Increase the GC allocation trigger
Created attachment 142402 [details] Patch
Benchmark report for SunSpider, V8, V8Real, Kraken, JSBench, JSRegress, and DSP on garen (MacPro5,1). VMs tested: "BASE" at /Volumes/Big/ggaren/baseline-webkit/WebKitBuild/Release/DumpRenderTree (r117340) "PATCH" at /Volumes/Big/ggaren/webkit/WebKitBuild/Release/DumpRenderTree (r117340) 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. BASE PATCH SunSpider: 3d-cube 9.1433+-0.4905 8.9365+-0.1552 might be 1.0231x faster 3d-morph 8.5952+-0.0687 8.5577+-0.0824 3d-raytrace 11.8690+-0.5976 ? 11.8790+-0.5670 ? access-binary-trees 2.1468+-0.0370 2.1260+-0.0232 access-fannkuch 8.7502+-0.0888 8.7431+-0.0420 access-nbody 4.5497+-0.0791 4.5395+-0.0787 access-nsieve 4.2917+-0.0591 ? 4.2952+-0.0339 ? bitops-3bit-bits-in-byte 1.6609+-0.0192 1.6551+-0.0166 bitops-bits-in-byte 6.3039+-0.0766 ? 6.3764+-0.0931 ? might be 1.0115x slower bitops-bitwise-and 3.9581+-0.0156 ? 3.9739+-0.0226 ? bitops-nsieve-bits 3.8389+-0.0169 3.8167+-0.0199 controlflow-recursive 2.8414+-0.0257 2.8394+-0.0301 crypto-aes 9.2146+-0.1980 9.1801+-0.1905 crypto-md5 3.9568+-0.1015 ? 3.9606+-0.1122 ? crypto-sha1 3.2399+-0.0602 3.2295+-0.0660 date-format-tofte 15.2260+-1.4562 ? 15.3482+-1.5438 ? date-format-xparb 13.4378+-1.0855 13.2417+-1.1211 might be 1.0148x faster math-cordic 5.0385+-0.0985 ? 5.0475+-0.1028 ? math-partial-sums 10.5282+-0.0396 ? 10.5295+-0.0592 ? math-spectral-norm 3.4076+-0.0532 3.4040+-0.0553 regexp-dna 11.3938+-0.1030 ? 11.4267+-0.1354 ? string-base64 6.5444+-0.8145 6.5410+-0.8427 string-fasta 9.4850+-0.4746 9.4188+-0.4454 string-tagcloud 15.5875+-0.4166 ? 15.6230+-0.4387 ? string-unpack-code 25.5526+-1.0364 ? 25.6402+-1.0722 ? string-validate-input 9.5765+-0.7677 9.5076+-0.7106 <arithmetic> * 8.0823+-0.1290 8.0706+-0.1258 might be 1.0014x faster <geometric> 6.5586+-0.0476 6.5459+-0.0438 might be 1.0019x faster <harmonic> 5.2718+-0.0171 5.2596+-0.0126 might be 1.0023x faster BASE PATCH V8: crypto 86.9721+-0.8849 ? 87.1733+-0.7522 ? deltablue 179.1931+-3.0519 178.8170+-2.0419 earley-boyer 106.9763+-0.4241 ? 107.2566+-0.4439 ? raytrace 62.3463+-0.2830 ? 62.8009+-0.6932 ? regexp 107.1802+-0.7016 106.9179+-0.6501 richards 164.3974+-2.9438 ? 165.2053+-4.3568 ? splay 142.9082+-11.7480 113.4285+-27.2097 might be 1.2599x faster <arithmetic> 121.4248+-1.4485 117.3713+-3.5695 might be 1.0345x faster <geometric> * 114.6020+-1.1664 110.3548+-3.2700 might be 1.0385x faster <harmonic> 107.6982+-0.8101 ^ 103.7235+-2.7112 ^ definitely 1.0383x faster BASE PATCH V8Real: encrypt 0.49067+-0.00065 0.49049+-0.00065 decrypt 8.56374+-0.01154 8.55829+-0.00969 deltablue x2 1.07967+-0.01029 1.07840+-0.01058 earley 3.51312+-0.04181 ? 3.51875+-0.03536 ? boyer 18.09254+-0.10827 ? 18.09705+-0.08665 ? raytrace x2 8.12321+-0.12322 8.05406+-0.11584 regexp x2 31.11167+-0.21164 ? 31.21755+-0.23637 ? richards x2 0.43860+-0.00481 ? 0.43888+-0.00596 ? splay x2 1.07533+-0.01928 ^ 0.98978+-0.05197 ^ definitely 1.0864x faster <arithmetic> 8.16550+-0.03602 8.15871+-0.03136 might be 1.0008x faster <geometric> * 2.98293+-0.01280 2.94464+-0.02653 might be 1.0130x faster <harmonic> 1.26357+-0.00773 1.24439+-0.01591 might be 1.0154x faster BASE PATCH Kraken: ai-astar 946.016+-1.350 945.557+-2.629 audio-beat-detection 233.023+-2.256 232.035+-0.761 audio-dft 337.897+-0.934 ? 338.207+-0.864 ? audio-fft 137.814+-0.248 137.763+-0.247 audio-oscillator 370.306+-0.744 ? 370.369+-0.866 ? imaging-darkroom 328.158+-1.623 ? 328.457+-1.987 ? imaging-desaturate 252.217+-0.090 252.205+-0.090 imaging-gaussian-blur 525.323+-0.486 ? 526.589+-2.699 ? json-parse-financial 75.900+-0.108 ! 76.420+-0.244 ! definitely 1.0068x slower json-stringify-tinderbox 96.423+-0.294 ? 96.467+-0.517 ? stanford-crypto-aes 101.358+-0.532 ? 102.150+-0.950 ? stanford-crypto-ccm 109.494+-0.657 109.281+-0.482 stanford-crypto-pbkdf2 220.479+-0.506 ? 222.351+-1.539 ? stanford-crypto-sha256-iterative 113.058+-0.356 ^ 112.263+-0.300 ^ definitely 1.0071x faster <arithmetic> * 274.819+-0.346 ? 275.008+-0.503 ? might be 1.0007x slower <geometric> 210.463+-0.355 ? 210.669+-0.424 ? might be 1.0010x slower <harmonic> 168.455+-0.310 ? 168.691+-0.383 ? might be 1.0014x slower BASE PATCH JSBench: amazon 20.3333+-0.3128 ? 20.5833+-0.3272 ? might be 1.0123x slower facebook 77.7500+-1.8989 ? 78.7500+-2.1007 ? might be 1.0129x slower google 112.6667+-4.3713 112.5000+-4.4641 twitter 59.4167+-0.3272 59.0833+-0.4248 yahoo 25.7500+-0.2874 25.5000+-0.3318 <arithmetic> * 59.1833+-0.8541 ? 59.2833+-0.7339 ? might be 1.0017x slower <geometric> 48.6264+-0.4101 ? 48.7021+-0.3828 ? might be 1.0016x slower <harmonic> 39.4776+-0.2743 ? 39.5626+-0.3726 ? might be 1.0022x slower BASE PATCH JSRegress: adapt-to-double-divide 83.4064+-0.0418 ? 83.4120+-0.1016 ? aliased-arguments-getbyval 4.5681+-0.3241 ? 4.5745+-0.3354 ? arity-mismatch-inlining 1.4313+-0.0429 ? 1.4705+-0.0323 ? might be 1.0274x slower big-int-mul 33.3517+-0.5335 33.0791+-0.7708 boolean-test 4.5366+-0.0363 ? 4.5409+-0.0291 ? cast-int-to-double 16.2892+-0.0439 16.2604+-0.0306 cfg-simplify 7.5949+-0.0250 ? 7.5963+-0.0305 ? cmpeq-obj-to-obj-other 16.5950+-0.1553 ? 16.6248+-0.2072 ? constant-test 31.1825+-0.0948 ? 31.3069+-0.1127 ? direct-arguments-getbyval 0.8214+-0.0107 0.8160+-0.0184 double-pollution-getbyval 10.0719+-0.0171 10.0713+-0.0327 double-pollution-putbyoffset 5.4028+-0.0586 ? 5.4291+-0.0534 ? external-arguments-getbyval 5.3135+-0.3631 5.2594+-0.3364 might be 1.0103x faster external-arguments-putbyval 8.2696+-0.7352 8.2433+-0.7326 Float32Array-matrix-mult 13.9366+-1.1439 ? 14.1165+-1.1598 ? might be 1.0129x slower fold-double-to-int 40.1503+-1.3529 38.9903+-0.4293 might be 1.0298x faster function-test 5.3080+-0.0579 ^ 5.2223+-0.0204 ^ definitely 1.0164x faster inline-arguments-access 4.1445+-0.0268 4.1294+-0.0282 inline-arguments-local-escape 45.2946+-3.3097 ? 45.4430+-3.3960 ? int-overflow-local 117.9510+-0.1922 ? 118.1308+-0.3642 ? Int16Array-bubble-sort 79.4776+-2.3423 77.9072+-0.0882 might be 1.0202x faster Int16Array-load-int-mul 18.3273+-0.2299 18.1800+-0.1558 Int8Array-load 5.4426+-0.0619 ? 5.4678+-0.0984 ? integer-divide 17.0533+-0.0575 17.0064+-0.0308 method-on-number 218.4744+-2.6994 ? 218.5951+-3.5892 ? number-test 4.5220+-0.0305 4.4960+-0.0169 object-test 4.8691+-0.0329 4.8568+-0.0140 poly-stricteq 104.0616+-0.3097 ? 104.2639+-0.6984 ? rare-osr-exit-on-local 173.2848+-0.1687 ? 173.5993+-0.3586 ? simple-activation-demo 63.3235+-0.1624 63.1933+-0.1512 slow-convergence 103.9075+-0.5227 103.8209+-0.5834 string-hash 16.6754+-0.1885 ? 16.6772+-0.3758 ? string-test 4.2973+-0.0283 ? 4.3104+-0.0237 ? tear-off-arguments 4.6381+-0.3100 ? 4.6461+-0.3280 ? to-int32-boolean 33.4488+-0.1210 33.4030+-0.1476 undefined-test 4.8335+-0.0180 ? 4.8738+-0.0257 ? <arithmetic> 36.4516+-0.1872 36.3893+-0.1389 might be 1.0017x faster <geometric> * 15.0914+-0.1538 15.0737+-0.1472 might be 1.0012x faster <harmonic> 6.7012+-0.0439 ? 6.7092+-0.0440 ? might be 1.0012x slower BASE PATCH DSP: filtrr-posterize-tint 56.1393+-0.4359 56.0947+-0.3881 filtrr-tint-contrast-sat-bright 93.6196+-1.3939 93.5761+-1.3467 filtrr-tint-sat-adj-contr-mult 108.3652+-0.5375 ? 109.1542+-0.7810 ? filtrr-blur-overlay-sat-contr 269.4519+-2.9625 269.0424+-3.0036 filtrr-sat-blur-mult-sharpen-contr 334.9958+-5.5921 ? 336.9401+-5.7039 ? filtrr-sepia-bias 38.9325+-0.6758 ? 39.1924+-0.5278 ? route9-vp8 x5 1896.2909+-17.2587 1882.3232+-12.4624 <arithmetic> 943.9054+-7.7502 937.7833+-5.7899 might be 1.0065x faster <geometric> * 405.7016+-1.6824 404.9729+-1.3584 might be 1.0018x faster <harmonic> 151.1784+-1.1034 ? 151.6273+-0.9580 ? might be 1.0030x slower BASE PATCH All benchmarks: <arithmetic> 150.5578+-0.7351 149.7152+-0.6171 might be 1.0056x faster <geometric> 23.2367+-0.1062 23.1252+-0.0807 might be 1.0048x faster <harmonic> 5.1996+-0.0153 5.1571+-0.0348 might be 1.0082x faster BASE PATCH Geomean of preferred means: <scaled-result> 43.0705+-0.1715 ^ 42.7417+-0.1112 ^ definitely 1.0077x faster
Comment on attachment 142402 [details] Patch Sam Weinig reviewed this.
Committed r117496: <http://trac.webkit.org/changeset/117496>