| 1 | // Protocol Buffers - Google's data interchange format |
| 2 | // Copyright 2008 Google Inc. All rights reserved. |
| 3 | // https://developers.google.com/protocol-buffers/ |
| 4 | // |
| 5 | // Redistribution and use in source and binary forms, with or without |
| 6 | // modification, are permitted provided that the following conditions are |
| 7 | // met: |
| 8 | // |
| 9 | // * Redistributions of source code must retain the above copyright |
| 10 | // notice, this list of conditions and the following disclaimer. |
| 11 | // * Redistributions in binary form must reproduce the above |
| 12 | // copyright notice, this list of conditions and the following disclaimer |
| 13 | // in the documentation and/or other materials provided with the |
| 14 | // distribution. |
| 15 | // * Neither the name of Google Inc. nor the names of its |
| 16 | // contributors may be used to endorse or promote products derived from |
| 17 | // this software without specific prior written permission. |
| 18 | // |
| 19 | // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS |
| 20 | // "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT |
| 21 | // LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR |
| 22 | // A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT |
| 23 | // OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, |
| 24 | // SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT |
| 25 | // LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, |
| 26 | // DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY |
| 27 | // THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT |
| 28 | // (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE |
| 29 | // OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
| 30 | |
| 31 | // This file contains declarations needed in generated headers for messages |
| 32 | // that use tail-call table parsing. Everything in this file is for internal |
| 33 | // use only. |
| 34 | |
| 35 | #ifndef GOOGLE_PROTOBUF_GENERATED_MESSAGE_TCTABLE_DECL_H__ |
| 36 | #define GOOGLE_PROTOBUF_GENERATED_MESSAGE_TCTABLE_DECL_H__ |
| 37 | |
| 38 | #include <array> |
| 39 | #include <cstddef> |
| 40 | #include <cstdint> |
| 41 | #include <type_traits> |
| 42 | |
| 43 | #include <google/protobuf/message_lite.h> |
| 44 | #include <google/protobuf/parse_context.h> |
| 45 | |
| 46 | // Must come last: |
| 47 | #include <google/protobuf/port_def.inc> |
| 48 | |
| 49 | namespace google { |
| 50 | namespace protobuf { |
| 51 | namespace internal { |
| 52 | |
| 53 | // Additional information about this field: |
| 54 | struct TcFieldData { |
| 55 | constexpr TcFieldData() : data(0) {} |
| 56 | |
| 57 | // Fast table entry constructor: |
| 58 | constexpr TcFieldData(uint16_t coded_tag, uint8_t hasbit_idx, uint8_t aux_idx, |
| 59 | uint16_t offset) |
| 60 | : data(uint64_t{offset} << 48 | // |
| 61 | uint64_t{aux_idx} << 24 | // |
| 62 | uint64_t{hasbit_idx} << 16 | // |
| 63 | uint64_t{coded_tag}) {} |
| 64 | |
| 65 | // Fields used in fast table parsing: |
| 66 | // |
| 67 | // Bit: |
| 68 | // +-----------+-------------------+ |
| 69 | // |63 .. 32|31 .. 0| |
| 70 | // +---------------+---------------+ |
| 71 | // : . : . : . 16|=======| [16] coded_tag() |
| 72 | // : . : . : 24|===| . : [ 8] hasbit_idx() |
| 73 | // : . : . 32|===| : . : [ 8] aux_idx() |
| 74 | // : . 48:---.---: . : . : [16] (unused) |
| 75 | // |=======| . : . : . : [16] offset() |
| 76 | // +-----------+-------------------+ |
| 77 | // |63 .. 32|31 .. 0| |
| 78 | // +---------------+---------------+ |
| 79 | |
| 80 | template <typename TagType = uint16_t> |
| 81 | TagType coded_tag() const { |
| 82 | return static_cast<TagType>(data); |
| 83 | } |
| 84 | uint8_t hasbit_idx() const { return static_cast<uint8_t>(data >> 16); } |
| 85 | uint8_t aux_idx() const { return static_cast<uint8_t>(data >> 24); } |
| 86 | uint16_t offset() const { return static_cast<uint16_t>(data >> 48); } |
| 87 | |
| 88 | // Fields used in mini table parsing: |
| 89 | // |
| 90 | // Bit: |
| 91 | // +-----------+-------------------+ |
| 92 | // |63 .. 32|31 .. 0| |
| 93 | // +---------------+---------------+ |
| 94 | // : . : . |===============| [32] tag() (decoded) |
| 95 | // |===============| . : . : [32] entry_offset() |
| 96 | // +-----------+-------------------+ |
| 97 | // |63 .. 32|31 .. 0| |
| 98 | // +---------------+---------------+ |
| 99 | |
| 100 | uint32_t tag() const { return static_cast<uint32_t>(data); } |
| 101 | uint32_t entry_offset() const { return static_cast<uint32_t>(data >> 32); } |
| 102 | |
| 103 | uint64_t data; |
| 104 | }; |
| 105 | |
| 106 | struct TcParseTableBase; |
| 107 | |
| 108 | // TailCallParseFunc is the function pointer type used in the tailcall table. |
| 109 | typedef const char* (*TailCallParseFunc)(PROTOBUF_TC_PARAM_DECL); |
| 110 | |
| 111 | namespace field_layout { |
| 112 | struct Offset { |
| 113 | uint32_t off; |
| 114 | }; |
| 115 | } // namespace field_layout |
| 116 | |
| 117 | #if defined(_MSC_VER) && !defined(_WIN64) |
| 118 | #pragma warning(push) |
| 119 | // TcParseTableBase is intentionally overaligned on 32 bit targets. |
| 120 | #pragma warning(disable : 4324) |
| 121 | #endif |
| 122 | |
| 123 | // Base class for message-level table with info for the tail-call parser. |
| 124 | struct alignas(uint64_t) TcParseTableBase { |
| 125 | // Common attributes for message layout: |
| 126 | uint16_t has_bits_offset; |
| 127 | uint16_t extension_offset; |
| 128 | uint32_t extension_range_low; |
| 129 | uint32_t extension_range_high; |
| 130 | uint32_t max_field_number; |
| 131 | uint8_t fast_idx_mask; |
| 132 | uint16_t lookup_table_offset; |
| 133 | uint32_t skipmap32; |
| 134 | uint32_t field_entries_offset; |
| 135 | uint16_t num_field_entries; |
| 136 | |
| 137 | uint16_t num_aux_entries; |
| 138 | uint32_t aux_offset; |
| 139 | |
| 140 | const MessageLite* default_instance; |
| 141 | |
| 142 | // Handler for fields which are not handled by table dispatch. |
| 143 | TailCallParseFunc fallback; |
| 144 | |
| 145 | // This constructor exactly follows the field layout, so it's technically |
| 146 | // not necessary. However, it makes it much much easier to add or re-arrange |
| 147 | // fields, because it can be overloaded with an additional constructor, |
| 148 | // temporarily allowing both old and new protocol buffer headers to be |
| 149 | // compiled. |
| 150 | constexpr TcParseTableBase( |
| 151 | uint16_t has_bits_offset, uint16_t extension_offset, |
| 152 | uint32_t extension_range_low, uint32_t extension_range_high, |
| 153 | uint32_t max_field_number, uint8_t fast_idx_mask, |
| 154 | uint16_t lookup_table_offset, uint32_t skipmap32, |
| 155 | uint32_t field_entries_offset, uint16_t num_field_entries, |
| 156 | uint16_t num_aux_entries, uint32_t aux_offset, |
| 157 | const MessageLite* default_instance, TailCallParseFunc fallback) |
| 158 | : has_bits_offset(has_bits_offset), |
| 159 | extension_offset(extension_offset), |
| 160 | extension_range_low(extension_range_low), |
| 161 | extension_range_high(extension_range_high), |
| 162 | max_field_number(max_field_number), |
| 163 | fast_idx_mask(fast_idx_mask), |
| 164 | lookup_table_offset(lookup_table_offset), |
| 165 | skipmap32(skipmap32), |
| 166 | field_entries_offset(field_entries_offset), |
| 167 | num_field_entries(num_field_entries), |
| 168 | num_aux_entries(num_aux_entries), |
| 169 | aux_offset(aux_offset), |
| 170 | default_instance(default_instance), |
| 171 | fallback(fallback) {} |
| 172 | |
| 173 | // Table entry for fast-path tailcall dispatch handling. |
| 174 | struct FastFieldEntry { |
| 175 | // Target function for dispatch: |
| 176 | TailCallParseFunc target; |
| 177 | // Field data used during parse: |
| 178 | TcFieldData bits; |
| 179 | }; |
| 180 | // There is always at least one table entry. |
| 181 | const FastFieldEntry* fast_entry(size_t idx) const { |
| 182 | return reinterpret_cast<const FastFieldEntry*>(this + 1) + idx; |
| 183 | } |
| 184 | |
| 185 | // Returns a begin iterator (pointer) to the start of the field lookup table. |
| 186 | const uint16_t* field_lookup_begin() const { |
| 187 | return reinterpret_cast<const uint16_t*>(reinterpret_cast<uintptr_t>(this) + |
| 188 | lookup_table_offset); |
| 189 | } |
| 190 | |
| 191 | // Field entry for all fields. |
| 192 | struct FieldEntry { |
| 193 | uint32_t offset; // offset in the message object |
| 194 | int32_t has_idx; // has-bit index |
| 195 | uint16_t aux_idx; // index for `field_aux`. |
| 196 | uint16_t type_card; // `FieldType` and `Cardinality` (see _impl.h) |
| 197 | }; |
| 198 | |
| 199 | // Returns a begin iterator (pointer) to the start of the field entries array. |
| 200 | const FieldEntry* field_entries_begin() const { |
| 201 | return reinterpret_cast<const FieldEntry*>( |
| 202 | reinterpret_cast<uintptr_t>(this) + field_entries_offset); |
| 203 | } |
| 204 | |
| 205 | // Auxiliary entries for field types that need extra information. |
| 206 | union FieldAux { |
| 207 | constexpr FieldAux() : message_default(nullptr) {} |
| 208 | constexpr FieldAux(bool (*enum_validator)(int)) |
| 209 | : enum_validator(enum_validator) {} |
| 210 | constexpr FieldAux(field_layout::Offset off) : offset(off.off) {} |
| 211 | constexpr FieldAux(int16_t range_start, uint16_t range_length) |
| 212 | : enum_range{.start: range_start, .length: range_length} {} |
| 213 | constexpr FieldAux(const MessageLite* msg) : message_default(msg) {} |
| 214 | bool (*enum_validator)(int); |
| 215 | struct { |
| 216 | int16_t start; // minimum enum number (if it fits) |
| 217 | uint16_t length; // length of range (i.e., max = start + length - 1) |
| 218 | } enum_range; |
| 219 | uint32_t offset; |
| 220 | const MessageLite* message_default; |
| 221 | }; |
| 222 | const FieldAux* field_aux(uint32_t idx) const { |
| 223 | return reinterpret_cast<const FieldAux*>(reinterpret_cast<uintptr_t>(this) + |
| 224 | aux_offset) + |
| 225 | idx; |
| 226 | } |
| 227 | const FieldAux* field_aux(const FieldEntry* entry) const { |
| 228 | return field_aux(idx: entry->aux_idx); |
| 229 | } |
| 230 | |
| 231 | // Field name data |
| 232 | const char* name_data() const { |
| 233 | return reinterpret_cast<const char*>(reinterpret_cast<uintptr_t>(this) + |
| 234 | aux_offset + |
| 235 | num_aux_entries * sizeof(FieldAux)); |
| 236 | } |
| 237 | }; |
| 238 | |
| 239 | #if defined(_MSC_VER) && !defined(_WIN64) |
| 240 | #pragma warning(pop) |
| 241 | #endif |
| 242 | |
| 243 | static_assert(sizeof(TcParseTableBase::FastFieldEntry) <= 16, |
| 244 | "Fast field entry is too big." ); |
| 245 | static_assert(sizeof(TcParseTableBase::FieldEntry) <= 16, |
| 246 | "Field entry is too big." ); |
| 247 | |
| 248 | template <size_t kFastTableSizeLog2, size_t kNumFieldEntries = 0, |
| 249 | size_t kNumFieldAux = 0, size_t kNameTableSize = 0, |
| 250 | size_t kFieldLookupSize = 2> |
| 251 | struct TcParseTable { |
| 252 | TcParseTableBase ; |
| 253 | |
| 254 | // Entries for each field. |
| 255 | // |
| 256 | // Fields are indexed by the lowest bits of their field number. The field |
| 257 | // number is masked to fit inside the table. Note that the parsing logic |
| 258 | // generally calls `TailCallParseTableBase::fast_entry()` instead of accessing |
| 259 | // this field directly. |
| 260 | std::array<TcParseTableBase::FastFieldEntry, (1 << kFastTableSizeLog2)> |
| 261 | fast_entries; |
| 262 | |
| 263 | // Just big enough to find all the field entries. |
| 264 | std::array<uint16_t, kFieldLookupSize> field_lookup_table; |
| 265 | // Entries for all fields: |
| 266 | std::array<TcParseTableBase::FieldEntry, kNumFieldEntries> field_entries; |
| 267 | std::array<TcParseTableBase::FieldAux, kNumFieldAux> aux_entries; |
| 268 | std::array<char, kNameTableSize> field_names; |
| 269 | }; |
| 270 | |
| 271 | // Partial specialization: if there are no aux entries, there will be no array. |
| 272 | // In C++, arrays cannot have length 0, but (C++11) std::array<T, 0> is valid. |
| 273 | // However, different implementations have different sizeof(std::array<T, 0>). |
| 274 | // Skipping the member makes offset computations portable. |
| 275 | template <size_t kFastTableSizeLog2, size_t kNumFieldEntries, |
| 276 | size_t kNameTableSize, size_t kFieldLookupSize> |
| 277 | struct TcParseTable<kFastTableSizeLog2, kNumFieldEntries, 0, kNameTableSize, |
| 278 | kFieldLookupSize> { |
| 279 | TcParseTableBase ; |
| 280 | std::array<TcParseTableBase::FastFieldEntry, (1 << kFastTableSizeLog2)> |
| 281 | fast_entries; |
| 282 | std::array<uint16_t, kFieldLookupSize> field_lookup_table; |
| 283 | std::array<TcParseTableBase::FieldEntry, kNumFieldEntries> field_entries; |
| 284 | std::array<char, kNameTableSize> field_names; |
| 285 | }; |
| 286 | |
| 287 | // Partial specialization: if there are no fields at all, then we can save space |
| 288 | // by skipping the field numbers and entries. |
| 289 | template <size_t kNameTableSize, size_t kFieldLookupSize> |
| 290 | struct TcParseTable<0, 0, 0, kNameTableSize, kFieldLookupSize> { |
| 291 | TcParseTableBase ; |
| 292 | // N.B.: the fast entries are sized by log2, so 2**0 fields = 1 entry. |
| 293 | // The fast parsing loop will always use this entry, so it must be present. |
| 294 | std::array<TcParseTableBase::FastFieldEntry, 1> fast_entries; |
| 295 | std::array<uint16_t, kFieldLookupSize> field_lookup_table; |
| 296 | std::array<char, kNameTableSize> field_names; |
| 297 | }; |
| 298 | |
| 299 | static_assert(std::is_standard_layout<TcParseTable<1>>::value, |
| 300 | "TcParseTable must be standard layout." ); |
| 301 | |
| 302 | static_assert(offsetof(TcParseTable<1>, fast_entries) == |
| 303 | sizeof(TcParseTableBase), |
| 304 | "Table entries must be laid out after TcParseTableBase." ); |
| 305 | |
| 306 | } // namespace internal |
| 307 | } // namespace protobuf |
| 308 | } // namespace google |
| 309 | |
| 310 | #include <google/protobuf/port_undef.inc> |
| 311 | |
| 312 | #endif // GOOGLE_PROTOBUF_GENERATED_MESSAGE_TCTABLE_DECL_H__ |
| 313 | |