Bug 162132 - [JSC][LLInt] Introduce is_cell_with_type
Summary: [JSC][LLInt] Introduce is_cell_with_type
Status: RESOLVED FIXED
Alias: None
Product: WebKit
Classification: Unclassified
Component: JavaScriptCore (show other bugs)
Version: WebKit Nightly Build
Hardware: Unspecified Unspecified
: P2 Normal
Assignee: Yusuke Suzuki
URL:
Keywords:
Depends on:
Blocks: 162142
  Show dependency tree / graph
 
Reported: 2016-09-17 23:18 PDT by Yusuke Suzuki
Modified: 2016-09-19 16:29 PDT (History)
5 users (show)

See Also:

Attachments
Patch (35.05 KB, patch)
2016-09-18 23:50 PDT, Yusuke Suzuki 		no flags Details | Formatted Diff | Diff
Patch (35.93 KB, patch)
2016-09-18 23:53 PDT, Yusuke Suzuki 		no flags Details | Formatted Diff | Diff
Patch (35.87 KB, patch)
2016-09-19 01:06 PDT, Yusuke Suzuki 		sam: review+
Details | Formatted Diff | Diff
Show Obsolete (2) 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 Yusuke Suzuki 2016-09-17 23:18:04 PDT 
If it does not pose any perf regression!
Comment 1 Yusuke Suzuki 2016-09-18 23:15:53 PDT 
Quick note: is-array-xxx microbenchmarks does not pose any perf regression.
Comment 2 Yusuke Suzuki 2016-09-18 23:20:31 PDT 
Kraken 10x does not pose any perf regression.
                                           baseline                  patched                                      

ai-astar                               104.779+-5.396            104.388+-7.231         
audio-beat-detection                    43.249+-0.479      ?      45.301+-3.136         ? might be 1.0475x slower
audio-dft                              122.091+-0.207            121.975+-0.120         
audio-fft                               36.526+-2.884             35.894+-0.336           might be 1.0176x faster
audio-oscillator                        50.530+-0.125      ?      50.622+-0.245         ?
imaging-darkroom                        87.718+-0.181             87.644+-0.127         
imaging-desaturate                      55.689+-0.057      ?      55.714+-0.089         ?
imaging-gaussian-blur                   79.802+-5.441             76.817+-13.184          might be 1.0389x faster
json-parse-financial                    45.172+-1.699             44.412+-0.483           might be 1.0171x faster
json-stringify-tinderbox                27.742+-2.602             26.060+-0.269           might be 1.0645x faster
stanford-crypto-aes                     41.248+-0.297      ?      41.319+-0.517         ?
stanford-crypto-ccm                     40.269+-3.120             40.129+-0.829         
stanford-crypto-pbkdf2                  99.160+-2.688      ?      99.512+-2.383         ?
stanford-crypto-sha256-iterative        33.518+-0.055      ^      33.391+-0.071         ^ definitely 1.0038x faster

<arithmetic>                            61.964+-0.517             61.656+-1.027           might be 1.0050x faster
Comment 3 Yusuke Suzuki 2016-09-18 23:32:06 PDT 
OK! Octane also does not pose perf regressions!

                            baseline                  patched                                      

encrypt                 0.18176+-0.00113          0.18137+-0.00095       
decrypt                 2.94496+-0.02083          2.94092+-0.01175       
deltablue      x2       0.14056+-0.00248    ?     0.14811+-0.01230       ? might be 1.0537x slower
earley                  0.27780+-0.00071    ?     0.27826+-0.00056       ?
boyer                   4.72845+-0.01276          4.72487+-0.00844       
navier-stokes  x2       4.78319+-0.01079    ?     4.79980+-0.03981       ?
raytrace       x2       0.82850+-0.01110    ?     0.83306+-0.00671       ?
richards       x2       0.08974+-0.00250    ?     0.09145+-0.00223       ? might be 1.0190x slower
splay          x2       0.36832+-0.00193    ?     0.37042+-0.00178       ?
regexp         x2      19.45685+-0.13716    ?    19.61233+-0.12050       ?
pdfjs          x2      42.33251+-0.49102    ?    43.09881+-0.87100       ? might be 1.0181x slower
mandreel       x2      50.41776+-1.02654         49.49966+-0.36287         might be 1.0185x faster
gbemu          x2      35.68241+-0.56086    ?    36.03761+-0.88023       ?
closure                 0.62445+-0.01067          0.61458+-0.00300         might be 1.0161x faster
jquery                  8.15364+-0.19368          8.05949+-0.04403         might be 1.0117x faster
box2d          x2      14.47267+-0.16801    ?    14.60825+-0.14110       ?
zlib           x2     365.77266+-5.31775        357.42082+-5.87320         might be 1.0234x faster
typescript     x2     837.10984+-11.77692       835.97700+-13.96924      

<geometric>             5.71762+-0.01958    ?     5.74389+-0.03770       ? might be 1.0046x slower
Comment 4 Yusuke Suzuki 2016-09-18 23:50:41 PDT 
Created attachment 289217 [details]
Patch
Comment 5 Yusuke Suzuki 2016-09-18 23:51:12 PDT 
Comment on attachment 289217 [details]
Patch

I'll perform rebaseline.
Comment 6 Yusuke Suzuki 2016-09-18 23:53:18 PDT 
Created attachment 289218 [details]
Patch
Comment 7 Yusuke Suzuki 2016-09-18 23:54:31 PDT 
OK, ready for review!
Comment 8 Yusuke Suzuki 2016-09-19 01:06:16 PDT 
Created attachment 289219 [details]
Patch
Comment 9 Sam Weinig 2016-09-19 07:09:47 PDT 
Comment on attachment 289219 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=289219&action=review

> Source/JavaScriptCore/ChangeLog:11
> +        This automatically offers optimization to previous IsSring node: dropping cell check by using CellUse edge filter.

Typo, IsSring -> IsString
Comment 10 Yusuke Suzuki 2016-09-19 09:55:52 PDT 
Comment on attachment 289219 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=289219&action=review

Thanks!

>> Source/JavaScriptCore/ChangeLog:11
>> +        This automatically offers optimization to previous IsSring node: dropping cell check by using CellUse edge filter.
> 
> Typo, IsSring -> IsString

Oops! Fixed.
Comment 11 Yusuke Suzuki 2016-09-19 10:02:40 PDT 
Committed r206098: <http://trac.webkit.org/changeset/206098>
Comment 12 Yusuke Suzuki 2016-09-19 16:29:45 PDT 
Committed r206128: <http://trac.webkit.org/changeset/206128>