Speed up ASCII, LATIN1, UTF8 converters, for single-byte codeunits.

sdk/lib/convert/utf.dart

Index: sdk/lib/convert/utf.dart
diff --git a/sdk/lib/convert/utf.dart b/sdk/lib/convert/utf.dart
index 819013913540a3e403d1cfee98bb9c78ec03419c..34b94dae5fcdbd17c41087af62da079152a040a7 100644
--- a/sdk/lib/convert/utf.dart
+++ b/sdk/lib/convert/utf.dart
@@ -420,10 +420,24 @@ class _Utf8Decoder {
     int value = _value;
     int expectedUnits = _expectedUnits;
     int extraUnits = _extraUnits;
+    int singleBytesCount = 0;
     _value = 0;
     _expectedUnits = 0;
     _extraUnits = 0;
 
+    void addSingleBytes(int from, int to) {
+      assert(singleBytesCount > 0);
+      assert(from >= startIndex && from <= endIndex);
+      assert(to >= startIndex && to <= endIndex);
+      if (from == 0 && to == codeUnits.length) {
+        _stringSink.write(new String.fromCharCodes(codeUnits));
+      } else {
+        _stringSink.write(
+            new String.fromCharCodes(codeUnits.sublist(from, to)));
+      }
+      singleBytesCount = 0;
+    }
+
     int i = startIndex;
     loop: while (true) {
       multibyte: if (expectedUnits > 0) {
@@ -479,6 +493,9 @@ class _Utf8Decoder {
         // https://codereview.chromium.org/22929022/diff/1/sdk/lib/convert/utf.dart?column_width=80
         if (unit < 0) {
           // TODO(floitsch): should this be unit <= 0 ?
+          if (singleBytesCount > 0) {
+            addSingleBytes(i - singleBytesCount - 1, i - 1);
+          }
           if (!_allowMalformed) {
             throw new FormatException(
                 "Negative UTF-8 code unit: -0x${(-unit).toRadixString(16)}");
@@ -486,8 +503,11 @@ class _Utf8Decoder {
           _stringSink.writeCharCode(UNICODE_REPLACEMENT_CHARACTER_RUNE);
         } else if (unit <= _ONE_BYTE_LIMIT) {
           _isFirstCharacter = false;
-          _stringSink.writeCharCode(unit);
+          singleBytesCount++;
         } else {
+          if (singleBytesCount > 0) {
+            addSingleBytes(i - singleBytesCount - 1, i - 1);

[Lasse Reichstein Nielsen, 2014/03/24 09:21:01]

i - 1 - singleBytesCount

Allows reuse of the (i - 1) computation.
Ditto above.

[Anders Johnsen, 2014/03/24 12:01:01]

Done.

+          }
           if ((unit & 0xE0) == 0xC0) {
             value = unit & 0x1F;
             expectedUnits = extraUnits = 1;
@@ -516,6 +536,9 @@ class _Utf8Decoder {
       }
       break loop;
     }
+    if (singleBytesCount > 0) {
+      addSingleBytes(i - singleBytesCount, endIndex);
+    }
     if (expectedUnits > 0) {
       _value = value;
       _expectedUnits = expectedUnits;