Changeset 38632 in webkit for trunk/JavaScriptCore/bytecompiler/SegmentedVector.h

Timestamp:

Nov 20, 2008, 2:27:34 PM (17 years ago)

Author:

[email protected]

Message:

2008-11-20 Sam Weinig <​[email protected]>

Reviewed by Geoffrey Garen.

Speedup the lexer to offset coming re-parsing patch.

• .6% progression on Sunspider.

• bytecompiler/SegmentedVector.h: (JSC::SegmentedVector::shrink): Fixed bug where m_size would not be set when shrinking to 0.

• parser/Lexer.cpp: (JSC::Lexer::Lexer): (JSC::Lexer::isIdentStart): Use isASCIIAlpha and isASCII to avoid going into ICU in the common cases. (JSC::Lexer::isIdentPart): Use isASCIIAlphanumeric and isASCII to avoid going into ICU in the common cases (JSC::isDecimalDigit): Use version in ASCIICType.h. Inlining it was a regression. (JSC::Lexer::isHexDigit): Ditto. (JSC::Lexer::isOctalDigit): Ditto. (JSC::Lexer::clear): Resize the m_identifiers SegmentedVector to initial capacity
• parser/Lexer.h: Remove unused m_strings vector. Make m_identifiers a SegmentedVector<Identifier> to avoid allocating a new Identifier* for each identifier found. The SegmentedVector is need so we can passes references to the Identifier to the parser, which remain valid even when the vector is resized. (JSC::Lexer::makeIdentifier): Inline and return a reference to the added Identifier.

File:

• trunk/JavaScriptCore/bytecompiler/SegmentedVector.h (modified) (1 diff)

trunk/JavaScriptCore/bytecompiler/SegmentedVector.h

@@ -108 +108 @@
                 m_segments.resize(1);
                 m_inlineSegment.resize(0);
+                m_size = size;
                 return;
             }