...
+1
Created attachment 261835 [details] work in progress
Created attachment 261843 [details] sort of passing some tests, I guess
Created attachment 261897 [details] maybe the patch I'm still running some tests. But it looks pretty good so far.
Attachment 261897 [details] did not pass style-queue: ERROR: Source/WTF/wtf/ParallelHelperPool.cpp:176: Place brace on its own line for function definitions. [whitespace/braces] [4] ERROR: Source/JavaScriptCore/runtime/CodeCache.h:38: Alphabetical sorting problem. [build/include_order] [4] ERROR: Source/JavaScriptCore/heap/HeapHelperPool.cpp:30: Bad include order. Mixing system and custom headers. [build/include_order] [4] ERROR: Source/JavaScriptCore/heap/Heap.cpp:548: Place brace on its own line for function definitions. [whitespace/braces] [4] ERROR: Source/JavaScriptCore/heap/Heap.cpp:633: Place brace on its own line for function definitions. [whitespace/braces] [4] ERROR: Source/JavaScriptCore/runtime/VM.h:68: Alphabetical sorting problem. [build/include_order] [4] Total errors found: 6 in 17 files If any of these errors are false positives, please file a bug against check-webkit-style.
Performance is OK: Benchmark report for SunSpider, LongSpider, V8Spider, Octane, Kraken, JSRegress, AsmBench, and CompressionBench on shakezilla (MacBookPro11,3). VMs tested: "TipOfTree" at /Volumes/Data/secondary/OpenSource/WebKitBuild/Release/jsc (r190217) "ParallelHelperPool" at /Volumes/Data/quartary/OpenSource/WebKitBuild/Release/jsc (r190217) Collected 7 samples per benchmark/VM, with 7 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. TipOfTree ParallelHelperPool SunSpider: 3d-cube 4.8238+-0.4948 4.4259+-0.1112 might be 1.0899x faster 3d-morph 5.3545+-0.2622 5.2012+-0.1000 might be 1.0295x faster 3d-raytrace 5.3653+-0.1587 5.2177+-0.2311 might be 1.0283x faster access-binary-trees 2.1610+-0.1281 ? 2.2318+-0.2849 ? might be 1.0327x slower access-fannkuch 5.4899+-0.1947 ? 5.5426+-0.1371 ? access-nbody 2.5440+-0.1140 ? 2.5831+-0.1806 ? might be 1.0154x slower access-nsieve 3.0718+-0.0730 ? 3.0813+-0.1555 ? bitops-3bit-bits-in-byte 1.1482+-0.0227 ? 1.1590+-0.0418 ? bitops-bits-in-byte 3.1865+-0.0434 ? 3.2571+-0.1015 ? might be 1.0222x slower bitops-bitwise-and 1.9811+-0.0172 ? 2.0200+-0.1145 ? might be 1.0196x slower bitops-nsieve-bits 2.9163+-0.0590 ? 2.9771+-0.1171 ? might be 1.0209x slower controlflow-recursive 2.3349+-0.0745 2.3206+-0.0430 crypto-aes 3.8618+-0.0679 ? 3.9394+-0.2317 ? might be 1.0201x slower crypto-md5 2.4306+-0.0634 ? 2.5313+-0.1136 ? might be 1.0414x slower crypto-sha1 2.4447+-0.1530 2.4429+-0.1040 date-format-tofte 6.6284+-0.1152 ? 6.7748+-0.2771 ? might be 1.0221x slower date-format-xparb 4.7655+-0.1693 4.5829+-0.1946 might be 1.0398x faster math-cordic 2.8865+-0.1388 2.8842+-0.1256 math-partial-sums 4.7640+-0.0731 ? 4.7845+-0.2007 ? math-spectral-norm 1.9986+-0.1796 1.9368+-0.0857 might be 1.0319x faster regexp-dna 6.4210+-0.1994 ^ 6.0502+-0.1595 ^ definitely 1.0613x faster string-base64 4.4975+-0.2186 4.4594+-0.1558 string-fasta 5.8044+-0.1032 ? 5.8735+-0.1774 ? might be 1.0119x slower string-tagcloud 7.8370+-0.0885 ? 7.8931+-0.0769 ? string-unpack-code 18.5619+-0.5746 ? 18.7104+-0.4670 ? string-validate-input 4.5297+-0.1527 ? 4.5637+-0.4200 ? <arithmetic> 4.5311+-0.0236 4.5171+-0.0506 might be 1.0031x faster TipOfTree ParallelHelperPool LongSpider: 3d-cube 800.6923+-9.6531 799.8176+-2.4010 3d-morph 1485.1298+-2.1368 1484.4645+-3.0170 3d-raytrace 589.2321+-4.3652 ? 590.7834+-5.1502 ? access-binary-trees 785.2797+-3.3423 ? 787.6647+-6.4111 ? access-fannkuch 273.3021+-3.7824 ? 274.9242+-4.6492 ? access-nbody 504.3896+-0.7826 503.6922+-0.6525 access-nsieve 365.2946+-11.0123 ? 366.2232+-9.7443 ? bitops-3bit-bits-in-byte 33.7894+-0.7732 33.6817+-0.4180 bitops-bits-in-byte 73.9784+-1.3815 73.8790+-0.6959 bitops-nsieve-bits 398.2521+-2.1836 ? 399.3892+-3.1114 ? controlflow-recursive 425.8974+-4.6171 423.2567+-0.6182 crypto-aes 558.9363+-12.6580 553.3800+-10.7937 might be 1.0100x faster crypto-md5 431.8727+-3.2255 ? 441.8658+-26.4278 ? might be 1.0231x slower crypto-sha1 632.6772+-5.6073 628.0345+-10.0830 date-format-tofte 500.2225+-6.0276 ? 504.2688+-6.1823 ? date-format-xparb 658.1238+-1.8384 ? 700.6277+-119.7695 ? might be 1.0646x slower hash-map 148.6700+-1.6807 ? 149.1348+-1.4700 ? math-cordic 474.6093+-1.2511 473.9519+-0.5247 math-partial-sums 457.5010+-0.7586 ^ 455.0258+-1.5119 ^ definitely 1.0054x faster math-spectral-norm 545.2324+-0.4119 ? 547.1794+-2.7131 ? string-base64 349.8429+-2.4775 ! 357.3467+-2.8458 ! definitely 1.0214x slower string-fasta 362.0520+-2.0066 ? 364.2637+-1.2252 ? string-tagcloud 174.2948+-1.6912 ? 175.5440+-2.0972 ? <geometric> 380.9535+-0.9168 ? 382.6343+-3.2999 ? might be 1.0044x slower TipOfTree ParallelHelperPool V8Spider: crypto 47.7898+-1.8614 47.6309+-1.1581 deltablue 79.6906+-3.3930 ? 79.8958+-2.8654 ? earley-boyer 41.1604+-1.1718 40.7277+-0.7002 might be 1.0106x faster raytrace 30.8145+-1.3700 ? 31.0355+-1.3871 ? regexp 61.8230+-1.0898 61.5825+-1.4560 richards 54.3931+-1.1869 54.3314+-1.7836 splay 34.6575+-1.2398 ? 36.0469+-0.9957 ? might be 1.0401x slower <geometric> 47.6916+-0.7557 ? 47.8969+-0.2706 ? might be 1.0043x slower TipOfTree ParallelHelperPool Octane: encrypt 0.16343+-0.00267 0.16310+-0.00195 decrypt 2.94817+-0.05449 2.89314+-0.00637 might be 1.0190x faster deltablue x2 0.13779+-0.00205 ? 0.13879+-0.00254 ? earley 0.30340+-0.00789 0.30223+-0.00443 boyer 4.30783+-0.02409 4.29217+-0.03977 navier-stokes x2 4.82042+-0.00795 4.80779+-0.01098 raytrace x2 0.87676+-0.05835 0.85806+-0.00420 might be 1.0218x faster richards x2 0.08839+-0.00068 ? 0.08871+-0.00080 ? splay x2 0.33154+-0.00416 0.33099+-0.00137 regexp x2 23.95248+-0.37488 ? 24.35841+-0.29745 ? might be 1.0169x slower pdfjs x2 36.95942+-0.52543 36.75775+-0.45513 mandreel x2 41.85882+-0.49120 ? 42.35157+-0.36371 ? might be 1.0118x slower gbemu x2 33.13506+-5.06346 32.03116+-2.35316 might be 1.0345x faster closure 0.57249+-0.00310 0.57035+-0.00235 jquery 7.23814+-0.02116 7.21110+-0.02052 box2d x2 9.10411+-0.06195 9.06213+-0.06407 zlib x2 386.51659+-4.52678 385.91389+-5.79452 typescript x2 652.22147+-8.99379 ? 654.91033+-13.74540 ? <geometric> 5.31726+-0.07076 5.30404+-0.02898 might be 1.0025x faster TipOfTree ParallelHelperPool Kraken: ai-astar 126.137+-0.702 126.066+-0.502 audio-beat-detection 49.756+-0.358 49.647+-0.209 audio-dft 95.089+-0.585 ? 95.293+-0.163 ? audio-fft 35.076+-0.672 35.034+-0.461 audio-oscillator 55.413+-0.717 ? 55.541+-0.980 ? imaging-darkroom 59.730+-0.164 ? 59.848+-0.183 ? imaging-desaturate 47.560+-0.227 47.531+-0.376 imaging-gaussian-blur 85.038+-0.487 ? 88.892+-6.168 ? might be 1.0453x slower json-parse-financial 37.283+-0.456 ? 38.749+-1.274 ? might be 1.0393x slower json-stringify-tinderbox 23.404+-1.308 21.919+-0.694 might be 1.0678x faster stanford-crypto-aes 40.918+-0.919 ? 40.918+-1.438 ? stanford-crypto-ccm 35.441+-1.153 ? 36.464+-1.711 ? might be 1.0289x slower stanford-crypto-pbkdf2 93.509+-0.208 ! 93.882+-0.145 ! definitely 1.0040x slower stanford-crypto-sha256-iterative 36.001+-0.640 ? 36.955+-1.952 ? might be 1.0265x slower <arithmetic> 58.597+-0.130 ? 59.053+-0.559 ? might be 1.0078x slower TipOfTree ParallelHelperPool JSRegress: abc-forward-loop-equal 29.3565+-1.0327 ? 31.0222+-1.8528 ? might be 1.0567x slower abc-postfix-backward-loop 29.7912+-0.5513 29.3668+-0.5314 might be 1.0145x faster abc-simple-backward-loop 29.4310+-1.0981 29.3534+-0.7962 abc-simple-forward-loop 29.7808+-1.5927 29.0624+-0.4607 might be 1.0247x faster abc-skippy-loop 21.6326+-0.3919 21.0678+-0.5134 might be 1.0268x faster abs-boolean 2.4848+-0.1656 2.4536+-0.0799 might be 1.0127x faster adapt-to-double-divide 16.2429+-0.1942 ? 16.4609+-0.5834 ? might be 1.0134x slower aliased-arguments-getbyval 1.1821+-0.0281 ? 1.3010+-0.3082 ? might be 1.1006x slower allocate-big-object 2.5917+-0.2934 2.4178+-0.1343 might be 1.0719x faster arguments-named-and-reflective 10.6881+-0.1507 10.4298+-0.2074 might be 1.0248x faster arguments-out-of-bounds 9.2026+-0.2504 ? 9.2072+-0.3817 ? arguments-strict-mode 9.5392+-0.6549 ? 9.5483+-0.3914 ? arguments 8.2852+-0.2653 ? 8.4510+-0.3825 ? might be 1.0200x slower arity-mismatch-inlining 0.8222+-0.0106 ? 0.8616+-0.0376 ? might be 1.0480x slower array-access-polymorphic-structure 7.0726+-0.3831 7.0293+-0.2144 array-nonarray-polymorhpic-access 23.6395+-0.5155 ? 23.7437+-0.8642 ? array-prototype-every 74.5624+-0.6370 ? 75.6059+-1.0363 ? might be 1.0140x slower array-prototype-forEach 73.7615+-1.0234 ? 73.8851+-0.9513 ? array-prototype-map 79.7467+-0.9320 ? 80.2307+-0.6464 ? array-prototype-reduce 70.2190+-0.8770 ? 71.3728+-1.1883 ? might be 1.0164x slower array-prototype-reduceRight 70.8346+-1.5496 ? 71.4663+-1.3753 ? array-prototype-some 74.4315+-0.8592 ? 74.7282+-0.8082 ? array-splice-contiguous 21.0511+-0.5749 20.3212+-0.6597 might be 1.0359x faster array-with-double-add 3.3735+-0.0578 ? 3.4865+-0.1107 ? might be 1.0335x slower array-with-double-increment 3.1378+-0.1577 3.0481+-0.0639 might be 1.0294x faster array-with-double-mul-add 4.1968+-0.0944 ? 4.2897+-0.2413 ? might be 1.0221x slower array-with-double-sum 3.2120+-0.1081 ? 3.2314+-0.0786 ? array-with-int32-add-sub 5.8442+-0.1968 ? 5.8567+-0.2952 ? array-with-int32-or-double-sum 3.2271+-0.0461 ? 3.3396+-0.1125 ? might be 1.0349x slower ArrayBuffer-DataView-alloc-large-long-lived 25.7962+-0.8155 ? 26.2505+-1.3011 ? might be 1.0176x slower ArrayBuffer-DataView-alloc-long-lived 11.8668+-0.5699 ? 12.0131+-0.5715 ? might be 1.0123x slower ArrayBuffer-Int32Array-byteOffset 3.5920+-0.1720 3.4819+-0.0327 might be 1.0316x faster ArrayBuffer-Int8Array-alloc-large-long-lived 26.4563+-0.9379 25.8607+-0.7490 might be 1.0230x faster ArrayBuffer-Int8Array-alloc-long-lived-buffer 19.2873+-0.8727 ? 19.7725+-0.5821 ? might be 1.0252x slower ArrayBuffer-Int8Array-alloc-long-lived 11.2677+-0.5516 ? 11.9112+-0.9822 ? might be 1.0571x slower ArrayBuffer-Int8Array-alloc 9.3154+-0.2028 ? 9.5407+-0.3360 ? might be 1.0242x slower arrowfunction-call 10.4272+-0.1500 ? 10.7147+-0.1907 ? might be 1.0276x slower asmjs_bool_bug 7.4775+-0.1389 ? 7.6511+-0.2795 ? might be 1.0232x slower assign-custom-setter-polymorphic 2.4796+-0.0398 2.4779+-0.1156 assign-custom-setter 3.4798+-0.0856 ^ 3.2400+-0.0417 ^ definitely 1.0740x faster basic-set 7.1322+-0.2514 ? 7.4105+-0.5650 ? might be 1.0390x slower big-int-mul 3.4929+-0.0375 ? 3.5125+-0.0412 ? boolean-test 3.0570+-0.0552 3.0556+-0.0806 branch-fold 3.5759+-0.0483 ? 3.6265+-0.0553 ? might be 1.0141x slower branch-on-string-as-boolean 17.1311+-0.6498 16.7138+-0.4266 might be 1.0250x faster by-val-generic 2.4874+-0.0899 ? 2.5834+-0.2736 ? might be 1.0386x slower call-spread-apply 27.0971+-0.8009 ? 27.5159+-1.2324 ? might be 1.0155x slower call-spread-call 21.4626+-0.4063 ? 21.5661+-0.6216 ? captured-assignments 0.5231+-0.2430 0.4132+-0.0053 might be 1.2662x faster cast-int-to-double 5.0504+-0.0742 ? 5.2079+-0.2798 ? might be 1.0312x slower cell-argument 5.9909+-0.3708 5.8680+-0.1939 might be 1.0210x faster cfg-simplify 2.8473+-0.0468 ? 2.8748+-0.0681 ? chain-getter-access 8.3989+-0.1335 8.2409+-0.1293 might be 1.0192x faster cmpeq-obj-to-obj-other 12.6030+-1.0444 11.9971+-1.2182 might be 1.0505x faster constant-test 4.8196+-0.0917 ? 4.8638+-0.1182 ? create-lots-of-functions 9.4813+-0.4761 9.2423+-0.4429 might be 1.0259x faster cse-new-array-buffer 2.3040+-0.1265 ? 2.3085+-0.0937 ? cse-new-array 2.4287+-0.0913 ? 2.6374+-0.3116 ? might be 1.0859x slower DataView-custom-properties 30.8669+-1.1588 30.6830+-1.0097 delay-tear-off-arguments-strictmode 12.3640+-0.4357 ? 12.3947+-0.4274 ? deltablue-varargs 167.8660+-3.5312 ? 172.7589+-3.2891 ? might be 1.0291x slower destructuring-arguments 159.1486+-0.6103 ? 160.2453+-1.2824 ? destructuring-parameters-overridden-by-function 0.4861+-0.0836 ? 0.5148+-0.1033 ? might be 1.0590x slower destructuring-swap 4.8428+-0.2323 4.7849+-0.0678 might be 1.0121x faster direct-arguments-getbyval 1.1347+-0.0166 ! 1.2075+-0.0500 ! definitely 1.0642x slower div-boolean-double 5.2413+-0.1098 ? 5.3645+-0.1080 ? might be 1.0235x slower div-boolean 8.2072+-0.2929 ? 8.2744+-0.2825 ? double-get-by-val-out-of-bounds 4.4515+-0.0672 ? 4.4810+-0.1781 ? double-pollution-getbyval 8.7360+-0.0920 8.6736+-0.1610 double-pollution-putbyoffset 4.0387+-0.5099 3.5668+-0.1647 might be 1.1323x faster double-real-use 24.5575+-1.4997 23.6573+-0.5163 might be 1.0381x faster double-to-int32-typed-array-no-inline 2.2629+-0.1695 2.2012+-0.1551 might be 1.0280x faster double-to-int32-typed-array 2.0365+-0.0276 ? 2.0759+-0.0821 ? might be 1.0194x slower double-to-uint32-typed-array-no-inline 2.2050+-0.0554 ? 2.3474+-0.2306 ? might be 1.0646x slower double-to-uint32-typed-array 2.0994+-0.0866 2.0953+-0.0165 elidable-new-object-dag 33.9170+-0.7471 33.5063+-0.5747 might be 1.0123x faster elidable-new-object-roflcopter 32.4588+-0.2753 ? 32.4715+-1.1477 ? elidable-new-object-then-call 31.9255+-1.1010 31.6553+-0.9619 elidable-new-object-tree 37.6223+-1.0741 36.9362+-0.4600 might be 1.0186x faster empty-string-plus-int 4.7040+-0.0544 ? 4.7133+-0.0937 ? emscripten-cube2hash 27.6944+-1.7620 27.2388+-0.6391 might be 1.0167x faster exit-length-on-plain-object 16.3444+-1.2514 15.6522+-0.5977 might be 1.0442x faster external-arguments-getbyval 1.2690+-0.0883 ? 1.3101+-0.0735 ? might be 1.0324x slower external-arguments-putbyval 2.2081+-0.1511 ? 2.2174+-0.1200 ? fixed-typed-array-storage-var-index 1.3475+-0.3490 1.2414+-0.0788 might be 1.0855x faster fixed-typed-array-storage 0.9848+-0.0544 0.9353+-0.0766 might be 1.0529x faster Float32Array-matrix-mult 3.9841+-0.0731 ? 4.0958+-0.4146 ? might be 1.0281x slower Float32Array-to-Float64Array-set 46.8357+-0.8639 46.7759+-0.7813 Float64Array-alloc-long-lived 59.5727+-1.9765 ? 60.0234+-1.3327 ? Float64Array-to-Int16Array-set 56.7699+-1.0771 56.2013+-0.6626 might be 1.0101x faster fold-double-to-int 12.1760+-0.1663 12.1534+-0.1360 fold-get-by-id-to-multi-get-by-offset-rare-int 10.4807+-0.8839 ? 11.1325+-1.0295 ? might be 1.0622x slower fold-get-by-id-to-multi-get-by-offset 10.6729+-1.5534 10.0416+-0.5594 might be 1.0629x faster fold-multi-get-by-offset-to-get-by-offset 9.2352+-0.5870 9.1383+-0.7223 might be 1.0106x faster fold-multi-get-by-offset-to-poly-get-by-offset 9.2409+-1.0132 9.0216+-0.5751 might be 1.0243x faster fold-multi-put-by-offset-to-poly-put-by-offset 9.1681+-1.2521 ? 9.9054+-1.4483 ? might be 1.0804x slower fold-multi-put-by-offset-to-put-by-offset 10.8474+-1.7473 10.8092+-0.9561 fold-multi-put-by-offset-to-replace-or-transition-put-by-offset 9.1004+-0.2344 ? 9.3192+-0.5691 ? might be 1.0240x slower fold-put-by-id-to-multi-put-by-offset 11.0763+-0.9677 ? 12.6783+-2.5683 ? might be 1.1446x slower fold-put-by-val-with-string-to-multi-put-by-offset 10.8389+-0.6064 ? 11.0068+-0.7783 ? might be 1.0155x slower fold-put-by-val-with-symbol-to-multi-put-by-offset 10.4877+-0.7344 ? 10.9761+-0.7447 ? might be 1.0466x slower fold-put-structure 7.4094+-1.1789 ? 8.1991+-0.3357 ? might be 1.1066x slower for-of-iterate-array-entries 10.7965+-0.1067 10.7600+-0.2025 for-of-iterate-array-keys 3.4501+-0.0748 ? 3.5382+-0.4152 ? might be 1.0255x slower for-of-iterate-array-values 3.6132+-0.4172 3.3318+-0.1059 might be 1.0845x faster fround 18.2111+-0.4730 18.1421+-1.1409 ftl-library-inlining-dataview 56.4142+-0.2190 55.8224+-0.6345 might be 1.0106x faster ftl-library-inlining 95.5581+-0.5499 95.2924+-0.4687 function-call 10.8105+-0.2711 ? 10.8851+-0.1592 ? function-dot-apply 2.1627+-0.1397 2.0641+-0.1477 might be 1.0478x faster function-test 2.7878+-0.2491 2.7455+-0.0944 might be 1.0154x faster function-with-eval 91.5505+-0.3059 ? 93.4033+-1.9011 ? might be 1.0202x slower gcse-poly-get-less-obvious 20.3363+-0.3851 ? 20.3853+-0.3501 ? gcse-poly-get 22.4937+-1.0273 ? 22.6642+-0.8397 ? gcse 3.5188+-0.1518 3.4017+-0.0348 might be 1.0344x faster get-by-id-bimorphic-check-structure-elimination-simple 2.6298+-0.0891 2.6172+-0.0825 get-by-id-bimorphic-check-structure-elimination 4.7826+-0.1820 4.7096+-0.0998 might be 1.0155x faster get-by-id-chain-from-try-block 2.5400+-0.2495 ? 2.5651+-0.3229 ? get-by-id-check-structure-elimination 3.9961+-0.1522 3.9094+-0.0709 might be 1.0222x faster get-by-id-proto-or-self 15.1374+-0.3066 ? 15.5473+-0.5754 ? might be 1.0271x slower get-by-id-quadmorphic-check-structure-elimination-simple 2.9500+-0.1272 2.8846+-0.0726 might be 1.0227x faster get-by-id-self-or-proto 15.7680+-1.0238 ? 15.8248+-0.5483 ? get-by-val-out-of-bounds 4.3104+-0.1146 ? 4.3688+-0.2105 ? might be 1.0136x slower get-by-val-with-string-bimorphic-check-structure-elimination-simple 2.7129+-0.0379 2.7096+-0.1101 get-by-val-with-string-bimorphic-check-structure-elimination 5.8228+-0.0281 ? 5.8603+-0.1184 ? get-by-val-with-string-chain-from-try-block 2.4559+-0.1264 2.3769+-0.0045 might be 1.0332x faster get-by-val-with-string-check-structure-elimination 5.2392+-0.0832 5.2046+-0.1634 get-by-val-with-string-proto-or-self 15.4528+-0.6247 ? 15.7096+-0.5122 ? might be 1.0166x slower get-by-val-with-string-quadmorphic-check-structure-elimination-simple 3.1660+-0.2759 3.1572+-0.1307 get-by-val-with-string-self-or-proto 15.9467+-0.6466 15.7133+-0.8377 might be 1.0149x faster get-by-val-with-symbol-bimorphic-check-structure-elimination-simple 3.0986+-0.0303 3.0863+-0.0356 get-by-val-with-symbol-bimorphic-check-structure-elimination 12.3212+-0.0864 12.2513+-0.0412 get-by-val-with-symbol-chain-from-try-block 2.4749+-0.1629 ? 2.5348+-0.2313 ? might be 1.0242x slower get-by-val-with-symbol-check-structure-elimination 11.4203+-0.4333 11.2365+-0.1053 might be 1.0164x faster get-by-val-with-symbol-proto-or-self 15.5873+-0.7831 15.5483+-0.7293 get-by-val-with-symbol-quadmorphic-check-structure-elimination-simple 4.0209+-0.4042 3.7020+-0.1585 might be 1.0861x faster get-by-val-with-symbol-self-or-proto 15.8728+-0.7238 15.8668+-1.0346 get_callee_monomorphic 2.3732+-0.1104 ? 2.5224+-0.5199 ? might be 1.0629x slower get_callee_polymorphic 3.4611+-0.3077 3.3736+-0.3556 might be 1.0259x faster getter-no-activation 4.7616+-0.1553 4.7422+-0.2508 getter-prototype 7.8935+-0.1788 ? 7.9819+-0.2259 ? might be 1.0112x slower getter-richards 119.0602+-9.9869 114.7559+-8.3484 might be 1.0375x faster getter 5.3080+-0.4459 ? 5.3659+-0.4727 ? might be 1.0109x slower global-object-access-with-mutating-structure 5.6175+-0.1955 5.5761+-0.1240 global-var-const-infer-fire-from-opt 0.8607+-0.0801 ? 0.8703+-0.0821 ? might be 1.0112x slower global-var-const-infer 0.6394+-0.0318 ? 0.6530+-0.0203 ? might be 1.0214x slower hard-overflow-check-equal 26.2897+-0.5881 ? 26.7191+-0.8121 ? might be 1.0163x slower hard-overflow-check 26.6345+-1.0546 ? 26.6432+-0.2225 ? HashMap-put-get-iterate-keys 25.7031+-1.9463 ? 26.5482+-1.8008 ? might be 1.0329x slower HashMap-put-get-iterate 28.2694+-1.1328 27.5800+-0.6482 might be 1.0250x faster HashMap-string-put-get-iterate 24.4550+-1.0591 23.7402+-0.6055 might be 1.0301x faster hoist-make-rope 9.1133+-0.7831 ? 9.2104+-0.9708 ? might be 1.0107x slower hoist-poly-check-structure-effectful-loop 3.7140+-0.1865 3.5447+-0.0450 might be 1.0478x faster hoist-poly-check-structure 3.1646+-0.0978 ? 3.1914+-0.1601 ? imul-double-only 7.8158+-0.1566 7.7408+-0.1084 imul-int-only 9.0356+-0.3665 8.7873+-0.8767 might be 1.0283x faster imul-mixed 6.8087+-0.7229 ? 6.9012+-0.2755 ? might be 1.0136x slower in-four-cases 17.0128+-0.4362 ? 17.1189+-0.5557 ? in-one-case-false 9.4655+-0.2469 ? 10.0965+-0.8116 ? might be 1.0667x slower in-one-case-true 9.9453+-0.8318 9.5506+-0.4274 might be 1.0413x faster in-two-cases 10.1339+-0.6603 9.5269+-0.1093 might be 1.0637x faster indexed-properties-in-objects 2.7344+-0.0289 ? 2.7851+-0.1134 ? might be 1.0185x slower infer-closure-const-then-mov-no-inline 3.6414+-0.1464 3.5841+-0.0362 might be 1.0160x faster infer-closure-const-then-mov 18.8162+-0.6625 ? 18.8378+-0.4839 ? infer-closure-const-then-put-to-scope-no-inline 10.8014+-0.2542 10.7742+-0.1370 infer-closure-const-then-put-to-scope 22.6386+-0.9609 ? 23.0738+-1.1797 ? might be 1.0192x slower infer-closure-const-then-reenter-no-inline 44.9486+-0.7900 ? 45.2117+-0.5123 ? infer-closure-const-then-reenter 22.6184+-0.5018 ? 23.8609+-0.9626 ? might be 1.0549x slower infer-constant-global-property 3.3894+-0.0511 3.3758+-0.0732 infer-constant-property 2.7150+-0.2658 2.6262+-0.0830 might be 1.0338x faster infer-one-time-closure-ten-vars 7.9693+-0.5260 7.7623+-0.1160 might be 1.0267x faster infer-one-time-closure-two-vars 7.6501+-0.5744 7.6404+-0.6334 infer-one-time-closure 7.2405+-0.1119 7.1614+-0.3940 might be 1.0111x faster infer-one-time-deep-closure 10.6788+-0.2469 ? 10.7645+-0.3737 ? inline-arguments-access 3.6673+-0.2454 3.6465+-0.1146 inline-arguments-aliased-access 3.6794+-0.0924 ? 3.8422+-0.3734 ? might be 1.0442x slower inline-arguments-local-escape 3.7056+-0.3390 3.5457+-0.0605 might be 1.0451x faster inline-get-scoped-var 4.5418+-0.1585 4.5095+-0.1070 inlined-put-by-id-transition 9.7186+-1.6817 8.9215+-0.4415 might be 1.0893x faster inlined-put-by-val-with-string-transition 42.6241+-1.1743 42.0691+-1.5529 might be 1.0132x faster inlined-put-by-val-with-symbol-transition 43.3025+-2.2422 42.4315+-1.4180 might be 1.0205x faster int-or-other-abs-then-get-by-val 4.4167+-0.0576 ? 4.5500+-0.1379 ? might be 1.0302x slower int-or-other-abs-zero-then-get-by-val 15.3950+-0.3113 15.3335+-0.2753 int-or-other-add-then-get-by-val 4.0481+-0.1070 ? 4.0682+-0.1162 ? int-or-other-add 4.9826+-0.2172 ? 5.0635+-0.1760 ? might be 1.0162x slower int-or-other-div-then-get-by-val 3.8110+-0.1744 3.7661+-0.0845 might be 1.0119x faster int-or-other-max-then-get-by-val 3.8939+-0.0909 ? 4.0365+-0.2729 ? might be 1.0366x slower int-or-other-min-then-get-by-val 3.7689+-0.0564 3.7610+-0.0532 int-or-other-mod-then-get-by-val 3.5532+-0.1756 3.5467+-0.1878 int-or-other-mul-then-get-by-val 3.6596+-0.0891 3.5919+-0.0789 might be 1.0189x faster int-or-other-neg-then-get-by-val 3.9577+-0.0902 ? 3.9670+-0.0326 ? int-or-other-neg-zero-then-get-by-val 15.6159+-0.3045 15.4045+-0.3922 might be 1.0137x faster int-or-other-sub-then-get-by-val 3.9610+-0.0415 ? 4.0350+-0.0530 ? might be 1.0187x slower int-or-other-sub 3.5075+-0.1072 3.4165+-0.1048 might be 1.0267x faster int-overflow-local 4.0496+-0.0315 ? 4.1213+-0.0915 ? might be 1.0177x slower Int16Array-alloc-long-lived 43.2995+-0.8622 42.8477+-1.1305 might be 1.0105x faster Int16Array-bubble-sort-with-byteLength 17.1314+-0.4944 16.6690+-0.1648 might be 1.0277x faster Int16Array-bubble-sort 16.5985+-0.3058 ? 16.9073+-0.9682 ? might be 1.0186x slower Int16Array-load-int-mul 1.4413+-0.0634 ? 1.5373+-0.1972 ? might be 1.0666x slower Int16Array-to-Int32Array-set 44.4871+-0.5405 43.1043+-0.9482 might be 1.0321x faster Int32Array-alloc-large 11.9762+-0.7968 11.9623+-0.6352 Int32Array-alloc-long-lived 48.5172+-1.3361 48.1144+-1.2252 Int32Array-alloc 2.8298+-0.2230 ? 2.8300+-0.1564 ? Int32Array-Int8Array-view-alloc 5.8675+-0.1196 ? 5.9424+-0.2765 ? might be 1.0128x slower int52-spill 4.6311+-0.3230 ? 4.6459+-0.2634 ? Int8Array-alloc-long-lived 38.1646+-1.0949 ? 38.3647+-1.1864 ? Int8Array-load-with-byteLength 3.3700+-0.0562 ? 3.7047+-0.5697 ? might be 1.0993x slower Int8Array-load 3.4621+-0.1709 3.4060+-0.1011 might be 1.0165x faster integer-divide 10.3601+-0.1310 10.3494+-0.2497 integer-modulo 1.5836+-0.0377 ? 1.5914+-0.0411 ? is-boolean-fold-tricky 3.8313+-0.2410 ? 3.8393+-0.2044 ? is-boolean-fold 2.7354+-0.2065 2.6140+-0.0324 might be 1.0464x faster is-function-fold-tricky-internal-function 9.5305+-0.1155 ? 9.6943+-0.3340 ? might be 1.0172x slower is-function-fold-tricky 4.2410+-0.4476 4.0590+-0.0645 might be 1.0448x faster is-function-fold 2.6386+-0.0755 ? 2.6755+-0.0855 ? might be 1.0140x slower is-number-fold-tricky 3.9211+-0.0345 ? 3.9917+-0.1681 ? might be 1.0180x slower is-number-fold 2.5924+-0.0381 ? 2.6210+-0.0675 ? might be 1.0110x slower is-object-or-null-fold-functions 2.7622+-0.1320 2.7206+-0.1188 might be 1.0153x faster is-object-or-null-fold-less-tricky 3.9877+-0.0408 ? 4.0920+-0.1602 ? might be 1.0262x slower is-object-or-null-fold-tricky 5.5365+-1.1037 4.7620+-0.1858 might be 1.1626x faster is-object-or-null-fold 2.7199+-0.0986 2.6864+-0.0216 might be 1.0125x faster is-object-or-null-trickier-function 3.9971+-0.0499 ? 4.0245+-0.0481 ? is-object-or-null-trickier-internal-function 9.8190+-0.1828 ! 10.1550+-0.1470 ! definitely 1.0342x slower is-object-or-null-tricky-function 4.0066+-0.0463 ? 4.2652+-0.4422 ? might be 1.0645x slower is-object-or-null-tricky-internal-function 7.3351+-0.1328 ! 7.6739+-0.0982 ! definitely 1.0462x slower is-string-fold-tricky 3.9874+-0.0531 ? 3.9973+-0.1352 ? is-string-fold 2.6693+-0.1171 2.6491+-0.0238 is-undefined-fold-tricky 3.3232+-0.0672 ? 3.3423+-0.0531 ? is-undefined-fold 2.6231+-0.0520 ? 2.6419+-0.0519 ? JSONP-negative-0 0.2604+-0.0330 0.2484+-0.0081 might be 1.0484x faster large-int-captured 4.2973+-0.3169 4.2202+-0.2129 might be 1.0183x faster large-int-neg 14.7954+-0.8013 13.9581+-0.1734 might be 1.0600x faster large-int 13.4844+-0.7982 13.0754+-0.3290 might be 1.0313x faster load-varargs-elimination 20.6490+-0.6290 20.3594+-0.8424 might be 1.0142x faster logical-not-weird-types 2.9992+-0.0327 ? 3.0846+-0.1091 ? might be 1.0285x slower logical-not 4.4164+-0.1109 4.3474+-0.1050 might be 1.0159x faster lots-of-fields 9.5087+-0.5004 9.4936+-0.5264 make-indexed-storage 2.7808+-0.2482 2.7718+-0.2194 make-rope-cse 3.9319+-0.4784 3.5344+-0.0372 might be 1.1125x faster marsaglia-larger-ints 32.9067+-1.5389 32.3362+-1.6979 might be 1.0176x faster marsaglia-osr-entry 21.2575+-0.5225 ? 21.3472+-0.4127 ? math-with-out-of-bounds-array-values 21.5580+-0.3866 ? 21.6646+-1.2230 ? max-boolean 2.6102+-0.1065 ? 2.6421+-0.0489 ? might be 1.0122x slower method-on-number 15.4083+-0.4428 ? 15.5264+-0.4749 ? min-boolean 2.5866+-0.0313 ? 2.6546+-0.0955 ? might be 1.0263x slower minus-boolean-double 3.1091+-0.0569 ? 3.1620+-0.1375 ? might be 1.0170x slower minus-boolean 2.3064+-0.0218 ? 2.3627+-0.0922 ? might be 1.0244x slower misc-strict-eq 28.2740+-0.8269 ? 29.5530+-1.9318 ? might be 1.0452x slower mod-boolean-double 11.1554+-0.3663 11.1373+-0.1828 mod-boolean 8.3034+-0.1700 ? 8.4493+-0.3886 ? might be 1.0176x slower mul-boolean-double 3.6550+-0.1095 3.5960+-0.0607 might be 1.0164x faster mul-boolean 2.8509+-0.1725 2.8058+-0.0727 might be 1.0161x faster neg-boolean 3.0974+-0.0596 3.0641+-0.0341 might be 1.0108x faster negative-zero-divide 0.3608+-0.0335 0.3603+-0.0762 negative-zero-modulo 0.3303+-0.0161 ? 0.3787+-0.0794 ? might be 1.1465x slower negative-zero-negate 0.3476+-0.0566 0.3187+-0.0024 might be 1.0904x faster nested-function-parsing 44.3195+-0.5547 ? 44.3249+-0.6894 ? new-array-buffer-dead 86.6065+-0.3025 ? 87.4377+-0.9717 ? new-array-buffer-push 6.6640+-1.0166 6.1062+-0.1240 might be 1.0914x faster new-array-dead 15.1968+-1.0878 14.7859+-0.6634 might be 1.0278x faster new-array-push 3.4681+-0.1090 3.3714+-0.0456 might be 1.0287x faster no-inline-constructor 30.9118+-0.7755 ? 30.9155+-0.7164 ? number-test 3.0039+-0.0382 ? 3.0677+-0.0561 ? might be 1.0213x slower object-closure-call 4.8431+-0.0256 ? 4.8966+-0.1231 ? might be 1.0111x slower object-get-own-property-symbols-on-large-array 4.2254+-0.2372 4.0466+-0.1704 might be 1.0442x faster object-test 2.6862+-0.0459 ? 2.7699+-0.0484 ? might be 1.0311x slower obvious-sink-pathology-taken 97.5746+-0.6149 ? 97.6515+-0.6388 ? obvious-sink-pathology 92.8748+-1.1597 ? 92.9670+-1.2502 ? obviously-elidable-new-object 28.5708+-0.6228 ? 28.8054+-0.7981 ? plus-boolean-arith 2.3924+-0.1264 ? 2.4629+-0.1670 ? might be 1.0294x slower plus-boolean-double 3.1053+-0.0514 3.0969+-0.0473 plus-boolean 2.5821+-0.1540 2.5259+-0.0501 might be 1.0222x faster poly-chain-access-different-prototypes-simple 3.1890+-0.0427 3.1623+-0.0310 poly-chain-access-different-prototypes 3.2613+-0.0501 ? 3.2665+-0.0944 ? poly-chain-access-simpler 3.2534+-0.1208 ? 3.6232+-0.8147 ? might be 1.1137x slower poly-chain-access 3.2681+-0.0795 ? 3.3290+-0.1545 ? might be 1.0186x slower poly-stricteq 48.8830+-0.1800 ? 48.9623+-0.4356 ? polymorphic-array-call 1.3306+-0.0548 1.2556+-0.0601 might be 1.0597x faster polymorphic-get-by-id 2.8398+-0.0416 ? 2.8664+-0.0854 ? polymorphic-put-by-id 28.2633+-1.4489 28.1894+-1.7182 polymorphic-put-by-val-with-string 28.5860+-0.4248 ? 29.0419+-1.6790 ? might be 1.0160x slower polymorphic-put-by-val-with-symbol 29.0743+-0.5832 29.0348+-1.0786 polymorphic-structure 14.0579+-3.0330 12.4952+-0.4418 might be 1.1251x faster polyvariant-monomorphic-get-by-id 7.5096+-1.2529 6.3908+-0.8487 might be 1.1751x faster proto-getter-access 8.3421+-0.2234 ? 8.3998+-0.3137 ? prototype-access-with-mutating-prototype 5.4835+-0.2098 5.4172+-0.2232 might be 1.0122x faster put-by-id-replace-and-transition 7.6007+-0.4711 ? 7.7330+-0.7646 ? might be 1.0174x slower put-by-id-slightly-polymorphic 2.6693+-0.0202 ? 2.7060+-0.0292 ? might be 1.0138x slower put-by-id 10.1565+-0.5094 9.8857+-0.4965 might be 1.0274x faster put-by-val-direct 0.3335+-0.0106 ? 0.3800+-0.0618 ? might be 1.1396x slower put-by-val-large-index-blank-indexing-type 5.2738+-0.2928 ? 5.7930+-0.8042 ? might be 1.0985x slower put-by-val-machine-int 2.5371+-0.1509 2.4472+-0.0720 might be 1.0367x faster put-by-val-with-string-replace-and-transition 10.4202+-0.2969 ? 10.7699+-0.4777 ? might be 1.0336x slower put-by-val-with-string-slightly-polymorphic 2.9110+-0.0316 ? 3.0379+-0.1894 ? might be 1.0436x slower put-by-val-with-string 10.1071+-0.3918 ? 10.5052+-0.6443 ? might be 1.0394x slower put-by-val-with-symbol-replace-and-transition 11.8419+-0.4317 11.7856+-0.4352 put-by-val-with-symbol-slightly-polymorphic 3.2135+-0.1890 ? 3.3250+-0.2135 ? might be 1.0347x slower put-by-val-with-symbol 10.2437+-0.4533 ? 10.9224+-0.8477 ? might be 1.0663x slower rare-osr-exit-on-local 13.8951+-0.4876 13.3469+-0.0870 might be 1.0411x faster raytrace-with-empty-try-catch 5.1696+-0.1871 5.0948+-0.0404 might be 1.0147x faster raytrace-with-try-catch 9.9105+-0.1863 ? 9.9575+-0.2816 ? register-pressure-from-osr 16.4480+-0.2775 ? 16.5224+-0.4097 ? repeat-multi-get-by-offset 20.4210+-0.7246 ? 20.4836+-0.5127 ? richards-empty-try-catch 74.6801+-0.7977 74.1591+-1.0262 richards-try-catch 241.4964+-2.1454 240.9025+-1.9898 setter-prototype 7.8885+-0.3935 ? 7.9046+-0.3373 ? setter 5.6474+-0.4507 5.5835+-0.6598 might be 1.0114x faster simple-activation-demo 24.1150+-0.4533 24.0153+-0.4877 simple-getter-access 10.7463+-0.4197 ? 10.8721+-0.3903 ? might be 1.0117x slower simple-poly-call-nested 9.1348+-0.3713 8.5956+-0.6041 might be 1.0627x faster simple-poly-call 1.2584+-0.0267 ? 1.3909+-0.1834 ? might be 1.1053x slower sin-boolean 20.2623+-2.0284 18.7060+-1.6909 might be 1.0832x faster singleton-scope 55.4791+-0.5753 ? 56.1047+-0.7566 ? might be 1.0113x slower sink-function 10.0931+-0.5567 ? 10.4807+-1.2054 ? might be 1.0384x slower sink-huge-activation 16.9272+-0.9447 16.8241+-0.9365 sinkable-new-object-dag 53.3750+-0.5963 ? 55.5417+-2.7107 ? might be 1.0406x slower sinkable-new-object-taken 44.1523+-0.8778 43.8390+-1.0670 sinkable-new-object 29.5261+-0.3168 ? 29.6653+-0.6363 ? slow-array-profile-convergence 2.5245+-0.1113 2.5113+-0.0671 slow-convergence 2.3451+-0.0537 2.3012+-0.0371 might be 1.0191x faster slow-ternaries 17.9653+-1.0141 17.4788+-0.4184 might be 1.0278x faster sorting-benchmark 16.9148+-0.7482 ? 17.0519+-0.8636 ? sparse-conditional 1.1746+-0.0539 ? 1.1852+-0.1140 ? splice-to-remove 12.2147+-0.4477 ? 12.3866+-0.4591 ? might be 1.0141x slower string-char-code-at 12.9472+-0.1404 ? 13.1440+-0.2408 ? might be 1.0152x slower string-concat-object 2.2061+-0.1055 2.1649+-0.0612 might be 1.0190x faster string-concat-pair-object 2.1234+-0.1601 2.1231+-0.0818 string-concat-pair-simple 9.3406+-0.4451 ? 9.7102+-0.9358 ? might be 1.0396x slower string-concat-simple 9.7193+-0.4496 ? 10.5088+-2.1247 ? might be 1.0812x slower string-cons-repeat 6.3597+-0.1519 ? 6.4084+-0.1793 ? string-cons-tower 6.6364+-0.1780 6.5843+-0.3248 string-equality 15.0114+-0.3327 14.9855+-0.3474 string-get-by-val-big-char 6.5044+-0.0943 6.4815+-0.0621 string-get-by-val-out-of-bounds-insane 3.2540+-0.3156 3.0799+-0.1214 might be 1.0565x faster string-get-by-val-out-of-bounds 4.1152+-0.3561 3.9595+-0.2400 might be 1.0393x faster string-get-by-val 2.8127+-0.0427 2.7889+-0.0166 string-hash 1.9093+-0.1614 1.8412+-0.0526 might be 1.0370x faster string-long-ident-equality 13.5992+-0.9678 13.2061+-0.2427 might be 1.0298x faster string-out-of-bounds 10.2600+-0.6198 10.2426+-0.5000 string-repeat-arith 27.1294+-1.9850 26.4701+-0.5211 might be 1.0249x faster string-sub 53.9452+-0.5260 ? 54.1053+-0.5833 ? string-test 2.8825+-0.0913 ? 2.8994+-0.1128 ? string-var-equality 26.5474+-0.6164 ? 27.1690+-1.5696 ? might be 1.0234x slower structure-hoist-over-transitions 2.3041+-0.0899 ? 2.3405+-0.1026 ? might be 1.0158x slower substring-concat-weird 35.9076+-0.7507 ? 36.0751+-1.8344 ? substring-concat 39.2244+-0.9722 ? 39.3648+-0.9156 ? substring 44.4754+-0.6751 ? 44.7293+-1.0340 ? switch-char-constant 2.6519+-0.0535 ? 2.7227+-0.0715 ? might be 1.0267x slower switch-char 5.9419+-0.9988 ? 6.2082+-0.9391 ? might be 1.0448x slower switch-constant 8.3178+-1.0719 ? 8.3855+-0.8487 ? switch-string-basic-big-var 14.5553+-0.4162 14.2636+-0.1317 might be 1.0204x faster switch-string-basic-big 15.0913+-0.3046 ? 15.1725+-0.1661 ? switch-string-basic-var 13.4137+-0.2847 ? 13.5043+-0.4941 ? switch-string-basic 12.5241+-0.5112 12.5144+-0.3866 switch-string-big-length-tower-var 18.2237+-0.8138 18.1006+-0.4993 switch-string-length-tower-var 13.2668+-0.3905 13.2432+-0.3911 switch-string-length-tower 11.2128+-0.1285 ? 11.3698+-0.0545 ? might be 1.0140x slower switch-string-short 11.3896+-0.2214 11.2258+-0.1083 might be 1.0146x faster switch 11.0766+-0.8017 ? 11.1951+-0.8196 ? might be 1.0107x slower tear-off-arguments-simple 3.1184+-0.1647 ? 3.4198+-0.7459 ? might be 1.0967x slower tear-off-arguments 4.2604+-0.4418 3.9475+-0.0788 might be 1.0793x faster temporal-structure 12.2130+-0.7872 11.6562+-0.1047 might be 1.0478x faster to-int32-boolean 12.8454+-0.6960 12.5456+-0.2090 might be 1.0239x faster try-catch-get-by-val-cloned-arguments 9.6792+-0.4189 9.5939+-0.2517 try-catch-get-by-val-direct-arguments 2.2733+-0.2505 2.2324+-0.2693 might be 1.0183x faster try-catch-get-by-val-scoped-arguments 4.9942+-0.2145 4.6955+-0.1419 might be 1.0636x faster typed-array-get-set-by-val-profiling 27.4358+-0.5585 26.8394+-0.4618 might be 1.0222x faster undefined-property-access 216.0598+-0.6995 ? 216.4635+-1.4428 ? undefined-test 2.9856+-0.0898 ? 3.0613+-0.0809 ? might be 1.0254x slower unprofiled-licm 9.6752+-0.3929 9.3211+-0.4896 might be 1.0380x faster v8-raytrace-with-empty-try-catch 49.2170+-0.4580 ? 49.5795+-1.9464 ? v8-raytrace-with-try-catch 61.7539+-0.5760 61.5331+-0.7321 varargs-call 12.7864+-0.1292 ? 13.0188+-0.2522 ? might be 1.0182x slower varargs-construct-inline 21.8156+-0.8146 21.5449+-0.6859 might be 1.0126x faster varargs-construct 20.5729+-0.6290 ? 21.1206+-0.7438 ? might be 1.0266x slower varargs-inline 8.6338+-0.1354 ? 8.6540+-0.2113 ? varargs-strict-mode 9.6808+-0.3116 9.6558+-0.3358 varargs 9.4506+-0.1032 9.4299+-0.0304 weird-inlining-const-prop 2.1017+-0.0881 ? 2.1657+-0.1970 ? might be 1.0305x slower <geometric> 7.9726+-0.0181 7.9709+-0.0132 might be 1.0002x faster TipOfTree ParallelHelperPool AsmBench: bigfib.cpp 442.1815+-6.3538 ? 453.0791+-13.7661 ? might be 1.0246x slower cray.c 392.3013+-3.3147 392.1994+-4.0855 dry.c 417.9313+-7.8421 415.9226+-7.4424 FloatMM.c 681.3876+-3.5137 ? 681.4820+-2.3939 ? gcc-loops.cpp 3408.7941+-16.8878 ? 3415.2036+-33.6737 ? n-body.c 819.6172+-0.6425 ? 862.8707+-105.3475 ? might be 1.0528x slower Quicksort.c 404.9588+-4.2404 403.1143+-4.7886 stepanov_container.cpp 3476.1975+-33.0324 3468.5560+-26.9007 Towers.c 233.8833+-2.2038 232.3964+-1.1485 <geometric> 709.2614+-2.3591 ? 713.4667+-7.4889 ? might be 1.0059x slower TipOfTree ParallelHelperPool CompressionBench: huffman 59.1054+-1.4146 ? 61.6995+-6.2506 ? might be 1.0439x slower arithmetic-simple 272.2116+-1.4485 ^ 269.0524+-0.4276 ^ definitely 1.0117x faster arithmetic-precise 241.2559+-1.2868 ? 241.3825+-0.7072 ? arithmetic-complex-precise 243.7722+-2.3799 242.4104+-1.9398 arithmetic-precise-order-0 278.2844+-1.9047 278.0792+-1.0005 arithmetic-precise-order-1 297.6564+-2.1971 ? 299.7858+-5.4466 ? arithmetic-precise-order-2 341.1390+-3.0461 ? 342.3434+-2.3227 ? arithmetic-simple-order-1 317.5185+-2.1918 317.3589+-1.2325 arithmetic-simple-order-2 369.9989+-5.6354 ? 370.6637+-5.8619 ? lz-string 310.2207+-5.0437 306.0996+-1.4824 might be 1.0135x faster <geometric> 250.5231+-0.8076 ? 251.0277+-2.0638 ? might be 1.0020x slower TipOfTree ParallelHelperPool Geomean of preferred means: <scaled-result> 49.5484+-0.1158 ? 49.6636+-0.1532 ? might be 1.0023x slower
Created attachment 261907 [details] getter better It passes tests. It has build things for VS/CMake. It performs well. Now I just need to check that it works with two VMs.
Attachment 261907 [details] did not pass style-queue: ERROR: Source/JavaScriptCore/heap/HeapHelperPool.cpp:30: Bad include order. Mixing system and custom headers. [build/include_order] [4] ERROR: Source/WTF/wtf/ParallelHelperPool.cpp:181: Place brace on its own line for function definitions. [whitespace/braces] [4] ERROR: Source/JavaScriptCore/runtime/CodeCache.h:38: Alphabetical sorting problem. [build/include_order] [4] ERROR: Source/JavaScriptCore/heap/Heap.cpp:550: Place brace on its own line for function definitions. [whitespace/braces] [4] ERROR: Source/JavaScriptCore/heap/Heap.cpp:635: Place brace on its own line for function definitions. [whitespace/braces] [4] ERROR: Source/JavaScriptCore/runtime/VM.h:68: Alphabetical sorting problem. [build/include_order] [4] Total errors found: 6 in 23 files If any of these errors are false positives, please file a bug against check-webkit-style.
Created attachment 261910 [details] the patch
Attachment 261910 [details] did not pass style-queue: ERROR: Source/JavaScriptCore/API/tests/testapi.mm:495: Weird number of spaces at line-start. Are you using a 4-space indent? [whitespace/indent] [3] ERROR: Source/JavaScriptCore/API/tests/testapi.mm:496: Weird number of spaces at line-start. Are you using a 4-space indent? [whitespace/indent] [3] ERROR: Source/JavaScriptCore/API/tests/testapi.mm:497: Weird number of spaces at line-start. Are you using a 4-space indent? [whitespace/indent] [3] ERROR: Source/JavaScriptCore/API/tests/testapi.mm:498: Weird number of spaces at line-start. Are you using a 4-space indent? [whitespace/indent] [3] ERROR: Source/JavaScriptCore/API/tests/testapi.mm:499: Weird number of spaces at line-start. Are you using a 4-space indent? [whitespace/indent] [3] ERROR: Source/JavaScriptCore/API/tests/testapi.mm:500: Weird number of spaces at line-start. Are you using a 4-space indent? [whitespace/indent] [3] ERROR: Source/JavaScriptCore/heap/HeapHelperPool.cpp:30: Bad include order. Mixing system and custom headers. [build/include_order] [4] ERROR: Source/WTF/wtf/ParallelHelperPool.cpp:181: Place brace on its own line for function definitions. [whitespace/braces] [4] ERROR: Source/JavaScriptCore/runtime/CodeCache.h:38: Alphabetical sorting problem. [build/include_order] [4] ERROR: Source/JavaScriptCore/heap/Heap.cpp:550: Place brace on its own line for function definitions. [whitespace/braces] [4] ERROR: Source/JavaScriptCore/heap/Heap.cpp:635: Place brace on its own line for function definitions. [whitespace/braces] [4] ERROR: Source/JavaScriptCore/runtime/VM.h:68: Alphabetical sorting problem. [build/include_order] [4] Total errors found: 12 in 24 files If any of these errors are false positives, please file a bug against check-webkit-style.
Are there any tests to unskip with this? I think that some worker tests may have been skipped since GC started to use many threads.
Comment on attachment 261910 [details] the patch View in context: https://bugs.webkit.org/attachment.cgi?id=261910&action=review > Source/JavaScriptCore/heap/Heap.cpp:604 > + m_helperClient.finish(); Shouldn't we doSomeHelping at some point? It seems like the main thread will not participate in marking unless we do some helping. > Source/WTF/ChangeLog:39 > + join after the other threads have already started. it's not advisable to make algorithmic It's > Source/WTF/wtf/ParallelHelperPool.cpp:56 > +void ParallelHelperClient::setTask(RefPtr<ParallelHelperTask> task) I don't think the setter naming idiom works here. The post condition for a setter is that the value you set should now be visible in the member variable. But this function will clear the value you set before returning. Let's call this runTask. > Source/WTF/wtf/ParallelHelperPool.cpp:83 > +void ParallelHelperClient::runTaskInParallel(RefPtr<ParallelHelperTask> task) Let's call this runTaskSynchronously. We don't need parallel in the function name since it's in the class name. > Source/WTF/wtf/ParallelHelperPool.cpp:180 > + "Parallel Helper Thread", Let's namespace this since it will show up in process-global output. WTF? WebKit? > Source/WTF/wtf/ParallelHelperPool.h:61 > +// run this task on potentially many threads. The pool may not run the task on any threads, of all of => if > Source/WTF/wtf/ParallelHelperPool.h:169 > +class ParallelHelperClient { One class per file please. > Source/WTF/wtf/ParallelHelperPool.h:170 > + WTF_MAKE_NONCOPYABLE(ParallelHelperClient); WTF_MAKE_FAST_ALLOCATED; Two lines please. > Source/WTF/wtf/ParallelHelperPool.h:181 > + void setTask(RefPtr<ParallelHelperTask>); > + > + template<typename Functor> > + void setFunction(const Functor& functor) > + { > + setTask(createParallelHelperTask(functor)); > + } I think you can eliminate this task vs function distinction and the ParallelHelperTask and ParallelHelperTaskFunctor abstraction by using std::function. > Source/WTF/wtf/ParallelHelperPool.h:218 > +class ParallelHelperPool : public ThreadSafeRefCounted<ParallelHelperPool> { One class per file please. "Helper" is a bit vague, as is "Helper client". Maybe we can call these "ParallelTaskRunner (the pool) and ParallelTask (the 'client')". (Those names are currently sort of taken by the functor classes, but see my suggestion to remove those classes, or we could just rename them to ParallelTaskRunnable and ParallelTaskFunctor.) > Source/WTF/wtf/ParallelHelperPool.h:224 > + void addThreads(unsigned numThreads); > + void ensureThreads(unsigned numThreads); Nobody seems to call these :(. > Source/WTF/wtf/WeakRandom.h:100 > + uint64_t cutoff = (static_cast<uint64_t>(UINT_MAX) + 1) / limit * limit; > + for (;;) { > + uint64_t value = getUint32(); > + if (value >= cutoff) > + continue; > + return value % limit; > + } I don't get this. What values are we trying to defend against? Should % always be safe as long as we're dealing with uint32?
(In reply to comment #12) > Comment on attachment 261910 [details] > the patch > > View in context: > https://bugs.webkit.org/attachment.cgi?id=261910&action=review > > > Source/JavaScriptCore/heap/Heap.cpp:604 > > + m_helperClient.finish(); > > Shouldn't we doSomeHelping at some point? It seems like the main thread will > not participate in marking unless we do some helping. No, that would be wrong. Parallel marking uses a different marking algorithm on the helper threads than on the main thread. doSomeHelping() would run marking-in-helper instead of marking-on-main-thread. Note that this is not a change in behavior from what we do on trunk. Parallel GC has always used different code for the marking helpers. > > > Source/WTF/ChangeLog:39 > > + join after the other threads have already started. it's not advisable to make algorithmic > > It's > > > Source/WTF/wtf/ParallelHelperPool.cpp:56 > > +void ParallelHelperClient::setTask(RefPtr<ParallelHelperTask> task) > > I don't think the setter naming idiom works here. The post condition for a > setter is that the value you set should now be visible in the member > variable. But this function will clear the value you set before returning. No, it won't. setTask() doesn't clear the task. > > Let's call this runTask. That would be wrong, because setTask() does not run the task. I believe that setTask() is the correct name, since it just sets m_task and notifies the helper threads. > > > Source/WTF/wtf/ParallelHelperPool.cpp:83 > > +void ParallelHelperClient::runTaskInParallel(RefPtr<ParallelHelperTask> task) > > Let's call this runTaskSynchronously. We don't need parallel in the function > name since it's in the class name. I really don't like calling it runTaskSynchronously. "Synchronously" implies that there is some synchronization being performed, and could easily confuse people into thinking that the task will only run in one thread at a time. In fact, the task will experience parallelism - the same task instance will most likely be run from more than one thread. I don't mind that "Parallel" is both in the method name and the class name, since it's a useful reminder. Perhaps a better name is "runTask()", if you really think that the redundant use of "Parallel" should be avoided. > > > Source/WTF/wtf/ParallelHelperPool.cpp:180 > > + "Parallel Helper Thread", > > Let's namespace this since it will show up in process-global output. WTF? > WebKit? Good point, I'll use "WTF Parallel Helper Thread". > > > Source/WTF/wtf/ParallelHelperPool.h:61 > > +// run this task on potentially many threads. The pool may not run the task on any threads, of all > > of => if > > > Source/WTF/wtf/ParallelHelperPool.h:169 > > +class ParallelHelperClient { > > One class per file please. I'm not sure if that's a good idea. This header is included from everything that includes Heap.h, which is basically everything in the whole project. I've heard that build times are somewhat related to the number of headers that need to be included, and I've had requests in the past to coalesce headers that are included everywhere for the sake of WebCore build performance. That's one of the reasons why I used one header file. I also think that because of the tight coupling between these classes, one file simplifies things a bit. > > > Source/WTF/wtf/ParallelHelperPool.h:170 > > + WTF_MAKE_NONCOPYABLE(ParallelHelperClient); WTF_MAKE_FAST_ALLOCATED; > > Two lines please. > > > Source/WTF/wtf/ParallelHelperPool.h:181 > > + void setTask(RefPtr<ParallelHelperTask>); > > + > > + template<typename Functor> > > + void setFunction(const Functor& functor) > > + { > > + setTask(createParallelHelperTask(functor)); > > + } > > I think you can eliminate this task vs function distinction and the > ParallelHelperTask and ParallelHelperTaskFunctor abstraction by using > std::function. I don't think that's a good idea, unless you are proposing wrapping std::function inside a ref-counted object, which is exactly what ParallelHelperTaskFunctor does. Notice that a key part of the helper thread algorithm is: get a reference to the current task while holding the lock, then release the lock, then run the task. If we were using std::function, the "getting a reference" would require doing a copy of the std::function's payload. I don't like that, since it means that the helper thread worklist algorithm's performance would be related to the number of variables that you closed over. Also, copying std::function's that have more than ~3 captured variables requires system malloc. That would mean that we'd be calling system malloc while holding the single lock that protects the helper pool. That seems like it will cause scalability problems in the future. > > > Source/WTF/wtf/ParallelHelperPool.h:218 > > +class ParallelHelperPool : public ThreadSafeRefCounted<ParallelHelperPool> { > > One class per file please. > > "Helper" is a bit vague, as is "Helper client". > > Maybe we can call these "ParallelTaskRunner (the pool) and ParallelTask (the > 'client')". (Those names are currently sort of taken by the functor classes, > but see my suggestion to remove those classes, or we could just rename them > to ParallelTaskRunnable and ParallelTaskFunctor.) I'm fine with maybe changing the names of things, but I'd like to preserve the performance guarantees by using a reference-counted functor rather than std::function. > > > Source/WTF/wtf/ParallelHelperPool.h:224 > > + void addThreads(unsigned numThreads); > > + void ensureThreads(unsigned numThreads); > > Nobody seems to call these :(. See heap/HeapHelperPool.cpp. You're right that it only calls one of them. I can remove the one that it doesn't call. > > > Source/WTF/wtf/WeakRandom.h:100 > > + uint64_t cutoff = (static_cast<uint64_t>(UINT_MAX) + 1) / limit * limit; > > + for (;;) { > > + uint64_t value = getUint32(); > > + if (value >= cutoff) > > + continue; > > + return value % limit; > > + } > > I don't get this. What values are we trying to defend against? Should % > always be safe as long as we're dealing with uint32? The % is "safe" but it would lead to bias. For example, if you just did random.getUint32() % 3, then you'd be more likely to get 0 than 2. One easy way to defend against this is to ignore return values from getUint32() that are above the cutoff above which the modulo experiences a bias. That's what I'm doing here. It doesn't matter *too* much but I figured that if I'm putting WeakRandom into WTF and adding a getUint32(limit) API, I might as well do it the right way.
Geoff and I talked in person. I'll move ParallelHelperTask into its own header, called Task. I'll make some of the other fixes. (In reply to comment #13) > (In reply to comment #12) > > Comment on attachment 261910 [details] > > the patch > > > > View in context: > > https://bugs.webkit.org/attachment.cgi?id=261910&action=review > > > > > Source/JavaScriptCore/heap/Heap.cpp:604 > > > + m_helperClient.finish(); > > > > Shouldn't we doSomeHelping at some point? It seems like the main thread will > > not participate in marking unless we do some helping. > > No, that would be wrong. Parallel marking uses a different marking > algorithm on the helper threads than on the main thread. doSomeHelping() > would run marking-in-helper instead of marking-on-main-thread. > > Note that this is not a change in behavior from what we do on trunk. > Parallel GC has always used different code for the marking helpers. We'll keep this the same. > > > > > > Source/WTF/ChangeLog:39 > > > + join after the other threads have already started. it's not advisable to make algorithmic > > > > It's I'll fix this. > > > > > Source/WTF/wtf/ParallelHelperPool.cpp:56 > > > +void ParallelHelperClient::setTask(RefPtr<ParallelHelperTask> task) > > > > I don't think the setter naming idiom works here. The post condition for a > > setter is that the value you set should now be visible in the member > > variable. But this function will clear the value you set before returning. > > No, it won't. setTask() doesn't clear the task. > > > > > Let's call this runTask. > > That would be wrong, because setTask() does not run the task. > > I believe that setTask() is the correct name, since it just sets m_task and > notifies the helper threads. > We'll keep this the same. > > > > > Source/WTF/wtf/ParallelHelperPool.cpp:83 > > > +void ParallelHelperClient::runTaskInParallel(RefPtr<ParallelHelperTask> task) > > > > Let's call this runTaskSynchronously. We don't need parallel in the function > > name since it's in the class name. > > I really don't like calling it runTaskSynchronously. "Synchronously" > implies that there is some synchronization being performed, and could easily > confuse people into thinking that the task will only run in one thread at a > time. In fact, the task will experience parallelism - the same task > instance will most likely be run from more than one thread. I don't mind > that "Parallel" is both in the method name and the class name, since it's a > useful reminder. > > Perhaps a better name is "runTask()", if you really think that the redundant > use of "Parallel" should be avoided. > We'll keep this the same. > > > > > Source/WTF/wtf/ParallelHelperPool.cpp:180 > > > + "Parallel Helper Thread", > > > > Let's namespace this since it will show up in process-global output. WTF? > > WebKit? > > Good point, I'll use "WTF Parallel Helper Thread". > I'll fix this. > > > > > Source/WTF/wtf/ParallelHelperPool.h:61 > > > +// run this task on potentially many threads. The pool may not run the task on any threads, of all > > > > of => if I'll fix this. > > > > > Source/WTF/wtf/ParallelHelperPool.h:169 > > > +class ParallelHelperClient { > > > > One class per file please. > > I'm not sure if that's a good idea. This header is included from everything > that includes Heap.h, which is basically everything in the whole project. > I've heard that build times are somewhat related to the number of headers > that need to be included, and I've had requests in the past to coalesce > headers that are included everywhere for the sake of WebCore build > performance. > > That's one of the reasons why I used one header file. I also think that > because of the tight coupling between these classes, one file simplifies > things a bit. I'll move task into its own header. > > > > > > Source/WTF/wtf/ParallelHelperPool.h:170 > > > + WTF_MAKE_NONCOPYABLE(ParallelHelperClient); WTF_MAKE_FAST_ALLOCATED; > > > > Two lines please. I'll fix this. > > > > > Source/WTF/wtf/ParallelHelperPool.h:181 > > > + void setTask(RefPtr<ParallelHelperTask>); > > > + > > > + template<typename Functor> > > > + void setFunction(const Functor& functor) > > > + { > > > + setTask(createParallelHelperTask(functor)); > > > + } > > > > I think you can eliminate this task vs function distinction and the > > ParallelHelperTask and ParallelHelperTaskFunctor abstraction by using > > std::function. > > I don't think that's a good idea, unless you are proposing wrapping > std::function inside a ref-counted object, which is exactly what > ParallelHelperTaskFunctor does. > > Notice that a key part of the helper thread algorithm is: get a reference to > the current task while holding the lock, then release the lock, then run the > task. If we were using std::function, the "getting a reference" would > require doing a copy of the std::function's payload. I don't like that, > since it means that the helper thread worklist algorithm's performance would > be related to the number of variables that you closed over. > > Also, copying std::function's that have more than ~3 captured variables > requires system malloc. That would mean that we'd be calling system malloc > while holding the single lock that protects the helper pool. That seems > like it will cause scalability problems in the future. We'll keep this the same. > > > > > > Source/WTF/wtf/ParallelHelperPool.h:218 > > > +class ParallelHelperPool : public ThreadSafeRefCounted<ParallelHelperPool> { > > > > One class per file please. > > > > "Helper" is a bit vague, as is "Helper client". > > > > Maybe we can call these "ParallelTaskRunner (the pool) and ParallelTask (the > > 'client')". (Those names are currently sort of taken by the functor classes, > > but see my suggestion to remove those classes, or we could just rename them > > to ParallelTaskRunnable and ParallelTaskFunctor.) > > I'm fine with maybe changing the names of things, but I'd like to preserve > the performance guarantees by using a reference-counted functor rather than > std::function. > We'll keep this the same. > > > > > Source/WTF/wtf/ParallelHelperPool.h:224 > > > + void addThreads(unsigned numThreads); > > > + void ensureThreads(unsigned numThreads); > > > > Nobody seems to call these :(. > > See heap/HeapHelperPool.cpp. You're right that it only calls one of them. > I can remove the one that it doesn't call. > I'll remove addThreads(). > > > > > Source/WTF/wtf/WeakRandom.h:100 > > > + uint64_t cutoff = (static_cast<uint64_t>(UINT_MAX) + 1) / limit * limit; > > > + for (;;) { > > > + uint64_t value = getUint32(); > > > + if (value >= cutoff) > > > + continue; > > > + return value % limit; > > > + } > > > > I don't get this. What values are we trying to defend against? Should % > > always be safe as long as we're dealing with uint32? > > The % is "safe" but it would lead to bias. For example, if you just did > random.getUint32() % 3, then you'd be more likely to get 0 than 2. One easy > way to defend against this is to ignore return values from getUint32() that > are above the cutoff above which the modulo experiences a bias. That's what > I'm doing here. It doesn't matter *too* much but I figured that if I'm > putting WeakRandom into WTF and adding a getUint32(limit) API, I might as > well do it the right way. We'll keep this the same.
Created attachment 261934 [details] patch for landing
Attachment 261934 [details] did not pass style-queue: ERROR: Source/JavaScriptCore/API/tests/testapi.mm:495: Weird number of spaces at line-start. Are you using a 4-space indent? [whitespace/indent] [3] ERROR: Source/JavaScriptCore/API/tests/testapi.mm:496: Weird number of spaces at line-start. Are you using a 4-space indent? [whitespace/indent] [3] ERROR: Source/JavaScriptCore/API/tests/testapi.mm:497: Weird number of spaces at line-start. Are you using a 4-space indent? [whitespace/indent] [3] ERROR: Source/JavaScriptCore/API/tests/testapi.mm:498: Weird number of spaces at line-start. Are you using a 4-space indent? [whitespace/indent] [3] ERROR: Source/JavaScriptCore/API/tests/testapi.mm:499: Weird number of spaces at line-start. Are you using a 4-space indent? [whitespace/indent] [3] ERROR: Source/JavaScriptCore/API/tests/testapi.mm:500: Weird number of spaces at line-start. Are you using a 4-space indent? [whitespace/indent] [3] ERROR: Source/JavaScriptCore/heap/HeapHelperPool.cpp:30: Bad include order. Mixing system and custom headers. [build/include_order] [4] ERROR: Source/WTF/wtf/ParallelHelperPool.cpp:173: Place brace on its own line for function definitions. [whitespace/braces] [4] ERROR: Source/JavaScriptCore/runtime/CodeCache.h:38: Alphabetical sorting problem. [build/include_order] [4] ERROR: Source/JavaScriptCore/heap/Heap.cpp:550: Place brace on its own line for function definitions. [whitespace/braces] [4] ERROR: Source/JavaScriptCore/heap/Heap.cpp:635: Place brace on its own line for function definitions. [whitespace/braces] [4] ERROR: Source/JavaScriptCore/runtime/VM.h:68: Alphabetical sorting problem. [build/include_order] [4] Total errors found: 12 in 25 files If any of these errors are false positives, please file a bug against check-webkit-style.
Landed in http://trac.webkit.org/changeset/190267
This broke Windows builds: https://build.webkit.org/builders/Apple%20Win%20Debug%20%28Build%29/builds/91251/steps/compile-webkit/logs/stdio
(In reply to comment #18) > This broke Windows builds: > https://build.webkit.org/builders/Apple%20Win%20Debug%20%28Build%29/builds/ > 91251/steps/compile-webkit/logs/stdio Fixing it.
(In reply to comment #19) > (In reply to comment #18) > > This broke Windows builds: > > https://build.webkit.org/builders/Apple%20Win%20Debug%20%28Build%29/builds/ > > 91251/steps/compile-webkit/logs/stdio > > Fixing it. Fixed in http://trac.webkit.org/changeset/190268
Now lots of layout tests are crashing in Debug builds: https://build.webkit.org/results/Apple%20Mavericks%20Debug%20WK1%20(Tests)/r190267%20(15885)/results.html e.g. stderr: ASSERTION FAILED: isMainThreadOrGCThread() /Volumes/Data/slave/mavericks-debug/build/Source/WebCore/dom/ShadowRoot.h(134) : WebCore::ContainerNode *WebCore::Node::parentOrShadowHostNode() const 1 0x109b75830 WTFCrash 2 0x10e461995 WebCore::Node::parentOrShadowHostNode() const 3 0x10eff6e42 WebCore::root(WebCore::Node*) 4 0x10f1e8375 WebCore::root(WebCore::Node&) 5 0x10f35c9f9 WebCore::JSNode::visitAdditionalChildren(JSC::SlotVisitor&) 6 0x10f35af76 WebCore::JSNode::visitChildren(JSC::JSCell*, JSC::SlotVisitor&) 7 0x109a94688 JSC::visitChildren(JSC::SlotVisitor&, JSC::JSCell const*) 8 0x109a94558 JSC::SlotVisitor::drain() 9 0x109a94acc JSC::SlotVisitor::drainFromShared(JSC::SlotVisitor::SharedDrainMode) 10 0x109638e58 JSC::Heap::markRoots(double, void*, void*, int (&) [37])::$_0::operator()() const 11 0x109638c5c WTF::SharedTaskFunctor<JSC::Heap::markRoots(double, void*, void*, int (&) [37])::$_0>::run() 12 0x109bc3497 WTF::ParallelHelperClient::runTask(WTF::RefPtr<WTF::SharedTask>) 13 0x109bc3c52 WTF::ParallelHelperPool::helperThreadBody() 14 0x109bc4ca8 WTF::ParallelHelperPool::didMakeWorkAvailable(WTF::Locker<WTF::LockBase> const&)::$_0::operator()() const 15 0x109bc4c7c std::__1::__function::__func<WTF::ParallelHelperPool::didMakeWorkAvailable(WTF::Locker<WTF::LockBase> const&)::$_0, std::__1::allocator<WTF::ParallelHelperPool::didMakeWorkAvailable(WTF::Locker<WTF::LockBase> const&)::$_0>, void ()>::operator()() 16 0x10963b7da std::__1::function<void ()>::operator()() const 17 0x109be1c2e WTF::threadEntryPoint(void*) 18 0x109be3548 WTF::wtfThreadEntryPoint(void*) 19 0x7fff8b58e899 _pthread_body 20 0x7fff8b58e72a _pthread_struct_init 21 0x7fff8b592fc9 thread_start
(In reply to comment #21) > Now lots of layout tests are crashing in Debug builds: > https://build.webkit.org/results/Apple%20Mavericks%20Debug%20WK1%20(Tests)/ > r190267%20(15885)/results.html > > e.g. > stderr: > ASSERTION FAILED: isMainThreadOrGCThread() > /Volumes/Data/slave/mavericks-debug/build/Source/WebCore/dom/ShadowRoot. > h(134) : WebCore::ContainerNode *WebCore::Node::parentOrShadowHostNode() > const > 1 0x109b75830 WTFCrash > 2 0x10e461995 WebCore::Node::parentOrShadowHostNode() const > 3 0x10eff6e42 WebCore::root(WebCore::Node*) > 4 0x10f1e8375 WebCore::root(WebCore::Node&) > 5 0x10f35c9f9 WebCore::JSNode::visitAdditionalChildren(JSC::SlotVisitor&) > 6 0x10f35af76 WebCore::JSNode::visitChildren(JSC::JSCell*, > JSC::SlotVisitor&) > 7 0x109a94688 JSC::visitChildren(JSC::SlotVisitor&, JSC::JSCell const*) > 8 0x109a94558 JSC::SlotVisitor::drain() > 9 0x109a94acc > JSC::SlotVisitor::drainFromShared(JSC::SlotVisitor::SharedDrainMode) > 10 0x109638e58 JSC::Heap::markRoots(double, void*, void*, int (&) > [37])::$_0::operator()() const > 11 0x109638c5c WTF::SharedTaskFunctor<JSC::Heap::markRoots(double, void*, > void*, int (&) [37])::$_0>::run() > 12 0x109bc3497 > WTF::ParallelHelperClient::runTask(WTF::RefPtr<WTF::SharedTask>) > 13 0x109bc3c52 WTF::ParallelHelperPool::helperThreadBody() > 14 0x109bc4ca8 > WTF::ParallelHelperPool::didMakeWorkAvailable(WTF::Locker<WTF::LockBase> > const&)::$_0::operator()() const > 15 0x109bc4c7c > std::__1::__function::__func<WTF::ParallelHelperPool:: > didMakeWorkAvailable(WTF::Locker<WTF::LockBase> const&)::$_0, > std::__1::allocator<WTF::ParallelHelperPool::didMakeWorkAvailable(WTF:: > Locker<WTF::LockBase> const&)::$_0>, void ()>::operator()() > 16 0x10963b7da std::__1::function<void ()>::operator()() const > 17 0x109be1c2e WTF::threadEntryPoint(void*) > 18 0x109be3548 WTF::wtfThreadEntryPoint(void*) > 19 0x7fff8b58e899 _pthread_body > 20 0x7fff8b58e72a _pthread_struct_init > 21 0x7fff8b592fc9 thread_start Looking...
(In reply to comment #22) > (In reply to comment #21) > > Now lots of layout tests are crashing in Debug builds: > > https://build.webkit.org/results/Apple%20Mavericks%20Debug%20WK1%20(Tests)/ > > r190267%20(15885)/results.html > > > > e.g. > > stderr: > > ASSERTION FAILED: isMainThreadOrGCThread() > > /Volumes/Data/slave/mavericks-debug/build/Source/WebCore/dom/ShadowRoot. > > h(134) : WebCore::ContainerNode *WebCore::Node::parentOrShadowHostNode() > > const > > 1 0x109b75830 WTFCrash > > 2 0x10e461995 WebCore::Node::parentOrShadowHostNode() const > > 3 0x10eff6e42 WebCore::root(WebCore::Node*) > > 4 0x10f1e8375 WebCore::root(WebCore::Node&) > > 5 0x10f35c9f9 WebCore::JSNode::visitAdditionalChildren(JSC::SlotVisitor&) > > 6 0x10f35af76 WebCore::JSNode::visitChildren(JSC::JSCell*, > > JSC::SlotVisitor&) > > 7 0x109a94688 JSC::visitChildren(JSC::SlotVisitor&, JSC::JSCell const*) > > 8 0x109a94558 JSC::SlotVisitor::drain() > > 9 0x109a94acc > > JSC::SlotVisitor::drainFromShared(JSC::SlotVisitor::SharedDrainMode) > > 10 0x109638e58 JSC::Heap::markRoots(double, void*, void*, int (&) > > [37])::$_0::operator()() const > > 11 0x109638c5c WTF::SharedTaskFunctor<JSC::Heap::markRoots(double, void*, > > void*, int (&) [37])::$_0>::run() > > 12 0x109bc3497 > > WTF::ParallelHelperClient::runTask(WTF::RefPtr<WTF::SharedTask>) > > 13 0x109bc3c52 WTF::ParallelHelperPool::helperThreadBody() > > 14 0x109bc4ca8 > > WTF::ParallelHelperPool::didMakeWorkAvailable(WTF::Locker<WTF::LockBase> > > const&)::$_0::operator()() const > > 15 0x109bc4c7c > > std::__1::__function::__func<WTF::ParallelHelperPool:: > > didMakeWorkAvailable(WTF::Locker<WTF::LockBase> const&)::$_0, > > std::__1::allocator<WTF::ParallelHelperPool::didMakeWorkAvailable(WTF:: > > Locker<WTF::LockBase> const&)::$_0>, void ()>::operator()() > > 16 0x10963b7da std::__1::function<void ()>::operator()() const > > 17 0x109be1c2e WTF::threadEntryPoint(void*) > > 18 0x109be3548 WTF::wtfThreadEntryPoint(void*) > > 19 0x7fff8b58e899 _pthread_body > > 20 0x7fff8b58e72a _pthread_struct_init > > 21 0x7fff8b592fc9 thread_start > > Looking... Fix on the way.
(In reply to comment #23) > (In reply to comment #22) > > (In reply to comment #21) > > > Now lots of layout tests are crashing in Debug builds: > > > https://build.webkit.org/results/Apple%20Mavericks%20Debug%20WK1%20(Tests)/ > > > r190267%20(15885)/results.html > > > > > > e.g. > > > stderr: > > > ASSERTION FAILED: isMainThreadOrGCThread() > > > /Volumes/Data/slave/mavericks-debug/build/Source/WebCore/dom/ShadowRoot. > > > h(134) : WebCore::ContainerNode *WebCore::Node::parentOrShadowHostNode() > > > const > > > 1 0x109b75830 WTFCrash > > > 2 0x10e461995 WebCore::Node::parentOrShadowHostNode() const > > > 3 0x10eff6e42 WebCore::root(WebCore::Node*) > > > 4 0x10f1e8375 WebCore::root(WebCore::Node&) > > > 5 0x10f35c9f9 WebCore::JSNode::visitAdditionalChildren(JSC::SlotVisitor&) > > > 6 0x10f35af76 WebCore::JSNode::visitChildren(JSC::JSCell*, > > > JSC::SlotVisitor&) > > > 7 0x109a94688 JSC::visitChildren(JSC::SlotVisitor&, JSC::JSCell const*) > > > 8 0x109a94558 JSC::SlotVisitor::drain() > > > 9 0x109a94acc > > > JSC::SlotVisitor::drainFromShared(JSC::SlotVisitor::SharedDrainMode) > > > 10 0x109638e58 JSC::Heap::markRoots(double, void*, void*, int (&) > > > [37])::$_0::operator()() const > > > 11 0x109638c5c WTF::SharedTaskFunctor<JSC::Heap::markRoots(double, void*, > > > void*, int (&) [37])::$_0>::run() > > > 12 0x109bc3497 > > > WTF::ParallelHelperClient::runTask(WTF::RefPtr<WTF::SharedTask>) > > > 13 0x109bc3c52 WTF::ParallelHelperPool::helperThreadBody() > > > 14 0x109bc4ca8 > > > WTF::ParallelHelperPool::didMakeWorkAvailable(WTF::Locker<WTF::LockBase> > > > const&)::$_0::operator()() const > > > 15 0x109bc4c7c > > > std::__1::__function::__func<WTF::ParallelHelperPool:: > > > didMakeWorkAvailable(WTF::Locker<WTF::LockBase> const&)::$_0, > > > std::__1::allocator<WTF::ParallelHelperPool::didMakeWorkAvailable(WTF:: > > > Locker<WTF::LockBase> const&)::$_0>, void ()>::operator()() > > > 16 0x10963b7da std::__1::function<void ()>::operator()() const > > > 17 0x109be1c2e WTF::threadEntryPoint(void*) > > > 18 0x109be3548 WTF::wtfThreadEntryPoint(void*) > > > 19 0x7fff8b58e899 _pthread_body > > > 20 0x7fff8b58e72a _pthread_struct_init > > > 21 0x7fff8b592fc9 thread_start > > > > Looking... > > Fix on the way. Here is a fairly principled fix: https://bugs.webkit.org/show_bug.cgi?id=149582 It's got enough code that it should probably be reviewed. If I don't see a review soon, I will land a simpler fix that just calls WTF::registerGCThread().
(In reply to comment #24) > (In reply to comment #23) > > (In reply to comment #22) > > > (In reply to comment #21) > > > > Now lots of layout tests are crashing in Debug builds: > > > > https://build.webkit.org/results/Apple%20Mavericks%20Debug%20WK1%20(Tests)/ > > > > r190267%20(15885)/results.html > > > > > > > > e.g. > > > > stderr: > > > > ASSERTION FAILED: isMainThreadOrGCThread() > > > > /Volumes/Data/slave/mavericks-debug/build/Source/WebCore/dom/ShadowRoot. > > > > h(134) : WebCore::ContainerNode *WebCore::Node::parentOrShadowHostNode() > > > > const > > > > 1 0x109b75830 WTFCrash > > > > 2 0x10e461995 WebCore::Node::parentOrShadowHostNode() const > > > > 3 0x10eff6e42 WebCore::root(WebCore::Node*) > > > > 4 0x10f1e8375 WebCore::root(WebCore::Node&) > > > > 5 0x10f35c9f9 WebCore::JSNode::visitAdditionalChildren(JSC::SlotVisitor&) > > > > 6 0x10f35af76 WebCore::JSNode::visitChildren(JSC::JSCell*, > > > > JSC::SlotVisitor&) > > > > 7 0x109a94688 JSC::visitChildren(JSC::SlotVisitor&, JSC::JSCell const*) > > > > 8 0x109a94558 JSC::SlotVisitor::drain() > > > > 9 0x109a94acc > > > > JSC::SlotVisitor::drainFromShared(JSC::SlotVisitor::SharedDrainMode) > > > > 10 0x109638e58 JSC::Heap::markRoots(double, void*, void*, int (&) > > > > [37])::$_0::operator()() const > > > > 11 0x109638c5c WTF::SharedTaskFunctor<JSC::Heap::markRoots(double, void*, > > > > void*, int (&) [37])::$_0>::run() > > > > 12 0x109bc3497 > > > > WTF::ParallelHelperClient::runTask(WTF::RefPtr<WTF::SharedTask>) > > > > 13 0x109bc3c52 WTF::ParallelHelperPool::helperThreadBody() > > > > 14 0x109bc4ca8 > > > > WTF::ParallelHelperPool::didMakeWorkAvailable(WTF::Locker<WTF::LockBase> > > > > const&)::$_0::operator()() const > > > > 15 0x109bc4c7c > > > > std::__1::__function::__func<WTF::ParallelHelperPool:: > > > > didMakeWorkAvailable(WTF::Locker<WTF::LockBase> const&)::$_0, > > > > std::__1::allocator<WTF::ParallelHelperPool::didMakeWorkAvailable(WTF:: > > > > Locker<WTF::LockBase> const&)::$_0>, void ()>::operator()() > > > > 16 0x10963b7da std::__1::function<void ()>::operator()() const > > > > 17 0x109be1c2e WTF::threadEntryPoint(void*) > > > > 18 0x109be3548 WTF::wtfThreadEntryPoint(void*) > > > > 19 0x7fff8b58e899 _pthread_body > > > > 20 0x7fff8b58e72a _pthread_struct_init > > > > 21 0x7fff8b592fc9 thread_start > > > > > > Looking... > > > > Fix on the way. > > Here is a fairly principled fix: > https://bugs.webkit.org/show_bug.cgi?id=149582 > > It's got enough code that it should probably be reviewed. If I don't see a > review soon, I will land a simpler fix that just calls > WTF::registerGCThread(). Landed the simpler fix in http://trac.webkit.org/changeset/190269