[Webkit-unassigned] [Bug 66363] The executable allocator makes it difficult to free individual chunks of executable memory
bugzilla-daemon at webkit.org
bugzilla-daemon at webkit.org
Wed Aug 17 18:56:50 PDT 2011
https://bugs.webkit.org/show_bug.cgi?id=66363
--- Comment #8 from Filip Pizlo <fpizlo at apple.com> 2011-08-17 18:56:50 PST ---
This is the performance when right-allocation is enabled. This produces sub-optimal code layout as it's possible for two things allocated consecutively to lie on opposite sides of the heap. But, it minimizes the number of pages in use. It's not clear if this is a win, or not.
Benchmark report for SunSpider, V8, and Kraken.
VMs tested:
"TipOfTree" at /Volumes/Data/pizlo/quinary/OpenSource/WebKitBuild/Release/jsc
"BestFitExecAlloc" at /Volumes/Data/pizlo/senary/OpenSource/WebKitBuild/Release/jsc
Collected 45 samples per benchmark/VM, with 15 VM invocations per benchmark. 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 BestFitExecAlloc
SunSpider:
3d-cube 7.5292+-0.0860 ? 7.6575+-0.0832 ? might be 1.0171x slower
3d-morph 7.4882+-0.0696 7.4571+-0.0709
3d-raytrace 7.6368+-0.0993 7.5628+-0.0799
access-binary-trees 2.2285+-0.0342 2.1824+-0.0182 might be 1.0211x faster
access-fannkuch 11.4710+-0.1021 11.3988+-0.0922
access-nbody 4.1968+-0.0381 ? 4.2456+-0.0731 ? might be 1.0116x slower
access-nsieve 2.4720+-0.0354 2.4574+-0.0371
bitops-3bit-bits-in-byte 1.7080+-0.0368 ? 1.7586+-0.0251 ? might be 1.0296x slower
bitops-bits-in-byte 4.1622+-0.0615 ! 4.4769+-0.0770 ! definitely 1.0756x slower
bitops-bitwise-and 3.6113+-0.0459 3.5752+-0.0413 might be 1.0101x faster
bitops-nsieve-bits 5.4251+-0.0643 ? 5.4367+-0.0533 ?
controlflow-recursive 2.0054+-0.0202 1.9957+-0.0266
crypto-aes 6.5658+-0.0766 6.5456+-0.0811
crypto-md5 2.7402+-0.0324 2.7273+-0.0247
crypto-sha1 2.1952+-0.0228 ? 2.2520+-0.0423 ? might be 1.0259x slower
date-format-tofte 9.9693+-0.1228 ? 10.1310+-0.1342 ? might be 1.0162x slower
date-format-xparb 8.5465+-0.1080 ? 8.6580+-0.1145 ? might be 1.0130x slower
math-cordic 6.2492+-0.0529 ? 6.2668+-0.0624 ?
math-partial-sums 7.5516+-0.0834 ? 7.5654+-0.0807 ?
math-spectral-norm 2.5182+-0.0329 2.4926+-0.0309 might be 1.0103x faster
regexp-dna 10.2535+-0.0721 ! 10.8177+-0.0894 ! definitely 1.0550x slower
string-base64 6.0013+-0.0908 ? 6.0495+-0.0992 ?
string-fasta 7.3581+-0.0810 7.3069+-0.0761
string-tagcloud 13.3322+-0.1478 ? 13.4441+-0.1326 ?
string-unpack-code 18.3448+-0.1520 ? 18.6119+-0.1853 ? might be 1.0146x slower
string-validate-input 7.0115+-0.1431 ? 7.0499+-0.1044 ?
<arithmetic> 6.4835+-0.0143 ! 6.5432+-0.0165 ! definitely 1.0092x slower
<geometric> 5.3735+-0.0131 ! 5.4140+-0.0135 ! definitely 1.0075x slower
<harmonic> 4.3890+-0.0166 ? 4.4165+-0.0144 ?
TipOfTree BestFitExecAlloc
V8:
crypto 88.3855+-0.2304 88.0580+-0.3105
deltablue 262.9253+-1.2663 ? 263.8095+-0.6202 ?
earley-boyer 100.0719+-0.2928 ? 100.3218+-0.2207 ?
raytrace 75.9575+-0.2486 75.4541+-0.3980
regexp 108.7350+-0.5273 ^ 107.2281+-0.4855 ^ definitely 1.0141x faster
richards 251.9579+-1.0113 ^ 241.2311+-1.2318 ^ definitely 1.0445x faster
splay 108.6511+-0.3619 107.9194+-0.4180
<arithmetic> 142.3834+-0.2936 ^ 140.5746+-0.2242 ^ definitely 1.0129x faster
<geometric> 126.7503+-0.2138 ^ 125.5102+-0.1853 ^ definitely 1.0099x faster
<harmonic> 115.4135+-0.1802 ^ 114.5392+-0.1817 ^ definitely 1.0076x faster
TipOfTree BestFitExecAlloc
Kraken:
ai-astar 1083.0171+-3.6911 1081.3665+-3.2078
audio-beat-detection 457.3596+-1.1298 ^ 454.2206+-1.4376 ^ definitely 1.0069x faster
audio-dft 413.2815+-2.8334 ? 416.3008+-2.3072 ?
audio-fft 357.1645+-1.2802 356.8773+-1.0993
audio-oscillator 378.9674+-1.0617 ! 381.9522+-1.1016 ! definitely 1.0079x slower
imaging-darkroom 575.5395+-2.2349 ! 609.5229+-3.0018 ! definitely 1.0590x slower
imaging-desaturate 592.2349+-2.9636 ? 593.5229+-1.6949 ?
imaging-gaussian-blur 1716.1634+-3.0601 1710.2664+-4.5857
json-parse-financial 48.2717+-0.1376 ? 48.3075+-0.1697 ?
json-stringify-tinderbox 61.8694+-0.1754 61.5657+-0.2802
stanford-crypto-aes 141.7960+-0.3397 ! 142.6480+-0.3741 ! definitely 1.0060x slower
stanford-crypto-ccm 110.2550+-0.2987 ? 110.4126+-0.2987 ?
stanford-crypto-pbkdf2 364.9750+-1.1447 ? 370.8834+-5.3735 ? might be 1.0162x slower
stanford-crypto-sha256-iterative 137.8285+-0.3687 ! 138.7276+-0.2695 ! definitely 1.0065x slower
<arithmetic> 459.9088+-0.5808 ! 462.6125+-0.6659 ! definitely 1.0059x slower
<geometric> 295.1312+-0.3490 ! 296.9687+-0.4547 ! definitely 1.0062x slower
<harmonic> 179.5327+-0.2189 ? 179.9774+-0.2983 ?
TipOfTree BestFitExecAlloc
All benchmarks:
<arithmetic> 161.7868+-0.1712 ! 162.3557+-0.2025 ! definitely 1.0035x slower
<geometric> 28.3740+-0.0410 ! 28.5031+-0.0414 ! definitely 1.0046x slower
<harmonic> 7.7524+-0.0288 ? 7.7995+-0.0248 ?
--
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.
More information about the webkit-unassigned
mailing list