optimization: Add single byte writes

Single byte writes are used very often (used for every (u)int8_t or std::byte or bool and for every VarInt's first byte which is also needed for every (pre)Vector).
It makes sense to avoid the generalized serialization infrastructure that isn't needed:
* AutoFile write doesn't need to allocate 4k buffer for a single byte now;
* `VectorWriter` and `DataStream` avoids memcpy/insert calls.

> cmake -B build -DBUILD_BENCH=ON -DCMAKE_BUILD_TYPE=Release && cmake --build build -j$(nproc) && build/src/bench/bench_bitcoin -filter='SizeComputerBlock|SerializeBlock|DeserializeBlock' --min-time=10000

> C compiler ............................ AppleClang 16.0.0.16000026

|            ns/block |             block/s |    err% |     total | benchmark
|--------------------:|--------------------:|--------:|----------:|:----------
|          847,934.56 |            1,179.34 |    0.2% |     10.82 | `DeserializeBlock`
|          178,475.58 |            5,603.01 |    0.3% |     11.03 | `SerializeBlock`
|           10,234.19 |           97,711.73 |    0.1% |     11.02 | `SizeComputerBlock`

> C++ compiler .......................... GNU 13.3.0

TODO

src/bench/checkblock.cpp

@@ -52,7 +52,7 @@ static void DeserializeBlock(benchmark::Bench& bench)
 {
     DataStream stream(benchmark::data::block413567);
     std::byte a{0};
-    stream.write({&a, 1}); // Prevent compaction
+    stream.write(std::span{&a, 1}); // Prevent compaction
 
     bench.unit("block").run([&] {
         CBlock block;
@@ -66,7 +66,7 @@ static void DeserializeAndCheckBlock(benchmark::Bench& bench)
 {
     DataStream stream(benchmark::data::block413567);
     std::byte a{0};
-    stream.write({&a, 1}); // Prevent compaction
+    stream.write(std::span{&a, 1}); // Prevent compaction
 
     ArgsManager bench_args;
     const auto chainParams = CreateChainParams(bench_args, ChainType::MAIN);

src/bench/rpc_blockchain.cpp

@@ -33,7 +33,7 @@ struct TestBlockAndIndex {
     {
         DataStream stream{benchmark::data::block413567};
         std::byte a{0};
-        stream.write({&a, 1}); // Prevent compaction
+        stream.write(std::span{&a, 1}); // Prevent compaction
 
         stream >> TX_WITH_WITNESS(block);
 

src/crypto/sha256.cpp

@@ -723,6 +723,21 @@ CSHA256& CSHA256::Write(const unsigned char* data, size_t len)
     }
     return *this;
 }
+CSHA256& CSHA256::Write(unsigned char data)
+{
+    size_t bufsize = bytes % 64;
+
+    // Add the single byte to the buffer
+    buf[bufsize] = data;
+    bytes += 1;
+
+    if (bufsize == 63) {
+        // Process the buffer if full
+        Transform(s, buf, 1);
+    }
+
+    return *this;
+}
 
 void CSHA256::Finalize(unsigned char hash[OUTPUT_SIZE])
 {

src/crypto/sha256.h

@@ -22,6 +22,7 @@ public:
 
     CSHA256();
     CSHA256& Write(const unsigned char* data, size_t len);
+    CSHA256& Write(unsigned char data);
     void Finalize(unsigned char hash[OUTPUT_SIZE]);
     CSHA256& Reset();
 };

src/hash.h

@@ -38,6 +38,10 @@ public:
         sha.Write(input.data(), input.size());
         return *this;
     }
+    CHash256& Write(std::span<const unsigned char, 1> input) {
+        sha.Write(input[0]);
+        return *this;
+    }
 
     CHash256& Reset() {
         sha.Reset();
@@ -63,6 +67,10 @@ public:
         sha.Write(input.data(), input.size());
         return *this;
     }
+    CHash160& Write(std::span<const unsigned char, 1> input) {
+        sha.Write(input[0]);
+        return *this;
+    }
 
     CHash160& Reset() {
         sha.Reset();
@@ -107,6 +115,10 @@ public:
     {
         ctx.Write(UCharCast(src.data()), src.size());
     }
+    void write(std::span<const std::byte, 1> src)
+    {
+        ctx.Write(*UCharCast(&src[0]));
+    }
 
     /** Compute the double-SHA256 hash of all data written to this object.
      *
@@ -160,13 +172,18 @@ public:
         m_source.read(dst);
         this->write(dst);
     }
+    void read(std::span<std::byte, 1> dst)
+    {
+        m_source.read(dst);
+        this->write(std::span<const std::byte, 1>{dst});
+    }
 
     void ignore(size_t num_bytes)
     {
         std::byte data[1024];
         while (num_bytes > 0) {
             size_t now = std::min<size_t>(num_bytes, 1024);
-            read({data, now});
+            read(std::span{data, now});
             num_bytes -= now;
         }
     }
@@ -194,6 +211,11 @@ public:
         m_source.write(src);
         HashWriter::write(src);
     }
+    void write(std::span<const std::byte, 1> src)
+    {
+        m_source.write(src);
+        HashWriter::write(src);
+    }
 
     template <typename T>
     HashedSourceWriter& operator<<(const T& obj)

src/serialize.h

@@ -63,56 +63,56 @@ concept ContainsSizeComputer = ContainsStream<T> &&
  */
 template<typename Stream> void ser_writedata8(Stream &s, uint8_t obj)
 {
-    s.write(std::as_bytes(std::span{&obj, 1}));
+    s.write(std::as_bytes(std::span<uint8_t, 1>{&obj, 1}));
 }
 template<typename Stream> void ser_writedata16(Stream &s, uint16_t obj)
 {
     obj = htole16_internal(obj);
-    s.write(std::as_bytes(std::span{&obj, 1}));
+    s.write(std::as_bytes(std::span<uint16_t, 1>{&obj, 1}));
 }
 template<typename Stream> void ser_writedata32(Stream &s, uint32_t obj)
 {
     obj = htole32_internal(obj);
-    s.write(std::as_bytes(std::span{&obj, 1}));
+    s.write(std::as_bytes(std::span<uint32_t, 1>{&obj, 1}));
 }
 template<typename Stream> void ser_writedata32be(Stream &s, uint32_t obj)
 {
     obj = htobe32_internal(obj);
-    s.write(std::as_bytes(std::span{&obj, 1}));
+    s.write(std::as_bytes(std::span<uint32_t, 1>{&obj, 1}));
 }
 template<typename Stream> void ser_writedata64(Stream &s, uint64_t obj)
 {
     obj = htole64_internal(obj);
-    s.write(std::as_bytes(std::span{&obj, 1}));
+    s.write(std::as_bytes(std::span<uint64_t, 1>{&obj, 1}));
 }
 template<typename Stream> uint8_t ser_readdata8(Stream &s)
 {
     uint8_t obj;
-    s.read(std::as_writable_bytes(std::span{&obj, 1}));
+    s.read(std::as_writable_bytes(std::span<uint8_t, 1>{&obj, 1}));
     return obj;
 }
 template<typename Stream> uint16_t ser_readdata16(Stream &s)
 {
     uint16_t obj;
-    s.read(std::as_writable_bytes(std::span{&obj, 1}));
+    s.read(std::as_writable_bytes(std::span<uint16_t, 1>{&obj, 1}));
     return le16toh_internal(obj);
 }
 template<typename Stream> uint32_t ser_readdata32(Stream &s)
 {
     uint32_t obj;
-    s.read(std::as_writable_bytes(std::span{&obj, 1}));
+    s.read(std::as_writable_bytes(std::span<uint32_t, 1>{&obj, 1}));
     return le32toh_internal(obj);
 }
 template<typename Stream> uint32_t ser_readdata32be(Stream &s)
 {
     uint32_t obj;
-    s.read(std::as_writable_bytes(std::span{&obj, 1}));
+    s.read(std::as_writable_bytes(std::span<uint32_t, 1>{&obj, 1}));
     return be32toh_internal(obj);
 }
 template<typename Stream> uint64_t ser_readdata64(Stream &s)
 {
     uint64_t obj;
-    s.read(std::as_writable_bytes(std::span{&obj, 1}));
+    s.read(std::as_writable_bytes(std::span<uint64_t, 1>{&obj, 1}));
     return le64toh_internal(obj);
 }
 
@@ -319,7 +319,6 @@ template <typename Stream, ByteOrIntegral T> void Unserialize(Stream& s, T& a)
 template <typename Stream, BasicByte B, int N> void Unserialize(Stream& s, B (&a)[N]) { s.read(MakeWritableByteSpan(a)); }
 template <typename Stream, BasicByte B, std::size_t N> void Unserialize(Stream& s, std::array<B, N>& a) { s.read(MakeWritableByteSpan(a)); }
 template <typename Stream, BasicByte B, std::size_t N> void Unserialize(Stream& s, std::span<B, N> span) { s.read(std::as_writable_bytes(span)); }
-template <typename Stream, BasicByte B> void Unserialize(Stream& s, std::span<B> span) { s.read(std::as_writable_bytes(span)); }
 // clang-format on
 
 
@@ -577,10 +576,10 @@ struct CustomUintFormatter
             s.GetStream().seek(Bytes);
         } else if (BigEndian) {
             uint64_t raw = htobe64_internal(v);
-            s.write(std::as_bytes(std::span{&raw, 1}).last(Bytes));
+            s.write(std::as_bytes(std::span{&raw, 1}).template last<Bytes>());
         } else {
             uint64_t raw = htole64_internal(v);
-            s.write(std::as_bytes(std::span{&raw, 1}).first(Bytes));
+            s.write(std::as_bytes(std::span{&raw, 1}).template first<Bytes>());
         }
     }
 
@@ -590,10 +589,10 @@ struct CustomUintFormatter
         static_assert(std::numeric_limits<U>::max() >= MAX && std::numeric_limits<U>::min() <= 0, "Assigned type too small");
         uint64_t raw = 0;
         if (BigEndian) {
-            s.read(std::as_writable_bytes(std::span{&raw, 1}).last(Bytes));
+            s.read(std::as_writable_bytes(std::span{&raw, 1}).last<Bytes>());
             v = static_cast<I>(be64toh_internal(raw));
         } else {
-            s.read(std::as_writable_bytes(std::span{&raw, 1}).first(Bytes));
+            s.read(std::as_writable_bytes(std::span{&raw, 1}).first<Bytes>());
             v = static_cast<I>(le64toh_internal(raw));
         }
     }
@@ -1113,6 +1112,10 @@ public:
     {
         this->nSize += src.size();
     }
+    void write(std::span<const std::byte, 1>)
+    {
+        this->nSize += 1;
+    }
 
     /** Pretend _nSize bytes are written, without specifying them. */
     void seek(size_t _nSize)
@@ -1162,7 +1165,9 @@ public:
     template <typename U> ParamsStream& operator<<(const U& obj) { ::Serialize(*this, obj); return *this; }
     template <typename U> ParamsStream& operator>>(U&& obj) { ::Unserialize(*this, obj); return *this; }
     void write(std::span<const std::byte> src) { GetStream().write(src); }
+    void write(std::span<const std::byte, 1> src) { GetStream().write(src); }
     void read(std::span<std::byte> dst) { GetStream().read(dst); }
+    void read(std::span<std::byte, 1> dst) { GetStream().read(dst); }
     void ignore(size_t num) { GetStream().ignore(num); }
     bool eof() const { return GetStream().eof(); }
     size_t size() const { return GetStream().size(); }

src/streams.cpp

@@ -63,6 +63,12 @@ void AutoFile::read(std::span<std::byte> dst)
         throw std::ios_base::failure(feof() ? "AutoFile::read: end of file" : "AutoFile::read: fread failed");
     }
 }
+void AutoFile::read(std::span<std::byte, 1> dst)
+{
+    if (detail_fread(dst) != 1) {
+        throw std::ios_base::failure(feof() ? "AutoFile::read: end of file" : "AutoFile::read: fread failed");
+    }
+}
 
 void AutoFile::ignore(size_t nSize)
 {
@@ -101,6 +107,25 @@ void AutoFile::write(std::span<const std::byte> src)
         }
     }
 }
+void AutoFile::write(std::span<const std::byte, 1> src)
+{
+    if (!m_file) throw std::ios_base::failure("AutoFile::write: file handle is nullptr");
+    if (m_xor.empty()) {
+        if (std::fwrite(src.data(), 1, 1, m_file) != 1) {
+            throw std::ios_base::failure("AutoFile::write: write failed");
+        }
+        if (m_position.has_value()) *m_position += 1;
+    } else {
+        if (!m_position.has_value()) throw std::ios_base::failure("AutoFile::write: position unknown");
+        std::byte temp_byte = src[0];
+        std::span val(&temp_byte, 1);
+        util::Xor(val, m_xor, *m_position);
+        if (fwrite(val.data(), 1, 1, m_file) != 1) {
+            throw std::ios_base::failure{"XorFile::write: failed"};
+        }
+        *m_position += 1;
+    }
+}
 
 bool AutoFile::Commit()
 {

src/streams.h

@@ -83,6 +83,17 @@ public:
         }
         nPos += src.size();
     }
+    void write(std::span<const std::byte, 1> src)
+    {
+        assert(nPos <= vchData.size());
+        const auto byte{*UCharCast(&src[0])};
+        if (nPos < vchData.size()) {
+            vchData[nPos] = byte;
+        } else {
+            vchData.push_back(byte);
+        }
+        nPos += 1;
+    }
     template <typename T>
     VectorWriter& operator<<(const T& obj)
     {
@@ -254,6 +265,10 @@ public:
         // Write to the end of the buffer
         vch.insert(vch.end(), src.begin(), src.end());
     }
+    void write(std::span<const value_type, 1> src)
+    {
+        vch.push_back(src[0]);
+    }
 
     template<typename T>
     DataStream& operator<<(const T& obj)
@@ -450,8 +465,10 @@ public:
     // Stream subset
     //
     void read(std::span<std::byte> dst);
+    void read(std::span<std::byte, 1> dst);
     void ignore(size_t nSize);
     void write(std::span<const std::byte> src);
+    void write(std::span<const std::byte, 1> src);
 
     template <typename T>
     AutoFile& operator<<(const T& obj)

src/test/crypto_tests.cpp

@@ -1079,7 +1079,7 @@ BOOST_AUTO_TEST_CASE(sha256d64)
             in[j] = m_rng.randbits(8);
         }
         for (int j = 0; j < i; ++j) {
-            CHash256().Write({in + 64 * j, 64}).Finalize({out1 + 32 * j, 32});
+            CHash256().Write(std::span{in + 64 * j, 64}).Finalize({out1 + 32 * j, 32});
         }
         SHA256D64(out2, in, i);
         BOOST_CHECK(memcmp(out1, out2, 32 * i) == 0);

src/test/fuzz/autofile.cpp

@@ -29,14 +29,14 @@ FUZZ_TARGET(autofile)
             [&] {
                 std::array<std::byte, 4096> arr{};
                 try {
-                    auto_file.read({arr.data(), fuzzed_data_provider.ConsumeIntegralInRange<size_t>(0, 4096)});
+                    auto_file.read(std::span{arr.data(), fuzzed_data_provider.ConsumeIntegralInRange<size_t>(0, 4096)});
                 } catch (const std::ios_base::failure&) {
                 }
             },
             [&] {
                 const std::array<std::byte, 4096> arr{};
                 try {
-                    auto_file.write({arr.data(), fuzzed_data_provider.ConsumeIntegralInRange<size_t>(0, 4096)});
+                    auto_file.write(std::span{arr.data(), fuzzed_data_provider.ConsumeIntegralInRange<size_t>(0, 4096)});
                 } catch (const std::ios_base::failure&) {
                 }
             },