Bug 160346

Summary: [JSC] Use the same data structures for DFG and Air Liveness Analysis
Product: WebKit Reporter: Benjamin Poulain <benjamin>
Component: New BugsAssignee: 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 Flags
Patch
none
Patch none

Description Benjamin Poulain 2016-07-29 10:48:52 PDT
[JSC] Use the same data structures for DFG and Air Liveness Analysis
Comment 1 Benjamin Poulain 2016-07-29 10:53:33 PDT
Created attachment 284873 [details]
Patch
Comment 2 Geoffrey Garen 2016-07-29 10:58:31 PDT
Comment on attachment 284873 [details]
Patch

r=me

Do we have benchmark results for this?
Comment 3 Benjamin Poulain 2016-07-29 11:01:11 PDT
(In reply to comment #2)
> Comment on attachment 284873 [details]
> Patch
> 
> r=me
> 
> Do we have benchmark results for this?

Thanks for the review.

Here are the early result. Next I need to do something about CFA.

                                                  Conf#1                    Conf#2                                      
SunSpider:
   3d-cube                                    4.5709+-0.1010            4.5246+-0.0949          might be 1.0102x faster
   3d-morph                                   4.9709+-0.0442     ?      4.9763+-0.0455        ?
   3d-raytrace                                4.8954+-0.0640            4.8828+-0.0354        
   access-binary-trees                        1.9768+-0.0339            1.9554+-0.0198          might be 1.0110x faster
   access-fannkuch                            5.8099+-0.0619            5.7292+-0.0736          might be 1.0141x faster
   access-nbody                               2.3602+-0.0318            2.3507+-0.0324        
   access-nsieve                              2.9633+-0.0571            2.9355+-0.0242        
   bitops-3bit-bits-in-byte                   1.1042+-0.0441            1.0652+-0.0138          might be 1.0366x faster
   bitops-bits-in-byte                        2.6202+-0.0578            2.5688+-0.0250          might be 1.0200x faster
   bitops-bitwise-and                         1.9791+-0.0247     ?      1.9861+-0.0251        ?
   bitops-nsieve-bits                         3.1294+-0.0525     ?      3.2140+-0.1669        ? might be 1.0270x slower
   controlflow-recursive                      2.2983+-0.0241     ?      2.2986+-0.0290        ?
   crypto-aes                                 4.3926+-0.0464            4.3645+-0.0277        
   crypto-md5                                 2.6512+-0.0238            2.6446+-0.0476        
   crypto-sha1                                2.7672+-0.0559            2.7201+-0.0170          might be 1.0173x faster
   date-format-tofte                          6.4525+-0.0452     ?      6.4914+-0.1537        ?
   date-format-xparb                          4.7507+-0.0641     ?      4.7671+-0.0763        ?
   math-cordic                                2.7117+-0.0209     ?      2.7347+-0.0336        ?
   math-partial-sums                          4.0073+-0.1059            3.9786+-0.0658        
   math-spectral-norm                         2.0708+-0.0569            2.0469+-0.0383          might be 1.0117x faster
   regexp-dna                                 6.4805+-0.3018            6.2571+-0.0770          might be 1.0357x faster
   string-base64                              3.9325+-0.0215     ?      3.9911+-0.1128        ? might be 1.0149x slower
   string-fasta                               5.3708+-0.0120     ?      5.4720+-0.1385        ? might be 1.0188x slower
   string-tagcloud                            8.3679+-0.1476            8.3539+-0.1753        
   string-unpack-code                        18.0326+-0.3545           17.6974+-0.1306          might be 1.0189x faster
   string-validate-input                      4.0669+-0.0194            4.0493+-0.0409        

   <arithmetic>                               4.4128+-0.0158            4.3868+-0.0129          might be 1.0059x faster

                                                  Conf#1                    Conf#2                                      
Octane:
   encrypt                                   0.16192+-0.00158          0.16042+-0.00103       
   decrypt                                   2.70025+-0.01958    ?     2.72996+-0.03202       ? might be 1.0110x slower
   deltablue                        x2       0.13108+-0.00122    ?     0.13496+-0.00758       ? might be 1.0296x slower
   earley                                    0.28488+-0.00115    ?     0.28534+-0.00125       ?
   boyer                                     5.01169+-0.01921          4.96573+-0.07191       
   navier-stokes                    x2       4.93577+-0.00827          4.92569+-0.00858       
   raytrace                         x2       0.79704+-0.00663          0.79586+-0.00223       
   richards                         x2       0.08318+-0.00067          0.08292+-0.00099       
   splay                            x2       0.33854+-0.00095    ?     0.33914+-0.00204       ?
   regexp                           x2      15.84690+-0.17565    ?    15.92367+-0.22181       ?
   pdfjs                            x2      39.08439+-0.15289         38.86791+-0.19234       
   mandreel                         x2      42.72122+-0.16330    ?    42.84572+-0.19546       ?
   gbemu                            x2      29.64741+-0.13145         29.59276+-0.07079       
   closure                                   0.48228+-0.00189          0.48068+-0.00200       
   jquery                                    6.41661+-0.02983    ?     6.43618+-0.03195       ?
   box2d                            x2       9.25132+-0.04067          9.21141+-0.03270       
   zlib                             x2     371.11875+-6.40502        367.60278+-8.00245       
   typescript                       x2     604.87042+-3.76956        600.18077+-3.77438       

   <geometric>                               5.02463+-0.01060          5.02344+-0.01584         might be 1.0002x faster

                                                  Conf#1                    Conf#2                                      
Kraken:
   ai-astar                                   87.638+-1.246             87.086+-0.771         
   audio-beat-detection                       38.844+-0.197             38.762+-0.114         
   audio-dft                                  99.105+-2.158             97.684+-1.589           might be 1.0145x faster
   audio-fft                                  30.229+-0.050             30.168+-0.029         
   audio-oscillator                           47.659+-0.109      ?      47.814+-0.219         ?
   imaging-darkroom                           61.527+-0.720      ?      61.560+-0.824         ?
   imaging-desaturate                         44.139+-0.500             43.682+-0.117           might be 1.0104x faster
   imaging-gaussian-blur                      61.829+-2.742      ?      63.319+-2.334         ? might be 1.0241x slower
   json-parse-financial                       33.583+-0.869      ?      33.779+-0.152         ?
   json-stringify-tinderbox                   22.726+-0.513             22.201+-0.541           might be 1.0236x faster
   stanford-crypto-aes                        36.984+-0.786             36.652+-0.149         
   stanford-crypto-ccm                        33.881+-0.956      ?      34.121+-0.897         ?
   stanford-crypto-pbkdf2                     92.377+-0.224      ?      92.415+-0.189         ?
   stanford-crypto-sha256-iterative           30.209+-0.130      ?      30.303+-0.181         ?

   <arithmetic>                               51.481+-0.396             51.396+-0.177           might be 1.0016x faster

                                                  Conf#1                    Conf#2                                      
AsmBench:
   bigfib.cpp                               427.9843+-2.9245     ?    428.2946+-1.4885        ?
   cray.c                                   388.8401+-1.7058          387.5606+-1.4859        
   dry.c                                    445.6685+-38.6640         423.4977+-4.3387          might be 1.0524x faster
   FloatMM.c                                738.9616+-22.8370    ?    767.9878+-25.9539       ? might be 1.0393x slower
   gcc-loops.cpp                           3594.7563+-11.6217    ?   3599.8395+-15.4293       ?
   n-body.c                                 801.5333+-3.0331          799.2980+-2.4383        
   Quicksort.c                              394.9852+-3.2691     ?    396.3970+-2.5728        ?
   stepanov_container.cpp                  3309.3185+-26.7553        3305.2373+-12.1616       
   Towers.c                                 264.5931+-0.4013     ?    264.8891+-0.9552        ?

   <geometric>                              723.3475+-6.5352          722.7770+-2.3164          might be 1.0008x faster

                                                  Conf#1                    Conf#2                                      
Geomean of preferred means:
   <scaled-result>                           30.1433+-0.1118           30.0793+-0.0609          might be 1.0021x faster
Comment 4 Benjamin Poulain 2016-07-29 11:44:55 PDT
Created attachment 284878 [details]
Patch
Comment 5 WebKit Commit Bot 2016-07-29 13:59:51 PDT
Comment on attachment 284878 [details]
Patch

Clearing flags on attachment: 284878

Committed r203921: <http://trac.webkit.org/changeset/203921>
Comment 6 WebKit Commit Bot 2016-07-29 13:59:54 PDT
All reviewed patches have been landed.  Closing bug.