Internal changes

PiperOrigin-RevId: 822002697

snappy.cc

@@ -280,6 +280,8 @@ inline char* IncrementalCopySlow(const char* src, char* op,
 // 4, 5, 0, 1, 2, 3, 4, 5, 0, 1}. These byte index sequences are generated by
 // calling MakePatternMaskBytes(0, 6, index_sequence<16>()) and
 // MakePatternMaskBytes(16, 6, index_sequence<16>()) respectively.
+
+
 template <size_t... indexes>
 inline constexpr std::array<char, sizeof...(indexes)> MakePatternMaskBytes(
     int index_offset, int pattern_size, index_sequence<indexes...>) {
@@ -297,7 +299,6 @@ MakePatternMaskBytesTable(int index_offset,
       MakePatternMaskBytes(index_offset, pattern_sizes_minus_one + 1,
                            make_index_sequence</*indexes=*/sizeof(V128)>())...};
 }
-
 // This is an array of shuffle control masks that can be used as the source
 // operand for PSHUFB to permute the contents of the destination XMM register
 // into a repeating byte pattern.
@@ -493,7 +494,6 @@ inline char* IncrementalCopy(const char* src, char* op, char* const op_limit,
           LoadPatternAndReshuffleMask(src, pattern_size);
       V128 pattern = pattern_and_reshuffle_mask.first;
       V128 reshuffle_mask = pattern_and_reshuffle_mask.second;
-
       // There is at least one, and at most four 16-byte blocks. Writing four
       // conditionals instead of a loop allows FDO to layout the code with
       // respect to the actual probabilities of each length.
@@ -520,7 +520,6 @@ inline char* IncrementalCopy(const char* src, char* op, char* const op_limit,
           LoadPatternAndReshuffleMask(src, pattern_size);
       V128 pattern = pattern_and_reshuffle_mask.first;
       V128 reshuffle_mask = pattern_and_reshuffle_mask.second;
-
       // This code path is relatively cold however so we save code size
       // by avoiding unrolling and vectorizing.
       //
@@ -959,6 +958,8 @@ emit_remainder:
 char* CompressFragmentDoubleHash(const char* input, size_t input_size, char* op,
                                  uint16_t* table, const int table_size,
                                  uint16_t* table2, const int table_size2) {
+  (void)table_size2;
+  assert(table_size == table_size2);
   // "ip" is the input pointer, and "op" is the output pointer.
   const char* ip = input;
   assert(input_size <= kBlockSize);
@@ -1224,7 +1225,7 @@ inline bool Copy64BytesWithPatternExtension(ptrdiff_t dst, size_t offset) {
 void MemCopy64(char* dst, const void* src, size_t size) {
   // Always copy this many bytes.  If that's below size then copy the full 64.
   constexpr int kShortMemCopy = 32;
-
+  (void)kShortMemCopy;
   assert(size <= 64);
   assert(std::less_equal<const void*>()(static_cast<const char*>(src) + size,
                                         dst) ||
@@ -1243,6 +1244,27 @@ void MemCopy64(char* dst, const void* src, size_t size) {
     data = _mm256_lddqu_si256(static_cast<const __m256i *>(src) + 1);
     _mm256_storeu_si256(reinterpret_cast<__m256i *>(dst) + 1, data);
   }
+  // RVV acceleration available on RISC-V when compiled with -march=rv64gcv
+#elif defined(__riscv) && SNAPPY_HAVE_RVV
+  // Cast pointers to the type we will operate on.
+  unsigned char* dst_ptr = reinterpret_cast<unsigned char*>(dst);
+  const unsigned char* src_ptr = reinterpret_cast<const unsigned char*>(src);
+  size_t remaining_bytes = size;
+  // Loop as long as there are bytes remaining to be copied.
+  while (remaining_bytes > 0) {
+    // Set vector configuration: e8 (8-bit elements), m2 (LMUL=2).
+    // Use e8m2 configuration to maximize throughput.
+    size_t vl = VSETVL_E8M2(remaining_bytes);
+    // Load data from the current source pointer.
+    vuint8m2_t vec = VLE8_V_U8M2(src_ptr, vl);
+    // Store data to the current destination pointer.
+    VSE8_V_U8M2(dst_ptr, vec, vl);
+    // Update pointers and the remaining count.
+    src_ptr += vl;
+    dst_ptr += vl;
+    remaining_bytes -= vl;
+  }
+
 #else
   std::memmove(dst, src, kShortMemCopy);
   // Profiling shows that nearly all copies are short.
@@ -1687,7 +1709,8 @@ constexpr uint32_t CalculateNeeded(uint8_t tag) {
 #if __cplusplus >= 201402L
 constexpr bool VerifyCalculateNeeded() {
   for (int i = 0; i < 1; i++) {
-    if (CalculateNeeded(i) != (char_table[i] >> 11) + 1) return false;
+    if (CalculateNeeded(i) != static_cast<uint32_t>((char_table[i] >> 11)) + 1)
+      return false;
   }
   return true;
 }
@@ -1775,6 +1798,7 @@ static bool InternalUncompressAllTags(SnappyDecompressor* decompressor,
                                       Writer* writer, uint32_t compressed_len,
                                       uint32_t uncompressed_len) {
     int token = 0;
+
   writer->SetExpectedLength(uncompressed_len);
 
   // Process the entire input
@@ -1885,16 +1909,16 @@ class SnappyIOVecReader : public Source {
     if (total_size > 0 && curr_size_remaining_ == 0) Advance();
   }
 
-  ~SnappyIOVecReader() = default;
+  ~SnappyIOVecReader() override = default;
 
-  size_t Available() const { return total_size_remaining_; }
+  size_t Available() const override { return total_size_remaining_; }
 
-  const char* Peek(size_t* len) {
+  const char* Peek(size_t* len) override {
     *len = curr_size_remaining_;
     return curr_pos_;
   }
 
-  void Skip(size_t n) {
+  void Skip(size_t n) override {
     while (n >= curr_size_remaining_ && n > 0) {
       n -= curr_size_remaining_;
       Advance();
@@ -2532,7 +2556,6 @@ bool SnappyScatteredWriter<Allocator>::SlowAppendFromSelf(size_t offset,
 class SnappySinkAllocator {
  public:
   explicit SnappySinkAllocator(Sink* dest) : dest_(dest) {}
-  ~SnappySinkAllocator() {}
 
   char* Allocate(int size) {
     Datablock block(new char[size], size);