Changeset 77297 in webkit

Timestamp:

Feb 1, 2011, 2:03:04 PM (15 years ago)

Author:

Patrick Gansterer

Message:

2011-02-01 Patrick Gansterer <Patrick Gansterer>

Reviewed by Darin Adler.

Avoid strlen() in AtomicString::fromUTF8
​https://bugs.webkit.org/show_bug.cgi?id=50516

Add an overload to calculateStringHashFromUTF8 to get
strlen() of the input data with only one call.

This change shows about 3% performance win on the xml-parser benchmark.

• JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.def:
• wtf/text/AtomicString.cpp: (WTF::AtomicString::fromUTF8):
• wtf/unicode/UTF8.cpp: (WTF::Unicode::calculateStringHashAndLengthFromUTF8Internal): (WTF::Unicode::calculateStringHashFromUTF8): (WTF::Unicode::calculateStringHashAndLengthFromUTF8):
• wtf/unicode/UTF8.h:

Location:

trunk/Source/JavaScriptCore

Files:

• ChangeLog (modified) (1 diff)
• JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.def (modified) (1 diff)
• wtf/text/AtomicString.cpp (modified) (1 diff)
• wtf/unicode/UTF8.cpp (modified) (4 diffs)
• wtf/unicode/UTF8.h (modified) (1 diff)

trunk/Source/JavaScriptCore/ChangeLog

@@ +1 @@
+2011-02-01  Patrick Gansterer  <[email protected]>
+
+        Reviewed by Darin Adler.
+
+        Avoid strlen() in AtomicString::fromUTF8
+        https://bugs.webkit.org/show_bug.cgi?id=50516
+
+        Add an overload to calculateStringHashFromUTF8 to get
+        strlen() of the input data with only one call.
+
+        This change shows about 3% performance win on the xml-parser benchmark.
+
+        * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.def:
+        * wtf/text/AtomicString.cpp:
+        (WTF::AtomicString::fromUTF8):
+        * wtf/unicode/UTF8.cpp:
+        (WTF::Unicode::calculateStringHashAndLengthFromUTF8Internal):
+        (WTF::Unicode::calculateStringHashFromUTF8):
+        (WTF::Unicode::calculateStringHashAndLengthFromUTF8):
+        * wtf/unicode/UTF8.h:
+
 2011-02-01  Sam Weinig  <[email protected]>
 

trunk/Source/JavaScriptCore/JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.def

@@ -65 +65 @@
     ?calculateDSTOffset@WTF@@YANNN@Z
     ?calculateStringHashFromUTF8@Unicode@WTF@@YAIPBD0AAI@Z
+    ?calculateStringHashAndLengthFromUTF8@Unicode@WTF@@YAIPBDAAI1@Z
     ?calculateUTCOffset@WTF@@YAHXZ
     ?calculatedFunctionName@DebuggerCallFrame@JSC@@QBE?AVUString@2@XZ

trunk/Source/JavaScriptCore/wtf/text/AtomicString.cpp

@@ -377 +377 @@
     if (!characters)
         return AtomicString();
-    return fromUTF8(characters, strlen(characters));
+
+    HashAndUTF8Characters buffer;
+    buffer.characters = characters;
+    buffer.hash = calculateStringHashAndLengthFromUTF8(characters, buffer.length, buffer.utf16Length);
+
+    if (!buffer.hash)
+        return AtomicString();
+
+    AtomicString atomicString;
+    atomicString.m_string = addToStringTable<HashAndUTF8Characters, HashAndUTF8CharactersTranslator>(buffer);
+    return atomicString;
 }
 

trunk/Source/JavaScriptCore/wtf/unicode/UTF8.cpp

@@ -315 +315 @@
 }
 
-unsigned calculateStringHashFromUTF8(const char* data, const char* dataEnd, unsigned& utf16Length)
+static inline unsigned calculateStringHashAndLengthFromUTF8Internal(const char* data, const char* dataEnd, unsigned& dataLength, unsigned& utf16Length)
 {
     if (!data)
@@ -321 +321 @@
 
     WTF::StringHasher stringHasher;
+    dataLength = 0;
     utf16Length = 0;
 
-    while (data < dataEnd) {
+    while (data < dataEnd || (!dataEnd && *data)) {
         if (isASCII(*data)) {
             stringHasher.addCharacter(*data++);
+            dataLength++;
             utf16Length++;
             continue;
@@ -331 +333 @@
 
         int utf8SequenceLength = inlineUTF8SequenceLengthNonASCII(*data);
-
-        if (dataEnd - data < utf8SequenceLength)
-            return false;
+        dataLength += utf8SequenceLength;
+
+        if (!dataEnd) {
+            for (int i = 1; i < utf8SequenceLength; ++i) {
+                if (!data[i])
+                    return 0;
+            }
+        } else if (dataEnd - data < utf8SequenceLength)
+            return 0;
 
         if (!isLegalUTF8(reinterpret_cast<const unsigned char*>(data), utf8SequenceLength))
@@ -356 +364 @@
 
     return stringHasher.hash();
+}
+
+unsigned calculateStringHashFromUTF8(const char* data, const char* dataEnd, unsigned& utf16Length)
+{
+    unsigned dataLength;
+    return calculateStringHashAndLengthFromUTF8Internal(data, dataEnd, dataLength, utf16Length);
+}
+
+unsigned calculateStringHashAndLengthFromUTF8(const char* data, unsigned& dataLength, unsigned& utf16Length)
+{
+    return calculateStringHashAndLengthFromUTF8Internal(data, 0, dataLength, utf16Length);
 }
 

trunk/Source/JavaScriptCore/wtf/unicode/UTF8.h

@@ -72 +72 @@
 
     unsigned calculateStringHashFromUTF8(const char* data, const char* dataEnd, unsigned& utf16Length);
+    unsigned calculateStringHashAndLengthFromUTF8(const char* data, unsigned& dataLength, unsigned& utf16Length);
 
     bool equalUTF16WithUTF8(const UChar* a, const UChar* aEnd, const char* b, const char* bEnd);