108414 – DFG should allow phases to break Phi's and then have one phase to rebuild them

Bug 108414 - DFG should allow phases to break Phi's and then have one phase to rebuild them

Summary: DFG should allow phases to break Phi's and then have one phase to rebuild them

Status:	RESOLVED FIXED

Alias:	None

Product:	WebKit
Classification:	Unclassified
Component:	JavaScriptCore (show other bugs)
Version:	528+ (Nightly build)
Hardware:	All All

Importance:	P2 Normal
Assignee:	Filip Pizlo

URL:
Keywords:

Depends on:	108417 108868 109000
Blocks:
	Show dependency tree / graph

Reported:	2013-01-30 18:18 PST by Filip Pizlo
Modified:	2013-02-09 11:35 PST (History)
CC List:	10 users (show)

See Also:

Attachments
it begins (20.96 KB, patch) 2013-01-30 18:19 PST, Filip Pizlo	no flags	Details \| Formatted Diff \| Diff
work in progress (31.98 KB, patch) 2013-01-31 13:59 PST, Filip Pizlo	no flags	Details \| Formatted Diff \| Diff
work in progress (57.45 KB, patch) 2013-01-31 16:10 PST, Filip Pizlo	no flags	Details \| Formatted Diff \| Diff
work in progress (82.74 KB, patch) 2013-02-01 01:35 PST, Filip Pizlo	no flags	Details \| Formatted Diff \| Diff
it might be done (95.02 KB, patch) 2013-02-01 16:02 PST, Filip Pizlo	no flags	Details \| Formatted Diff \| Diff
it builds! (106.00 KB, patch) 2013-02-01 17:25 PST, Filip Pizlo	no flags	Details \| Formatted Diff \| Diff
it runs 3d-cube! (107.94 KB, patch) 2013-02-02 14:13 PST, Filip Pizlo	no flags	Details \| Formatted Diff \| Diff
passes all of the simple tests (109.50 KB, patch) 2013-02-03 16:22 PST, Filip Pizlo	no flags	Details \| Formatted Diff \| Diff
It runs all of the things (111.08 KB, patch) 2013-02-03 20:31 PST, Filip Pizlo	no flags	Details \| Formatted Diff \| Diff
improved performance (113.72 KB, patch) 2013-02-03 23:15 PST, Filip Pizlo	no flags	Details \| Formatted Diff \| Diff
it might be done (125.66 KB, patch) 2013-02-07 13:26 PST, Filip Pizlo	no flags	Details \| Formatted Diff \| Diff
almost there (133.23 KB, patch) 2013-02-07 14:53 PST, Filip Pizlo	no flags	Details \| Formatted Diff \| Diff
the patch (138.24 KB, patch) 2013-02-08 16:55 PST, Filip Pizlo	mhahnenberg: review+	Details \| Formatted Diff \| Diff
Show Obsolete (12) View All Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.

Description Filip Pizlo 2013-01-30 18:18:20 PST

This will be painful.

Comment 1 Filip Pizlo 2013-01-30 18:19:10 PST

Created attachment 185646 [details]
it begins

Comment 2 Filip Pizlo 2013-01-31 13:59:51 PST

Created attachment 185856 [details]
work in progress

I believe that I've written a completely sensible and easy-to-understand Phi simplifier.  Now I need to carefully glue it in and hopefully not get a massive performance regression in the process...

Comment 3 Filip Pizlo 2013-01-31 16:10:50 PST

Created attachment 185888 [details]
work in progress

Slowly weaning other phases off of their Phi maintenance addiction.  Also starting to work on the unification phase.

Comment 4 Filip Pizlo 2013-02-01 01:35:02 PST

Created attachment 185975 [details]
work in progress

Comment 5 Filip Pizlo 2013-02-01 16:02:55 PST

Created attachment 186179 [details]
it might be done

Now I just have to compile it...

Comment 6 Filip Pizlo 2013-02-01 17:25:41 PST

Created attachment 186192 [details]
it builds!

Comment 7 Filip Pizlo 2013-02-02 14:13:54 PST

Created attachment 186237 [details]
it runs 3d-cube!

It can run some things now, like the 3d-cube benchmark, which is the most important benchmark, ever.

Soon, it will run other things, too.

Comment 8 Filip Pizlo 2013-02-03 16:22:06 PST

Created attachment 186271 [details]
passes all of the simple tests

Still need more debugging.

Comment 9 Filip Pizlo 2013-02-03 20:31:56 PST

Created attachment 186292 [details]
It runs all of the things

I'm still working on performance issues.

Comment 10 Filip Pizlo 2013-02-03 23:15:38 PST

Created attachment 186313 [details]
improved performance

Reintroduced some of the optimizations that the bytecode parser's phi management had: in particular we template-specialize so that we don't have to branch on operand kind.

Comment 11 Filip Pizlo 2013-02-07 13:26:46 PST

Created attachment 187166 [details]
it might be done

Comment 12 Filip Pizlo 2013-02-07 14:53:03 PST

Created attachment 187181 [details]
almost there

Still need to do build system hackage.

Comment 13 Filip Pizlo 2013-02-08 16:55:33 PST

Created attachment 187387 [details]
the patch

Comment 14 WebKit Review Bot 2013-02-08 16:58:41 PST

Attachment 187387 [details] did not pass style-queue:

Failed to run "['Tools/Scripts/check-webkit-style', '--diff-files', u'Source/JavaScriptCore/CMakeLists.txt', u'Source/JavaScriptCore/ChangeLog', u'Source/JavaScriptCore/GNUmakefile.list.am', u'Source/JavaScriptCore/JavaScriptCore.xcodeproj/project.pbxproj', u'Source/JavaScriptCore/Target.pri', u'Source/JavaScriptCore/bytecode/Operands.h', u'Source/JavaScriptCore/dfg/DFGAbstractState.cpp', u'Source/JavaScriptCore/dfg/DFGAllocator.h', u'Source/JavaScriptCore/dfg/DFGArgumentsSimplificationPhase.cpp', u'Source/JavaScriptCore/dfg/DFGBasicBlockInlines.h', u'Source/JavaScriptCore/dfg/DFGByteCodeParser.cpp', u'Source/JavaScriptCore/dfg/DFGCFGSimplificationPhase.cpp', u'Source/JavaScriptCore/dfg/DFGCPSRethreadingPhase.cpp', u'Source/JavaScriptCore/dfg/DFGCPSRethreadingPhase.h', u'Source/JavaScriptCore/dfg/DFGCSEPhase.cpp', u'Source/JavaScriptCore/dfg/DFGCommon.cpp', u'Source/JavaScriptCore/dfg/DFGCommon.h', u'Source/JavaScriptCore/dfg/DFGConstantFoldingPhase.cpp', u'Source/JavaScriptCore/dfg/DFGDriver.cpp', u'Source/JavaScriptCore/dfg/DFGGraph.cpp', u'Source/JavaScriptCore/dfg/DFGGraph.h', u'Source/JavaScriptCore/dfg/DFGNode.h', u'Source/JavaScriptCore/dfg/DFGNodeType.h', u'Source/JavaScriptCore/dfg/DFGPhase.cpp', u'Source/JavaScriptCore/dfg/DFGPhase.h', u'Source/JavaScriptCore/dfg/DFGPredictionInjectionPhase.cpp', u'Source/JavaScriptCore/dfg/DFGPredictionInjectionPhase.h', u'Source/JavaScriptCore/dfg/DFGPredictionPropagationPhase.cpp', u'Source/JavaScriptCore/dfg/DFGSpeculativeJIT32_64.cpp', u'Source/JavaScriptCore/dfg/DFGSpeculativeJIT64.cpp', u'Source/JavaScriptCore/dfg/DFGStructureCheckHoistingPhase.cpp', u'Source/JavaScriptCore/dfg/DFGUnificationPhase.cpp', u'Source/JavaScriptCore/dfg/DFGUnificationPhase.h', u'Source/JavaScriptCore/dfg/DFGValidate.cpp', u'Source/JavaScriptCore/dfg/DFGVirtualRegisterAllocationPhase.cpp', u'Source/JavaScriptCore/llint/LLIntSlowPaths.cpp', u'Source/JavaScriptCore/runtime/JSCJSValue.cpp', u'Source/JavaScriptCore/runtime/JSString.h', u'Source/JavaScriptCore/runtime/Options.h']" exit_code: 1
Source/JavaScriptCore/dfg/DFGArgumentsSimplificationPhase.cpp:242:  One space before end of line comments  [whitespace/comments] [5]
Source/JavaScriptCore/dfg/DFGArgumentsSimplificationPhase.cpp:242:  Should have a space between // and comment  [whitespace/comments] [4]
Source/JavaScriptCore/dfg/DFGCPSRethreadingPhase.cpp:158:  An else statement can be removed when the prior "if" concludes with a return, break, continue or goto statement.  [readability/control_flow] [4]
Source/JavaScriptCore/dfg/DFGCPSRethreadingPhase.cpp:432:  Weird number of spaces at line-start.  Are you using a 4-space indent?  [whitespace/indent] [3]
Source/JavaScriptCore/dfg/DFGConstantFoldingPhase.cpp:333:  An else statement can be removed when the prior "if" concludes with a return, break, continue or goto statement.  [readability/control_flow] [4]
Total errors found: 5 in 39 files


If any of these errors are false positives, please file a bug against check-webkit-style.

Comment 15 Mark Hahnenberg 2013-02-08 18:19:54 PST

Comment on attachment 187387 [details]
the patch

View in context: https://bugs.webkit.org/attachment.cgi?id=187387&action=review

r=me. Just for due diligence, what's the performance numbers for this currently?

> Source/JavaScriptCore/dfg/DFGCPSRethreadingPhase.cpp:342
> +            // following from from predecessor blocks: SetLocal, Phi, or SetArgument.

from from

> Source/JavaScriptCore/dfg/DFGGraph.h:-83
> -    

Stray whitespace.

Comment 16 Filip Pizlo 2013-02-08 21:32:57 PST

Performance:


Benchmark report for SunSpider, V8Spider, Octane, Kraken, and JSRegress on wartooth (MacBookPro8,2).

VMs tested:
"TipOfTree" at /Volumes/Data/pizlo/quinary/OpenSource/WebKitBuild/Release/jsc (r142321)
"CPSRethreading" at /Volumes/Data/pizlo/secondary/OpenSource/WebKitBuild/Release/jsc (r142321)

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.

                                                     TipOfTree               CPSRethreading                                  
SunSpider:
   3d-cube                                         7.3513+-0.0962     !      7.5958+-0.0772        ! definitely 1.0333x slower
   3d-morph                                        7.2424+-0.0946            7.2288+-0.0586        
   3d-raytrace                                     8.1068+-0.1291     ?      8.3262+-0.0990        ? might be 1.0271x slower
   access-binary-trees                             1.4855+-0.0513     ?      1.5246+-0.0969        ? might be 1.0263x slower
   access-fannkuch                                 5.7826+-0.0783     ?      5.9039+-0.1111        ? might be 1.0210x slower
   access-nbody                                    3.2072+-0.0785     ?      3.2796+-0.0757        ? might be 1.0226x slower
   access-nsieve                                   3.5960+-0.0697     ?      3.6101+-0.0991        ?
   bitops-3bit-bits-in-byte                        1.7174+-0.0334     ?      1.7606+-0.0776        ? might be 1.0251x slower
   bitops-bits-in-byte                             2.2026+-0.0260            2.1293+-0.0556          might be 1.0344x faster
   bitops-bitwise-and                              2.0469+-0.0497     ?      2.1184+-0.1290        ? might be 1.0349x slower
   bitops-nsieve-bits                              3.3347+-0.0828     ?      3.3657+-0.0514        ?
   controlflow-recursive                           2.0731+-0.0206            2.0585+-0.0218        
   crypto-aes                                      6.1142+-0.0730     !      6.4354+-0.0943        ! definitely 1.0525x slower
   crypto-md5                                      2.9443+-0.0919     ?      3.0622+-0.1059        ? might be 1.0400x slower
   crypto-sha1                                     2.5351+-0.0664     !      2.6746+-0.0719        ! definitely 1.0551x slower
   date-format-tofte                              11.8771+-0.2156     ?     11.9804+-0.2246        ?
   date-format-xparb                              12.1287+-0.2728           11.9756+-0.1811          might be 1.0128x faster
   math-cordic                                     3.2758+-0.0850            3.2494+-0.0845        
   math-partial-sums                               7.6486+-0.1564            7.6229+-0.1564        
   math-spectral-norm                              2.5782+-0.0682            2.4860+-0.0514          might be 1.0371x faster
   regexp-dna                                      7.3667+-0.0854     ?      7.4244+-0.0844        ?
   string-base64                                   4.1721+-0.0956            4.0864+-0.0669          might be 1.0210x faster
   string-fasta                                    7.5388+-0.1207            7.4819+-0.1130        
   string-tagcloud                                10.9886+-0.2348     ?     11.0610+-0.1851        ?
   string-unpack-code                             19.0560+-0.2886           19.0204+-0.2474        
   string-validate-input                           5.9433+-0.1340     ?      6.0162+-0.1287        ? might be 1.0123x slower

   <arithmetic> *                                  5.8582+-0.0486     ?      5.9030+-0.0384        ? might be 1.0076x slower
   <geometric>                                     4.6869+-0.0408     ?      4.7303+-0.0409        ? might be 1.0093x slower
   <harmonic>                                      3.7928+-0.0357     ?      3.8295+-0.0462        ? might be 1.0097x slower

                                                     TipOfTree               CPSRethreading                                  
V8Spider:
   crypto                                         60.0306+-0.5406     !     61.5250+-0.6450        ! definitely 1.0249x slower
   deltablue                                      93.6007+-1.3571     !     95.7814+-0.4248        ! definitely 1.0233x slower
   earley-boyer                                   56.4647+-1.3491     ?     57.1343+-0.2794        ? might be 1.0119x slower
   raytrace                                       48.3663+-0.5129     ?     48.7322+-0.3738        ?
   regexp                                         70.1942+-0.4902     ?     70.5960+-0.9391        ?
   richards                                       82.9286+-0.3618     ?     83.2773+-0.6056        ?
   splay                                          40.8410+-0.2034           40.6449+-0.2960        

   <arithmetic>                                   64.6323+-0.2715     !     65.3844+-0.2346        ! definitely 1.0116x slower
   <geometric> *                                  62.3043+-0.2590     !     62.9541+-0.2444        ! definitely 1.0104x slower
   <harmonic>                                     60.0599+-0.2682     !     60.6024+-0.2542        ! definitely 1.0090x slower

                                                     TipOfTree               CPSRethreading                                  
Octane and V8v7:
   encrypt                                        0.32440+-0.00131    ?     0.32445+-0.00324       ?
   decrypt                                        5.67955+-0.01808          5.67041+-0.04890       
   deltablue                             x2       0.41085+-0.00128    ?     0.41176+-0.00142       ?
   earley                                         0.61496+-0.00440    ^     0.60046+-0.00208       ^ definitely 1.0242x faster
   boyer                                          7.99015+-0.01987    ^     7.95597+-0.01227       ^ definitely 1.0043x faster
   raytrace                              x2       3.35482+-0.00664          3.34612+-0.01250       
   regexp                                x2      22.59610+-0.10528         22.37049+-0.17507         might be 1.0101x faster
   richards                              x2       0.20972+-0.00124    ^     0.20724+-0.00054       ^ definitely 1.0120x faster
   splay                                 x2       0.56772+-0.02342    ?     0.59818+-0.02702       ? might be 1.0536x slower
   navier-stokes                         x2       8.40020+-0.02680    ?     8.60942+-0.22640       ? might be 1.0249x slower
   closure                                        0.39911+-0.00308    ?     0.40018+-0.00352       ?
   jquery                                         5.59731+-0.07091          5.49686+-0.09469         might be 1.0183x faster
   gbemu                                 x2     175.50045+-11.90205   ?   175.73621+-11.01902      ?
   box2d                                 x2      25.70952+-0.44401    ?    26.08464+-0.27045       ? might be 1.0146x slower

V8v7:
   <arithmetic>                                   5.35549+-0.01409          5.35236+-0.04041         might be 1.0006x faster
   <geometric> *                                  1.76354+-0.00948    ?     1.77209+-0.01286       ? might be 1.0049x slower
   <harmonic>                                     0.67025+-0.00475    ?     0.67129+-0.00453       ? might be 1.0016x slower

Octane including V8v7:
   <arithmetic>                                  22.45928+-1.08070    ?    22.50802+-1.00030       ? might be 1.0022x slower
   <geometric> *                                  3.36544+-0.02018    ?     3.38004+-0.02112       ? might be 1.0043x slower
   <harmonic>                                     0.82564+-0.00537    ?     0.82693+-0.00514       ? might be 1.0016x slower

                                                     TipOfTree               CPSRethreading                                  
Kraken:
   ai-astar                                       279.159+-1.644            277.098+-1.723         
   audio-beat-detection                           177.951+-1.766      ^     174.559+-1.516         ^ definitely 1.0194x faster
   audio-dft                                      246.525+-2.334      ?     247.634+-3.464         ?
   audio-fft                                      101.516+-0.934      ?     101.654+-0.668         ?
   audio-oscillator                               179.380+-2.052      ?     182.280+-2.190         ? might be 1.0162x slower
   imaging-darkroom                               207.871+-0.942      ?     211.039+-2.750         ? might be 1.0152x slower
   imaging-desaturate                             104.979+-0.669      ^     103.324+-0.229         ^ definitely 1.0160x faster
   imaging-gaussian-blur                          289.391+-2.327            288.607+-1.083         
   json-parse-financial                            54.813+-0.395             54.742+-0.748         
   json-stringify-tinderbox                        69.497+-1.918      ?      69.525+-0.395         ?
   stanford-crypto-aes                             69.130+-0.620      ?      69.707+-0.813         ?
   stanford-crypto-ccm                             75.030+-3.241      ?      75.516+-3.416         ?
   stanford-crypto-pbkdf2                         191.796+-2.000      ^     187.433+-0.847         ^ definitely 1.0233x faster
   stanford-crypto-sha256-iterative                82.991+-1.399             81.974+-0.482           might be 1.0124x faster

   <arithmetic> *                                 152.145+-0.465            151.792+-0.639           might be 1.0023x faster
   <geometric>                                    131.061+-0.584            130.790+-0.713           might be 1.0021x faster
   <harmonic>                                     112.669+-0.645            112.518+-0.809           might be 1.0013x faster

                                                     TipOfTree               CPSRethreading                                  
JSRegress:
   adapt-to-double-divide                         75.3465+-0.4350           75.2567+-0.2146        
   aliased-arguments-getbyval                      0.7673+-0.0058     !      0.7949+-0.0084        ! definitely 1.0360x slower
   allocate-big-object                             1.8407+-0.0228     ?      1.8958+-0.0434        ? might be 1.0300x slower
   arity-mismatch-inlining                         0.6015+-0.0110     ?      0.6078+-0.0065        ? might be 1.0106x slower
   array-access-polymorphic-structure             43.8935+-0.3524           43.6009+-0.2770        
   array-with-double-add                           4.1707+-0.0995            4.1263+-0.1007          might be 1.0108x faster
   array-with-double-increment                     3.5498+-0.1178     ?      3.6295+-0.0799        ? might be 1.0225x slower
   array-with-double-mul-add                       4.7339+-0.1115     ?      4.8435+-0.0807        ? might be 1.0231x slower
   array-with-double-sum                           4.5948+-0.0981     ?      4.6396+-0.0927        ?
   array-with-int32-add-sub                        7.4036+-0.1119     ?      7.4149+-0.1027        ?
   array-with-int32-or-double-sum                  4.6357+-0.0733     ?      4.7393+-0.0743        ? might be 1.0224x slower
   big-int-mul                                     3.7645+-0.0943     ?      3.8126+-0.0784        ? might be 1.0128x slower
   boolean-test                                    3.2422+-0.0849            3.2396+-0.0997        
   cast-int-to-double                             11.4394+-0.1556     ?     11.6818+-0.4150        ? might be 1.0212x slower
   cell-argument                                  12.8855+-0.1206           12.8213+-0.1928        
   cfg-simplify                                    3.0132+-0.0751            2.9762+-0.0638          might be 1.0124x faster
   cmpeq-obj-to-obj-other                          9.0928+-0.1743            8.8768+-0.1384          might be 1.0243x faster
   constant-test                                   4.8821+-0.0918     ?      4.9113+-0.0754        ?
   direct-arguments-getbyval                       0.6972+-0.0059     !      0.7237+-0.0075        ! definitely 1.0380x slower
   double-pollution-getbyval                       9.2472+-0.0666     ?      9.3956+-0.1112        ? might be 1.0160x slower
   double-pollution-putbyoffset                    3.9324+-0.0683     ?      3.9327+-0.1081        ?
   external-arguments-getbyval                     1.6322+-0.0266            1.6288+-0.0292        
   external-arguments-putbyval                     2.4096+-0.0617     ?      2.4294+-0.0451        ?
   Float32Array-matrix-mult                        9.4745+-0.0996            9.3447+-0.1307          might be 1.0139x faster
   fold-double-to-int                             17.1006+-0.4984           16.6145+-0.1413          might be 1.0293x faster
   function-dot-apply                              2.1912+-0.0274            2.1897+-0.0425        
   function-test                                   3.2597+-0.0610            3.2355+-0.0537        
   get-by-id-chain-from-try-block                  5.2250+-0.1271            5.1440+-0.1210          might be 1.0158x faster
   indexed-properties-in-objects                   3.2903+-0.0792     ?      3.3470+-0.0729        ? might be 1.0173x slower
   inline-arguments-access                         1.0281+-0.0104     !      1.0463+-0.0072        ! definitely 1.0176x slower
   inline-arguments-local-escape                  15.6365+-0.2735           15.2822+-0.2337          might be 1.0232x faster
   inline-get-scoped-var                           5.3761+-0.0750            5.3507+-0.0734        
   inlined-put-by-id-transition                    9.9811+-0.1655     ?     10.0022+-0.0992        ?
   int-or-other-abs-then-get-by-val                6.8123+-0.1196            6.7650+-0.1050        
   int-or-other-abs-zero-then-get-by-val          26.6592+-0.2342     !     27.5452+-0.2677        ! definitely 1.0332x slower
   int-or-other-add-then-get-by-val                8.2947+-0.1519     ?      8.2955+-0.1063        ?
   int-or-other-add                                7.5124+-0.1262     ?      7.5786+-0.1138        ?
   int-or-other-div-then-get-by-val                6.3039+-0.4014     ?      6.4763+-0.4647        ? might be 1.0273x slower
   int-or-other-max-then-get-by-val                7.0281+-0.0847            6.9958+-0.1307        
   int-or-other-min-then-get-by-val                6.8951+-0.1001     ?      6.9659+-0.1008        ? might be 1.0103x slower
   int-or-other-mod-then-get-by-val                5.8966+-0.0766     ?      5.9336+-0.0612        ?
   int-or-other-mul-then-get-by-val                5.8023+-0.0775     ?      5.8165+-0.1070        ?
   int-or-other-neg-then-get-by-val                5.9645+-0.0816     ?      6.0107+-0.0797        ?
   int-or-other-neg-zero-then-get-by-val          25.9048+-0.3786     ?     26.4683+-0.5408        ? might be 1.0218x slower
   int-or-other-sub-then-get-by-val                7.9825+-0.1494     ?      7.9956+-0.1221        ?
   int-or-other-sub                                7.1840+-0.1035            7.1656+-0.1249        
   int-overflow-local                            197.5204+-0.8437          196.9987+-1.0043        
   Int16Array-bubble-sort                         24.5623+-0.5116     ?     24.6167+-0.4209        ?
   Int16Array-load-int-mul                         1.4892+-0.0161     ?      1.5082+-0.0093        ? might be 1.0128x slower
   Int8Array-load                                  3.6639+-0.0890     ?      3.7345+-0.0597        ? might be 1.0193x slower
   integer-divide                                 13.0407+-0.4650           12.9980+-0.4247        
   integer-modulo                                  1.5775+-0.0201     ?      1.5900+-0.0139        ?
   make-indexed-storage                            2.8866+-0.0609     ?      2.9906+-0.0973        ? might be 1.0360x slower
   method-on-number                              187.3597+-2.5455          185.6352+-1.9131        
   new-array-buffer-dead                           2.8265+-0.0582            2.8215+-0.1022        
   new-array-buffer-push                           7.2039+-0.0722            7.2025+-0.0520        
   new-array-dead                                 22.2644+-0.1936     ?     22.6008+-0.2406        ? might be 1.0151x slower
   new-array-push                                  4.9765+-0.0697            4.9400+-0.0368        
   number-test                                     3.1589+-0.0858     ?      3.2116+-0.0615        ? might be 1.0167x slower
   object-closure-call                             6.7482+-0.0781            6.5998+-0.0932          might be 1.0225x faster
   object-test                                     3.4391+-0.0843     ?      3.5332+-0.0652        ? might be 1.0274x slower
   poly-stricteq                                  73.5902+-0.5459           73.2536+-0.3875        
   polymorphic-structure                          14.8530+-0.1938           14.8409+-0.1913        
   polyvariant-monomorphic-get-by-id              10.4995+-0.5741     ?     10.6131+-0.5376        ? might be 1.0108x slower
   rare-osr-exit-on-local                        152.3211+-0.7098     ?    153.1462+-0.7105        ?
   register-pressure-from-osr                     24.6226+-0.2096           24.4576+-0.2020        
   simple-activation-demo                         27.4539+-0.2322           27.2610+-0.2945        
   slow-array-profile-convergence                  3.3130+-0.0759     ?      3.3945+-0.0818        ? might be 1.0246x slower
   slow-convergence                               58.8908+-0.5100           58.3886+-0.3044        
   sparse-conditional                              0.9993+-0.0045     ?      1.0072+-0.0135        ?
   splice-to-remove                               39.1461+-0.2717     ?     39.2787+-0.3761        ?
   string-hash                                     2.1361+-0.0370            2.1073+-0.0492          might be 1.0137x faster
   string-repeat-arith                            32.4702+-0.4610     ?     32.6677+-0.3229        ?
   string-sub                                     64.9460+-0.4867     ?     65.8427+-0.5323        ? might be 1.0138x slower
   string-test                                     3.0620+-0.0771     ?      3.0846+-0.0809        ?
   structure-hoist-over-transitions                2.5320+-0.0470            2.5118+-0.0489        
   tear-off-arguments-simple                       1.4474+-0.0174     ?      1.4628+-0.0173        ? might be 1.0107x slower
   tear-off-arguments                              2.6686+-0.0568     ?      2.7120+-0.0536        ? might be 1.0163x slower
   temporal-structure                             15.3105+-0.1692     !     15.6571+-0.0916        ! definitely 1.0226x slower
   to-int32-boolean                               19.9505+-0.1784     ?     19.9568+-0.1780        ?
   undefined-test                                  3.2583+-0.0948            3.1846+-0.0662          might be 1.0231x faster

   <arithmetic>                                   17.9601+-0.0570     ?     17.9674+-0.0520        ? might be 1.0004x slower
   <geometric> *                                   7.0863+-0.0397     ?      7.1200+-0.0395        ? might be 1.0048x slower
   <harmonic>                                      3.7870+-0.0191     !      3.8273+-0.0206        ! definitely 1.0106x slower

                                                     TipOfTree               CPSRethreading                                  
All benchmarks:
   <arithmetic>                                   31.2243+-0.2229     ?     31.2453+-0.2274        ? might be 1.0007x slower
   <geometric>                                     8.5938+-0.0467     ?      8.6376+-0.0458        ? might be 1.0051x slower
   <harmonic>                                      2.7206+-0.0097     ?      2.7372+-0.0107        ? might be 1.0061x slower

                                                     TipOfTree               CPSRethreading                                  
Geomean of preferred means:
   <scaled-result>                                16.7647+-0.0757     ?     16.8478+-0.0688        ? might be 1.0050x slower

Comment 17 Filip Pizlo 2013-02-09 11:35:08 PST

Landed in http://trac.webkit.org/changeset/142377