Speed up WriteUtf8 in the case where the output buffer is large enough.

src/api.cc

 // Copyright 2012 the V8 project authors. All rights reserved.
 // Redistribution and use in source and binary forms, with or without
 // modification, are permitted provided that the following conditions are
 // met:
 //
 //     * Redistributions of source code must retain the above copyright
 //       notice, this list of conditions and the following disclaimer.
 //     * Redistributions in binary form must reproduce the above
 //       copyright notice, this list of conditions and the following
 //       disclaimer in the documentation and/or other materials provided
@@ skipping 3676 matching lines @@
 }
 
 
 int String::Utf8Length() const {
   i::Handle<i::String> str = Utils::OpenHandle(this);
   if (IsDeadCheck(str->GetIsolate(), "v8::String::Utf8Length()")) return 0;
   return i::Utf8Length(str);
 }
 
 
+// Will fail with a negative answer if the recursion depth is too high.
+static int RecursivelySerializeToUtf8(i::String* string,
+                                      char* buffer,
+                                      int start,
+                                      int end,
+                                      int recursion_budget,
+                                      int32_t previous_character,
+                                      int32_t* last_character) {
+  int utf8_bytes = 0;
+  while (true) {
+    if (string->IsAsciiRepresentation()) {
+      i::String::WriteToFlat(string, buffer, start, end);
+      *last_character = unibrow::Utf16::kNoPreviousCharacter;
+      return utf8_bytes + end - start;
+    }
+    switch (i::StringShape(string).representation_tag()) {
+      case i::kExternalStringTag: {
+        const uint16_t* data = i::ExternalTwoByteString::cast(string)->
+          ExternalTwoByteStringGetData(start);
+        char* current = buffer;
+        for (int i = start; i < end; i++) {
+          uint16_t character = data[i];
+          current +=
+              unibrow::Utf8::Encode(current, character, previous_character);
+          previous_character = character;
+        }
+        *last_character = previous_character;
+        return utf8_bytes + current - buffer;
+      }
+      case i::kSeqStringTag: {
+        const uint16_t* data =
+            i::SeqTwoByteString::cast(string)->SeqTwoByteStringGetData(start);
+        char* current = buffer;
+        for (int i = start; i < end; i++) {
+          uint16_t character = data[i];
+          current +=
+              unibrow::Utf8::Encode(current, character, previous_character);
+          previous_character = character;
+        }
+        *last_character = previous_character;
+        return utf8_bytes + current - buffer;
+      }
+      case i::kSlicedStringTag: {
+        i::SlicedString* slice = i::SlicedString::cast(string);
+        unsigned offset = slice->offset();
+        string = slice->parent();
+        start += offset;
+        end += offset;
+      }

[fschneider, 2012/03/20 09:45:49]

Accidental fall-through here?

+      case i::kConsStringTag: {
+        i::ConsString* cons_string = i::ConsString::cast(string);
+        i::String* first = cons_string->first();
+        int boundary = first->length();
+        if (start >= boundary) {
+          // Only need RHS.
+          string = cons_string->second();
+          start -= boundary;
+          end -= boundary;
+          continue;
+        } else if (end <= boundary) {
+          // Only need LHS.
+          string = first;
+        } else {
+          if (recursion_budget == 0) return -1;
+          int extra_utf8_bytes =
+              RecursivelySerializeToUtf8(first,
+                                         buffer,
+                                         start,
+                                         boundary,
+                                         recursion_budget - 1,
+                                         previous_character,
+                                         &previous_character);
+          if (extra_utf8_bytes < 0) return extra_utf8_bytes;
+          buffer += extra_utf8_bytes;
+          utf8_bytes += extra_utf8_bytes;
+          string = cons_string->second();
+          start = 0;
+          end -= boundary;
+        }
+      }
+    }
+  }
+  UNREACHABLE();
+  return 0;
+}
+
+
 int String::WriteUtf8(char* buffer,
                       int capacity,
                       int* nchars_ref,
                       int options) const {
   i::Isolate* isolate = Utils::OpenHandle(this)->GetIsolate();
   if (IsDeadCheck(isolate, "v8::String::WriteUtf8()")) return 0;
   LOG_API(isolate, "String::WriteUtf8");
   ENTER_V8(isolate);
   i::Handle<i::String> str = Utils::OpenHandle(this);
+  int string_length = str->length();
   if (str->IsAsciiRepresentation()) {
     int len;
     if (capacity == -1) {
       capacity = str->length() + 1;
-      len = str->length();
+      len = string_length;
     } else {
       len = i::Min(capacity, str->length());
     }
     i::String::WriteToFlat(*str, buffer, 0, len);
     if (nchars_ref != NULL) *nchars_ref = len;
     if (!(options & NO_NULL_TERMINATION) && capacity > len) {
       buffer[len] = '\0';
       return len + 1;
     }
     return len;
   }
 
+  if (capacity == -1 || capacity >= string_length * 3) {
+    int32_t previous = unibrow::Utf16::kNoPreviousCharacter;
+    const int kMaxRecursion = 100;
+    int utf8_bytes =
+        RecursivelySerializeToUtf8(*str,
+                                   buffer,
+                                   0,
+                                   string_length,
+                                   kMaxRecursion,
+                                   previous,
+                                   &previous);
+    if (utf8_bytes >= 0) {
+      // Success serializing with recursion.
+      if ((options & NO_NULL_TERMINATION) == 0 &&
+          (capacity > utf8_bytes || capacity == -1)) {
+        buffer[utf8_bytes++] = '\0';
+      }
+      if (nchars_ref != NULL) *nchars_ref = string_length;
+      return utf8_bytes;
+    }
+    FlattenString(str);
+    // Recurse once.  This time around the string is flat and the serializing
+    // with recursion will certainly succeed.
+    return WriteUtf8(buffer, capacity, nchars_ref, options);
+  } else if (capacity >= string_length) {
+    // First check that the buffer is large enough.  If it is, then recurse
+    // once without a capacity limit, which will get into the other branch of
+    // this 'if'.
+    int utf8_bytes = i::Utf8Length(str);
+    if ((options & NO_NULL_TERMINATION) == 0) utf8_bytes++;
+    if (utf8_bytes <= capacity) {
+      return WriteUtf8(buffer, -1, nchars_ref, options);
+    }
+  }
+
+  // Slow case.
   i::StringInputBuffer& write_input_buffer = *isolate->write_input_buffer();
   isolate->string_tracker()->RecordWrite(str);
   if (options & HINT_MANY_WRITES_EXPECTED) {
     // Flatten the string for efficiency.  This applies whether we are
     // using StringInputBuffer or Get(i) to access the characters.
     FlattenString(str);
   }
   write_input_buffer.Reset(0, *str);
   int len = str->length();
   // Encode the first K - 3 bytes directly into the buffer since we
@@ skipping 2489 matching lines @@
 
 
 char* HandleScopeImplementer::Iterate(ObjectVisitor* v, char* storage) {
   HandleScopeImplementer* scope_implementer =
       reinterpret_cast<HandleScopeImplementer*>(storage);
   scope_implementer->IterateThis(v);
   return storage + ArchiveSpacePerThread();
 }
 
 } }  // namespace v8::internal