| Summary: | [JSC] CFA's valuesAtHead should be a list, not a map | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Product: | WebKit | Reporter: | Benjamin Poulain <benjamin> | ||||||||||
| Component: | New Bugs | Assignee: | Benjamin Poulain <benjamin> | ||||||||||
| Status: | RESOLVED FIXED | ||||||||||||
| Severity: | Normal | CC: | commit-queue, keith_miller, mark.lam, msaboff, saam | ||||||||||
| Priority: | P2 | ||||||||||||
| Version: | WebKit Nightly Build | ||||||||||||
| Hardware: | Unspecified | ||||||||||||
| OS: | Unspecified | ||||||||||||
| Attachments: |
|
||||||||||||
|
Description
Benjamin Poulain
2016-03-31 15:39:30 PDT
Created attachment 275343 [details]
Patch
Conf#1 Conf#2
SunSpider:
3d-cube 5.0124+-0.1095 ? 5.0495+-0.1649 ?
3d-morph 5.2697+-0.0952 ? 5.3258+-0.1489 ? might be 1.0107x slower
3d-raytrace 5.6004+-0.0559 ? 5.6452+-0.0747 ?
access-binary-trees 2.1611+-0.0345 ? 2.2603+-0.1425 ? might be 1.0459x slower
access-fannkuch 5.9909+-0.1478 ? 5.9946+-0.1222 ?
access-nbody 2.5929+-0.0700 2.5786+-0.0741
access-nsieve 3.2500+-0.0844 ? 3.2807+-0.0344 ?
bitops-3bit-bits-in-byte 1.1734+-0.0652 1.1367+-0.0111 might be 1.0323x faster
bitops-bits-in-byte 2.7953+-0.0645 ? 2.8072+-0.0750 ?
bitops-bitwise-and 2.0582+-0.0202 ? 2.0760+-0.0156 ?
bitops-nsieve-bits 3.1624+-0.0833 3.1225+-0.0143 might be 1.0128x faster
controlflow-recursive 2.3448+-0.0275 ? 2.3459+-0.0204 ?
crypto-aes 4.0596+-0.0582 ? 4.0630+-0.0314 ?
crypto-md5 2.4968+-0.0493 ? 2.5639+-0.0964 ? might be 1.0269x slower
crypto-sha1 2.3092+-0.0268 ? 2.3446+-0.0294 ? might be 1.0153x slower
date-format-tofte 6.6700+-0.2007 ? 6.7207+-0.3148 ?
date-format-xparb 4.8888+-0.0453 ? 4.8957+-0.0996 ?
math-cordic 2.8755+-0.0680 2.8470+-0.0354 might be 1.0100x faster
math-partial-sums 4.8273+-0.1244 4.8156+-0.1142
math-spectral-norm 2.0146+-0.0121 2.0107+-0.0211
regexp-dna 6.2893+-0.0666 ? 6.3829+-0.2097 ? might be 1.0149x slower
string-base64 4.4897+-0.0622 4.4598+-0.0668
string-fasta 5.9860+-0.1249 ? 6.0070+-0.1505 ?
string-tagcloud 8.2003+-0.1059 ? 8.2580+-0.1416 ?
string-unpack-code 19.7223+-0.6065 18.9724+-0.5338 might be 1.0395x faster
string-validate-input 4.3760+-0.0587 ? 4.5129+-0.1157 ? might be 1.0313x slower
<arithmetic> 4.6391+-0.0293 4.6337+-0.0290 might be 1.0012x faster
Conf#1 Conf#2
Octane:
encrypt 0.16570+-0.00128 0.16392+-0.00077 might be 1.0109x faster
decrypt 2.84671+-0.00361 ? 2.84968+-0.00424 ?
deltablue x2 0.14134+-0.00145 0.13976+-0.00201 might be 1.0113x faster
earley 0.28660+-0.00147 0.28592+-0.00127
boyer 5.00582+-0.06893 ? 5.01403+-0.04398 ?
navier-stokes x2 5.00782+-0.00921 5.00033+-0.00779
raytrace x2 0.90237+-0.00319 0.89788+-0.00260
richards x2 0.08282+-0.00056 ? 0.08318+-0.00058 ?
splay x2 0.35541+-0.00203 0.35212+-0.00271
regexp x2 18.88801+-0.17953 18.81963+-0.15750
pdfjs x2 39.45263+-0.35712 39.24387+-0.21780
mandreel x2 42.92651+-0.15591 ^ 42.63580+-0.11702 ^ definitely 1.0068x faster
gbemu x2 24.22296+-0.12918 24.16282+-0.09183
closure 0.55798+-0.00163 0.55783+-0.00170
jquery 7.23966+-0.02246 7.21009+-0.01784
box2d x2 9.28564+-0.04629 ? 9.29676+-0.05766 ?
zlib x2 364.82151+-1.48672 364.29013+-4.55952
typescript x2 637.38728+-2.41539 636.77105+-3.40731
<geometric> 5.18058+-0.00643 ^ 5.16347+-0.00824 ^ definitely 1.0033x faster
Conf#1 Conf#2
Kraken:
ai-astar 88.454+-0.513 ? 89.055+-1.202 ?
audio-beat-detection 42.457+-0.103 42.290+-0.086
audio-dft 100.360+-1.766 98.818+-0.891 might be 1.0156x faster
audio-fft 33.039+-0.471 32.827+-0.044
audio-oscillator 48.470+-0.141 ? 48.548+-0.355 ?
imaging-darkroom 60.220+-0.110 ? 60.281+-0.332 ?
imaging-desaturate 45.902+-1.206 45.296+-0.211 might be 1.0134x faster
imaging-gaussian-blur 61.880+-0.920 61.762+-1.189
json-parse-financial 38.174+-0.293 ^ 37.275+-0.255 ^ definitely 1.0241x faster
json-stringify-tinderbox 24.590+-0.683 ^ 22.721+-0.625 ^ definitely 1.0822x faster
stanford-crypto-aes 40.209+-0.666 40.152+-0.663
stanford-crypto-ccm 36.426+-1.295 35.396+-1.116 might be 1.0291x faster
stanford-crypto-pbkdf2 100.125+-0.498 99.911+-0.193
stanford-crypto-sha256-iterative 39.341+-0.639 39.175+-0.194
<arithmetic> 54.260+-0.244 53.822+-0.214 might be 1.0081x faster
Conf#1 Conf#2
AsmBench:
bigfib.cpp 442.4133+-4.7837 ? 446.6549+-2.0427 ?
cray.c 357.1573+-1.8432 356.9566+-2.2301
dry.c 447.4224+-33.0504 ? 463.2113+-40.5820 ? might be 1.0353x slower
FloatMM.c 731.9868+-1.8324 ? 733.0239+-1.8757 ?
gcc-loops.cpp 3719.1458+-12.1226 3713.4890+-3.4340
n-body.c 810.4151+-1.5719 ? 811.5347+-2.5391 ?
Quicksort.c 398.8821+-1.8514 398.2057+-1.3629
stepanov_container.cpp 3326.1429+-14.8132 3307.3530+-14.7376
Towers.c 272.6502+-1.2299 272.1253+-0.5854
<geometric> 726.0555+-5.1672 ? 728.7385+-6.4754 ? might be 1.0037x slower
Conf#1 Conf#2
Geomean of preferred means:
<scaled-result> 31.1929+-0.0960 31.1233+-0.0931 might be 1.0022x faster
Comment on attachment 275343 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=275343&action=review r=me > Source/JavaScriptCore/ChangeLog:8 > + One more step toward moving to the Air-style of livness analysis: typo: /livness/liveness/. > Source/JavaScriptCore/dfg/DFGNode.h:2354 > + T sortedMap = nodeValuePairList; > + std::sort(sortedMap.begin(), sortedMap.end(), nodeValuePairComparator<decltype(*sortedMap.begin())>); Why not call it sortedList instead of sortedMap? Created attachment 275345 [details]
Patch for landing
Created attachment 275348 [details]
Patch
Created attachment 275353 [details]
Patch
Comment on attachment 275353 [details] Patch Clearing flags on attachment: 275353 Committed r198935: <http://trac.webkit.org/changeset/198935> All reviewed patches have been landed. Closing bug. Comment on attachment 275353 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=275353&action=review > Source/JavaScriptCore/dfg/DFGInPlaceAbstractState.cpp:89 > + values.resize(0); I think this can be shrink(0) and that might save one branch. (In reply to comment #9) > Comment on attachment 275353 [details] > Patch > > View in context: > https://bugs.webkit.org/attachment.cgi?id=275353&action=review > > > Source/JavaScriptCore/dfg/DFGInPlaceAbstractState.cpp:89 > > + values.resize(0); > > I think this can be shrink(0) and that might save one branch. If Clang really generates a branch here, we should file a bug. |