Bug 157814 - REGRESSION: JetStream crashes on some iPhones
Summary: REGRESSION: JetStream crashes on some iPhones
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: Geoffrey Garen
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2016-05-17 15:15 PDT by Geoffrey Garen
Modified: 2016-05-17 15:26 PDT (History)
2 users (show)

See Also:


Attachments
Patch (1.87 KB, patch)
2016-05-17 15:20 PDT, Geoffrey Garen
msaboff: review+
Details | Formatted Diff | Diff

Note You need to log in before you can comment on or make changes to this bug.
Description Geoffrey Garen 2016-05-17 15:15:59 PDT
REGRESSION: JetStream crashes on some iPhones
Comment 1 Geoffrey Garen 2016-05-17 15:20:46 PDT
Created attachment 279170 [details]
Patch
Comment 2 Geoffrey Garen 2016-05-17 15:21:34 PDT
Here's the memory win:

~/OpenSource/PerformanceTests/MallocBench> ./run-malloc-benchmarks --benchmark "jetstream --runs 1" Baseline:/Users/ggaren/OpenSource/WebKitBuildBaseline/ Patch:/Users/ggaren/OpenSource/WebKitBuild/Release/
                                                                                
                                    Baseline                 Patch                     Δ
Execution Time:
    jetstream --runs 1               5,269ms               5,285ms         ! 1.0x slower

    <geometric mean>                 5,269ms               5,285ms         ! 1.0x slower
    <arithmetic mean>                5,269ms               5,285ms         ! 1.0x slower
    <harmonic mean>                  5,269ms               5,285ms         ! 1.0x slower

Peak Memory:
    jetstream --runs 1           1,269,532kB           1,137,040kB       ^ 1.12x smaller

    <geometric mean>             1,269,532kB           1,137,040kB       ^ 1.12x smaller
    <arithmetic mean>            1,269,532kB           1,137,040kB       ^ 1.12x smaller
    <harmonic mean>              1,269,532kB           1,137,040kB       ^ 1.12x smaller
Comment 3 Geoffrey Garen 2016-05-17 15:22:07 PDT
Octane splay (and Octane overall) is OK with this:

                            Baseline                   Patch                                       

encrypt                 0.19483+-0.00094    ?     0.19562+-0.00130       ?
decrypt                 3.43557+-0.05172    ?     3.44904+-0.02291       ?
deltablue      x2       0.15733+-0.00478    ?     0.16024+-0.00694       ? might be 1.0185x slower
earley                  0.36196+-0.00224    ?     0.36252+-0.00153       ?
boyer                   5.79808+-0.01359          5.79511+-0.06196       
navier-stokes  x2       5.36963+-0.00812    ?     5.37160+-0.01836       ?
raytrace       x2       0.96771+-0.01209          0.95949+-0.00898       
richards       x2       0.10679+-0.00229          0.10582+-0.00125       
splay          x2       0.38585+-0.00202    ?     0.38934+-0.00491       ?
regexp         x2      19.73298+-0.07684    ?    19.86209+-0.08336       ?
pdfjs          x2      46.21900+-0.80682         46.07309+-0.69264       
mandreel       x2      51.22570+-0.48006         51.11198+-0.21479       
gbemu          x2      43.63421+-8.62700         38.98809+-0.31607         might be 1.1192x faster
closure                 0.68696+-0.00272          0.68326+-0.00444       
jquery                  8.56558+-0.03242    ?     8.59034+-0.05483       ?
box2d          x2      12.24223+-0.13089         12.21356+-0.14817       
zlib           x2     411.37925+-2.31866        402.43274+-24.12873        might be 1.0222x faster
typescript     x2     870.34131+-3.97328        867.75671+-12.64601      

<geometric>             6.24665+-0.08206          6.19660+-0.02566         might be 1.0081x faster
Comment 4 Michael Saboff 2016-05-17 15:23:38 PDT
Comment on attachment 279170 [details]
Patch

r=me
Comment 5 Geoffrey Garen 2016-05-17 15:26:24 PDT
Committed r201050: <http://trac.webkit.org/changeset/201050>