equalIgnoringCase(StringImpl*, StringImpl*) is implemented by calling CaseFoldingHash::equal(a, b) which always up converts 8 bit strings.
Created attachment 162606 [details] Patch
SunSpider and V8 Performance results of posted patch. Seems neutral. Benchmark report for SunSpider and V8Real on msaboff-pro (MacPro5,1). VMs tested: "BaseLine" at /Volumes/Data/src/webkit.baseline/WebKitBuild/Release/DumpRenderTree (r127525) "96028" at /Volumes/Data/src/webkit/WebKitBuild/Release/DumpRenderTree (r127762) Collected 12 samples per benchmark/VM, with 4 VM invocations per benchmark. Emitted a call to gc() between sample measurements. Used 1 benchmark iteration per VM invocation for warm-up. Used the jsc-specific preciseTime() function to get microsecond-level timing. Reporting benchmark execution times with 95% confidence intervals in milliseconds. BaseLine 96028 SunSpider: 3d-cube 8.6759+-0.4965 ? 8.6795+-0.4975 ? 3d-morph 8.1207+-0.0540 ? 8.2320+-0.1022 ? might be 1.0137x slower 3d-raytrace 12.4447+-0.3669 ? 12.7408+-0.2827 ? might be 1.0238x slower access-binary-trees 2.3152+-0.4143 ? 2.3615+-0.4420 ? might be 1.0200x slower access-fannkuch 6.9096+-0.0546 6.8998+-0.0728 access-nbody 4.1547+-0.0387 ? 4.3159+-0.1864 ? might be 1.0388x slower access-nsieve 3.4240+-0.0603 3.4100+-0.0361 bitops-3bit-bits-in-byte 1.3655+-0.0132 ? 1.3787+-0.0229 ? bitops-bits-in-byte 5.7560+-0.0294 ? 5.7835+-0.0546 ? bitops-bitwise-and 2.3053+-0.0559 2.1998+-0.1298 might be 1.0480x faster bitops-nsieve-bits 3.5439+-0.0450 3.5358+-0.0341 controlflow-recursive 2.5098+-0.0199 2.4791+-0.0399 might be 1.0124x faster crypto-aes 9.7697+-0.5576 9.7663+-0.5242 crypto-md5 3.9863+-0.1273 3.9119+-0.1399 might be 1.0190x faster crypto-sha1 3.2103+-0.0622 3.1315+-0.0479 might be 1.0252x faster date-format-tofte 14.9243+-1.1552 14.1839+-0.6570 might be 1.0522x faster date-format-xparb 12.1777+-0.6900 11.5475+-0.2127 might be 1.0546x faster math-cordic 4.2617+-0.0855 ? 4.2981+-0.0387 ? math-partial-sums 10.3130+-0.2316 10.1840+-0.0735 might be 1.0127x faster math-spectral-norm 3.0180+-0.0339 3.0162+-0.0184 regexp-dna 10.4587+-0.3506 ? 10.4712+-0.3430 ? string-base64 5.5117+-0.0758 ? 5.8878+-0.5140 ? might be 1.0682x slower string-fasta 8.1848+-0.2835 ? 8.3168+-0.2465 ? might be 1.0161x slower string-tagcloud 13.7753+-0.2121 13.6639+-0.2892 string-unpack-code 23.5257+-0.5769 23.2775+-0.4931 might be 1.0107x faster string-validate-input 8.8146+-0.5770 8.7567+-0.5374 <arithmetic> * 7.4407+-0.1668 7.4012+-0.1276 might be 1.0053x faster <geometric> 5.9266+-0.1177 5.9129+-0.1067 might be 1.0023x faster <harmonic> 4.6557+-0.0864 4.6442+-0.0923 might be 1.0025x faster BaseLine 96028 V8Real: encrypt 0.40173+-0.00063 ? 0.40197+-0.00114 ? decrypt 6.93700+-0.01300 ? 6.94133+-0.01487 ? deltablue x2 0.56658+-0.00450 ? 0.56968+-0.00429 ? earley 1.23068+-0.00589 ^ 1.21994+-0.00352 ^ definitely 1.0088x faster boyer 13.66712+-0.05575 ? 13.67038+-0.05082 ? raytrace x2 4.55525+-0.03219 ? 4.55528+-0.02875 ? regexp x2 26.79253+-0.07617 26.62287+-0.09437 richards x2 0.26984+-0.00132 0.26955+-0.00128 splay x2 0.74498+-0.00724 0.74428+-0.00661 navier-stokes x2 12.73651+-0.01486 ? 12.73830+-0.00966 ? <arithmetic> 7.09799+-0.01237 7.07710+-0.01179 might be 1.0030x faster <geometric> * 2.42901+-0.00434 2.42707+-0.00476 might be 1.0008x faster <harmonic> 0.89800+-0.00301 ? 0.89814+-0.00259 ? might be 1.0002x slower BaseLine 96028 All benchmarks: <arithmetic> 7.3101+-0.1044 7.2777+-0.0809 might be 1.0045x faster <geometric> 4.2188+-0.0531 4.2116+-0.0481 might be 1.0017x faster <harmonic> 1.7945+-0.0108 1.7936+-0.0104 might be 1.0005x faster BaseLine 96028 Geomean of preferred means: <scaled-result> 13.4420+-0.1556 13.4016+-0.1194 might be 1.0030x faster
Comment on attachment 162606 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=162606&action=review > Source/WTF/wtf/text/StringHash.h:66 > - if (b->is8Bit()) { > - // We know that a is 8 bit and b is 16 bit. > + if (b->is8Bit()) > return WTF::equal(a->characters16(), b->characters8(), aLength); Nice comment indeed :) > Source/WTF/wtf/text/StringHash.h:135 > return WTF::Unicode::umemcasecmp(a->characters(), b->characters(), length) == 0; For clarity, I would also use equalIgnoringCase here. And characters() could be characters16().
(In reply to comment #3) > (From update of attachment 162606 [details]) > View in context: https://bugs.webkit.org/attachment.cgi?id=162606&action=review > > > Source/WTF/wtf/text/StringHash.h:66 > > - if (b->is8Bit()) { > > - // We know that a is 8 bit and b is 16 bit. > > + if (b->is8Bit()) > > return WTF::equal(a->characters16(), b->characters8(), aLength); > > Nice comment indeed :) > > > Source/WTF/wtf/text/StringHash.h:135 > > return WTF::Unicode::umemcasecmp(a->characters(), b->characters(), length) == 0; > > For clarity, I would also use equalIgnoringCase here. > > And characters() could be characters16(). WTF::equalIgnoringCase(UChar*, UChar*) is currently a static inline StringImpl.cpp that calls umemcasecmp(). I moved it to StringImpl.h and changed StringHash.h to call it (with characters16()). Want a new patch or you fine with this change?
> WTF::equalIgnoringCase(UChar*, UChar*) is currently a static inline StringImpl.cpp that calls umemcasecmp(). I moved it to StringImpl.h and changed StringHash.h to call it (with characters16()). Want a new patch or you fine with this change? I did not know equalIgnoringCase() was only in the cpp file. Your change sounds good, no need for another review.
Committed r127887: <http://trac.webkit.org/changeset/127887>