Branch: refs/heads/main
Home: https://github.com/WebKit/WebKit
Commit: aaee2a6f166a1cda4ee3876c665bceb8a7988d3f
https://github.com/WebKit/WebKit/commit/aaee2a6f166a1cda4ee3876c665bceb8a7988d3f
Author: Michael Saboff <[email protected]>
Date: 2025-04-23 (Wed, 23 Apr 2025)
Changed paths:
A JSTests/microbenchmarks/regexp-non-bmp-matching.js
M Source/JavaScriptCore/assembler/MacroAssemblerARM64.h
M Source/JavaScriptCore/jit/JITThunks.cpp
M Source/JavaScriptCore/jit/JITThunks.h
M Source/JavaScriptCore/yarr/YarrJIT.cpp
M Source/JavaScriptCore/yarr/YarrJIT.h
M Source/JavaScriptCore/yarr/YarrJITRegisters.h
Log Message:
-----------
[Yarr] Improve reading of Surrogate Pairs in Unicode Regular Expressions
https://bugs.webkit.org/show_bug.cgi?id=291923
rdar://149811541
Reviewed by Daniel Liu.
Restructured the read possible surrogate pairs helper
Yarr::tryReadUnicodeCharImpl() to try to read two
characters at once, then check to see if those two characters are well formed
surrogate pair and then
return the resulting non-BMP codepoint. If they are a surrogate pair, we
return either the first single
character or an error if we happened to read in a trailing surrogate that was
preceded by a leading
surrogate.
Turned the helper into an on demand thunk to reduce the amount of JIT'ed code
when we have more than one
unicode enabled RegExp's. Also made the areCanonicallyEquivalentThunk an on
demand thunk.
Added the set flags variant of the ARM64 AND instruction as it appears that
setting the flags as part
of the AND operation may be slightly more performant than doing a separate
compare to zero.
Added a new micro benchmark, regexp-non-bmp-matching.js, to verify the
performance improvement.
This change improves that benchmark by about 2%.
Baseline TuneReadUnicode
regexp-non-bmp-matching 38.6839+-0.2168 ^ 37.9023+-0.1007 ^
definitely 1.0206x faster
<geometric> 38.6839+-0.2168 ^ 37.9023+-0.1007 ^
definitely 1.0206x faster
* JSTests/microbenchmarks/regexp-non-bmp-matching.js: Added.
(arrayToString):
(objectToString):
(dumpValue):
(compareArray):
(compareGroups):
(testRegExp):
(testRegExpSyntaxError):
* Source/JavaScriptCore/assembler/MacroAssemblerARM64.h:
(JSC::MacroAssemblerARM64::add32AndSetFlags):
(JSC::MacroAssemblerARM64::add64AndSetFlags):
(JSC::MacroAssemblerARM64::and32AndSetFlags):
* Source/JavaScriptCore/jit/JITThunks.cpp:
(JSC::JITThunks::initialize):
* Source/JavaScriptCore/jit/JITThunks.h:
* Source/JavaScriptCore/yarr/YarrJIT.cpp:
(JSC::Yarr::tryReadUnicodeCharImpl):
(JSC::Yarr::tryReadUnicodeCharThunkGenerator):
(JSC::Yarr::tryReadUnicodeCharIncForNonBMPThunkGenerator):
(JSC::Yarr::areCanonicallyEquivalentThunkGenerator):
* Source/JavaScriptCore/yarr/YarrJIT.h:
* Source/JavaScriptCore/yarr/YarrJITRegisters.h:
Canonical link: https://commits.webkit.org/294046@main
To unsubscribe from these emails, change your notification settings at
https://github.com/WebKit/WebKit/settings/notifications
_______________________________________________
webkit-changes mailing list
[email protected]
https://lists.webkit.org/mailman/listinfo/webkit-changes
