libazo 0.1.1

Decompression library for the AZO format
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175

//! Adaptive probability models.
use crate::range::RangeDecoder;

const BOOL_TOTAL_BIT: u32 = 12;
const BOOL_TOTAL_COUNT: u32 = 4096;
const BOOL_SHIFT_BIT: u32 = 6;

pub struct BoolState {
    state: usize,
    prob: Vec<u32>,
    n_bits: u32,
}

impl BoolState {
    pub fn new(n_bits: u32) -> Self {
        let size = 1 << n_bits;
        BoolState {
            state: 0,
            prob: vec![BOOL_TOTAL_COUNT / 2; size],
            n_bits,
        }
    }

    pub fn decode(&mut self, entropy: &mut RangeDecoder) -> u32 {
        let bit = entropy.decode_boolean(self.prob[self.state], BOOL_TOTAL_BIT);
        if bit == 0 {
            self.prob[self.state] += (BOOL_TOTAL_COUNT - self.prob[self.state]) >> BOOL_SHIFT_BIT;
        } else {
            self.prob[self.state] -= self.prob[self.state] >> BOOL_SHIFT_BIT;
        }
        let mask = (1 << self.n_bits) - 1;
        self.state = ((self.state << 1) | bit as usize) & mask;
        bit
    }
}

const EBP_TOTAL_BIT: u32 = 10;
const EBP_TOTAL_COUNT: u32 = 1024;
const EBP_SHIFT_BIT: u32 = 4;

pub struct EntropyBitProb {
    prob: Vec<u32>,
    bit_n: u32,
}

impl EntropyBitProb {
    pub fn new(n: usize) -> Self {
        let bit_n = ceil_log2(n);
        let array_n = 1 << bit_n;
        EntropyBitProb {
            prob: vec![EBP_TOTAL_COUNT / 2; array_n],
            bit_n,
        }
    }

    pub fn decode(&mut self, entropy: &mut RangeDecoder) -> u32 {
        let mut value = 0u32;
        let mut pre = 1usize;

        for i in (0..self.bit_n).rev() {
            let bit = entropy.decode_boolean(self.prob[pre], EBP_TOTAL_BIT);
            self.update_prob(pre, bit);
            if bit != 0 {
                value |= 1 << i;
            }
            pre = (pre << 1) | bit as usize;
        }

        value
    }

    /// Update probabilities for a known value without decoding.
    pub fn update(&mut self, value: u32) {
        let mut pre = 1usize;
        for i in (0..self.bit_n).rev() {
            let v = (value >> i) & 1;
            self.update_prob(pre, v);
            pre = (pre << 1) | v as usize;
        }
    }

    fn update_prob(&mut self, idx: usize, bit: u32) {
        if bit == 0 {
            self.prob[idx] += (EBP_TOTAL_COUNT - self.prob[idx]) >> EBP_SHIFT_BIT;
        } else {
            self.prob[idx] -= self.prob[idx] >> EBP_SHIFT_BIT;
        }
    }

    /// Compare which model better predicted the value.
    pub fn compare_with(m1: &EntropyBitProb, m2: &EntropyBitProb, value: u32) -> i32 {
        let mut prod1 = 1u32;
        let mut prod2 = 1u32;
        let mut pre = 1usize;
        let overflow_mask = (EBP_TOTAL_COUNT - 1) << (32 - EBP_TOTAL_BIT);

        for i in (0..m1.bit_n).rev() {
            let p1 = m1.prob[pre];
            let p2 = m2.prob[pre];
            let v = (value >> i) & 1;

            if (prod1 | prod2) & overflow_mask != 0 {
                prod1 >>= EBP_TOTAL_BIT;
                prod2 >>= EBP_TOTAL_BIT;
            }

            if v != 0 {
                prod1 = prod1.wrapping_mul(EBP_TOTAL_COUNT - p1);
                prod2 = prod2.wrapping_mul(EBP_TOTAL_COUNT - p2);
            } else {
                prod1 = prod1.wrapping_mul(p1);
                prod2 = prod2.wrapping_mul(p2);
            }

            pre = (pre << 1) | v as usize;
        }

        if prod1 > prod2 {
            1
        } else if prod1 < prod2 {
            -1
        } else {
            0
        }
    }
}

pub struct PredictProb {
    prob1: Vec<EntropyBitProb>,
    prob2: Vec<EntropyBitProb>,
    lucky: Vec<i32>,
    shift: u32,
}

impl PredictProb {
    pub fn new(key: usize, n: usize, shift: u32) -> Self {
        let coarse_size = key >> shift;
        PredictProb {
            prob1: (0..key).map(|_| EntropyBitProb::new(n)).collect(),
            prob2: (0..coarse_size).map(|_| EntropyBitProb::new(n)).collect(),
            lucky: vec![0; key],
            shift,
        }
    }

    pub fn decode(&mut self, entropy: &mut RangeDecoder, context: usize) -> u32 {
        let coarse = context >> self.shift;
        let value = if self.lucky[context] >= 0 {
            let v = self.prob1[context].decode(entropy);
            self.prob2[coarse].update(v);
            v
        } else {
            let v = self.prob2[coarse].decode(entropy);
            self.prob1[context].update(v);
            v
        };

        let r = EntropyBitProb::compare_with(&self.prob1[context], &self.prob2[coarse], value);
        if r > 0 {
            self.lucky[context] += 1;
        }
        if r < 0 {
            self.lucky[context] -= 1;
        }

        value
    }
}

fn ceil_log2(n: usize) -> u32 {
    if n <= 1 {
        return 0;
    }
    usize::BITS - (n - 1).leading_zeros()
}