Bug 156095 - [JSC][x86] Add the 3 operands form of floating point substraction
Summary: [JSC][x86] Add the 3 operands form of floating point substraction
Status: RESOLVED FIXED
Alias: None
Product: WebKit
Classification: Unclassified
Component: New Bugs (show other bugs)
Version: WebKit Nightly Build
Hardware: Unspecified Unspecified
: P2 Normal
Assignee: Benjamin Poulain
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2016-03-31 21:37 PDT by Benjamin Poulain
Modified: 2016-04-01 12:20 PDT (History)
5 users (show)

See Also:


Attachments
Patch (10.06 KB, patch)
2016-03-31 21:51 PDT, Benjamin Poulain
no flags Details | Formatted Diff | Diff

Note You need to log in before you can comment on or make changes to this bug.
Description Benjamin Poulain 2016-03-31 21:37:47 PDT
[JSC][x86] Add the 3 operands form of floating point substraction
Comment 1 Benjamin Poulain 2016-03-31 21:51:51 PDT
Created attachment 275370 [details]
Patch
Comment 2 Benjamin Poulain 2016-03-31 22:23:01 PDT
                                                  Conf#1                    Conf#2                                      
SunSpider:
   3d-cube                                    4.9970+-0.1295            4.9917+-0.1754        
   3d-morph                                   5.3338+-0.0866            5.3217+-0.1527        
   3d-raytrace                                5.5747+-0.0451     ?      5.6360+-0.0541        ? might be 1.0110x slower
   access-binary-trees                        2.1744+-0.0456            2.1653+-0.0278        
   access-fannkuch                            5.9211+-0.1037     ?      6.0417+-0.1639        ? might be 1.0204x slower
   access-nbody                               2.5431+-0.0279     ?      2.5847+-0.0812        ? might be 1.0164x slower
   access-nsieve                              3.2416+-0.0835     ?      3.2741+-0.1313        ? might be 1.0100x slower
   bitops-3bit-bits-in-byte                   1.1284+-0.0137     ?      1.1570+-0.0534        ? might be 1.0253x slower
   bitops-bits-in-byte                        2.7648+-0.0324     ?      2.7672+-0.0237        ?
   bitops-bitwise-and                         2.0601+-0.0367     ?      2.1373+-0.0786        ? might be 1.0374x slower
   bitops-nsieve-bits                         3.0945+-0.0179     ?      3.1166+-0.0317        ?
   controlflow-recursive                      2.3816+-0.0336            2.3494+-0.0183          might be 1.0137x faster
   crypto-aes                                 4.0685+-0.0934            4.0280+-0.0276          might be 1.0100x faster
   crypto-md5                                 2.4896+-0.0161     ?      2.5144+-0.0596        ?
   crypto-sha1                                2.3378+-0.0522            2.3255+-0.0156        
   date-format-tofte                          6.4474+-0.0822     ?      6.5493+-0.2450        ? might be 1.0158x slower
   date-format-xparb                          4.9510+-0.1300            4.9251+-0.0931        
   math-cordic                                2.8579+-0.0865     ?      2.8680+-0.0196        ?
   math-partial-sums                          4.8621+-0.1135     ?      4.8950+-0.1396        ?
   math-spectral-norm                         1.9929+-0.0117     ?      2.0178+-0.0334        ? might be 1.0125x slower
   regexp-dna                                 6.3340+-0.1591     ?      6.3446+-0.0875        ?
   string-base64                              4.5489+-0.1775     ?      4.5627+-0.1784        ?
   string-fasta                               5.9884+-0.1140            5.9432+-0.0883        
   string-tagcloud                            8.1668+-0.0777            8.1238+-0.0256        
   string-unpack-code                        19.8111+-0.6574           18.9221+-0.2907          might be 1.0470x faster
   string-validate-input                      4.3708+-0.0211            4.3475+-0.0493        

   <arithmetic>                               4.6324+-0.0317            4.6119+-0.0163          might be 1.0044x faster

                                                  Conf#1                    Conf#2                                      
Octane:
   encrypt                                   0.16477+-0.00100          0.16448+-0.00115       
   decrypt                                   2.84480+-0.00263    ?     2.84828+-0.00532       ?
   deltablue                        x2       0.14104+-0.00110    ?     0.14133+-0.00130       ?
   earley                                    0.28699+-0.00136          0.28684+-0.00149       
   boyer                                     5.02378+-0.04956    ?     5.03049+-0.03587       ?
   navier-stokes                    x2       5.00439+-0.01176          4.99746+-0.01198       
   raytrace                         x2       0.89838+-0.00419    ?     0.90220+-0.00350       ?
   richards                         x2       0.08295+-0.00075          0.08284+-0.00068       
   splay                            x2       0.35417+-0.00249          0.35290+-0.00303       
   regexp                           x2      18.76751+-0.07713    ?    19.03021+-0.19185       ? might be 1.0140x slower
   pdfjs                            x2      39.46797+-0.39092         39.35072+-0.33608       
   mandreel                         x2      42.60600+-0.08343    ?    42.63567+-0.11660       ?
   gbemu                            x2      24.29384+-0.18998         24.20727+-0.16128       
   closure                                   0.55930+-0.00165          0.55891+-0.00204       
   jquery                                    7.23269+-0.01593    ?     7.24100+-0.01952       ?
   box2d                            x2       9.26587+-0.03753    ?     9.27044+-0.06774       ?
   zlib                             x2     364.33754+-1.71950        359.40654+-4.85495         might be 1.0137x faster
   typescript                       x2     642.88659+-4.95458        640.18424+-3.37852       

   <geometric>                               5.17552+-0.00681          5.17235+-0.00896         might be 1.0006x faster

                                                  Conf#1                    Conf#2                                      
Kraken:
   ai-astar                                   89.230+-1.300             88.568+-0.872         
   audio-beat-detection                       42.768+-0.453             42.493+-0.073         
   audio-dft                                  99.339+-1.040      ?      99.406+-1.083         ?
   audio-fft                                  32.832+-0.043      ?      32.882+-0.138         ?
   audio-oscillator                           48.676+-0.646             48.410+-0.061         
   imaging-darkroom                           60.652+-0.715             60.364+-0.290         
   imaging-desaturate                         45.396+-0.192             45.152+-0.131         
   imaging-gaussian-blur                      62.463+-0.863      ?      62.716+-0.566         ?
   json-parse-financial                       38.368+-0.657             38.156+-0.356         
   json-stringify-tinderbox                   24.770+-0.663      ?      25.124+-1.036         ? might be 1.0143x slower
   stanford-crypto-aes                        39.790+-0.243      ^      39.388+-0.095         ^ definitely 1.0102x faster
   stanford-crypto-ccm                        35.249+-1.214      ?      35.414+-0.545         ?
   stanford-crypto-pbkdf2                    100.020+-0.200      ?     100.479+-0.765         ?
   stanford-crypto-sha256-iterative           38.930+-0.192      ?      38.937+-0.293         ?

   <arithmetic>                               54.177+-0.158             54.106+-0.183           might be 1.0013x faster

                                                  Conf#1                    Conf#2                                      
AsmBench:
   bigfib.cpp                               441.9200+-4.5823     ?    446.3416+-1.6150        ? might be 1.0100x slower
   cray.c                                   358.0092+-1.4886          357.6292+-2.1206        
   dry.c                                    448.0542+-23.9222         436.5019+-24.7843         might be 1.0265x faster
   FloatMM.c                                732.3499+-3.3762     ?    735.2382+-4.9886        ?
   gcc-loops.cpp                           3710.8789+-4.8850     ?   3711.0612+-6.4472        ?
   n-body.c                                 808.5736+-2.3940          806.8895+-1.5501        
   Quicksort.c                              398.9364+-1.3773     ?    399.2765+-1.5109        ?
   stepanov_container.cpp                  3320.1520+-12.5985        3318.6777+-11.7300       
   Towers.c                                 273.7396+-1.7797          273.2300+-1.4172        

   <geometric>                              726.3499+-4.8021          724.9713+-3.7337          might be 1.0019x faster

                                                  Conf#1                    Conf#2                                      
Geomean of preferred means:
   <scaled-result>                           31.1652+-0.0950           31.1013+-0.0626          might be 1.0021x faster
Comment 3 Geoffrey Garen 2016-04-01 10:58:23 PDT
Comment on attachment 275370 [details]
Patch

r=me
Comment 4 WebKit Commit Bot 2016-04-01 12:20:05 PDT
Comment on attachment 275370 [details]
Patch

Clearing flags on attachment: 275370

Committed r198953: <http://trac.webkit.org/changeset/198953>
Comment 5 WebKit Commit Bot 2016-04-01 12:20:08 PDT
All reviewed patches have been landed.  Closing bug.