Move Android WebView aosp perf bot from fyi to main waterfall |
||||
Issue descriptionWebView perf bot is now having consistent results with the main perf bot's Android Chrome perf bot(Nexus5). I propose to move Android WebView perf bot to the main perf waterfall so it can be part of the normal perf rotation. Android WebView Perf https://build.chromium.org/p/chromium.perf.fyi/builders/android_webview_aosp_perf (Currently red) Android Chrome Perf https://build.chromium.org/p/chromium.perf/builders/Android%20Nexus5%20Perf%20(1) https://build.chromium.org/p/chromium.perf/builders/Android%20Nexus5%20Perf%20(2) (Currently red for the same test) https://build.chromium.org/p/chromium.perf/builders/Android%20Nexus5%20Perf%20(3) Caveat: The buildslave for WebView Perf bot is a1(build15-a1) while all the perf bot on the main perf waterfall is b1. Would there be any problem for this?
,
Mar 30 2016
I have a CL that makes the needed recipe changes to move the webview perf bots to the main waterfall. https://codereview.chromium.org/1777863005 We are just waiting on getting additional capacity to run the webview tests so that they can we run (and have more issues bisected) more quickly.
,
Apr 5 2016
,
Apr 18 2016
The following revision refers to this bug: https://chromium.googlesource.com/chromium/tools/build.git/+/1bc43f2edf6bffa87baea20c12a4d48161cb01ae commit 1bc43f2edf6bffa87baea20c12a4d48161cb01ae Author: mikecase@chromium.org <mikecase@chromium.org> Date: Mon Apr 18 18:41:41 2016 Recipe changes to move android webiew perf bot to the android/perf recipe. This change is motivated by us getting 2 new WebView perf bots and I would like to use the sharding logic in the android/perf recipe for the new bots. BUG= 590800 Review URL: https://codereview.chromium.org/1777863005 git-svn-id: svn://svn.chromium.org/chrome/trunk/tools/build@299994 0039d316-1c4b-4281-b951-d872f2087c98 [modify] https://crrev.com/1bc43f2edf6bffa87baea20c12a4d48161cb01ae/masters/master.chromium.perf.fyi/master.cfg [modify] https://crrev.com/1bc43f2edf6bffa87baea20c12a4d48161cb01ae/masters/master.chromium.perf.fyi/slaves.cfg [modify] https://crrev.com/1bc43f2edf6bffa87baea20c12a4d48161cb01ae/scripts/slave/recipe_modules/chromium_android/api.py [modify] https://crrev.com/1bc43f2edf6bffa87baea20c12a4d48161cb01ae/scripts/slave/recipe_modules/chromium_android/chromium_config.py [modify] https://crrev.com/1bc43f2edf6bffa87baea20c12a4d48161cb01ae/scripts/slave/recipe_modules/chromium_android/config.py [modify] https://crrev.com/1bc43f2edf6bffa87baea20c12a4d48161cb01ae/scripts/slave/recipes/android/builder.expected/full_chromium_perf_Android_Builder.json [modify] https://crrev.com/1bc43f2edf6bffa87baea20c12a4d48161cb01ae/scripts/slave/recipes/android/builder.py [rename] https://crrev.com/1bc43f2edf6bffa87baea20c12a4d48161cb01ae/scripts/slave/recipes/android/perf.expected/full_chromium_perf_fyi_Android_Nexus5_WebView_Perf__1_.json [copy] https://crrev.com/1bc43f2edf6bffa87baea20c12a4d48161cb01ae/scripts/slave/recipes/android/perf.expected/full_chromium_perf_fyi_Android_Nexus5x_WebView_Perf__1_.json [copy] https://crrev.com/1bc43f2edf6bffa87baea20c12a4d48161cb01ae/scripts/slave/recipes/android/perf.expected/full_chromium_perf_fyi_Android_Nexus5x_WebView_Perf__2_.json [modify] https://crrev.com/1bc43f2edf6bffa87baea20c12a4d48161cb01ae/scripts/slave/recipes/android/perf.py [delete] https://crrev.com/5b894af76d7a1f3a6ccb5c6204e0c416add78e46/scripts/slave/recipes/android_webview_aosp_perf.py
,
Apr 19 2016
The following revision refers to this bug: http://goto.ext.google.com/viewvc/chrome-internal?view=rev&revision=86750 ------------------------------------------------------------------ r86750 | recipe-roller@chromium.org | 2016-04-19T00:03:06.829470Z -----------------------------------------------------------------
,
Jun 6 2016
Is the Webview bot working? The graphs appear to be out of date. https://chromeperf.appspot.com/report?sid=8aa74c7973f7ceff953a68427e4d66a5c2e6508a4bebf1106776d512222ef246
,
Jun 6 2016
That link is for the fyi bot, and it got moved from there. https://build.chromium.org/p/chromium.perf/builders/Android%20Nexus5X%20WebView%20Perf%20(1) https://build.chromium.org/p/chromium.perf/builders/Android%20Nexus5X%20WebView%20Perf%20(2) https://build.chromium.org/p/chromium.perf/builders/Android%20Nexus5X%20WebView%20Perf%20(3) It looks like that metric is no longer being shown though :/ https://chromeperf.appspot.com/new_points?pattern=ChromiumPerf%2Fandroid-webview-nexus5X%2Fsmoothness.top_25_smooth%2Fmean_input_event_latency&num_points=&max_tests= But ther metrics are: https://chromeperf.appspot.com/new_points?pattern=ChromiumPerf%2Fandroid-webview-nexus5X%2Fsmoothness.top_25_smooth%2F*&num_points=100&max_tests=100
,
Jun 6 2016
Old trace with other metrics: aGexpEAPJKsl5hm6WFr6nZd0BGxskTaei4oZjAuE9v77imiNH1pwzdirHaY9H6pum6k64l5gD9kRpLJxJF5Ey7dfG9MEUxKHO0K8M New trace missing metrics: https://00e9e64bac83c6c863c5d2a5c4d5afb7e6868ec6efd4f13324-apidata.googleusercontent.com/download/storage/v1_internal/b/chrome-telemetry-output/o/trace-file-id_7-2016-06-05_13-34-40-24295.html?qk=AD5uMEt3UU73NEO9R5cVc8YmbWxWmB1NB8FQLbWsYr6auBq_YDXnOPt6IXFhpOYkEFZJenPcmCE4n7orxMf3Cxz33-8cI75Kt3Tz3vL1dev13Im6p376aB1ZHziTlU0Gno_ZCwR_65CPzzS3CZcVuNIw54FR7n1-kWr8_GcCR0N1WlBpWGFVDWxKmgcFL1xIPHi186JgBpY026zu_Y0iMktoAYOoqnlMatibLqKnKP4u4Xf--PTdf1xu1uLkXe_jkRZpgAcdgsDJ4qxIne2hxvnuD8z0SjE4OLCMTjvKuzry7bz_JrzyzH2t2auzX3xZACfjRTdhV1ypoWjfpJBkEX4qdv6F6CsDu2r8heRWwIwxvsANLzK-rA0-Rd_SbhEhxLKdSwwPsvaPiFbCZEa2gXNCuMzhqQSA8b46lis_Ue-oyFKlV4UHIeLuWmAqhrPvLOYxQciKeKn6ckxHh5tj3VQotecl91xwkeFijfatF4OI_wUzU-wx3wenE5KrB0Db_8fVGZjkwIYI5Y3T3xt9ZrojoR3FMTdhDSKN4w5Sz9sPbTV0n91rTkCv9iDHQ3HASnvMlMk1BuK1WgtjiVncO4X1-T8_l9k9YTMBAfcM4oHWMVBObsdz4irrAA-ZRs2_mc9jvjYDaIeHSYhzhmfnscdH7ySjkDuHmIAOR4wxI-8QDafpheK5xYQYtH6J66sMyXnwta0SVkoCo7hG-sxUOReLORrftBZqigGDXWXsQnyJJkwzOe1RaCNKxvFe1Mfx4kjGz4yyABOL4sCf-4WsXRZB69YtcgioMsc3VWHhBA8AsCeuo-4hYSj92C_WWjsosuW3T8yHzF_VUQdEx_c_9q6hbre5mulbll4v-2wKEM4HCvLPnE-UaCsWhLReOs1wyrjgSwH40rjH The second trace does not appear to want to open.
,
Jun 6 2016
My links were bad and I should feel bad for it... New trace: https://console.developers.google.com/m/cloudstorage/b/chrome-telemetry-output/o/trace-file-id_7-2016-06-05_16-06-06-23111.html old trace: https://console.developers.google.com/m/cloudstorage/b/chrome-telemetry-output/o/trace-file-id_7-2016-05-13_10-06-33-84979.html
,
Jun 6 2016
Thanks for the trace links, Randy. tdresser@: any idea what's going on here? Are we missing some critical trace events to compute mean_input_event_latency?
,
Jun 6 2016
The gesture scroll update latency traces are present. That should be all that's needed for first_gesture_scroll_update_latency. I'm not sure what's going on here. https://chromeperf.appspot.com/report?sid=8dadacf6b13adf73c9b2cac143c7506e46da4c4e49806d6cc609cbec7dd8f0a0
,
Jun 6 2016
It is getting None for some reason. @@@STEP_LOG_LINE@json.output@ "http://sports.yahoo.com/": {@@@ @@@STEP_LOG_LINE@json.output@ "description": "First gesture scroll update latency measures the time it takes to process the very first gesture scroll update input event. The first scroll gesture can often get delayed by work related to page loading.", @@@ @@@STEP_LOG_LINE@json.output@ "important": true, @@@ @@@STEP_LOG_LINE@json.output@ "improvement_direction": "down", @@@ @@@STEP_LOG_LINE@json.output@ "name": "first_gesture_scroll_update_latency", @@@ @@@STEP_LOG_LINE@json.output@ "none_value_reason": "Merging values containing a None value results in a None value.", @@@ @@@STEP_LOG_LINE@json.output@ "page_id": 22, @@@ @@@STEP_LOG_LINE@json.output@ "std": null, @@@ @@@STEP_LOG_LINE@json.output@ "type": "list_of_scalar_values", @@@ @@@STEP_LOG_LINE@json.output@ "units": "ms", @@@ @@@STEP_LOG_LINE@json.output@ "values": null@@@ @@@STEP_LOG_LINE@json.output@ }, @@@
,
Jun 6 2016
Oh, we don't have InputLatency::GestureScrollUpdate slices in the new trace, only Latency::ScrollUpdate. I wonder if this is due to the change to make webview input async. I think that happened fairly recently. +boliu@
,
Jun 6 2016
tdresser: can you point out the trace event in code? "async input" means using the same input path as chrome, so if webview is missing this event, then it probably never had it to begin with
,
Jun 6 2016
async LatencyInfo stuff, kinda complicated where start and end happens actually happens :/
,
Jun 6 2016
There are InputLatency::GestureScrollUpdate start events, but no end events in webview traces I think. Now where does the end one happen in chrome that's missing...
,
Jun 6 2016
Webview never calls RenderWidgetHostLatencyTracker::OnFrameSwapped. Swaps are still different between webview and chrome.
,
Jun 6 2016
> Webview never calls RenderWidgetHostLatencyTracker::OnFrameSwapped. Swaps are still different between webview and chrome. And that's the browser compositor swap. Ok, this can't easily be made to work because webview doesn't have a browser compositor, and the thing that receives renderer frames does not live on the UI thread. So this whole RenderWidgetHostLatencyTracker thing needs to be re-written to work on webview. On the other hand, swap scheduling is not really in webview's control anyway, so maybe not all that important to keep track.
,
Jun 7 2016
Thanks for digging into this. I think we should address this at some point, but it's a separate issue. (Filed issue 617942). Despite the fact that the webview doesn't schedule the swap, the timing of the swap impacts the user experience of scroll latency, so we should keep track of it.
,
Jan 17 2017
|
||||
►
Sign in to add a comment |
||||
Comment 1 by tdres...@chromium.org
, Mar 30 2016