RegExp bitmap test for word character.

RegExp bitmap test for word character.

[Lasse Reichstein, 2010-01-12 16:58:01]

Smallish review.

[Erik Corry, 2010-01-13 11:52:53]

I think this would be faster and simpler with a byte map instead of a bit map.

[Lasse Reichstein, 2010-01-15 10:18:48]

Now uses a byte map. Please re-review.

[Erik Corry, 2010-01-15 12:02:09]

LGTM

http://codereview.chromium.org/547024/diff/2002/2003
File src/arm/regexp-macro-assembler-arm.cc (right):

http://codereview.chromium.org/547024/diff/2002/2003#newcode551
src/arm/regexp-macro-assembler-arm.cc:551: // Table is 128 bits, so all ASCII
characters can be tested.
128 bytes.

http://codereview.chromium.org/547024/diff/2002/2007
File src/ia32/assembler-ia32.h (right):

http://codereview.chromium.org/547024/diff/2002/2007#newcode231
src/ia32/assembler-ia32.h:231: times_8 = 3,
This is unused

http://codereview.chromium.org/547024/diff/2002/2009
File src/ia32/regexp-macro-assembler-ia32.cc (right):

http://codereview.chromium.org/547024/diff/2002/2009#newcode543
src/ia32/regexp-macro-assembler-ia32.cc:543: // Table is 128 bits, so all ASCII
characters can be tested.
bytes.

http://codereview.chromium.org/547024/diff/2002/2009#newcode547
src/ia32/regexp-macro-assembler-ia32.cc:547: ASSERT_EQ(0,
word_character_map[0]);  // Character '\0' is not a word char.
Perhaps also assert that word_character_map['a'] is 255.

http://codereview.chromium.org/547024/diff/2002/2009#newcode557
src/ia32/regexp-macro-assembler-ia32.cc:557: // Table is 128 bits, so all ASCII
characters can be tested.
bytes

http://codereview.chromium.org/547024/diff/2002/2012
File src/serialize.cc (right):

http://codereview.chromium.org/547024/diff/2002/2012#newcode489
src/serialize.cc:489: "NativeRegExpMacroAssembler::word_character_bitmap");
byte!

http://codereview.chromium.org/547024/diff/2002/2015
File src/x64/regexp-macro-assembler-x64.cc (right):

http://codereview.chromium.org/547024/diff/2002/2015#newcode586
src/x64/regexp-macro-assembler-x64.cc:586: // Table is 128 bits, so all ASCII
characters can be tested.
bytes!

http://codereview.chromium.org/547024/diff/2002/2015#newcode601
src/x64/regexp-macro-assembler-x64.cc:601: // Table is 128 bits, so all ASCII
characters can be tested.
bytes!

[Lasse Reichstein, 2010-01-18 09:56:52]

http://codereview.chromium.org/547024/diff/2002/2009
File src/ia32/regexp-macro-assembler-ia32.cc (right):

http://codereview.chromium.org/547024/diff/2002/2009#newcode547
src/ia32/regexp-macro-assembler-ia32.cc:547: ASSERT_EQ(0,
word_character_map[0]);  // Character '\0' is not a word char.
I'll put it in the comment at the table declaration that entries must be 0 or
255.
The important test here is that '\0' isn't a word character, so it doesn't
matter that:
  is_word_char(character) ::= (character & table[character]) != 0
gives false independently of the table.