artful 0.1.1

Low-memory overhead and high-performance adaptive radix tree
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
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304

#![allow(unused_variables)]
#![allow(unused_imports)]
use crate::node::ArtNode;
use crate::node4::Node4;
use crate::node48::Node48;
use crate::ArtKey;
use crate::Header;
#[cfg(target_arch = "aarch64")]
use std::arch::aarch64::*;
#[cfg(target_arch = "x86")]
use std::arch::x86::*;
#[cfg(target_arch = "x86_64")]
use std::arch::x86_64::*;

pub(crate) struct Node16<K: ArtKey, V, const MAX_PARTIAL_LEN: usize> {
    pub(crate) header: Header<MAX_PARTIAL_LEN>,
    pub(crate) key: [u8; 16],
    pub(crate) children: [ArtNode<K, V, MAX_PARTIAL_LEN>; 16],
    pub(crate) prefixed_child: ArtNode<K, V, MAX_PARTIAL_LEN>,
}

impl<K: ArtKey, V, const MAX_PARTIAL_LEN: usize> Default for Node16<K, V, MAX_PARTIAL_LEN> {
    fn default() -> Node16<K, V, MAX_PARTIAL_LEN> {
        // Why dont' i use macro `vec![]` initialize the children?
        // just like, you know `vec![ArtNode::none(); 16].try_into()...`.
        // because it need clone and our initialization with occur on
        // the insert critical performance path. so, we manual do it.
        Node16 {
            header: Default::default(),
            key: [0; 16],
            prefixed_child: ArtNode::default(),
            children: [
                ArtNode::none(),
                ArtNode::none(),
                ArtNode::none(),
                ArtNode::none(),
                ArtNode::none(),
                ArtNode::none(),
                ArtNode::none(),
                ArtNode::none(),
                ArtNode::none(),
                ArtNode::none(),
                ArtNode::none(),
                ArtNode::none(),
                ArtNode::none(),
                ArtNode::none(),
                ArtNode::none(),
                ArtNode::none(),
            ],
        }
    }
}

impl<K: ArtKey, V, const MAX_PARTIAL_LEN: usize> Node16<K, V, MAX_PARTIAL_LEN> {
    #[inline(always)]
    pub(crate) fn is_full(&self) -> bool {
        self.header.non_null_children == 16
    }

    #[inline(always)]
    pub(crate) fn minimum_child(&self) -> Option<&ArtNode<K, V, MAX_PARTIAL_LEN>> {
        if !self.prefixed_child.is_none() {
            Some(&self.prefixed_child)
        } else {
            Some(&self.children[0])
        }
    }

    #[inline]
    fn find_less_than_index(&mut self, key: u8) -> u16 {
        let mask = (1 << self.header.non_null_children) - 1;
        #[cfg(feature = "simd")]
        #[cfg(any(target_arch = "x86", target_arch = "x86_64"))]
        if is_x86_feature_detected!("sse2") {
            unsafe {
                let lt = _mm_cmplt_epi8(
                    _mm_set1_epi8(std::mem::transmute::<u8, i8>(key)),
                    _mm_loadu_si128(self.key.as_ptr() as *const __m128i),
                );
                let bit_fields = _mm_movemask_epi8(lt) & mask;
                if bit_fields != 0 {
                    return bit_fields.trailing_zeros() as u16;
                }

                return self.header.non_null_children;
            }
        }

        #[cfg(feature = "simd")]
        #[cfg(any(target_arch = "aarch64", target_arch = "arm"))]
        if std::arch::is_aarch64_feature_detected!("neon") {
            unsafe {
                let lt = vcltq_u8(vdupq_n_u8(key), vld1q_u8(self.key.as_ptr()));
                let bit_fields = _mm_movemask_epi8(lt) & mask;
                if bit_fields != 0 {
                    return bit_fields.trailing_zeros() as u16;
                }

                return self.header.non_null_children;
            }
        }

        let mut index = 0;
        while (index < self.header.non_null_children) && self.key[index as usize] < key {
            index += 1;
        }

        index
    }

    /// Safety: grow.
    pub(crate) fn insert_child(
        &mut self,
        valid_key: (u8, bool),
        mut new_child: ArtNode<K, V, MAX_PARTIAL_LEN>,
    ) {
        assert!(self.header.non_null_children < 16);
        if !valid_key.1 {
            assert!(self.prefixed_child.is_none());
            std::mem::swap(&mut self.prefixed_child, &mut new_child);
            return;
        }

        let index = self.find_less_than_index(valid_key.0);
        if !self.children[index as usize].is_none() {
            let mut i = self.header.non_null_children;
            while i > index {
                let mut moved = std::mem::take(&mut self.children[i as usize - 1]);
                std::mem::swap(&mut self.children[i as usize], &mut moved);
                // self.children[i as usize] = std::mem::take(&mut self.children[i as usize - 1]);
                self.key[i as usize] = self.key[i as usize - 1];
                i -= 1;
            }
        }

        self.key[index as usize] = valid_key.0;
        std::mem::swap(&mut self.children[index as usize], &mut new_child);
        // self.children[index as usize] = new_child;
        self.header.non_null_children += 1;
    }

    #[inline]
    pub(crate) fn get_child(&self, key: (u8, bool)) -> Option<&ArtNode<K, V, MAX_PARTIAL_LEN>> {
        if !key.1 {
            return Some(&self.prefixed_child);
        }

        Some(&self.children[self.find_child_index(key.0)?])
    }

    pub(crate) fn get_mut_child(
        &mut self,
        key: (u8, bool),
    ) -> Option<&mut ArtNode<K, V, MAX_PARTIAL_LEN>> {
        if !key.1 {
            return Some(&mut self.prefixed_child);
        }

        Some(&mut self.children[self.find_child_index(key.0)?])
    }

    #[inline]
    fn find_child_index(&self, key: u8) -> Option<usize> {
        // simd is preferred
        #[cfg(feature = "simd")]
        #[cfg(any(target_arch = "x86", target_arch = "x86_64"))]
        if is_x86_feature_detected!("sse2") {
            unsafe {
                // vectorized node16 keys.
                // create selection mask by search byte
                let mask = _mm_set1_epi8(std::mem::transmute::<u8, i8>(key));
                // if byte in node16, where all elem in simd_keys that are equal to simd_mask equal
                // one, else 0.
                // vectorized node16 keys.
                let cmp =
                    _mm_cmpeq_epi8(mask, _mm_loadu_si128(self.key.as_ptr() as *const __m128i));
                return match _mm_movemask_epi8(cmp) & ((1 << self.header.non_null_children) - 1) {
                    0 => None,
                    bit_fields => Some(bit_fields.trailing_zeros() as usize),
                };
            };
        }

        #[cfg(feature = "simd")]
        #[cfg(any(target_arch = "aarch64", target_arch = "arm"))]
        if std::arch::is_aarch64_feature_detected!("neon") {
            unsafe {
                let simd_keys = vld1q_u8(self.key.as_ptr());
                let simd_mask = vdupq_n_u8(key);
                let eq = vceqq_u8(simd_mask, simd_keys);
                let mask = (1 << self.header.non_null_children) - 1;
                let bit_fields = _mm_movemask_epi8(eq) & mask;
                if bit_fields == 0 {
                    return None;
                }
                return Some(bit_fields.trailing_zeros() as usize);
            }
        }
        // slow path, binary search used.
        match self.key[0..self.header.non_null_children as usize].binary_search(&key) {
            Ok(idx) => Some(idx),
            Err(_) => None,
        }
    }

    #[inline(always)]
    pub(crate) fn grow(&mut self) -> Box<Node48<K, V, MAX_PARTIAL_LEN>> {
        let mut node48: Box<Node48<K, V, MAX_PARTIAL_LEN>> = Box::default();
        // copy invalid child
        std::mem::swap(&mut self.prefixed_child, &mut node48.prefixed_child);
        // copy children and key
        for i in 0..self.header.non_null_children {
            std::mem::swap(
                &mut self.children[i as usize],
                &mut node48.children[i as usize],
            );
            node48.child_index[self.key[i as usize] as usize] = i as u8;
        }
        // copy the old node header to the new grown node.
        node48.header.partial.clone_from(&self.header.partial);
        node48.header.non_null_children = self.header.non_null_children;
        // node48.header = self.header;
        node48
    }

    #[inline(always)]
    pub fn is_few(&self) -> bool {
        self.header.non_null_children < 5
    }

    pub(crate) fn remove_child(
        &mut self,
        valid_key: (u8, bool),
    ) -> Option<ArtNode<K, V, MAX_PARTIAL_LEN>> {
        if !valid_key.1 {
            assert_eq!(self.prefixed_child.is_none(), false);
            return Some(std::mem::take(&mut self.prefixed_child));
        }

        // TODO: 这一次的查找可以优化为从外部传递.
        let mut idx = self.find_child_index(valid_key.0)?;
        let child = std::mem::take(&mut self.children[idx]);
        self.key[idx] = 0;
        self.header.non_null_children -= 1;

        // to keep order
        while idx < self.header.non_null_children as usize {
            self.key[idx] = self.key[idx + 1];
            let mut moved = std::mem::take(&mut self.children[idx + 1]);
            std::mem::swap(&mut self.children[idx], &mut moved);
            idx += 1
        }

        // again remaining
        while idx < 16 {
            if !self.children[idx].is_none() {
                break;
            }
            let _ = std::mem::take(&mut self.children[idx]);
            idx += 1
        }

        Some(child)
    }

    pub(crate) fn shrink_to_fit(&mut self) -> Box<Node4<K, V, MAX_PARTIAL_LEN>> {
        let mut node4: Box<Node4<K, V, MAX_PARTIAL_LEN>> = Box::default();
        let mut node4_index = 0;
        for i in 0..self.header.non_null_children as usize {
            std::mem::swap(&mut self.children[i], &mut node4.children[node4_index]);
            node4.key[node4_index] = self.key[i];
            node4_index += 1;
        }

        std::mem::swap(&mut self.prefixed_child, &mut node4.prefixed_child);
        // node4.header = self.header;
        node4.header.partial.clone_from(&self.header.partial);
        node4.header.non_null_children = node4_index as u16;
        node4
    }
}

#[inline]
#[cfg(feature = "simd")]
#[cfg(any(target_arch = "aarch64", target_arch = "arm"))]
unsafe fn _mm_movemask_epi8(input: uint8x16_t) -> i32 {
    // Example input (half scale):
    // 0x89 FF 1D C0 00 10 99 33
    // Shift out everything but the sign bits
    // 0x01 01 00 01 00 00 01 00
    let high_bits = vreinterpretq_u16_u8(vshrq_n_u8::<7>(input));
    // Merge the even lanes together with vsra. The '??' bytes are garbage.
    // vsri could also be used, but it is slightly slower on aarch64.
    // 0x??03 ??02 ??00 ??01
    let paired16 = vreinterpretq_u32_u16(vsraq_n_u16::<7>(high_bits, high_bits));
    // Repeat with wider lanes.
    // 0x??????0B ??????04
    let paired32 = vreinterpretq_u64_u32(vsraq_n_u32::<14>(paired16, paired16));
    // 0x??????????????4B
    let paired64 = vreinterpretq_u8_u64(vsraq_n_u64::<28>(paired32, paired32));
    // Extract the low 8 bits from each lane and join.
    // 0x4B
    vgetq_lane_u8::<0>(paired64) as i32 | (vgetq_lane_u8::<8>(paired64) as i32) << 8
}