Bug 155217

Summary: [JSC] Pick how to OSR Enter to FTL at runtime instead of compile time
Product: WebKit Reporter: Benjamin Poulain <benjamin>
Component: New BugsAssignee: Benjamin Poulain <benjamin>
Status: RESOLVED FIXED    
Severity: Normal CC: commit-queue, fpizlo, keith_miller, mark.lam, msaboff, saam
Priority: P2    
Version: WebKit Nightly Build   
Hardware: Unspecified   
OS: Unspecified   
Attachments:
Description Flags
Patch
none
Patch
none
Patch none

Description Benjamin Poulain 2016-03-08 21:37:44 PST
[JSC] Pick how to OSR Enter to FTL at runtime instead of compile time
Comment 1 Benjamin Poulain 2016-03-08 21:38:08 PST
Created attachment 273390 [details]
Patch
Comment 2 Benjamin Poulain 2016-03-08 22:04:45 PST
Created attachment 273392 [details]
Patch
Comment 3 Benjamin Poulain 2016-03-09 00:32:06 PST
Created attachment 273406 [details]
Patch
Comment 4 Benjamin Poulain 2016-03-09 01:02:19 PST
On Haswell:

                                                  Conf#1                    Conf#2                                      
SunSpider:
   3d-cube                                    4.7818+-0.1586     ?      4.9487+-0.4769        ? might be 1.0349x slower
   3d-morph                                   5.1752+-0.0789            5.1709+-0.0258        
   3d-raytrace                                5.6498+-0.0935     ?      5.7074+-0.1271        ? might be 1.0102x slower
   access-binary-trees                        2.1734+-0.0289     ?      2.1929+-0.0953        ?
   access-fannkuch                            5.8580+-0.1877     ?      6.2639+-0.2354        ? might be 1.0693x slower
   access-nbody                               2.6979+-0.1572            2.6641+-0.0291          might be 1.0127x faster
   access-nsieve                              3.0267+-0.0570     ?      3.1146+-0.1131        ? might be 1.0291x slower
   bitops-3bit-bits-in-byte                   1.1736+-0.0171            1.1694+-0.0097        
   bitops-bits-in-byte                        3.3361+-0.0566     ^      2.7948+-0.0250        ^ definitely 1.1937x faster
   bitops-bitwise-and                         2.0667+-0.0949            2.0118+-0.0280          might be 1.0273x faster
   bitops-nsieve-bits                         3.2236+-0.2418            3.1553+-0.0092          might be 1.0216x faster
   controlflow-recursive                      2.3496+-0.0185     ?      2.3804+-0.0663        ? might be 1.0131x slower
   crypto-aes                                 3.9884+-0.0184     !      4.0337+-0.0194        ! definitely 1.0114x slower
   crypto-md5                                 2.7276+-0.2583            2.6060+-0.0490          might be 1.0466x faster
   crypto-sha1                                2.4071+-0.0787            2.3528+-0.0981          might be 1.0231x faster
   date-format-tofte                          6.8128+-0.0754            6.7476+-0.1613        
   date-format-xparb                          4.9510+-0.0993     ?      5.0350+-0.2771        ? might be 1.0170x slower
   math-cordic                                2.8824+-0.0726     ?      2.9388+-0.1121        ? might be 1.0196x slower
   math-partial-sums                          4.9696+-0.1537            4.8574+-0.0796          might be 1.0231x faster
   math-spectral-norm                         2.0487+-0.0443            2.0464+-0.0204        
   regexp-dna                                 6.1712+-0.4185            5.9733+-0.1118          might be 1.0331x faster
   string-base64                              4.5110+-0.1032            4.4902+-0.0137        
   string-fasta                               5.8586+-0.1448     ?      6.1360+-0.5954        ? might be 1.0474x slower
   string-tagcloud                            8.2051+-0.2147     ?      8.2271+-0.2378        ?
   string-unpack-code                        19.3395+-0.8338           18.8068+-0.4569          might be 1.0283x faster
   string-validate-input                      4.2962+-0.1992            4.2766+-0.0788        

   <arithmetic>                               4.6416+-0.0437            4.6193+-0.0420          might be 1.0048x faster

                                                  Conf#1                    Conf#2                                      
Octane:
   encrypt                                   0.16005+-0.00144          0.15975+-0.00193       
   decrypt                                   2.90814+-0.01990          2.90627+-0.02178       
   deltablue                        x2       0.13722+-0.00144    ?     0.13741+-0.00214       ?
   earley                                    0.28578+-0.00109    ?     0.28593+-0.00178       ?
   boyer                                     4.97481+-0.06238          4.96097+-0.05369       
   navier-stokes                    x2       4.94279+-0.01247    ?     4.94304+-0.01130       ?
   raytrace                         x2       0.89288+-0.00314          0.89004+-0.00345       
   richards                         x2       0.08258+-0.00065          0.08237+-0.00098       
   splay                            x2       0.34555+-0.00233          0.34476+-0.00231       
   regexp                           x2      20.69545+-0.19894    ?    20.70842+-0.13202       ?
   pdfjs                            x2      38.93763+-0.42943    ?    39.20909+-0.30882       ?
   mandreel                         x2      42.52644+-0.18505    ?    42.65767+-0.69928       ?
   gbemu                            x2      24.55816+-0.09063    ?    24.76803+-0.21996       ?
   closure                                   0.56404+-0.00227          0.56392+-0.00190       
   jquery                                    7.41735+-0.03448          7.38266+-0.03637       
   box2d                            x2       9.35439+-0.04863          9.28616+-0.04939       
   zlib                             x2     388.44289+-2.89113        380.34795+-5.78622         might be 1.0213x faster
   typescript                       x2     656.85282+-5.90715    ?   660.89775+-8.55374       ?

   <geometric>                               5.21695+-0.00960          5.21163+-0.01203         might be 1.0010x faster

                                                  Conf#1                    Conf#2                                      
Kraken:
   ai-astar                                   93.425+-0.894      ^      87.995+-1.652         ^ definitely 1.0617x faster
   audio-beat-detection                       47.502+-0.138      ^      45.242+-0.185         ^ definitely 1.0500x faster
   audio-dft                                  97.386+-1.297      ?      97.939+-1.332         ?
   audio-fft                                  35.960+-0.413             35.709+-0.021         
   audio-oscillator                           47.633+-0.124             47.589+-0.129         
   imaging-darkroom                           60.452+-0.752             60.211+-0.106         
   imaging-desaturate                         44.334+-0.084      ?      45.127+-1.717         ? might be 1.0179x slower
   imaging-gaussian-blur                      67.618+-0.072             64.667+-3.045           might be 1.0456x faster
   json-parse-financial                       37.237+-0.858      ?      37.323+-0.403         ?
   json-stringify-tinderbox                   25.510+-0.885      ?      25.603+-0.680         ?
   stanford-crypto-aes                        40.435+-0.363             40.431+-0.335         
   stanford-crypto-ccm                        36.403+-1.428             36.342+-0.387         
   stanford-crypto-pbkdf2                    101.465+-0.978            101.228+-0.885         
   stanford-crypto-sha256-iterative           38.552+-0.170      ?      38.837+-0.443         ?

   <arithmetic>                               55.279+-0.219      ^      54.589+-0.357         ^ definitely 1.0127x faster

                                                  Conf#1                    Conf#2                                      
AsmBench:
   bigfib.cpp                               437.0777+-5.0931          435.2994+-4.3668        
   cray.c                                   363.5383+-2.1272          359.7468+-1.9729          might be 1.0105x faster
   dry.c                                    430.1062+-19.4462         421.9460+-5.1615          might be 1.0193x faster
   FloatMM.c                                717.2010+-2.2742     ?    717.3281+-1.9894        ?
   gcc-loops.cpp                           3676.3863+-16.2518    ?   3679.1685+-19.1307       ?
   n-body.c                                 808.7624+-1.9947     ?    808.7940+-2.5295        ?
   Quicksort.c                              394.7614+-1.8896          391.7815+-1.4073        
   stepanov_container.cpp                  3279.1700+-20.4512    ?   3298.1549+-18.6087       ?
   Towers.c                                 269.0207+-1.8971          267.8508+-0.9749        

   <geometric>                              717.8877+-3.6137          714.9325+-1.4096          might be 1.0041x faster

                                                  Conf#1                    Conf#2                                      
Geomean of preferred means:
   <scaled-result>                           31.3086+-0.0951           31.1327+-0.0890          might be 1.0057x faster
Comment 5 WebKit Commit Bot 2016-03-09 09:51:27 PST
Comment on attachment 273406 [details]
Patch

Clearing flags on attachment: 273406

Committed r197861: <http://trac.webkit.org/changeset/197861>
Comment 6 WebKit Commit Bot 2016-03-09 09:51:31 PST
All reviewed patches have been landed.  Closing bug.