Issue metadata
Sign in to add a comment
|
11.4%-13.1% regression in blink_perf.bindings at 522800:522932 |
||||||||||||||||||||
Issue descriptionSee the link to graphs below.
,
Dec 14 2017
📍 Pinpoint job started. https://pinpoint-dot-chromeperf.appspot.com/job/12ecedaa040000
,
Jan 22 2018
+dtu, the job in #2 still says "running" even though it was started Dec 14. Kicking off another bisect.
,
Jan 22 2018
📍 Pinpoint job started. https://pinpoint-dot-chromeperf.appspot.com/job/14c81224840000
,
Jan 22 2018
📍 Pinpoint job started. https://pinpoint-dot-chromeperf.appspot.com/job/12f27c24840000
,
Jan 23 2018
📍 Found significant differences after each of 3 commits. https://pinpoint-dot-chromeperf.appspot.com/job/12f27c24840000 Fix autofill focus handling for webview By changwan@chromium.org · Fri Dec 08 19:22:33 2017 chromium @ eb8a057e1e373652f61585b12313db4fe57ab8fd Update strings in the DICE turn-on-sync UI By tangltom@chromium.org · Fri Dec 08 20:19:46 2017 chromium @ e0f4019ad4bf7285af7638c013f24c9dac62b925 [ios] Fixes Omnibox CleanToolbar Animations. By sczs@chromium.org · Fri Dec 08 20:22:05 2017 chromium @ ce3cea470d376caf206e8b8200d7b0eda5548126 Understanding performance regressions: http://g.co/ChromePerformanceRegressions
,
Jan 23 2018
Removing myself as owner since ce3cea470d376caf206e8b8200d7b0eda5548126 is an iOS UI only change, and this looks like its related to Blink performance. Kicking off another bisect.
,
Jan 23 2018
📍 Pinpoint job started. https://pinpoint-dot-chromeperf.appspot.com/job/14912cd4840000
,
Jan 23 2018
My CL should only target Android WebView: https://chromium-review.googlesource.com/c/chromium/src/+/748887. It's guarded by focus_requires_scroll_, which is set to false in content autofill driver when autofill provider is non-null. In production, this is only set in aw_contents.cc.
,
Jan 23 2018
😿 Pinpoint job stopped with an error. https://pinpoint-dot-chromeperf.appspot.com/job/14c81224840000
,
Jan 23 2018
😿 Pinpoint job stopped with an error. https://pinpoint-dot-chromeperf.appspot.com/job/14912cd4840000
,
Feb 23 2018
Re-running after the fix for the errors.
,
Feb 23 2018
📍 Pinpoint job started. https://pinpoint-dot-chromeperf.appspot.com/job/12fd2b2f840000
,
Feb 23 2018
📍 Pinpoint job started. https://pinpoint-dot-chromeperf.appspot.com/job/12bb8d9f840000
,
Feb 25 2018
📍 Found significant differences after each of 9 commits. https://pinpoint-dot-chromeperf.appspot.com/job/12fd2b2f840000 MD Extensions: fix error message styling in error-console. by scottchen@chromium.org https://chromium.googlesource.com/chromium/src/+/70df9a14f7b4dc8edbed69be2eeac423f6ecadef Migrate BitmapFetcher to SimpleURLLoader by pilgrim@chromium.org https://chromium.googlesource.com/chromium/src/+/445bb78e03bf0352dd344afa3bb4d92e628c5b57 safe_browsing_service_browsertest: protect data which is accessed from multiple threads. by yzshen@chromium.org https://chromium.googlesource.com/chromium/src/+/c535dfb4753effb51ee45b1889f098223d72003b Remove AMD-style module support from gin/. by yzshen@chromium.org https://chromium.googlesource.com/chromium/src/+/c1ab49a9e23036de2f94da601782da9985a9a481 Delete x509_util::ParseCertificateSandboxed, use X509Certificate in WebURLLoaderImpl by mattm@chromium.org https://chromium.googlesource.com/chromium/src/+/1ff29573457ad07f198126582e3a22d39de8f041 [Offline Pages] Adding UMA for consistency checks in OPMTaskified. by romax@chromium.org https://chromium.googlesource.com/chromium/src/+/73e94b9aeb8e440db2929b9f7be1dfd0d8ee6cc2 Don't clear DnsAttempts that have received a response by mgersh@chromium.org https://chromium.googlesource.com/chromium/src/+/9069772b10e2796e4a09d6248a81b3c4ea4506d5 [heap] make SupportsInlineAllocation virtual in SpaceWithLinearArea by ofrobots@google.com https://chromium.googlesource.com/v8/v8/+/eb62a4aba4b329592979d80ab89e981d59f06344 [api] Make api-arguments.h interface more obvious by cbruni@chromium.org https://chromium.googlesource.com/v8/v8/+/bbf43d8488bb3d5b5a5d5189c41fc58bd5f729e6 Understanding performance regressions: http://g.co/ChromePerformanceRegressions
,
Feb 25 2018
📍 Found significant differences after each of 7 commits. https://pinpoint-dot-chromeperf.appspot.com/job/12bb8d9f840000 Replace ScriptSourceCode::GetResource() with CacheHandler() by hiroshige@chromium.org https://chromium.googlesource.com/chromium/src/+/945431d84342213951f61cadde27c5192fac1f70 Bindings: Use std::move on SerializedScriptValue references. by pwnall@chromium.org https://chromium.googlesource.com/chromium/src/+/6f1096decbadfaec4867f9d9d2217886123e961f Delete x509_util::ParseCertificateSandboxed, use X509Certificate in WebURLLoaderImpl by mattm@chromium.org https://chromium.googlesource.com/chromium/src/+/1ff29573457ad07f198126582e3a22d39de8f041 [heap] Minor MC: Pause concurrent marking by mlippautz@chromium.org https://chromium.googlesource.com/v8/v8/+/4c7f2d814f40cab847f069c9f957dad2ace7fa70 [heap] make SupportsInlineAllocation virtual in SpaceWithLinearArea by ofrobots@google.com https://chromium.googlesource.com/v8/v8/+/eb62a4aba4b329592979d80ab89e981d59f06344 [api] Make api-arguments.h interface more obvious by cbruni@chromium.org https://chromium.googlesource.com/v8/v8/+/bbf43d8488bb3d5b5a5d5189c41fc58bd5f729e6 Update V8 to version 6.5.33. by v8-autoroll@chromium.org https://chromium.googlesource.com/chromium/src/+/b908f3c1018c991647179b5aa395bccc0c59069f Understanding performance regressions: http://g.co/ChromePerformanceRegressions
,
Feb 26 2018
,
Feb 26 2018
,
Feb 26 2018
,
Feb 26 2018
,
Feb 26 2018
,
Feb 26 2018
,
Feb 26 2018
+cbruni, ishell In both jobs, I see the biggest jump at cbruni's change: [api] Make api-arguments.h interface more obvious by cbruni https://chromium.googlesource.com/v8/v8/+/bbf43d8488bb3d5b5a5d5189c41fc58bd5f729e6 Also +jbroman, yukishiino, haraken, the owners of blink_perf.bindings. There seems to be a lot of variability in this benchmark. +sullivan, I hope that having more devices will help mitigate the noise in the results in jobs like these, by amortizing the variability across many devices. (The jobs in c15 and c16 ran on 2 and 3 devices, respectively.)
,
Feb 27 2018
My CL is mostly a refactoring... said every chrome engineer ever :) I did factor out certain snippets into separate methods. Maybe that drastically changed the performance due to additional function calls? I will look into it.
,
Mar 26 2018
,
Mar 27 2018
I cannot detect the same regression on other than the mac bots. Will have to repro on my macbook later today.
,
Apr 3 2018
The following revision refers to this bug: https://chromium.googlesource.com/v8/v8.git/+/bd93135e4ed21ce20268963cfd5db787b2646519 commit bd93135e4ed21ce20268963cfd5db787b2646519 Author: Camillo Bruni <cbruni@chromium.org> Date: Tue Apr 03 08:25:17 2018 [api] Make more Callback helpers inlineable Bug: chromium:794998 Change-Id: Ib607bc891625db686fe37cfe416c3abf4ddf9a2b Reviewed-on: https://chromium-review.googlesource.com/983777 Reviewed-by: Igor Sheludko <ishell@chromium.org> Commit-Queue: Camillo Bruni <cbruni@chromium.org> Cr-Commit-Position: refs/heads/master@{#52317} [modify] https://crrev.com/bd93135e4ed21ce20268963cfd5db787b2646519/src/api-arguments-inl.h [modify] https://crrev.com/bd93135e4ed21ce20268963cfd5db787b2646519/src/api-arguments.cc [modify] https://crrev.com/bd93135e4ed21ce20268963cfd5db787b2646519/src/api-arguments.h [modify] https://crrev.com/bd93135e4ed21ce20268963cfd5db787b2646519/src/builtins/builtins-api.cc
,
Apr 12 2018
My CL didn't move the needle on this. While profiling on the mac I didn't see anything else popping up. I might run another profile on my new machine to see if I can spot something else.
,
May 3 2018
I didn't see much else to do while profiling and the performance is roughly back to what it was before possibly due to other work. |
|||||||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||||||
Comment 1 by 42576172...@developer.gserviceaccount.com
, Dec 14 2017