Patch forthcoming.
Created attachment 199896 [details] work in progress
Created attachment 199897 [details] the patch
Comment on attachment 199897 [details] the patch Attachment 199897 [details] did not pass qt-wk2-ews (qt-wk2): Output: http://webkit-queues.appspot.com/results/56828
Comment on attachment 199897 [details] the patch Attachment 199897 [details] did not pass qt-ews (qt): Output: http://webkit-queues.appspot.com/results/173248
Comment on attachment 199897 [details] the patch Attachment 199897 [details] did not pass efl-ews (efl): Output: http://webkit-queues.appspot.com/results/183476
Comment on attachment 199897 [details] the patch Attachment 199897 [details] did not pass efl-wk2-ews (efl-wk2): Output: http://webkit-queues.appspot.com/results/239024
Comment on attachment 199897 [details] the patch Attachment 199897 [details] did not pass mac-wk2-ews (mac-wk2): Output: http://webkit-queues.appspot.com/results/58697
Comment on attachment 199897 [details] the patch Attachment 199897 [details] did not pass win-ews (win): Output: http://webkit-queues.appspot.com/results/37478
Comment on attachment 199897 [details] the patch Attachment 199897 [details] did not pass mac-ews (mac): Output: http://webkit-queues.appspot.com/results/51283
Comment on attachment 199897 [details] the patch Attachment 199897 [details] did not pass gtk-ews (gtk): Output: http://webkit-queues.appspot.com/results/153472
A near flawless victory of EWS destruction! ;-) Too bad this patch is for the FTL branch.
As expected, this has neutral performance: Benchmark report for SunSpider, V8Spider, Octane, Kraken, and JSRegress on oldmac (MacPro4,1). VMs tested: "TipOfTree" at /Volumes/Data/pizlo/fourthTier/OpenSource/WebKitBuild/Release/jsc (r149236) "LockedSymTab" at /Volumes/Data/fromMiniMe/fourthTier/secondary/OpenSource/WebKitBuild/Release/jsc (r149236) 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. TipOfTree LockedSymTab SunSpider: 3d-cube 9.4853+-0.1687 9.4708+-0.1534 3d-morph 8.8450+-0.1732 ? 8.8513+-0.1364 ? 3d-raytrace 10.5321+-0.1532 ? 10.5752+-0.1833 ? access-binary-trees 1.8999+-0.0079 ? 1.9032+-0.0063 ? access-fannkuch 7.8547+-0.1363 7.8521+-0.1087 access-nbody 4.7085+-0.0093 4.6971+-0.0403 access-nsieve 4.9781+-0.0629 4.9242+-0.0830 might be 1.0110x faster bitops-3bit-bits-in-byte 1.8504+-0.0099 ? 1.8538+-0.0110 ? bitops-bits-in-byte 7.1578+-0.1013 7.1349+-0.1033 bitops-bitwise-and 2.8015+-0.0589 2.7942+-0.0783 bitops-nsieve-bits 4.7981+-0.0170 4.7400+-0.0833 might be 1.0122x faster controlflow-recursive 3.1646+-0.0494 3.1490+-0.0484 crypto-aes 7.8315+-0.1485 ? 7.9136+-0.1312 ? might be 1.0105x slower crypto-md5 4.2016+-0.0411 ? 4.2054+-0.0336 ? crypto-sha1 3.3061+-0.0154 ? 3.3242+-0.0264 ? date-format-tofte 15.3321+-0.2272 15.1823+-0.1580 date-format-xparb 9.4408+-0.1932 ? 9.8244+-0.3991 ? might be 1.0406x slower math-cordic 4.0593+-0.0104 ? 4.0606+-0.0073 ? math-partial-sums 12.5430+-0.1325 ? 12.6080+-0.1354 ? math-spectral-norm 3.1746+-0.0110 ? 3.1756+-0.0128 ? regexp-dna 12.6929+-0.1829 ? 12.9468+-0.1918 ? might be 1.0200x slower string-base64 5.1346+-0.0648 5.0202+-0.1020 might be 1.0228x faster string-fasta 10.8201+-0.1367 ? 10.8693+-0.1156 ? string-tagcloud 14.7043+-0.4384 14.2581+-0.2455 might be 1.0313x faster string-unpack-code 28.2888+-0.1585 ? 28.3793+-0.1470 ? string-validate-input 7.4476+-0.1724 7.2448+-0.1799 might be 1.0280x faster <arithmetic> * 7.9636+-0.0633 7.9599+-0.0661 might be 1.0005x faster <geometric> 6.4131+-0.0391 6.4053+-0.0456 might be 1.0012x faster <harmonic> 5.1716+-0.0223 5.1638+-0.0278 might be 1.0015x faster TipOfTree LockedSymTab V8Spider: crypto 87.4193+-0.2684 ? 87.8282+-0.2366 ? deltablue 132.0028+-0.4315 ? 132.2209+-0.8015 ? earley-boyer 84.0875+-0.4977 83.3908+-0.3244 raytrace 62.1315+-0.2003 ? 62.4411+-0.2595 ? regexp 102.8483+-0.7213 102.5691+-0.6543 richards 118.9806+-0.6528 ? 119.8726+-0.5321 ? splay 49.1056+-0.3912 48.9985+-0.2025 <arithmetic> 90.9394+-0.1446 ? 91.0459+-0.2269 ? might be 1.0012x slower <geometric> * 86.4951+-0.1148 ? 86.5646+-0.1915 ? might be 1.0008x slower <harmonic> 81.8715+-0.1351 ? 81.9102+-0.1678 ? might be 1.0005x slower TipOfTree LockedSymTab Octane and V8v7: encrypt 0.47012+-0.00122 ? 0.47013+-0.00098 ? decrypt 8.70860+-0.02957 8.68786+-0.02463 deltablue x2 0.61039+-0.00436 0.60705+-0.00092 earley 0.91475+-0.00305 ? 0.91992+-0.01022 ? boyer 12.85130+-0.04347 ? 12.85830+-0.04491 ? raytrace x2 4.42646+-0.00697 ? 4.43799+-0.00919 ? regexp x2 32.50308+-0.18186 ? 32.82752+-0.19594 ? richards x2 0.30974+-0.00053 0.30908+-0.00072 splay x2 0.66233+-0.03966 0.66164+-0.03646 navier-stokes x2 10.83210+-0.03376 10.80770+-0.02305 closure 0.30775+-0.03515 ? 0.30954+-0.03500 ? jquery 4.42657+-0.55281 ? 4.42716+-0.54569 ? gbemu x2 251.55537+-17.48037 251.45155+-17.33567 box2d x2 31.71124+-0.11451 ? 31.71913+-0.08318 ? V8v7: <arithmetic> 7.60206+-0.02771 ? 7.63989+-0.02754 ? might be 1.0050x slower <geometric> * 2.46183+-0.01854 ? 2.46305+-0.01721 ? might be 1.0005x slower <harmonic> 0.94767+-0.00964 0.94625+-0.00881 might be 1.0015x faster Octane including V8v7: <arithmetic> 31.49548+-1.59774 ? 31.51437+-1.58646 ? might be 1.0006x slower <geometric> * 4.41154+-0.06019 ? 4.41442+-0.05873 ? might be 1.0007x slower <harmonic> 1.07220+-0.01789 1.07200+-0.01732 might be 1.0002x faster TipOfTree LockedSymTab Kraken: ai-astar 493.343+-0.742 ? 494.385+-1.354 ? audio-beat-detection 245.441+-0.939 244.624+-1.173 audio-dft 323.669+-1.803 323.412+-2.170 audio-fft 143.650+-0.405 143.629+-0.387 audio-oscillator 234.800+-0.983 ? 235.006+-1.184 ? imaging-darkroom 294.261+-1.558 292.652+-1.150 imaging-desaturate 160.446+-0.280 ? 160.465+-0.479 ? imaging-gaussian-blur 397.865+-0.406 ! 399.320+-0.900 ! definitely 1.0037x slower json-parse-financial 80.970+-0.222 80.818+-0.159 json-stringify-tinderbox 101.720+-0.414 101.709+-0.316 stanford-crypto-aes 238.505+-1.177 236.723+-1.854 stanford-crypto-ccm 110.943+-0.344 ? 111.361+-0.210 ? stanford-crypto-pbkdf2 274.398+-1.227 ? 276.021+-1.161 ? stanford-crypto-sha256-iterative 115.787+-0.181 ! 116.302+-0.271 ! definitely 1.0044x slower <arithmetic> * 229.700+-0.245 ? 229.745+-0.426 ? might be 1.0002x slower <geometric> 200.687+-0.215 ? 200.710+-0.352 ? might be 1.0001x slower <harmonic> 174.507+-0.186 ? 174.550+-0.265 ? might be 1.0002x slower TipOfTree LockedSymTab JSRegress: adapt-to-double-divide 22.6772+-0.1209 22.5509+-0.1286 aliased-arguments-getbyval 0.9256+-0.0106 0.9221+-0.0097 allocate-big-object 2.5439+-0.0353 ? 2.5527+-0.0388 ? arity-mismatch-inlining 0.7839+-0.0128 0.7770+-0.0156 array-access-polymorphic-structure 7.1247+-0.1087 7.0338+-0.1338 might be 1.0129x faster array-with-double-add 5.8482+-0.0954 ? 5.8645+-0.0584 ? array-with-double-increment 4.2005+-0.0894 4.1053+-0.0103 might be 1.0232x faster array-with-double-mul-add 7.0661+-0.0948 ? 7.1388+-0.0993 ? might be 1.0103x slower array-with-double-sum 7.9927+-0.0909 7.8818+-0.1171 might be 1.0141x faster array-with-int32-add-sub 10.5986+-0.1282 10.4619+-0.1108 might be 1.0131x faster array-with-int32-or-double-sum 8.1077+-0.0851 8.0732+-0.0899 big-int-mul 4.9740+-0.0615 ? 4.9908+-0.0487 ? boolean-test 4.4304+-0.0610 ? 4.4344+-0.0619 ? cast-int-to-double 14.0254+-0.1403 ? 14.0896+-0.1509 ? cell-argument 14.5879+-0.1147 14.4835+-0.0913 cfg-simplify 4.0328+-0.0845 4.0114+-0.0317 cmpeq-obj-to-obj-other 11.5532+-0.1712 ^ 11.0827+-0.1869 ^ definitely 1.0425x faster constant-test 8.5107+-0.1471 ? 8.5506+-0.1349 ? direct-arguments-getbyval 0.8496+-0.0109 ? 0.8500+-0.0094 ? double-pollution-getbyval 10.7973+-0.1073 ? 10.8190+-0.1513 ? double-pollution-putbyoffset 5.0945+-0.0896 5.0583+-0.0340 empty-string-plus-int 11.0503+-0.2076 ? 11.0643+-0.2151 ? external-arguments-getbyval 2.2301+-0.0334 ? 2.2312+-0.0322 ? external-arguments-putbyval 3.2857+-0.0201 ? 3.3349+-0.0764 ? might be 1.0150x slower Float32Array-matrix-mult 14.6364+-0.1432 ^ 14.3423+-0.1131 ^ definitely 1.0205x faster fold-double-to-int 22.3196+-0.2973 22.1905+-0.2795 function-dot-apply 3.1942+-0.0105 ? 3.1969+-0.0082 ? function-test 5.0442+-0.0747 ? 5.0656+-0.0415 ? get-by-id-chain-from-try-block 7.4422+-0.0989 7.3854+-0.0906 HashMap-put-get-iterate-keys 96.5506+-1.2191 ? 97.1141+-0.6602 ? HashMap-put-get-iterate 98.4338+-0.7239 ? 99.0688+-0.8952 ? HashMap-string-put-get-iterate 73.8042+-0.3793 ? 74.4451+-0.8219 ? indexed-properties-in-objects 4.5787+-0.0689 ? 4.5916+-0.0189 ? inline-arguments-access 1.2662+-0.0079 1.2643+-0.0074 inline-arguments-local-escape 23.3145+-0.1568 ? 23.4913+-0.1096 ? inline-get-scoped-var 6.5831+-0.1067 ? 6.6033+-0.1011 ? inlined-put-by-id-transition 16.7702+-0.2042 16.4514+-0.1639 might be 1.0194x faster int-or-other-abs-then-get-by-val 8.9707+-0.0925 8.8569+-0.1044 might be 1.0129x faster int-or-other-abs-zero-then-get-by-val 37.2793+-0.1553 ? 37.9880+-0.9653 ? might be 1.0190x slower int-or-other-add-then-get-by-val 10.3761+-0.1206 10.2648+-0.1026 might be 1.0108x faster int-or-other-add 10.5281+-0.1058 ? 10.5379+-0.1308 ? int-or-other-div-then-get-by-val 8.0005+-0.0959 ? 8.0417+-0.1075 ? int-or-other-max-then-get-by-val 10.0047+-0.2603 ? 10.1012+-0.1366 ? int-or-other-min-then-get-by-val 8.2734+-0.1158 8.1754+-0.0921 might be 1.0120x faster int-or-other-mod-then-get-by-val 8.1264+-0.0845 8.0642+-0.0857 int-or-other-mul-then-get-by-val 7.2610+-0.0979 ? 7.2645+-0.0973 ? int-or-other-neg-then-get-by-val 8.1614+-0.1402 8.0649+-0.0891 might be 1.0120x faster int-or-other-neg-zero-then-get-by-val 36.5605+-0.2255 36.5384+-0.1223 int-or-other-sub-then-get-by-val 10.2852+-0.1406 ? 10.3753+-0.1161 ? int-or-other-sub 8.3548+-0.0704 8.2582+-0.1099 might be 1.0117x faster int-overflow-local 12.9406+-0.0864 ? 13.0106+-0.0855 ? Int16Array-bubble-sort 49.5605+-0.1955 49.5395+-0.1528 Int16Array-load-int-mul 1.8958+-0.0062 ? 1.9040+-0.0081 ? Int8Array-load 4.8306+-0.0633 ? 4.8629+-0.0594 ? integer-divide 15.2532+-0.1342 15.1747+-0.1242 integer-modulo 2.1141+-0.0159 2.1138+-0.0158 make-indexed-storage 3.9144+-0.0368 3.9099+-0.0372 method-on-number 23.7926+-0.2728 ? 24.0895+-0.3485 ? might be 1.0125x slower nested-function-parsing-random 382.1898+-13.2124 381.5689+-13.1304 nested-function-parsing 47.7541+-1.1281 ? 47.8574+-1.0886 ? new-array-buffer-dead 3.6334+-0.0120 ? 3.6366+-0.0114 ? new-array-buffer-push 10.5347+-0.1709 ? 10.6120+-0.1056 ? new-array-dead 28.3748+-0.1203 28.3088+-0.0863 new-array-push 7.0678+-0.1366 7.0317+-0.1022 number-test 4.3368+-0.0582 ? 4.3399+-0.0536 ? object-closure-call 8.3207+-0.0937 ? 8.4598+-0.1156 ? might be 1.0167x slower object-test 4.8699+-0.0864 ? 4.9424+-0.0532 ? might be 1.0149x slower poly-stricteq 90.9276+-0.1412 ! 92.0319+-0.8518 ! definitely 1.0121x slower polymorphic-structure 20.1326+-0.0886 ? 20.1810+-0.1050 ? polyvariant-monomorphic-get-by-id 12.6289+-0.0950 12.6071+-0.1238 rare-osr-exit-on-local 20.8056+-0.1292 20.7675+-0.1310 register-pressure-from-osr 31.6120+-0.1071 ? 31.7152+-0.1346 ? simple-activation-demo 34.5070+-0.1194 ? 34.5641+-0.1351 ? slow-array-profile-convergence 4.3345+-0.0165 4.3195+-0.0142 slow-convergence 3.8316+-0.0110 3.8230+-0.0098 sparse-conditional 1.3315+-0.0174 ? 1.3420+-0.0094 ? splice-to-remove 50.4484+-0.2022 ? 51.0209+-0.5762 ? might be 1.0113x slower string-concat-object 2.7388+-0.0412 ? 2.7569+-0.0337 ? string-concat-pair-object 2.6782+-0.0143 ! 2.7395+-0.0464 ! definitely 1.0229x slower string-concat-pair-simple 17.3284+-0.2873 17.2926+-0.1884 string-concat-simple 17.3269+-0.2673 17.3249+-0.2672 string-cons-repeat 10.1490+-0.0308 ? 10.1797+-0.0453 ? string-cons-tower 10.9721+-0.0344 ? 11.0561+-0.0840 ? string-equality 106.0078+-1.1518 ? 106.4500+-1.6264 ? string-hash 2.6662+-0.0089 ? 2.6750+-0.0100 ? string-repeat-arith 47.0378+-0.2420 ? 47.3012+-0.3184 ? string-sub 90.0675+-0.4973 89.6221+-0.6720 string-test 4.3258+-0.0472 4.2769+-0.0302 might be 1.0114x faster structure-hoist-over-transitions 3.2836+-0.0259 ? 3.3721+-0.1236 ? might be 1.0270x slower tear-off-arguments-simple 1.7930+-0.0112 ? 1.8036+-0.0132 ? tear-off-arguments 3.4020+-0.0108 ? 3.4037+-0.0088 ? temporal-structure 20.9846+-0.1256 ? 21.1803+-0.4286 ? to-int32-boolean 30.7233+-0.1404 30.7000+-0.1225 undefined-test 4.5769+-0.0413 ? 4.5834+-0.0371 ? <arithmetic> 21.0363+-0.1585 ? 21.0698+-0.1436 ? might be 1.0016x slower <geometric> * 9.4827+-0.0265 ? 9.4852+-0.0225 ? might be 1.0003x slower <harmonic> 5.2306+-0.0181 ? 5.2325+-0.0156 ? might be 1.0003x slower TipOfTree LockedSymTab All benchmarks: <arithmetic> 41.2868+-0.3174 ? 41.3164+-0.3178 ? might be 1.0007x slower <geometric> 11.4830+-0.0498 ? 11.4841+-0.0493 ? might be 1.0001x slower <harmonic> 3.7309+-0.0326 3.7305+-0.0312 might be 1.0001x faster TipOfTree LockedSymTab Geomean of preferred means: <scaled-result> 23.1283+-0.1145 ? 23.1351+-0.1191 ? might be 1.0003x slower
The SymbolTable itself -- excluding Watchpoints -- is immutable after byte-compilation. It must be so in order to maintain lots of different invariants held by arguments objects, closures, and compiled code. It feels a little strange to go out of our way to make mutation thread-safe when it's not safe at all. Can take advantage of this by ASSERTing that mutation doesn't happen, instead of adding guards for thread-safe mutation?
(In reply to comment #13) > The SymbolTable itself -- excluding Watchpoints -- is immutable after byte-compilation. It must be so in order to maintain lots of different invariants held by arguments objects, closures, and compiled code. It feels a little strange to go out of our way to make mutation thread-safe when it's not safe at all. Really? What about a second script tag? > > Can take advantage of this by ASSERTing that mutation doesn't happen, instead of adding guards for thread-safe mutation?
Oops! I forgot about the global object's symbol table. That is mutable.
Comment on attachment 199897 [details] the patch r=me
Landed in http://trac.webkit.org/changeset/149248