[Webkit-unassigned] [Bug 74804] DFG 32_64 should eagerly speculate Int32 and Cell on heap loads
bugzilla-daemon at webkit.org
bugzilla-daemon at webkit.org
Sat Dec 17 23:17:30 PST 2011
https://bugs.webkit.org/show_bug.cgi?id=74804
Filip Pizlo <fpizlo at apple.com> changed:
What |Removed |Added
----------------------------------------------------------------------------
Attachment #119751| |review?
Flag| |
--- Comment #1 from Filip Pizlo <fpizlo at apple.com> 2011-12-17 23:17:30 PST ---
Created an attachment (id=119751)
--> (https://bugs.webkit.org/attachment.cgi?id=119751&action=review)
the patch
Benchmark report for SunSpider, V8, and Kraken on nitroflex (MacBookPro8,2).
VMs tested:
"TipOfTree32" at /Volumes/Data/pizlo/quinary/OpenSource/WebKitBuild/Release/jsc (r103166)
"EagerIntCell32" at /Volumes/Data/pizlo/quartary/OpenSource/WebKitBuild/Release/jsc (r103166)
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.
TipOfTree32 EagerIntCell32
SunSpider:
3d-cube 6.5152+-0.1034 ! 6.9520+-0.1399 ! definitely 1.0671x slower
3d-morph 10.5681+-0.1076 10.5453+-0.0857
3d-raytrace 9.2411+-0.1250 ? 9.2665+-0.1191 ?
access-binary-trees 1.7557+-0.0412 1.7321+-0.0354 might be 1.0136x faster
access-fannkuch 7.9326+-0.1010 ^ 7.2286+-0.0862 ^ definitely 1.0974x faster
access-nbody 5.1912+-0.0406 5.1522+-0.0602
access-nsieve 2.6116+-0.0499 2.5841+-0.0483 might be 1.0106x faster
bitops-3bit-bits-in-byte 1.2822+-0.0254 ^ 1.2343+-0.0084 ^ definitely 1.0389x faster
bitops-bits-in-byte 2.4810+-0.0537 ? 2.5142+-0.0676 ? might be 1.0134x slower
bitops-bitwise-and 2.8281+-0.0258 ? 2.8735+-0.0348 ? might be 1.0161x slower
bitops-nsieve-bits 6.2589+-0.1028 ? 6.3116+-0.0845 ?
controlflow-recursive 2.5813+-0.0530 2.5580+-0.0432
crypto-aes 9.2850+-0.1267 9.1234+-0.1779 might be 1.0177x faster
crypto-md5 3.0472+-0.0484 3.0451+-0.0386
crypto-sha1 2.6046+-0.0640 2.5974+-0.0537
date-format-tofte 10.6983+-0.1870 10.5845+-0.1181 might be 1.0108x faster
date-format-xparb 10.4971+-0.1022 ? 10.5607+-0.1040 ?
math-cordic 8.2823+-0.0975 8.2167+-0.1438
math-partial-sums 9.7506+-0.0925 9.6517+-0.0738 might be 1.0102x faster
math-spectral-norm 2.5231+-0.0492 2.4600+-0.0363 might be 1.0257x faster
regexp-dna 7.9393+-0.1087 ? 7.9501+-0.1697 ?
string-base64 4.8136+-0.1001 4.7835+-0.0954
string-fasta 7.8033+-0.1175 ? 7.8557+-0.1087 ?
string-tagcloud 12.3709+-0.2043 ? 12.5135+-0.2845 ? might be 1.0115x slower
string-unpack-code 20.2817+-0.2596 19.9612+-0.1532 might be 1.0161x faster
string-validate-input 5.9654+-0.1052 ? 6.0222+-0.1078 ?
<arithmetic> * 6.7350+-0.0203 6.7030+-0.0182 might be 1.0048x faster
<geometric> 5.4351+-0.0137 ^ 5.4078+-0.0126 ^ definitely 1.0050x faster
<harmonic> 4.2599+-0.0170 ^ 4.2245+-0.0112 ^ definitely 1.0084x faster
TipOfTree32 EagerIntCell32
V8:
crypto 95.5250+-1.3538 ? 96.4640+-0.5734 ?
deltablue 158.2678+-1.7653 ^ 154.4879+-0.5682 ^ definitely 1.0245x faster
earley-boyer 94.1228+-1.1494 ? 94.3708+-0.9196 ?
raytrace 51.6229+-0.5621 ? 51.8910+-1.1234 ?
regexp 109.8100+-0.8648 108.6379+-0.3674 might be 1.0108x faster
richards 169.3354+-0.4180 ^ 163.4216+-1.8519 ^ definitely 1.0362x faster
splay 81.9789+-1.2006 80.7980+-0.8816 might be 1.0146x faster
<arithmetic> 108.6661+-0.4893 ^ 107.1530+-0.4445 ^ definitely 1.0141x faster
<geometric> * 101.6176+-0.4695 100.6451+-0.5093 might be 1.0097x faster
<harmonic> 94.5797+-0.4440 94.0507+-0.6585 might be 1.0056x faster
TipOfTree32 EagerIntCell32
Kraken:
ai-astar 532.2392+-3.9922 ? 534.4831+-4.9855 ?
audio-beat-detection 350.2439+-9.9496 336.6175+-4.8982 might be 1.0405x faster
audio-dft 377.7458+-3.6279 ! 388.8345+-7.3560 ! definitely 1.0294x slower
audio-fft 208.5070+-1.7687 208.3388+-0.9079
audio-oscillator 416.5024+-6.2843 411.9362+-1.4885 might be 1.0111x faster
imaging-darkroom 367.0025+-4.2294 ? 370.3441+-3.8998 ?
imaging-desaturate 266.8889+-1.2328 ^ 260.8575+-1.9483 ^ definitely 1.0231x faster
imaging-gaussian-blur 600.8885+-2.4145 ^ 545.8639+-0.9871 ^ definitely 1.1008x faster
json-parse-financial 61.4667+-0.2622 ^ 60.7386+-0.4464 ^ definitely 1.0120x faster
json-stringify-tinderbox 94.9131+-0.9780 ^ 92.4179+-0.7361 ^ definitely 1.0270x faster
stanford-crypto-aes 115.3358+-1.3787 ^ 110.1062+-0.4967 ^ definitely 1.0475x faster
stanford-crypto-ccm 111.7851+-0.8190 111.2815+-0.7619
stanford-crypto-pbkdf2 219.6226+-0.8071 ? 220.5950+-0.8563 ?
stanford-crypto-sha256-iterative 98.3436+-0.5314 97.5897+-0.3802
<arithmetic> * 272.9632+-1.1985 ^ 267.8575+-0.7008 ^ definitely 1.0191x faster
<geometric> 218.9986+-0.8108 ^ 215.5792+-0.4689 ^ definitely 1.0159x faster
<harmonic> 171.0000+-0.5759 ^ 168.3718+-0.3686 ^ definitely 1.0156x faster
TipOfTree32 EagerIntCell32
All benchmarks:
<arithmetic> 101.2182+-0.3439 ^ 99.4543+-0.2435 ^ definitely 1.0177x faster
<geometric> 25.2796+-0.0549 ^ 25.0555+-0.0324 ^ definitely 1.0089x faster
<harmonic> 7.5088+-0.0295 ^ 7.4460+-0.0194 ^ definitely 1.0084x faster
TipOfTree32 EagerIntCell32
Geomean of preferred means:
<scaled-result> 57.1650+-0.1089 ^ 56.5350+-0.1051 ^ definitely 1.0111x faster
--
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