Line | Count | Source |
1 | | // Licensed to the Apache Software Foundation (ASF) under one |
2 | | // or more contributor license agreements. See the NOTICE file |
3 | | // distributed with this work for additional information |
4 | | // regarding copyright ownership. The ASF licenses this file |
5 | | // to you under the Apache License, Version 2.0 (the |
6 | | // "License"); you may not use this file except in compliance |
7 | | // with the License. You may obtain a copy of the License at |
8 | | // |
9 | | // http://www.apache.org/licenses/LICENSE-2.0 |
10 | | // |
11 | | // Unless required by applicable law or agreed to in writing, |
12 | | // software distributed under the License is distributed on an |
13 | | // "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY |
14 | | // KIND, either express or implied. See the License for the |
15 | | // specific language governing permissions and limitations |
16 | | // under the License. |
17 | | // This file is copied from |
18 | | // https://github.com/apache/impala/blob/branch-2.9.0/be/src/util/bitmap.h |
19 | | // and modified by Doris |
20 | | |
21 | | #pragma once |
22 | | |
23 | | #include <glog/logging.h> |
24 | | #include <stdint.h> |
25 | | #include <string.h> |
26 | | |
27 | | #include <algorithm> |
28 | | #include <string> |
29 | | #include <vector> |
30 | | |
31 | | #include "common/cast_set.h" |
32 | | #include "util/bit_util.h" |
33 | | |
34 | | namespace doris { |
35 | | |
36 | | // Return the number of bytes necessary to store the given number of bits. |
37 | 30.3k | inline size_t BitmapSize(size_t num_bits) { |
38 | 30.3k | return (num_bits + 7) / 8; |
39 | 30.3k | } |
40 | | |
41 | | // Set the given bit. |
42 | 2.80k | inline void BitmapSet(uint8_t* bitmap, size_t idx) { |
43 | 2.80k | bitmap[idx >> 3] |= 1 << (idx & 7); |
44 | 2.80k | } |
45 | | |
46 | | // Switch the given bit to the specified value. |
47 | 651k | inline void BitmapChange(uint8_t* bitmap, size_t idx, bool value) { |
48 | 651k | bitmap[idx >> 3] = |
49 | 651k | cast_set<uint8_t>((bitmap[idx >> 3] & ~(1 << (idx & 7))) | ((!!value) << (idx & 7))); |
50 | 651k | } |
51 | | |
52 | | // Clear the given bit. |
53 | 1 | inline void BitmapClear(uint8_t* bitmap, size_t idx) { |
54 | 1 | bitmap[idx >> 3] &= ~(1 << (idx & 7)); |
55 | 1 | } |
56 | | |
57 | | // Test/get the given bit. |
58 | 1.30M | inline bool BitmapTest(const uint8_t* bitmap, size_t idx) { |
59 | 1.30M | return bitmap[idx >> 3] & (1 << (idx & 7)); |
60 | 1.30M | } |
61 | | |
62 | | // Merge the two bitmaps using bitwise or. Both bitmaps should have at least |
63 | | // n_bits valid bits. |
64 | 0 | inline void BitmapMergeOr(uint8_t* dst, const uint8_t* src, size_t n_bits) { |
65 | 0 | size_t n_bytes = BitmapSize(n_bits); |
66 | 0 | for (size_t i = 0; i < n_bytes; i++) { |
67 | 0 | *dst++ |= *src++; |
68 | 0 | } |
69 | 0 | } |
70 | | |
71 | | // Set bits from offset to (offset + num_bits) to the specified value |
72 | | void BitmapChangeBits(uint8_t* bitmap, size_t offset, size_t num_bits, bool value); |
73 | | |
74 | | // Find the first bit of the specified value, starting from the specified offset. |
75 | | bool BitmapFindFirst(const uint8_t* bitmap, size_t offset, size_t bitmap_size, bool value, |
76 | | size_t* idx); |
77 | | |
78 | | // Find the first set bit in the bitmap, at the specified offset. |
79 | | inline bool BitmapFindFirstSet(const uint8_t* bitmap, size_t offset, size_t bitmap_size, |
80 | 2 | size_t* idx) { |
81 | 2 | return BitmapFindFirst(bitmap, offset, bitmap_size, true, idx); |
82 | 2 | } |
83 | | |
84 | | // Find the first zero bit in the bitmap, at the specified offset. |
85 | | inline bool BitmapFindFirstZero(const uint8_t* bitmap, size_t offset, size_t bitmap_size, |
86 | 2 | size_t* idx) { |
87 | 2 | return BitmapFindFirst(bitmap, offset, bitmap_size, false, idx); |
88 | 2 | } |
89 | | |
90 | | // Returns true if the bitmap contains only ones. |
91 | 0 | inline bool BitMapIsAllSet(const uint8_t* bitmap, size_t offset, size_t bitmap_size) { |
92 | 0 | DCHECK_LT(offset, bitmap_size); |
93 | 0 | size_t idx; |
94 | 0 | return !BitmapFindFirstZero(bitmap, offset, bitmap_size, &idx); |
95 | 0 | } |
96 | | |
97 | | // Returns true if the bitmap contains only zeros. |
98 | 0 | inline bool BitmapIsAllZero(const uint8_t* bitmap, size_t offset, size_t bitmap_size) { |
99 | 0 | DCHECK_LT(offset, bitmap_size); |
100 | 0 | size_t idx; |
101 | 0 | return !BitmapFindFirstSet(bitmap, offset, bitmap_size, &idx); |
102 | 0 | } |
103 | | |
104 | | // Returns true if the two bitmaps are equal. |
105 | | // |
106 | | // It is assumed that both bitmaps have 'bitmap_size' number of bits. |
107 | 0 | inline bool BitmapEquals(const uint8_t* bm1, const uint8_t* bm2, size_t bitmap_size) { |
108 | 0 | size_t num_full_bytes = bitmap_size >> 3; |
109 | 0 | if (memcmp(bm1, bm2, num_full_bytes)) { |
110 | 0 | return false; |
111 | 0 | } |
112 | 0 |
|
113 | 0 | // Check any remaining bits in one extra operation. |
114 | 0 | size_t num_remaining_bits = bitmap_size - (num_full_bytes << 3); |
115 | 0 | if (num_remaining_bits == 0) { |
116 | 0 | return true; |
117 | 0 | } |
118 | 0 | DCHECK_LT(num_remaining_bits, 8); |
119 | 0 | auto mask = (1 << num_remaining_bits) - 1; |
120 | 0 | return (bm1[num_full_bytes] & mask) == (bm2[num_full_bytes] & mask); |
121 | 0 | } |
122 | | |
123 | | // Iterator which yields ranges of set and unset bits. |
124 | | // Example usage: |
125 | | // bool value; |
126 | | // size_t size; |
127 | | // BitmapIterator iter(bitmap, n_bits); |
128 | | // while ((size = iter.Next(&value))) { |
129 | | // printf("bitmap block len=%lu value=%d\n", size, value); |
130 | | // } |
131 | | class BitmapIterator { |
132 | | public: |
133 | | BitmapIterator(const uint8_t* map, size_t num_bits) |
134 | 651k | : offset_(0), num_bits_(num_bits), map_(map) {} |
135 | | |
136 | 0 | void Reset(const uint8_t* map, size_t num_bits) { |
137 | 0 | offset_ = 0; |
138 | 0 | num_bits_ = num_bits; |
139 | 0 | map_ = map; |
140 | 0 | } |
141 | | |
142 | 0 | void Reset() { |
143 | 0 | offset_ = 0; |
144 | 0 | num_bits_ = 0; |
145 | 0 | map_ = nullptr; |
146 | 0 | } |
147 | | |
148 | 3 | bool done() const { return (num_bits_ - offset_) == 0; } |
149 | | |
150 | 2 | void SeekTo(size_t bit) { |
151 | 2 | DCHECK_LE(bit, num_bits_); |
152 | 2 | offset_ = bit; |
153 | 2 | } |
154 | | |
155 | 1 | size_t Next(bool* value, size_t max_run) { |
156 | 1 | return NextWithLimit(value, std::min(num_bits_, max_run + offset_)); |
157 | 1 | } |
158 | | |
159 | 1.30M | size_t Next(bool* value) { return NextWithLimit(value, num_bits_); } |
160 | | |
161 | | private: |
162 | 1.30M | size_t NextWithLimit(bool* value, size_t limit) { |
163 | 1.30M | size_t len = limit - offset_; |
164 | 1.30M | if (len == 0) [[unlikely]] { |
165 | 651k | return (0); |
166 | 651k | } |
167 | | |
168 | 651k | *value = BitmapTest(map_, offset_); |
169 | | |
170 | 651k | size_t index; |
171 | 651k | if (BitmapFindFirst(map_, offset_, limit, !(*value), &index)) { |
172 | 5 | len = index - offset_; |
173 | 651k | } else { |
174 | 651k | index = limit; |
175 | 651k | } |
176 | | |
177 | 651k | offset_ = index; |
178 | 651k | return len; |
179 | 1.30M | } |
180 | | |
181 | | private: |
182 | | size_t offset_; |
183 | | size_t num_bits_; |
184 | | const uint8_t* map_ = nullptr; |
185 | | }; |
186 | | |
187 | | /// Bitmap vector utility class. |
188 | | /// TODO: investigate perf. |
189 | | /// - Precomputed bitmap |
190 | | /// - Explicit Set/Unset() apis |
191 | | /// - Bigger words |
192 | | /// - size bitmap to Mersenne prime. |
193 | | class Bitmap { |
194 | | public: |
195 | 9 | Bitmap(int64_t num_bits) { |
196 | 9 | DCHECK_GE(num_bits, 0); |
197 | 9 | buffer_.resize(BitUtil::round_up_numi_64(cast_set<uint32_t>(num_bits))); |
198 | 9 | num_bits_ = num_bits; |
199 | 9 | } |
200 | | |
201 | | /// Resize bitmap and set all bits to zero. |
202 | 11 | void Reset(int64_t num_bits) { |
203 | 11 | DCHECK_GE(num_bits, 0); |
204 | 11 | buffer_.resize(BitUtil::round_up_numi_64(cast_set<uint32_t>(num_bits))); |
205 | 11 | num_bits_ = num_bits; |
206 | 11 | SetAllBits(false); |
207 | 11 | } |
208 | | |
209 | | /// Compute memory usage of a bitmap, not including the Bitmap object itself. |
210 | 0 | static int64_t MemUsage(int64_t num_bits) { |
211 | 0 | DCHECK_GE(num_bits, 0); |
212 | 0 | return BitUtil::round_up_numi_64(cast_set<uint32_t>(num_bits)) * sizeof(int64_t); |
213 | 0 | } |
214 | | |
215 | | /// Compute memory usage of this bitmap, not including the Bitmap object itself. |
216 | 0 | int64_t MemUsage() const { return MemUsage(num_bits_); } |
217 | | |
218 | | /// Sets the bit at 'bit_index' to v. |
219 | 0 | void Set(int64_t bit_index, bool v) { |
220 | 0 | int64_t word_index = bit_index >> NUM_OFFSET_BITS; |
221 | 0 | bit_index &= BIT_INDEX_MASK; |
222 | 0 | DCHECK_LT(word_index, buffer_.size()); |
223 | 0 | if (v) { |
224 | 0 | buffer_[word_index] |= (1LL << bit_index); |
225 | 0 | } else { |
226 | 0 | buffer_[word_index] &= ~(1LL << bit_index); |
227 | 0 | } |
228 | 0 | } |
229 | | |
230 | | /// Returns true if the bit at 'bit_index' is set. |
231 | 0 | bool Get(int64_t bit_index) const { |
232 | 0 | int64_t word_index = bit_index >> NUM_OFFSET_BITS; |
233 | 0 | bit_index &= BIT_INDEX_MASK; |
234 | 0 | DCHECK_LT(word_index, buffer_.size()); |
235 | 0 | return (buffer_[word_index] & (1LL << bit_index)) != 0; |
236 | 0 | } |
237 | | |
238 | 11 | void SetAllBits(bool b) { memset(&buffer_[0], 255 * b, buffer_.size() * sizeof(uint64_t)); } |
239 | | |
240 | 0 | int64_t num_bits() const { return num_bits_; } |
241 | | |
242 | | /// If 'print_bits' prints 0/1 per bit, otherwise it prints the int64_t value. |
243 | | std::string DebugString(bool print_bits) const; |
244 | | |
245 | | private: |
246 | | std::vector<uint64_t> buffer_; |
247 | | int64_t num_bits_; |
248 | | |
249 | | /// Used for bit shifting and masking for the word and offset calculation. |
250 | | static const int64_t NUM_OFFSET_BITS = 6; |
251 | | static const int64_t BIT_INDEX_MASK = 63; |
252 | | }; |
253 | | } // namespace doris |