RESOLVED FIXED Bug 126722
DFG and FTL should have separate compilation worklists
https://bugs.webkit.org/show_bug.cgi?id=126722
Summary DFG and FTL should have separate compilation worklists
Filip Pizlo
Reported 2014-01-09 15:05:31 PST
Patch forthcoming.
Attachments
the patch (18.20 KB, patch)
2014-01-09 15:08 PST, Filip Pizlo
mhahnenberg: review+
Filip Pizlo
Comment 1 2014-01-09 15:08:37 PST
Created attachment 220773 [details] the patch
Mark Hahnenberg
Comment 2 2014-01-09 15:15:34 PST
Comment on attachment 220773 [details] the patch r=me
Filip Pizlo
Comment 3 2014-01-09 16:26:19 PST
Benchmark report for SunSpider, Octane, and Kraken on oldmac (MacPro4,1). VMs tested: "TipOfTree" at /Volumes/Data/pizlo/cStack/OpenSource/WebKitBuild/Release/jsc (r161590) "ToTParallel" at /Volumes/Data/pizlo/cStack/OpenSource/WebKitBuild/Release/jsc (r161590) export JSC_numberOfCompilerThreads=7 "DualFTL" at /Volumes/Data/fromMiniMe/cStack/OpenSource/WebKitBuild/Release/jsc (r161590) Collected 10 samples per benchmark/VM, with 10 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 ToTParallel DualFTL DualFTL v. TipOfTree SunSpider: 3d-cube 7.7350+-0.1403 ? 7.8031+-0.0543 ? 7.8924+-0.0669 ? might be 1.0204x slower 3d-morph 8.7733+-0.0991 ! 9.1814+-0.1007 ^ 8.7825+-0.0606 ? 3d-raytrace 9.5156+-0.1249 9.2736+-0.1206 ? 9.3440+-0.1619 might be 1.0184x faster access-binary-trees 2.4751+-0.0171 ^ 2.3861+-0.0198 ^ 2.1944+-0.0394 ^ definitely 1.1279x faster access-fannkuch 7.9955+-0.1110 ? 8.2837+-0.5175 8.0945+-0.0772 ? might be 1.0124x slower access-nbody 4.2644+-0.0142 ! 4.4524+-0.0255 ^ 4.2758+-0.0098 ? access-nsieve 5.5880+-0.0206 ? 5.6016+-0.0964 5.5880+-0.1668 bitops-3bit-bits-in-byte 1.9945+-0.0207 ! 2.1833+-0.0133 ^ 1.9876+-0.0198 bitops-bits-in-byte 6.7168+-0.0741 ! 6.8758+-0.0213 ^ 6.6912+-0.0102 bitops-bitwise-and 3.0426+-0.1102 ? 3.1417+-0.0885 2.9948+-0.0926 might be 1.0159x faster bitops-nsieve-bits 5.7832+-0.0789 ! 6.1569+-0.0582 ^ 5.7990+-0.0970 ? controlflow-recursive 4.5550+-0.0541 ^ 3.9242+-0.0130 ^ 3.8724+-0.0140 ^ definitely 1.1763x faster crypto-aes 5.8840+-0.0771 ! 6.0512+-0.0807 ? 6.0727+-0.0630 ! definitely 1.0321x slower crypto-md5 3.5651+-0.0237 ? 3.6012+-0.0418 ? 3.6126+-0.0722 ? might be 1.0133x slower crypto-sha1 3.6039+-0.0455 ! 3.8102+-0.0323 ^ 3.7198+-0.0110 ! definitely 1.0321x slower date-format-tofte 11.4091+-0.1328 ! 11.6409+-0.0693 11.4883+-0.1030 ? date-format-xparb 8.4204+-0.0703 ! 8.5730+-0.0675 8.5148+-0.1166 ? might be 1.0112x slower math-cordic 4.7391+-0.0530 ! 4.9023+-0.0500 4.8443+-0.0160 ! definitely 1.0222x slower math-partial-sums 10.2762+-0.1186 ? 10.4455+-0.0977 10.2392+-0.1667 math-spectral-norm 4.4365+-0.3661 ^ 3.1942+-0.0144 ! 3.2475+-0.0260 ^ definitely 1.3661x faster regexp-dna 13.0538+-0.1486 12.9485+-0.1302 ? 13.1896+-0.2219 ? might be 1.0104x slower string-base64 5.5600+-0.0266 ! 5.9501+-0.0452 ^ 5.6209+-0.0939 ? might be 1.0109x slower string-fasta 11.2043+-0.2024 10.8907+-0.1598 ? 10.8925+-0.1670 might be 1.0286x faster string-tagcloud 15.0982+-0.1287 ! 15.4909+-0.1335 ^ 15.1574+-0.1170 ? string-unpack-code 32.1125+-0.3270 ? 32.6096+-0.2257 32.2217+-0.1813 ? string-validate-input 7.0659+-0.0967 ? 7.2557+-0.0951 ^ 7.0590+-0.0798 <arithmetic> * 7.8795+-0.0217 ! 7.9472+-0.0231 ^ 7.8230+-0.0155 ^ definitely 1.0072x faster <geometric> 6.5084+-0.0227 ? 6.5309+-0.0209 ^ 6.3898+-0.0163 ^ definitely 1.0186x faster <harmonic> 5.5083+-0.0320 ? 5.5089+-0.0198 ^ 5.3392+-0.0228 ^ definitely 1.0317x faster TipOfTree ToTParallel DualFTL DualFTL v. TipOfTree Octane and V8v7: encrypt 0.44227+-0.00445 0.43756+-0.00049 ? 0.43765+-0.00074 might be 1.0106x faster decrypt 8.09333+-0.02411 ? 8.14397+-0.04516 ? 8.17768+-0.07785 ? might be 1.0104x slower deltablue x2 0.55937+-0.00265 ^ 0.54515+-0.00929 0.53778+-0.00746 ^ definitely 1.0402x faster earley 0.89108+-0.01754 0.88955+-0.01030 0.88891+-0.01445 boyer 11.98773+-0.03993 ? 12.08035+-0.09569 12.01329+-0.05560 ? raytrace x2 3.97495+-0.04962 3.93134+-0.01134 ^ 3.88757+-0.03056 ^ definitely 1.0225x faster regexp x2 30.94620+-0.10105 ? 31.18226+-0.14836 ? 31.28541+-0.05398 ! definitely 1.0110x slower richards x2 0.22445+-0.00253 ? 0.22446+-0.00196 0.22423+-0.00170 splay x2 0.63010+-0.00352 0.62697+-0.00234 ? 0.62863+-0.00345 navier-stokes x2 8.24634+-0.01810 ? 8.27680+-0.09913 8.22878+-0.01162 closure 0.77319+-0.00194 0.77150+-0.00241 ? 0.78109+-0.01136 ? might be 1.0102x slower jquery 10.86971+-0.09816 ? 10.89717+-0.09670 ? 10.90944+-0.06909 ? gbemu x2 94.55735+-2.15531 ^ 85.47404+-2.31036 84.20366+-0.56867 ^ definitely 1.1230x faster mandreel x2 102.09544+-0.41028 101.44793+-0.37334 ? 101.68476+-0.91844 pdfjs x2 99.10754+-0.29963 ? 99.56893+-0.27883 ! 100.22624+-0.18232 ! definitely 1.0113x slower box2d x2 31.14315+-0.08152 ! 31.46904+-0.09198 ? 31.52653+-0.19350 ! definitely 1.0123x slower V8v7: <arithmetic> 6.91108+-0.01459 ? 6.94534+-0.02585 6.94390+-0.00663 ! definitely 1.0047x slower <geometric> * 2.17237+-0.00685 2.16430+-0.00537 2.15712+-0.00525 ^ definitely 1.0071x faster <harmonic> 0.79748+-0.00451 0.79203+-0.00381 0.78975+-0.00277 ^ definitely 1.0098x faster Octane including V8v7: <arithmetic> 29.84720+-0.16703 ^ 29.18130+-0.18892 29.15674+-0.07819 ^ definitely 1.0237x faster <geometric> * 6.57389+-0.02082 ^ 6.51243+-0.01967 6.50058+-0.01290 ^ definitely 1.0113x faster <harmonic> 1.20517+-0.00633 1.19727+-0.00544 1.19493+-0.00436 might be 1.0086x faster TipOfTree ToTParallel DualFTL DualFTL v. TipOfTree Kraken: ai-astar 494.054+-0.566 493.943+-0.519 ? 495.509+-4.113 ? audio-beat-detection 222.228+-1.196 221.917+-0.813 ? 222.988+-2.524 ? audio-dft 292.768+-2.611 ? 293.431+-2.820 ? 294.920+-4.407 ? audio-fft 129.644+-0.126 ? 129.781+-0.128 129.740+-0.139 ? audio-oscillator 530.579+-0.590 ? 531.590+-0.647 ? 533.247+-3.535 ? imaging-darkroom 298.699+-7.128 289.801+-2.708 ! 298.392+-1.162 imaging-desaturate 109.765+-0.061 ? 109.943+-0.201 109.736+-0.179 imaging-gaussian-blur 191.039+-1.253 ? 192.812+-0.961 ? 203.382+-26.139 ? might be 1.0646x slower json-parse-financial 79.933+-0.454 79.550+-0.288 ? 79.895+-0.347 json-stringify-tinderbox 103.529+-0.459 103.476+-0.412 ? 103.812+-0.424 ? stanford-crypto-aes 92.541+-0.573 92.278+-0.384 ? 92.791+-0.704 ? stanford-crypto-ccm 99.185+-1.443 ! 102.089+-1.355 ? 102.411+-0.800 ! definitely 1.0325x slower stanford-crypto-pbkdf2 259.257+-2.044 258.160+-1.551 ? 263.137+-3.793 ? might be 1.0150x slower stanford-crypto-sha256-iterative 113.674+-0.582 ? 114.611+-0.749 114.286+-0.367 ? <arithmetic> * 215.492+-0.609 215.242+-0.378 ! 217.446+-1.794 ? might be 1.0091x slower <geometric> 177.564+-0.415 ? 177.684+-0.345 ? 179.188+-1.369 ? might be 1.0091x slower <harmonic> 150.614+-0.300 ? 150.978+-0.330 ? 151.841+-0.780 ! definitely 1.0081x slower TipOfTree ToTParallel DualFTL DualFTL v. TipOfTree All benchmarks: <arithmetic> 60.5726+-0.1297 ^ 60.2837+-0.1281 ? 60.6927+-0.4013 ? might be 1.0020x slower <geometric> 13.1755+-0.0302 13.1465+-0.0182 ^ 13.0477+-0.0217 ^ definitely 1.0098x faster <harmonic> 2.5012+-0.0118 2.4879+-0.0098 ^ 2.4700+-0.0058 ^ definitely 1.0126x faster TipOfTree ToTParallel DualFTL DualFTL v. TipOfTree Geomean of preferred means: <scaled-result> 22.3486+-0.0367 22.3336+-0.0299 22.2785+-0.0631 might be 1.0031x faster
Filip Pizlo
Comment 4 2014-01-10 13:22:08 PST
Note You need to log in before you can comment on or make changes to this bug.