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
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195

#![allow(clippy::excessive_precision)]

use crate::enc::log_table_16::logs_16;
use crate::enc::log_table_8::logs_8;

#[cfg(feature = "float64")]
pub type floatX = f64;

#[cfg(not(feature = "float64"))]
pub type floatX = f32;

#[inline(always)]
pub fn FastLog2u16(v: u16) -> floatX {
    logs_16[v as usize]
}

#[cfg(feature = "std")]
#[inline(always)]
pub fn FastLog2(v: u64) -> floatX {
    if v < 256 {
        logs_8[v as usize]
    } else {
        (v as f32).log2() as floatX
    }
}

#[cfg(not(feature = "std"))]
#[inline(always)]
pub fn FastLog2(v: u64) -> floatX {
    if v < 256 {
        logs_8[v as usize]
    } else {
        FastLog2u64(v)
    }
}

#[cfg(feature = "std")]
#[inline(always)]
pub fn FastLog2f64(v: u64) -> floatX {
    if v < 256 {
        logs_8[v as usize]
    } else {
        (v as floatX).log2()
    }
}

#[cfg(not(feature = "std"))]
#[inline(always)]
pub fn FastLog2f64(v: u64) -> floatX {
    FastLog2(v) as floatX
}

#[inline]
pub fn FastLog2u64(v: u64) -> floatX {
    let bsr_8 = 56i8 - v.leading_zeros() as i8;
    let offset = bsr_8 & -((bsr_8 >= 0) as i8);
    (offset as floatX) + logs_8[(v >> offset) as usize]
}

#[inline(always)]
pub fn FastLog2u32(v: i32) -> floatX {
    let bsr_8 = 24i8 - v.leading_zeros() as i8;
    let offset = bsr_8 & -((bsr_8 >= 0) as i8);
    (offset as floatX) + logs_8[(v >> offset) as usize]
}

#[inline(always)]
pub fn xFastLog2u16(v: u16) -> floatX {
    let bsr_8 = 8i8 - v.leading_zeros() as i8;
    let offset = (bsr_8 & -((bsr_8 >= 0) as i8));
    (offset as floatX) + logs_8[(v >> offset) as usize]
}

#[cfg(feature = "std")]
#[inline(always)]
pub fn FastPow2(v: floatX) -> floatX {
    (2 as floatX).powf(v)
}

#[cfg(not(feature = "std"))]
#[inline(always)]
pub fn FastPow2(v: floatX) -> floatX {
    assert!(v >= 0 as floatX);
    let round_down = v as i32;
    let remainder = v - round_down as floatX;
    let mut x = 1 as floatX;
    // (1 + (x/n) * ln2) ^ n
    // let n = 8
    x += remainder * (0.693147180559945309417232121458 / 256.0) as floatX;
    x *= x;
    x *= x;
    x *= x;
    x *= x;
    x *= x;
    x *= x;
    x *= x;
    x *= x;

    (1 << round_down) as floatX * x
}

#[inline(always)]
pub fn Log2FloorNonZero(v: u64) -> u32 {
    63u32 ^ v.leading_zeros()
}

#[cfg(test)]
mod test {
    fn baseline_log2_floor_non_zero(mut n: u64) -> u32 {
        let mut result: u32 = 0;
        while {
            n >>= 1i32;
            n
        } != 0
        {
            result = result.wrapping_add(1);
        }
        result
    }

    #[test]
    fn log2floor_non_zero_works() {
        let examples = [
            4u64,
            254,
            256,
            1428,
            25412509,
            21350891256,
            65536,
            1258912591,
            60968101,
            1,
            12589125190825,
            105912059215091,
            0,
        ];
        for example in examples.iter() {
            let fast_version = super::Log2FloorNonZero(*example);
            let baseline_version = baseline_log2_floor_non_zero(*example);
            if *example != 0 {
                // make sure we don't panic when computing...but don't care about result
                assert_eq!(fast_version, baseline_version);
            }
        }
    }
    pub fn approx_eq(a: f64, b: f64, tol: f64) {
        let mut t0 = a - b;
        let mut t1 = b - a;
        if t0 < 0.0 {
            t0 = -t0;
        }
        if t1 < 0.0 {
            t1 = -t1;
        }
        if (!(t1 < tol)) {
            assert_eq!(a, b);
        }
        if (!(t0 < tol)) {
            assert_eq!(a, b);
        }
    }
    #[test]
    fn fast_log2_works() {
        let examples = [
            4u64,
            254,
            256,
            1428,
            25412509,
            21350891256,
            65536,
            1258912591,
            60968101,
            1,
            12589125190825,
            105912059215091,
            0,
        ];
        let tol = [
            0.00001, 0.0001, 0.0001, 0.005, 0.007, 0.008, 0.01, 0.01, 0.01, 0.000001, 0.01, 0.01,
            0.0001,
        ];
        for (index, example) in examples.iter().enumerate() {
            let fast_version = super::FastLog2(*example);
            if *example != 0 {
                // make sure we don't panic when computing...but don't care about result
                let baseline_version = (*example as f64).log2();
                approx_eq(fast_version as f64, baseline_version, tol[index]);
            } else {
                //assert_eq!(fast_version as f64, 0.0 as f64);
            }
        }
    }
}