rev() {
             let prob = step as u32 & 32767;
             let (start, prob) = if step & 32768 != 0 {
                 (0, prob)
             } else {
                 (prob, ONE_PROB - prob)
             };
-            let max_state = MAX_STATE_FACTOR * prob;
+            let max_state = max_state_factor * prob;
             while state >= max_state {
-                buffer.push(state as u8);
-                state >>= 8;
+                flush_state(&mut state);
             }
             state = ((state / prob) << PROB_BITS) + (state % prob) + start;
         }
 
         while state > 0 {
-            buffer.push(state as u8);
-            state >>= 8;
+            flush_state(&mut state);
+        }
+
+        drop(flush_state);
+
+        if self.use_bitstream && byte != 0 {
+            buffer.push(byte);
         }
 
         buffer.reverse();
@@ -99,14 +130,22 @@ impl EntropyCoder for CostCounter {
 pub struct RansDecoder<'a> {
     data: &'a [u8],
     state: u32,
+    use_bitstream: bool,
+    byte: u8,
+    bits_left: u8,
 }
 
 const PROB_MASK: u32 = ONE_PROB - 1;
-const L: u32 = 1 << L_BITS;
 
 impl<'a> RansDecoder<'a> {
-    pub fn new(data: &'a [u8]) -> RansDecoder<'a> {
-        RansDecoder { data, state: 0 }
+    pub fn new(data: &'a [u8], use_bitstream: bool) -> RansDecoder<'a> {
+        RansDecoder {
+            data,
+            state: 0,
+            use_bitstream,
+            byte: 0,
+            bits_left: 0,
+        }
     }
 
     pub fn decode_with_context(&mut self, context: &mut Context) -> bool {
@@ -117,10 +156,23 @@ impl<'a> RansDecoder<'a> {
 
     pub fn decode_bit(&mut self, prob: u16) -> bool {
         let prob = prob as u32;
-        while self.state < L {
+        if self.use_bitstream {
+            while self.state < 32768 {
+                if self.bits_left == 0 {
+                    self.byte = self.data[0];
+                    self.data = &self.data[1..];
+                    self.bits_left = 8;
+                }
+                self.state = (self.state << 1) | (self.byte & 1) as u32;
+                self.byte >>= 1;
+                self.bits_left -= 1;
+            }
+        } else {
+            while self.state < 4096 {
                 self.state = (self.state << 8) | self.data[0] as u32;
                 self.data = &self.data[1..];
             }
+        }
 
         let bit = (self.state & PROB_MASK) < prob;