brotli 0.1.1 → 0.1.2

data/vendor/brotli/enc/backward_references.h

@@ -23,17 +23,93 @@ namespace brotli {
 // by this call.
 void CreateBackwardReferences(size_t num_bytes,
                               size_t position,
+                              bool is_last,
                               const uint8_t* ringbuffer,
                               size_t ringbuffer_mask,
-                              const size_t max_backward_limit,
                               const int quality,
+                              const int lgwin,
                               Hashers* hashers,
                               int hash_type,
                               int* dist_cache,
-                              int* last_insert_len,
+                              size_t* last_insert_len,
                               Command* commands,
                               size_t* num_commands,
-                              int* num_literals);
+                              size_t* num_literals);
+
+static const float kInfinity = std::numeric_limits<float>::infinity();
+
+struct ZopfliNode {
+  ZopfliNode(void) : length(1),
+                     distance(0),
+                     insert_length(0),
+                     cost(kInfinity) {}
+
+  inline uint32_t copy_length() const {
+    return length & 0xffffff;
+  }
+
+  inline uint32_t length_code() const {
+    const uint32_t modifier = length >> 24;
+    return copy_length() + 9u - modifier;
+  }
+
+  inline uint32_t copy_distance() const {
+    return distance & 0x1ffffff;
+  }
+
+  inline uint32_t distance_code() const {
+    const uint32_t short_code = distance >> 25;
+    return short_code == 0 ? copy_distance() + 15 : short_code - 1;
+  }
+
+  inline uint32_t command_length() const {
+    return copy_length() + insert_length;
+  }
+
+  // best length to get up to this byte (not including this byte itself)
+  // highest 8 bit is used to reconstruct the length code
+  uint32_t length;
+  // distance associated with the length
+  // highest 7 bit contains distance short code + 1 (or zero if no short code)
+  uint32_t distance;
+  // number of literal inserts before this copy
+  uint32_t insert_length;
+  // smallest cost to get to this byte from the beginning, as found so far
+  float cost;
+};
+
+// Computes the shortest path of commands from position to at most
+// position + num_bytes.
+//
+// On return, path->size() is the number of commands found and path[i] is the
+// length of the ith command (copy length plus insert length).
+// Note that the sum of the lengths of all commands can be less than num_bytes.
+//
+// On return, the nodes[0..num_bytes] array will have the following
+// "ZopfliNode array invariant":
+// For each i in [1..num_bytes], if nodes[i].cost < kInfinity, then
+//   (1) nodes[i].copy_length() >= 2
+//   (2) nodes[i].command_length() <= i and
+//   (3) nodes[i - nodes[i].command_length()].cost < kInfinity
+void ZopfliComputeShortestPath(size_t num_bytes,
+                               size_t position,
+                               const uint8_t* ringbuffer,
+                               size_t ringbuffer_mask,
+                               const size_t max_backward_limit,
+                               const int* dist_cache,
+                               Hashers::H10* hasher,
+                               ZopfliNode* nodes,
+                               std::vector<uint32_t>* path);
+
+void ZopfliCreateCommands(const size_t num_bytes,
+                          const size_t block_start,
+                          const size_t max_backward_limit,
+                          const std::vector<uint32_t>& path,
+                          const ZopfliNode* nodes,
+                          int* dist_cache,
+                          size_t* last_insert_len,
+                          Command* commands,
+                          size_t* num_literals);
 
 }  // namespace brotli
 

data/vendor/brotli/enc/bit_cost.h

@@ -9,89 +9,111 @@
 #ifndef BROTLI_ENC_BIT_COST_H_
 #define BROTLI_ENC_BIT_COST_H_
 
-
-
 #include "./entropy_encode.h"
 #include "./fast_log.h"
 #include "./types.h"
 
 namespace brotli {
 
-static inline double ShannonEntropy(const int *population, int size,
-                                    int *total) {
-  int sum = 0;
+static inline double ShannonEntropy(const uint32_t *population, size_t size,
+                                    size_t *total) {
+  size_t sum = 0;
   double retval = 0;
-  const int *population_end = population + size;
-  int p;
+  const uint32_t *population_end = population + size;
+  size_t p;
   if (size & 1) {
     goto odd_number_of_elements_left;
   }
   while (population < population_end) {
     p = *population++;
     sum += p;
-    retval -= p * FastLog2(p);
+    retval -= static_cast<double>(p) * FastLog2(p);
  odd_number_of_elements_left:
     p = *population++;
     sum += p;
-    retval -= p * FastLog2(p);
+    retval -= static_cast<double>(p) * FastLog2(p);
   }
-  if (sum) retval += sum * FastLog2(sum);
+  if (sum) retval += static_cast<double>(sum) * FastLog2(sum);
   *total = sum;
   return retval;
 }
 
-static inline double BitsEntropy(const int *population, int size) {
-  int sum;
+static inline double BitsEntropy(const uint32_t *population, size_t size) {
+  size_t sum;
   double retval = ShannonEntropy(population, size, &sum);
   if (retval < sum) {
     // At least one bit per literal is needed.
-    retval = sum;
+    retval = static_cast<double>(sum);
   }
   return retval;
 }
 
-
 template<int kSize>
 double PopulationCost(const Histogram<kSize>& histogram) {
+  static const double kOneSymbolHistogramCost = 12;
+  static const double kTwoSymbolHistogramCost = 20;
+  static const double kThreeSymbolHistogramCost = 28;
+  static const double kFourSymbolHistogramCost = 37;
   if (histogram.total_count_ == 0) {
-    return 12;
+    return kOneSymbolHistogramCost;
   }
   int count = 0;
+  int s[5];
   for (int i = 0; i < kSize; ++i) {
     if (histogram.data_[i] > 0) {
+      s[count] = i;
       ++count;
+      if (count > 4) break;
     }
   }
   if (count == 1) {
-    return 12;
+    return kOneSymbolHistogramCost;
   }
   if (count == 2) {
-    return 20 + histogram.total_count_;
+    return (kTwoSymbolHistogramCost +
+            static_cast<double>(histogram.total_count_));
   }
-  double bits = 0;
-  uint8_t depth_array[kSize] = { 0 };
-  if (count <= 4) {
-    // For very low symbol count we build the Huffman tree.
-    CreateHuffmanTree(&histogram.data_[0], kSize, 15, depth_array);
-    for (int i = 0; i < kSize; ++i) {
-      bits += histogram.data_[i] * depth_array[i];
+  if (count == 3) {
+    const uint32_t histo0 = histogram.data_[s[0]];
+    const uint32_t histo1 = histogram.data_[s[1]];
+    const uint32_t histo2 = histogram.data_[s[2]];
+    const uint32_t histomax = std::max(histo0, std::max(histo1, histo2));
+    return (kThreeSymbolHistogramCost +
+            2 * (histo0 + histo1 + histo2) - histomax);
+  }
+  if (count == 4) {
+    uint32_t histo[4];
+    for (int i = 0; i < 4; ++i) {
+      histo[i] = histogram.data_[s[i]];
     }
-    return count == 3 ? bits + 28 : bits + 37;
+    // Sort
+    for (int i = 0; i < 4; ++i) {
+      for (int j = i + 1; j < 4; ++j) {
+        if (histo[j] > histo[i]) {
+          std::swap(histo[j], histo[i]);
+        }
+      }
+    }
+    const uint32_t h23 = histo[2] + histo[3];
+    const uint32_t histomax = std::max(h23, histo[0]);
+    return (kFourSymbolHistogramCost +
+            3 * h23 + 2 * (histo[0] + histo[1]) - histomax);
   }
 
   // In this loop we compute the entropy of the histogram and simultaneously
   // build a simplified histogram of the code length codes where we use the
   // zero repeat code 17, but we don't use the non-zero repeat code 16.
-  int max_depth = 1;
-  int depth_histo[kCodeLengthCodes] = { 0 };
+  double bits = 0;
+  size_t max_depth = 1;
+  uint32_t depth_histo[kCodeLengthCodes] = { 0 };
   const double log2total = FastLog2(histogram.total_count_);
-  for (int i = 0; i < kSize;) {
+  for (size_t i = 0; i < kSize;) {
     if (histogram.data_[i] > 0) {
       // Compute -log2(P(symbol)) = -log2(count(symbol)/total_count) =
       //                          =  log2(total_count) - log2(count(symbol))
       double log2p = log2total - FastLog2(histogram.data_[i]);
       // Approximate the bit depth by round(-log2(P(symbol)))
-      int depth = static_cast<int>(log2p + 0.5);
+      size_t depth = static_cast<size_t>(log2p + 0.5);
       bits += histogram.data_[i] * log2p;
       if (depth > 15) {
         depth = 15;
@@ -104,8 +126,8 @@ double PopulationCost(const Histogram<kSize>& histogram) {
     } else {
       // Compute the run length of zeros and add the appropriate number of 0 and
       // 17 code length codes to the code length code histogram.
-      int reps = 1;
-      for (int k = i + 1; k < kSize && histogram.data_[k] == 0; ++k) {
+      uint32_t reps = 1;
+      for (size_t k = i + 1; k < kSize && histogram.data_[k] == 0; ++k) {
         ++reps;
       }
       i += reps;
@@ -128,7 +150,7 @@ double PopulationCost(const Histogram<kSize>& histogram) {
     }
   }
   // Add the estimated encoding cost of the code length code histogram.
-  bits += 18 + 2 * max_depth;
+  bits += static_cast<double>(18 + 2 * max_depth);
   // Add the entropy of the code length code histogram.
   bits += BitsEntropy(depth_histo, kCodeLengthCodes);
   return bits;