kiddo 6.0.1

A high-performance, flexible, ergonomic k-d tree library. Ideal for geo- and astro- nearest-neighbour and k-nearest-neighbor queries
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

use std::arch::aarch64::*;

use crate::dist::distance_metric_neon::{NeonF32LeafOps, NeonF64LeafOps};

#[inline(always)]
unsafe fn pow_f64x2<const P: u32>(x: float64x2_t) -> float64x2_t {
    if P == 0 {
        return vdupq_n_f64(1.0);
    }
    if P == 1 {
        return x;
    }
    if P == 2 {
        return vmulq_f64(x, x);
    }
    if P == 3 {
        let x2 = vmulq_f64(x, x);
        return vmulq_f64(x2, x);
    }
    if P == 4 {
        let x2 = vmulq_f64(x, x);
        return vmulq_f64(x2, x2);
    }

    let mut acc = vdupq_n_f64(1.0);
    let mut base = x;
    let mut exp = P;
    while exp != 0 {
        if exp & 1 == 1 {
            acc = vmulq_f64(acc, base);
        }
        exp >>= 1;
        if exp != 0 {
            base = vmulq_f64(base, base);
        }
    }
    acc
}

#[inline(always)]
unsafe fn pow_f32x4<const P: u32>(x: float32x4_t) -> float32x4_t {
    if P == 0 {
        return vdupq_n_f32(1.0);
    }
    if P == 1 {
        return x;
    }
    if P == 2 {
        return vmulq_f32(x, x);
    }
    if P == 3 {
        let x2 = vmulq_f32(x, x);
        return vmulq_f32(x2, x);
    }
    if P == 4 {
        let x2 = vmulq_f32(x, x);
        return vmulq_f32(x2, x2);
    }

    let mut acc = vdupq_n_f32(1.0);
    let mut base = x;
    let mut exp = P;
    while exp != 0 {
        if exp & 1 == 1 {
            acc = vmulq_f32(acc, base);
        }
        exp >>= 1;
        if exp != 0 {
            base = vmulq_f32(base, base);
        }
    }
    acc
}

#[inline(always)]
fn pow_f64<const P: u32>(x: f64) -> f64 {
    if P == 0 {
        return 1.0;
    }
    if P == 1 {
        return x;
    }
    if P == 2 {
        return x * x;
    }
    if P == 3 {
        return x * x * x;
    }
    if P == 4 {
        let x2 = x * x;
        return x2 * x2;
    }
    x.powi(P as i32)
}

#[inline(always)]
fn pow_f32<const P: u32>(x: f32) -> f32 {
    if P == 0 {
        return 1.0;
    }
    if P == 1 {
        return x;
    }
    if P == 2 {
        return x * x;
    }
    if P == 3 {
        return x * x * x;
    }
    if P == 4 {
        let x2 = x * x;
        return x2 * x2;
    }
    x.powi(P as i32)
}

pub struct MinkowskiNeonF64LeafOps<const P: u32>;

impl<const P: u32> NeonF64LeafOps for MinkowskiNeonF64LeafOps<P> {
    #[inline(always)]
    unsafe fn dist_k0_f64x2(delta: float64x2_t) -> float64x2_t {
        pow_f64x2::<P>(vabsq_f64(delta))
    }

    #[inline(always)]
    unsafe fn dist_kn_f64x2(acc: float64x2_t, delta: float64x2_t) -> float64x2_t {
        vaddq_f64(acc, pow_f64x2::<P>(vabsq_f64(delta)))
    }

    #[inline(always)]
    fn dist_k0_f64x1(delta: f64) -> f64 {
        pow_f64::<P>(delta.abs())
    }

    #[inline(always)]
    fn dist_kn_f64x1(acc: f64, delta: f64) -> f64 {
        acc + pow_f64::<P>(delta.abs())
    }
}

pub struct MinkowskiNeonF32LeafOps<const P: u32>;

impl<const P: u32> NeonF32LeafOps for MinkowskiNeonF32LeafOps<P> {
    #[inline(always)]
    unsafe fn dist_k0_f32x4(delta: float32x4_t) -> float32x4_t {
        pow_f32x4::<P>(vabsq_f32(delta))
    }

    #[inline(always)]
    unsafe fn dist_kn_f32x4(acc: float32x4_t, delta: float32x4_t) -> float32x4_t {
        vaddq_f32(acc, pow_f32x4::<P>(vabsq_f32(delta)))
    }

    #[inline(always)]
    fn dist_k0_f32x1(delta: f32) -> f32 {
        pow_f32::<P>(delta.abs())
    }

    #[inline(always)]
    fn dist_kn_f32x1(acc: f32, delta: f32) -> f32 {
        acc + pow_f32::<P>(delta.abs())
    }
}