// Protocol Buffers - Google's data interchange format // Copyright 2023 Google Inc. All rights reserved. // // Use of this source code is governed by a BSD-style // license that can be found in the LICENSE file or at // https://developers.google.com/open-source/licenses/bsd #ifndef GOOGLE_PROTOBUF_VARINT_SHUFFLE_H__ #define GOOGLE_PROTOBUF_VARINT_SHUFFLE_H__ #include #include #include #include #include // Must be included last. #include "google/protobuf/port_def.inc" namespace google { namespace protobuf { namespace internal { // Shifts "byte" left by n * 7 bits, filling vacated bits from `ones`. template inline PROTOBUF_ALWAYS_INLINE int64_t VarintShlByte(int8_t byte, int64_t ones) { return static_cast((static_cast(byte) << n * 7) | (static_cast(ones) >> (64 - n * 7))); } // Shifts "byte" left by n * 7 bits, filling vacated bits from `ones` and // bitwise ANDs the resulting value into the input/output `res` parameter. // Returns true if the result was not negative. template inline PROTOBUF_ALWAYS_INLINE bool VarintShlAnd(int8_t byte, int64_t ones, int64_t& res) { res &= VarintShlByte(byte, ones); return res >= 0; } // Shifts `byte` left by n * 7 bits, filling vacated bits with ones, and // puts the new value in the output only parameter `res`. // Returns true if the result was not negative. template inline PROTOBUF_ALWAYS_INLINE bool VarintShl(int8_t byte, int64_t ones, int64_t& res) { res = VarintShlByte(byte, ones); return res >= 0; } template inline PROTOBUF_ALWAYS_INLINE const char* ShiftMixParseVarint(const char* p, int64_t& res1) { using Signed = std::make_signed_t; constexpr bool kIs64BitVarint = std::is_same::value; constexpr bool kIs32BitVarint = std::is_same::value; static_assert(kIs64BitVarint || kIs32BitVarint, ""); // The algorithm relies on sign extension for each byte to set all high bits // when the varint continues. It also relies on asserting all of the lower // bits for each successive byte read. This allows the result to be aggregated // using a bitwise AND. For example: // // 8 1 64 57 ... 24 17 16 9 8 1 // ptr[0] = 1aaa aaaa ; res1 = 1111 1111 ... 1111 1111 1111 1111 1aaa aaaa // ptr[1] = 1bbb bbbb ; res2 = 1111 1111 ... 1111 1111 11bb bbbb b111 1111 // ptr[2] = 0ccc cccc ; res3 = 0000 0000 ... 000c cccc cc11 1111 1111 1111 // --------------------------------------------- // res1 & res2 & res3 = 0000 0000 ... 000c cccc ccbb bbbb baaa aaaa // // On x86-64, a shld from a single register filled with enough 1s in the high // bits can accomplish all this in one instruction. It so happens that res1 // has 57 high bits of ones, which is enough for the largest shift done. // // Just as importantly, by keeping results in res1, res2, and res3, we take // advantage of the superscalar abilities of the CPU. const auto next = [&p] { return static_cast(*p++); }; const auto last = [&p] { return static_cast(p[-1]); }; int64_t res2, res3; // accumulated result chunks res1 = next(); if (PROTOBUF_PREDICT_TRUE(res1 >= 0)) return p; if (limit <= 1) goto limit0; // Densify all ops with explicit FALSE predictions from here on, except that // we predict length = 5 as a common length for fields like timestamp. if (PROTOBUF_PREDICT_FALSE(VarintShl<1>(next(), res1, res2))) goto done1; if (limit <= 2) goto limit1; if (PROTOBUF_PREDICT_FALSE(VarintShl<2>(next(), res1, res3))) goto done2; if (limit <= 3) goto limit2; if (PROTOBUF_PREDICT_FALSE(VarintShlAnd<3>(next(), res1, res2))) goto done2; if (limit <= 4) goto limit2; if (PROTOBUF_PREDICT_TRUE(VarintShlAnd<4>(next(), res1, res3))) goto done2; if (limit <= 5) goto limit2; if (kIs64BitVarint) { if (PROTOBUF_PREDICT_FALSE(VarintShlAnd<5>(next(), res1, res2))) goto done2; if (limit <= 6) goto limit2; if (PROTOBUF_PREDICT_FALSE(VarintShlAnd<6>(next(), res1, res3))) goto done2; if (limit <= 7) goto limit2; if (PROTOBUF_PREDICT_FALSE(VarintShlAnd<7>(next(), res1, res2))) goto done2; if (limit <= 8) goto limit2; if (PROTOBUF_PREDICT_FALSE(VarintShlAnd<8>(next(), res1, res3))) goto done2; if (limit <= 9) goto limit2; } else { // An overlong int32 is expected to span the full 10 bytes if (PROTOBUF_PREDICT_FALSE(!(next() & 0x80))) goto done2; if (limit <= 6) goto limit2; if (PROTOBUF_PREDICT_FALSE(!(next() & 0x80))) goto done2; if (limit <= 7) goto limit2; if (PROTOBUF_PREDICT_FALSE(!(next() & 0x80))) goto done2; if (limit <= 8) goto limit2; if (PROTOBUF_PREDICT_FALSE(!(next() & 0x80))) goto done2; if (limit <= 9) goto limit2; } // For valid 64bit varints, the 10th byte/ptr[9] should be exactly 1. In this // case, the continuation bit of ptr[8] already set the top bit of res3 // correctly, so all we have to do is check that the expected case is true. if (PROTOBUF_PREDICT_TRUE(next() == 1)) goto done2; if (PROTOBUF_PREDICT_FALSE(last() & 0x80)) { // If the continue bit is set, it is an unterminated varint. return nullptr; } // A zero value of the first bit of the 10th byte represents an // over-serialized varint. This case should not happen, but if does (say, due // to a nonconforming serializer), deassert the continuation bit that came // from ptr[8]. if (kIs64BitVarint && (last() & 1) == 0) { static constexpr int bits = 64 - 1; #if defined(__GCC_ASM_FLAG_OUTPUTS__) && defined(__x86_64__) // Use a small instruction since this is an uncommon code path. asm("btc %[bits], %[res3]" : [res3] "+r"(res3) : [bits] "i"(bits)); #else res3 ^= int64_t{1} << bits; #endif } done2: res2 &= res3; done1: res1 &= res2; PROTOBUF_ASSUME(p != nullptr); return p; limit2: res2 &= res3; limit1: res1 &= res2; limit0: PROTOBUF_ASSUME(p != nullptr); PROTOBUF_ASSUME(res1 < 0); return p; } } // namespace internal } // namespace protobuf } // namespace google #include "google/protobuf/port_undef.inc" #endif // GOOGLE_PROTOBUF_VARINT_SHUFFLE_H__