RESOLVED FIXED 74984
Value Profiles for arguments should be more easily accessible to the interpreter
https://bugs.webkit.org/show_bug.cgi?id=74984
Summary Value Profiles for arguments should be more easily accessible to the interpreter
Filip Pizlo
Reported 2011-12-20 17:50:25 PST
Patch forthcoming.
Attachments
the patch (13.69 KB, patch)
2011-12-20 17:53 PST, Filip Pizlo
barraclough: review+
Filip Pizlo
Comment 1 2011-12-20 17:50:34 PST
Filip Pizlo
Comment 2 2011-12-20 17:53:39 PST
Created attachment 120127 [details] the patch Looks neutral. Benchmark report for SunSpider, V8, and Kraken on bigmac (MacPro5,1). VMs tested: "TipOfTree" at /Volumes/Data/pizlo/quinary/OpenSource/WebKitBuild/Release/jsc (r103374) "SepArgProfile" at /Volumes/Data/pizlo/OpenSource/WebKitBuild/Release/jsc (r103374) 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 SepArgProfile SunSpider: 3d-cube 5.8983+-0.0210 5.8960+-0.0245 3d-morph 8.3063+-0.0316 8.2932+-0.0350 3d-raytrace 7.5780+-0.0573 ? 7.6596+-0.1066 ? might be 1.0108x slower access-binary-trees 1.6330+-0.0072 1.6215+-0.0070 access-fannkuch 7.5015+-0.0082 ? 7.5078+-0.0119 ? access-nbody 3.9341+-0.0112 3.9320+-0.0136 access-nsieve 3.1339+-0.0492 ? 3.1361+-0.0546 ? bitops-3bit-bits-in-byte 1.2492+-0.0136 ? 1.2570+-0.0168 ? bitops-bits-in-byte 5.0888+-0.0257 ? 5.0891+-0.0241 ? bitops-bitwise-and 3.2908+-0.0042 3.2845+-0.0025 bitops-nsieve-bits 5.6564+-0.0360 5.6386+-0.0355 controlflow-recursive 2.3338+-0.0126 ? 2.3657+-0.0312 ? might be 1.0137x slower crypto-aes 7.5375+-0.0357 ! 7.9208+-0.2434 ! definitely 1.0509x slower crypto-md5 2.5011+-0.0324 2.4883+-0.0352 crypto-sha1 2.2588+-0.0305 2.2437+-0.0317 date-format-tofte 10.5184+-0.0684 ? 10.6383+-0.1559 ? might be 1.0114x slower date-format-xparb 9.7473+-0.0681 ! 9.9303+-0.0646 ! definitely 1.0188x slower math-cordic 7.2035+-0.0465 7.1695+-0.0183 math-partial-sums 10.4090+-0.0329 10.3948+-0.0197 math-spectral-norm 2.6357+-0.0044 ? 2.6493+-0.0215 ? regexp-dna 8.7873+-0.0520 ? 8.8476+-0.0411 ? string-base64 4.3422+-0.0172 ^ 4.3053+-0.0095 ^ definitely 1.0086x faster string-fasta 7.2648+-0.0211 ? 7.2716+-0.0196 ? string-tagcloud 12.3247+-0.0608 ? 12.3749+-0.0768 ? string-unpack-code 20.4331+-0.1069 20.4294+-0.0926 string-validate-input 5.5607+-0.0347 5.5410+-0.0350 <arithmetic> * 6.4280+-0.0164 ? 6.4571+-0.0192 ? might be 1.0045x slower <geometric> 5.2455+-0.0156 ? 5.2621+-0.0207 ? might be 1.0032x slower <harmonic> 4.1757+-0.0166 ? 4.1826+-0.0239 ? might be 1.0016x slower TipOfTree SepArgProfile V8: crypto 77.3966+-0.8162 ? 77.5275+-0.2308 ? deltablue 170.2357+-0.6947 ! 172.8179+-0.9768 ! definitely 1.0152x slower earley-boyer 99.7197+-0.9777 ? 100.0812+-1.6073 ? raytrace 52.6695+-0.8751 ? 53.0945+-0.8225 ? regexp 122.1514+-0.3019 121.4377+-0.5533 richards 137.6613+-0.3169 ? 137.9168+-0.6718 ? splay 92.0140+-1.1472 91.9703+-1.4986 <arithmetic> 107.4069+-0.2343 ? 107.8351+-0.5295 ? might be 1.0040x slower <geometric> * 100.9715+-0.3417 ? 101.3161+-0.5456 ? might be 1.0034x slower <harmonic> 94.4104+-0.4930 ? 94.7387+-0.5820 ? might be 1.0035x slower TipOfTree SepArgProfile Kraken: ai-astar 808.1426+-12.7223 ? 818.4768+-11.3447 ? might be 1.0128x slower audio-beat-detection 191.5014+-0.4783 191.0791+-0.3519 audio-dft 274.6687+-2.8362 ? 277.7278+-4.5364 ? might be 1.0111x slower audio-fft 119.7009+-1.0617 119.3369+-0.3265 audio-oscillator 284.3136+-4.1024 281.9721+-2.4438 imaging-darkroom 298.9661+-5.3908 ? 300.5002+-6.2156 ? imaging-desaturate 230.6104+-0.2709 230.5099+-0.1834 imaging-gaussian-blur 520.4534+-0.3280 520.2765+-0.2146 json-parse-financial 70.8143+-0.2055 ? 70.8822+-0.1997 ? json-stringify-tinderbox 81.4055+-0.2200 ? 81.7770+-0.3007 ? stanford-crypto-aes 117.8352+-1.4268 ? 117.9396+-0.6121 ? stanford-crypto-ccm 115.2356+-0.7613 114.6706+-1.0170 stanford-crypto-pbkdf2 232.3635+-0.4912 ? 233.9215+-3.2915 ? stanford-crypto-sha256-iterative 96.6566+-0.3117 ^ 95.9592+-0.1967 ^ definitely 1.0073x faster <arithmetic> * 245.9048+-1.2667 ? 246.7878+-0.8757 ? might be 1.0036x slower <geometric> 191.7964+-0.6193 ? 192.0069+-0.5309 ? might be 1.0011x slower <harmonic> 155.8955+-0.3985 155.8811+-0.3892 might be 1.0001x faster TipOfTree SepArgProfile All benchmarks: <arithmetic> 92.8009+-0.4021 ? 93.1439+-0.3090 ? might be 1.0037x slower <geometric> 23.8051+-0.0645 ? 23.8666+-0.0779 ? might be 1.0026x slower <harmonic> 7.3547+-0.0288 ? 7.3668+-0.0414 ? might be 1.0016x slower TipOfTree SepArgProfile Geomean of preferred means: <scaled-result> 54.2431+-0.1665 ? 54.4516+-0.1735 ? might be 1.0038x slower
Filip Pizlo
Comment 3 2011-12-20 18:29:37 PST
Note You need to log in before you can comment on or make changes to this bug.