Bug 152579 - FTL B3 should be able to run all of the old V8v7 tests
Summary: FTL B3 should be able to run all of the old V8v7 tests
Status: RESOLVED FIXED
Alias: None
Product: WebKit
Classification: Unclassified
Component: JavaScriptCore (show other bugs)
Version: WebKit Nightly Build
Hardware: All All
: P2 Normal
Assignee: Filip Pizlo
URL:
Keywords:
Depends on:
Blocks: 150279
  Show dependency treegraph
 
Reported: 2015-12-28 17:55 PST by Filip Pizlo
Modified: 2016-01-01 22:04 PST (History)
12 users (show)

See Also:


Attachments
the patch (4.12 KB, patch)
2015-12-28 18:00 PST, Filip Pizlo
no flags Details | Formatted Diff | Diff
the patch (7.86 KB, patch)
2015-12-28 20:06 PST, Filip Pizlo
no flags Details | Formatted Diff | Diff
the patch (6.59 KB, patch)
2015-12-28 20:07 PST, Filip Pizlo
no flags Details | Formatted Diff | Diff
the patch (6.77 KB, patch)
2015-12-28 20:08 PST, Filip Pizlo
no flags Details | Formatted Diff | Diff
the patch (6.33 KB, patch)
2015-12-28 20:08 PST, Filip Pizlo
sbarati: review+
Details | Formatted Diff | Diff

Note You need to log in before you can comment on or make changes to this bug.
Description Filip Pizlo 2015-12-28 17:55:02 PST
Patch forthcoming.
Comment 1 Filip Pizlo 2015-12-28 18:00:10 PST
Created attachment 267970 [details]
the patch
Comment 2 Filip Pizlo 2015-12-28 20:06:32 PST
Created attachment 267974 [details]
the patch
Comment 3 Filip Pizlo 2015-12-28 20:07:22 PST
Created attachment 267975 [details]
the patch
Comment 4 Filip Pizlo 2015-12-28 20:08:23 PST
Created attachment 267976 [details]
the patch
Comment 5 Filip Pizlo 2015-12-28 20:08:59 PST
Created attachment 267977 [details]
the patch
Comment 6 WebKit Commit Bot 2015-12-28 20:11:12 PST
Attachment 267977 [details] did not pass style-queue:


ERROR: Source/JavaScriptCore/ftl/FTLB3Output.h:192:  The parameter name "value" adds no information, so it should be removed.  [readability/parameter_name] [5]
Total errors found: 1 in 7 files


If any of these errors are false positives, please file a bug against check-webkit-style.
Comment 7 Filip Pizlo 2015-12-28 20:30:26 PST
With these fixes, we can almost run all of Octane.  The missing piece is some binary snippet that pdfjs uses.


Benchmark report for SunSpider and Octane on shakezilla (MacBookPro11,3).

VMs tested:
"Conf#1" at /Volumes/Data/secondary/OpenSource/WebKitBuild/Release/jsc (r194431)
"Conf#2" at /Volumes/Data/quartary/OpenSource/WebKitBuild/Release/jsc (r194431)

Collected 6 samples per benchmark/VM, with 6 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.

                                          Conf#1                    Conf#2                                      
SunSpider:
   3d-cube                            4.7627+-0.2505            4.7311+-0.3383        
   3d-morph                           5.4824+-0.1328            5.3780+-0.1186          might be 1.0194x faster
   3d-raytrace                        5.2353+-0.1183     ?      5.3410+-0.0713        ? might be 1.0202x slower
   access-binary-trees                2.2360+-0.1027     ?      2.2848+-0.1606        ? might be 1.0218x slower
   access-fannkuch                    5.6672+-0.3503            5.6071+-0.3714          might be 1.0107x faster
   access-nbody                       2.5324+-0.0460     ?      2.9852+-0.6937        ? might be 1.1788x slower
   access-nsieve                      3.3470+-0.2375     ?      3.4620+-0.1418        ? might be 1.0343x slower
   bitops-3bit-bits-in-byte           1.1838+-0.0156            1.1731+-0.0285        
   bitops-bits-in-byte                3.1543+-0.0775     ?      3.1703+-0.0339        ?
   bitops-bitwise-and                 2.0222+-0.0505     ?      2.1941+-0.3710        ? might be 1.0850x slower
   bitops-nsieve-bits                 3.0497+-0.0535     ?      3.1704+-0.2097        ? might be 1.0396x slower
   controlflow-recursive              2.4990+-0.2380            2.4720+-0.2172          might be 1.0109x faster
   crypto-aes                         4.1477+-0.1653     ?      4.6705+-1.3541        ? might be 1.1261x slower
   crypto-md5                         2.8105+-0.8127            2.6470+-0.1065          might be 1.0618x faster
   crypto-sha1                        2.3752+-0.0793     ?      2.4100+-0.1073        ? might be 1.0146x slower
   date-format-tofte                  7.1764+-0.2404            6.9337+-0.4981          might be 1.0350x faster
   date-format-xparb                  4.8194+-0.4084            4.7233+-0.1740          might be 1.0203x faster
   math-cordic                        2.9875+-0.1005            2.9228+-0.0397          might be 1.0222x faster
   math-partial-sums                  4.8293+-0.1177            4.7778+-0.0920          might be 1.0108x faster
   math-spectral-norm                 1.9503+-0.0339     ?      1.9633+-0.0057        ?
   regexp-dna                         6.1565+-0.2353            6.0293+-0.0857          might be 1.0211x faster
   string-base64                      4.5140+-0.2303     ?      4.5533+-0.1427        ?
   string-fasta                       6.1608+-0.5222            5.6602+-0.0546          might be 1.0885x faster
   string-tagcloud                    8.0947+-0.4238            7.8454+-0.1006          might be 1.0318x faster
   string-unpack-code                18.3186+-0.4630     ?     18.8571+-1.9620        ? might be 1.0294x slower
   string-validate-input              5.2292+-2.3270            4.3978+-0.1639          might be 1.1890x faster

   <arithmetic>                       4.6439+-0.0608            4.6292+-0.1105          might be 1.0032x faster

                                          Conf#1                    Conf#2                                      
Octane:
   encrypt                           0.15747+-0.00375    !     0.18074+-0.00209       ! definitely 1.1478x slower
   decrypt                           2.89590+-0.02773    !     3.22444+-0.04698       ! definitely 1.1135x slower
   deltablue                x2       0.13536+-0.00260    !     0.15556+-0.00897       ! definitely 1.1492x slower
   earley                            0.28128+-0.00260    ?     0.28321+-0.00157       ?
   boyer                             4.27392+-0.03602    !     4.39448+-0.06543       ! definitely 1.0282x slower
   navier-stokes            x2       4.82600+-0.03717    !     4.94272+-0.03744       ! definitely 1.0242x slower
   raytrace                 x2       0.85101+-0.00512    !     0.88857+-0.00483       ! definitely 1.0441x slower
   richards                 x2       0.08786+-0.00150    !     0.10193+-0.00157       ! definitely 1.1601x slower
   splay                    x2       0.35165+-0.00142    ?     0.35598+-0.00756       ? might be 1.0123x slower
   regexp                   x2      24.35582+-0.26063         24.28389+-0.26061       
   pdfjs                    x2      38.20492+-0.56883               ERROR             
   mandreel                 x2      43.39686+-0.63562    !    54.05068+-0.41852       ! definitely 1.2455x slower
   gbemu                    x2      29.29804+-0.22415    ?    33.68748+-4.19246       ? might be 1.1498x slower
   closure                           0.58824+-0.00600    ?     0.58916+-0.00412       ?
   jquery                            7.39158+-0.08692    ?     7.43347+-0.05415       ?
   box2d                    x2       9.19709+-0.16573    ?     9.34087+-0.08174       ? might be 1.0156x slower
   zlib                     x2     378.24365+-20.14061   ?   402.73398+-7.39207       ? might be 1.0647x slower
   typescript               x2     672.37345+-11.21920       658.66976+-5.46906         might be 1.0208x faster

   <geometric>                       5.30050+-0.02704               ERROR             

                                          Conf#1                    Conf#2                                      
Geomean of preferred means:
   <scaled-result>                   15.6889+-0.1093                ERROR
Comment 8 Saam Barati 2016-01-01 11:29:54 PST
Comment on attachment 267977 [details]
the patch

r=me
Comment 9 Filip Pizlo 2016-01-01 22:04:12 PST
Landed in http://trac.webkit.org/changeset/194493